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, 2010, 2011
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, 2010, 2011
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 1.3 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} -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
168 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-abi-version=@var{n} @gol
212 -fobjc-call-cxx-cdtors @gol
213 -fobjc-direct-dispatch @gol
214 -fobjc-exceptions @gol
217 -fobjc-std=objc1 @gol
218 -freplace-objc-classes @gol
221 -Wassign-intercept @gol
222 -Wno-protocol -Wselector @gol
223 -Wstrict-selector-match @gol
224 -Wundeclared-selector}
226 @item Language Independent Options
227 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
228 @gccoptlist{-fmessage-length=@var{n} @gol
229 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
230 -fno-diagnostics-show-option}
232 @item Warning Options
233 @xref{Warning Options,,Options to Request or Suppress Warnings}.
234 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
235 -pedantic-errors @gol
236 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
237 -Wno-attributes -Wno-builtin-macro-redefined @gol
238 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
239 -Wchar-subscripts -Wclobbered -Wcomment @gol
240 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
241 -Wno-deprecated-declarations -Wdisabled-optimization @gol
242 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
243 -Wno-endif-labels -Werror -Werror=* @gol
244 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
245 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
246 -Wformat-security -Wformat-y2k @gol
247 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
248 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
249 -Winit-self -Winline -Wmaybe-uninitialized @gol
250 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
251 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
252 -Wlogical-op -Wlong-long @gol
253 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
254 -Wmissing-format-attribute -Wmissing-include-dirs @gol
256 -Wno-multichar -Wnonnull -Wno-overflow @gol
257 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
258 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
259 -Wpointer-arith -Wno-pointer-to-int-cast @gol
260 -Wredundant-decls @gol
261 -Wreturn-type -Wsequence-point -Wshadow @gol
262 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
263 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
264 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
265 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
266 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
267 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
268 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
269 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
270 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
271 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
272 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
274 @item C and Objective-C-only Warning Options
275 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
276 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
277 -Wold-style-declaration -Wold-style-definition @gol
278 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
279 -Wdeclaration-after-statement -Wpointer-sign}
281 @item Debugging Options
282 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
283 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
284 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
285 -fdisable-ipa-@var{pass_name} @gol
286 -fdisable-rtl-@var{pass_name} @gol
287 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
288 -fdisable-tree-@var{pass_name} @gol
289 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
290 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
291 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
292 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
293 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
295 -fdump-statistics @gol
297 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
301 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-nrv -fdump-tree-vect @gol
311 -fdump-tree-sink @gol
312 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
316 -ftree-vectorizer-verbose=@var{n} @gol
317 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
318 -fdump-final-insns=@var{file} @gol
319 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
320 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
321 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
322 -fenable-@var{kind}-@var{pass} @gol
323 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
324 -fdebug-types-section @gol
325 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
326 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
327 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
328 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
329 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
330 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
331 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
332 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
333 -gvms -gxcoff -gxcoff+ @gol
334 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
335 -fdebug-prefix-map=@var{old}=@var{new} @gol
336 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
337 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
338 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
339 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
340 -print-prog-name=@var{program} -print-search-dirs -Q @gol
341 -print-sysroot -print-sysroot-headers-suffix @gol
342 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
344 @item Optimization Options
345 @xref{Optimize Options,,Options that Control Optimization}.
346 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
347 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
348 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
349 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
350 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
351 -fcompare-elim -fcprop-registers -fcrossjumping @gol
352 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
353 -fcx-limited-range @gol
354 -fdata-sections -fdce -fdce -fdelayed-branch @gol
355 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
356 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
357 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
358 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
359 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
360 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
361 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
362 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
363 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
364 -fira-algorithm=@var{algorithm} @gol
365 -fira-region=@var{region} @gol
366 -fira-loop-pressure -fno-ira-share-save-slots @gol
367 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
368 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
369 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
370 -floop-parallelize-all -flto -flto-compression-level @gol
371 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
372 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
373 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
374 -fno-default-inline @gol
375 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
376 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
377 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
378 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
379 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
380 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
381 -fprefetch-loop-arrays @gol
382 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
383 -fprofile-generate=@var{path} @gol
384 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
385 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
386 -freorder-blocks-and-partition -freorder-functions @gol
387 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
388 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
389 -fsched-spec-load -fsched-spec-load-dangerous @gol
390 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
391 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
392 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
393 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
394 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
395 -fselective-scheduling -fselective-scheduling2 @gol
396 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
397 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
398 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
399 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
401 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
402 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
403 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
404 -ftree-loop-if-convert-stores -ftree-loop-im @gol
405 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
406 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
407 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
408 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
409 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
410 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
411 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
412 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
413 -fwhole-program -fwpa -fuse-linker-plugin @gol
414 --param @var{name}=@var{value}
415 -O -O0 -O1 -O2 -O3 -Os -Ofast}
417 @item Preprocessor Options
418 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
419 @gccoptlist{-A@var{question}=@var{answer} @gol
420 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
421 -C -dD -dI -dM -dN @gol
422 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
423 -idirafter @var{dir} @gol
424 -include @var{file} -imacros @var{file} @gol
425 -iprefix @var{file} -iwithprefix @var{dir} @gol
426 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
427 -imultilib @var{dir} -isysroot @var{dir} @gol
428 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
429 -P -fworking-directory -remap @gol
430 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
431 -Xpreprocessor @var{option}}
433 @item Assembler Option
434 @xref{Assembler Options,,Passing Options to the Assembler}.
435 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
438 @xref{Link Options,,Options for Linking}.
439 @gccoptlist{@var{object-file-name} -l@var{library} @gol
440 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
441 -s -static -static-libgcc -static-libstdc++ -shared @gol
442 -shared-libgcc -symbolic @gol
443 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
446 @item Directory Options
447 @xref{Directory Options,,Options for Directory Search}.
448 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
449 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
452 @item Machine Dependent Options
453 @xref{Submodel Options,,Hardware Models and Configurations}.
454 @c This list is ordered alphanumerically by subsection name.
455 @c Try and put the significant identifier (CPU or system) first,
456 @c so users have a clue at guessing where the ones they want will be.
459 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
460 -mabi=@var{name} @gol
461 -mapcs-stack-check -mno-apcs-stack-check @gol
462 -mapcs-float -mno-apcs-float @gol
463 -mapcs-reentrant -mno-apcs-reentrant @gol
464 -msched-prolog -mno-sched-prolog @gol
465 -mlittle-endian -mbig-endian -mwords-little-endian @gol
466 -mfloat-abi=@var{name} -mfpe @gol
467 -mfp16-format=@var{name}
468 -mthumb-interwork -mno-thumb-interwork @gol
469 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
470 -mstructure-size-boundary=@var{n} @gol
471 -mabort-on-noreturn @gol
472 -mlong-calls -mno-long-calls @gol
473 -msingle-pic-base -mno-single-pic-base @gol
474 -mpic-register=@var{reg} @gol
475 -mnop-fun-dllimport @gol
476 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
477 -mpoke-function-name @gol
479 -mtpcs-frame -mtpcs-leaf-frame @gol
480 -mcaller-super-interworking -mcallee-super-interworking @gol
481 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
482 -mword-relocations @gol
483 -mfix-cortex-m3-ldrd}
486 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
487 -mcall-prologues -mtiny-stack -mint8}
489 @emph{Blackfin Options}
490 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
491 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
492 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
493 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
494 -mno-id-shared-library -mshared-library-id=@var{n} @gol
495 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
496 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
497 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
501 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
502 -msim -msdata=@var{sdata-type}}
505 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
506 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
507 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
508 -mstack-align -mdata-align -mconst-align @gol
509 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
510 -melf -maout -melinux -mlinux -sim -sim2 @gol
511 -mmul-bug-workaround -mno-mul-bug-workaround}
513 @emph{Darwin Options}
514 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
515 -arch_only -bind_at_load -bundle -bundle_loader @gol
516 -client_name -compatibility_version -current_version @gol
518 -dependency-file -dylib_file -dylinker_install_name @gol
519 -dynamic -dynamiclib -exported_symbols_list @gol
520 -filelist -flat_namespace -force_cpusubtype_ALL @gol
521 -force_flat_namespace -headerpad_max_install_names @gol
523 -image_base -init -install_name -keep_private_externs @gol
524 -multi_module -multiply_defined -multiply_defined_unused @gol
525 -noall_load -no_dead_strip_inits_and_terms @gol
526 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
527 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
528 -private_bundle -read_only_relocs -sectalign @gol
529 -sectobjectsymbols -whyload -seg1addr @gol
530 -sectcreate -sectobjectsymbols -sectorder @gol
531 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
532 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
533 -segprot -segs_read_only_addr -segs_read_write_addr @gol
534 -single_module -static -sub_library -sub_umbrella @gol
535 -twolevel_namespace -umbrella -undefined @gol
536 -unexported_symbols_list -weak_reference_mismatches @gol
537 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
538 -mkernel -mone-byte-bool}
540 @emph{DEC Alpha Options}
541 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
542 -mieee -mieee-with-inexact -mieee-conformant @gol
543 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
544 -mtrap-precision=@var{mode} -mbuild-constants @gol
545 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
546 -mbwx -mmax -mfix -mcix @gol
547 -mfloat-vax -mfloat-ieee @gol
548 -mexplicit-relocs -msmall-data -mlarge-data @gol
549 -msmall-text -mlarge-text @gol
550 -mmemory-latency=@var{time}}
552 @emph{DEC Alpha/VMS Options}
553 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
556 @gccoptlist{-msmall-model -mno-lsim}
559 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
560 -mhard-float -msoft-float @gol
561 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
562 -mdouble -mno-double @gol
563 -mmedia -mno-media -mmuladd -mno-muladd @gol
564 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
565 -mlinked-fp -mlong-calls -malign-labels @gol
566 -mlibrary-pic -macc-4 -macc-8 @gol
567 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
568 -moptimize-membar -mno-optimize-membar @gol
569 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
570 -mvliw-branch -mno-vliw-branch @gol
571 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
572 -mno-nested-cond-exec -mtomcat-stats @gol
576 @emph{GNU/Linux Options}
577 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
578 -tno-android-cc -tno-android-ld}
580 @emph{H8/300 Options}
581 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
584 @gccoptlist{-march=@var{architecture-type} @gol
585 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
586 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
587 -mfixed-range=@var{register-range} @gol
588 -mjump-in-delay -mlinker-opt -mlong-calls @gol
589 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
590 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
591 -mno-jump-in-delay -mno-long-load-store @gol
592 -mno-portable-runtime -mno-soft-float @gol
593 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
594 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
595 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
596 -munix=@var{unix-std} -nolibdld -static -threads}
598 @emph{i386 and x86-64 Options}
599 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
600 -mfpmath=@var{unit} @gol
601 -masm=@var{dialect} -mno-fancy-math-387 @gol
602 -mno-fp-ret-in-387 -msoft-float @gol
603 -mno-wide-multiply -mrtd -malign-double @gol
604 -mpreferred-stack-boundary=@var{num} @gol
605 -mincoming-stack-boundary=@var{num} @gol
606 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
607 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
608 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
609 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
610 -mthreads -mno-align-stringops -minline-all-stringops @gol
611 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
612 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
613 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
614 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
615 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
616 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
617 -mcmodel=@var{code-model} -mabi=@var{name} @gol
618 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
619 -msse2avx -mfentry -m8bit-idiv @gol
620 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
622 @emph{i386 and x86-64 Windows Options}
623 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
624 -mnop-fun-dllimport -mthread @gol
625 -municode -mwin32 -mwindows -fno-set-stack-executable}
628 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
629 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
630 -mconstant-gp -mauto-pic -mfused-madd @gol
631 -minline-float-divide-min-latency @gol
632 -minline-float-divide-max-throughput @gol
633 -mno-inline-float-divide @gol
634 -minline-int-divide-min-latency @gol
635 -minline-int-divide-max-throughput @gol
636 -mno-inline-int-divide @gol
637 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
638 -mno-inline-sqrt @gol
639 -mdwarf2-asm -mearly-stop-bits @gol
640 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
641 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
642 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
643 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
644 -msched-spec-ldc -msched-spec-control-ldc @gol
645 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
646 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
647 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
648 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
650 @emph{IA-64/VMS Options}
651 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
654 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
655 -msign-extend-enabled -muser-enabled}
657 @emph{M32R/D Options}
658 @gccoptlist{-m32r2 -m32rx -m32r @gol
660 -malign-loops -mno-align-loops @gol
661 -missue-rate=@var{number} @gol
662 -mbranch-cost=@var{number} @gol
663 -mmodel=@var{code-size-model-type} @gol
664 -msdata=@var{sdata-type} @gol
665 -mno-flush-func -mflush-func=@var{name} @gol
666 -mno-flush-trap -mflush-trap=@var{number} @gol
670 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
672 @emph{M680x0 Options}
673 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
674 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
675 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
676 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
677 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
678 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
679 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
680 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
684 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
685 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
686 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
687 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
688 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
691 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
692 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
693 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
694 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
697 @emph{MicroBlaze Options}
698 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
699 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
700 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
701 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
702 -mxl-mode-@var{app-model}}
705 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
706 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
707 -mips64 -mips64r2 @gol
708 -mips16 -mno-mips16 -mflip-mips16 @gol
709 -minterlink-mips16 -mno-interlink-mips16 @gol
710 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
711 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
712 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
713 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
714 -mfpu=@var{fpu-type} @gol
715 -msmartmips -mno-smartmips @gol
716 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
717 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
718 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
719 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
720 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
721 -membedded-data -mno-embedded-data @gol
722 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
723 -mcode-readable=@var{setting} @gol
724 -msplit-addresses -mno-split-addresses @gol
725 -mexplicit-relocs -mno-explicit-relocs @gol
726 -mcheck-zero-division -mno-check-zero-division @gol
727 -mdivide-traps -mdivide-breaks @gol
728 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
729 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
730 -mfix-24k -mno-fix-24k @gol
731 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
732 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
733 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
734 -mflush-func=@var{func} -mno-flush-func @gol
735 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
736 -mfp-exceptions -mno-fp-exceptions @gol
737 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
738 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
741 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
742 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
743 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
744 -mno-base-addresses -msingle-exit -mno-single-exit}
746 @emph{MN10300 Options}
747 @gccoptlist{-mmult-bug -mno-mult-bug @gol
748 -mno-am33 -mam33 -mam33-2 -mam34 @gol
749 -mtune=@var{cpu-type} @gol
750 -mreturn-pointer-on-d0 @gol
751 -mno-crt0 -mrelax -mliw -msetlb}
753 @emph{PDP-11 Options}
754 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
755 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
756 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
757 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
758 -mbranch-expensive -mbranch-cheap @gol
759 -munix-asm -mdec-asm}
761 @emph{picoChip Options}
762 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
763 -msymbol-as-address -mno-inefficient-warnings}
765 @emph{PowerPC Options}
766 See RS/6000 and PowerPC Options.
768 @emph{RS/6000 and PowerPC Options}
769 @gccoptlist{-mcpu=@var{cpu-type} @gol
770 -mtune=@var{cpu-type} @gol
771 -mcmodel=@var{code-model} @gol
772 -mpower -mno-power -mpower2 -mno-power2 @gol
773 -mpowerpc -mpowerpc64 -mno-powerpc @gol
774 -maltivec -mno-altivec @gol
775 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
776 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
777 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
778 -mfprnd -mno-fprnd @gol
779 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
780 -mnew-mnemonics -mold-mnemonics @gol
781 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
782 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
783 -malign-power -malign-natural @gol
784 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
785 -msingle-float -mdouble-float -msimple-fpu @gol
786 -mstring -mno-string -mupdate -mno-update @gol
787 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
788 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
789 -mstrict-align -mno-strict-align -mrelocatable @gol
790 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
791 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
792 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
793 -mprioritize-restricted-insns=@var{priority} @gol
794 -msched-costly-dep=@var{dependence_type} @gol
795 -minsert-sched-nops=@var{scheme} @gol
796 -mcall-sysv -mcall-netbsd @gol
797 -maix-struct-return -msvr4-struct-return @gol
798 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
799 -mblock-move-inline-limit=@var{num} @gol
800 -misel -mno-isel @gol
801 -misel=yes -misel=no @gol
803 -mspe=yes -mspe=no @gol
805 -mgen-cell-microcode -mwarn-cell-microcode @gol
806 -mvrsave -mno-vrsave @gol
807 -mmulhw -mno-mulhw @gol
808 -mdlmzb -mno-dlmzb @gol
809 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
810 -mprototype -mno-prototype @gol
811 -msim -mmvme -mads -myellowknife -memb -msdata @gol
812 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
813 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
814 -mno-recip-precision @gol
815 -mveclibabi=@var{type} -mfriz -mno-friz @gol
816 -mpointers-to-nested-functions -mno-pointers-to-nested-functions}
819 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
821 -mbig-endian-data -mlittle-endian-data @gol
824 -mas100-syntax -mno-as100-syntax@gol
826 -mmax-constant-size=@gol
828 -msave-acc-in-interrupts}
830 @emph{S/390 and zSeries Options}
831 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
832 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
833 -mlong-double-64 -mlong-double-128 @gol
834 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
835 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
836 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
837 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
838 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
841 @gccoptlist{-meb -mel @gol
845 -mscore5 -mscore5u -mscore7 -mscore7d}
848 @gccoptlist{-m1 -m2 -m2e @gol
849 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
851 -m4-nofpu -m4-single-only -m4-single -m4 @gol
852 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
853 -m5-64media -m5-64media-nofpu @gol
854 -m5-32media -m5-32media-nofpu @gol
855 -m5-compact -m5-compact-nofpu @gol
856 -mb -ml -mdalign -mrelax @gol
857 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
858 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
859 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
860 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
861 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
862 -maccumulate-outgoing-args -minvalid-symbols}
864 @emph{Solaris 2 Options}
865 @gccoptlist{-mimpure-text -mno-impure-text @gol
869 @gccoptlist{-mcpu=@var{cpu-type} @gol
870 -mtune=@var{cpu-type} @gol
871 -mcmodel=@var{code-model} @gol
872 -m32 -m64 -mapp-regs -mno-app-regs @gol
873 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
874 -mfpu -mno-fpu -mhard-float -msoft-float @gol
875 -mhard-quad-float -msoft-quad-float @gol
877 -mstack-bias -mno-stack-bias @gol
878 -munaligned-doubles -mno-unaligned-doubles @gol
879 -mv8plus -mno-v8plus -mvis -mno-vis}
882 @gccoptlist{-mwarn-reloc -merror-reloc @gol
883 -msafe-dma -munsafe-dma @gol
885 -msmall-mem -mlarge-mem -mstdmain @gol
886 -mfixed-range=@var{register-range} @gol
888 -maddress-space-conversion -mno-address-space-conversion @gol
889 -mcache-size=@var{cache-size} @gol
890 -matomic-updates -mno-atomic-updates}
892 @emph{System V Options}
893 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
896 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
897 -mprolog-function -mno-prolog-function -mspace @gol
898 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
899 -mapp-regs -mno-app-regs @gol
900 -mdisable-callt -mno-disable-callt @gol
903 -mv850e1 -mv850es @gol
908 @gccoptlist{-mg -mgnu -munix}
910 @emph{VxWorks Options}
911 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
912 -Xbind-lazy -Xbind-now}
914 @emph{x86-64 Options}
915 See i386 and x86-64 Options.
917 @emph{Xstormy16 Options}
920 @emph{Xtensa Options}
921 @gccoptlist{-mconst16 -mno-const16 @gol
922 -mfused-madd -mno-fused-madd @gol
924 -mserialize-volatile -mno-serialize-volatile @gol
925 -mtext-section-literals -mno-text-section-literals @gol
926 -mtarget-align -mno-target-align @gol
927 -mlongcalls -mno-longcalls}
929 @emph{zSeries Options}
930 See S/390 and zSeries Options.
932 @item Code Generation Options
933 @xref{Code Gen Options,,Options for Code Generation Conventions}.
934 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
935 -ffixed-@var{reg} -fexceptions @gol
936 -fnon-call-exceptions -funwind-tables @gol
937 -fasynchronous-unwind-tables @gol
938 -finhibit-size-directive -finstrument-functions @gol
939 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
940 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
941 -fno-common -fno-ident @gol
942 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
943 -fno-jump-tables @gol
944 -frecord-gcc-switches @gol
945 -freg-struct-return -fshort-enums @gol
946 -fshort-double -fshort-wchar @gol
947 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
948 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
949 -fno-stack-limit -fsplit-stack @gol
950 -fleading-underscore -ftls-model=@var{model} @gol
951 -ftrapv -fwrapv -fbounds-check @gol
952 -fvisibility -fstrict-volatile-bitfields}
956 * Overall Options:: Controlling the kind of output:
957 an executable, object files, assembler files,
958 or preprocessed source.
959 * C Dialect Options:: Controlling the variant of C language compiled.
960 * C++ Dialect Options:: Variations on C++.
961 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
963 * Language Independent Options:: Controlling how diagnostics should be
965 * Warning Options:: How picky should the compiler be?
966 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
967 * Optimize Options:: How much optimization?
968 * Preprocessor Options:: Controlling header files and macro definitions.
969 Also, getting dependency information for Make.
970 * Assembler Options:: Passing options to the assembler.
971 * Link Options:: Specifying libraries and so on.
972 * Directory Options:: Where to find header files and libraries.
973 Where to find the compiler executable files.
974 * Spec Files:: How to pass switches to sub-processes.
975 * Target Options:: Running a cross-compiler, or an old version of GCC.
978 @node Overall Options
979 @section Options Controlling the Kind of Output
981 Compilation can involve up to four stages: preprocessing, compilation
982 proper, assembly and linking, always in that order. GCC is capable of
983 preprocessing and compiling several files either into several
984 assembler input files, or into one assembler input file; then each
985 assembler input file produces an object file, and linking combines all
986 the object files (those newly compiled, and those specified as input)
987 into an executable file.
989 @cindex file name suffix
990 For any given input file, the file name suffix determines what kind of
995 C source code which must be preprocessed.
998 C source code which should not be preprocessed.
1001 C++ source code which should not be preprocessed.
1004 Objective-C source code. Note that you must link with the @file{libobjc}
1005 library to make an Objective-C program work.
1008 Objective-C source code which should not be preprocessed.
1012 Objective-C++ source code. Note that you must link with the @file{libobjc}
1013 library to make an Objective-C++ program work. Note that @samp{.M} refers
1014 to a literal capital M@.
1016 @item @var{file}.mii
1017 Objective-C++ source code which should not be preprocessed.
1020 C, C++, Objective-C or Objective-C++ header file to be turned into a
1021 precompiled header (default), or C, C++ header file to be turned into an
1022 Ada spec (via the @option{-fdump-ada-spec} switch).
1025 @itemx @var{file}.cp
1026 @itemx @var{file}.cxx
1027 @itemx @var{file}.cpp
1028 @itemx @var{file}.CPP
1029 @itemx @var{file}.c++
1031 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1032 the last two letters must both be literally @samp{x}. Likewise,
1033 @samp{.C} refers to a literal capital C@.
1037 Objective-C++ source code which must be preprocessed.
1039 @item @var{file}.mii
1040 Objective-C++ source code which should not be preprocessed.
1044 @itemx @var{file}.hp
1045 @itemx @var{file}.hxx
1046 @itemx @var{file}.hpp
1047 @itemx @var{file}.HPP
1048 @itemx @var{file}.h++
1049 @itemx @var{file}.tcc
1050 C++ header file to be turned into a precompiled header or Ada spec.
1053 @itemx @var{file}.for
1054 @itemx @var{file}.ftn
1055 Fixed form Fortran source code which should not be preprocessed.
1058 @itemx @var{file}.FOR
1059 @itemx @var{file}.fpp
1060 @itemx @var{file}.FPP
1061 @itemx @var{file}.FTN
1062 Fixed form Fortran source code which must be preprocessed (with the traditional
1065 @item @var{file}.f90
1066 @itemx @var{file}.f95
1067 @itemx @var{file}.f03
1068 @itemx @var{file}.f08
1069 Free form Fortran source code which should not be preprocessed.
1071 @item @var{file}.F90
1072 @itemx @var{file}.F95
1073 @itemx @var{file}.F03
1074 @itemx @var{file}.F08
1075 Free form Fortran source code which must be preprocessed (with the
1076 traditional preprocessor).
1081 @c FIXME: Descriptions of Java file types.
1087 @item @var{file}.ads
1088 Ada source code file which contains a library unit declaration (a
1089 declaration of a package, subprogram, or generic, or a generic
1090 instantiation), or a library unit renaming declaration (a package,
1091 generic, or subprogram renaming declaration). Such files are also
1094 @item @var{file}.adb
1095 Ada source code file containing a library unit body (a subprogram or
1096 package body). Such files are also called @dfn{bodies}.
1098 @c GCC also knows about some suffixes for languages not yet included:
1109 @itemx @var{file}.sx
1110 Assembler code which must be preprocessed.
1113 An object file to be fed straight into linking.
1114 Any file name with no recognized suffix is treated this way.
1118 You can specify the input language explicitly with the @option{-x} option:
1121 @item -x @var{language}
1122 Specify explicitly the @var{language} for the following input files
1123 (rather than letting the compiler choose a default based on the file
1124 name suffix). This option applies to all following input files until
1125 the next @option{-x} option. Possible values for @var{language} are:
1127 c c-header cpp-output
1128 c++ c++-header c++-cpp-output
1129 objective-c objective-c-header objective-c-cpp-output
1130 objective-c++ objective-c++-header objective-c++-cpp-output
1131 assembler assembler-with-cpp
1133 f77 f77-cpp-input f95 f95-cpp-input
1139 Turn off any specification of a language, so that subsequent files are
1140 handled according to their file name suffixes (as they are if @option{-x}
1141 has not been used at all).
1143 @item -pass-exit-codes
1144 @opindex pass-exit-codes
1145 Normally the @command{gcc} program will exit with the code of 1 if any
1146 phase of the compiler returns a non-success return code. If you specify
1147 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1148 numerically highest error produced by any phase that returned an error
1149 indication. The C, C++, and Fortran frontends return 4, if an internal
1150 compiler error is encountered.
1153 If you only want some of the stages of compilation, you can use
1154 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1155 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1156 @command{gcc} is to stop. Note that some combinations (for example,
1157 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1162 Compile or assemble the source files, but do not link. The linking
1163 stage simply is not done. The ultimate output is in the form of an
1164 object file for each source file.
1166 By default, the object file name for a source file is made by replacing
1167 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1169 Unrecognized input files, not requiring compilation or assembly, are
1174 Stop after the stage of compilation proper; do not assemble. The output
1175 is in the form of an assembler code file for each non-assembler input
1178 By default, the assembler file name for a source file is made by
1179 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1181 Input files that don't require compilation are ignored.
1185 Stop after the preprocessing stage; do not run the compiler proper. The
1186 output is in the form of preprocessed source code, which is sent to the
1189 Input files which don't require preprocessing are ignored.
1191 @cindex output file option
1194 Place output in file @var{file}. This applies regardless to whatever
1195 sort of output is being produced, whether it be an executable file,
1196 an object file, an assembler file or preprocessed C code.
1198 If @option{-o} is not specified, the default is to put an executable
1199 file in @file{a.out}, the object file for
1200 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1201 assembler file in @file{@var{source}.s}, a precompiled header file in
1202 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1207 Print (on standard error output) the commands executed to run the stages
1208 of compilation. Also print the version number of the compiler driver
1209 program and of the preprocessor and the compiler proper.
1213 Like @option{-v} except the commands are not executed and arguments
1214 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1215 This is useful for shell scripts to capture the driver-generated command lines.
1219 Use pipes rather than temporary files for communication between the
1220 various stages of compilation. This fails to work on some systems where
1221 the assembler is unable to read from a pipe; but the GNU assembler has
1226 Print (on the standard output) a description of the command line options
1227 understood by @command{gcc}. If the @option{-v} option is also specified
1228 then @option{--help} will also be passed on to the various processes
1229 invoked by @command{gcc}, so that they can display the command line options
1230 they accept. If the @option{-Wextra} option has also been specified
1231 (prior to the @option{--help} option), then command line options which
1232 have no documentation associated with them will also be displayed.
1235 @opindex target-help
1236 Print (on the standard output) a description of target-specific command
1237 line options for each tool. For some targets extra target-specific
1238 information may also be printed.
1240 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1241 Print (on the standard output) a description of the command line
1242 options understood by the compiler that fit into all specified classes
1243 and qualifiers. These are the supported classes:
1246 @item @samp{optimizers}
1247 This will display all of the optimization options supported by the
1250 @item @samp{warnings}
1251 This will display all of the options controlling warning messages
1252 produced by the compiler.
1255 This will display target-specific options. Unlike the
1256 @option{--target-help} option however, target-specific options of the
1257 linker and assembler will not be displayed. This is because those
1258 tools do not currently support the extended @option{--help=} syntax.
1261 This will display the values recognized by the @option{--param}
1264 @item @var{language}
1265 This will display the options supported for @var{language}, where
1266 @var{language} is the name of one of the languages supported in this
1270 This will display the options that are common to all languages.
1273 These are the supported qualifiers:
1276 @item @samp{undocumented}
1277 Display only those options which are undocumented.
1280 Display options which take an argument that appears after an equal
1281 sign in the same continuous piece of text, such as:
1282 @samp{--help=target}.
1284 @item @samp{separate}
1285 Display options which take an argument that appears as a separate word
1286 following the original option, such as: @samp{-o output-file}.
1289 Thus for example to display all the undocumented target-specific
1290 switches supported by the compiler the following can be used:
1293 --help=target,undocumented
1296 The sense of a qualifier can be inverted by prefixing it with the
1297 @samp{^} character, so for example to display all binary warning
1298 options (i.e., ones that are either on or off and that do not take an
1299 argument), which have a description the following can be used:
1302 --help=warnings,^joined,^undocumented
1305 The argument to @option{--help=} should not consist solely of inverted
1308 Combining several classes is possible, although this usually
1309 restricts the output by so much that there is nothing to display. One
1310 case where it does work however is when one of the classes is
1311 @var{target}. So for example to display all the target-specific
1312 optimization options the following can be used:
1315 --help=target,optimizers
1318 The @option{--help=} option can be repeated on the command line. Each
1319 successive use will display its requested class of options, skipping
1320 those that have already been displayed.
1322 If the @option{-Q} option appears on the command line before the
1323 @option{--help=} option, then the descriptive text displayed by
1324 @option{--help=} is changed. Instead of describing the displayed
1325 options, an indication is given as to whether the option is enabled,
1326 disabled or set to a specific value (assuming that the compiler
1327 knows this at the point where the @option{--help=} option is used).
1329 Here is a truncated example from the ARM port of @command{gcc}:
1332 % gcc -Q -mabi=2 --help=target -c
1333 The following options are target specific:
1335 -mabort-on-noreturn [disabled]
1339 The output is sensitive to the effects of previous command line
1340 options, so for example it is possible to find out which optimizations
1341 are enabled at @option{-O2} by using:
1344 -Q -O2 --help=optimizers
1347 Alternatively you can discover which binary optimizations are enabled
1348 by @option{-O3} by using:
1351 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1352 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1353 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1356 @item -no-canonical-prefixes
1357 @opindex no-canonical-prefixes
1358 Do not expand any symbolic links, resolve references to @samp{/../}
1359 or @samp{/./}, or make the path absolute when generating a relative
1364 Display the version number and copyrights of the invoked GCC@.
1368 Invoke all subcommands under a wrapper program. The name of the
1369 wrapper program and its parameters are passed as a comma separated
1373 gcc -c t.c -wrapper gdb,--args
1376 This will invoke all subprograms of @command{gcc} under
1377 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1378 @samp{gdb --args cc1 @dots{}}.
1380 @item -fplugin=@var{name}.so
1381 Load the plugin code in file @var{name}.so, assumed to be a
1382 shared object to be dlopen'd by the compiler. The base name of
1383 the shared object file is used to identify the plugin for the
1384 purposes of argument parsing (See
1385 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1386 Each plugin should define the callback functions specified in the
1389 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1390 Define an argument called @var{key} with a value of @var{value}
1391 for the plugin called @var{name}.
1393 @item -fdump-ada-spec@r{[}-slim@r{]}
1394 For C and C++ source and include files, generate corresponding Ada
1395 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1396 GNAT User's Guide}, which provides detailed documentation on this feature.
1398 @item -fdump-go-spec=@var{file}
1399 For input files in any language, generate corresponding Go
1400 declarations in @var{file}. This generates Go @code{const},
1401 @code{type}, @code{var}, and @code{func} declarations which may be a
1402 useful way to start writing a Go interface to code written in some
1405 @include @value{srcdir}/../libiberty/at-file.texi
1409 @section Compiling C++ Programs
1411 @cindex suffixes for C++ source
1412 @cindex C++ source file suffixes
1413 C++ source files conventionally use one of the suffixes @samp{.C},
1414 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1415 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1416 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1417 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1418 files with these names and compiles them as C++ programs even if you
1419 call the compiler the same way as for compiling C programs (usually
1420 with the name @command{gcc}).
1424 However, the use of @command{gcc} does not add the C++ library.
1425 @command{g++} is a program that calls GCC and treats @samp{.c},
1426 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1427 files unless @option{-x} is used, and automatically specifies linking
1428 against the C++ library. This program is also useful when
1429 precompiling a C header file with a @samp{.h} extension for use in C++
1430 compilations. On many systems, @command{g++} is also installed with
1431 the name @command{c++}.
1433 @cindex invoking @command{g++}
1434 When you compile C++ programs, you may specify many of the same
1435 command-line options that you use for compiling programs in any
1436 language; or command-line options meaningful for C and related
1437 languages; or options that are meaningful only for C++ programs.
1438 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1439 explanations of options for languages related to C@.
1440 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1441 explanations of options that are meaningful only for C++ programs.
1443 @node C Dialect Options
1444 @section Options Controlling C Dialect
1445 @cindex dialect options
1446 @cindex language dialect options
1447 @cindex options, dialect
1449 The following options control the dialect of C (or languages derived
1450 from C, such as C++, Objective-C and Objective-C++) that the compiler
1454 @cindex ANSI support
1458 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1459 equivalent to @samp{-std=c++98}.
1461 This turns off certain features of GCC that are incompatible with ISO
1462 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1463 such as the @code{asm} and @code{typeof} keywords, and
1464 predefined macros such as @code{unix} and @code{vax} that identify the
1465 type of system you are using. It also enables the undesirable and
1466 rarely used ISO trigraph feature. For the C compiler,
1467 it disables recognition of C++ style @samp{//} comments as well as
1468 the @code{inline} keyword.
1470 The alternate keywords @code{__asm__}, @code{__extension__},
1471 @code{__inline__} and @code{__typeof__} continue to work despite
1472 @option{-ansi}. You would not want to use them in an ISO C program, of
1473 course, but it is useful to put them in header files that might be included
1474 in compilations done with @option{-ansi}. Alternate predefined macros
1475 such as @code{__unix__} and @code{__vax__} are also available, with or
1476 without @option{-ansi}.
1478 The @option{-ansi} option does not cause non-ISO programs to be
1479 rejected gratuitously. For that, @option{-pedantic} is required in
1480 addition to @option{-ansi}. @xref{Warning Options}.
1482 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1483 option is used. Some header files may notice this macro and refrain
1484 from declaring certain functions or defining certain macros that the
1485 ISO standard doesn't call for; this is to avoid interfering with any
1486 programs that might use these names for other things.
1488 Functions that would normally be built in but do not have semantics
1489 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1490 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1491 built-in functions provided by GCC}, for details of the functions
1496 Determine the language standard. @xref{Standards,,Language Standards
1497 Supported by GCC}, for details of these standard versions. This option
1498 is currently only supported when compiling C or C++.
1500 The compiler can accept several base standards, such as @samp{c90} or
1501 @samp{c++98}, and GNU dialects of those standards, such as
1502 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1503 compiler will accept all programs following that standard and those
1504 using GNU extensions that do not contradict it. For example,
1505 @samp{-std=c90} turns off certain features of GCC that are
1506 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1507 keywords, but not other GNU extensions that do not have a meaning in
1508 ISO C90, such as omitting the middle term of a @code{?:}
1509 expression. On the other hand, by specifying a GNU dialect of a
1510 standard, all features the compiler support are enabled, even when
1511 those features change the meaning of the base standard and some
1512 strict-conforming programs may be rejected. The particular standard
1513 is used by @option{-pedantic} to identify which features are GNU
1514 extensions given that version of the standard. For example
1515 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1516 comments, while @samp{-std=gnu99 -pedantic} would not.
1518 A value for this option must be provided; possible values are
1524 Support all ISO C90 programs (certain GNU extensions that conflict
1525 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1527 @item iso9899:199409
1528 ISO C90 as modified in amendment 1.
1534 ISO C99. Note that this standard is not yet fully supported; see
1535 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1536 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1539 ISO C1X, the draft of the next revision of the ISO C standard.
1540 Support is limited and experimental and features enabled by this
1541 option may be changed or removed if changed in or removed from the
1546 GNU dialect of ISO C90 (including some C99 features). This
1547 is the default for C code.
1551 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1552 this will become the default. The name @samp{gnu9x} is deprecated.
1555 GNU dialect of ISO C1X. Support is limited and experimental and
1556 features enabled by this option may be changed or removed if changed
1557 in or removed from the standard draft.
1560 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1564 GNU dialect of @option{-std=c++98}. This is the default for
1568 The working draft of the upcoming ISO C++0x standard. This option
1569 enables experimental features that are likely to be included in
1570 C++0x. The working draft is constantly changing, and any feature that is
1571 enabled by this flag may be removed from future versions of GCC if it is
1572 not part of the C++0x standard.
1575 GNU dialect of @option{-std=c++0x}. This option enables
1576 experimental features that may be removed in future versions of GCC.
1579 @item -fgnu89-inline
1580 @opindex fgnu89-inline
1581 The option @option{-fgnu89-inline} tells GCC to use the traditional
1582 GNU semantics for @code{inline} functions when in C99 mode.
1583 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1584 is accepted and ignored by GCC versions 4.1.3 up to but not including
1585 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1586 C99 mode. Using this option is roughly equivalent to adding the
1587 @code{gnu_inline} function attribute to all inline functions
1588 (@pxref{Function Attributes}).
1590 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1591 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1592 specifies the default behavior). This option was first supported in
1593 GCC 4.3. This option is not supported in @option{-std=c90} or
1594 @option{-std=gnu90} mode.
1596 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1597 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1598 in effect for @code{inline} functions. @xref{Common Predefined
1599 Macros,,,cpp,The C Preprocessor}.
1601 @item -aux-info @var{filename}
1603 Output to the given filename prototyped declarations for all functions
1604 declared and/or defined in a translation unit, including those in header
1605 files. This option is silently ignored in any language other than C@.
1607 Besides declarations, the file indicates, in comments, the origin of
1608 each declaration (source file and line), whether the declaration was
1609 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1610 @samp{O} for old, respectively, in the first character after the line
1611 number and the colon), and whether it came from a declaration or a
1612 definition (@samp{C} or @samp{F}, respectively, in the following
1613 character). In the case of function definitions, a K&R-style list of
1614 arguments followed by their declarations is also provided, inside
1615 comments, after the declaration.
1619 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1620 keyword, so that code can use these words as identifiers. You can use
1621 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1622 instead. @option{-ansi} implies @option{-fno-asm}.
1624 In C++, this switch only affects the @code{typeof} keyword, since
1625 @code{asm} and @code{inline} are standard keywords. You may want to
1626 use the @option{-fno-gnu-keywords} flag instead, which has the same
1627 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1628 switch only affects the @code{asm} and @code{typeof} keywords, since
1629 @code{inline} is a standard keyword in ISO C99.
1632 @itemx -fno-builtin-@var{function}
1633 @opindex fno-builtin
1634 @cindex built-in functions
1635 Don't recognize built-in functions that do not begin with
1636 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1637 functions provided by GCC}, for details of the functions affected,
1638 including those which are not built-in functions when @option{-ansi} or
1639 @option{-std} options for strict ISO C conformance are used because they
1640 do not have an ISO standard meaning.
1642 GCC normally generates special code to handle certain built-in functions
1643 more efficiently; for instance, calls to @code{alloca} may become single
1644 instructions that adjust the stack directly, and calls to @code{memcpy}
1645 may become inline copy loops. The resulting code is often both smaller
1646 and faster, but since the function calls no longer appear as such, you
1647 cannot set a breakpoint on those calls, nor can you change the behavior
1648 of the functions by linking with a different library. In addition,
1649 when a function is recognized as a built-in function, GCC may use
1650 information about that function to warn about problems with calls to
1651 that function, or to generate more efficient code, even if the
1652 resulting code still contains calls to that function. For example,
1653 warnings are given with @option{-Wformat} for bad calls to
1654 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1655 known not to modify global memory.
1657 With the @option{-fno-builtin-@var{function}} option
1658 only the built-in function @var{function} is
1659 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1660 function is named that is not built-in in this version of GCC, this
1661 option is ignored. There is no corresponding
1662 @option{-fbuiltin-@var{function}} option; if you wish to enable
1663 built-in functions selectively when using @option{-fno-builtin} or
1664 @option{-ffreestanding}, you may define macros such as:
1667 #define abs(n) __builtin_abs ((n))
1668 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1673 @cindex hosted environment
1675 Assert that compilation takes place in a hosted environment. This implies
1676 @option{-fbuiltin}. A hosted environment is one in which the
1677 entire standard library is available, and in which @code{main} has a return
1678 type of @code{int}. Examples are nearly everything except a kernel.
1679 This is equivalent to @option{-fno-freestanding}.
1681 @item -ffreestanding
1682 @opindex ffreestanding
1683 @cindex hosted environment
1685 Assert that compilation takes place in a freestanding environment. This
1686 implies @option{-fno-builtin}. A freestanding environment
1687 is one in which the standard library may not exist, and program startup may
1688 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1689 This is equivalent to @option{-fno-hosted}.
1691 @xref{Standards,,Language Standards Supported by GCC}, for details of
1692 freestanding and hosted environments.
1696 @cindex OpenMP parallel
1697 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1698 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1699 compiler generates parallel code according to the OpenMP Application
1700 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1701 implies @option{-pthread}, and thus is only supported on targets that
1702 have support for @option{-pthread}.
1704 @item -fms-extensions
1705 @opindex fms-extensions
1706 Accept some non-standard constructs used in Microsoft header files.
1708 In C++ code, this allows member names in structures to be similar
1709 to previous types declarations.
1718 Some cases of unnamed fields in structures and unions are only
1719 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1720 fields within structs/unions}, for details.
1722 @item -fplan9-extensions
1723 Accept some non-standard constructs used in Plan 9 code.
1725 This enables @option{-fms-extensions}, permits passing pointers to
1726 structures with anonymous fields to functions which expect pointers to
1727 elements of the type of the field, and permits referring to anonymous
1728 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1729 struct/union fields within structs/unions}, for details. This is only
1730 supported for C, not C++.
1734 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1735 options for strict ISO C conformance) implies @option{-trigraphs}.
1737 @item -no-integrated-cpp
1738 @opindex no-integrated-cpp
1739 Performs a compilation in two passes: preprocessing and compiling. This
1740 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1741 @option{-B} option. The user supplied compilation step can then add in
1742 an additional preprocessing step after normal preprocessing but before
1743 compiling. The default is to use the integrated cpp (internal cpp)
1745 The semantics of this option will change if "cc1", "cc1plus", and
1746 "cc1obj" are merged.
1748 @cindex traditional C language
1749 @cindex C language, traditional
1751 @itemx -traditional-cpp
1752 @opindex traditional-cpp
1753 @opindex traditional
1754 Formerly, these options caused GCC to attempt to emulate a pre-standard
1755 C compiler. They are now only supported with the @option{-E} switch.
1756 The preprocessor continues to support a pre-standard mode. See the GNU
1757 CPP manual for details.
1759 @item -fcond-mismatch
1760 @opindex fcond-mismatch
1761 Allow conditional expressions with mismatched types in the second and
1762 third arguments. The value of such an expression is void. This option
1763 is not supported for C++.
1765 @item -flax-vector-conversions
1766 @opindex flax-vector-conversions
1767 Allow implicit conversions between vectors with differing numbers of
1768 elements and/or incompatible element types. This option should not be
1771 @item -funsigned-char
1772 @opindex funsigned-char
1773 Let the type @code{char} be unsigned, like @code{unsigned char}.
1775 Each kind of machine has a default for what @code{char} should
1776 be. It is either like @code{unsigned char} by default or like
1777 @code{signed char} by default.
1779 Ideally, a portable program should always use @code{signed char} or
1780 @code{unsigned char} when it depends on the signedness of an object.
1781 But many programs have been written to use plain @code{char} and
1782 expect it to be signed, or expect it to be unsigned, depending on the
1783 machines they were written for. This option, and its inverse, let you
1784 make such a program work with the opposite default.
1786 The type @code{char} is always a distinct type from each of
1787 @code{signed char} or @code{unsigned char}, even though its behavior
1788 is always just like one of those two.
1791 @opindex fsigned-char
1792 Let the type @code{char} be signed, like @code{signed char}.
1794 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1795 the negative form of @option{-funsigned-char}. Likewise, the option
1796 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1798 @item -fsigned-bitfields
1799 @itemx -funsigned-bitfields
1800 @itemx -fno-signed-bitfields
1801 @itemx -fno-unsigned-bitfields
1802 @opindex fsigned-bitfields
1803 @opindex funsigned-bitfields
1804 @opindex fno-signed-bitfields
1805 @opindex fno-unsigned-bitfields
1806 These options control whether a bit-field is signed or unsigned, when the
1807 declaration does not use either @code{signed} or @code{unsigned}. By
1808 default, such a bit-field is signed, because this is consistent: the
1809 basic integer types such as @code{int} are signed types.
1812 @node C++ Dialect Options
1813 @section Options Controlling C++ Dialect
1815 @cindex compiler options, C++
1816 @cindex C++ options, command line
1817 @cindex options, C++
1818 This section describes the command-line options that are only meaningful
1819 for C++ programs; but you can also use most of the GNU compiler options
1820 regardless of what language your program is in. For example, you
1821 might compile a file @code{firstClass.C} like this:
1824 g++ -g -frepo -O -c firstClass.C
1828 In this example, only @option{-frepo} is an option meant
1829 only for C++ programs; you can use the other options with any
1830 language supported by GCC@.
1832 Here is a list of options that are @emph{only} for compiling C++ programs:
1836 @item -fabi-version=@var{n}
1837 @opindex fabi-version
1838 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1839 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1840 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1841 the version that conforms most closely to the C++ ABI specification.
1842 Therefore, the ABI obtained using version 0 will change as ABI bugs
1845 The default is version 2.
1847 Version 3 corrects an error in mangling a constant address as a
1850 Version 4 implements a standard mangling for vector types.
1852 Version 5 corrects the mangling of attribute const/volatile on
1853 function pointer types, decltype of a plain decl, and use of a
1854 function parameter in the declaration of another parameter.
1856 Version 6 corrects the promotion behavior of C++0x scoped enums.
1858 See also @option{-Wabi}.
1860 @item -fno-access-control
1861 @opindex fno-access-control
1862 Turn off all access checking. This switch is mainly useful for working
1863 around bugs in the access control code.
1867 Check that the pointer returned by @code{operator new} is non-null
1868 before attempting to modify the storage allocated. This check is
1869 normally unnecessary because the C++ standard specifies that
1870 @code{operator new} will only return @code{0} if it is declared
1871 @samp{throw()}, in which case the compiler will always check the
1872 return value even without this option. In all other cases, when
1873 @code{operator new} has a non-empty exception specification, memory
1874 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1875 @samp{new (nothrow)}.
1877 @item -fconserve-space
1878 @opindex fconserve-space
1879 Put uninitialized or runtime-initialized global variables into the
1880 common segment, as C does. This saves space in the executable at the
1881 cost of not diagnosing duplicate definitions. If you compile with this
1882 flag and your program mysteriously crashes after @code{main()} has
1883 completed, you may have an object that is being destroyed twice because
1884 two definitions were merged.
1886 This option is no longer useful on most targets, now that support has
1887 been added for putting variables into BSS without making them common.
1889 @item -fconstexpr-depth=@var{n}
1890 @opindex fconstexpr-depth
1891 Set the maximum nested evaluation depth for C++0x constexpr functions
1892 to @var{n}. A limit is needed to detect endless recursion during
1893 constant expression evaluation. The minimum specified by the standard
1896 @item -fno-deduce-init-list
1897 @opindex fno-deduce-init-list
1898 Disable deduction of a template type parameter as
1899 std::initializer_list from a brace-enclosed initializer list, i.e.
1902 template <class T> auto forward(T t) -> decltype (realfn (t))
1909 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1913 This option is present because this deduction is an extension to the
1914 current specification in the C++0x working draft, and there was
1915 some concern about potential overload resolution problems.
1917 @item -ffriend-injection
1918 @opindex ffriend-injection
1919 Inject friend functions into the enclosing namespace, so that they are
1920 visible outside the scope of the class in which they are declared.
1921 Friend functions were documented to work this way in the old Annotated
1922 C++ Reference Manual, and versions of G++ before 4.1 always worked
1923 that way. However, in ISO C++ a friend function which is not declared
1924 in an enclosing scope can only be found using argument dependent
1925 lookup. This option causes friends to be injected as they were in
1928 This option is for compatibility, and may be removed in a future
1931 @item -fno-elide-constructors
1932 @opindex fno-elide-constructors
1933 The C++ standard allows an implementation to omit creating a temporary
1934 which is only used to initialize another object of the same type.
1935 Specifying this option disables that optimization, and forces G++ to
1936 call the copy constructor in all cases.
1938 @item -fno-enforce-eh-specs
1939 @opindex fno-enforce-eh-specs
1940 Don't generate code to check for violation of exception specifications
1941 at runtime. This option violates the C++ standard, but may be useful
1942 for reducing code size in production builds, much like defining
1943 @samp{NDEBUG}. This does not give user code permission to throw
1944 exceptions in violation of the exception specifications; the compiler
1945 will still optimize based on the specifications, so throwing an
1946 unexpected exception will result in undefined behavior.
1949 @itemx -fno-for-scope
1951 @opindex fno-for-scope
1952 If @option{-ffor-scope} is specified, the scope of variables declared in
1953 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1954 as specified by the C++ standard.
1955 If @option{-fno-for-scope} is specified, the scope of variables declared in
1956 a @i{for-init-statement} extends to the end of the enclosing scope,
1957 as was the case in old versions of G++, and other (traditional)
1958 implementations of C++.
1960 The default if neither flag is given to follow the standard,
1961 but to allow and give a warning for old-style code that would
1962 otherwise be invalid, or have different behavior.
1964 @item -fno-gnu-keywords
1965 @opindex fno-gnu-keywords
1966 Do not recognize @code{typeof} as a keyword, so that code can use this
1967 word as an identifier. You can use the keyword @code{__typeof__} instead.
1968 @option{-ansi} implies @option{-fno-gnu-keywords}.
1970 @item -fno-implicit-templates
1971 @opindex fno-implicit-templates
1972 Never emit code for non-inline templates which are instantiated
1973 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1974 @xref{Template Instantiation}, for more information.
1976 @item -fno-implicit-inline-templates
1977 @opindex fno-implicit-inline-templates
1978 Don't emit code for implicit instantiations of inline templates, either.
1979 The default is to handle inlines differently so that compiles with and
1980 without optimization will need the same set of explicit instantiations.
1982 @item -fno-implement-inlines
1983 @opindex fno-implement-inlines
1984 To save space, do not emit out-of-line copies of inline functions
1985 controlled by @samp{#pragma implementation}. This will cause linker
1986 errors if these functions are not inlined everywhere they are called.
1988 @item -fms-extensions
1989 @opindex fms-extensions
1990 Disable pedantic warnings about constructs used in MFC, such as implicit
1991 int and getting a pointer to member function via non-standard syntax.
1993 @item -fno-nonansi-builtins
1994 @opindex fno-nonansi-builtins
1995 Disable built-in declarations of functions that are not mandated by
1996 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1997 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2000 @opindex fnothrow-opt
2001 Treat a @code{throw()} exception specification as though it were a
2002 @code{noexcept} specification to reduce or eliminate the text size
2003 overhead relative to a function with no exception specification. If
2004 the function has local variables of types with non-trivial
2005 destructors, the exception specification will actually make the
2006 function smaller because the EH cleanups for those variables can be
2007 optimized away. The semantic effect is that an exception thrown out of
2008 a function with such an exception specification will result in a call
2009 to @code{terminate} rather than @code{unexpected}.
2011 @item -fno-operator-names
2012 @opindex fno-operator-names
2013 Do not treat the operator name keywords @code{and}, @code{bitand},
2014 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2015 synonyms as keywords.
2017 @item -fno-optional-diags
2018 @opindex fno-optional-diags
2019 Disable diagnostics that the standard says a compiler does not need to
2020 issue. Currently, the only such diagnostic issued by G++ is the one for
2021 a name having multiple meanings within a class.
2024 @opindex fpermissive
2025 Downgrade some diagnostics about nonconformant code from errors to
2026 warnings. Thus, using @option{-fpermissive} will allow some
2027 nonconforming code to compile.
2029 @item -fno-pretty-templates
2030 @opindex fno-pretty-templates
2031 When an error message refers to a specialization of a function
2032 template, the compiler will normally print the signature of the
2033 template followed by the template arguments and any typedefs or
2034 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2035 rather than @code{void f(int)}) so that it's clear which template is
2036 involved. When an error message refers to a specialization of a class
2037 template, the compiler will omit any template arguments which match
2038 the default template arguments for that template. If either of these
2039 behaviors make it harder to understand the error message rather than
2040 easier, using @option{-fno-pretty-templates} will disable them.
2044 Enable automatic template instantiation at link time. This option also
2045 implies @option{-fno-implicit-templates}. @xref{Template
2046 Instantiation}, for more information.
2050 Disable generation of information about every class with virtual
2051 functions for use by the C++ runtime type identification features
2052 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2053 of the language, you can save some space by using this flag. Note that
2054 exception handling uses the same information, but it will generate it as
2055 needed. The @samp{dynamic_cast} operator can still be used for casts that
2056 do not require runtime type information, i.e.@: casts to @code{void *} or to
2057 unambiguous base classes.
2061 Emit statistics about front-end processing at the end of the compilation.
2062 This information is generally only useful to the G++ development team.
2064 @item -fstrict-enums
2065 @opindex fstrict-enums
2066 Allow the compiler to optimize using the assumption that a value of
2067 enumeration type can only be one of the values of the enumeration (as
2068 defined in the C++ standard; basically, a value which can be
2069 represented in the minimum number of bits needed to represent all the
2070 enumerators). This assumption may not be valid if the program uses a
2071 cast to convert an arbitrary integer value to the enumeration type.
2073 @item -ftemplate-depth=@var{n}
2074 @opindex ftemplate-depth
2075 Set the maximum instantiation depth for template classes to @var{n}.
2076 A limit on the template instantiation depth is needed to detect
2077 endless recursions during template class instantiation. ANSI/ISO C++
2078 conforming programs must not rely on a maximum depth greater than 17
2079 (changed to 1024 in C++0x). The default value is 900, as the compiler
2080 can run out of stack space before hitting 1024 in some situations.
2082 @item -fno-threadsafe-statics
2083 @opindex fno-threadsafe-statics
2084 Do not emit the extra code to use the routines specified in the C++
2085 ABI for thread-safe initialization of local statics. You can use this
2086 option to reduce code size slightly in code that doesn't need to be
2089 @item -fuse-cxa-atexit
2090 @opindex fuse-cxa-atexit
2091 Register destructors for objects with static storage duration with the
2092 @code{__cxa_atexit} function rather than the @code{atexit} function.
2093 This option is required for fully standards-compliant handling of static
2094 destructors, but will only work if your C library supports
2095 @code{__cxa_atexit}.
2097 @item -fno-use-cxa-get-exception-ptr
2098 @opindex fno-use-cxa-get-exception-ptr
2099 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2100 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2101 if the runtime routine is not available.
2103 @item -fvisibility-inlines-hidden
2104 @opindex fvisibility-inlines-hidden
2105 This switch declares that the user does not attempt to compare
2106 pointers to inline methods where the addresses of the two functions
2107 were taken in different shared objects.
2109 The effect of this is that GCC may, effectively, mark inline methods with
2110 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2111 appear in the export table of a DSO and do not require a PLT indirection
2112 when used within the DSO@. Enabling this option can have a dramatic effect
2113 on load and link times of a DSO as it massively reduces the size of the
2114 dynamic export table when the library makes heavy use of templates.
2116 The behavior of this switch is not quite the same as marking the
2117 methods as hidden directly, because it does not affect static variables
2118 local to the function or cause the compiler to deduce that
2119 the function is defined in only one shared object.
2121 You may mark a method as having a visibility explicitly to negate the
2122 effect of the switch for that method. For example, if you do want to
2123 compare pointers to a particular inline method, you might mark it as
2124 having default visibility. Marking the enclosing class with explicit
2125 visibility will have no effect.
2127 Explicitly instantiated inline methods are unaffected by this option
2128 as their linkage might otherwise cross a shared library boundary.
2129 @xref{Template Instantiation}.
2131 @item -fvisibility-ms-compat
2132 @opindex fvisibility-ms-compat
2133 This flag attempts to use visibility settings to make GCC's C++
2134 linkage model compatible with that of Microsoft Visual Studio.
2136 The flag makes these changes to GCC's linkage model:
2140 It sets the default visibility to @code{hidden}, like
2141 @option{-fvisibility=hidden}.
2144 Types, but not their members, are not hidden by default.
2147 The One Definition Rule is relaxed for types without explicit
2148 visibility specifications which are defined in more than one different
2149 shared object: those declarations are permitted if they would have
2150 been permitted when this option was not used.
2153 In new code it is better to use @option{-fvisibility=hidden} and
2154 export those classes which are intended to be externally visible.
2155 Unfortunately it is possible for code to rely, perhaps accidentally,
2156 on the Visual Studio behavior.
2158 Among the consequences of these changes are that static data members
2159 of the same type with the same name but defined in different shared
2160 objects will be different, so changing one will not change the other;
2161 and that pointers to function members defined in different shared
2162 objects may not compare equal. When this flag is given, it is a
2163 violation of the ODR to define types with the same name differently.
2167 Do not use weak symbol support, even if it is provided by the linker.
2168 By default, G++ will use weak symbols if they are available. This
2169 option exists only for testing, and should not be used by end-users;
2170 it will result in inferior code and has no benefits. This option may
2171 be removed in a future release of G++.
2175 Do not search for header files in the standard directories specific to
2176 C++, but do still search the other standard directories. (This option
2177 is used when building the C++ library.)
2180 In addition, these optimization, warning, and code generation options
2181 have meanings only for C++ programs:
2184 @item -fno-default-inline
2185 @opindex fno-default-inline
2186 Do not assume @samp{inline} for functions defined inside a class scope.
2187 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2188 functions will have linkage like inline functions; they just won't be
2191 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2194 Warn when G++ generates code that is probably not compatible with the
2195 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2196 all such cases, there are probably some cases that are not warned about,
2197 even though G++ is generating incompatible code. There may also be
2198 cases where warnings are emitted even though the code that is generated
2201 You should rewrite your code to avoid these warnings if you are
2202 concerned about the fact that code generated by G++ may not be binary
2203 compatible with code generated by other compilers.
2205 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2210 A template with a non-type template parameter of reference type is
2211 mangled incorrectly:
2214 template <int &> struct S @{@};
2218 This is fixed in @option{-fabi-version=3}.
2221 SIMD vector types declared using @code{__attribute ((vector_size))} are
2222 mangled in a non-standard way that does not allow for overloading of
2223 functions taking vectors of different sizes.
2225 The mangling is changed in @option{-fabi-version=4}.
2228 The known incompatibilities in @option{-fabi-version=1} include:
2233 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2234 pack data into the same byte as a base class. For example:
2237 struct A @{ virtual void f(); int f1 : 1; @};
2238 struct B : public A @{ int f2 : 1; @};
2242 In this case, G++ will place @code{B::f2} into the same byte
2243 as@code{A::f1}; other compilers will not. You can avoid this problem
2244 by explicitly padding @code{A} so that its size is a multiple of the
2245 byte size on your platform; that will cause G++ and other compilers to
2246 layout @code{B} identically.
2249 Incorrect handling of tail-padding for virtual bases. G++ does not use
2250 tail padding when laying out virtual bases. For example:
2253 struct A @{ virtual void f(); char c1; @};
2254 struct B @{ B(); char c2; @};
2255 struct C : public A, public virtual B @{@};
2259 In this case, G++ will not place @code{B} into the tail-padding for
2260 @code{A}; other compilers will. You can avoid this problem by
2261 explicitly padding @code{A} so that its size is a multiple of its
2262 alignment (ignoring virtual base classes); that will cause G++ and other
2263 compilers to layout @code{C} identically.
2266 Incorrect handling of bit-fields with declared widths greater than that
2267 of their underlying types, when the bit-fields appear in a union. For
2271 union U @{ int i : 4096; @};
2275 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2276 union too small by the number of bits in an @code{int}.
2279 Empty classes can be placed at incorrect offsets. For example:
2289 struct C : public B, public A @{@};
2293 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2294 it should be placed at offset zero. G++ mistakenly believes that the
2295 @code{A} data member of @code{B} is already at offset zero.
2298 Names of template functions whose types involve @code{typename} or
2299 template template parameters can be mangled incorrectly.
2302 template <typename Q>
2303 void f(typename Q::X) @{@}
2305 template <template <typename> class Q>
2306 void f(typename Q<int>::X) @{@}
2310 Instantiations of these templates may be mangled incorrectly.
2314 It also warns psABI related changes. The known psABI changes at this
2320 For SYSV/x86-64, when passing union with long double, it is changed to
2321 pass in memory as specified in psABI. For example:
2331 @code{union U} will always be passed in memory.
2335 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2336 @opindex Wctor-dtor-privacy
2337 @opindex Wno-ctor-dtor-privacy
2338 Warn when a class seems unusable because all the constructors or
2339 destructors in that class are private, and it has neither friends nor
2340 public static member functions.
2342 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2343 @opindex Wdelete-non-virtual-dtor
2344 @opindex Wno-delete-non-virtual-dtor
2345 Warn when @samp{delete} is used to destroy an instance of a class which
2346 has virtual functions and non-virtual destructor. It is unsafe to delete
2347 an instance of a derived class through a pointer to a base class if the
2348 base class does not have a virtual destructor. This warning is enabled
2351 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2353 @opindex Wno-narrowing
2354 With -std=c++0x, suppress the diagnostic required by the standard for
2355 narrowing conversions within @samp{@{ @}}, e.g.
2358 int i = @{ 2.2 @}; // error: narrowing from double to int
2361 This flag can be useful for compiling valid C++98 code in C++0x mode.
2363 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2365 @opindex Wno-noexcept
2366 Warn when a noexcept-expression evaluates to false because of a call
2367 to a function that does not have a non-throwing exception
2368 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2369 the compiler to never throw an exception.
2371 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2372 @opindex Wnon-virtual-dtor
2373 @opindex Wno-non-virtual-dtor
2374 Warn when a class has virtual functions and accessible non-virtual
2375 destructor, in which case it would be possible but unsafe to delete
2376 an instance of a derived class through a pointer to the base class.
2377 This warning is also enabled if -Weffc++ is specified.
2379 @item -Wreorder @r{(C++ and Objective-C++ only)}
2381 @opindex Wno-reorder
2382 @cindex reordering, warning
2383 @cindex warning for reordering of member initializers
2384 Warn when the order of member initializers given in the code does not
2385 match the order in which they must be executed. For instance:
2391 A(): j (0), i (1) @{ @}
2395 The compiler will rearrange the member initializers for @samp{i}
2396 and @samp{j} to match the declaration order of the members, emitting
2397 a warning to that effect. This warning is enabled by @option{-Wall}.
2400 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2403 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2406 Warn about violations of the following style guidelines from Scott Meyers'
2407 @cite{Effective C++} book:
2411 Item 11: Define a copy constructor and an assignment operator for classes
2412 with dynamically allocated memory.
2415 Item 12: Prefer initialization to assignment in constructors.
2418 Item 14: Make destructors virtual in base classes.
2421 Item 15: Have @code{operator=} return a reference to @code{*this}.
2424 Item 23: Don't try to return a reference when you must return an object.
2428 Also warn about violations of the following style guidelines from
2429 Scott Meyers' @cite{More Effective C++} book:
2433 Item 6: Distinguish between prefix and postfix forms of increment and
2434 decrement operators.
2437 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2441 When selecting this option, be aware that the standard library
2442 headers do not obey all of these guidelines; use @samp{grep -v}
2443 to filter out those warnings.
2445 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2446 @opindex Wstrict-null-sentinel
2447 @opindex Wno-strict-null-sentinel
2448 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2449 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2450 to @code{__null}. Although it is a null pointer constant not a null pointer,
2451 it is guaranteed to be of the same size as a pointer. But this use is
2452 not portable across different compilers.
2454 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2455 @opindex Wno-non-template-friend
2456 @opindex Wnon-template-friend
2457 Disable warnings when non-templatized friend functions are declared
2458 within a template. Since the advent of explicit template specification
2459 support in G++, if the name of the friend is an unqualified-id (i.e.,
2460 @samp{friend foo(int)}), the C++ language specification demands that the
2461 friend declare or define an ordinary, nontemplate function. (Section
2462 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2463 could be interpreted as a particular specialization of a templatized
2464 function. Because this non-conforming behavior is no longer the default
2465 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2466 check existing code for potential trouble spots and is on by default.
2467 This new compiler behavior can be turned off with
2468 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2469 but disables the helpful warning.
2471 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2472 @opindex Wold-style-cast
2473 @opindex Wno-old-style-cast
2474 Warn if an old-style (C-style) cast to a non-void type is used within
2475 a C++ program. The new-style casts (@samp{dynamic_cast},
2476 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2477 less vulnerable to unintended effects and much easier to search for.
2479 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2480 @opindex Woverloaded-virtual
2481 @opindex Wno-overloaded-virtual
2482 @cindex overloaded virtual function, warning
2483 @cindex warning for overloaded virtual function
2484 Warn when a function declaration hides virtual functions from a
2485 base class. For example, in:
2492 struct B: public A @{
2497 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2505 will fail to compile.
2507 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2508 @opindex Wno-pmf-conversions
2509 @opindex Wpmf-conversions
2510 Disable the diagnostic for converting a bound pointer to member function
2513 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2514 @opindex Wsign-promo
2515 @opindex Wno-sign-promo
2516 Warn when overload resolution chooses a promotion from unsigned or
2517 enumerated type to a signed type, over a conversion to an unsigned type of
2518 the same size. Previous versions of G++ would try to preserve
2519 unsignedness, but the standard mandates the current behavior.
2524 A& operator = (int);
2534 In this example, G++ will synthesize a default @samp{A& operator =
2535 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2538 @node Objective-C and Objective-C++ Dialect Options
2539 @section Options Controlling Objective-C and Objective-C++ Dialects
2541 @cindex compiler options, Objective-C and Objective-C++
2542 @cindex Objective-C and Objective-C++ options, command line
2543 @cindex options, Objective-C and Objective-C++
2544 (NOTE: This manual does not describe the Objective-C and Objective-C++
2545 languages themselves. @xref{Standards,,Language Standards
2546 Supported by GCC}, for references.)
2548 This section describes the command-line options that are only meaningful
2549 for Objective-C and Objective-C++ programs, but you can also use most of
2550 the language-independent GNU compiler options.
2551 For example, you might compile a file @code{some_class.m} like this:
2554 gcc -g -fgnu-runtime -O -c some_class.m
2558 In this example, @option{-fgnu-runtime} is an option meant only for
2559 Objective-C and Objective-C++ programs; you can use the other options with
2560 any language supported by GCC@.
2562 Note that since Objective-C is an extension of the C language, Objective-C
2563 compilations may also use options specific to the C front-end (e.g.,
2564 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2565 C++-specific options (e.g., @option{-Wabi}).
2567 Here is a list of options that are @emph{only} for compiling Objective-C
2568 and Objective-C++ programs:
2571 @item -fconstant-string-class=@var{class-name}
2572 @opindex fconstant-string-class
2573 Use @var{class-name} as the name of the class to instantiate for each
2574 literal string specified with the syntax @code{@@"@dots{}"}. The default
2575 class name is @code{NXConstantString} if the GNU runtime is being used, and
2576 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2577 @option{-fconstant-cfstrings} option, if also present, will override the
2578 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2579 to be laid out as constant CoreFoundation strings.
2582 @opindex fgnu-runtime
2583 Generate object code compatible with the standard GNU Objective-C
2584 runtime. This is the default for most types of systems.
2586 @item -fnext-runtime
2587 @opindex fnext-runtime
2588 Generate output compatible with the NeXT runtime. This is the default
2589 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2590 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2593 @item -fno-nil-receivers
2594 @opindex fno-nil-receivers
2595 Assume that all Objective-C message dispatches (@code{[receiver
2596 message:arg]}) in this translation unit ensure that the receiver is
2597 not @code{nil}. This allows for more efficient entry points in the
2598 runtime to be used. This option is only available in conjunction with
2599 the NeXT runtime and ABI version 0 or 1.
2601 @item -fobjc-abi-version=@var{n}
2602 @opindex fobjc-abi-version
2603 Use version @var{n} of the Objective-C ABI for the selected runtime.
2604 This option is currently supported only for the NeXT runtime. In that
2605 case, Version 0 is the traditional (32-bit) ABI without support for
2606 properties and other Objective-C 2.0 additions. Version 1 is the
2607 traditional (32-bit) ABI with support for properties and other
2608 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2609 nothing is specified, the default is Version 0 on 32-bit target
2610 machines, and Version 2 on 64-bit target machines.
2612 @item -fobjc-call-cxx-cdtors
2613 @opindex fobjc-call-cxx-cdtors
2614 For each Objective-C class, check if any of its instance variables is a
2615 C++ object with a non-trivial default constructor. If so, synthesize a
2616 special @code{- (id) .cxx_construct} instance method that will run
2617 non-trivial default constructors on any such instance variables, in order,
2618 and then return @code{self}. Similarly, check if any instance variable
2619 is a C++ object with a non-trivial destructor, and if so, synthesize a
2620 special @code{- (void) .cxx_destruct} method that will run
2621 all such default destructors, in reverse order.
2623 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2624 methods thusly generated will only operate on instance variables
2625 declared in the current Objective-C class, and not those inherited
2626 from superclasses. It is the responsibility of the Objective-C
2627 runtime to invoke all such methods in an object's inheritance
2628 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2629 by the runtime immediately after a new object instance is allocated;
2630 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2631 before the runtime deallocates an object instance.
2633 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2634 support for invoking the @code{- (id) .cxx_construct} and
2635 @code{- (void) .cxx_destruct} methods.
2637 @item -fobjc-direct-dispatch
2638 @opindex fobjc-direct-dispatch
2639 Allow fast jumps to the message dispatcher. On Darwin this is
2640 accomplished via the comm page.
2642 @item -fobjc-exceptions
2643 @opindex fobjc-exceptions
2644 Enable syntactic support for structured exception handling in
2645 Objective-C, similar to what is offered by C++ and Java. This option
2646 is required to use the Objective-C keywords @code{@@try},
2647 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2648 @code{@@synchronized}. This option is available with both the GNU
2649 runtime and the NeXT runtime (but not available in conjunction with
2650 the NeXT runtime on Mac OS X 10.2 and earlier).
2654 Enable garbage collection (GC) in Objective-C and Objective-C++
2655 programs. This option is only available with the NeXT runtime; the
2656 GNU runtime has a different garbage collection implementation that
2657 does not require special compiler flags.
2659 @item -fobjc-nilcheck
2660 @opindex fobjc-nilcheck
2661 For the NeXT runtime with version 2 of the ABI, check for a nil
2662 receiver in method invocations before doing the actual method call.
2663 This is the default and can be disabled using
2664 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2665 checked for nil in this way no matter what this flag is set to.
2666 Currently this flag does nothing when the GNU runtime, or an older
2667 version of the NeXT runtime ABI, is used.
2669 @item -fobjc-std=objc1
2671 Conform to the language syntax of Objective-C 1.0, the language
2672 recognized by GCC 4.0. This only affects the Objective-C additions to
2673 the C/C++ language; it does not affect conformance to C/C++ standards,
2674 which is controlled by the separate C/C++ dialect option flags. When
2675 this option is used with the Objective-C or Objective-C++ compiler,
2676 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2677 This is useful if you need to make sure that your Objective-C code can
2678 be compiled with older versions of GCC.
2680 @item -freplace-objc-classes
2681 @opindex freplace-objc-classes
2682 Emit a special marker instructing @command{ld(1)} not to statically link in
2683 the resulting object file, and allow @command{dyld(1)} to load it in at
2684 run time instead. This is used in conjunction with the Fix-and-Continue
2685 debugging mode, where the object file in question may be recompiled and
2686 dynamically reloaded in the course of program execution, without the need
2687 to restart the program itself. Currently, Fix-and-Continue functionality
2688 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2693 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2694 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2695 compile time) with static class references that get initialized at load time,
2696 which improves run-time performance. Specifying the @option{-fzero-link} flag
2697 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2698 to be retained. This is useful in Zero-Link debugging mode, since it allows
2699 for individual class implementations to be modified during program execution.
2700 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2701 regardless of command line options.
2705 Dump interface declarations for all classes seen in the source file to a
2706 file named @file{@var{sourcename}.decl}.
2708 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2709 @opindex Wassign-intercept
2710 @opindex Wno-assign-intercept
2711 Warn whenever an Objective-C assignment is being intercepted by the
2714 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2715 @opindex Wno-protocol
2717 If a class is declared to implement a protocol, a warning is issued for
2718 every method in the protocol that is not implemented by the class. The
2719 default behavior is to issue a warning for every method not explicitly
2720 implemented in the class, even if a method implementation is inherited
2721 from the superclass. If you use the @option{-Wno-protocol} option, then
2722 methods inherited from the superclass are considered to be implemented,
2723 and no warning is issued for them.
2725 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2727 @opindex Wno-selector
2728 Warn if multiple methods of different types for the same selector are
2729 found during compilation. The check is performed on the list of methods
2730 in the final stage of compilation. Additionally, a check is performed
2731 for each selector appearing in a @code{@@selector(@dots{})}
2732 expression, and a corresponding method for that selector has been found
2733 during compilation. Because these checks scan the method table only at
2734 the end of compilation, these warnings are not produced if the final
2735 stage of compilation is not reached, for example because an error is
2736 found during compilation, or because the @option{-fsyntax-only} option is
2739 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2740 @opindex Wstrict-selector-match
2741 @opindex Wno-strict-selector-match
2742 Warn if multiple methods with differing argument and/or return types are
2743 found for a given selector when attempting to send a message using this
2744 selector to a receiver of type @code{id} or @code{Class}. When this flag
2745 is off (which is the default behavior), the compiler will omit such warnings
2746 if any differences found are confined to types which share the same size
2749 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2750 @opindex Wundeclared-selector
2751 @opindex Wno-undeclared-selector
2752 Warn if a @code{@@selector(@dots{})} expression referring to an
2753 undeclared selector is found. A selector is considered undeclared if no
2754 method with that name has been declared before the
2755 @code{@@selector(@dots{})} expression, either explicitly in an
2756 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2757 an @code{@@implementation} section. This option always performs its
2758 checks as soon as a @code{@@selector(@dots{})} expression is found,
2759 while @option{-Wselector} only performs its checks in the final stage of
2760 compilation. This also enforces the coding style convention
2761 that methods and selectors must be declared before being used.
2763 @item -print-objc-runtime-info
2764 @opindex print-objc-runtime-info
2765 Generate C header describing the largest structure that is passed by
2770 @node Language Independent Options
2771 @section Options to Control Diagnostic Messages Formatting
2772 @cindex options to control diagnostics formatting
2773 @cindex diagnostic messages
2774 @cindex message formatting
2776 Traditionally, diagnostic messages have been formatted irrespective of
2777 the output device's aspect (e.g.@: its width, @dots{}). The options described
2778 below can be used to control the diagnostic messages formatting
2779 algorithm, e.g.@: how many characters per line, how often source location
2780 information should be reported. Right now, only the C++ front end can
2781 honor these options. However it is expected, in the near future, that
2782 the remaining front ends would be able to digest them correctly.
2785 @item -fmessage-length=@var{n}
2786 @opindex fmessage-length
2787 Try to format error messages so that they fit on lines of about @var{n}
2788 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2789 the front ends supported by GCC@. If @var{n} is zero, then no
2790 line-wrapping will be done; each error message will appear on a single
2793 @opindex fdiagnostics-show-location
2794 @item -fdiagnostics-show-location=once
2795 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2796 reporter to emit @emph{once} source location information; that is, in
2797 case the message is too long to fit on a single physical line and has to
2798 be wrapped, the source location won't be emitted (as prefix) again,
2799 over and over, in subsequent continuation lines. This is the default
2802 @item -fdiagnostics-show-location=every-line
2803 Only meaningful in line-wrapping mode. Instructs the diagnostic
2804 messages reporter to emit the same source location information (as
2805 prefix) for physical lines that result from the process of breaking
2806 a message which is too long to fit on a single line.
2808 @item -fno-diagnostics-show-option
2809 @opindex fno-diagnostics-show-option
2810 @opindex fdiagnostics-show-option
2811 By default, each diagnostic emitted includes text which indicates the
2812 command line option that directly controls the diagnostic (if such an
2813 option is known to the diagnostic machinery). Specifying the
2814 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2816 @item -Wcoverage-mismatch
2817 @opindex Wcoverage-mismatch
2818 Warn if feedback profiles do not match when using the
2819 @option{-fprofile-use} option.
2820 If a source file was changed between @option{-fprofile-gen} and
2821 @option{-fprofile-use}, the files with the profile feedback can fail
2822 to match the source file and GCC can not use the profile feedback
2823 information. By default, this warning is enabled and is treated as an
2824 error. @option{-Wno-coverage-mismatch} can be used to disable the
2825 warning or @option{-Wno-error=coverage-mismatch} can be used to
2826 disable the error. Disable the error for this warning can result in
2827 poorly optimized code, so disabling the error is useful only in the
2828 case of very minor changes such as bug fixes to an existing code-base.
2829 Completely disabling the warning is not recommended.
2833 @node Warning Options
2834 @section Options to Request or Suppress Warnings
2835 @cindex options to control warnings
2836 @cindex warning messages
2837 @cindex messages, warning
2838 @cindex suppressing warnings
2840 Warnings are diagnostic messages that report constructions which
2841 are not inherently erroneous but which are risky or suggest there
2842 may have been an error.
2844 The following language-independent options do not enable specific
2845 warnings but control the kinds of diagnostics produced by GCC.
2848 @cindex syntax checking
2850 @opindex fsyntax-only
2851 Check the code for syntax errors, but don't do anything beyond that.
2853 @item -fmax-errors=@var{n}
2854 @opindex fmax-errors
2855 Limits the maximum number of error messages to @var{n}, at which point
2856 GCC bails out rather than attempting to continue processing the source
2857 code. If @var{n} is 0 (the default), there is no limit on the number
2858 of error messages produced. If @option{-Wfatal-errors} is also
2859 specified, then @option{-Wfatal-errors} takes precedence over this
2864 Inhibit all warning messages.
2869 Make all warnings into errors.
2874 Make the specified warning into an error. The specifier for a warning
2875 is appended, for example @option{-Werror=switch} turns the warnings
2876 controlled by @option{-Wswitch} into errors. This switch takes a
2877 negative form, to be used to negate @option{-Werror} for specific
2878 warnings, for example @option{-Wno-error=switch} makes
2879 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2882 The warning message for each controllable warning includes the
2883 option which controls the warning. That option can then be used with
2884 @option{-Werror=} and @option{-Wno-error=} as described above.
2885 (Printing of the option in the warning message can be disabled using the
2886 @option{-fno-diagnostics-show-option} flag.)
2888 Note that specifying @option{-Werror=}@var{foo} automatically implies
2889 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2892 @item -Wfatal-errors
2893 @opindex Wfatal-errors
2894 @opindex Wno-fatal-errors
2895 This option causes the compiler to abort compilation on the first error
2896 occurred rather than trying to keep going and printing further error
2901 You can request many specific warnings with options beginning
2902 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2903 implicit declarations. Each of these specific warning options also
2904 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2905 example, @option{-Wno-implicit}. This manual lists only one of the
2906 two forms, whichever is not the default. For further,
2907 language-specific options also refer to @ref{C++ Dialect Options} and
2908 @ref{Objective-C and Objective-C++ Dialect Options}.
2910 When an unrecognized warning option is requested (e.g.,
2911 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2912 that the option is not recognized. However, if the @option{-Wno-} form
2913 is used, the behavior is slightly different: No diagnostic will be
2914 produced for @option{-Wno-unknown-warning} unless other diagnostics
2915 are being produced. This allows the use of new @option{-Wno-} options
2916 with old compilers, but if something goes wrong, the compiler will
2917 warn that an unrecognized option was used.
2922 Issue all the warnings demanded by strict ISO C and ISO C++;
2923 reject all programs that use forbidden extensions, and some other
2924 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2925 version of the ISO C standard specified by any @option{-std} option used.
2927 Valid ISO C and ISO C++ programs should compile properly with or without
2928 this option (though a rare few will require @option{-ansi} or a
2929 @option{-std} option specifying the required version of ISO C)@. However,
2930 without this option, certain GNU extensions and traditional C and C++
2931 features are supported as well. With this option, they are rejected.
2933 @option{-pedantic} does not cause warning messages for use of the
2934 alternate keywords whose names begin and end with @samp{__}. Pedantic
2935 warnings are also disabled in the expression that follows
2936 @code{__extension__}. However, only system header files should use
2937 these escape routes; application programs should avoid them.
2938 @xref{Alternate Keywords}.
2940 Some users try to use @option{-pedantic} to check programs for strict ISO
2941 C conformance. They soon find that it does not do quite what they want:
2942 it finds some non-ISO practices, but not all---only those for which
2943 ISO C @emph{requires} a diagnostic, and some others for which
2944 diagnostics have been added.
2946 A feature to report any failure to conform to ISO C might be useful in
2947 some instances, but would require considerable additional work and would
2948 be quite different from @option{-pedantic}. We don't have plans to
2949 support such a feature in the near future.
2951 Where the standard specified with @option{-std} represents a GNU
2952 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2953 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2954 extended dialect is based. Warnings from @option{-pedantic} are given
2955 where they are required by the base standard. (It would not make sense
2956 for such warnings to be given only for features not in the specified GNU
2957 C dialect, since by definition the GNU dialects of C include all
2958 features the compiler supports with the given option, and there would be
2959 nothing to warn about.)
2961 @item -pedantic-errors
2962 @opindex pedantic-errors
2963 Like @option{-pedantic}, except that errors are produced rather than
2969 This enables all the warnings about constructions that some users
2970 consider questionable, and that are easy to avoid (or modify to
2971 prevent the warning), even in conjunction with macros. This also
2972 enables some language-specific warnings described in @ref{C++ Dialect
2973 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2975 @option{-Wall} turns on the following warning flags:
2977 @gccoptlist{-Waddress @gol
2978 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2980 -Wchar-subscripts @gol
2981 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2982 -Wimplicit-int @r{(C and Objective-C only)} @gol
2983 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2986 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2987 -Wmaybe-uninitialized @gol
2988 -Wmissing-braces @gol
2994 -Wsequence-point @gol
2995 -Wsign-compare @r{(only in C++)} @gol
2996 -Wstrict-aliasing @gol
2997 -Wstrict-overflow=1 @gol
3000 -Wuninitialized @gol
3001 -Wunknown-pragmas @gol
3002 -Wunused-function @gol
3005 -Wunused-variable @gol
3006 -Wvolatile-register-var @gol
3009 Note that some warning flags are not implied by @option{-Wall}. Some of
3010 them warn about constructions that users generally do not consider
3011 questionable, but which occasionally you might wish to check for;
3012 others warn about constructions that are necessary or hard to avoid in
3013 some cases, and there is no simple way to modify the code to suppress
3014 the warning. Some of them are enabled by @option{-Wextra} but many of
3015 them must be enabled individually.
3021 This enables some extra warning flags that are not enabled by
3022 @option{-Wall}. (This option used to be called @option{-W}. The older
3023 name is still supported, but the newer name is more descriptive.)
3025 @gccoptlist{-Wclobbered @gol
3027 -Wignored-qualifiers @gol
3028 -Wmissing-field-initializers @gol
3029 -Wmissing-parameter-type @r{(C only)} @gol
3030 -Wold-style-declaration @r{(C only)} @gol
3031 -Woverride-init @gol
3034 -Wuninitialized @gol
3035 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3036 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3039 The option @option{-Wextra} also prints warning messages for the
3045 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3046 @samp{>}, or @samp{>=}.
3049 (C++ only) An enumerator and a non-enumerator both appear in a
3050 conditional expression.
3053 (C++ only) Ambiguous virtual bases.
3056 (C++ only) Subscripting an array which has been declared @samp{register}.
3059 (C++ only) Taking the address of a variable which has been declared
3063 (C++ only) A base class is not initialized in a derived class' copy
3068 @item -Wchar-subscripts
3069 @opindex Wchar-subscripts
3070 @opindex Wno-char-subscripts
3071 Warn if an array subscript has type @code{char}. This is a common cause
3072 of error, as programmers often forget that this type is signed on some
3074 This warning is enabled by @option{-Wall}.
3078 @opindex Wno-comment
3079 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3080 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3081 This warning is enabled by @option{-Wall}.
3084 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3086 Suppress warning messages emitted by @code{#warning} directives.
3088 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3089 @opindex Wdouble-promotion
3090 @opindex Wno-double-promotion
3091 Give a warning when a value of type @code{float} is implicitly
3092 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3093 floating-point unit implement @code{float} in hardware, but emulate
3094 @code{double} in software. On such a machine, doing computations
3095 using @code{double} values is much more expensive because of the
3096 overhead required for software emulation.
3098 It is easy to accidentally do computations with @code{double} because
3099 floating-point literals are implicitly of type @code{double}. For
3103 float area(float radius)
3105 return 3.14159 * radius * radius;
3109 the compiler will perform the entire computation with @code{double}
3110 because the floating-point literal is a @code{double}.
3115 @opindex ffreestanding
3116 @opindex fno-builtin
3117 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3118 the arguments supplied have types appropriate to the format string
3119 specified, and that the conversions specified in the format string make
3120 sense. This includes standard functions, and others specified by format
3121 attributes (@pxref{Function Attributes}), in the @code{printf},
3122 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3123 not in the C standard) families (or other target-specific families).
3124 Which functions are checked without format attributes having been
3125 specified depends on the standard version selected, and such checks of
3126 functions without the attribute specified are disabled by
3127 @option{-ffreestanding} or @option{-fno-builtin}.
3129 The formats are checked against the format features supported by GNU
3130 libc version 2.2. These include all ISO C90 and C99 features, as well
3131 as features from the Single Unix Specification and some BSD and GNU
3132 extensions. Other library implementations may not support all these
3133 features; GCC does not support warning about features that go beyond a
3134 particular library's limitations. However, if @option{-pedantic} is used
3135 with @option{-Wformat}, warnings will be given about format features not
3136 in the selected standard version (but not for @code{strfmon} formats,
3137 since those are not in any version of the C standard). @xref{C Dialect
3138 Options,,Options Controlling C Dialect}.
3140 Since @option{-Wformat} also checks for null format arguments for
3141 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3143 @option{-Wformat} is included in @option{-Wall}. For more control over some
3144 aspects of format checking, the options @option{-Wformat-y2k},
3145 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3146 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3147 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3150 @opindex Wformat-y2k
3151 @opindex Wno-format-y2k
3152 If @option{-Wformat} is specified, also warn about @code{strftime}
3153 formats which may yield only a two-digit year.
3155 @item -Wno-format-contains-nul
3156 @opindex Wno-format-contains-nul
3157 @opindex Wformat-contains-nul
3158 If @option{-Wformat} is specified, do not warn about format strings that
3161 @item -Wno-format-extra-args
3162 @opindex Wno-format-extra-args
3163 @opindex Wformat-extra-args
3164 If @option{-Wformat} is specified, do not warn about excess arguments to a
3165 @code{printf} or @code{scanf} format function. The C standard specifies
3166 that such arguments are ignored.
3168 Where the unused arguments lie between used arguments that are
3169 specified with @samp{$} operand number specifications, normally
3170 warnings are still given, since the implementation could not know what
3171 type to pass to @code{va_arg} to skip the unused arguments. However,
3172 in the case of @code{scanf} formats, this option will suppress the
3173 warning if the unused arguments are all pointers, since the Single
3174 Unix Specification says that such unused arguments are allowed.
3176 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3177 @opindex Wno-format-zero-length
3178 @opindex Wformat-zero-length
3179 If @option{-Wformat} is specified, do not warn about zero-length formats.
3180 The C standard specifies that zero-length formats are allowed.
3182 @item -Wformat-nonliteral
3183 @opindex Wformat-nonliteral
3184 @opindex Wno-format-nonliteral
3185 If @option{-Wformat} is specified, also warn if the format string is not a
3186 string literal and so cannot be checked, unless the format function
3187 takes its format arguments as a @code{va_list}.
3189 @item -Wformat-security
3190 @opindex Wformat-security
3191 @opindex Wno-format-security
3192 If @option{-Wformat} is specified, also warn about uses of format
3193 functions that represent possible security problems. At present, this
3194 warns about calls to @code{printf} and @code{scanf} functions where the
3195 format string is not a string literal and there are no format arguments,
3196 as in @code{printf (foo);}. This may be a security hole if the format
3197 string came from untrusted input and contains @samp{%n}. (This is
3198 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3199 in future warnings may be added to @option{-Wformat-security} that are not
3200 included in @option{-Wformat-nonliteral}.)
3204 @opindex Wno-format=2
3205 Enable @option{-Wformat} plus format checks not included in
3206 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3207 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3209 @item -Wnonnull @r{(C and Objective-C only)}
3211 @opindex Wno-nonnull
3212 Warn about passing a null pointer for arguments marked as
3213 requiring a non-null value by the @code{nonnull} function attribute.
3215 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3216 can be disabled with the @option{-Wno-nonnull} option.
3218 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3220 @opindex Wno-init-self
3221 Warn about uninitialized variables which are initialized with themselves.
3222 Note this option can only be used with the @option{-Wuninitialized} option.
3224 For example, GCC will warn about @code{i} being uninitialized in the
3225 following snippet only when @option{-Winit-self} has been specified:
3236 @item -Wimplicit-int @r{(C and Objective-C only)}
3237 @opindex Wimplicit-int
3238 @opindex Wno-implicit-int
3239 Warn when a declaration does not specify a type.
3240 This warning is enabled by @option{-Wall}.
3242 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3243 @opindex Wimplicit-function-declaration
3244 @opindex Wno-implicit-function-declaration
3245 Give a warning whenever a function is used before being declared. In
3246 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3247 enabled by default and it is made into an error by
3248 @option{-pedantic-errors}. This warning is also enabled by
3251 @item -Wimplicit @r{(C and Objective-C only)}
3253 @opindex Wno-implicit
3254 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3255 This warning is enabled by @option{-Wall}.
3257 @item -Wignored-qualifiers @r{(C and C++ only)}
3258 @opindex Wignored-qualifiers
3259 @opindex Wno-ignored-qualifiers
3260 Warn if the return type of a function has a type qualifier
3261 such as @code{const}. For ISO C such a type qualifier has no effect,
3262 since the value returned by a function is not an lvalue.
3263 For C++, the warning is only emitted for scalar types or @code{void}.
3264 ISO C prohibits qualified @code{void} return types on function
3265 definitions, so such return types always receive a warning
3266 even without this option.
3268 This warning is also enabled by @option{-Wextra}.
3273 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3274 a function with external linkage, returning int, taking either zero
3275 arguments, two, or three arguments of appropriate types. This warning
3276 is enabled by default in C++ and is enabled by either @option{-Wall}
3277 or @option{-pedantic}.
3279 @item -Wmissing-braces
3280 @opindex Wmissing-braces
3281 @opindex Wno-missing-braces
3282 Warn if an aggregate or union initializer is not fully bracketed. In
3283 the following example, the initializer for @samp{a} is not fully
3284 bracketed, but that for @samp{b} is fully bracketed.
3287 int a[2][2] = @{ 0, 1, 2, 3 @};
3288 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3291 This warning is enabled by @option{-Wall}.
3293 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3294 @opindex Wmissing-include-dirs
3295 @opindex Wno-missing-include-dirs
3296 Warn if a user-supplied include directory does not exist.
3299 @opindex Wparentheses
3300 @opindex Wno-parentheses
3301 Warn if parentheses are omitted in certain contexts, such
3302 as when there is an assignment in a context where a truth value
3303 is expected, or when operators are nested whose precedence people
3304 often get confused about.
3306 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3307 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3308 interpretation from that of ordinary mathematical notation.
3310 Also warn about constructions where there may be confusion to which
3311 @code{if} statement an @code{else} branch belongs. Here is an example of
3326 In C/C++, every @code{else} branch belongs to the innermost possible
3327 @code{if} statement, which in this example is @code{if (b)}. This is
3328 often not what the programmer expected, as illustrated in the above
3329 example by indentation the programmer chose. When there is the
3330 potential for this confusion, GCC will issue a warning when this flag
3331 is specified. To eliminate the warning, add explicit braces around
3332 the innermost @code{if} statement so there is no way the @code{else}
3333 could belong to the enclosing @code{if}. The resulting code would
3350 Also warn for dangerous uses of the
3351 ?: with omitted middle operand GNU extension. When the condition
3352 in the ?: operator is a boolean expression the omitted value will
3353 be always 1. Often the user expects it to be a value computed
3354 inside the conditional expression instead.
3356 This warning is enabled by @option{-Wall}.
3358 @item -Wsequence-point
3359 @opindex Wsequence-point
3360 @opindex Wno-sequence-point
3361 Warn about code that may have undefined semantics because of violations
3362 of sequence point rules in the C and C++ standards.
3364 The C and C++ standards defines the order in which expressions in a C/C++
3365 program are evaluated in terms of @dfn{sequence points}, which represent
3366 a partial ordering between the execution of parts of the program: those
3367 executed before the sequence point, and those executed after it. These
3368 occur after the evaluation of a full expression (one which is not part
3369 of a larger expression), after the evaluation of the first operand of a
3370 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3371 function is called (but after the evaluation of its arguments and the
3372 expression denoting the called function), and in certain other places.
3373 Other than as expressed by the sequence point rules, the order of
3374 evaluation of subexpressions of an expression is not specified. All
3375 these rules describe only a partial order rather than a total order,
3376 since, for example, if two functions are called within one expression
3377 with no sequence point between them, the order in which the functions
3378 are called is not specified. However, the standards committee have
3379 ruled that function calls do not overlap.
3381 It is not specified when between sequence points modifications to the
3382 values of objects take effect. Programs whose behavior depends on this
3383 have undefined behavior; the C and C++ standards specify that ``Between
3384 the previous and next sequence point an object shall have its stored
3385 value modified at most once by the evaluation of an expression.
3386 Furthermore, the prior value shall be read only to determine the value
3387 to be stored.''. If a program breaks these rules, the results on any
3388 particular implementation are entirely unpredictable.
3390 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3391 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3392 diagnosed by this option, and it may give an occasional false positive
3393 result, but in general it has been found fairly effective at detecting
3394 this sort of problem in programs.
3396 The standard is worded confusingly, therefore there is some debate
3397 over the precise meaning of the sequence point rules in subtle cases.
3398 Links to discussions of the problem, including proposed formal
3399 definitions, may be found on the GCC readings page, at
3400 @uref{http://gcc.gnu.org/@/readings.html}.
3402 This warning is enabled by @option{-Wall} for C and C++.
3405 @opindex Wreturn-type
3406 @opindex Wno-return-type
3407 Warn whenever a function is defined with a return-type that defaults
3408 to @code{int}. Also warn about any @code{return} statement with no
3409 return-value in a function whose return-type is not @code{void}
3410 (falling off the end of the function body is considered returning
3411 without a value), and about a @code{return} statement with an
3412 expression in a function whose return-type is @code{void}.
3414 For C++, a function without return type always produces a diagnostic
3415 message, even when @option{-Wno-return-type} is specified. The only
3416 exceptions are @samp{main} and functions defined in system headers.
3418 This warning is enabled by @option{-Wall}.
3423 Warn whenever a @code{switch} statement has an index of enumerated type
3424 and lacks a @code{case} for one or more of the named codes of that
3425 enumeration. (The presence of a @code{default} label prevents this
3426 warning.) @code{case} labels outside the enumeration range also
3427 provoke warnings when this option is used (even if there is a
3428 @code{default} label).
3429 This warning is enabled by @option{-Wall}.
3431 @item -Wswitch-default
3432 @opindex Wswitch-default
3433 @opindex Wno-switch-default
3434 Warn whenever a @code{switch} statement does not have a @code{default}
3438 @opindex Wswitch-enum
3439 @opindex Wno-switch-enum
3440 Warn whenever a @code{switch} statement has an index of enumerated type
3441 and lacks a @code{case} for one or more of the named codes of that
3442 enumeration. @code{case} labels outside the enumeration range also
3443 provoke warnings when this option is used. The only difference
3444 between @option{-Wswitch} and this option is that this option gives a
3445 warning about an omitted enumeration code even if there is a
3446 @code{default} label.
3448 @item -Wsync-nand @r{(C and C++ only)}
3450 @opindex Wno-sync-nand
3451 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3452 built-in functions are used. These functions changed semantics in GCC 4.4.
3456 @opindex Wno-trigraphs
3457 Warn if any trigraphs are encountered that might change the meaning of
3458 the program (trigraphs within comments are not warned about).
3459 This warning is enabled by @option{-Wall}.
3461 @item -Wunused-but-set-parameter
3462 @opindex Wunused-but-set-parameter
3463 @opindex Wno-unused-but-set-parameter
3464 Warn whenever a function parameter is assigned to, but otherwise unused
3465 (aside from its declaration).
3467 To suppress this warning use the @samp{unused} attribute
3468 (@pxref{Variable Attributes}).
3470 This warning is also enabled by @option{-Wunused} together with
3473 @item -Wunused-but-set-variable
3474 @opindex Wunused-but-set-variable
3475 @opindex Wno-unused-but-set-variable
3476 Warn whenever a local variable is assigned to, but otherwise unused
3477 (aside from its declaration).
3478 This warning is enabled by @option{-Wall}.
3480 To suppress this warning use the @samp{unused} attribute
3481 (@pxref{Variable Attributes}).
3483 This warning is also enabled by @option{-Wunused}, which is enabled
3486 @item -Wunused-function
3487 @opindex Wunused-function
3488 @opindex Wno-unused-function
3489 Warn whenever a static function is declared but not defined or a
3490 non-inline static function is unused.
3491 This warning is enabled by @option{-Wall}.
3493 @item -Wunused-label
3494 @opindex Wunused-label
3495 @opindex Wno-unused-label
3496 Warn whenever a label is declared but not used.
3497 This warning is enabled by @option{-Wall}.
3499 To suppress this warning use the @samp{unused} attribute
3500 (@pxref{Variable Attributes}).
3502 @item -Wunused-parameter
3503 @opindex Wunused-parameter
3504 @opindex Wno-unused-parameter
3505 Warn whenever a function parameter is unused aside from its declaration.
3507 To suppress this warning use the @samp{unused} attribute
3508 (@pxref{Variable Attributes}).
3510 @item -Wno-unused-result
3511 @opindex Wunused-result
3512 @opindex Wno-unused-result
3513 Do not warn if a caller of a function marked with attribute
3514 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3515 its return value. The default is @option{-Wunused-result}.
3517 @item -Wunused-variable
3518 @opindex Wunused-variable
3519 @opindex Wno-unused-variable
3520 Warn whenever a local variable or non-constant static variable is unused
3521 aside from its declaration.
3522 This warning is enabled by @option{-Wall}.
3524 To suppress this warning use the @samp{unused} attribute
3525 (@pxref{Variable Attributes}).
3527 @item -Wunused-value
3528 @opindex Wunused-value
3529 @opindex Wno-unused-value
3530 Warn whenever a statement computes a result that is explicitly not
3531 used. To suppress this warning cast the unused expression to
3532 @samp{void}. This includes an expression-statement or the left-hand
3533 side of a comma expression that contains no side effects. For example,
3534 an expression such as @samp{x[i,j]} will cause a warning, while
3535 @samp{x[(void)i,j]} will not.
3537 This warning is enabled by @option{-Wall}.
3542 All the above @option{-Wunused} options combined.
3544 In order to get a warning about an unused function parameter, you must
3545 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3546 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3548 @item -Wuninitialized
3549 @opindex Wuninitialized
3550 @opindex Wno-uninitialized
3551 Warn if an automatic variable is used without first being initialized
3552 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3553 warn if a non-static reference or non-static @samp{const} member
3554 appears in a class without constructors.
3556 If you want to warn about code which uses the uninitialized value of the
3557 variable in its own initializer, use the @option{-Winit-self} option.
3559 These warnings occur for individual uninitialized or clobbered
3560 elements of structure, union or array variables as well as for
3561 variables which are uninitialized or clobbered as a whole. They do
3562 not occur for variables or elements declared @code{volatile}. Because
3563 these warnings depend on optimization, the exact variables or elements
3564 for which there are warnings will depend on the precise optimization
3565 options and version of GCC used.
3567 Note that there may be no warning about a variable that is used only
3568 to compute a value that itself is never used, because such
3569 computations may be deleted by data flow analysis before the warnings
3572 @item -Wmaybe-uninitialized
3573 @opindex Wmaybe-uninitialized
3574 @opindex Wno-maybe-uninitialized
3575 For an automatic variable, if there exists a path from the function
3576 entry to a use of the variable that is initialized, but there exist
3577 some other paths the variable is not initialized, the compiler will
3578 emit a warning if it can not prove the uninitialized paths do not
3579 happen at runtime. These warnings are made optional because GCC is
3580 not smart enough to see all the reasons why the code might be correct
3581 despite appearing to have an error. Here is one example of how
3602 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3603 always initialized, but GCC doesn't know this. To suppress the
3604 warning, the user needs to provide a default case with assert(0) or
3607 @cindex @code{longjmp} warnings
3608 This option also warns when a non-volatile automatic variable might be
3609 changed by a call to @code{longjmp}. These warnings as well are possible
3610 only in optimizing compilation.
3612 The compiler sees only the calls to @code{setjmp}. It cannot know
3613 where @code{longjmp} will be called; in fact, a signal handler could
3614 call it at any point in the code. As a result, you may get a warning
3615 even when there is in fact no problem because @code{longjmp} cannot
3616 in fact be called at the place which would cause a problem.
3618 Some spurious warnings can be avoided if you declare all the functions
3619 you use that never return as @code{noreturn}. @xref{Function
3622 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3624 @item -Wunknown-pragmas
3625 @opindex Wunknown-pragmas
3626 @opindex Wno-unknown-pragmas
3627 @cindex warning for unknown pragmas
3628 @cindex unknown pragmas, warning
3629 @cindex pragmas, warning of unknown
3630 Warn when a #pragma directive is encountered which is not understood by
3631 GCC@. If this command line option is used, warnings will even be issued
3632 for unknown pragmas in system header files. This is not the case if
3633 the warnings were only enabled by the @option{-Wall} command line option.
3636 @opindex Wno-pragmas
3638 Do not warn about misuses of pragmas, such as incorrect parameters,
3639 invalid syntax, or conflicts between pragmas. See also
3640 @samp{-Wunknown-pragmas}.
3642 @item -Wstrict-aliasing
3643 @opindex Wstrict-aliasing
3644 @opindex Wno-strict-aliasing
3645 This option is only active when @option{-fstrict-aliasing} is active.
3646 It warns about code which might break the strict aliasing rules that the
3647 compiler is using for optimization. The warning does not catch all
3648 cases, but does attempt to catch the more common pitfalls. It is
3649 included in @option{-Wall}.
3650 It is equivalent to @option{-Wstrict-aliasing=3}
3652 @item -Wstrict-aliasing=n
3653 @opindex Wstrict-aliasing=n
3654 @opindex Wno-strict-aliasing=n
3655 This option is only active when @option{-fstrict-aliasing} is active.
3656 It warns about code which might break the strict aliasing rules that the
3657 compiler is using for optimization.
3658 Higher levels correspond to higher accuracy (fewer false positives).
3659 Higher levels also correspond to more effort, similar to the way -O works.
3660 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3663 Level 1: Most aggressive, quick, least accurate.
3664 Possibly useful when higher levels
3665 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3666 false negatives. However, it has many false positives.
3667 Warns for all pointer conversions between possibly incompatible types,
3668 even if never dereferenced. Runs in the frontend only.
3670 Level 2: Aggressive, quick, not too precise.
3671 May still have many false positives (not as many as level 1 though),
3672 and few false negatives (but possibly more than level 1).
3673 Unlike level 1, it only warns when an address is taken. Warns about
3674 incomplete types. Runs in the frontend only.
3676 Level 3 (default for @option{-Wstrict-aliasing}):
3677 Should have very few false positives and few false
3678 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3679 Takes care of the common pun+dereference pattern in the frontend:
3680 @code{*(int*)&some_float}.
3681 If optimization is enabled, it also runs in the backend, where it deals
3682 with multiple statement cases using flow-sensitive points-to information.
3683 Only warns when the converted pointer is dereferenced.
3684 Does not warn about incomplete types.
3686 @item -Wstrict-overflow
3687 @itemx -Wstrict-overflow=@var{n}
3688 @opindex Wstrict-overflow
3689 @opindex Wno-strict-overflow
3690 This option is only active when @option{-fstrict-overflow} is active.
3691 It warns about cases where the compiler optimizes based on the
3692 assumption that signed overflow does not occur. Note that it does not
3693 warn about all cases where the code might overflow: it only warns
3694 about cases where the compiler implements some optimization. Thus
3695 this warning depends on the optimization level.
3697 An optimization which assumes that signed overflow does not occur is
3698 perfectly safe if the values of the variables involved are such that
3699 overflow never does, in fact, occur. Therefore this warning can
3700 easily give a false positive: a warning about code which is not
3701 actually a problem. To help focus on important issues, several
3702 warning levels are defined. No warnings are issued for the use of
3703 undefined signed overflow when estimating how many iterations a loop
3704 will require, in particular when determining whether a loop will be
3708 @item -Wstrict-overflow=1
3709 Warn about cases which are both questionable and easy to avoid. For
3710 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3711 compiler will simplify this to @code{1}. This level of
3712 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3713 are not, and must be explicitly requested.
3715 @item -Wstrict-overflow=2
3716 Also warn about other cases where a comparison is simplified to a
3717 constant. For example: @code{abs (x) >= 0}. This can only be
3718 simplified when @option{-fstrict-overflow} is in effect, because
3719 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3720 zero. @option{-Wstrict-overflow} (with no level) is the same as
3721 @option{-Wstrict-overflow=2}.
3723 @item -Wstrict-overflow=3
3724 Also warn about other cases where a comparison is simplified. For
3725 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3727 @item -Wstrict-overflow=4
3728 Also warn about other simplifications not covered by the above cases.
3729 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3731 @item -Wstrict-overflow=5
3732 Also warn about cases where the compiler reduces the magnitude of a
3733 constant involved in a comparison. For example: @code{x + 2 > y} will
3734 be simplified to @code{x + 1 >= y}. This is reported only at the
3735 highest warning level because this simplification applies to many
3736 comparisons, so this warning level will give a very large number of
3740 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3741 @opindex Wsuggest-attribute=
3742 @opindex Wno-suggest-attribute=
3743 Warn for cases where adding an attribute may be beneficial. The
3744 attributes currently supported are listed below.
3747 @item -Wsuggest-attribute=pure
3748 @itemx -Wsuggest-attribute=const
3749 @itemx -Wsuggest-attribute=noreturn
3750 @opindex Wsuggest-attribute=pure
3751 @opindex Wno-suggest-attribute=pure
3752 @opindex Wsuggest-attribute=const
3753 @opindex Wno-suggest-attribute=const
3754 @opindex Wsuggest-attribute=noreturn
3755 @opindex Wno-suggest-attribute=noreturn
3757 Warn about functions which might be candidates for attributes
3758 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3759 functions visible in other compilation units or (in the case of @code{pure} and
3760 @code{const}) if it cannot prove that the function returns normally. A function
3761 returns normally if it doesn't contain an infinite loop nor returns abnormally
3762 by throwing, calling @code{abort()} or trapping. This analysis requires option
3763 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3764 higher. Higher optimization levels improve the accuracy of the analysis.
3767 @item -Warray-bounds
3768 @opindex Wno-array-bounds
3769 @opindex Warray-bounds
3770 This option is only active when @option{-ftree-vrp} is active
3771 (default for @option{-O2} and above). It warns about subscripts to arrays
3772 that are always out of bounds. This warning is enabled by @option{-Wall}.
3774 @item -Wno-div-by-zero
3775 @opindex Wno-div-by-zero
3776 @opindex Wdiv-by-zero
3777 Do not warn about compile-time integer division by zero. Floating point
3778 division by zero is not warned about, as it can be a legitimate way of
3779 obtaining infinities and NaNs.
3781 @item -Wsystem-headers
3782 @opindex Wsystem-headers
3783 @opindex Wno-system-headers
3784 @cindex warnings from system headers
3785 @cindex system headers, warnings from
3786 Print warning messages for constructs found in system header files.
3787 Warnings from system headers are normally suppressed, on the assumption
3788 that they usually do not indicate real problems and would only make the
3789 compiler output harder to read. Using this command line option tells
3790 GCC to emit warnings from system headers as if they occurred in user
3791 code. However, note that using @option{-Wall} in conjunction with this
3792 option will @emph{not} warn about unknown pragmas in system
3793 headers---for that, @option{-Wunknown-pragmas} must also be used.
3796 @opindex Wtrampolines
3797 @opindex Wno-trampolines
3798 Warn about trampolines generated for pointers to nested functions.
3800 A trampoline is a small piece of data or code that is created at run
3801 time on the stack when the address of a nested function is taken, and
3802 is used to call the nested function indirectly. For some targets, it
3803 is made up of data only and thus requires no special treatment. But,
3804 for most targets, it is made up of code and thus requires the stack
3805 to be made executable in order for the program to work properly.
3808 @opindex Wfloat-equal
3809 @opindex Wno-float-equal
3810 Warn if floating point values are used in equality comparisons.
3812 The idea behind this is that sometimes it is convenient (for the
3813 programmer) to consider floating-point values as approximations to
3814 infinitely precise real numbers. If you are doing this, then you need
3815 to compute (by analyzing the code, or in some other way) the maximum or
3816 likely maximum error that the computation introduces, and allow for it
3817 when performing comparisons (and when producing output, but that's a
3818 different problem). In particular, instead of testing for equality, you
3819 would check to see whether the two values have ranges that overlap; and
3820 this is done with the relational operators, so equality comparisons are
3823 @item -Wtraditional @r{(C and Objective-C only)}
3824 @opindex Wtraditional
3825 @opindex Wno-traditional
3826 Warn about certain constructs that behave differently in traditional and
3827 ISO C@. Also warn about ISO C constructs that have no traditional C
3828 equivalent, and/or problematic constructs which should be avoided.
3832 Macro parameters that appear within string literals in the macro body.
3833 In traditional C macro replacement takes place within string literals,
3834 but does not in ISO C@.
3837 In traditional C, some preprocessor directives did not exist.
3838 Traditional preprocessors would only consider a line to be a directive
3839 if the @samp{#} appeared in column 1 on the line. Therefore
3840 @option{-Wtraditional} warns about directives that traditional C
3841 understands but would ignore because the @samp{#} does not appear as the
3842 first character on the line. It also suggests you hide directives like
3843 @samp{#pragma} not understood by traditional C by indenting them. Some
3844 traditional implementations would not recognize @samp{#elif}, so it
3845 suggests avoiding it altogether.
3848 A function-like macro that appears without arguments.
3851 The unary plus operator.
3854 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3855 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3856 constants.) Note, these suffixes appear in macros defined in the system
3857 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3858 Use of these macros in user code might normally lead to spurious
3859 warnings, however GCC's integrated preprocessor has enough context to
3860 avoid warning in these cases.
3863 A function declared external in one block and then used after the end of
3867 A @code{switch} statement has an operand of type @code{long}.
3870 A non-@code{static} function declaration follows a @code{static} one.
3871 This construct is not accepted by some traditional C compilers.
3874 The ISO type of an integer constant has a different width or
3875 signedness from its traditional type. This warning is only issued if
3876 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3877 typically represent bit patterns, are not warned about.
3880 Usage of ISO string concatenation is detected.
3883 Initialization of automatic aggregates.
3886 Identifier conflicts with labels. Traditional C lacks a separate
3887 namespace for labels.
3890 Initialization of unions. If the initializer is zero, the warning is
3891 omitted. This is done under the assumption that the zero initializer in
3892 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3893 initializer warnings and relies on default initialization to zero in the
3897 Conversions by prototypes between fixed/floating point values and vice
3898 versa. The absence of these prototypes when compiling with traditional
3899 C would cause serious problems. This is a subset of the possible
3900 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3903 Use of ISO C style function definitions. This warning intentionally is
3904 @emph{not} issued for prototype declarations or variadic functions
3905 because these ISO C features will appear in your code when using
3906 libiberty's traditional C compatibility macros, @code{PARAMS} and
3907 @code{VPARAMS}. This warning is also bypassed for nested functions
3908 because that feature is already a GCC extension and thus not relevant to
3909 traditional C compatibility.
3912 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3913 @opindex Wtraditional-conversion
3914 @opindex Wno-traditional-conversion
3915 Warn if a prototype causes a type conversion that is different from what
3916 would happen to the same argument in the absence of a prototype. This
3917 includes conversions of fixed point to floating and vice versa, and
3918 conversions changing the width or signedness of a fixed point argument
3919 except when the same as the default promotion.
3921 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3922 @opindex Wdeclaration-after-statement
3923 @opindex Wno-declaration-after-statement
3924 Warn when a declaration is found after a statement in a block. This
3925 construct, known from C++, was introduced with ISO C99 and is by default
3926 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3927 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3932 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3934 @item -Wno-endif-labels
3935 @opindex Wno-endif-labels
3936 @opindex Wendif-labels
3937 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3942 Warn whenever a local variable or type declaration shadows another variable,
3943 parameter, type, or class member (in C++), or whenever a built-in function
3944 is shadowed. Note that in C++, the compiler will not warn if a local variable
3945 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3947 @item -Wlarger-than=@var{len}
3948 @opindex Wlarger-than=@var{len}
3949 @opindex Wlarger-than-@var{len}
3950 Warn whenever an object of larger than @var{len} bytes is defined.
3952 @item -Wframe-larger-than=@var{len}
3953 @opindex Wframe-larger-than
3954 Warn if the size of a function frame is larger than @var{len} bytes.
3955 The computation done to determine the stack frame size is approximate
3956 and not conservative.
3957 The actual requirements may be somewhat greater than @var{len}
3958 even if you do not get a warning. In addition, any space allocated
3959 via @code{alloca}, variable-length arrays, or related constructs
3960 is not included by the compiler when determining
3961 whether or not to issue a warning.
3963 @item -Wstack-usage=@var{len}
3964 @opindex Wstack-usage
3965 Warn if the stack usage of a function might be larger than @var{len} bytes.
3966 The computation done to determine the stack usage is conservative.
3967 Any space allocated via @code{alloca}, variable-length arrays, or related
3968 constructs is included by the compiler when determining whether or not to
3971 The message is in keeping with the output of @option{-fstack-usage}.
3975 If the stack usage is fully static but exceeds the specified amount, it's:
3978 Â warning: stack usage is 1120 bytes
3981 If the stack usage is (partly) dynamic but bounded, it's:
3984 Â warning: stack usage might be 1648 bytes
3987 If the stack usage is (partly) dynamic and not bounded, it's:
3990 Â warning: stack usage might be unbounded
3994 @item -Wunsafe-loop-optimizations
3995 @opindex Wunsafe-loop-optimizations
3996 @opindex Wno-unsafe-loop-optimizations
3997 Warn if the loop cannot be optimized because the compiler could not
3998 assume anything on the bounds of the loop indices. With
3999 @option{-funsafe-loop-optimizations} warn if the compiler made
4002 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4003 @opindex Wno-pedantic-ms-format
4004 @opindex Wpedantic-ms-format
4005 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4006 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4007 depending on the MS runtime, when you are using the options @option{-Wformat}
4008 and @option{-pedantic} without gnu-extensions.
4010 @item -Wpointer-arith
4011 @opindex Wpointer-arith
4012 @opindex Wno-pointer-arith
4013 Warn about anything that depends on the ``size of'' a function type or
4014 of @code{void}. GNU C assigns these types a size of 1, for
4015 convenience in calculations with @code{void *} pointers and pointers
4016 to functions. In C++, warn also when an arithmetic operation involves
4017 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4020 @opindex Wtype-limits
4021 @opindex Wno-type-limits
4022 Warn if a comparison is always true or always false due to the limited
4023 range of the data type, but do not warn for constant expressions. For
4024 example, warn if an unsigned variable is compared against zero with
4025 @samp{<} or @samp{>=}. This warning is also enabled by
4028 @item -Wbad-function-cast @r{(C and Objective-C only)}
4029 @opindex Wbad-function-cast
4030 @opindex Wno-bad-function-cast
4031 Warn whenever a function call is cast to a non-matching type.
4032 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4034 @item -Wc++-compat @r{(C and Objective-C only)}
4035 Warn about ISO C constructs that are outside of the common subset of
4036 ISO C and ISO C++, e.g.@: request for implicit conversion from
4037 @code{void *} to a pointer to non-@code{void} type.
4039 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
4040 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
4041 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
4042 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
4046 @opindex Wno-cast-qual
4047 Warn whenever a pointer is cast so as to remove a type qualifier from
4048 the target type. For example, warn if a @code{const char *} is cast
4049 to an ordinary @code{char *}.
4051 Also warn when making a cast which introduces a type qualifier in an
4052 unsafe way. For example, casting @code{char **} to @code{const char **}
4053 is unsafe, as in this example:
4056 /* p is char ** value. */
4057 const char **q = (const char **) p;
4058 /* Assignment of readonly string to const char * is OK. */
4060 /* Now char** pointer points to read-only memory. */
4065 @opindex Wcast-align
4066 @opindex Wno-cast-align
4067 Warn whenever a pointer is cast such that the required alignment of the
4068 target is increased. For example, warn if a @code{char *} is cast to
4069 an @code{int *} on machines where integers can only be accessed at
4070 two- or four-byte boundaries.
4072 @item -Wwrite-strings
4073 @opindex Wwrite-strings
4074 @opindex Wno-write-strings
4075 When compiling C, give string constants the type @code{const
4076 char[@var{length}]} so that copying the address of one into a
4077 non-@code{const} @code{char *} pointer will get a warning. These
4078 warnings will help you find at compile time code that can try to write
4079 into a string constant, but only if you have been very careful about
4080 using @code{const} in declarations and prototypes. Otherwise, it will
4081 just be a nuisance. This is why we did not make @option{-Wall} request
4084 When compiling C++, warn about the deprecated conversion from string
4085 literals to @code{char *}. This warning is enabled by default for C++
4090 @opindex Wno-clobbered
4091 Warn for variables that might be changed by @samp{longjmp} or
4092 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4095 @opindex Wconversion
4096 @opindex Wno-conversion
4097 Warn for implicit conversions that may alter a value. This includes
4098 conversions between real and integer, like @code{abs (x)} when
4099 @code{x} is @code{double}; conversions between signed and unsigned,
4100 like @code{unsigned ui = -1}; and conversions to smaller types, like
4101 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4102 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4103 changed by the conversion like in @code{abs (2.0)}. Warnings about
4104 conversions between signed and unsigned integers can be disabled by
4105 using @option{-Wno-sign-conversion}.
4107 For C++, also warn for confusing overload resolution for user-defined
4108 conversions; and conversions that will never use a type conversion
4109 operator: conversions to @code{void}, the same type, a base class or a
4110 reference to them. Warnings about conversions between signed and
4111 unsigned integers are disabled by default in C++ unless
4112 @option{-Wsign-conversion} is explicitly enabled.
4114 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4115 @opindex Wconversion-null
4116 @opindex Wno-conversion-null
4117 Do not warn for conversions between @code{NULL} and non-pointer
4118 types. @option{-Wconversion-null} is enabled by default.
4121 @opindex Wempty-body
4122 @opindex Wno-empty-body
4123 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4124 while} statement. This warning is also enabled by @option{-Wextra}.
4126 @item -Wenum-compare
4127 @opindex Wenum-compare
4128 @opindex Wno-enum-compare
4129 Warn about a comparison between values of different enum types. In C++
4130 this warning is enabled by default. In C this warning is enabled by
4133 @item -Wjump-misses-init @r{(C, Objective-C only)}
4134 @opindex Wjump-misses-init
4135 @opindex Wno-jump-misses-init
4136 Warn if a @code{goto} statement or a @code{switch} statement jumps
4137 forward across the initialization of a variable, or jumps backward to a
4138 label after the variable has been initialized. This only warns about
4139 variables which are initialized when they are declared. This warning is
4140 only supported for C and Objective C; in C++ this sort of branch is an
4143 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4144 can be disabled with the @option{-Wno-jump-misses-init} option.
4146 @item -Wsign-compare
4147 @opindex Wsign-compare
4148 @opindex Wno-sign-compare
4149 @cindex warning for comparison of signed and unsigned values
4150 @cindex comparison of signed and unsigned values, warning
4151 @cindex signed and unsigned values, comparison warning
4152 Warn when a comparison between signed and unsigned values could produce
4153 an incorrect result when the signed value is converted to unsigned.
4154 This warning is also enabled by @option{-Wextra}; to get the other warnings
4155 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4157 @item -Wsign-conversion
4158 @opindex Wsign-conversion
4159 @opindex Wno-sign-conversion
4160 Warn for implicit conversions that may change the sign of an integer
4161 value, like assigning a signed integer expression to an unsigned
4162 integer variable. An explicit cast silences the warning. In C, this
4163 option is enabled also by @option{-Wconversion}.
4167 @opindex Wno-address
4168 Warn about suspicious uses of memory addresses. These include using
4169 the address of a function in a conditional expression, such as
4170 @code{void func(void); if (func)}, and comparisons against the memory
4171 address of a string literal, such as @code{if (x == "abc")}. Such
4172 uses typically indicate a programmer error: the address of a function
4173 always evaluates to true, so their use in a conditional usually
4174 indicate that the programmer forgot the parentheses in a function
4175 call; and comparisons against string literals result in unspecified
4176 behavior and are not portable in C, so they usually indicate that the
4177 programmer intended to use @code{strcmp}. This warning is enabled by
4181 @opindex Wlogical-op
4182 @opindex Wno-logical-op
4183 Warn about suspicious uses of logical operators in expressions.
4184 This includes using logical operators in contexts where a
4185 bit-wise operator is likely to be expected.
4187 @item -Waggregate-return
4188 @opindex Waggregate-return
4189 @opindex Wno-aggregate-return
4190 Warn if any functions that return structures or unions are defined or
4191 called. (In languages where you can return an array, this also elicits
4194 @item -Wno-attributes
4195 @opindex Wno-attributes
4196 @opindex Wattributes
4197 Do not warn if an unexpected @code{__attribute__} is used, such as
4198 unrecognized attributes, function attributes applied to variables,
4199 etc. This will not stop errors for incorrect use of supported
4202 @item -Wno-builtin-macro-redefined
4203 @opindex Wno-builtin-macro-redefined
4204 @opindex Wbuiltin-macro-redefined
4205 Do not warn if certain built-in macros are redefined. This suppresses
4206 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4207 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4209 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4210 @opindex Wstrict-prototypes
4211 @opindex Wno-strict-prototypes
4212 Warn if a function is declared or defined without specifying the
4213 argument types. (An old-style function definition is permitted without
4214 a warning if preceded by a declaration which specifies the argument
4217 @item -Wold-style-declaration @r{(C and Objective-C only)}
4218 @opindex Wold-style-declaration
4219 @opindex Wno-old-style-declaration
4220 Warn for obsolescent usages, according to the C Standard, in a
4221 declaration. For example, warn if storage-class specifiers like
4222 @code{static} are not the first things in a declaration. This warning
4223 is also enabled by @option{-Wextra}.
4225 @item -Wold-style-definition @r{(C and Objective-C only)}
4226 @opindex Wold-style-definition
4227 @opindex Wno-old-style-definition
4228 Warn if an old-style function definition is used. A warning is given
4229 even if there is a previous prototype.
4231 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4232 @opindex Wmissing-parameter-type
4233 @opindex Wno-missing-parameter-type
4234 A function parameter is declared without a type specifier in K&R-style
4241 This warning is also enabled by @option{-Wextra}.
4243 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4244 @opindex Wmissing-prototypes
4245 @opindex Wno-missing-prototypes
4246 Warn if a global function is defined without a previous prototype
4247 declaration. This warning is issued even if the definition itself
4248 provides a prototype. The aim is to detect global functions that fail
4249 to be declared in header files.
4251 @item -Wmissing-declarations
4252 @opindex Wmissing-declarations
4253 @opindex Wno-missing-declarations
4254 Warn if a global function is defined without a previous declaration.
4255 Do so even if the definition itself provides a prototype.
4256 Use this option to detect global functions that are not declared in
4257 header files. In C++, no warnings are issued for function templates,
4258 or for inline functions, or for functions in anonymous namespaces.
4260 @item -Wmissing-field-initializers
4261 @opindex Wmissing-field-initializers
4262 @opindex Wno-missing-field-initializers
4266 Warn if a structure's initializer has some fields missing. For
4267 example, the following code would cause such a warning, because
4268 @code{x.h} is implicitly zero:
4271 struct s @{ int f, g, h; @};
4272 struct s x = @{ 3, 4 @};
4275 This option does not warn about designated initializers, so the following
4276 modification would not trigger a warning:
4279 struct s @{ int f, g, h; @};
4280 struct s x = @{ .f = 3, .g = 4 @};
4283 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4284 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4286 @item -Wmissing-format-attribute
4287 @opindex Wmissing-format-attribute
4288 @opindex Wno-missing-format-attribute
4291 Warn about function pointers which might be candidates for @code{format}
4292 attributes. Note these are only possible candidates, not absolute ones.
4293 GCC will guess that function pointers with @code{format} attributes that
4294 are used in assignment, initialization, parameter passing or return
4295 statements should have a corresponding @code{format} attribute in the
4296 resulting type. I.e.@: the left-hand side of the assignment or
4297 initialization, the type of the parameter variable, or the return type
4298 of the containing function respectively should also have a @code{format}
4299 attribute to avoid the warning.
4301 GCC will also warn about function definitions which might be
4302 candidates for @code{format} attributes. Again, these are only
4303 possible candidates. GCC will guess that @code{format} attributes
4304 might be appropriate for any function that calls a function like
4305 @code{vprintf} or @code{vscanf}, but this might not always be the
4306 case, and some functions for which @code{format} attributes are
4307 appropriate may not be detected.
4309 @item -Wno-multichar
4310 @opindex Wno-multichar
4312 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4313 Usually they indicate a typo in the user's code, as they have
4314 implementation-defined values, and should not be used in portable code.
4316 @item -Wnormalized=<none|id|nfc|nfkc>
4317 @opindex Wnormalized=
4320 @cindex character set, input normalization
4321 In ISO C and ISO C++, two identifiers are different if they are
4322 different sequences of characters. However, sometimes when characters
4323 outside the basic ASCII character set are used, you can have two
4324 different character sequences that look the same. To avoid confusion,
4325 the ISO 10646 standard sets out some @dfn{normalization rules} which
4326 when applied ensure that two sequences that look the same are turned into
4327 the same sequence. GCC can warn you if you are using identifiers which
4328 have not been normalized; this option controls that warning.
4330 There are four levels of warning that GCC supports. The default is
4331 @option{-Wnormalized=nfc}, which warns about any identifier which is
4332 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4333 recommended form for most uses.
4335 Unfortunately, there are some characters which ISO C and ISO C++ allow
4336 in identifiers that when turned into NFC aren't allowable as
4337 identifiers. That is, there's no way to use these symbols in portable
4338 ISO C or C++ and have all your identifiers in NFC@.
4339 @option{-Wnormalized=id} suppresses the warning for these characters.
4340 It is hoped that future versions of the standards involved will correct
4341 this, which is why this option is not the default.
4343 You can switch the warning off for all characters by writing
4344 @option{-Wnormalized=none}. You would only want to do this if you
4345 were using some other normalization scheme (like ``D''), because
4346 otherwise you can easily create bugs that are literally impossible to see.
4348 Some characters in ISO 10646 have distinct meanings but look identical
4349 in some fonts or display methodologies, especially once formatting has
4350 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4351 LETTER N'', will display just like a regular @code{n} which has been
4352 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4353 normalization scheme to convert all these into a standard form as
4354 well, and GCC will warn if your code is not in NFKC if you use
4355 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4356 about every identifier that contains the letter O because it might be
4357 confused with the digit 0, and so is not the default, but may be
4358 useful as a local coding convention if the programming environment is
4359 unable to be fixed to display these characters distinctly.
4361 @item -Wno-deprecated
4362 @opindex Wno-deprecated
4363 @opindex Wdeprecated
4364 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4366 @item -Wno-deprecated-declarations
4367 @opindex Wno-deprecated-declarations
4368 @opindex Wdeprecated-declarations
4369 Do not warn about uses of functions (@pxref{Function Attributes}),
4370 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4371 Attributes}) marked as deprecated by using the @code{deprecated}
4375 @opindex Wno-overflow
4377 Do not warn about compile-time overflow in constant expressions.
4379 @item -Woverride-init @r{(C and Objective-C only)}
4380 @opindex Woverride-init
4381 @opindex Wno-override-init
4385 Warn if an initialized field without side effects is overridden when
4386 using designated initializers (@pxref{Designated Inits, , Designated
4389 This warning is included in @option{-Wextra}. To get other
4390 @option{-Wextra} warnings without this one, use @samp{-Wextra
4391 -Wno-override-init}.
4396 Warn if a structure is given the packed attribute, but the packed
4397 attribute has no effect on the layout or size of the structure.
4398 Such structures may be mis-aligned for little benefit. For
4399 instance, in this code, the variable @code{f.x} in @code{struct bar}
4400 will be misaligned even though @code{struct bar} does not itself
4401 have the packed attribute:
4408 @} __attribute__((packed));
4416 @item -Wpacked-bitfield-compat
4417 @opindex Wpacked-bitfield-compat
4418 @opindex Wno-packed-bitfield-compat
4419 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4420 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4421 the change can lead to differences in the structure layout. GCC
4422 informs you when the offset of such a field has changed in GCC 4.4.
4423 For example there is no longer a 4-bit padding between field @code{a}
4424 and @code{b} in this structure:
4431 @} __attribute__ ((packed));
4434 This warning is enabled by default. Use
4435 @option{-Wno-packed-bitfield-compat} to disable this warning.
4440 Warn if padding is included in a structure, either to align an element
4441 of the structure or to align the whole structure. Sometimes when this
4442 happens it is possible to rearrange the fields of the structure to
4443 reduce the padding and so make the structure smaller.
4445 @item -Wredundant-decls
4446 @opindex Wredundant-decls
4447 @opindex Wno-redundant-decls
4448 Warn if anything is declared more than once in the same scope, even in
4449 cases where multiple declaration is valid and changes nothing.
4451 @item -Wnested-externs @r{(C and Objective-C only)}
4452 @opindex Wnested-externs
4453 @opindex Wno-nested-externs
4454 Warn if an @code{extern} declaration is encountered within a function.
4459 Warn if a function can not be inlined and it was declared as inline.
4460 Even with this option, the compiler will not warn about failures to
4461 inline functions declared in system headers.
4463 The compiler uses a variety of heuristics to determine whether or not
4464 to inline a function. For example, the compiler takes into account
4465 the size of the function being inlined and the amount of inlining
4466 that has already been done in the current function. Therefore,
4467 seemingly insignificant changes in the source program can cause the
4468 warnings produced by @option{-Winline} to appear or disappear.
4470 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4471 @opindex Wno-invalid-offsetof
4472 @opindex Winvalid-offsetof
4473 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4474 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4475 to a non-POD type is undefined. In existing C++ implementations,
4476 however, @samp{offsetof} typically gives meaningful results even when
4477 applied to certain kinds of non-POD types. (Such as a simple
4478 @samp{struct} that fails to be a POD type only by virtue of having a
4479 constructor.) This flag is for users who are aware that they are
4480 writing nonportable code and who have deliberately chosen to ignore the
4483 The restrictions on @samp{offsetof} may be relaxed in a future version
4484 of the C++ standard.
4486 @item -Wno-int-to-pointer-cast
4487 @opindex Wno-int-to-pointer-cast
4488 @opindex Wint-to-pointer-cast
4489 Suppress warnings from casts to pointer type of an integer of a
4490 different size. In C++, casting to a pointer type of smaller size is
4491 an error. @option{Wint-to-pointer-cast} is enabled by default.
4494 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4495 @opindex Wno-pointer-to-int-cast
4496 @opindex Wpointer-to-int-cast
4497 Suppress warnings from casts from a pointer to an integer type of a
4501 @opindex Winvalid-pch
4502 @opindex Wno-invalid-pch
4503 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4504 the search path but can't be used.
4508 @opindex Wno-long-long
4509 Warn if @samp{long long} type is used. This is enabled by either
4510 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4511 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4513 @item -Wvariadic-macros
4514 @opindex Wvariadic-macros
4515 @opindex Wno-variadic-macros
4516 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4517 alternate syntax when in pedantic ISO C99 mode. This is default.
4518 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4523 Warn if variable length array is used in the code.
4524 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4525 the variable length array.
4527 @item -Wvolatile-register-var
4528 @opindex Wvolatile-register-var
4529 @opindex Wno-volatile-register-var
4530 Warn if a register variable is declared volatile. The volatile
4531 modifier does not inhibit all optimizations that may eliminate reads
4532 and/or writes to register variables. This warning is enabled by
4535 @item -Wdisabled-optimization
4536 @opindex Wdisabled-optimization
4537 @opindex Wno-disabled-optimization
4538 Warn if a requested optimization pass is disabled. This warning does
4539 not generally indicate that there is anything wrong with your code; it
4540 merely indicates that GCC's optimizers were unable to handle the code
4541 effectively. Often, the problem is that your code is too big or too
4542 complex; GCC will refuse to optimize programs when the optimization
4543 itself is likely to take inordinate amounts of time.
4545 @item -Wpointer-sign @r{(C and Objective-C only)}
4546 @opindex Wpointer-sign
4547 @opindex Wno-pointer-sign
4548 Warn for pointer argument passing or assignment with different signedness.
4549 This option is only supported for C and Objective-C@. It is implied by
4550 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4551 @option{-Wno-pointer-sign}.
4553 @item -Wstack-protector
4554 @opindex Wstack-protector
4555 @opindex Wno-stack-protector
4556 This option is only active when @option{-fstack-protector} is active. It
4557 warns about functions that will not be protected against stack smashing.
4560 @opindex Wno-mudflap
4561 Suppress warnings about constructs that cannot be instrumented by
4564 @item -Woverlength-strings
4565 @opindex Woverlength-strings
4566 @opindex Wno-overlength-strings
4567 Warn about string constants which are longer than the ``minimum
4568 maximum'' length specified in the C standard. Modern compilers
4569 generally allow string constants which are much longer than the
4570 standard's minimum limit, but very portable programs should avoid
4571 using longer strings.
4573 The limit applies @emph{after} string constant concatenation, and does
4574 not count the trailing NUL@. In C90, the limit was 509 characters; in
4575 C99, it was raised to 4095. C++98 does not specify a normative
4576 minimum maximum, so we do not diagnose overlength strings in C++@.
4578 This option is implied by @option{-pedantic}, and can be disabled with
4579 @option{-Wno-overlength-strings}.
4581 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4582 @opindex Wunsuffixed-float-constants
4584 GCC will issue a warning for any floating constant that does not have
4585 a suffix. When used together with @option{-Wsystem-headers} it will
4586 warn about such constants in system header files. This can be useful
4587 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4588 from the decimal floating-point extension to C99.
4591 @node Debugging Options
4592 @section Options for Debugging Your Program or GCC
4593 @cindex options, debugging
4594 @cindex debugging information options
4596 GCC has various special options that are used for debugging
4597 either your program or GCC:
4602 Produce debugging information in the operating system's native format
4603 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4606 On most systems that use stabs format, @option{-g} enables use of extra
4607 debugging information that only GDB can use; this extra information
4608 makes debugging work better in GDB but will probably make other debuggers
4610 refuse to read the program. If you want to control for certain whether
4611 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4612 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4614 GCC allows you to use @option{-g} with
4615 @option{-O}. The shortcuts taken by optimized code may occasionally
4616 produce surprising results: some variables you declared may not exist
4617 at all; flow of control may briefly move where you did not expect it;
4618 some statements may not be executed because they compute constant
4619 results or their values were already at hand; some statements may
4620 execute in different places because they were moved out of loops.
4622 Nevertheless it proves possible to debug optimized output. This makes
4623 it reasonable to use the optimizer for programs that might have bugs.
4625 The following options are useful when GCC is generated with the
4626 capability for more than one debugging format.
4630 Produce debugging information for use by GDB@. This means to use the
4631 most expressive format available (DWARF 2, stabs, or the native format
4632 if neither of those are supported), including GDB extensions if at all
4637 Produce debugging information in stabs format (if that is supported),
4638 without GDB extensions. This is the format used by DBX on most BSD
4639 systems. On MIPS, Alpha and System V Release 4 systems this option
4640 produces stabs debugging output which is not understood by DBX or SDB@.
4641 On System V Release 4 systems this option requires the GNU assembler.
4643 @item -feliminate-unused-debug-symbols
4644 @opindex feliminate-unused-debug-symbols
4645 Produce debugging information in stabs format (if that is supported),
4646 for only symbols that are actually used.
4648 @item -femit-class-debug-always
4649 Instead of emitting debugging information for a C++ class in only one
4650 object file, emit it in all object files using the class. This option
4651 should be used only with debuggers that are unable to handle the way GCC
4652 normally emits debugging information for classes because using this
4653 option will increase the size of debugging information by as much as a
4656 @item -fno-debug-types-section
4657 @opindex fno-debug-types-section
4658 @opindex fdebug-types-section
4659 By default when using DWARF v4 or higher type DIEs will be put into
4660 their own .debug_types section instead of making them part of the
4661 .debug_info section. It is more efficient to put them in a separate
4662 comdat sections since the linker will then be able to remove duplicates.
4663 But not all DWARF consumers support .debug_types sections yet.
4667 Produce debugging information in stabs format (if that is supported),
4668 using GNU extensions understood only by the GNU debugger (GDB)@. The
4669 use of these extensions is likely to make other debuggers crash or
4670 refuse to read the program.
4674 Produce debugging information in COFF format (if that is supported).
4675 This is the format used by SDB on most System V systems prior to
4680 Produce debugging information in XCOFF format (if that is supported).
4681 This is the format used by the DBX debugger on IBM RS/6000 systems.
4685 Produce debugging information in XCOFF format (if that is supported),
4686 using GNU extensions understood only by the GNU debugger (GDB)@. The
4687 use of these extensions is likely to make other debuggers crash or
4688 refuse to read the program, and may cause assemblers other than the GNU
4689 assembler (GAS) to fail with an error.
4691 @item -gdwarf-@var{version}
4692 @opindex gdwarf-@var{version}
4693 Produce debugging information in DWARF format (if that is
4694 supported). This is the format used by DBX on IRIX 6. The value
4695 of @var{version} may be either 2, 3 or 4; the default version is 2.
4697 Note that with DWARF version 2 some ports require, and will always
4698 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4700 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4701 for maximum benefit.
4703 @item -grecord-gcc-switches
4704 @opindex grecord-gcc-switches
4705 This switch causes the command line options, that were used to invoke the
4706 compiler and may affect code generation, to be appended to the
4707 DW_AT_producer attribute in DWARF debugging information. The options
4708 are concatenated with spaces separating them from each other and from
4709 the compiler version. See also @option{-frecord-gcc-switches} for another
4710 way of storing compiler options into the object file.
4712 @item -gno-record-gcc-switches
4713 @opindex gno-record-gcc-switches
4714 Disallow appending command line options to the DW_AT_producer attribute
4715 in DWARF debugging information. This is the default.
4717 @item -gstrict-dwarf
4718 @opindex gstrict-dwarf
4719 Disallow using extensions of later DWARF standard version than selected
4720 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4721 DWARF extensions from later standard versions is allowed.
4723 @item -gno-strict-dwarf
4724 @opindex gno-strict-dwarf
4725 Allow using extensions of later DWARF standard version than selected with
4726 @option{-gdwarf-@var{version}}.
4730 Produce debugging information in VMS debug format (if that is
4731 supported). This is the format used by DEBUG on VMS systems.
4734 @itemx -ggdb@var{level}
4735 @itemx -gstabs@var{level}
4736 @itemx -gcoff@var{level}
4737 @itemx -gxcoff@var{level}
4738 @itemx -gvms@var{level}
4739 Request debugging information and also use @var{level} to specify how
4740 much information. The default level is 2.
4742 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4745 Level 1 produces minimal information, enough for making backtraces in
4746 parts of the program that you don't plan to debug. This includes
4747 descriptions of functions and external variables, but no information
4748 about local variables and no line numbers.
4750 Level 3 includes extra information, such as all the macro definitions
4751 present in the program. Some debuggers support macro expansion when
4752 you use @option{-g3}.
4754 @option{-gdwarf-2} does not accept a concatenated debug level, because
4755 GCC used to support an option @option{-gdwarf} that meant to generate
4756 debug information in version 1 of the DWARF format (which is very
4757 different from version 2), and it would have been too confusing. That
4758 debug format is long obsolete, but the option cannot be changed now.
4759 Instead use an additional @option{-g@var{level}} option to change the
4760 debug level for DWARF.
4764 Turn off generation of debug info, if leaving out this option would have
4765 generated it, or turn it on at level 2 otherwise. The position of this
4766 argument in the command line does not matter, it takes effect after all
4767 other options are processed, and it does so only once, no matter how
4768 many times it is given. This is mainly intended to be used with
4769 @option{-fcompare-debug}.
4771 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4772 @opindex fdump-final-insns
4773 Dump the final internal representation (RTL) to @var{file}. If the
4774 optional argument is omitted (or if @var{file} is @code{.}), the name
4775 of the dump file will be determined by appending @code{.gkd} to the
4776 compilation output file name.
4778 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4779 @opindex fcompare-debug
4780 @opindex fno-compare-debug
4781 If no error occurs during compilation, run the compiler a second time,
4782 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4783 passed to the second compilation. Dump the final internal
4784 representation in both compilations, and print an error if they differ.
4786 If the equal sign is omitted, the default @option{-gtoggle} is used.
4788 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4789 and nonzero, implicitly enables @option{-fcompare-debug}. If
4790 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4791 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4794 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4795 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4796 of the final representation and the second compilation, preventing even
4797 @env{GCC_COMPARE_DEBUG} from taking effect.
4799 To verify full coverage during @option{-fcompare-debug} testing, set
4800 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4801 which GCC will reject as an invalid option in any actual compilation
4802 (rather than preprocessing, assembly or linking). To get just a
4803 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4804 not overridden} will do.
4806 @item -fcompare-debug-second
4807 @opindex fcompare-debug-second
4808 This option is implicitly passed to the compiler for the second
4809 compilation requested by @option{-fcompare-debug}, along with options to
4810 silence warnings, and omitting other options that would cause
4811 side-effect compiler outputs to files or to the standard output. Dump
4812 files and preserved temporary files are renamed so as to contain the
4813 @code{.gk} additional extension during the second compilation, to avoid
4814 overwriting those generated by the first.
4816 When this option is passed to the compiler driver, it causes the
4817 @emph{first} compilation to be skipped, which makes it useful for little
4818 other than debugging the compiler proper.
4820 @item -feliminate-dwarf2-dups
4821 @opindex feliminate-dwarf2-dups
4822 Compress DWARF2 debugging information by eliminating duplicated
4823 information about each symbol. This option only makes sense when
4824 generating DWARF2 debugging information with @option{-gdwarf-2}.
4826 @item -femit-struct-debug-baseonly
4827 Emit debug information for struct-like types
4828 only when the base name of the compilation source file
4829 matches the base name of file in which the struct was defined.
4831 This option substantially reduces the size of debugging information,
4832 but at significant potential loss in type information to the debugger.
4833 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4834 See @option{-femit-struct-debug-detailed} for more detailed control.
4836 This option works only with DWARF 2.
4838 @item -femit-struct-debug-reduced
4839 Emit debug information for struct-like types
4840 only when the base name of the compilation source file
4841 matches the base name of file in which the type was defined,
4842 unless the struct is a template or defined in a system header.
4844 This option significantly reduces the size of debugging information,
4845 with some potential loss in type information to the debugger.
4846 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4847 See @option{-femit-struct-debug-detailed} for more detailed control.
4849 This option works only with DWARF 2.
4851 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4852 Specify the struct-like types
4853 for which the compiler will generate debug information.
4854 The intent is to reduce duplicate struct debug information
4855 between different object files within the same program.
4857 This option is a detailed version of
4858 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4859 which will serve for most needs.
4861 A specification has the syntax@*
4862 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4864 The optional first word limits the specification to
4865 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4866 A struct type is used directly when it is the type of a variable, member.
4867 Indirect uses arise through pointers to structs.
4868 That is, when use of an incomplete struct would be legal, the use is indirect.
4870 @samp{struct one direct; struct two * indirect;}.
4872 The optional second word limits the specification to
4873 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4874 Generic structs are a bit complicated to explain.
4875 For C++, these are non-explicit specializations of template classes,
4876 or non-template classes within the above.
4877 Other programming languages have generics,
4878 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4880 The third word specifies the source files for those
4881 structs for which the compiler will emit debug information.
4882 The values @samp{none} and @samp{any} have the normal meaning.
4883 The value @samp{base} means that
4884 the base of name of the file in which the type declaration appears
4885 must match the base of the name of the main compilation file.
4886 In practice, this means that
4887 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4888 but types declared in other header will not.
4889 The value @samp{sys} means those types satisfying @samp{base}
4890 or declared in system or compiler headers.
4892 You may need to experiment to determine the best settings for your application.
4894 The default is @samp{-femit-struct-debug-detailed=all}.
4896 This option works only with DWARF 2.
4898 @item -fno-merge-debug-strings
4899 @opindex fmerge-debug-strings
4900 @opindex fno-merge-debug-strings
4901 Direct the linker to not merge together strings in the debugging
4902 information which are identical in different object files. Merging is
4903 not supported by all assemblers or linkers. Merging decreases the size
4904 of the debug information in the output file at the cost of increasing
4905 link processing time. Merging is enabled by default.
4907 @item -fdebug-prefix-map=@var{old}=@var{new}
4908 @opindex fdebug-prefix-map
4909 When compiling files in directory @file{@var{old}}, record debugging
4910 information describing them as in @file{@var{new}} instead.
4912 @item -fno-dwarf2-cfi-asm
4913 @opindex fdwarf2-cfi-asm
4914 @opindex fno-dwarf2-cfi-asm
4915 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4916 instead of using GAS @code{.cfi_*} directives.
4918 @cindex @command{prof}
4921 Generate extra code to write profile information suitable for the
4922 analysis program @command{prof}. You must use this option when compiling
4923 the source files you want data about, and you must also use it when
4926 @cindex @command{gprof}
4929 Generate extra code to write profile information suitable for the
4930 analysis program @command{gprof}. You must use this option when compiling
4931 the source files you want data about, and you must also use it when
4936 Makes the compiler print out each function name as it is compiled, and
4937 print some statistics about each pass when it finishes.
4940 @opindex ftime-report
4941 Makes the compiler print some statistics about the time consumed by each
4942 pass when it finishes.
4945 @opindex fmem-report
4946 Makes the compiler print some statistics about permanent memory
4947 allocation when it finishes.
4949 @item -fpre-ipa-mem-report
4950 @opindex fpre-ipa-mem-report
4951 @item -fpost-ipa-mem-report
4952 @opindex fpost-ipa-mem-report
4953 Makes the compiler print some statistics about permanent memory
4954 allocation before or after interprocedural optimization.
4957 @opindex fstack-usage
4958 Makes the compiler output stack usage information for the program, on a
4959 per-function basis. The filename for the dump is made by appending
4960 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4961 the output file, if explicitly specified and it is not an executable,
4962 otherwise it is the basename of the source file. An entry is made up
4967 The name of the function.
4971 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4974 The qualifier @code{static} means that the function manipulates the stack
4975 statically: a fixed number of bytes are allocated for the frame on function
4976 entry and released on function exit; no stack adjustments are otherwise made
4977 in the function. The second field is this fixed number of bytes.
4979 The qualifier @code{dynamic} means that the function manipulates the stack
4980 dynamically: in addition to the static allocation described above, stack
4981 adjustments are made in the body of the function, for example to push/pop
4982 arguments around function calls. If the qualifier @code{bounded} is also
4983 present, the amount of these adjustments is bounded at compile-time and
4984 the second field is an upper bound of the total amount of stack used by
4985 the function. If it is not present, the amount of these adjustments is
4986 not bounded at compile-time and the second field only represents the
4989 @item -fprofile-arcs
4990 @opindex fprofile-arcs
4991 Add code so that program flow @dfn{arcs} are instrumented. During
4992 execution the program records how many times each branch and call is
4993 executed and how many times it is taken or returns. When the compiled
4994 program exits it saves this data to a file called
4995 @file{@var{auxname}.gcda} for each source file. The data may be used for
4996 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4997 test coverage analysis (@option{-ftest-coverage}). Each object file's
4998 @var{auxname} is generated from the name of the output file, if
4999 explicitly specified and it is not the final executable, otherwise it is
5000 the basename of the source file. In both cases any suffix is removed
5001 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5002 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5003 @xref{Cross-profiling}.
5005 @cindex @command{gcov}
5009 This option is used to compile and link code instrumented for coverage
5010 analysis. The option is a synonym for @option{-fprofile-arcs}
5011 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5012 linking). See the documentation for those options for more details.
5017 Compile the source files with @option{-fprofile-arcs} plus optimization
5018 and code generation options. For test coverage analysis, use the
5019 additional @option{-ftest-coverage} option. You do not need to profile
5020 every source file in a program.
5023 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5024 (the latter implies the former).
5027 Run the program on a representative workload to generate the arc profile
5028 information. This may be repeated any number of times. You can run
5029 concurrent instances of your program, and provided that the file system
5030 supports locking, the data files will be correctly updated. Also
5031 @code{fork} calls are detected and correctly handled (double counting
5035 For profile-directed optimizations, compile the source files again with
5036 the same optimization and code generation options plus
5037 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5038 Control Optimization}).
5041 For test coverage analysis, use @command{gcov} to produce human readable
5042 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5043 @command{gcov} documentation for further information.
5047 With @option{-fprofile-arcs}, for each function of your program GCC
5048 creates a program flow graph, then finds a spanning tree for the graph.
5049 Only arcs that are not on the spanning tree have to be instrumented: the
5050 compiler adds code to count the number of times that these arcs are
5051 executed. When an arc is the only exit or only entrance to a block, the
5052 instrumentation code can be added to the block; otherwise, a new basic
5053 block must be created to hold the instrumentation code.
5056 @item -ftest-coverage
5057 @opindex ftest-coverage
5058 Produce a notes file that the @command{gcov} code-coverage utility
5059 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5060 show program coverage. Each source file's note file is called
5061 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5062 above for a description of @var{auxname} and instructions on how to
5063 generate test coverage data. Coverage data will match the source files
5064 more closely, if you do not optimize.
5066 @item -fdbg-cnt-list
5067 @opindex fdbg-cnt-list
5068 Print the name and the counter upper bound for all debug counters.
5071 @item -fdbg-cnt=@var{counter-value-list}
5073 Set the internal debug counter upper bound. @var{counter-value-list}
5074 is a comma-separated list of @var{name}:@var{value} pairs
5075 which sets the upper bound of each debug counter @var{name} to @var{value}.
5076 All debug counters have the initial upper bound of @var{UINT_MAX},
5077 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5078 e.g. With -fdbg-cnt=dce:10,tail_call:0
5079 dbg_cnt(dce) will return true only for first 10 invocations
5081 @itemx -fenable-@var{kind}-@var{pass}
5082 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5086 This is a set of debugging options that are used to explicitly disable/enable
5087 optimization passes. For compiler users, regular options for enabling/disabling
5088 passes should be used instead.
5092 @item -fdisable-ipa-@var{pass}
5093 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5094 statically invoked in the compiler multiple times, the pass name should be
5095 appended with a sequential number starting from 1.
5097 @item -fdisable-rtl-@var{pass}
5098 @item -fdisable-rtl-@var{pass}=@var{range-list}
5099 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5100 statically invoked in the compiler multiple times, the pass name should be
5101 appended with a sequential number starting from 1. @var{range-list} is a comma
5102 seperated list of function ranges or assembler names. Each range is a number
5103 pair seperated by a colon. The range is inclusive in both ends. If the range
5104 is trivial, the number pair can be simplified as a single number. If the
5105 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5106 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5107 function header of a dump file, and the pass names can be dumped by using
5108 option @option{-fdump-passes}.
5110 @item -fdisable-tree-@var{pass}
5111 @item -fdisable-tree-@var{pass}=@var{range-list}
5112 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5115 @item -fenable-ipa-@var{pass}
5116 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5117 statically invoked in the compiler multiple times, the pass name should be
5118 appended with a sequential number starting from 1.
5120 @item -fenable-rtl-@var{pass}
5121 @item -fenable-rtl-@var{pass}=@var{range-list}
5122 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5123 description and examples.
5125 @item -fenable-tree-@var{pass}
5126 @item -fenable-tree-@var{pass}=@var{range-list}
5127 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5128 of option arguments.
5132 # disable ccp1 for all functions
5134 # disable complete unroll for function whose cgraph node uid is 1
5135 -fenable-tree-cunroll=1
5136 # disable gcse2 for functions at the following ranges [1,1],
5137 # [300,400], and [400,1000]
5138 # disable gcse2 for functions foo and foo2
5139 -fdisable-rtl-gcse2=foo,foo2
5140 # disable early inlining
5141 -fdisable-tree-einline
5142 # disable ipa inlining
5143 -fdisable-ipa-inline
5144 # enable tree full unroll
5145 -fenable-tree-unroll
5151 @item -d@var{letters}
5152 @itemx -fdump-rtl-@var{pass}
5154 Says to make debugging dumps during compilation at times specified by
5155 @var{letters}. This is used for debugging the RTL-based passes of the
5156 compiler. The file names for most of the dumps are made by appending
5157 a pass number and a word to the @var{dumpname}, and the files are
5158 created in the directory of the output file. Note that the pass
5159 number is computed statically as passes get registered into the pass
5160 manager. Thus the numbering is not related to the dynamic order of
5161 execution of passes. In particular, a pass installed by a plugin
5162 could have a number over 200 even if it executed quite early.
5163 @var{dumpname} is generated from the name of the output file, if
5164 explicitly specified and it is not an executable, otherwise it is the
5165 basename of the source file. These switches may have different effects
5166 when @option{-E} is used for preprocessing.
5168 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5169 @option{-d} option @var{letters}. Here are the possible
5170 letters for use in @var{pass} and @var{letters}, and their meanings:
5174 @item -fdump-rtl-alignments
5175 @opindex fdump-rtl-alignments
5176 Dump after branch alignments have been computed.
5178 @item -fdump-rtl-asmcons
5179 @opindex fdump-rtl-asmcons
5180 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5182 @item -fdump-rtl-auto_inc_dec
5183 @opindex fdump-rtl-auto_inc_dec
5184 Dump after auto-inc-dec discovery. This pass is only run on
5185 architectures that have auto inc or auto dec instructions.
5187 @item -fdump-rtl-barriers
5188 @opindex fdump-rtl-barriers
5189 Dump after cleaning up the barrier instructions.
5191 @item -fdump-rtl-bbpart
5192 @opindex fdump-rtl-bbpart
5193 Dump after partitioning hot and cold basic blocks.
5195 @item -fdump-rtl-bbro
5196 @opindex fdump-rtl-bbro
5197 Dump after block reordering.
5199 @item -fdump-rtl-btl1
5200 @itemx -fdump-rtl-btl2
5201 @opindex fdump-rtl-btl2
5202 @opindex fdump-rtl-btl2
5203 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5204 after the two branch
5205 target load optimization passes.
5207 @item -fdump-rtl-bypass
5208 @opindex fdump-rtl-bypass
5209 Dump after jump bypassing and control flow optimizations.
5211 @item -fdump-rtl-combine
5212 @opindex fdump-rtl-combine
5213 Dump after the RTL instruction combination pass.
5215 @item -fdump-rtl-compgotos
5216 @opindex fdump-rtl-compgotos
5217 Dump after duplicating the computed gotos.
5219 @item -fdump-rtl-ce1
5220 @itemx -fdump-rtl-ce2
5221 @itemx -fdump-rtl-ce3
5222 @opindex fdump-rtl-ce1
5223 @opindex fdump-rtl-ce2
5224 @opindex fdump-rtl-ce3
5225 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5226 @option{-fdump-rtl-ce3} enable dumping after the three
5227 if conversion passes.
5229 @itemx -fdump-rtl-cprop_hardreg
5230 @opindex fdump-rtl-cprop_hardreg
5231 Dump after hard register copy propagation.
5233 @itemx -fdump-rtl-csa
5234 @opindex fdump-rtl-csa
5235 Dump after combining stack adjustments.
5237 @item -fdump-rtl-cse1
5238 @itemx -fdump-rtl-cse2
5239 @opindex fdump-rtl-cse1
5240 @opindex fdump-rtl-cse2
5241 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5242 the two common sub-expression elimination passes.
5244 @itemx -fdump-rtl-dce
5245 @opindex fdump-rtl-dce
5246 Dump after the standalone dead code elimination passes.
5248 @itemx -fdump-rtl-dbr
5249 @opindex fdump-rtl-dbr
5250 Dump after delayed branch scheduling.
5252 @item -fdump-rtl-dce1
5253 @itemx -fdump-rtl-dce2
5254 @opindex fdump-rtl-dce1
5255 @opindex fdump-rtl-dce2
5256 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5257 the two dead store elimination passes.
5260 @opindex fdump-rtl-eh
5261 Dump after finalization of EH handling code.
5263 @item -fdump-rtl-eh_ranges
5264 @opindex fdump-rtl-eh_ranges
5265 Dump after conversion of EH handling range regions.
5267 @item -fdump-rtl-expand
5268 @opindex fdump-rtl-expand
5269 Dump after RTL generation.
5271 @item -fdump-rtl-fwprop1
5272 @itemx -fdump-rtl-fwprop2
5273 @opindex fdump-rtl-fwprop1
5274 @opindex fdump-rtl-fwprop2
5275 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5276 dumping after the two forward propagation passes.
5278 @item -fdump-rtl-gcse1
5279 @itemx -fdump-rtl-gcse2
5280 @opindex fdump-rtl-gcse1
5281 @opindex fdump-rtl-gcse2
5282 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5283 after global common subexpression elimination.
5285 @item -fdump-rtl-init-regs
5286 @opindex fdump-rtl-init-regs
5287 Dump after the initialization of the registers.
5289 @item -fdump-rtl-initvals
5290 @opindex fdump-rtl-initvals
5291 Dump after the computation of the initial value sets.
5293 @itemx -fdump-rtl-into_cfglayout
5294 @opindex fdump-rtl-into_cfglayout
5295 Dump after converting to cfglayout mode.
5297 @item -fdump-rtl-ira
5298 @opindex fdump-rtl-ira
5299 Dump after iterated register allocation.
5301 @item -fdump-rtl-jump
5302 @opindex fdump-rtl-jump
5303 Dump after the second jump optimization.
5305 @item -fdump-rtl-loop2
5306 @opindex fdump-rtl-loop2
5307 @option{-fdump-rtl-loop2} enables dumping after the rtl
5308 loop optimization passes.
5310 @item -fdump-rtl-mach
5311 @opindex fdump-rtl-mach
5312 Dump after performing the machine dependent reorganization pass, if that
5315 @item -fdump-rtl-mode_sw
5316 @opindex fdump-rtl-mode_sw
5317 Dump after removing redundant mode switches.
5319 @item -fdump-rtl-rnreg
5320 @opindex fdump-rtl-rnreg
5321 Dump after register renumbering.
5323 @itemx -fdump-rtl-outof_cfglayout
5324 @opindex fdump-rtl-outof_cfglayout
5325 Dump after converting from cfglayout mode.
5327 @item -fdump-rtl-peephole2
5328 @opindex fdump-rtl-peephole2
5329 Dump after the peephole pass.
5331 @item -fdump-rtl-postreload
5332 @opindex fdump-rtl-postreload
5333 Dump after post-reload optimizations.
5335 @itemx -fdump-rtl-pro_and_epilogue
5336 @opindex fdump-rtl-pro_and_epilogue
5337 Dump after generating the function pro and epilogues.
5339 @item -fdump-rtl-regmove
5340 @opindex fdump-rtl-regmove
5341 Dump after the register move pass.
5343 @item -fdump-rtl-sched1
5344 @itemx -fdump-rtl-sched2
5345 @opindex fdump-rtl-sched1
5346 @opindex fdump-rtl-sched2
5347 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5348 after the basic block scheduling passes.
5350 @item -fdump-rtl-see
5351 @opindex fdump-rtl-see
5352 Dump after sign extension elimination.
5354 @item -fdump-rtl-seqabstr
5355 @opindex fdump-rtl-seqabstr
5356 Dump after common sequence discovery.
5358 @item -fdump-rtl-shorten
5359 @opindex fdump-rtl-shorten
5360 Dump after shortening branches.
5362 @item -fdump-rtl-sibling
5363 @opindex fdump-rtl-sibling
5364 Dump after sibling call optimizations.
5366 @item -fdump-rtl-split1
5367 @itemx -fdump-rtl-split2
5368 @itemx -fdump-rtl-split3
5369 @itemx -fdump-rtl-split4
5370 @itemx -fdump-rtl-split5
5371 @opindex fdump-rtl-split1
5372 @opindex fdump-rtl-split2
5373 @opindex fdump-rtl-split3
5374 @opindex fdump-rtl-split4
5375 @opindex fdump-rtl-split5
5376 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5377 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5378 @option{-fdump-rtl-split5} enable dumping after five rounds of
5379 instruction splitting.
5381 @item -fdump-rtl-sms
5382 @opindex fdump-rtl-sms
5383 Dump after modulo scheduling. This pass is only run on some
5386 @item -fdump-rtl-stack
5387 @opindex fdump-rtl-stack
5388 Dump after conversion from GCC's "flat register file" registers to the
5389 x87's stack-like registers. This pass is only run on x86 variants.
5391 @item -fdump-rtl-subreg1
5392 @itemx -fdump-rtl-subreg2
5393 @opindex fdump-rtl-subreg1
5394 @opindex fdump-rtl-subreg2
5395 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5396 the two subreg expansion passes.
5398 @item -fdump-rtl-unshare
5399 @opindex fdump-rtl-unshare
5400 Dump after all rtl has been unshared.
5402 @item -fdump-rtl-vartrack
5403 @opindex fdump-rtl-vartrack
5404 Dump after variable tracking.
5406 @item -fdump-rtl-vregs
5407 @opindex fdump-rtl-vregs
5408 Dump after converting virtual registers to hard registers.
5410 @item -fdump-rtl-web
5411 @opindex fdump-rtl-web
5412 Dump after live range splitting.
5414 @item -fdump-rtl-regclass
5415 @itemx -fdump-rtl-subregs_of_mode_init
5416 @itemx -fdump-rtl-subregs_of_mode_finish
5417 @itemx -fdump-rtl-dfinit
5418 @itemx -fdump-rtl-dfinish
5419 @opindex fdump-rtl-regclass
5420 @opindex fdump-rtl-subregs_of_mode_init
5421 @opindex fdump-rtl-subregs_of_mode_finish
5422 @opindex fdump-rtl-dfinit
5423 @opindex fdump-rtl-dfinish
5424 These dumps are defined but always produce empty files.
5426 @item -fdump-rtl-all
5427 @opindex fdump-rtl-all
5428 Produce all the dumps listed above.
5432 Annotate the assembler output with miscellaneous debugging information.
5436 Dump all macro definitions, at the end of preprocessing, in addition to
5441 Produce a core dump whenever an error occurs.
5445 Print statistics on memory usage, at the end of the run, to
5450 Annotate the assembler output with a comment indicating which
5451 pattern and alternative was used. The length of each instruction is
5456 Dump the RTL in the assembler output as a comment before each instruction.
5457 Also turns on @option{-dp} annotation.
5461 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5462 dump a representation of the control flow graph suitable for viewing with VCG
5463 to @file{@var{file}.@var{pass}.vcg}.
5467 Just generate RTL for a function instead of compiling it. Usually used
5468 with @option{-fdump-rtl-expand}.
5472 @opindex fdump-noaddr
5473 When doing debugging dumps, suppress address output. This makes it more
5474 feasible to use diff on debugging dumps for compiler invocations with
5475 different compiler binaries and/or different
5476 text / bss / data / heap / stack / dso start locations.
5478 @item -fdump-unnumbered
5479 @opindex fdump-unnumbered
5480 When doing debugging dumps, suppress instruction numbers and address output.
5481 This makes it more feasible to use diff on debugging dumps for compiler
5482 invocations with different options, in particular with and without
5485 @item -fdump-unnumbered-links
5486 @opindex fdump-unnumbered-links
5487 When doing debugging dumps (see @option{-d} option above), suppress
5488 instruction numbers for the links to the previous and next instructions
5491 @item -fdump-translation-unit @r{(C++ only)}
5492 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5493 @opindex fdump-translation-unit
5494 Dump a representation of the tree structure for the entire translation
5495 unit to a file. The file name is made by appending @file{.tu} to the
5496 source file name, and the file is created in the same directory as the
5497 output file. If the @samp{-@var{options}} form is used, @var{options}
5498 controls the details of the dump as described for the
5499 @option{-fdump-tree} options.
5501 @item -fdump-class-hierarchy @r{(C++ only)}
5502 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5503 @opindex fdump-class-hierarchy
5504 Dump a representation of each class's hierarchy and virtual function
5505 table layout to a file. The file name is made by appending
5506 @file{.class} to the source file name, and the file is created in the
5507 same directory as the output file. If the @samp{-@var{options}} form
5508 is used, @var{options} controls the details of the dump as described
5509 for the @option{-fdump-tree} options.
5511 @item -fdump-ipa-@var{switch}
5513 Control the dumping at various stages of inter-procedural analysis
5514 language tree to a file. The file name is generated by appending a
5515 switch specific suffix to the source file name, and the file is created
5516 in the same directory as the output file. The following dumps are
5521 Enables all inter-procedural analysis dumps.
5524 Dumps information about call-graph optimization, unused function removal,
5525 and inlining decisions.
5528 Dump after function inlining.
5533 @opindex fdump-passes
5534 Dump the list of optimization passes that are turned on and off by
5535 the current command line options.
5537 @item -fdump-statistics-@var{option}
5538 @opindex fdump-statistics
5539 Enable and control dumping of pass statistics in a separate file. The
5540 file name is generated by appending a suffix ending in
5541 @samp{.statistics} to the source file name, and the file is created in
5542 the same directory as the output file. If the @samp{-@var{option}}
5543 form is used, @samp{-stats} will cause counters to be summed over the
5544 whole compilation unit while @samp{-details} will dump every event as
5545 the passes generate them. The default with no option is to sum
5546 counters for each function compiled.
5548 @item -fdump-tree-@var{switch}
5549 @itemx -fdump-tree-@var{switch}-@var{options}
5551 Control the dumping at various stages of processing the intermediate
5552 language tree to a file. The file name is generated by appending a
5553 switch specific suffix to the source file name, and the file is
5554 created in the same directory as the output file. If the
5555 @samp{-@var{options}} form is used, @var{options} is a list of
5556 @samp{-} separated options that control the details of the dump. Not
5557 all options are applicable to all dumps, those which are not
5558 meaningful will be ignored. The following options are available
5562 Print the address of each node. Usually this is not meaningful as it
5563 changes according to the environment and source file. Its primary use
5564 is for tying up a dump file with a debug environment.
5566 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5567 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5568 use working backward from mangled names in the assembly file.
5570 Inhibit dumping of members of a scope or body of a function merely
5571 because that scope has been reached. Only dump such items when they
5572 are directly reachable by some other path. When dumping pretty-printed
5573 trees, this option inhibits dumping the bodies of control structures.
5575 Print a raw representation of the tree. By default, trees are
5576 pretty-printed into a C-like representation.
5578 Enable more detailed dumps (not honored by every dump option).
5580 Enable dumping various statistics about the pass (not honored by every dump
5583 Enable showing basic block boundaries (disabled in raw dumps).
5585 Enable showing virtual operands for every statement.
5587 Enable showing line numbers for statements.
5589 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5591 Enable showing the tree dump for each statement.
5593 Enable showing the EH region number holding each statement.
5595 Enable showing scalar evolution analysis details.
5597 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5598 and @option{lineno}.
5601 The following tree dumps are possible:
5605 @opindex fdump-tree-original
5606 Dump before any tree based optimization, to @file{@var{file}.original}.
5609 @opindex fdump-tree-optimized
5610 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5613 @opindex fdump-tree-gimple
5614 Dump each function before and after the gimplification pass to a file. The
5615 file name is made by appending @file{.gimple} to the source file name.
5618 @opindex fdump-tree-cfg
5619 Dump the control flow graph of each function to a file. The file name is
5620 made by appending @file{.cfg} to the source file name.
5623 @opindex fdump-tree-vcg
5624 Dump the control flow graph of each function to a file in VCG format. The
5625 file name is made by appending @file{.vcg} to the source file name. Note
5626 that if the file contains more than one function, the generated file cannot
5627 be used directly by VCG@. You will need to cut and paste each function's
5628 graph into its own separate file first.
5631 @opindex fdump-tree-ch
5632 Dump each function after copying loop headers. The file name is made by
5633 appending @file{.ch} to the source file name.
5636 @opindex fdump-tree-ssa
5637 Dump SSA related information to a file. The file name is made by appending
5638 @file{.ssa} to the source file name.
5641 @opindex fdump-tree-alias
5642 Dump aliasing information for each function. The file name is made by
5643 appending @file{.alias} to the source file name.
5646 @opindex fdump-tree-ccp
5647 Dump each function after CCP@. The file name is made by appending
5648 @file{.ccp} to the source file name.
5651 @opindex fdump-tree-storeccp
5652 Dump each function after STORE-CCP@. The file name is made by appending
5653 @file{.storeccp} to the source file name.
5656 @opindex fdump-tree-pre
5657 Dump trees after partial redundancy elimination. The file name is made
5658 by appending @file{.pre} to the source file name.
5661 @opindex fdump-tree-fre
5662 Dump trees after full redundancy elimination. The file name is made
5663 by appending @file{.fre} to the source file name.
5666 @opindex fdump-tree-copyprop
5667 Dump trees after copy propagation. The file name is made
5668 by appending @file{.copyprop} to the source file name.
5670 @item store_copyprop
5671 @opindex fdump-tree-store_copyprop
5672 Dump trees after store copy-propagation. The file name is made
5673 by appending @file{.store_copyprop} to the source file name.
5676 @opindex fdump-tree-dce
5677 Dump each function after dead code elimination. The file name is made by
5678 appending @file{.dce} to the source file name.
5681 @opindex fdump-tree-mudflap
5682 Dump each function after adding mudflap instrumentation. The file name is
5683 made by appending @file{.mudflap} to the source file name.
5686 @opindex fdump-tree-sra
5687 Dump each function after performing scalar replacement of aggregates. The
5688 file name is made by appending @file{.sra} to the source file name.
5691 @opindex fdump-tree-sink
5692 Dump each function after performing code sinking. The file name is made
5693 by appending @file{.sink} to the source file name.
5696 @opindex fdump-tree-dom
5697 Dump each function after applying dominator tree optimizations. The file
5698 name is made by appending @file{.dom} to the source file name.
5701 @opindex fdump-tree-dse
5702 Dump each function after applying dead store elimination. The file
5703 name is made by appending @file{.dse} to the source file name.
5706 @opindex fdump-tree-phiopt
5707 Dump each function after optimizing PHI nodes into straightline code. The file
5708 name is made by appending @file{.phiopt} to the source file name.
5711 @opindex fdump-tree-forwprop
5712 Dump each function after forward propagating single use variables. The file
5713 name is made by appending @file{.forwprop} to the source file name.
5716 @opindex fdump-tree-copyrename
5717 Dump each function after applying the copy rename optimization. The file
5718 name is made by appending @file{.copyrename} to the source file name.
5721 @opindex fdump-tree-nrv
5722 Dump each function after applying the named return value optimization on
5723 generic trees. The file name is made by appending @file{.nrv} to the source
5727 @opindex fdump-tree-vect
5728 Dump each function after applying vectorization of loops. The file name is
5729 made by appending @file{.vect} to the source file name.
5732 @opindex fdump-tree-slp
5733 Dump each function after applying vectorization of basic blocks. The file name
5734 is made by appending @file{.slp} to the source file name.
5737 @opindex fdump-tree-vrp
5738 Dump each function after Value Range Propagation (VRP). The file name
5739 is made by appending @file{.vrp} to the source file name.
5742 @opindex fdump-tree-all
5743 Enable all the available tree dumps with the flags provided in this option.
5746 @item -ftree-vectorizer-verbose=@var{n}
5747 @opindex ftree-vectorizer-verbose
5748 This option controls the amount of debugging output the vectorizer prints.
5749 This information is written to standard error, unless
5750 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5751 in which case it is output to the usual dump listing file, @file{.vect}.
5752 For @var{n}=0 no diagnostic information is reported.
5753 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5754 and the total number of loops that got vectorized.
5755 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5756 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5757 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5758 level that @option{-fdump-tree-vect-stats} uses.
5759 Higher verbosity levels mean either more information dumped for each
5760 reported loop, or same amount of information reported for more loops:
5761 if @var{n}=3, vectorizer cost model information is reported.
5762 If @var{n}=4, alignment related information is added to the reports.
5763 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5764 memory access-patterns) is added to the reports.
5765 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5766 that did not pass the first analysis phase (i.e., may not be countable, or
5767 may have complicated control-flow).
5768 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5769 If @var{n}=8, SLP related information is added to the reports.
5770 For @var{n}=9, all the information the vectorizer generates during its
5771 analysis and transformation is reported. This is the same verbosity level
5772 that @option{-fdump-tree-vect-details} uses.
5774 @item -frandom-seed=@var{string}
5775 @opindex frandom-seed
5776 This option provides a seed that GCC uses when it would otherwise use
5777 random numbers. It is used to generate certain symbol names
5778 that have to be different in every compiled file. It is also used to
5779 place unique stamps in coverage data files and the object files that
5780 produce them. You can use the @option{-frandom-seed} option to produce
5781 reproducibly identical object files.
5783 The @var{string} should be different for every file you compile.
5785 @item -fsched-verbose=@var{n}
5786 @opindex fsched-verbose
5787 On targets that use instruction scheduling, this option controls the
5788 amount of debugging output the scheduler prints. This information is
5789 written to standard error, unless @option{-fdump-rtl-sched1} or
5790 @option{-fdump-rtl-sched2} is specified, in which case it is output
5791 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5792 respectively. However for @var{n} greater than nine, the output is
5793 always printed to standard error.
5795 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5796 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5797 For @var{n} greater than one, it also output basic block probabilities,
5798 detailed ready list information and unit/insn info. For @var{n} greater
5799 than two, it includes RTL at abort point, control-flow and regions info.
5800 And for @var{n} over four, @option{-fsched-verbose} also includes
5804 @itemx -save-temps=cwd
5806 Store the usual ``temporary'' intermediate files permanently; place them
5807 in the current directory and name them based on the source file. Thus,
5808 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5809 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5810 preprocessed @file{foo.i} output file even though the compiler now
5811 normally uses an integrated preprocessor.
5813 When used in combination with the @option{-x} command line option,
5814 @option{-save-temps} is sensible enough to avoid over writing an
5815 input source file with the same extension as an intermediate file.
5816 The corresponding intermediate file may be obtained by renaming the
5817 source file before using @option{-save-temps}.
5819 If you invoke GCC in parallel, compiling several different source
5820 files that share a common base name in different subdirectories or the
5821 same source file compiled for multiple output destinations, it is
5822 likely that the different parallel compilers will interfere with each
5823 other, and overwrite the temporary files. For instance:
5826 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5827 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5830 may result in @file{foo.i} and @file{foo.o} being written to
5831 simultaneously by both compilers.
5833 @item -save-temps=obj
5834 @opindex save-temps=obj
5835 Store the usual ``temporary'' intermediate files permanently. If the
5836 @option{-o} option is used, the temporary files are based on the
5837 object file. If the @option{-o} option is not used, the
5838 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5843 gcc -save-temps=obj -c foo.c
5844 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5845 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5848 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5849 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5850 @file{dir2/yfoobar.o}.
5852 @item -time@r{[}=@var{file}@r{]}
5854 Report the CPU time taken by each subprocess in the compilation
5855 sequence. For C source files, this is the compiler proper and assembler
5856 (plus the linker if linking is done).
5858 Without the specification of an output file, the output looks like this:
5865 The first number on each line is the ``user time'', that is time spent
5866 executing the program itself. The second number is ``system time'',
5867 time spent executing operating system routines on behalf of the program.
5868 Both numbers are in seconds.
5870 With the specification of an output file, the output is appended to the
5871 named file, and it looks like this:
5874 0.12 0.01 cc1 @var{options}
5875 0.00 0.01 as @var{options}
5878 The ``user time'' and the ``system time'' are moved before the program
5879 name, and the options passed to the program are displayed, so that one
5880 can later tell what file was being compiled, and with which options.
5882 @item -fvar-tracking
5883 @opindex fvar-tracking
5884 Run variable tracking pass. It computes where variables are stored at each
5885 position in code. Better debugging information is then generated
5886 (if the debugging information format supports this information).
5888 It is enabled by default when compiling with optimization (@option{-Os},
5889 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5890 the debug info format supports it.
5892 @item -fvar-tracking-assignments
5893 @opindex fvar-tracking-assignments
5894 @opindex fno-var-tracking-assignments
5895 Annotate assignments to user variables early in the compilation and
5896 attempt to carry the annotations over throughout the compilation all the
5897 way to the end, in an attempt to improve debug information while
5898 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5900 It can be enabled even if var-tracking is disabled, in which case
5901 annotations will be created and maintained, but discarded at the end.
5903 @item -fvar-tracking-assignments-toggle
5904 @opindex fvar-tracking-assignments-toggle
5905 @opindex fno-var-tracking-assignments-toggle
5906 Toggle @option{-fvar-tracking-assignments}, in the same way that
5907 @option{-gtoggle} toggles @option{-g}.
5909 @item -print-file-name=@var{library}
5910 @opindex print-file-name
5911 Print the full absolute name of the library file @var{library} that
5912 would be used when linking---and don't do anything else. With this
5913 option, GCC does not compile or link anything; it just prints the
5916 @item -print-multi-directory
5917 @opindex print-multi-directory
5918 Print the directory name corresponding to the multilib selected by any
5919 other switches present in the command line. This directory is supposed
5920 to exist in @env{GCC_EXEC_PREFIX}.
5922 @item -print-multi-lib
5923 @opindex print-multi-lib
5924 Print the mapping from multilib directory names to compiler switches
5925 that enable them. The directory name is separated from the switches by
5926 @samp{;}, and each switch starts with an @samp{@@} instead of the
5927 @samp{-}, without spaces between multiple switches. This is supposed to
5928 ease shell-processing.
5930 @item -print-multi-os-directory
5931 @opindex print-multi-os-directory
5932 Print the path to OS libraries for the selected
5933 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5934 present in the @file{lib} subdirectory and no multilibs are used, this is
5935 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5936 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5937 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5938 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5940 @item -print-prog-name=@var{program}
5941 @opindex print-prog-name
5942 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5944 @item -print-libgcc-file-name
5945 @opindex print-libgcc-file-name
5946 Same as @option{-print-file-name=libgcc.a}.
5948 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5949 but you do want to link with @file{libgcc.a}. You can do
5952 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5955 @item -print-search-dirs
5956 @opindex print-search-dirs
5957 Print the name of the configured installation directory and a list of
5958 program and library directories @command{gcc} will search---and don't do anything else.
5960 This is useful when @command{gcc} prints the error message
5961 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5962 To resolve this you either need to put @file{cpp0} and the other compiler
5963 components where @command{gcc} expects to find them, or you can set the environment
5964 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5965 Don't forget the trailing @samp{/}.
5966 @xref{Environment Variables}.
5968 @item -print-sysroot
5969 @opindex print-sysroot
5970 Print the target sysroot directory that will be used during
5971 compilation. This is the target sysroot specified either at configure
5972 time or using the @option{--sysroot} option, possibly with an extra
5973 suffix that depends on compilation options. If no target sysroot is
5974 specified, the option prints nothing.
5976 @item -print-sysroot-headers-suffix
5977 @opindex print-sysroot-headers-suffix
5978 Print the suffix added to the target sysroot when searching for
5979 headers, or give an error if the compiler is not configured with such
5980 a suffix---and don't do anything else.
5983 @opindex dumpmachine
5984 Print the compiler's target machine (for example,
5985 @samp{i686-pc-linux-gnu})---and don't do anything else.
5988 @opindex dumpversion
5989 Print the compiler version (for example, @samp{3.0})---and don't do
5994 Print the compiler's built-in specs---and don't do anything else. (This
5995 is used when GCC itself is being built.) @xref{Spec Files}.
5997 @item -feliminate-unused-debug-types
5998 @opindex feliminate-unused-debug-types
5999 Normally, when producing DWARF2 output, GCC will emit debugging
6000 information for all types declared in a compilation
6001 unit, regardless of whether or not they are actually used
6002 in that compilation unit. Sometimes this is useful, such as
6003 if, in the debugger, you want to cast a value to a type that is
6004 not actually used in your program (but is declared). More often,
6005 however, this results in a significant amount of wasted space.
6006 With this option, GCC will avoid producing debug symbol output
6007 for types that are nowhere used in the source file being compiled.
6010 @node Optimize Options
6011 @section Options That Control Optimization
6012 @cindex optimize options
6013 @cindex options, optimization
6015 These options control various sorts of optimizations.
6017 Without any optimization option, the compiler's goal is to reduce the
6018 cost of compilation and to make debugging produce the expected
6019 results. Statements are independent: if you stop the program with a
6020 breakpoint between statements, you can then assign a new value to any
6021 variable or change the program counter to any other statement in the
6022 function and get exactly the results you would expect from the source
6025 Turning on optimization flags makes the compiler attempt to improve
6026 the performance and/or code size at the expense of compilation time
6027 and possibly the ability to debug the program.
6029 The compiler performs optimization based on the knowledge it has of the
6030 program. Compiling multiple files at once to a single output file mode allows
6031 the compiler to use information gained from all of the files when compiling
6034 Not all optimizations are controlled directly by a flag. Only
6035 optimizations that have a flag are listed in this section.
6037 Most optimizations are only enabled if an @option{-O} level is set on
6038 the command line. Otherwise they are disabled, even if individual
6039 optimization flags are specified.
6041 Depending on the target and how GCC was configured, a slightly different
6042 set of optimizations may be enabled at each @option{-O} level than
6043 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6044 to find out the exact set of optimizations that are enabled at each level.
6045 @xref{Overall Options}, for examples.
6052 Optimize. Optimizing compilation takes somewhat more time, and a lot
6053 more memory for a large function.
6055 With @option{-O}, the compiler tries to reduce code size and execution
6056 time, without performing any optimizations that take a great deal of
6059 @option{-O} turns on the following optimization flags:
6063 -fcprop-registers @gol
6066 -fdelayed-branch @gol
6068 -fguess-branch-probability @gol
6069 -fif-conversion2 @gol
6070 -fif-conversion @gol
6071 -fipa-pure-const @gol
6073 -fipa-reference @gol
6075 -fsplit-wide-types @gol
6077 -ftree-builtin-call-dce @gol
6080 -ftree-copyrename @gol
6082 -ftree-dominator-opts @gol
6084 -ftree-forwprop @gol
6092 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6093 where doing so does not interfere with debugging.
6097 Optimize even more. GCC performs nearly all supported optimizations
6098 that do not involve a space-speed tradeoff.
6099 As compared to @option{-O}, this option increases both compilation time
6100 and the performance of the generated code.
6102 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6103 also turns on the following optimization flags:
6104 @gccoptlist{-fthread-jumps @gol
6105 -falign-functions -falign-jumps @gol
6106 -falign-loops -falign-labels @gol
6109 -fcse-follow-jumps -fcse-skip-blocks @gol
6110 -fdelete-null-pointer-checks @gol
6112 -fexpensive-optimizations @gol
6113 -fgcse -fgcse-lm @gol
6114 -finline-small-functions @gol
6115 -findirect-inlining @gol
6117 -foptimize-sibling-calls @gol
6118 -fpartial-inlining @gol
6121 -freorder-blocks -freorder-functions @gol
6122 -frerun-cse-after-loop @gol
6123 -fsched-interblock -fsched-spec @gol
6124 -fschedule-insns -fschedule-insns2 @gol
6125 -fstrict-aliasing -fstrict-overflow @gol
6126 -ftree-switch-conversion @gol
6130 Please note the warning under @option{-fgcse} about
6131 invoking @option{-O2} on programs that use computed gotos.
6135 Optimize yet more. @option{-O3} turns on all optimizations specified
6136 by @option{-O2} and also turns on the @option{-finline-functions},
6137 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6138 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6139 @option{-fipa-cp-clone} options.
6143 Reduce compilation time and make debugging produce the expected
6144 results. This is the default.
6148 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6149 do not typically increase code size. It also performs further
6150 optimizations designed to reduce code size.
6152 @option{-Os} disables the following optimization flags:
6153 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6154 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6155 -fprefetch-loop-arrays -ftree-vect-loop-version}
6159 Disregard strict standards compliance. @option{-Ofast} enables all
6160 @option{-O3} optimizations. It also enables optimizations that are not
6161 valid for all standard compliant programs.
6162 It turns on @option{-ffast-math} and the Fortran-specific
6163 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6165 If you use multiple @option{-O} options, with or without level numbers,
6166 the last such option is the one that is effective.
6169 Options of the form @option{-f@var{flag}} specify machine-independent
6170 flags. Most flags have both positive and negative forms; the negative
6171 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6172 below, only one of the forms is listed---the one you typically will
6173 use. You can figure out the other form by either removing @samp{no-}
6176 The following options control specific optimizations. They are either
6177 activated by @option{-O} options or are related to ones that are. You
6178 can use the following flags in the rare cases when ``fine-tuning'' of
6179 optimizations to be performed is desired.
6182 @item -fno-default-inline
6183 @opindex fno-default-inline
6184 Do not make member functions inline by default merely because they are
6185 defined inside the class scope (C++ only). Otherwise, when you specify
6186 @w{@option{-O}}, member functions defined inside class scope are compiled
6187 inline by default; i.e., you don't need to add @samp{inline} in front of
6188 the member function name.
6190 @item -fno-defer-pop
6191 @opindex fno-defer-pop
6192 Always pop the arguments to each function call as soon as that function
6193 returns. For machines which must pop arguments after a function call,
6194 the compiler normally lets arguments accumulate on the stack for several
6195 function calls and pops them all at once.
6197 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6199 @item -fforward-propagate
6200 @opindex fforward-propagate
6201 Perform a forward propagation pass on RTL@. The pass tries to combine two
6202 instructions and checks if the result can be simplified. If loop unrolling
6203 is active, two passes are performed and the second is scheduled after
6206 This option is enabled by default at optimization levels @option{-O},
6207 @option{-O2}, @option{-O3}, @option{-Os}.
6209 @item -ffp-contract=@var{style}
6210 @opindex ffp-contract
6211 @option{-ffp-contract=off} disables floating-point expression contraction.
6212 @option{-ffp-contract=fast} enables floating-point expression contraction
6213 such as forming of fused multiply-add operations if the target has
6214 native support for them.
6215 @option{-ffp-contract=on} enables floating-point expression contraction
6216 if allowed by the language standard. This is currently not implemented
6217 and treated equal to @option{-ffp-contract=off}.
6219 The default is @option{-ffp-contract=fast}.
6221 @item -fomit-frame-pointer
6222 @opindex fomit-frame-pointer
6223 Don't keep the frame pointer in a register for functions that
6224 don't need one. This avoids the instructions to save, set up and
6225 restore frame pointers; it also makes an extra register available
6226 in many functions. @strong{It also makes debugging impossible on
6229 On some machines, such as the VAX, this flag has no effect, because
6230 the standard calling sequence automatically handles the frame pointer
6231 and nothing is saved by pretending it doesn't exist. The
6232 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6233 whether a target machine supports this flag. @xref{Registers,,Register
6234 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6236 Starting with GCC version 4.6, the default setting (when not optimizing for
6237 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6238 @option{-fomit-frame-pointer}. The default can be reverted to
6239 @option{-fno-omit-frame-pointer} by configuring GCC with the
6240 @option{--enable-frame-pointer} configure option.
6242 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6244 @item -foptimize-sibling-calls
6245 @opindex foptimize-sibling-calls
6246 Optimize sibling and tail recursive calls.
6248 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6252 Don't pay attention to the @code{inline} keyword. Normally this option
6253 is used to keep the compiler from expanding any functions inline.
6254 Note that if you are not optimizing, no functions can be expanded inline.
6256 @item -finline-small-functions
6257 @opindex finline-small-functions
6258 Integrate functions into their callers when their body is smaller than expected
6259 function call code (so overall size of program gets smaller). The compiler
6260 heuristically decides which functions are simple enough to be worth integrating
6263 Enabled at level @option{-O2}.
6265 @item -findirect-inlining
6266 @opindex findirect-inlining
6267 Inline also indirect calls that are discovered to be known at compile
6268 time thanks to previous inlining. This option has any effect only
6269 when inlining itself is turned on by the @option{-finline-functions}
6270 or @option{-finline-small-functions} options.
6272 Enabled at level @option{-O2}.
6274 @item -finline-functions
6275 @opindex finline-functions
6276 Integrate all simple functions into their callers. The compiler
6277 heuristically decides which functions are simple enough to be worth
6278 integrating in this way.
6280 If all calls to a given function are integrated, and the function is
6281 declared @code{static}, then the function is normally not output as
6282 assembler code in its own right.
6284 Enabled at level @option{-O3}.
6286 @item -finline-functions-called-once
6287 @opindex finline-functions-called-once
6288 Consider all @code{static} functions called once for inlining into their
6289 caller even if they are not marked @code{inline}. If a call to a given
6290 function is integrated, then the function is not output as assembler code
6293 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6295 @item -fearly-inlining
6296 @opindex fearly-inlining
6297 Inline functions marked by @code{always_inline} and functions whose body seems
6298 smaller than the function call overhead early before doing
6299 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6300 makes profiling significantly cheaper and usually inlining faster on programs
6301 having large chains of nested wrapper functions.
6307 Perform interprocedural scalar replacement of aggregates, removal of
6308 unused parameters and replacement of parameters passed by reference
6309 by parameters passed by value.
6311 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6313 @item -finline-limit=@var{n}
6314 @opindex finline-limit
6315 By default, GCC limits the size of functions that can be inlined. This flag
6316 allows coarse control of this limit. @var{n} is the size of functions that
6317 can be inlined in number of pseudo instructions.
6319 Inlining is actually controlled by a number of parameters, which may be
6320 specified individually by using @option{--param @var{name}=@var{value}}.
6321 The @option{-finline-limit=@var{n}} option sets some of these parameters
6325 @item max-inline-insns-single
6326 is set to @var{n}/2.
6327 @item max-inline-insns-auto
6328 is set to @var{n}/2.
6331 See below for a documentation of the individual
6332 parameters controlling inlining and for the defaults of these parameters.
6334 @emph{Note:} there may be no value to @option{-finline-limit} that results
6335 in default behavior.
6337 @emph{Note:} pseudo instruction represents, in this particular context, an
6338 abstract measurement of function's size. In no way does it represent a count
6339 of assembly instructions and as such its exact meaning might change from one
6340 release to an another.
6342 @item -fno-keep-inline-dllexport
6343 @opindex -fno-keep-inline-dllexport
6344 This is a more fine-grained version of @option{-fkeep-inline-functions},
6345 which applies only to functions that are declared using the @code{dllexport}
6346 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6349 @item -fkeep-inline-functions
6350 @opindex fkeep-inline-functions
6351 In C, emit @code{static} functions that are declared @code{inline}
6352 into the object file, even if the function has been inlined into all
6353 of its callers. This switch does not affect functions using the
6354 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6355 inline functions into the object file.
6357 @item -fkeep-static-consts
6358 @opindex fkeep-static-consts
6359 Emit variables declared @code{static const} when optimization isn't turned
6360 on, even if the variables aren't referenced.
6362 GCC enables this option by default. If you want to force the compiler to
6363 check if the variable was referenced, regardless of whether or not
6364 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6366 @item -fmerge-constants
6367 @opindex fmerge-constants
6368 Attempt to merge identical constants (string constants and floating point
6369 constants) across compilation units.
6371 This option is the default for optimized compilation if the assembler and
6372 linker support it. Use @option{-fno-merge-constants} to inhibit this
6375 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6377 @item -fmerge-all-constants
6378 @opindex fmerge-all-constants
6379 Attempt to merge identical constants and identical variables.
6381 This option implies @option{-fmerge-constants}. In addition to
6382 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6383 arrays or initialized constant variables with integral or floating point
6384 types. Languages like C or C++ require each variable, including multiple
6385 instances of the same variable in recursive calls, to have distinct locations,
6386 so using this option will result in non-conforming
6389 @item -fmodulo-sched
6390 @opindex fmodulo-sched
6391 Perform swing modulo scheduling immediately before the first scheduling
6392 pass. This pass looks at innermost loops and reorders their
6393 instructions by overlapping different iterations.
6395 @item -fmodulo-sched-allow-regmoves
6396 @opindex fmodulo-sched-allow-regmoves
6397 Perform more aggressive SMS based modulo scheduling with register moves
6398 allowed. By setting this flag certain anti-dependences edges will be
6399 deleted which will trigger the generation of reg-moves based on the
6400 life-range analysis. This option is effective only with
6401 @option{-fmodulo-sched} enabled.
6403 @item -fno-branch-count-reg
6404 @opindex fno-branch-count-reg
6405 Do not use ``decrement and branch'' instructions on a count register,
6406 but instead generate a sequence of instructions that decrement a
6407 register, compare it against zero, then branch based upon the result.
6408 This option is only meaningful on architectures that support such
6409 instructions, which include x86, PowerPC, IA-64 and S/390.
6411 The default is @option{-fbranch-count-reg}.
6413 @item -fno-function-cse
6414 @opindex fno-function-cse
6415 Do not put function addresses in registers; make each instruction that
6416 calls a constant function contain the function's address explicitly.
6418 This option results in less efficient code, but some strange hacks
6419 that alter the assembler output may be confused by the optimizations
6420 performed when this option is not used.
6422 The default is @option{-ffunction-cse}
6424 @item -fno-zero-initialized-in-bss
6425 @opindex fno-zero-initialized-in-bss
6426 If the target supports a BSS section, GCC by default puts variables that
6427 are initialized to zero into BSS@. This can save space in the resulting
6430 This option turns off this behavior because some programs explicitly
6431 rely on variables going to the data section. E.g., so that the
6432 resulting executable can find the beginning of that section and/or make
6433 assumptions based on that.
6435 The default is @option{-fzero-initialized-in-bss}.
6437 @item -fmudflap -fmudflapth -fmudflapir
6441 @cindex bounds checking
6443 For front-ends that support it (C and C++), instrument all risky
6444 pointer/array dereferencing operations, some standard library
6445 string/heap functions, and some other associated constructs with
6446 range/validity tests. Modules so instrumented should be immune to
6447 buffer overflows, invalid heap use, and some other classes of C/C++
6448 programming errors. The instrumentation relies on a separate runtime
6449 library (@file{libmudflap}), which will be linked into a program if
6450 @option{-fmudflap} is given at link time. Run-time behavior of the
6451 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6452 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6455 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6456 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6457 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6458 instrumentation should ignore pointer reads. This produces less
6459 instrumentation (and therefore faster execution) and still provides
6460 some protection against outright memory corrupting writes, but allows
6461 erroneously read data to propagate within a program.
6463 @item -fthread-jumps
6464 @opindex fthread-jumps
6465 Perform optimizations where we check to see if a jump branches to a
6466 location where another comparison subsumed by the first is found. If
6467 so, the first branch is redirected to either the destination of the
6468 second branch or a point immediately following it, depending on whether
6469 the condition is known to be true or false.
6471 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6473 @item -fsplit-wide-types
6474 @opindex fsplit-wide-types
6475 When using a type that occupies multiple registers, such as @code{long
6476 long} on a 32-bit system, split the registers apart and allocate them
6477 independently. This normally generates better code for those types,
6478 but may make debugging more difficult.
6480 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6483 @item -fcse-follow-jumps
6484 @opindex fcse-follow-jumps
6485 In common subexpression elimination (CSE), scan through jump instructions
6486 when the target of the jump is not reached by any other path. For
6487 example, when CSE encounters an @code{if} statement with an
6488 @code{else} clause, CSE will follow the jump when the condition
6491 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6493 @item -fcse-skip-blocks
6494 @opindex fcse-skip-blocks
6495 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6496 follow jumps which conditionally skip over blocks. When CSE
6497 encounters a simple @code{if} statement with no else clause,
6498 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6499 body of the @code{if}.
6501 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6503 @item -frerun-cse-after-loop
6504 @opindex frerun-cse-after-loop
6505 Re-run common subexpression elimination after loop optimizations has been
6508 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6512 Perform a global common subexpression elimination pass.
6513 This pass also performs global constant and copy propagation.
6515 @emph{Note:} When compiling a program using computed gotos, a GCC
6516 extension, you may get better runtime performance if you disable
6517 the global common subexpression elimination pass by adding
6518 @option{-fno-gcse} to the command line.
6520 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6524 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6525 attempt to move loads which are only killed by stores into themselves. This
6526 allows a loop containing a load/store sequence to be changed to a load outside
6527 the loop, and a copy/store within the loop.
6529 Enabled by default when gcse is enabled.
6533 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6534 global common subexpression elimination. This pass will attempt to move
6535 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6536 loops containing a load/store sequence can be changed to a load before
6537 the loop and a store after the loop.
6539 Not enabled at any optimization level.
6543 When @option{-fgcse-las} is enabled, the global common subexpression
6544 elimination pass eliminates redundant loads that come after stores to the
6545 same memory location (both partial and full redundancies).
6547 Not enabled at any optimization level.
6549 @item -fgcse-after-reload
6550 @opindex fgcse-after-reload
6551 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6552 pass is performed after reload. The purpose of this pass is to cleanup
6555 @item -funsafe-loop-optimizations
6556 @opindex funsafe-loop-optimizations
6557 If given, the loop optimizer will assume that loop indices do not
6558 overflow, and that the loops with nontrivial exit condition are not
6559 infinite. This enables a wider range of loop optimizations even if
6560 the loop optimizer itself cannot prove that these assumptions are valid.
6561 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6562 if it finds this kind of loop.
6564 @item -fcrossjumping
6565 @opindex fcrossjumping
6566 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6567 resulting code may or may not perform better than without cross-jumping.
6569 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6571 @item -fauto-inc-dec
6572 @opindex fauto-inc-dec
6573 Combine increments or decrements of addresses with memory accesses.
6574 This pass is always skipped on architectures that do not have
6575 instructions to support this. Enabled by default at @option{-O} and
6576 higher on architectures that support this.
6580 Perform dead code elimination (DCE) on RTL@.
6581 Enabled by default at @option{-O} and higher.
6585 Perform dead store elimination (DSE) on RTL@.
6586 Enabled by default at @option{-O} and higher.
6588 @item -fif-conversion
6589 @opindex fif-conversion
6590 Attempt to transform conditional jumps into branch-less equivalents. This
6591 include use of conditional moves, min, max, set flags and abs instructions, and
6592 some tricks doable by standard arithmetics. The use of conditional execution
6593 on chips where it is available is controlled by @code{if-conversion2}.
6595 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6597 @item -fif-conversion2
6598 @opindex fif-conversion2
6599 Use conditional execution (where available) to transform conditional jumps into
6600 branch-less equivalents.
6602 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6604 @item -fdelete-null-pointer-checks
6605 @opindex fdelete-null-pointer-checks
6606 Assume that programs cannot safely dereference null pointers, and that
6607 no code or data element resides there. This enables simple constant
6608 folding optimizations at all optimization levels. In addition, other
6609 optimization passes in GCC use this flag to control global dataflow
6610 analyses that eliminate useless checks for null pointers; these assume
6611 that if a pointer is checked after it has already been dereferenced,
6614 Note however that in some environments this assumption is not true.
6615 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6616 for programs which depend on that behavior.
6618 Some targets, especially embedded ones, disable this option at all levels.
6619 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6620 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6621 are enabled independently at different optimization levels.
6623 @item -fdevirtualize
6624 @opindex fdevirtualize
6625 Attempt to convert calls to virtual functions to direct calls. This
6626 is done both within a procedure and interprocedurally as part of
6627 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6628 propagation (@option{-fipa-cp}).
6629 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6631 @item -fexpensive-optimizations
6632 @opindex fexpensive-optimizations
6633 Perform a number of minor optimizations that are relatively expensive.
6635 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6637 @item -foptimize-register-move
6639 @opindex foptimize-register-move
6641 Attempt to reassign register numbers in move instructions and as
6642 operands of other simple instructions in order to maximize the amount of
6643 register tying. This is especially helpful on machines with two-operand
6646 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6649 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6651 @item -fira-algorithm=@var{algorithm}
6652 Use specified coloring algorithm for the integrated register
6653 allocator. The @var{algorithm} argument should be @code{priority} or
6654 @code{CB}. The first algorithm specifies Chow's priority coloring,
6655 the second one specifies Chaitin-Briggs coloring. The second
6656 algorithm can be unimplemented for some architectures. If it is
6657 implemented, it is the default because Chaitin-Briggs coloring as a
6658 rule generates a better code.
6660 @item -fira-region=@var{region}
6661 Use specified regions for the integrated register allocator. The
6662 @var{region} argument should be one of @code{all}, @code{mixed}, or
6663 @code{one}. The first value means using all loops as register
6664 allocation regions, the second value which is the default means using
6665 all loops except for loops with small register pressure as the
6666 regions, and third one means using all function as a single region.
6667 The first value can give best result for machines with small size and
6668 irregular register set, the third one results in faster and generates
6669 decent code and the smallest size code, and the default value usually
6670 give the best results in most cases and for most architectures.
6672 @item -fira-loop-pressure
6673 @opindex fira-loop-pressure
6674 Use IRA to evaluate register pressure in loops for decision to move
6675 loop invariants. Usage of this option usually results in generation
6676 of faster and smaller code on machines with big register files (>= 32
6677 registers) but it can slow compiler down.
6679 This option is enabled at level @option{-O3} for some targets.
6681 @item -fno-ira-share-save-slots
6682 @opindex fno-ira-share-save-slots
6683 Switch off sharing stack slots used for saving call used hard
6684 registers living through a call. Each hard register will get a
6685 separate stack slot and as a result function stack frame will be
6688 @item -fno-ira-share-spill-slots
6689 @opindex fno-ira-share-spill-slots
6690 Switch off sharing stack slots allocated for pseudo-registers. Each
6691 pseudo-register which did not get a hard register will get a separate
6692 stack slot and as a result function stack frame will be bigger.
6694 @item -fira-verbose=@var{n}
6695 @opindex fira-verbose
6696 Set up how verbose dump file for the integrated register allocator
6697 will be. Default value is 5. If the value is greater or equal to 10,
6698 the dump file will be stderr as if the value were @var{n} minus 10.
6700 @item -fdelayed-branch
6701 @opindex fdelayed-branch
6702 If supported for the target machine, attempt to reorder instructions
6703 to exploit instruction slots available after delayed branch
6706 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6708 @item -fschedule-insns
6709 @opindex fschedule-insns
6710 If supported for the target machine, attempt to reorder instructions to
6711 eliminate execution stalls due to required data being unavailable. This
6712 helps machines that have slow floating point or memory load instructions
6713 by allowing other instructions to be issued until the result of the load
6714 or floating point instruction is required.
6716 Enabled at levels @option{-O2}, @option{-O3}.
6718 @item -fschedule-insns2
6719 @opindex fschedule-insns2
6720 Similar to @option{-fschedule-insns}, but requests an additional pass of
6721 instruction scheduling after register allocation has been done. This is
6722 especially useful on machines with a relatively small number of
6723 registers and where memory load instructions take more than one cycle.
6725 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6727 @item -fno-sched-interblock
6728 @opindex fno-sched-interblock
6729 Don't schedule instructions across basic blocks. This is normally
6730 enabled by default when scheduling before register allocation, i.e.@:
6731 with @option{-fschedule-insns} or at @option{-O2} or higher.
6733 @item -fno-sched-spec
6734 @opindex fno-sched-spec
6735 Don't allow speculative motion of non-load instructions. This is normally
6736 enabled by default when scheduling before register allocation, i.e.@:
6737 with @option{-fschedule-insns} or at @option{-O2} or higher.
6739 @item -fsched-pressure
6740 @opindex fsched-pressure
6741 Enable register pressure sensitive insn scheduling before the register
6742 allocation. This only makes sense when scheduling before register
6743 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6744 @option{-O2} or higher. Usage of this option can improve the
6745 generated code and decrease its size by preventing register pressure
6746 increase above the number of available hard registers and as a
6747 consequence register spills in the register allocation.
6749 @item -fsched-spec-load
6750 @opindex fsched-spec-load
6751 Allow speculative motion of some load instructions. This only makes
6752 sense when scheduling before register allocation, i.e.@: with
6753 @option{-fschedule-insns} or at @option{-O2} or higher.
6755 @item -fsched-spec-load-dangerous
6756 @opindex fsched-spec-load-dangerous
6757 Allow speculative motion of more load instructions. This only makes
6758 sense when scheduling before register allocation, i.e.@: with
6759 @option{-fschedule-insns} or at @option{-O2} or higher.
6761 @item -fsched-stalled-insns
6762 @itemx -fsched-stalled-insns=@var{n}
6763 @opindex fsched-stalled-insns
6764 Define how many insns (if any) can be moved prematurely from the queue
6765 of stalled insns into the ready list, during the second scheduling pass.
6766 @option{-fno-sched-stalled-insns} means that no insns will be moved
6767 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6768 on how many queued insns can be moved prematurely.
6769 @option{-fsched-stalled-insns} without a value is equivalent to
6770 @option{-fsched-stalled-insns=1}.
6772 @item -fsched-stalled-insns-dep
6773 @itemx -fsched-stalled-insns-dep=@var{n}
6774 @opindex fsched-stalled-insns-dep
6775 Define how many insn groups (cycles) will be examined for a dependency
6776 on a stalled insn that is candidate for premature removal from the queue
6777 of stalled insns. This has an effect only during the second scheduling pass,
6778 and only if @option{-fsched-stalled-insns} is used.
6779 @option{-fno-sched-stalled-insns-dep} is equivalent to
6780 @option{-fsched-stalled-insns-dep=0}.
6781 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6782 @option{-fsched-stalled-insns-dep=1}.
6784 @item -fsched2-use-superblocks
6785 @opindex fsched2-use-superblocks
6786 When scheduling after register allocation, do use superblock scheduling
6787 algorithm. Superblock scheduling allows motion across basic block boundaries
6788 resulting on faster schedules. This option is experimental, as not all machine
6789 descriptions used by GCC model the CPU closely enough to avoid unreliable
6790 results from the algorithm.
6792 This only makes sense when scheduling after register allocation, i.e.@: with
6793 @option{-fschedule-insns2} or at @option{-O2} or higher.
6795 @item -fsched-group-heuristic
6796 @opindex fsched-group-heuristic
6797 Enable the group heuristic in the scheduler. This heuristic favors
6798 the instruction that belongs to a schedule group. This is enabled
6799 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6800 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6802 @item -fsched-critical-path-heuristic
6803 @opindex fsched-critical-path-heuristic
6804 Enable the critical-path heuristic in the scheduler. This heuristic favors
6805 instructions on the critical path. This is enabled by default when
6806 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6807 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6809 @item -fsched-spec-insn-heuristic
6810 @opindex fsched-spec-insn-heuristic
6811 Enable the speculative instruction heuristic in the scheduler. This
6812 heuristic favors speculative instructions with greater dependency weakness.
6813 This is enabled by default when scheduling is enabled, i.e.@:
6814 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6815 or at @option{-O2} or higher.
6817 @item -fsched-rank-heuristic
6818 @opindex fsched-rank-heuristic
6819 Enable the rank heuristic in the scheduler. This heuristic favors
6820 the instruction belonging to a basic block with greater size or frequency.
6821 This is enabled by default when scheduling is enabled, i.e.@:
6822 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6823 at @option{-O2} or higher.
6825 @item -fsched-last-insn-heuristic
6826 @opindex fsched-last-insn-heuristic
6827 Enable the last-instruction heuristic in the scheduler. This heuristic
6828 favors the instruction that is less dependent on the last instruction
6829 scheduled. This is enabled by default when scheduling is enabled,
6830 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6831 at @option{-O2} or higher.
6833 @item -fsched-dep-count-heuristic
6834 @opindex fsched-dep-count-heuristic
6835 Enable the dependent-count heuristic in the scheduler. This heuristic
6836 favors the instruction that has more instructions depending on it.
6837 This is enabled by default when scheduling is enabled, i.e.@:
6838 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6839 at @option{-O2} or higher.
6841 @item -freschedule-modulo-scheduled-loops
6842 @opindex freschedule-modulo-scheduled-loops
6843 The modulo scheduling comes before the traditional scheduling, if a loop
6844 was modulo scheduled we may want to prevent the later scheduling passes
6845 from changing its schedule, we use this option to control that.
6847 @item -fselective-scheduling
6848 @opindex fselective-scheduling
6849 Schedule instructions using selective scheduling algorithm. Selective
6850 scheduling runs instead of the first scheduler pass.
6852 @item -fselective-scheduling2
6853 @opindex fselective-scheduling2
6854 Schedule instructions using selective scheduling algorithm. Selective
6855 scheduling runs instead of the second scheduler pass.
6857 @item -fsel-sched-pipelining
6858 @opindex fsel-sched-pipelining
6859 Enable software pipelining of innermost loops during selective scheduling.
6860 This option has no effect until one of @option{-fselective-scheduling} or
6861 @option{-fselective-scheduling2} is turned on.
6863 @item -fsel-sched-pipelining-outer-loops
6864 @opindex fsel-sched-pipelining-outer-loops
6865 When pipelining loops during selective scheduling, also pipeline outer loops.
6866 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6868 @item -fcaller-saves
6869 @opindex fcaller-saves
6870 Enable values to be allocated in registers that will be clobbered by
6871 function calls, by emitting extra instructions to save and restore the
6872 registers around such calls. Such allocation is done only when it
6873 seems to result in better code than would otherwise be produced.
6875 This option is always enabled by default on certain machines, usually
6876 those which have no call-preserved registers to use instead.
6878 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6880 @item -fcombine-stack-adjustments
6881 @opindex fcombine-stack-adjustments
6882 Tracks stack adjustments (pushes and pops) and stack memory references
6883 and then tries to find ways to combine them.
6885 Enabled by default at @option{-O1} and higher.
6887 @item -fconserve-stack
6888 @opindex fconserve-stack
6889 Attempt to minimize stack usage. The compiler will attempt to use less
6890 stack space, even if that makes the program slower. This option
6891 implies setting the @option{large-stack-frame} parameter to 100
6892 and the @option{large-stack-frame-growth} parameter to 400.
6894 @item -ftree-reassoc
6895 @opindex ftree-reassoc
6896 Perform reassociation on trees. This flag is enabled by default
6897 at @option{-O} and higher.
6901 Perform partial redundancy elimination (PRE) on trees. This flag is
6902 enabled by default at @option{-O2} and @option{-O3}.
6904 @item -ftree-forwprop
6905 @opindex ftree-forwprop
6906 Perform forward propagation on trees. This flag is enabled by default
6907 at @option{-O} and higher.
6911 Perform full redundancy elimination (FRE) on trees. The difference
6912 between FRE and PRE is that FRE only considers expressions
6913 that are computed on all paths leading to the redundant computation.
6914 This analysis is faster than PRE, though it exposes fewer redundancies.
6915 This flag is enabled by default at @option{-O} and higher.
6917 @item -ftree-phiprop
6918 @opindex ftree-phiprop
6919 Perform hoisting of loads from conditional pointers on trees. This
6920 pass is enabled by default at @option{-O} and higher.
6922 @item -ftree-copy-prop
6923 @opindex ftree-copy-prop
6924 Perform copy propagation on trees. This pass eliminates unnecessary
6925 copy operations. This flag is enabled by default at @option{-O} and
6928 @item -fipa-pure-const
6929 @opindex fipa-pure-const
6930 Discover which functions are pure or constant.
6931 Enabled by default at @option{-O} and higher.
6933 @item -fipa-reference
6934 @opindex fipa-reference
6935 Discover which static variables do not escape cannot escape the
6937 Enabled by default at @option{-O} and higher.
6941 Perform interprocedural pointer analysis and interprocedural modification
6942 and reference analysis. This option can cause excessive memory and
6943 compile-time usage on large compilation units. It is not enabled by
6944 default at any optimization level.
6947 @opindex fipa-profile
6948 Perform interprocedural profile propagation. The functions called only from
6949 cold functions are marked as cold. Also functions executed once (such as
6950 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6951 functions and loop less parts of functions executed once are then optimized for
6953 Enabled by default at @option{-O} and higher.
6957 Perform interprocedural constant propagation.
6958 This optimization analyzes the program to determine when values passed
6959 to functions are constants and then optimizes accordingly.
6960 This optimization can substantially increase performance
6961 if the application has constants passed to functions.
6962 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6964 @item -fipa-cp-clone
6965 @opindex fipa-cp-clone
6966 Perform function cloning to make interprocedural constant propagation stronger.
6967 When enabled, interprocedural constant propagation will perform function cloning
6968 when externally visible function can be called with constant arguments.
6969 Because this optimization can create multiple copies of functions,
6970 it may significantly increase code size
6971 (see @option{--param ipcp-unit-growth=@var{value}}).
6972 This flag is enabled by default at @option{-O3}.
6974 @item -fipa-matrix-reorg
6975 @opindex fipa-matrix-reorg
6976 Perform matrix flattening and transposing.
6977 Matrix flattening tries to replace an @math{m}-dimensional matrix
6978 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6979 This reduces the level of indirection needed for accessing the elements
6980 of the matrix. The second optimization is matrix transposing that
6981 attempts to change the order of the matrix's dimensions in order to
6982 improve cache locality.
6983 Both optimizations need the @option{-fwhole-program} flag.
6984 Transposing is enabled only if profiling information is available.
6988 Perform forward store motion on trees. This flag is
6989 enabled by default at @option{-O} and higher.
6991 @item -ftree-bit-ccp
6992 @opindex ftree-bit-ccp
6993 Perform sparse conditional bit constant propagation on trees and propagate
6994 pointer alignment information.
6995 This pass only operates on local scalar variables and is enabled by default
6996 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7000 Perform sparse conditional constant propagation (CCP) on trees. This
7001 pass only operates on local scalar variables and is enabled by default
7002 at @option{-O} and higher.
7004 @item -ftree-switch-conversion
7005 Perform conversion of simple initializations in a switch to
7006 initializations from a scalar array. This flag is enabled by default
7007 at @option{-O2} and higher.
7011 Perform dead code elimination (DCE) on trees. This flag is enabled by
7012 default at @option{-O} and higher.
7014 @item -ftree-builtin-call-dce
7015 @opindex ftree-builtin-call-dce
7016 Perform conditional dead code elimination (DCE) for calls to builtin functions
7017 that may set @code{errno} but are otherwise side-effect free. This flag is
7018 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7021 @item -ftree-dominator-opts
7022 @opindex ftree-dominator-opts
7023 Perform a variety of simple scalar cleanups (constant/copy
7024 propagation, redundancy elimination, range propagation and expression
7025 simplification) based on a dominator tree traversal. This also
7026 performs jump threading (to reduce jumps to jumps). This flag is
7027 enabled by default at @option{-O} and higher.
7031 Perform dead store elimination (DSE) on trees. A dead store is a store into
7032 a memory location which will later be overwritten by another store without
7033 any intervening loads. In this case the earlier store can be deleted. This
7034 flag is enabled by default at @option{-O} and higher.
7038 Perform loop header copying on trees. This is beneficial since it increases
7039 effectiveness of code motion optimizations. It also saves one jump. This flag
7040 is enabled by default at @option{-O} and higher. It is not enabled
7041 for @option{-Os}, since it usually increases code size.
7043 @item -ftree-loop-optimize
7044 @opindex ftree-loop-optimize
7045 Perform loop optimizations on trees. This flag is enabled by default
7046 at @option{-O} and higher.
7048 @item -ftree-loop-linear
7049 @opindex ftree-loop-linear
7050 Perform loop interchange transformations on tree. Same as
7051 @option{-floop-interchange}. To use this code transformation, GCC has
7052 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7053 enable the Graphite loop transformation infrastructure.
7055 @item -floop-interchange
7056 @opindex floop-interchange
7057 Perform loop interchange transformations on loops. Interchanging two
7058 nested loops switches the inner and outer loops. For example, given a
7063 A(J, I) = A(J, I) * C
7067 loop interchange will transform the loop as if the user had written:
7071 A(J, I) = A(J, I) * C
7075 which can be beneficial when @code{N} is larger than the caches,
7076 because in Fortran, the elements of an array are stored in memory
7077 contiguously by column, and the original loop iterates over rows,
7078 potentially creating at each access a cache miss. This optimization
7079 applies to all the languages supported by GCC and is not limited to
7080 Fortran. To use this code transformation, GCC has to be configured
7081 with @option{--with-ppl} and @option{--with-cloog} to enable the
7082 Graphite loop transformation infrastructure.
7084 @item -floop-strip-mine
7085 @opindex floop-strip-mine
7086 Perform loop strip mining transformations on loops. Strip mining
7087 splits a loop into two nested loops. The outer loop has strides
7088 equal to the strip size and the inner loop has strides of the
7089 original loop within a strip. The strip length can be changed
7090 using the @option{loop-block-tile-size} parameter. For example,
7097 loop strip mining will transform the loop as if the user had written:
7100 DO I = II, min (II + 50, N)
7105 This optimization applies to all the languages supported by GCC and is
7106 not limited to Fortran. To use this code transformation, GCC has to
7107 be configured with @option{--with-ppl} and @option{--with-cloog} to
7108 enable the Graphite loop transformation infrastructure.
7111 @opindex floop-block
7112 Perform loop blocking transformations on loops. Blocking strip mines
7113 each loop in the loop nest such that the memory accesses of the
7114 element loops fit inside caches. The strip length can be changed
7115 using the @option{loop-block-tile-size} parameter. For example, given
7120 A(J, I) = B(I) + C(J)
7124 loop blocking will transform the loop as if the user had written:
7128 DO I = II, min (II + 50, N)
7129 DO J = JJ, min (JJ + 50, M)
7130 A(J, I) = B(I) + C(J)
7136 which can be beneficial when @code{M} is larger than the caches,
7137 because the innermost loop will iterate over a smaller amount of data
7138 that can be kept in the caches. This optimization applies to all the
7139 languages supported by GCC and is not limited to Fortran. To use this
7140 code transformation, GCC has to be configured with @option{--with-ppl}
7141 and @option{--with-cloog} to enable the Graphite loop transformation
7144 @item -fgraphite-identity
7145 @opindex fgraphite-identity
7146 Enable the identity transformation for graphite. For every SCoP we generate
7147 the polyhedral representation and transform it back to gimple. Using
7148 @option{-fgraphite-identity} we can check the costs or benefits of the
7149 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7150 are also performed by the code generator CLooG, like index splitting and
7151 dead code elimination in loops.
7153 @item -floop-flatten
7154 @opindex floop-flatten
7155 Removes the loop nesting structure: transforms the loop nest into a
7156 single loop. This transformation can be useful to vectorize all the
7157 levels of the loop nest.
7159 @item -floop-parallelize-all
7160 @opindex floop-parallelize-all
7161 Use the Graphite data dependence analysis to identify loops that can
7162 be parallelized. Parallelize all the loops that can be analyzed to
7163 not contain loop carried dependences without checking that it is
7164 profitable to parallelize the loops.
7166 @item -fcheck-data-deps
7167 @opindex fcheck-data-deps
7168 Compare the results of several data dependence analyzers. This option
7169 is used for debugging the data dependence analyzers.
7171 @item -ftree-loop-if-convert
7172 Attempt to transform conditional jumps in the innermost loops to
7173 branch-less equivalents. The intent is to remove control-flow from
7174 the innermost loops in order to improve the ability of the
7175 vectorization pass to handle these loops. This is enabled by default
7176 if vectorization is enabled.
7178 @item -ftree-loop-if-convert-stores
7179 Attempt to also if-convert conditional jumps containing memory writes.
7180 This transformation can be unsafe for multi-threaded programs as it
7181 transforms conditional memory writes into unconditional memory writes.
7184 for (i = 0; i < N; i++)
7188 would be transformed to
7190 for (i = 0; i < N; i++)
7191 A[i] = cond ? expr : A[i];
7193 potentially producing data races.
7195 @item -ftree-loop-distribution
7196 Perform loop distribution. This flag can improve cache performance on
7197 big loop bodies and allow further loop optimizations, like
7198 parallelization or vectorization, to take place. For example, the loop
7215 @item -ftree-loop-distribute-patterns
7216 Perform loop distribution of patterns that can be code generated with
7217 calls to a library. This flag is enabled by default at @option{-O3}.
7219 This pass distributes the initialization loops and generates a call to
7220 memset zero. For example, the loop
7236 and the initialization loop is transformed into a call to memset zero.
7238 @item -ftree-loop-im
7239 @opindex ftree-loop-im
7240 Perform loop invariant motion on trees. This pass moves only invariants that
7241 would be hard to handle at RTL level (function calls, operations that expand to
7242 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7243 operands of conditions that are invariant out of the loop, so that we can use
7244 just trivial invariantness analysis in loop unswitching. The pass also includes
7247 @item -ftree-loop-ivcanon
7248 @opindex ftree-loop-ivcanon
7249 Create a canonical counter for number of iterations in the loop for that
7250 determining number of iterations requires complicated analysis. Later
7251 optimizations then may determine the number easily. Useful especially
7252 in connection with unrolling.
7256 Perform induction variable optimizations (strength reduction, induction
7257 variable merging and induction variable elimination) on trees.
7259 @item -ftree-parallelize-loops=n
7260 @opindex ftree-parallelize-loops
7261 Parallelize loops, i.e., split their iteration space to run in n threads.
7262 This is only possible for loops whose iterations are independent
7263 and can be arbitrarily reordered. The optimization is only
7264 profitable on multiprocessor machines, for loops that are CPU-intensive,
7265 rather than constrained e.g.@: by memory bandwidth. This option
7266 implies @option{-pthread}, and thus is only supported on targets
7267 that have support for @option{-pthread}.
7271 Perform function-local points-to analysis on trees. This flag is
7272 enabled by default at @option{-O} and higher.
7276 Perform scalar replacement of aggregates. This pass replaces structure
7277 references with scalars to prevent committing structures to memory too
7278 early. This flag is enabled by default at @option{-O} and higher.
7280 @item -ftree-copyrename
7281 @opindex ftree-copyrename
7282 Perform copy renaming on trees. This pass attempts to rename compiler
7283 temporaries to other variables at copy locations, usually resulting in
7284 variable names which more closely resemble the original variables. This flag
7285 is enabled by default at @option{-O} and higher.
7289 Perform temporary expression replacement during the SSA->normal phase. Single
7290 use/single def temporaries are replaced at their use location with their
7291 defining expression. This results in non-GIMPLE code, but gives the expanders
7292 much more complex trees to work on resulting in better RTL generation. This is
7293 enabled by default at @option{-O} and higher.
7295 @item -ftree-vectorize
7296 @opindex ftree-vectorize
7297 Perform loop vectorization on trees. This flag is enabled by default at
7300 @item -ftree-slp-vectorize
7301 @opindex ftree-slp-vectorize
7302 Perform basic block vectorization on trees. This flag is enabled by default at
7303 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7305 @item -ftree-vect-loop-version
7306 @opindex ftree-vect-loop-version
7307 Perform loop versioning when doing loop vectorization on trees. When a loop
7308 appears to be vectorizable except that data alignment or data dependence cannot
7309 be determined at compile time then vectorized and non-vectorized versions of
7310 the loop are generated along with runtime checks for alignment or dependence
7311 to control which version is executed. This option is enabled by default
7312 except at level @option{-Os} where it is disabled.
7314 @item -fvect-cost-model
7315 @opindex fvect-cost-model
7316 Enable cost model for vectorization.
7320 Perform Value Range Propagation on trees. This is similar to the
7321 constant propagation pass, but instead of values, ranges of values are
7322 propagated. This allows the optimizers to remove unnecessary range
7323 checks like array bound checks and null pointer checks. This is
7324 enabled by default at @option{-O2} and higher. Null pointer check
7325 elimination is only done if @option{-fdelete-null-pointer-checks} is
7330 Perform tail duplication to enlarge superblock size. This transformation
7331 simplifies the control flow of the function allowing other optimizations to do
7334 @item -funroll-loops
7335 @opindex funroll-loops
7336 Unroll loops whose number of iterations can be determined at compile
7337 time or upon entry to the loop. @option{-funroll-loops} implies
7338 @option{-frerun-cse-after-loop}. This option makes code larger,
7339 and may or may not make it run faster.
7341 @item -funroll-all-loops
7342 @opindex funroll-all-loops
7343 Unroll all loops, even if their number of iterations is uncertain when
7344 the loop is entered. This usually makes programs run more slowly.
7345 @option{-funroll-all-loops} implies the same options as
7346 @option{-funroll-loops},
7348 @item -fsplit-ivs-in-unroller
7349 @opindex fsplit-ivs-in-unroller
7350 Enables expressing of values of induction variables in later iterations
7351 of the unrolled loop using the value in the first iteration. This breaks
7352 long dependency chains, thus improving efficiency of the scheduling passes.
7354 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7355 same effect. However in cases the loop body is more complicated than
7356 a single basic block, this is not reliable. It also does not work at all
7357 on some of the architectures due to restrictions in the CSE pass.
7359 This optimization is enabled by default.
7361 @item -fvariable-expansion-in-unroller
7362 @opindex fvariable-expansion-in-unroller
7363 With this option, the compiler will create multiple copies of some
7364 local variables when unrolling a loop which can result in superior code.
7366 @item -fpartial-inlining
7367 @opindex fpartial-inlining
7368 Inline parts of functions. This option has any effect only
7369 when inlining itself is turned on by the @option{-finline-functions}
7370 or @option{-finline-small-functions} options.
7372 Enabled at level @option{-O2}.
7374 @item -fpredictive-commoning
7375 @opindex fpredictive-commoning
7376 Perform predictive commoning optimization, i.e., reusing computations
7377 (especially memory loads and stores) performed in previous
7378 iterations of loops.
7380 This option is enabled at level @option{-O3}.
7382 @item -fprefetch-loop-arrays
7383 @opindex fprefetch-loop-arrays
7384 If supported by the target machine, generate instructions to prefetch
7385 memory to improve the performance of loops that access large arrays.
7387 This option may generate better or worse code; results are highly
7388 dependent on the structure of loops within the source code.
7390 Disabled at level @option{-Os}.
7393 @itemx -fno-peephole2
7394 @opindex fno-peephole
7395 @opindex fno-peephole2
7396 Disable any machine-specific peephole optimizations. The difference
7397 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7398 are implemented in the compiler; some targets use one, some use the
7399 other, a few use both.
7401 @option{-fpeephole} is enabled by default.
7402 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7404 @item -fno-guess-branch-probability
7405 @opindex fno-guess-branch-probability
7406 Do not guess branch probabilities using heuristics.
7408 GCC will use heuristics to guess branch probabilities if they are
7409 not provided by profiling feedback (@option{-fprofile-arcs}). These
7410 heuristics are based on the control flow graph. If some branch probabilities
7411 are specified by @samp{__builtin_expect}, then the heuristics will be
7412 used to guess branch probabilities for the rest of the control flow graph,
7413 taking the @samp{__builtin_expect} info into account. The interactions
7414 between the heuristics and @samp{__builtin_expect} can be complex, and in
7415 some cases, it may be useful to disable the heuristics so that the effects
7416 of @samp{__builtin_expect} are easier to understand.
7418 The default is @option{-fguess-branch-probability} at levels
7419 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7421 @item -freorder-blocks
7422 @opindex freorder-blocks
7423 Reorder basic blocks in the compiled function in order to reduce number of
7424 taken branches and improve code locality.
7426 Enabled at levels @option{-O2}, @option{-O3}.
7428 @item -freorder-blocks-and-partition
7429 @opindex freorder-blocks-and-partition
7430 In addition to reordering basic blocks in the compiled function, in order
7431 to reduce number of taken branches, partitions hot and cold basic blocks
7432 into separate sections of the assembly and .o files, to improve
7433 paging and cache locality performance.
7435 This optimization is automatically turned off in the presence of
7436 exception handling, for linkonce sections, for functions with a user-defined
7437 section attribute and on any architecture that does not support named
7440 @item -freorder-functions
7441 @opindex freorder-functions
7442 Reorder functions in the object file in order to
7443 improve code locality. This is implemented by using special
7444 subsections @code{.text.hot} for most frequently executed functions and
7445 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7446 the linker so object file format must support named sections and linker must
7447 place them in a reasonable way.
7449 Also profile feedback must be available in to make this option effective. See
7450 @option{-fprofile-arcs} for details.
7452 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7454 @item -fstrict-aliasing
7455 @opindex fstrict-aliasing
7456 Allow the compiler to assume the strictest aliasing rules applicable to
7457 the language being compiled. For C (and C++), this activates
7458 optimizations based on the type of expressions. In particular, an
7459 object of one type is assumed never to reside at the same address as an
7460 object of a different type, unless the types are almost the same. For
7461 example, an @code{unsigned int} can alias an @code{int}, but not a
7462 @code{void*} or a @code{double}. A character type may alias any other
7465 @anchor{Type-punning}Pay special attention to code like this:
7478 The practice of reading from a different union member than the one most
7479 recently written to (called ``type-punning'') is common. Even with
7480 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7481 is accessed through the union type. So, the code above will work as
7482 expected. @xref{Structures unions enumerations and bit-fields
7483 implementation}. However, this code might not:
7494 Similarly, access by taking the address, casting the resulting pointer
7495 and dereferencing the result has undefined behavior, even if the cast
7496 uses a union type, e.g.:
7500 return ((union a_union *) &d)->i;
7504 The @option{-fstrict-aliasing} option is enabled at levels
7505 @option{-O2}, @option{-O3}, @option{-Os}.
7507 @item -fstrict-overflow
7508 @opindex fstrict-overflow
7509 Allow the compiler to assume strict signed overflow rules, depending
7510 on the language being compiled. For C (and C++) this means that
7511 overflow when doing arithmetic with signed numbers is undefined, which
7512 means that the compiler may assume that it will not happen. This
7513 permits various optimizations. For example, the compiler will assume
7514 that an expression like @code{i + 10 > i} will always be true for
7515 signed @code{i}. This assumption is only valid if signed overflow is
7516 undefined, as the expression is false if @code{i + 10} overflows when
7517 using twos complement arithmetic. When this option is in effect any
7518 attempt to determine whether an operation on signed numbers will
7519 overflow must be written carefully to not actually involve overflow.
7521 This option also allows the compiler to assume strict pointer
7522 semantics: given a pointer to an object, if adding an offset to that
7523 pointer does not produce a pointer to the same object, the addition is
7524 undefined. This permits the compiler to conclude that @code{p + u >
7525 p} is always true for a pointer @code{p} and unsigned integer
7526 @code{u}. This assumption is only valid because pointer wraparound is
7527 undefined, as the expression is false if @code{p + u} overflows using
7528 twos complement arithmetic.
7530 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7531 that integer signed overflow is fully defined: it wraps. When
7532 @option{-fwrapv} is used, there is no difference between
7533 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7534 integers. With @option{-fwrapv} certain types of overflow are
7535 permitted. For example, if the compiler gets an overflow when doing
7536 arithmetic on constants, the overflowed value can still be used with
7537 @option{-fwrapv}, but not otherwise.
7539 The @option{-fstrict-overflow} option is enabled at levels
7540 @option{-O2}, @option{-O3}, @option{-Os}.
7542 @item -falign-functions
7543 @itemx -falign-functions=@var{n}
7544 @opindex falign-functions
7545 Align the start of functions to the next power-of-two greater than
7546 @var{n}, skipping up to @var{n} bytes. For instance,
7547 @option{-falign-functions=32} aligns functions to the next 32-byte
7548 boundary, but @option{-falign-functions=24} would align to the next
7549 32-byte boundary only if this can be done by skipping 23 bytes or less.
7551 @option{-fno-align-functions} and @option{-falign-functions=1} are
7552 equivalent and mean that functions will not be aligned.
7554 Some assemblers only support this flag when @var{n} is a power of two;
7555 in that case, it is rounded up.
7557 If @var{n} is not specified or is zero, use a machine-dependent default.
7559 Enabled at levels @option{-O2}, @option{-O3}.
7561 @item -falign-labels
7562 @itemx -falign-labels=@var{n}
7563 @opindex falign-labels
7564 Align all branch targets to a power-of-two boundary, skipping up to
7565 @var{n} bytes like @option{-falign-functions}. This option can easily
7566 make code slower, because it must insert dummy operations for when the
7567 branch target is reached in the usual flow of the code.
7569 @option{-fno-align-labels} and @option{-falign-labels=1} are
7570 equivalent and mean that labels will not be aligned.
7572 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7573 are greater than this value, then their values are used instead.
7575 If @var{n} is not specified or is zero, use a machine-dependent default
7576 which is very likely to be @samp{1}, meaning no alignment.
7578 Enabled at levels @option{-O2}, @option{-O3}.
7581 @itemx -falign-loops=@var{n}
7582 @opindex falign-loops
7583 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7584 like @option{-falign-functions}. The hope is that the loop will be
7585 executed many times, which will make up for any execution of the dummy
7588 @option{-fno-align-loops} and @option{-falign-loops=1} are
7589 equivalent and mean that loops will not be aligned.
7591 If @var{n} is not specified or is zero, use a machine-dependent default.
7593 Enabled at levels @option{-O2}, @option{-O3}.
7596 @itemx -falign-jumps=@var{n}
7597 @opindex falign-jumps
7598 Align branch targets to a power-of-two boundary, for branch targets
7599 where the targets can only be reached by jumping, skipping up to @var{n}
7600 bytes like @option{-falign-functions}. In this case, no dummy operations
7603 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7604 equivalent and mean that loops will not be aligned.
7606 If @var{n} is not specified or is zero, use a machine-dependent default.
7608 Enabled at levels @option{-O2}, @option{-O3}.
7610 @item -funit-at-a-time
7611 @opindex funit-at-a-time
7612 This option is left for compatibility reasons. @option{-funit-at-a-time}
7613 has no effect, while @option{-fno-unit-at-a-time} implies
7614 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7618 @item -fno-toplevel-reorder
7619 @opindex fno-toplevel-reorder
7620 Do not reorder top-level functions, variables, and @code{asm}
7621 statements. Output them in the same order that they appear in the
7622 input file. When this option is used, unreferenced static variables
7623 will not be removed. This option is intended to support existing code
7624 which relies on a particular ordering. For new code, it is better to
7627 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7628 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7633 Constructs webs as commonly used for register allocation purposes and assign
7634 each web individual pseudo register. This allows the register allocation pass
7635 to operate on pseudos directly, but also strengthens several other optimization
7636 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7637 however, make debugging impossible, since variables will no longer stay in a
7640 Enabled by default with @option{-funroll-loops}.
7642 @item -fwhole-program
7643 @opindex fwhole-program
7644 Assume that the current compilation unit represents the whole program being
7645 compiled. All public functions and variables with the exception of @code{main}
7646 and those merged by attribute @code{externally_visible} become static functions
7647 and in effect are optimized more aggressively by interprocedural optimizers. If @command{gold} is used as the linker plugin, @code{externally_visible} attributes are automatically added to functions (not variable yet due to a current @command{gold} issue) that are accessed outside of LTO objects according to resolution file produced by @command{gold}. For other linkers that cannot generate resolution file, explicit @code{externally_visible} attributes are still necessary.
7648 While this option is equivalent to proper use of the @code{static} keyword for
7649 programs consisting of a single file, in combination with option
7650 @option{-flto} this flag can be used to
7651 compile many smaller scale programs since the functions and variables become
7652 local for the whole combined compilation unit, not for the single source file
7655 This option implies @option{-fwhole-file} for Fortran programs.
7657 @item -flto[=@var{n}]
7659 This option runs the standard link-time optimizer. When invoked
7660 with source code, it generates GIMPLE (one of GCC's internal
7661 representations) and writes it to special ELF sections in the object
7662 file. When the object files are linked together, all the function
7663 bodies are read from these ELF sections and instantiated as if they
7664 had been part of the same translation unit.
7666 To use the link-timer optimizer, @option{-flto} needs to be specified at
7667 compile time and during the final link. For example,
7670 gcc -c -O2 -flto foo.c
7671 gcc -c -O2 -flto bar.c
7672 gcc -o myprog -flto -O2 foo.o bar.o
7675 The first two invocations to GCC will save a bytecode representation
7676 of GIMPLE into special ELF sections inside @file{foo.o} and
7677 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7678 @file{foo.o} and @file{bar.o}, merge the two files into a single
7679 internal image, and compile the result as usual. Since both
7680 @file{foo.o} and @file{bar.o} are merged into a single image, this
7681 causes all the inter-procedural analyses and optimizations in GCC to
7682 work across the two files as if they were a single one. This means,
7683 for example, that the inliner will be able to inline functions in
7684 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7686 Another (simpler) way to enable link-time optimization is,
7689 gcc -o myprog -flto -O2 foo.c bar.c
7692 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7693 merge them together into a single GIMPLE representation and optimize
7694 them as usual to produce @file{myprog}.
7696 The only important thing to keep in mind is that to enable link-time
7697 optimizations the @option{-flto} flag needs to be passed to both the
7698 compile and the link commands.
7700 To make whole program optimization effective, it is necessary to make
7701 certain whole program assumptions. The compiler needs to know
7702 what functions and variables can be accessed by libraries and runtime
7703 outside of the link time optimized unit. When supported by the linker,
7704 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7705 compiler information about used and externally visible symbols. When
7706 the linker plugin is not available, @option{-fwhole-program} should be
7707 used to allow the compiler to make these assumptions, which will lead
7708 to more aggressive optimization decisions.
7710 Note that when a file is compiled with @option{-flto}, the generated
7711 object file will be larger than a regular object file because it will
7712 contain GIMPLE bytecodes and the usual final code. This means that
7713 object files with LTO information can be linked as a normal object
7714 file. So, in the previous example, if the final link is done with
7717 gcc -o myprog foo.o bar.o
7720 The only difference will be that no inter-procedural optimizations
7721 will be applied to produce @file{myprog}. The two object files
7722 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7725 Additionally, the optimization flags used to compile individual files
7726 are not necessarily related to those used at link-time. For instance,
7729 gcc -c -O0 -flto foo.c
7730 gcc -c -O0 -flto bar.c
7731 gcc -o myprog -flto -O3 foo.o bar.o
7734 This will produce individual object files with unoptimized assembler
7735 code, but the resulting binary @file{myprog} will be optimized at
7736 @option{-O3}. Now, if the final binary is generated without
7737 @option{-flto}, then @file{myprog} will not be optimized.
7739 When producing the final binary with @option{-flto}, GCC will only
7740 apply link-time optimizations to those files that contain bytecode.
7741 Therefore, you can mix and match object files and libraries with
7742 GIMPLE bytecodes and final object code. GCC will automatically select
7743 which files to optimize in LTO mode and which files to link without
7746 There are some code generation flags that GCC will preserve when
7747 generating bytecodes, as they need to be used during the final link
7748 stage. Currently, the following options are saved into the GIMPLE
7749 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7750 @option{-m} target flags.
7752 At link time, these options are read-in and reapplied. Note that the
7753 current implementation makes no attempt at recognizing conflicting
7754 values for these options. If two or more files have a conflicting
7755 value (e.g., one file is compiled with @option{-fPIC} and another
7756 isn't), the compiler will simply use the last value read from the
7757 bytecode files. It is recommended, then, that all the files
7758 participating in the same link be compiled with the same options.
7760 Another feature of LTO is that it is possible to apply interprocedural
7761 optimizations on files written in different languages. This requires
7762 some support in the language front end. Currently, the C, C++ and
7763 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7764 something like this should work
7769 gfortran -c -flto baz.f90
7770 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7773 Notice that the final link is done with @command{g++} to get the C++
7774 runtime libraries and @option{-lgfortran} is added to get the Fortran
7775 runtime libraries. In general, when mixing languages in LTO mode, you
7776 should use the same link command used when mixing languages in a
7777 regular (non-LTO) compilation. This means that if your build process
7778 was mixing languages before, all you need to add is @option{-flto} to
7779 all the compile and link commands.
7781 If LTO encounters objects with C linkage declared with incompatible
7782 types in separate translation units to be linked together (undefined
7783 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7784 issued. The behavior is still undefined at runtime.
7786 If object files containing GIMPLE bytecode are stored in a library archive, say
7787 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7788 are using a linker with linker plugin support. To enable this feature, use
7789 the flag @option{-fuse-linker-plugin} at link-time:
7792 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7795 With the linker plugin enabled, the linker will extract the needed
7796 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7797 to make them part of the aggregated GIMPLE image to be optimized.
7799 If you are not using a linker with linker plugin support and/or do not
7800 enable linker plugin then the objects inside @file{libfoo.a}
7801 will be extracted and linked as usual, but they will not participate
7802 in the LTO optimization process.
7804 Link time optimizations do not require the presence of the whole program to
7805 operate. If the program does not require any symbols to be exported, it is
7806 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7807 the interprocedural optimizers to use more aggressive assumptions which may
7808 lead to improved optimization opportunities.
7809 Use of @option{-fwhole-program} is not needed when linker plugin is
7810 active (see @option{-fuse-linker-plugin}).
7812 Regarding portability: the current implementation of LTO makes no
7813 attempt at generating bytecode that can be ported between different
7814 types of hosts. The bytecode files are versioned and there is a
7815 strict version check, so bytecode files generated in one version of
7816 GCC will not work with an older/newer version of GCC.
7818 Link time optimization does not play well with generating debugging
7819 information. Combining @option{-flto} with
7820 @option{-g} is currently experimental and expected to produce wrong
7823 If you specify the optional @var{n}, the optimization and code
7824 generation done at link time is executed in parallel using @var{n}
7825 parallel jobs by utilizing an installed @command{make} program. The
7826 environment variable @env{MAKE} may be used to override the program
7827 used. The default value for @var{n} is 1.
7829 You can also specify @option{-flto=jobserver} to use GNU make's
7830 job server mode to determine the number of parallel jobs. This
7831 is useful when the Makefile calling GCC is already executing in parallel.
7832 The parent Makefile will need a @samp{+} prepended to the command recipe
7833 for this to work. This will likely only work if @env{MAKE} is
7836 This option is disabled by default.
7838 @item -flto-partition=@var{alg}
7839 @opindex flto-partition
7840 Specify the partitioning algorithm used by the link time optimizer.
7841 The value is either @code{1to1} to specify a partitioning mirroring
7842 the original source files or @code{balanced} to specify partitioning
7843 into equally sized chunks (whenever possible). Specifying @code{none}
7844 as an algorithm disables partitioning and streaming completely. The
7845 default value is @code{balanced}.
7847 @item -flto-compression-level=@var{n}
7848 This option specifies the level of compression used for intermediate
7849 language written to LTO object files, and is only meaningful in
7850 conjunction with LTO mode (@option{-flto}). Valid
7851 values are 0 (no compression) to 9 (maximum compression). Values
7852 outside this range are clamped to either 0 or 9. If the option is not
7853 given, a default balanced compression setting is used.
7856 Prints a report with internal details on the workings of the link-time
7857 optimizer. The contents of this report vary from version to version,
7858 it is meant to be useful to GCC developers when processing object
7859 files in LTO mode (via @option{-flto}).
7861 Disabled by default.
7863 @item -fuse-linker-plugin
7864 Enables the use of a linker plugin during link time optimization. This
7865 option relies on plugin support in the linker, which is available in gold
7866 or in GNU ld 2.21 or newer.
7868 This option enables the extraction of object files with GIMPLE bytecode out
7869 of library archives. This improves the quality of optimization by exposing
7870 more code to the link time optimizer. This information specifies what
7871 symbols can be accessed externally (by non-LTO object or during dynamic
7872 linking). Resulting code quality improvements on binaries (and shared
7873 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7874 See @option{-flto} for a description of the effect of this flag and how to
7877 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7878 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7880 @item -fcompare-elim
7881 @opindex fcompare-elim
7882 After register allocation and post-register allocation instruction splitting,
7883 identify arithmetic instructions that compute processor flags similar to a
7884 comparison operation based on that arithmetic. If possible, eliminate the
7885 explicit comparison operation.
7887 This pass only applies to certain targets that cannot explicitly represent
7888 the comparison operation before register allocation is complete.
7890 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7892 @item -fcprop-registers
7893 @opindex fcprop-registers
7894 After register allocation and post-register allocation instruction splitting,
7895 we perform a copy-propagation pass to try to reduce scheduling dependencies
7896 and occasionally eliminate the copy.
7898 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7900 @item -fprofile-correction
7901 @opindex fprofile-correction
7902 Profiles collected using an instrumented binary for multi-threaded programs may
7903 be inconsistent due to missed counter updates. When this option is specified,
7904 GCC will use heuristics to correct or smooth out such inconsistencies. By
7905 default, GCC will emit an error message when an inconsistent profile is detected.
7907 @item -fprofile-dir=@var{path}
7908 @opindex fprofile-dir
7910 Set the directory to search for the profile data files in to @var{path}.
7911 This option affects only the profile data generated by
7912 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7913 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7914 and its related options. Both absolute and relative paths can be used.
7915 By default, GCC will use the current directory as @var{path}, thus the
7916 profile data file will appear in the same directory as the object file.
7918 @item -fprofile-generate
7919 @itemx -fprofile-generate=@var{path}
7920 @opindex fprofile-generate
7922 Enable options usually used for instrumenting application to produce
7923 profile useful for later recompilation with profile feedback based
7924 optimization. You must use @option{-fprofile-generate} both when
7925 compiling and when linking your program.
7927 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7929 If @var{path} is specified, GCC will look at the @var{path} to find
7930 the profile feedback data files. See @option{-fprofile-dir}.
7933 @itemx -fprofile-use=@var{path}
7934 @opindex fprofile-use
7935 Enable profile feedback directed optimizations, and optimizations
7936 generally profitable only with profile feedback available.
7938 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7939 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7941 By default, GCC emits an error message if the feedback profiles do not
7942 match the source code. This error can be turned into a warning by using
7943 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7946 If @var{path} is specified, GCC will look at the @var{path} to find
7947 the profile feedback data files. See @option{-fprofile-dir}.
7950 The following options control compiler behavior regarding floating
7951 point arithmetic. These options trade off between speed and
7952 correctness. All must be specifically enabled.
7956 @opindex ffloat-store
7957 Do not store floating point variables in registers, and inhibit other
7958 options that might change whether a floating point value is taken from a
7961 @cindex floating point precision
7962 This option prevents undesirable excess precision on machines such as
7963 the 68000 where the floating registers (of the 68881) keep more
7964 precision than a @code{double} is supposed to have. Similarly for the
7965 x86 architecture. For most programs, the excess precision does only
7966 good, but a few programs rely on the precise definition of IEEE floating
7967 point. Use @option{-ffloat-store} for such programs, after modifying
7968 them to store all pertinent intermediate computations into variables.
7970 @item -fexcess-precision=@var{style}
7971 @opindex fexcess-precision
7972 This option allows further control over excess precision on machines
7973 where floating-point registers have more precision than the IEEE
7974 @code{float} and @code{double} types and the processor does not
7975 support operations rounding to those types. By default,
7976 @option{-fexcess-precision=fast} is in effect; this means that
7977 operations are carried out in the precision of the registers and that
7978 it is unpredictable when rounding to the types specified in the source
7979 code takes place. When compiling C, if
7980 @option{-fexcess-precision=standard} is specified then excess
7981 precision will follow the rules specified in ISO C99; in particular,
7982 both casts and assignments cause values to be rounded to their
7983 semantic types (whereas @option{-ffloat-store} only affects
7984 assignments). This option is enabled by default for C if a strict
7985 conformance option such as @option{-std=c99} is used.
7988 @option{-fexcess-precision=standard} is not implemented for languages
7989 other than C, and has no effect if
7990 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7991 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7992 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7993 semantics apply without excess precision, and in the latter, rounding
7998 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7999 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8000 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8002 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8004 This option is not turned on by any @option{-O} option besides
8005 @option{-Ofast} since it can result in incorrect output for programs
8006 which depend on an exact implementation of IEEE or ISO rules/specifications
8007 for math functions. It may, however, yield faster code for programs
8008 that do not require the guarantees of these specifications.
8010 @item -fno-math-errno
8011 @opindex fno-math-errno
8012 Do not set ERRNO after calling math functions that are executed
8013 with a single instruction, e.g., sqrt. A program that relies on
8014 IEEE exceptions for math error handling may want to use this flag
8015 for speed while maintaining IEEE arithmetic compatibility.
8017 This option is not turned on by any @option{-O} option since
8018 it can result in incorrect output for programs which depend on
8019 an exact implementation of IEEE or ISO rules/specifications for
8020 math functions. It may, however, yield faster code for programs
8021 that do not require the guarantees of these specifications.
8023 The default is @option{-fmath-errno}.
8025 On Darwin systems, the math library never sets @code{errno}. There is
8026 therefore no reason for the compiler to consider the possibility that
8027 it might, and @option{-fno-math-errno} is the default.
8029 @item -funsafe-math-optimizations
8030 @opindex funsafe-math-optimizations
8032 Allow optimizations for floating-point arithmetic that (a) assume
8033 that arguments and results are valid and (b) may violate IEEE or
8034 ANSI standards. When used at link-time, it may include libraries
8035 or startup files that change the default FPU control word or other
8036 similar optimizations.
8038 This option is not turned on by any @option{-O} option since
8039 it can result in incorrect output for programs which depend on
8040 an exact implementation of IEEE or ISO rules/specifications for
8041 math functions. It may, however, yield faster code for programs
8042 that do not require the guarantees of these specifications.
8043 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8044 @option{-fassociative-math} and @option{-freciprocal-math}.
8046 The default is @option{-fno-unsafe-math-optimizations}.
8048 @item -fassociative-math
8049 @opindex fassociative-math
8051 Allow re-association of operands in series of floating-point operations.
8052 This violates the ISO C and C++ language standard by possibly changing
8053 computation result. NOTE: re-ordering may change the sign of zero as
8054 well as ignore NaNs and inhibit or create underflow or overflow (and
8055 thus cannot be used on a code which relies on rounding behavior like
8056 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
8057 and thus may not be used when ordered comparisons are required.
8058 This option requires that both @option{-fno-signed-zeros} and
8059 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8060 much sense with @option{-frounding-math}. For Fortran the option
8061 is automatically enabled when both @option{-fno-signed-zeros} and
8062 @option{-fno-trapping-math} are in effect.
8064 The default is @option{-fno-associative-math}.
8066 @item -freciprocal-math
8067 @opindex freciprocal-math
8069 Allow the reciprocal of a value to be used instead of dividing by
8070 the value if this enables optimizations. For example @code{x / y}
8071 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8072 is subject to common subexpression elimination. Note that this loses
8073 precision and increases the number of flops operating on the value.
8075 The default is @option{-fno-reciprocal-math}.
8077 @item -ffinite-math-only
8078 @opindex ffinite-math-only
8079 Allow optimizations for floating-point arithmetic that assume
8080 that arguments and results are not NaNs or +-Infs.
8082 This option is not turned on by any @option{-O} option since
8083 it can result in incorrect output for programs which depend on
8084 an exact implementation of IEEE or ISO rules/specifications for
8085 math functions. It may, however, yield faster code for programs
8086 that do not require the guarantees of these specifications.
8088 The default is @option{-fno-finite-math-only}.
8090 @item -fno-signed-zeros
8091 @opindex fno-signed-zeros
8092 Allow optimizations for floating point arithmetic that ignore the
8093 signedness of zero. IEEE arithmetic specifies the behavior of
8094 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8095 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8096 This option implies that the sign of a zero result isn't significant.
8098 The default is @option{-fsigned-zeros}.
8100 @item -fno-trapping-math
8101 @opindex fno-trapping-math
8102 Compile code assuming that floating-point operations cannot generate
8103 user-visible traps. These traps include division by zero, overflow,
8104 underflow, inexact result and invalid operation. This option requires
8105 that @option{-fno-signaling-nans} be in effect. Setting this option may
8106 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8108 This option should never be turned on by any @option{-O} option since
8109 it can result in incorrect output for programs which depend on
8110 an exact implementation of IEEE or ISO rules/specifications for
8113 The default is @option{-ftrapping-math}.
8115 @item -frounding-math
8116 @opindex frounding-math
8117 Disable transformations and optimizations that assume default floating
8118 point rounding behavior. This is round-to-zero for all floating point
8119 to integer conversions, and round-to-nearest for all other arithmetic
8120 truncations. This option should be specified for programs that change
8121 the FP rounding mode dynamically, or that may be executed with a
8122 non-default rounding mode. This option disables constant folding of
8123 floating point expressions at compile-time (which may be affected by
8124 rounding mode) and arithmetic transformations that are unsafe in the
8125 presence of sign-dependent rounding modes.
8127 The default is @option{-fno-rounding-math}.
8129 This option is experimental and does not currently guarantee to
8130 disable all GCC optimizations that are affected by rounding mode.
8131 Future versions of GCC may provide finer control of this setting
8132 using C99's @code{FENV_ACCESS} pragma. This command line option
8133 will be used to specify the default state for @code{FENV_ACCESS}.
8135 @item -fsignaling-nans
8136 @opindex fsignaling-nans
8137 Compile code assuming that IEEE signaling NaNs may generate user-visible
8138 traps during floating-point operations. Setting this option disables
8139 optimizations that may change the number of exceptions visible with
8140 signaling NaNs. This option implies @option{-ftrapping-math}.
8142 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8145 The default is @option{-fno-signaling-nans}.
8147 This option is experimental and does not currently guarantee to
8148 disable all GCC optimizations that affect signaling NaN behavior.
8150 @item -fsingle-precision-constant
8151 @opindex fsingle-precision-constant
8152 Treat floating point constant as single precision constant instead of
8153 implicitly converting it to double precision constant.
8155 @item -fcx-limited-range
8156 @opindex fcx-limited-range
8157 When enabled, this option states that a range reduction step is not
8158 needed when performing complex division. Also, there is no checking
8159 whether the result of a complex multiplication or division is @code{NaN
8160 + I*NaN}, with an attempt to rescue the situation in that case. The
8161 default is @option{-fno-cx-limited-range}, but is enabled by
8162 @option{-ffast-math}.
8164 This option controls the default setting of the ISO C99
8165 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8168 @item -fcx-fortran-rules
8169 @opindex fcx-fortran-rules
8170 Complex multiplication and division follow Fortran rules. Range
8171 reduction is done as part of complex division, but there is no checking
8172 whether the result of a complex multiplication or division is @code{NaN
8173 + I*NaN}, with an attempt to rescue the situation in that case.
8175 The default is @option{-fno-cx-fortran-rules}.
8179 The following options control optimizations that may improve
8180 performance, but are not enabled by any @option{-O} options. This
8181 section includes experimental options that may produce broken code.
8184 @item -fbranch-probabilities
8185 @opindex fbranch-probabilities
8186 After running a program compiled with @option{-fprofile-arcs}
8187 (@pxref{Debugging Options,, Options for Debugging Your Program or
8188 @command{gcc}}), you can compile it a second time using
8189 @option{-fbranch-probabilities}, to improve optimizations based on
8190 the number of times each branch was taken. When the program
8191 compiled with @option{-fprofile-arcs} exits it saves arc execution
8192 counts to a file called @file{@var{sourcename}.gcda} for each source
8193 file. The information in this data file is very dependent on the
8194 structure of the generated code, so you must use the same source code
8195 and the same optimization options for both compilations.
8197 With @option{-fbranch-probabilities}, GCC puts a
8198 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8199 These can be used to improve optimization. Currently, they are only
8200 used in one place: in @file{reorg.c}, instead of guessing which path a
8201 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8202 exactly determine which path is taken more often.
8204 @item -fprofile-values
8205 @opindex fprofile-values
8206 If combined with @option{-fprofile-arcs}, it adds code so that some
8207 data about values of expressions in the program is gathered.
8209 With @option{-fbranch-probabilities}, it reads back the data gathered
8210 from profiling values of expressions for usage in optimizations.
8212 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8216 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8217 a code to gather information about values of expressions.
8219 With @option{-fbranch-probabilities}, it reads back the data gathered
8220 and actually performs the optimizations based on them.
8221 Currently the optimizations include specialization of division operation
8222 using the knowledge about the value of the denominator.
8224 @item -frename-registers
8225 @opindex frename-registers
8226 Attempt to avoid false dependencies in scheduled code by making use
8227 of registers left over after register allocation. This optimization
8228 will most benefit processors with lots of registers. Depending on the
8229 debug information format adopted by the target, however, it can
8230 make debugging impossible, since variables will no longer stay in
8231 a ``home register''.
8233 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8237 Perform tail duplication to enlarge superblock size. This transformation
8238 simplifies the control flow of the function allowing other optimizations to do
8241 Enabled with @option{-fprofile-use}.
8243 @item -funroll-loops
8244 @opindex funroll-loops
8245 Unroll loops whose number of iterations can be determined at compile time or
8246 upon entry to the loop. @option{-funroll-loops} implies
8247 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8248 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8249 small constant number of iterations). This option makes code larger, and may
8250 or may not make it run faster.
8252 Enabled with @option{-fprofile-use}.
8254 @item -funroll-all-loops
8255 @opindex funroll-all-loops
8256 Unroll all loops, even if their number of iterations is uncertain when
8257 the loop is entered. This usually makes programs run more slowly.
8258 @option{-funroll-all-loops} implies the same options as
8259 @option{-funroll-loops}.
8262 @opindex fpeel-loops
8263 Peels the loops for that there is enough information that they do not
8264 roll much (from profile feedback). It also turns on complete loop peeling
8265 (i.e.@: complete removal of loops with small constant number of iterations).
8267 Enabled with @option{-fprofile-use}.
8269 @item -fmove-loop-invariants
8270 @opindex fmove-loop-invariants
8271 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8272 at level @option{-O1}
8274 @item -funswitch-loops
8275 @opindex funswitch-loops
8276 Move branches with loop invariant conditions out of the loop, with duplicates
8277 of the loop on both branches (modified according to result of the condition).
8279 @item -ffunction-sections
8280 @itemx -fdata-sections
8281 @opindex ffunction-sections
8282 @opindex fdata-sections
8283 Place each function or data item into its own section in the output
8284 file if the target supports arbitrary sections. The name of the
8285 function or the name of the data item determines the section's name
8288 Use these options on systems where the linker can perform optimizations
8289 to improve locality of reference in the instruction space. Most systems
8290 using the ELF object format and SPARC processors running Solaris 2 have
8291 linkers with such optimizations. AIX may have these optimizations in
8294 Only use these options when there are significant benefits from doing
8295 so. When you specify these options, the assembler and linker will
8296 create larger object and executable files and will also be slower.
8297 You will not be able to use @code{gprof} on all systems if you
8298 specify this option and you may have problems with debugging if
8299 you specify both this option and @option{-g}.
8301 @item -fbranch-target-load-optimize
8302 @opindex fbranch-target-load-optimize
8303 Perform branch target register load optimization before prologue / epilogue
8305 The use of target registers can typically be exposed only during reload,
8306 thus hoisting loads out of loops and doing inter-block scheduling needs
8307 a separate optimization pass.
8309 @item -fbranch-target-load-optimize2
8310 @opindex fbranch-target-load-optimize2
8311 Perform branch target register load optimization after prologue / epilogue
8314 @item -fbtr-bb-exclusive
8315 @opindex fbtr-bb-exclusive
8316 When performing branch target register load optimization, don't reuse
8317 branch target registers in within any basic block.
8319 @item -fstack-protector
8320 @opindex fstack-protector
8321 Emit extra code to check for buffer overflows, such as stack smashing
8322 attacks. This is done by adding a guard variable to functions with
8323 vulnerable objects. This includes functions that call alloca, and
8324 functions with buffers larger than 8 bytes. The guards are initialized
8325 when a function is entered and then checked when the function exits.
8326 If a guard check fails, an error message is printed and the program exits.
8328 @item -fstack-protector-all
8329 @opindex fstack-protector-all
8330 Like @option{-fstack-protector} except that all functions are protected.
8332 @item -fsection-anchors
8333 @opindex fsection-anchors
8334 Try to reduce the number of symbolic address calculations by using
8335 shared ``anchor'' symbols to address nearby objects. This transformation
8336 can help to reduce the number of GOT entries and GOT accesses on some
8339 For example, the implementation of the following function @code{foo}:
8343 int foo (void) @{ return a + b + c; @}
8346 would usually calculate the addresses of all three variables, but if you
8347 compile it with @option{-fsection-anchors}, it will access the variables
8348 from a common anchor point instead. The effect is similar to the
8349 following pseudocode (which isn't valid C):
8354 register int *xr = &x;
8355 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8359 Not all targets support this option.
8361 @item --param @var{name}=@var{value}
8363 In some places, GCC uses various constants to control the amount of
8364 optimization that is done. For example, GCC will not inline functions
8365 that contain more that a certain number of instructions. You can
8366 control some of these constants on the command-line using the
8367 @option{--param} option.
8369 The names of specific parameters, and the meaning of the values, are
8370 tied to the internals of the compiler, and are subject to change
8371 without notice in future releases.
8373 In each case, the @var{value} is an integer. The allowable choices for
8374 @var{name} are given in the following table:
8377 @item predictable-branch-outcome
8378 When branch is predicted to be taken with probability lower than this threshold
8379 (in percent), then it is considered well predictable. The default is 10.
8381 @item max-crossjump-edges
8382 The maximum number of incoming edges to consider for crossjumping.
8383 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8384 the number of edges incoming to each block. Increasing values mean
8385 more aggressive optimization, making the compile time increase with
8386 probably small improvement in executable size.
8388 @item min-crossjump-insns
8389 The minimum number of instructions which must be matched at the end
8390 of two blocks before crossjumping will be performed on them. This
8391 value is ignored in the case where all instructions in the block being
8392 crossjumped from are matched. The default value is 5.
8394 @item max-grow-copy-bb-insns
8395 The maximum code size expansion factor when copying basic blocks
8396 instead of jumping. The expansion is relative to a jump instruction.
8397 The default value is 8.
8399 @item max-goto-duplication-insns
8400 The maximum number of instructions to duplicate to a block that jumps
8401 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8402 passes, GCC factors computed gotos early in the compilation process,
8403 and unfactors them as late as possible. Only computed jumps at the
8404 end of a basic blocks with no more than max-goto-duplication-insns are
8405 unfactored. The default value is 8.
8407 @item max-delay-slot-insn-search
8408 The maximum number of instructions to consider when looking for an
8409 instruction to fill a delay slot. If more than this arbitrary number of
8410 instructions is searched, the time savings from filling the delay slot
8411 will be minimal so stop searching. Increasing values mean more
8412 aggressive optimization, making the compile time increase with probably
8413 small improvement in executable run time.
8415 @item max-delay-slot-live-search
8416 When trying to fill delay slots, the maximum number of instructions to
8417 consider when searching for a block with valid live register
8418 information. Increasing this arbitrarily chosen value means more
8419 aggressive optimization, increasing the compile time. This parameter
8420 should be removed when the delay slot code is rewritten to maintain the
8423 @item max-gcse-memory
8424 The approximate maximum amount of memory that will be allocated in
8425 order to perform the global common subexpression elimination
8426 optimization. If more memory than specified is required, the
8427 optimization will not be done.
8429 @item max-gcse-insertion-ratio
8430 If the ratio of expression insertions to deletions is larger than this value
8431 for any expression, then RTL PRE will insert or remove the expression and thus
8432 leave partially redundant computations in the instruction stream. The default value is 20.
8434 @item max-pending-list-length
8435 The maximum number of pending dependencies scheduling will allow
8436 before flushing the current state and starting over. Large functions
8437 with few branches or calls can create excessively large lists which
8438 needlessly consume memory and resources.
8440 @item max-inline-insns-single
8441 Several parameters control the tree inliner used in gcc.
8442 This number sets the maximum number of instructions (counted in GCC's
8443 internal representation) in a single function that the tree inliner
8444 will consider for inlining. This only affects functions declared
8445 inline and methods implemented in a class declaration (C++).
8446 The default value is 400.
8448 @item max-inline-insns-auto
8449 When you use @option{-finline-functions} (included in @option{-O3}),
8450 a lot of functions that would otherwise not be considered for inlining
8451 by the compiler will be investigated. To those functions, a different
8452 (more restrictive) limit compared to functions declared inline can
8454 The default value is 40.
8456 @item large-function-insns
8457 The limit specifying really large functions. For functions larger than this
8458 limit after inlining, inlining is constrained by
8459 @option{--param large-function-growth}. This parameter is useful primarily
8460 to avoid extreme compilation time caused by non-linear algorithms used by the
8462 The default value is 2700.
8464 @item large-function-growth
8465 Specifies maximal growth of large function caused by inlining in percents.
8466 The default value is 100 which limits large function growth to 2.0 times
8469 @item large-unit-insns
8470 The limit specifying large translation unit. Growth caused by inlining of
8471 units larger than this limit is limited by @option{--param inline-unit-growth}.
8472 For small units this might be too tight (consider unit consisting of function A
8473 that is inline and B that just calls A three time. If B is small relative to
8474 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8475 large units consisting of small inlineable functions however the overall unit
8476 growth limit is needed to avoid exponential explosion of code size. Thus for
8477 smaller units, the size is increased to @option{--param large-unit-insns}
8478 before applying @option{--param inline-unit-growth}. The default is 10000
8480 @item inline-unit-growth
8481 Specifies maximal overall growth of the compilation unit caused by inlining.
8482 The default value is 30 which limits unit growth to 1.3 times the original
8485 @item ipcp-unit-growth
8486 Specifies maximal overall growth of the compilation unit caused by
8487 interprocedural constant propagation. The default value is 10 which limits
8488 unit growth to 1.1 times the original size.
8490 @item large-stack-frame
8491 The limit specifying large stack frames. While inlining the algorithm is trying
8492 to not grow past this limit too much. Default value is 256 bytes.
8494 @item large-stack-frame-growth
8495 Specifies maximal growth of large stack frames caused by inlining in percents.
8496 The default value is 1000 which limits large stack frame growth to 11 times
8499 @item max-inline-insns-recursive
8500 @itemx max-inline-insns-recursive-auto
8501 Specifies maximum number of instructions out-of-line copy of self recursive inline
8502 function can grow into by performing recursive inlining.
8504 For functions declared inline @option{--param max-inline-insns-recursive} is
8505 taken into account. For function not declared inline, recursive inlining
8506 happens only when @option{-finline-functions} (included in @option{-O3}) is
8507 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8508 default value is 450.
8510 @item max-inline-recursive-depth
8511 @itemx max-inline-recursive-depth-auto
8512 Specifies maximum recursion depth used by the recursive inlining.
8514 For functions declared inline @option{--param max-inline-recursive-depth} is
8515 taken into account. For function not declared inline, recursive inlining
8516 happens only when @option{-finline-functions} (included in @option{-O3}) is
8517 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8520 @item min-inline-recursive-probability
8521 Recursive inlining is profitable only for function having deep recursion
8522 in average and can hurt for function having little recursion depth by
8523 increasing the prologue size or complexity of function body to other
8526 When profile feedback is available (see @option{-fprofile-generate}) the actual
8527 recursion depth can be guessed from probability that function will recurse via
8528 given call expression. This parameter limits inlining only to call expression
8529 whose probability exceeds given threshold (in percents). The default value is
8532 @item early-inlining-insns
8533 Specify growth that early inliner can make. In effect it increases amount of
8534 inlining for code having large abstraction penalty. The default value is 10.
8536 @item max-early-inliner-iterations
8537 @itemx max-early-inliner-iterations
8538 Limit of iterations of early inliner. This basically bounds number of nested
8539 indirect calls early inliner can resolve. Deeper chains are still handled by
8542 @item comdat-sharing-probability
8543 @itemx comdat-sharing-probability
8544 Probability (in percent) that C++ inline function with comdat visibility
8545 will be shared across multiple compilation units. The default value is 20.
8547 @item min-vect-loop-bound
8548 The minimum number of iterations under which a loop will not get vectorized
8549 when @option{-ftree-vectorize} is used. The number of iterations after
8550 vectorization needs to be greater than the value specified by this option
8551 to allow vectorization. The default value is 0.
8553 @item gcse-cost-distance-ratio
8554 Scaling factor in calculation of maximum distance an expression
8555 can be moved by GCSE optimizations. This is currently supported only in the
8556 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8557 will be with simple expressions, i.e., the expressions which have cost
8558 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8559 hoisting of simple expressions. The default value is 10.
8561 @item gcse-unrestricted-cost
8562 Cost, roughly measured as the cost of a single typical machine
8563 instruction, at which GCSE optimizations will not constrain
8564 the distance an expression can travel. This is currently
8565 supported only in the code hoisting pass. The lesser the cost,
8566 the more aggressive code hoisting will be. Specifying 0 will
8567 allow all expressions to travel unrestricted distances.
8568 The default value is 3.
8570 @item max-hoist-depth
8571 The depth of search in the dominator tree for expressions to hoist.
8572 This is used to avoid quadratic behavior in hoisting algorithm.
8573 The value of 0 will avoid limiting the search, but may slow down compilation
8574 of huge functions. The default value is 30.
8576 @item max-unrolled-insns
8577 The maximum number of instructions that a loop should have if that loop
8578 is unrolled, and if the loop is unrolled, it determines how many times
8579 the loop code is unrolled.
8581 @item max-average-unrolled-insns
8582 The maximum number of instructions biased by probabilities of their execution
8583 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8584 it determines how many times the loop code is unrolled.
8586 @item max-unroll-times
8587 The maximum number of unrollings of a single loop.
8589 @item max-peeled-insns
8590 The maximum number of instructions that a loop should have if that loop
8591 is peeled, and if the loop is peeled, it determines how many times
8592 the loop code is peeled.
8594 @item max-peel-times
8595 The maximum number of peelings of a single loop.
8597 @item max-completely-peeled-insns
8598 The maximum number of insns of a completely peeled loop.
8600 @item max-completely-peel-times
8601 The maximum number of iterations of a loop to be suitable for complete peeling.
8603 @item max-completely-peel-loop-nest-depth
8604 The maximum depth of a loop nest suitable for complete peeling.
8606 @item max-unswitch-insns
8607 The maximum number of insns of an unswitched loop.
8609 @item max-unswitch-level
8610 The maximum number of branches unswitched in a single loop.
8613 The minimum cost of an expensive expression in the loop invariant motion.
8615 @item iv-consider-all-candidates-bound
8616 Bound on number of candidates for induction variables below that
8617 all candidates are considered for each use in induction variable
8618 optimizations. Only the most relevant candidates are considered
8619 if there are more candidates, to avoid quadratic time complexity.
8621 @item iv-max-considered-uses
8622 The induction variable optimizations give up on loops that contain more
8623 induction variable uses.
8625 @item iv-always-prune-cand-set-bound
8626 If number of candidates in the set is smaller than this value,
8627 we always try to remove unnecessary ivs from the set during its
8628 optimization when a new iv is added to the set.
8630 @item scev-max-expr-size
8631 Bound on size of expressions used in the scalar evolutions analyzer.
8632 Large expressions slow the analyzer.
8634 @item scev-max-expr-complexity
8635 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8636 Complex expressions slow the analyzer.
8638 @item omega-max-vars
8639 The maximum number of variables in an Omega constraint system.
8640 The default value is 128.
8642 @item omega-max-geqs
8643 The maximum number of inequalities in an Omega constraint system.
8644 The default value is 256.
8647 The maximum number of equalities in an Omega constraint system.
8648 The default value is 128.
8650 @item omega-max-wild-cards
8651 The maximum number of wildcard variables that the Omega solver will
8652 be able to insert. The default value is 18.
8654 @item omega-hash-table-size
8655 The size of the hash table in the Omega solver. The default value is
8658 @item omega-max-keys
8659 The maximal number of keys used by the Omega solver. The default
8662 @item omega-eliminate-redundant-constraints
8663 When set to 1, use expensive methods to eliminate all redundant
8664 constraints. The default value is 0.
8666 @item vect-max-version-for-alignment-checks
8667 The maximum number of runtime checks that can be performed when
8668 doing loop versioning for alignment in the vectorizer. See option
8669 ftree-vect-loop-version for more information.
8671 @item vect-max-version-for-alias-checks
8672 The maximum number of runtime checks that can be performed when
8673 doing loop versioning for alias in the vectorizer. See option
8674 ftree-vect-loop-version for more information.
8676 @item max-iterations-to-track
8678 The maximum number of iterations of a loop the brute force algorithm
8679 for analysis of # of iterations of the loop tries to evaluate.
8681 @item hot-bb-count-fraction
8682 Select fraction of the maximal count of repetitions of basic block in program
8683 given basic block needs to have to be considered hot.
8685 @item hot-bb-frequency-fraction
8686 Select fraction of the entry block frequency of executions of basic block in
8687 function given basic block needs to have to be considered hot.
8689 @item max-predicted-iterations
8690 The maximum number of loop iterations we predict statically. This is useful
8691 in cases where function contain single loop with known bound and other loop
8692 with unknown. We predict the known number of iterations correctly, while
8693 the unknown number of iterations average to roughly 10. This means that the
8694 loop without bounds would appear artificially cold relative to the other one.
8696 @item align-threshold
8698 Select fraction of the maximal frequency of executions of basic block in
8699 function given basic block will get aligned.
8701 @item align-loop-iterations
8703 A loop expected to iterate at lest the selected number of iterations will get
8706 @item tracer-dynamic-coverage
8707 @itemx tracer-dynamic-coverage-feedback
8709 This value is used to limit superblock formation once the given percentage of
8710 executed instructions is covered. This limits unnecessary code size
8713 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8714 feedback is available. The real profiles (as opposed to statically estimated
8715 ones) are much less balanced allowing the threshold to be larger value.
8717 @item tracer-max-code-growth
8718 Stop tail duplication once code growth has reached given percentage. This is
8719 rather hokey argument, as most of the duplicates will be eliminated later in
8720 cross jumping, so it may be set to much higher values than is the desired code
8723 @item tracer-min-branch-ratio
8725 Stop reverse growth when the reverse probability of best edge is less than this
8726 threshold (in percent).
8728 @item tracer-min-branch-ratio
8729 @itemx tracer-min-branch-ratio-feedback
8731 Stop forward growth if the best edge do have probability lower than this
8734 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8735 compilation for profile feedback and one for compilation without. The value
8736 for compilation with profile feedback needs to be more conservative (higher) in
8737 order to make tracer effective.
8739 @item max-cse-path-length
8741 Maximum number of basic blocks on path that cse considers. The default is 10.
8744 The maximum instructions CSE process before flushing. The default is 1000.
8746 @item ggc-min-expand
8748 GCC uses a garbage collector to manage its own memory allocation. This
8749 parameter specifies the minimum percentage by which the garbage
8750 collector's heap should be allowed to expand between collections.
8751 Tuning this may improve compilation speed; it has no effect on code
8754 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8755 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8756 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8757 GCC is not able to calculate RAM on a particular platform, the lower
8758 bound of 30% is used. Setting this parameter and
8759 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8760 every opportunity. This is extremely slow, but can be useful for
8763 @item ggc-min-heapsize
8765 Minimum size of the garbage collector's heap before it begins bothering
8766 to collect garbage. The first collection occurs after the heap expands
8767 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8768 tuning this may improve compilation speed, and has no effect on code
8771 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8772 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8773 with a lower bound of 4096 (four megabytes) and an upper bound of
8774 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8775 particular platform, the lower bound is used. Setting this parameter
8776 very large effectively disables garbage collection. Setting this
8777 parameter and @option{ggc-min-expand} to zero causes a full collection
8778 to occur at every opportunity.
8780 @item max-reload-search-insns
8781 The maximum number of instruction reload should look backward for equivalent
8782 register. Increasing values mean more aggressive optimization, making the
8783 compile time increase with probably slightly better performance. The default
8786 @item max-cselib-memory-locations
8787 The maximum number of memory locations cselib should take into account.
8788 Increasing values mean more aggressive optimization, making the compile time
8789 increase with probably slightly better performance. The default value is 500.
8791 @item reorder-blocks-duplicate
8792 @itemx reorder-blocks-duplicate-feedback
8794 Used by basic block reordering pass to decide whether to use unconditional
8795 branch or duplicate the code on its destination. Code is duplicated when its
8796 estimated size is smaller than this value multiplied by the estimated size of
8797 unconditional jump in the hot spots of the program.
8799 The @option{reorder-block-duplicate-feedback} is used only when profile
8800 feedback is available and may be set to higher values than
8801 @option{reorder-block-duplicate} since information about the hot spots is more
8804 @item max-sched-ready-insns
8805 The maximum number of instructions ready to be issued the scheduler should
8806 consider at any given time during the first scheduling pass. Increasing
8807 values mean more thorough searches, making the compilation time increase
8808 with probably little benefit. The default value is 100.
8810 @item max-sched-region-blocks
8811 The maximum number of blocks in a region to be considered for
8812 interblock scheduling. The default value is 10.
8814 @item max-pipeline-region-blocks
8815 The maximum number of blocks in a region to be considered for
8816 pipelining in the selective scheduler. The default value is 15.
8818 @item max-sched-region-insns
8819 The maximum number of insns in a region to be considered for
8820 interblock scheduling. The default value is 100.
8822 @item max-pipeline-region-insns
8823 The maximum number of insns in a region to be considered for
8824 pipelining in the selective scheduler. The default value is 200.
8827 The minimum probability (in percents) of reaching a source block
8828 for interblock speculative scheduling. The default value is 40.
8830 @item max-sched-extend-regions-iters
8831 The maximum number of iterations through CFG to extend regions.
8832 0 - disable region extension,
8833 N - do at most N iterations.
8834 The default value is 0.
8836 @item max-sched-insn-conflict-delay
8837 The maximum conflict delay for an insn to be considered for speculative motion.
8838 The default value is 3.
8840 @item sched-spec-prob-cutoff
8841 The minimal probability of speculation success (in percents), so that
8842 speculative insn will be scheduled.
8843 The default value is 40.
8845 @item sched-mem-true-dep-cost
8846 Minimal distance (in CPU cycles) between store and load targeting same
8847 memory locations. The default value is 1.
8849 @item selsched-max-lookahead
8850 The maximum size of the lookahead window of selective scheduling. It is a
8851 depth of search for available instructions.
8852 The default value is 50.
8854 @item selsched-max-sched-times
8855 The maximum number of times that an instruction will be scheduled during
8856 selective scheduling. This is the limit on the number of iterations
8857 through which the instruction may be pipelined. The default value is 2.
8859 @item selsched-max-insns-to-rename
8860 The maximum number of best instructions in the ready list that are considered
8861 for renaming in the selective scheduler. The default value is 2.
8864 The minimum value of stage count that swing modulo scheduler will
8865 generate. The default value is 2.
8867 @item max-last-value-rtl
8868 The maximum size measured as number of RTLs that can be recorded in an expression
8869 in combiner for a pseudo register as last known value of that register. The default
8872 @item integer-share-limit
8873 Small integer constants can use a shared data structure, reducing the
8874 compiler's memory usage and increasing its speed. This sets the maximum
8875 value of a shared integer constant. The default value is 256.
8877 @item min-virtual-mappings
8878 Specifies the minimum number of virtual mappings in the incremental
8879 SSA updater that should be registered to trigger the virtual mappings
8880 heuristic defined by virtual-mappings-ratio. The default value is
8883 @item virtual-mappings-ratio
8884 If the number of virtual mappings is virtual-mappings-ratio bigger
8885 than the number of virtual symbols to be updated, then the incremental
8886 SSA updater switches to a full update for those symbols. The default
8889 @item ssp-buffer-size
8890 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8891 protection when @option{-fstack-protection} is used.
8893 @item max-jump-thread-duplication-stmts
8894 Maximum number of statements allowed in a block that needs to be
8895 duplicated when threading jumps.
8897 @item max-fields-for-field-sensitive
8898 Maximum number of fields in a structure we will treat in
8899 a field sensitive manner during pointer analysis. The default is zero
8900 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8902 @item prefetch-latency
8903 Estimate on average number of instructions that are executed before
8904 prefetch finishes. The distance we prefetch ahead is proportional
8905 to this constant. Increasing this number may also lead to less
8906 streams being prefetched (see @option{simultaneous-prefetches}).
8908 @item simultaneous-prefetches
8909 Maximum number of prefetches that can run at the same time.
8911 @item l1-cache-line-size
8912 The size of cache line in L1 cache, in bytes.
8915 The size of L1 cache, in kilobytes.
8918 The size of L2 cache, in kilobytes.
8920 @item min-insn-to-prefetch-ratio
8921 The minimum ratio between the number of instructions and the
8922 number of prefetches to enable prefetching in a loop.
8924 @item prefetch-min-insn-to-mem-ratio
8925 The minimum ratio between the number of instructions and the
8926 number of memory references to enable prefetching in a loop.
8928 @item use-canonical-types
8929 Whether the compiler should use the ``canonical'' type system. By
8930 default, this should always be 1, which uses a more efficient internal
8931 mechanism for comparing types in C++ and Objective-C++. However, if
8932 bugs in the canonical type system are causing compilation failures,
8933 set this value to 0 to disable canonical types.
8935 @item switch-conversion-max-branch-ratio
8936 Switch initialization conversion will refuse to create arrays that are
8937 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8938 branches in the switch.
8940 @item max-partial-antic-length
8941 Maximum length of the partial antic set computed during the tree
8942 partial redundancy elimination optimization (@option{-ftree-pre}) when
8943 optimizing at @option{-O3} and above. For some sorts of source code
8944 the enhanced partial redundancy elimination optimization can run away,
8945 consuming all of the memory available on the host machine. This
8946 parameter sets a limit on the length of the sets that are computed,
8947 which prevents the runaway behavior. Setting a value of 0 for
8948 this parameter will allow an unlimited set length.
8950 @item sccvn-max-scc-size
8951 Maximum size of a strongly connected component (SCC) during SCCVN
8952 processing. If this limit is hit, SCCVN processing for the whole
8953 function will not be done and optimizations depending on it will
8954 be disabled. The default maximum SCC size is 10000.
8956 @item ira-max-loops-num
8957 IRA uses a regional register allocation by default. If a function
8958 contains loops more than number given by the parameter, only at most
8959 given number of the most frequently executed loops will form regions
8960 for the regional register allocation. The default value of the
8963 @item ira-max-conflict-table-size
8964 Although IRA uses a sophisticated algorithm of compression conflict
8965 table, the table can be still big for huge functions. If the conflict
8966 table for a function could be more than size in MB given by the
8967 parameter, the conflict table is not built and faster, simpler, and
8968 lower quality register allocation algorithm will be used. The
8969 algorithm do not use pseudo-register conflicts. The default value of
8970 the parameter is 2000.
8972 @item ira-loop-reserved-regs
8973 IRA can be used to evaluate more accurate register pressure in loops
8974 for decision to move loop invariants (see @option{-O3}). The number
8975 of available registers reserved for some other purposes is described
8976 by this parameter. The default value of the parameter is 2 which is
8977 minimal number of registers needed for execution of typical
8978 instruction. This value is the best found from numerous experiments.
8980 @item loop-invariant-max-bbs-in-loop
8981 Loop invariant motion can be very expensive, both in compile time and
8982 in amount of needed compile time memory, with very large loops. Loops
8983 with more basic blocks than this parameter won't have loop invariant
8984 motion optimization performed on them. The default value of the
8985 parameter is 1000 for -O1 and 10000 for -O2 and above.
8987 @item max-vartrack-size
8988 Sets a maximum number of hash table slots to use during variable
8989 tracking dataflow analysis of any function. If this limit is exceeded
8990 with variable tracking at assignments enabled, analysis for that
8991 function is retried without it, after removing all debug insns from
8992 the function. If the limit is exceeded even without debug insns, var
8993 tracking analysis is completely disabled for the function. Setting
8994 the parameter to zero makes it unlimited.
8996 @item max-vartrack-expr-depth
8997 Sets a maximum number of recursion levels when attempting to map
8998 variable names or debug temporaries to value expressions. This trades
8999 compile time for more complete debug information. If this is set too
9000 low, value expressions that are available and could be represented in
9001 debug information may end up not being used; setting this higher may
9002 enable the compiler to find more complex debug expressions, but compile
9003 time may grow exponentially, and even then, it may fail to find more
9004 usable expressions. The default is 10.
9006 @item min-nondebug-insn-uid
9007 Use uids starting at this parameter for nondebug insns. The range below
9008 the parameter is reserved exclusively for debug insns created by
9009 @option{-fvar-tracking-assignments}, but debug insns may get
9010 (non-overlapping) uids above it if the reserved range is exhausted.
9012 @item ipa-sra-ptr-growth-factor
9013 IPA-SRA will replace a pointer to an aggregate with one or more new
9014 parameters only when their cumulative size is less or equal to
9015 @option{ipa-sra-ptr-growth-factor} times the size of the original
9018 @item graphite-max-nb-scop-params
9019 To avoid exponential effects in the Graphite loop transforms, the
9020 number of parameters in a Static Control Part (SCoP) is bounded. The
9021 default value is 10 parameters. A variable whose value is unknown at
9022 compile time and defined outside a SCoP is a parameter of the SCoP.
9024 @item graphite-max-bbs-per-function
9025 To avoid exponential effects in the detection of SCoPs, the size of
9026 the functions analyzed by Graphite is bounded. The default value is
9029 @item loop-block-tile-size
9030 Loop blocking or strip mining transforms, enabled with
9031 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9032 loop in the loop nest by a given number of iterations. The strip
9033 length can be changed using the @option{loop-block-tile-size}
9034 parameter. The default value is 51 iterations.
9036 @item ipa-cp-value-list-size
9037 IPA-CP attempts to track all possible values and types passed to a function's
9038 parameter in order to propagate them and perform devirtualization.
9039 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9040 stores per one formal parameter of a function.
9042 @item lto-partitions
9043 Specify desired number of partitions produced during WHOPR compilation.
9044 The number of partitions should exceed the number of CPUs used for compilation.
9045 The default value is 32.
9047 @item lto-minpartition
9048 Size of minimal partition for WHOPR (in estimated instructions).
9049 This prevents expenses of splitting very small programs into too many
9052 @item cxx-max-namespaces-for-diagnostic-help
9053 The maximum number of namespaces to consult for suggestions when C++
9054 name lookup fails for an identifier. The default is 1000.
9056 @item max-stores-to-sink
9057 The maximum number of conditional stores paires that can be sunk. Set to 0
9058 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9059 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9061 @item case-values-threshold
9062 The smallest number of different values for which it is best to use a
9063 jump-table instead of a tree of conditional branches. If the value is
9064 0, use the default for the machine. The default is 0.
9069 @node Preprocessor Options
9070 @section Options Controlling the Preprocessor
9071 @cindex preprocessor options
9072 @cindex options, preprocessor
9074 These options control the C preprocessor, which is run on each C source
9075 file before actual compilation.
9077 If you use the @option{-E} option, nothing is done except preprocessing.
9078 Some of these options make sense only together with @option{-E} because
9079 they cause the preprocessor output to be unsuitable for actual
9083 @item -Wp,@var{option}
9085 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9086 and pass @var{option} directly through to the preprocessor. If
9087 @var{option} contains commas, it is split into multiple options at the
9088 commas. However, many options are modified, translated or interpreted
9089 by the compiler driver before being passed to the preprocessor, and
9090 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9091 interface is undocumented and subject to change, so whenever possible
9092 you should avoid using @option{-Wp} and let the driver handle the
9095 @item -Xpreprocessor @var{option}
9096 @opindex Xpreprocessor
9097 Pass @var{option} as an option to the preprocessor. You can use this to
9098 supply system-specific preprocessor options which GCC does not know how to
9101 If you want to pass an option that takes an argument, you must use
9102 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9105 @include cppopts.texi
9107 @node Assembler Options
9108 @section Passing Options to the Assembler
9110 @c prevent bad page break with this line
9111 You can pass options to the assembler.
9114 @item -Wa,@var{option}
9116 Pass @var{option} as an option to the assembler. If @var{option}
9117 contains commas, it is split into multiple options at the commas.
9119 @item -Xassembler @var{option}
9121 Pass @var{option} as an option to the assembler. You can use this to
9122 supply system-specific assembler options which GCC does not know how to
9125 If you want to pass an option that takes an argument, you must use
9126 @option{-Xassembler} twice, once for the option and once for the argument.
9131 @section Options for Linking
9132 @cindex link options
9133 @cindex options, linking
9135 These options come into play when the compiler links object files into
9136 an executable output file. They are meaningless if the compiler is
9137 not doing a link step.
9141 @item @var{object-file-name}
9142 A file name that does not end in a special recognized suffix is
9143 considered to name an object file or library. (Object files are
9144 distinguished from libraries by the linker according to the file
9145 contents.) If linking is done, these object files are used as input
9154 If any of these options is used, then the linker is not run, and
9155 object file names should not be used as arguments. @xref{Overall
9159 @item -l@var{library}
9160 @itemx -l @var{library}
9162 Search the library named @var{library} when linking. (The second
9163 alternative with the library as a separate argument is only for
9164 POSIX compliance and is not recommended.)
9166 It makes a difference where in the command you write this option; the
9167 linker searches and processes libraries and object files in the order they
9168 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9169 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9170 to functions in @samp{z}, those functions may not be loaded.
9172 The linker searches a standard list of directories for the library,
9173 which is actually a file named @file{lib@var{library}.a}. The linker
9174 then uses this file as if it had been specified precisely by name.
9176 The directories searched include several standard system directories
9177 plus any that you specify with @option{-L}.
9179 Normally the files found this way are library files---archive files
9180 whose members are object files. The linker handles an archive file by
9181 scanning through it for members which define symbols that have so far
9182 been referenced but not defined. But if the file that is found is an
9183 ordinary object file, it is linked in the usual fashion. The only
9184 difference between using an @option{-l} option and specifying a file name
9185 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9186 and searches several directories.
9190 You need this special case of the @option{-l} option in order to
9191 link an Objective-C or Objective-C++ program.
9194 @opindex nostartfiles
9195 Do not use the standard system startup files when linking.
9196 The standard system libraries are used normally, unless @option{-nostdlib}
9197 or @option{-nodefaultlibs} is used.
9199 @item -nodefaultlibs
9200 @opindex nodefaultlibs
9201 Do not use the standard system libraries when linking.
9202 Only the libraries you specify will be passed to the linker, options
9203 specifying linkage of the system libraries, such as @code{-static-libgcc}
9204 or @code{-shared-libgcc}, will be ignored.
9205 The standard startup files are used normally, unless @option{-nostartfiles}
9206 is used. The compiler may generate calls to @code{memcmp},
9207 @code{memset}, @code{memcpy} and @code{memmove}.
9208 These entries are usually resolved by entries in
9209 libc. These entry points should be supplied through some other
9210 mechanism when this option is specified.
9214 Do not use the standard system startup files or libraries when linking.
9215 No startup files and only the libraries you specify will be passed to
9216 the linker, options specifying linkage of the system libraries, such as
9217 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9218 The compiler may generate calls to @code{memcmp}, @code{memset},
9219 @code{memcpy} and @code{memmove}.
9220 These entries are usually resolved by entries in
9221 libc. These entry points should be supplied through some other
9222 mechanism when this option is specified.
9224 @cindex @option{-lgcc}, use with @option{-nostdlib}
9225 @cindex @option{-nostdlib} and unresolved references
9226 @cindex unresolved references and @option{-nostdlib}
9227 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9228 @cindex @option{-nodefaultlibs} and unresolved references
9229 @cindex unresolved references and @option{-nodefaultlibs}
9230 One of the standard libraries bypassed by @option{-nostdlib} and
9231 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9232 that GCC uses to overcome shortcomings of particular machines, or special
9233 needs for some languages.
9234 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9235 Collection (GCC) Internals},
9236 for more discussion of @file{libgcc.a}.)
9237 In most cases, you need @file{libgcc.a} even when you want to avoid
9238 other standard libraries. In other words, when you specify @option{-nostdlib}
9239 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9240 This ensures that you have no unresolved references to internal GCC
9241 library subroutines. (For example, @samp{__main}, used to ensure C++
9242 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9243 GNU Compiler Collection (GCC) Internals}.)
9247 Produce a position independent executable on targets which support it.
9248 For predictable results, you must also specify the same set of options
9249 that were used to generate code (@option{-fpie}, @option{-fPIE},
9250 or model suboptions) when you specify this option.
9254 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9255 that support it. This instructs the linker to add all symbols, not
9256 only used ones, to the dynamic symbol table. This option is needed
9257 for some uses of @code{dlopen} or to allow obtaining backtraces
9258 from within a program.
9262 Remove all symbol table and relocation information from the executable.
9266 On systems that support dynamic linking, this prevents linking with the shared
9267 libraries. On other systems, this option has no effect.
9271 Produce a shared object which can then be linked with other objects to
9272 form an executable. Not all systems support this option. For predictable
9273 results, you must also specify the same set of options that were used to
9274 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9275 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9276 needs to build supplementary stub code for constructors to work. On
9277 multi-libbed systems, @samp{gcc -shared} must select the correct support
9278 libraries to link against. Failing to supply the correct flags may lead
9279 to subtle defects. Supplying them in cases where they are not necessary
9282 @item -shared-libgcc
9283 @itemx -static-libgcc
9284 @opindex shared-libgcc
9285 @opindex static-libgcc
9286 On systems that provide @file{libgcc} as a shared library, these options
9287 force the use of either the shared or static version respectively.
9288 If no shared version of @file{libgcc} was built when the compiler was
9289 configured, these options have no effect.
9291 There are several situations in which an application should use the
9292 shared @file{libgcc} instead of the static version. The most common
9293 of these is when the application wishes to throw and catch exceptions
9294 across different shared libraries. In that case, each of the libraries
9295 as well as the application itself should use the shared @file{libgcc}.
9297 Therefore, the G++ and GCJ drivers automatically add
9298 @option{-shared-libgcc} whenever you build a shared library or a main
9299 executable, because C++ and Java programs typically use exceptions, so
9300 this is the right thing to do.
9302 If, instead, you use the GCC driver to create shared libraries, you may
9303 find that they will not always be linked with the shared @file{libgcc}.
9304 If GCC finds, at its configuration time, that you have a non-GNU linker
9305 or a GNU linker that does not support option @option{--eh-frame-hdr},
9306 it will link the shared version of @file{libgcc} into shared libraries
9307 by default. Otherwise, it will take advantage of the linker and optimize
9308 away the linking with the shared version of @file{libgcc}, linking with
9309 the static version of libgcc by default. This allows exceptions to
9310 propagate through such shared libraries, without incurring relocation
9311 costs at library load time.
9313 However, if a library or main executable is supposed to throw or catch
9314 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9315 for the languages used in the program, or using the option
9316 @option{-shared-libgcc}, such that it is linked with the shared
9319 @item -static-libstdc++
9320 When the @command{g++} program is used to link a C++ program, it will
9321 normally automatically link against @option{libstdc++}. If
9322 @file{libstdc++} is available as a shared library, and the
9323 @option{-static} option is not used, then this will link against the
9324 shared version of @file{libstdc++}. That is normally fine. However, it
9325 is sometimes useful to freeze the version of @file{libstdc++} used by
9326 the program without going all the way to a fully static link. The
9327 @option{-static-libstdc++} option directs the @command{g++} driver to
9328 link @file{libstdc++} statically, without necessarily linking other
9329 libraries statically.
9333 Bind references to global symbols when building a shared object. Warn
9334 about any unresolved references (unless overridden by the link editor
9335 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9338 @item -T @var{script}
9340 @cindex linker script
9341 Use @var{script} as the linker script. This option is supported by most
9342 systems using the GNU linker. On some targets, such as bare-board
9343 targets without an operating system, the @option{-T} option may be required
9344 when linking to avoid references to undefined symbols.
9346 @item -Xlinker @var{option}
9348 Pass @var{option} as an option to the linker. You can use this to
9349 supply system-specific linker options which GCC does not know how to
9352 If you want to pass an option that takes a separate argument, you must use
9353 @option{-Xlinker} twice, once for the option and once for the argument.
9354 For example, to pass @option{-assert definitions}, you must write
9355 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9356 @option{-Xlinker "-assert definitions"}, because this passes the entire
9357 string as a single argument, which is not what the linker expects.
9359 When using the GNU linker, it is usually more convenient to pass
9360 arguments to linker options using the @option{@var{option}=@var{value}}
9361 syntax than as separate arguments. For example, you can specify
9362 @samp{-Xlinker -Map=output.map} rather than
9363 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9364 this syntax for command-line options.
9366 @item -Wl,@var{option}
9368 Pass @var{option} as an option to the linker. If @var{option} contains
9369 commas, it is split into multiple options at the commas. You can use this
9370 syntax to pass an argument to the option.
9371 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9372 linker. When using the GNU linker, you can also get the same effect with
9373 @samp{-Wl,-Map=output.map}.
9375 @item -u @var{symbol}
9377 Pretend the symbol @var{symbol} is undefined, to force linking of
9378 library modules to define it. You can use @option{-u} multiple times with
9379 different symbols to force loading of additional library modules.
9382 @node Directory Options
9383 @section Options for Directory Search
9384 @cindex directory options
9385 @cindex options, directory search
9388 These options specify directories to search for header files, for
9389 libraries and for parts of the compiler:
9394 Add the directory @var{dir} to the head of the list of directories to be
9395 searched for header files. This can be used to override a system header
9396 file, substituting your own version, since these directories are
9397 searched before the system header file directories. However, you should
9398 not use this option to add directories that contain vendor-supplied
9399 system header files (use @option{-isystem} for that). If you use more than
9400 one @option{-I} option, the directories are scanned in left-to-right
9401 order; the standard system directories come after.
9403 If a standard system include directory, or a directory specified with
9404 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9405 option will be ignored. The directory will still be searched but as a
9406 system directory at its normal position in the system include chain.
9407 This is to ensure that GCC's procedure to fix buggy system headers and
9408 the ordering for the include_next directive are not inadvertently changed.
9409 If you really need to change the search order for system directories,
9410 use the @option{-nostdinc} and/or @option{-isystem} options.
9412 @item -iplugindir=@var{dir}
9413 Set the directory to search for plugins which are passed
9414 by @option{-fplugin=@var{name}} instead of
9415 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9416 to be used by the user, but only passed by the driver.
9418 @item -iquote@var{dir}
9420 Add the directory @var{dir} to the head of the list of directories to
9421 be searched for header files only for the case of @samp{#include
9422 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9423 otherwise just like @option{-I}.
9427 Add directory @var{dir} to the list of directories to be searched
9430 @item -B@var{prefix}
9432 This option specifies where to find the executables, libraries,
9433 include files, and data files of the compiler itself.
9435 The compiler driver program runs one or more of the subprograms
9436 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9437 @var{prefix} as a prefix for each program it tries to run, both with and
9438 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9440 For each subprogram to be run, the compiler driver first tries the
9441 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9442 was not specified, the driver tries two standard prefixes, which are
9443 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9444 those results in a file name that is found, the unmodified program
9445 name is searched for using the directories specified in your
9446 @env{PATH} environment variable.
9448 The compiler will check to see if the path provided by the @option{-B}
9449 refers to a directory, and if necessary it will add a directory
9450 separator character at the end of the path.
9452 @option{-B} prefixes that effectively specify directory names also apply
9453 to libraries in the linker, because the compiler translates these
9454 options into @option{-L} options for the linker. They also apply to
9455 includes files in the preprocessor, because the compiler translates these
9456 options into @option{-isystem} options for the preprocessor. In this case,
9457 the compiler appends @samp{include} to the prefix.
9459 The run-time support file @file{libgcc.a} can also be searched for using
9460 the @option{-B} prefix, if needed. If it is not found there, the two
9461 standard prefixes above are tried, and that is all. The file is left
9462 out of the link if it is not found by those means.
9464 Another way to specify a prefix much like the @option{-B} prefix is to use
9465 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9468 As a special kludge, if the path provided by @option{-B} is
9469 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9470 9, then it will be replaced by @file{[dir/]include}. This is to help
9471 with boot-strapping the compiler.
9473 @item -specs=@var{file}
9475 Process @var{file} after the compiler reads in the standard @file{specs}
9476 file, in order to override the defaults that the @file{gcc} driver
9477 program uses when determining what switches to pass to @file{cc1},
9478 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9479 @option{-specs=@var{file}} can be specified on the command line, and they
9480 are processed in order, from left to right.
9482 @item --sysroot=@var{dir}
9484 Use @var{dir} as the logical root directory for headers and libraries.
9485 For example, if the compiler would normally search for headers in
9486 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9487 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9489 If you use both this option and the @option{-isysroot} option, then
9490 the @option{--sysroot} option will apply to libraries, but the
9491 @option{-isysroot} option will apply to header files.
9493 The GNU linker (beginning with version 2.16) has the necessary support
9494 for this option. If your linker does not support this option, the
9495 header file aspect of @option{--sysroot} will still work, but the
9496 library aspect will not.
9500 This option has been deprecated. Please use @option{-iquote} instead for
9501 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9502 Any directories you specify with @option{-I} options before the @option{-I-}
9503 option are searched only for the case of @samp{#include "@var{file}"};
9504 they are not searched for @samp{#include <@var{file}>}.
9506 If additional directories are specified with @option{-I} options after
9507 the @option{-I-}, these directories are searched for all @samp{#include}
9508 directives. (Ordinarily @emph{all} @option{-I} directories are used
9511 In addition, the @option{-I-} option inhibits the use of the current
9512 directory (where the current input file came from) as the first search
9513 directory for @samp{#include "@var{file}"}. There is no way to
9514 override this effect of @option{-I-}. With @option{-I.} you can specify
9515 searching the directory which was current when the compiler was
9516 invoked. That is not exactly the same as what the preprocessor does
9517 by default, but it is often satisfactory.
9519 @option{-I-} does not inhibit the use of the standard system directories
9520 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9527 @section Specifying subprocesses and the switches to pass to them
9530 @command{gcc} is a driver program. It performs its job by invoking a
9531 sequence of other programs to do the work of compiling, assembling and
9532 linking. GCC interprets its command-line parameters and uses these to
9533 deduce which programs it should invoke, and which command-line options
9534 it ought to place on their command lines. This behavior is controlled
9535 by @dfn{spec strings}. In most cases there is one spec string for each
9536 program that GCC can invoke, but a few programs have multiple spec
9537 strings to control their behavior. The spec strings built into GCC can
9538 be overridden by using the @option{-specs=} command-line switch to specify
9541 @dfn{Spec files} are plaintext files that are used to construct spec
9542 strings. They consist of a sequence of directives separated by blank
9543 lines. The type of directive is determined by the first non-whitespace
9544 character on the line and it can be one of the following:
9547 @item %@var{command}
9548 Issues a @var{command} to the spec file processor. The commands that can
9552 @item %include <@var{file}>
9553 @cindex @code{%include}
9554 Search for @var{file} and insert its text at the current point in the
9557 @item %include_noerr <@var{file}>
9558 @cindex @code{%include_noerr}
9559 Just like @samp{%include}, but do not generate an error message if the include
9560 file cannot be found.
9562 @item %rename @var{old_name} @var{new_name}
9563 @cindex @code{%rename}
9564 Rename the spec string @var{old_name} to @var{new_name}.
9568 @item *[@var{spec_name}]:
9569 This tells the compiler to create, override or delete the named spec
9570 string. All lines after this directive up to the next directive or
9571 blank line are considered to be the text for the spec string. If this
9572 results in an empty string then the spec will be deleted. (Or, if the
9573 spec did not exist, then nothing will happen.) Otherwise, if the spec
9574 does not currently exist a new spec will be created. If the spec does
9575 exist then its contents will be overridden by the text of this
9576 directive, unless the first character of that text is the @samp{+}
9577 character, in which case the text will be appended to the spec.
9579 @item [@var{suffix}]:
9580 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9581 and up to the next directive or blank line are considered to make up the
9582 spec string for the indicated suffix. When the compiler encounters an
9583 input file with the named suffix, it will processes the spec string in
9584 order to work out how to compile that file. For example:
9591 This says that any input file whose name ends in @samp{.ZZ} should be
9592 passed to the program @samp{z-compile}, which should be invoked with the
9593 command-line switch @option{-input} and with the result of performing the
9594 @samp{%i} substitution. (See below.)
9596 As an alternative to providing a spec string, the text that follows a
9597 suffix directive can be one of the following:
9600 @item @@@var{language}
9601 This says that the suffix is an alias for a known @var{language}. This is
9602 similar to using the @option{-x} command-line switch to GCC to specify a
9603 language explicitly. For example:
9610 Says that .ZZ files are, in fact, C++ source files.
9613 This causes an error messages saying:
9616 @var{name} compiler not installed on this system.
9620 GCC already has an extensive list of suffixes built into it.
9621 This directive will add an entry to the end of the list of suffixes, but
9622 since the list is searched from the end backwards, it is effectively
9623 possible to override earlier entries using this technique.
9627 GCC has the following spec strings built into it. Spec files can
9628 override these strings or create their own. Note that individual
9629 targets can also add their own spec strings to this list.
9632 asm Options to pass to the assembler
9633 asm_final Options to pass to the assembler post-processor
9634 cpp Options to pass to the C preprocessor
9635 cc1 Options to pass to the C compiler
9636 cc1plus Options to pass to the C++ compiler
9637 endfile Object files to include at the end of the link
9638 link Options to pass to the linker
9639 lib Libraries to include on the command line to the linker
9640 libgcc Decides which GCC support library to pass to the linker
9641 linker Sets the name of the linker
9642 predefines Defines to be passed to the C preprocessor
9643 signed_char Defines to pass to CPP to say whether @code{char} is signed
9645 startfile Object files to include at the start of the link
9648 Here is a small example of a spec file:
9654 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9657 This example renames the spec called @samp{lib} to @samp{old_lib} and
9658 then overrides the previous definition of @samp{lib} with a new one.
9659 The new definition adds in some extra command-line options before
9660 including the text of the old definition.
9662 @dfn{Spec strings} are a list of command-line options to be passed to their
9663 corresponding program. In addition, the spec strings can contain
9664 @samp{%}-prefixed sequences to substitute variable text or to
9665 conditionally insert text into the command line. Using these constructs
9666 it is possible to generate quite complex command lines.
9668 Here is a table of all defined @samp{%}-sequences for spec
9669 strings. Note that spaces are not generated automatically around the
9670 results of expanding these sequences. Therefore you can concatenate them
9671 together or combine them with constant text in a single argument.
9675 Substitute one @samp{%} into the program name or argument.
9678 Substitute the name of the input file being processed.
9681 Substitute the basename of the input file being processed.
9682 This is the substring up to (and not including) the last period
9683 and not including the directory.
9686 This is the same as @samp{%b}, but include the file suffix (text after
9690 Marks the argument containing or following the @samp{%d} as a
9691 temporary file name, so that that file will be deleted if GCC exits
9692 successfully. Unlike @samp{%g}, this contributes no text to the
9695 @item %g@var{suffix}
9696 Substitute a file name that has suffix @var{suffix} and is chosen
9697 once per compilation, and mark the argument in the same way as
9698 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9699 name is now chosen in a way that is hard to predict even when previously
9700 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9701 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9702 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9703 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9704 was simply substituted with a file name chosen once per compilation,
9705 without regard to any appended suffix (which was therefore treated
9706 just like ordinary text), making such attacks more likely to succeed.
9708 @item %u@var{suffix}
9709 Like @samp{%g}, but generates a new temporary file name even if
9710 @samp{%u@var{suffix}} was already seen.
9712 @item %U@var{suffix}
9713 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9714 new one if there is no such last file name. In the absence of any
9715 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9716 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9717 would involve the generation of two distinct file names, one
9718 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9719 simply substituted with a file name chosen for the previous @samp{%u},
9720 without regard to any appended suffix.
9722 @item %j@var{suffix}
9723 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9724 writable, and if save-temps is off; otherwise, substitute the name
9725 of a temporary file, just like @samp{%u}. This temporary file is not
9726 meant for communication between processes, but rather as a junk
9729 @item %|@var{suffix}
9730 @itemx %m@var{suffix}
9731 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9732 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9733 all. These are the two most common ways to instruct a program that it
9734 should read from standard input or write to standard output. If you
9735 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9736 construct: see for example @file{f/lang-specs.h}.
9738 @item %.@var{SUFFIX}
9739 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9740 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9741 terminated by the next space or %.
9744 Marks the argument containing or following the @samp{%w} as the
9745 designated output file of this compilation. This puts the argument
9746 into the sequence of arguments that @samp{%o} will substitute later.
9749 Substitutes the names of all the output files, with spaces
9750 automatically placed around them. You should write spaces
9751 around the @samp{%o} as well or the results are undefined.
9752 @samp{%o} is for use in the specs for running the linker.
9753 Input files whose names have no recognized suffix are not compiled
9754 at all, but they are included among the output files, so they will
9758 Substitutes the suffix for object files. Note that this is
9759 handled specially when it immediately follows @samp{%g, %u, or %U},
9760 because of the need for those to form complete file names. The
9761 handling is such that @samp{%O} is treated exactly as if it had already
9762 been substituted, except that @samp{%g, %u, and %U} do not currently
9763 support additional @var{suffix} characters following @samp{%O} as they would
9764 following, for example, @samp{.o}.
9767 Substitutes the standard macro predefinitions for the
9768 current target machine. Use this when running @code{cpp}.
9771 Like @samp{%p}, but puts @samp{__} before and after the name of each
9772 predefined macro, except for macros that start with @samp{__} or with
9773 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9777 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9778 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9779 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9780 and @option{-imultilib} as necessary.
9783 Current argument is the name of a library or startup file of some sort.
9784 Search for that file in a standard list of directories and substitute
9785 the full name found. The current working directory is included in the
9786 list of directories scanned.
9789 Current argument is the name of a linker script. Search for that file
9790 in the current list of directories to scan for libraries. If the file
9791 is located insert a @option{--script} option into the command line
9792 followed by the full path name found. If the file is not found then
9793 generate an error message. Note: the current working directory is not
9797 Print @var{str} as an error message. @var{str} is terminated by a newline.
9798 Use this when inconsistent options are detected.
9801 Substitute the contents of spec string @var{name} at this point.
9803 @item %x@{@var{option}@}
9804 Accumulate an option for @samp{%X}.
9807 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9811 Output the accumulated assembler options specified by @option{-Wa}.
9814 Output the accumulated preprocessor options specified by @option{-Wp}.
9817 Process the @code{asm} spec. This is used to compute the
9818 switches to be passed to the assembler.
9821 Process the @code{asm_final} spec. This is a spec string for
9822 passing switches to an assembler post-processor, if such a program is
9826 Process the @code{link} spec. This is the spec for computing the
9827 command line passed to the linker. Typically it will make use of the
9828 @samp{%L %G %S %D and %E} sequences.
9831 Dump out a @option{-L} option for each directory that GCC believes might
9832 contain startup files. If the target supports multilibs then the
9833 current multilib directory will be prepended to each of these paths.
9836 Process the @code{lib} spec. This is a spec string for deciding which
9837 libraries should be included on the command line to the linker.
9840 Process the @code{libgcc} spec. This is a spec string for deciding
9841 which GCC support library should be included on the command line to the linker.
9844 Process the @code{startfile} spec. This is a spec for deciding which
9845 object files should be the first ones passed to the linker. Typically
9846 this might be a file named @file{crt0.o}.
9849 Process the @code{endfile} spec. This is a spec string that specifies
9850 the last object files that will be passed to the linker.
9853 Process the @code{cpp} spec. This is used to construct the arguments
9854 to be passed to the C preprocessor.
9857 Process the @code{cc1} spec. This is used to construct the options to be
9858 passed to the actual C compiler (@samp{cc1}).
9861 Process the @code{cc1plus} spec. This is used to construct the options to be
9862 passed to the actual C++ compiler (@samp{cc1plus}).
9865 Substitute the variable part of a matched option. See below.
9866 Note that each comma in the substituted string is replaced by
9870 Remove all occurrences of @code{-S} from the command line. Note---this
9871 command is position dependent. @samp{%} commands in the spec string
9872 before this one will see @code{-S}, @samp{%} commands in the spec string
9873 after this one will not.
9875 @item %:@var{function}(@var{args})
9876 Call the named function @var{function}, passing it @var{args}.
9877 @var{args} is first processed as a nested spec string, then split
9878 into an argument vector in the usual fashion. The function returns
9879 a string which is processed as if it had appeared literally as part
9880 of the current spec.
9882 The following built-in spec functions are provided:
9886 The @code{getenv} spec function takes two arguments: an environment
9887 variable name and a string. If the environment variable is not
9888 defined, a fatal error is issued. Otherwise, the return value is the
9889 value of the environment variable concatenated with the string. For
9890 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9893 %:getenv(TOPDIR /include)
9896 expands to @file{/path/to/top/include}.
9898 @item @code{if-exists}
9899 The @code{if-exists} spec function takes one argument, an absolute
9900 pathname to a file. If the file exists, @code{if-exists} returns the
9901 pathname. Here is a small example of its usage:
9905 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9908 @item @code{if-exists-else}
9909 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9910 spec function, except that it takes two arguments. The first argument is
9911 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9912 returns the pathname. If it does not exist, it returns the second argument.
9913 This way, @code{if-exists-else} can be used to select one file or another,
9914 based on the existence of the first. Here is a small example of its usage:
9918 crt0%O%s %:if-exists(crti%O%s) \
9919 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9922 @item @code{replace-outfile}
9923 The @code{replace-outfile} spec function takes two arguments. It looks for the
9924 first argument in the outfiles array and replaces it with the second argument. Here
9925 is a small example of its usage:
9928 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9931 @item @code{remove-outfile}
9932 The @code{remove-outfile} spec function takes one argument. It looks for the
9933 first argument in the outfiles array and removes it. Here is a small example
9937 %:remove-outfile(-lm)
9940 @item @code{pass-through-libs}
9941 The @code{pass-through-libs} spec function takes any number of arguments. It
9942 finds any @option{-l} options and any non-options ending in ".a" (which it
9943 assumes are the names of linker input library archive files) and returns a
9944 result containing all the found arguments each prepended by
9945 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9946 intended to be passed to the LTO linker plugin.
9949 %:pass-through-libs(%G %L %G)
9952 @item @code{print-asm-header}
9953 The @code{print-asm-header} function takes no arguments and simply
9954 prints a banner like:
9960 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9963 It is used to separate compiler options from assembler options
9964 in the @option{--target-help} output.
9968 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9969 If that switch was not specified, this substitutes nothing. Note that
9970 the leading dash is omitted when specifying this option, and it is
9971 automatically inserted if the substitution is performed. Thus the spec
9972 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9973 and would output the command line option @option{-foo}.
9975 @item %W@{@code{S}@}
9976 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9979 @item %@{@code{S}*@}
9980 Substitutes all the switches specified to GCC whose names start
9981 with @code{-S}, but which also take an argument. This is used for
9982 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9983 GCC considers @option{-o foo} as being
9984 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9985 text, including the space. Thus two arguments would be generated.
9987 @item %@{@code{S}*&@code{T}*@}
9988 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9989 (the order of @code{S} and @code{T} in the spec is not significant).
9990 There can be any number of ampersand-separated variables; for each the
9991 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9993 @item %@{@code{S}:@code{X}@}
9994 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9996 @item %@{!@code{S}:@code{X}@}
9997 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9999 @item %@{@code{S}*:@code{X}@}
10000 Substitutes @code{X} if one or more switches whose names start with
10001 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10002 once, no matter how many such switches appeared. However, if @code{%*}
10003 appears somewhere in @code{X}, then @code{X} will be substituted once
10004 for each matching switch, with the @code{%*} replaced by the part of
10005 that switch that matched the @code{*}.
10007 @item %@{.@code{S}:@code{X}@}
10008 Substitutes @code{X}, if processing a file with suffix @code{S}.
10010 @item %@{!.@code{S}:@code{X}@}
10011 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10013 @item %@{,@code{S}:@code{X}@}
10014 Substitutes @code{X}, if processing a file for language @code{S}.
10016 @item %@{!,@code{S}:@code{X}@}
10017 Substitutes @code{X}, if not processing a file for language @code{S}.
10019 @item %@{@code{S}|@code{P}:@code{X}@}
10020 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10021 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10022 @code{*} sequences as well, although they have a stronger binding than
10023 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10024 alternatives must be starred, and only the first matching alternative
10027 For example, a spec string like this:
10030 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10033 will output the following command-line options from the following input
10034 command-line options:
10039 -d fred.c -foo -baz -boggle
10040 -d jim.d -bar -baz -boggle
10043 @item %@{S:X; T:Y; :D@}
10045 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10046 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10047 be as many clauses as you need. This may be combined with @code{.},
10048 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10053 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10054 construct may contain other nested @samp{%} constructs or spaces, or
10055 even newlines. They are processed as usual, as described above.
10056 Trailing white space in @code{X} is ignored. White space may also
10057 appear anywhere on the left side of the colon in these constructs,
10058 except between @code{.} or @code{*} and the corresponding word.
10060 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10061 handled specifically in these constructs. If another value of
10062 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10063 @option{-W} switch is found later in the command line, the earlier
10064 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10065 just one letter, which passes all matching options.
10067 The character @samp{|} at the beginning of the predicate text is used to
10068 indicate that a command should be piped to the following command, but
10069 only if @option{-pipe} is specified.
10071 It is built into GCC which switches take arguments and which do not.
10072 (You might think it would be useful to generalize this to allow each
10073 compiler's spec to say which switches take arguments. But this cannot
10074 be done in a consistent fashion. GCC cannot even decide which input
10075 files have been specified without knowing which switches take arguments,
10076 and it must know which input files to compile in order to tell which
10079 GCC also knows implicitly that arguments starting in @option{-l} are to be
10080 treated as compiler output files, and passed to the linker in their
10081 proper position among the other output files.
10083 @c man begin OPTIONS
10085 @node Target Options
10086 @section Specifying Target Machine and Compiler Version
10087 @cindex target options
10088 @cindex cross compiling
10089 @cindex specifying machine version
10090 @cindex specifying compiler version and target machine
10091 @cindex compiler version, specifying
10092 @cindex target machine, specifying
10094 The usual way to run GCC is to run the executable called @command{gcc}, or
10095 @command{@var{machine}-gcc} when cross-compiling, or
10096 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10097 one that was installed last.
10099 @node Submodel Options
10100 @section Hardware Models and Configurations
10101 @cindex submodel options
10102 @cindex specifying hardware config
10103 @cindex hardware models and configurations, specifying
10104 @cindex machine dependent options
10106 Each target machine types can have its own
10107 special options, starting with @samp{-m}, to choose among various
10108 hardware models or configurations---for example, 68010 vs 68020,
10109 floating coprocessor or none. A single installed version of the
10110 compiler can compile for any model or configuration, according to the
10113 Some configurations of the compiler also support additional special
10114 options, usually for compatibility with other compilers on the same
10117 @c This list is ordered alphanumerically by subsection name.
10118 @c It should be the same order and spelling as these options are listed
10119 @c in Machine Dependent Options
10124 * Blackfin Options::
10128 * DEC Alpha Options::
10129 * DEC Alpha/VMS Options::
10132 * GNU/Linux Options::
10135 * i386 and x86-64 Options::
10136 * i386 and x86-64 Windows Options::
10138 * IA-64/VMS Options::
10145 * MicroBlaze Options::
10148 * MN10300 Options::
10150 * picoChip Options::
10151 * PowerPC Options::
10152 * RS/6000 and PowerPC Options::
10154 * S/390 and zSeries Options::
10157 * Solaris 2 Options::
10160 * System V Options::
10163 * VxWorks Options::
10165 * Xstormy16 Options::
10167 * zSeries Options::
10171 @subsection ARM Options
10172 @cindex ARM options
10174 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10178 @item -mabi=@var{name}
10180 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10181 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10184 @opindex mapcs-frame
10185 Generate a stack frame that is compliant with the ARM Procedure Call
10186 Standard for all functions, even if this is not strictly necessary for
10187 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10188 with this option will cause the stack frames not to be generated for
10189 leaf functions. The default is @option{-mno-apcs-frame}.
10193 This is a synonym for @option{-mapcs-frame}.
10196 @c not currently implemented
10197 @item -mapcs-stack-check
10198 @opindex mapcs-stack-check
10199 Generate code to check the amount of stack space available upon entry to
10200 every function (that actually uses some stack space). If there is
10201 insufficient space available then either the function
10202 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10203 called, depending upon the amount of stack space required. The run time
10204 system is required to provide these functions. The default is
10205 @option{-mno-apcs-stack-check}, since this produces smaller code.
10207 @c not currently implemented
10209 @opindex mapcs-float
10210 Pass floating point arguments using the float point registers. This is
10211 one of the variants of the APCS@. This option is recommended if the
10212 target hardware has a floating point unit or if a lot of floating point
10213 arithmetic is going to be performed by the code. The default is
10214 @option{-mno-apcs-float}, since integer only code is slightly increased in
10215 size if @option{-mapcs-float} is used.
10217 @c not currently implemented
10218 @item -mapcs-reentrant
10219 @opindex mapcs-reentrant
10220 Generate reentrant, position independent code. The default is
10221 @option{-mno-apcs-reentrant}.
10224 @item -mthumb-interwork
10225 @opindex mthumb-interwork
10226 Generate code which supports calling between the ARM and Thumb
10227 instruction sets. Without this option, on pre-v5 architectures, the
10228 two instruction sets cannot be reliably used inside one program. The
10229 default is @option{-mno-thumb-interwork}, since slightly larger code
10230 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10231 configurations this option is meaningless.
10233 @item -mno-sched-prolog
10234 @opindex mno-sched-prolog
10235 Prevent the reordering of instructions in the function prolog, or the
10236 merging of those instruction with the instructions in the function's
10237 body. This means that all functions will start with a recognizable set
10238 of instructions (or in fact one of a choice from a small set of
10239 different function prologues), and this information can be used to
10240 locate the start if functions inside an executable piece of code. The
10241 default is @option{-msched-prolog}.
10243 @item -mfloat-abi=@var{name}
10244 @opindex mfloat-abi
10245 Specifies which floating-point ABI to use. Permissible values
10246 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10248 Specifying @samp{soft} causes GCC to generate output containing
10249 library calls for floating-point operations.
10250 @samp{softfp} allows the generation of code using hardware floating-point
10251 instructions, but still uses the soft-float calling conventions.
10252 @samp{hard} allows generation of floating-point instructions
10253 and uses FPU-specific calling conventions.
10255 The default depends on the specific target configuration. Note that
10256 the hard-float and soft-float ABIs are not link-compatible; you must
10257 compile your entire program with the same ABI, and link with a
10258 compatible set of libraries.
10260 @item -mlittle-endian
10261 @opindex mlittle-endian
10262 Generate code for a processor running in little-endian mode. This is
10263 the default for all standard configurations.
10266 @opindex mbig-endian
10267 Generate code for a processor running in big-endian mode; the default is
10268 to compile code for a little-endian processor.
10270 @item -mwords-little-endian
10271 @opindex mwords-little-endian
10272 This option only applies when generating code for big-endian processors.
10273 Generate code for a little-endian word order but a big-endian byte
10274 order. That is, a byte order of the form @samp{32107654}. Note: this
10275 option should only be used if you require compatibility with code for
10276 big-endian ARM processors generated by versions of the compiler prior to
10277 2.8. This option is now deprecated.
10279 @item -mcpu=@var{name}
10281 This specifies the name of the target ARM processor. GCC uses this name
10282 to determine what kind of instructions it can emit when generating
10283 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10284 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10285 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10286 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10287 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10289 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10290 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10291 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10292 @samp{strongarm1110},
10293 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10294 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10295 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10296 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10297 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10298 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10299 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10300 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10301 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10302 @samp{cortex-m4}, @samp{cortex-m3},
10305 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10306 @samp{fa526}, @samp{fa626},
10307 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10309 @item -mtune=@var{name}
10311 This option is very similar to the @option{-mcpu=} option, except that
10312 instead of specifying the actual target processor type, and hence
10313 restricting which instructions can be used, it specifies that GCC should
10314 tune the performance of the code as if the target were of the type
10315 specified in this option, but still choosing the instructions that it
10316 will generate based on the CPU specified by a @option{-mcpu=} option.
10317 For some ARM implementations better performance can be obtained by using
10320 @item -march=@var{name}
10322 This specifies the name of the target ARM architecture. GCC uses this
10323 name to determine what kind of instructions it can emit when generating
10324 assembly code. This option can be used in conjunction with or instead
10325 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10326 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10327 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10328 @samp{armv6}, @samp{armv6j},
10329 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10330 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10331 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10333 @item -mfpu=@var{name}
10334 @itemx -mfpe=@var{number}
10335 @itemx -mfp=@var{number}
10339 This specifies what floating point hardware (or hardware emulation) is
10340 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10341 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10342 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10343 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10344 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10345 @option{-mfp} and @option{-mfpe} are synonyms for
10346 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10349 If @option{-msoft-float} is specified this specifies the format of
10350 floating point values.
10352 If the selected floating-point hardware includes the NEON extension
10353 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10354 operations will not be used by GCC's auto-vectorization pass unless
10355 @option{-funsafe-math-optimizations} is also specified. This is
10356 because NEON hardware does not fully implement the IEEE 754 standard for
10357 floating-point arithmetic (in particular denormal values are treated as
10358 zero), so the use of NEON instructions may lead to a loss of precision.
10360 @item -mfp16-format=@var{name}
10361 @opindex mfp16-format
10362 Specify the format of the @code{__fp16} half-precision floating-point type.
10363 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10364 the default is @samp{none}, in which case the @code{__fp16} type is not
10365 defined. @xref{Half-Precision}, for more information.
10367 @item -mstructure-size-boundary=@var{n}
10368 @opindex mstructure-size-boundary
10369 The size of all structures and unions will be rounded up to a multiple
10370 of the number of bits set by this option. Permissible values are 8, 32
10371 and 64. The default value varies for different toolchains. For the COFF
10372 targeted toolchain the default value is 8. A value of 64 is only allowed
10373 if the underlying ABI supports it.
10375 Specifying the larger number can produce faster, more efficient code, but
10376 can also increase the size of the program. Different values are potentially
10377 incompatible. Code compiled with one value cannot necessarily expect to
10378 work with code or libraries compiled with another value, if they exchange
10379 information using structures or unions.
10381 @item -mabort-on-noreturn
10382 @opindex mabort-on-noreturn
10383 Generate a call to the function @code{abort} at the end of a
10384 @code{noreturn} function. It will be executed if the function tries to
10388 @itemx -mno-long-calls
10389 @opindex mlong-calls
10390 @opindex mno-long-calls
10391 Tells the compiler to perform function calls by first loading the
10392 address of the function into a register and then performing a subroutine
10393 call on this register. This switch is needed if the target function
10394 will lie outside of the 64 megabyte addressing range of the offset based
10395 version of subroutine call instruction.
10397 Even if this switch is enabled, not all function calls will be turned
10398 into long calls. The heuristic is that static functions, functions
10399 which have the @samp{short-call} attribute, functions that are inside
10400 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10401 definitions have already been compiled within the current compilation
10402 unit, will not be turned into long calls. The exception to this rule is
10403 that weak function definitions, functions with the @samp{long-call}
10404 attribute or the @samp{section} attribute, and functions that are within
10405 the scope of a @samp{#pragma long_calls} directive, will always be
10406 turned into long calls.
10408 This feature is not enabled by default. Specifying
10409 @option{-mno-long-calls} will restore the default behavior, as will
10410 placing the function calls within the scope of a @samp{#pragma
10411 long_calls_off} directive. Note these switches have no effect on how
10412 the compiler generates code to handle function calls via function
10415 @item -msingle-pic-base
10416 @opindex msingle-pic-base
10417 Treat the register used for PIC addressing as read-only, rather than
10418 loading it in the prologue for each function. The run-time system is
10419 responsible for initializing this register with an appropriate value
10420 before execution begins.
10422 @item -mpic-register=@var{reg}
10423 @opindex mpic-register
10424 Specify the register to be used for PIC addressing. The default is R10
10425 unless stack-checking is enabled, when R9 is used.
10427 @item -mcirrus-fix-invalid-insns
10428 @opindex mcirrus-fix-invalid-insns
10429 @opindex mno-cirrus-fix-invalid-insns
10430 Insert NOPs into the instruction stream to in order to work around
10431 problems with invalid Maverick instruction combinations. This option
10432 is only valid if the @option{-mcpu=ep9312} option has been used to
10433 enable generation of instructions for the Cirrus Maverick floating
10434 point co-processor. This option is not enabled by default, since the
10435 problem is only present in older Maverick implementations. The default
10436 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10439 @item -mpoke-function-name
10440 @opindex mpoke-function-name
10441 Write the name of each function into the text section, directly
10442 preceding the function prologue. The generated code is similar to this:
10446 .ascii "arm_poke_function_name", 0
10449 .word 0xff000000 + (t1 - t0)
10450 arm_poke_function_name
10452 stmfd sp!, @{fp, ip, lr, pc@}
10456 When performing a stack backtrace, code can inspect the value of
10457 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10458 location @code{pc - 12} and the top 8 bits are set, then we know that
10459 there is a function name embedded immediately preceding this location
10460 and has length @code{((pc[-3]) & 0xff000000)}.
10467 Select between generating code that executes in ARM and Thumb
10468 states. The default for most configurations is to generate code
10469 that executes in ARM state, but the default can be changed by
10470 configuring GCC with the @option{--with-mode=}@var{state}
10474 @opindex mtpcs-frame
10475 Generate a stack frame that is compliant with the Thumb Procedure Call
10476 Standard for all non-leaf functions. (A leaf function is one that does
10477 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10479 @item -mtpcs-leaf-frame
10480 @opindex mtpcs-leaf-frame
10481 Generate a stack frame that is compliant with the Thumb Procedure Call
10482 Standard for all leaf functions. (A leaf function is one that does
10483 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10485 @item -mcallee-super-interworking
10486 @opindex mcallee-super-interworking
10487 Gives all externally visible functions in the file being compiled an ARM
10488 instruction set header which switches to Thumb mode before executing the
10489 rest of the function. This allows these functions to be called from
10490 non-interworking code. This option is not valid in AAPCS configurations
10491 because interworking is enabled by default.
10493 @item -mcaller-super-interworking
10494 @opindex mcaller-super-interworking
10495 Allows calls via function pointers (including virtual functions) to
10496 execute correctly regardless of whether the target code has been
10497 compiled for interworking or not. There is a small overhead in the cost
10498 of executing a function pointer if this option is enabled. This option
10499 is not valid in AAPCS configurations because interworking is enabled
10502 @item -mtp=@var{name}
10504 Specify the access model for the thread local storage pointer. The valid
10505 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10506 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10507 (supported in the arm6k architecture), and @option{auto}, which uses the
10508 best available method for the selected processor. The default setting is
10511 @item -mtls-dialect=@var{dialect}
10512 @opindex mtls-dialect
10513 Specify the dialect to use for accessing thread local storage. Two
10514 dialects are supported --- @option{gnu} and @option{gnu2}. The
10515 @option{gnu} dialect selects the original GNU scheme for supporting
10516 local and global dynamic TLS models. The @option{gnu2} dialect
10517 selects the GNU descriptor scheme, which provides better performance
10518 for shared libraries. The GNU descriptor scheme is compatible with
10519 the original scheme, but does require new assembler, linker and
10520 library support. Initial and local exec TLS models are unaffected by
10521 this option and always use the original scheme.
10523 @item -mword-relocations
10524 @opindex mword-relocations
10525 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10526 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10527 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10530 @item -mfix-cortex-m3-ldrd
10531 @opindex mfix-cortex-m3-ldrd
10532 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10533 with overlapping destination and base registers are used. This option avoids
10534 generating these instructions. This option is enabled by default when
10535 @option{-mcpu=cortex-m3} is specified.
10540 @subsection AVR Options
10541 @cindex AVR Options
10543 These options are defined for AVR implementations:
10546 @item -mmcu=@var{mcu}
10548 Specify ATMEL AVR instruction set or MCU type.
10550 Instruction set avr1 is for the minimal AVR core, not supported by the C
10551 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10552 attiny11, attiny12, attiny15, attiny28).
10554 Instruction set avr2 (default) is for the classic AVR core with up to
10555 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10556 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10557 at90c8534, at90s8535).
10559 Instruction set avr3 is for the classic AVR core with up to 128K program
10560 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10562 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10563 memory space (MCU types: atmega8, atmega83, atmega85).
10565 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10566 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10567 atmega64, atmega128, at43usb355, at94k).
10569 @item -mno-interrupts
10570 @opindex mno-interrupts
10571 Generated code is not compatible with hardware interrupts.
10572 Code size will be smaller.
10574 @item -mcall-prologues
10575 @opindex mcall-prologues
10576 Functions prologues/epilogues expanded as call to appropriate
10577 subroutines. Code size will be smaller.
10580 @opindex mtiny-stack
10581 Change only the low 8 bits of the stack pointer.
10585 Assume int to be 8 bit integer. This affects the sizes of all types: A
10586 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10587 and long long will be 4 bytes. Please note that this option does not
10588 comply to the C standards, but it will provide you with smaller code
10592 @node Blackfin Options
10593 @subsection Blackfin Options
10594 @cindex Blackfin Options
10597 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10599 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10600 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10601 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10602 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10603 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10604 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10605 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10606 @samp{bf561}, @samp{bf592}.
10607 The optional @var{sirevision} specifies the silicon revision of the target
10608 Blackfin processor. Any workarounds available for the targeted silicon revision
10609 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10610 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10611 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10612 hexadecimal digits representing the major and minor numbers in the silicon
10613 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10614 is not defined. If @var{sirevision} is @samp{any}, the
10615 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10616 If this optional @var{sirevision} is not used, GCC assumes the latest known
10617 silicon revision of the targeted Blackfin processor.
10619 Support for @samp{bf561} is incomplete. For @samp{bf561},
10620 Only the processor macro is defined.
10621 Without this option, @samp{bf532} is used as the processor by default.
10622 The corresponding predefined processor macros for @var{cpu} is to
10623 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10624 provided by libgloss to be linked in if @option{-msim} is not given.
10628 Specifies that the program will be run on the simulator. This causes
10629 the simulator BSP provided by libgloss to be linked in. This option
10630 has effect only for @samp{bfin-elf} toolchain.
10631 Certain other options, such as @option{-mid-shared-library} and
10632 @option{-mfdpic}, imply @option{-msim}.
10634 @item -momit-leaf-frame-pointer
10635 @opindex momit-leaf-frame-pointer
10636 Don't keep the frame pointer in a register for leaf functions. This
10637 avoids the instructions to save, set up and restore frame pointers and
10638 makes an extra register available in leaf functions. The option
10639 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10640 which might make debugging harder.
10642 @item -mspecld-anomaly
10643 @opindex mspecld-anomaly
10644 When enabled, the compiler will ensure that the generated code does not
10645 contain speculative loads after jump instructions. If this option is used,
10646 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10648 @item -mno-specld-anomaly
10649 @opindex mno-specld-anomaly
10650 Don't generate extra code to prevent speculative loads from occurring.
10652 @item -mcsync-anomaly
10653 @opindex mcsync-anomaly
10654 When enabled, the compiler will ensure that the generated code does not
10655 contain CSYNC or SSYNC instructions too soon after conditional branches.
10656 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10658 @item -mno-csync-anomaly
10659 @opindex mno-csync-anomaly
10660 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10661 occurring too soon after a conditional branch.
10665 When enabled, the compiler is free to take advantage of the knowledge that
10666 the entire program fits into the low 64k of memory.
10669 @opindex mno-low-64k
10670 Assume that the program is arbitrarily large. This is the default.
10672 @item -mstack-check-l1
10673 @opindex mstack-check-l1
10674 Do stack checking using information placed into L1 scratchpad memory by the
10677 @item -mid-shared-library
10678 @opindex mid-shared-library
10679 Generate code that supports shared libraries via the library ID method.
10680 This allows for execute in place and shared libraries in an environment
10681 without virtual memory management. This option implies @option{-fPIC}.
10682 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10684 @item -mno-id-shared-library
10685 @opindex mno-id-shared-library
10686 Generate code that doesn't assume ID based shared libraries are being used.
10687 This is the default.
10689 @item -mleaf-id-shared-library
10690 @opindex mleaf-id-shared-library
10691 Generate code that supports shared libraries via the library ID method,
10692 but assumes that this library or executable won't link against any other
10693 ID shared libraries. That allows the compiler to use faster code for jumps
10696 @item -mno-leaf-id-shared-library
10697 @opindex mno-leaf-id-shared-library
10698 Do not assume that the code being compiled won't link against any ID shared
10699 libraries. Slower code will be generated for jump and call insns.
10701 @item -mshared-library-id=n
10702 @opindex mshared-library-id
10703 Specified the identification number of the ID based shared library being
10704 compiled. Specifying a value of 0 will generate more compact code, specifying
10705 other values will force the allocation of that number to the current
10706 library but is no more space or time efficient than omitting this option.
10710 Generate code that allows the data segment to be located in a different
10711 area of memory from the text segment. This allows for execute in place in
10712 an environment without virtual memory management by eliminating relocations
10713 against the text section.
10715 @item -mno-sep-data
10716 @opindex mno-sep-data
10717 Generate code that assumes that the data segment follows the text segment.
10718 This is the default.
10721 @itemx -mno-long-calls
10722 @opindex mlong-calls
10723 @opindex mno-long-calls
10724 Tells the compiler to perform function calls by first loading the
10725 address of the function into a register and then performing a subroutine
10726 call on this register. This switch is needed if the target function
10727 will lie outside of the 24 bit addressing range of the offset based
10728 version of subroutine call instruction.
10730 This feature is not enabled by default. Specifying
10731 @option{-mno-long-calls} will restore the default behavior. Note these
10732 switches have no effect on how the compiler generates code to handle
10733 function calls via function pointers.
10737 Link with the fast floating-point library. This library relaxes some of
10738 the IEEE floating-point standard's rules for checking inputs against
10739 Not-a-Number (NAN), in the interest of performance.
10742 @opindex minline-plt
10743 Enable inlining of PLT entries in function calls to functions that are
10744 not known to bind locally. It has no effect without @option{-mfdpic}.
10747 @opindex mmulticore
10748 Build standalone application for multicore Blackfin processor. Proper
10749 start files and link scripts will be used to support multicore.
10750 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10751 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10752 @option{-mcorea} or @option{-mcoreb}. If it's used without
10753 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10754 programming model is used. In this model, the main function of Core B
10755 should be named as coreb_main. If it's used with @option{-mcorea} or
10756 @option{-mcoreb}, one application per core programming model is used.
10757 If this option is not used, single core application programming
10762 Build standalone application for Core A of BF561 when using
10763 one application per core programming model. Proper start files
10764 and link scripts will be used to support Core A. This option
10765 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10769 Build standalone application for Core B of BF561 when using
10770 one application per core programming model. Proper start files
10771 and link scripts will be used to support Core B. This option
10772 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10773 should be used instead of main. It must be used with
10774 @option{-mmulticore}.
10778 Build standalone application for SDRAM. Proper start files and
10779 link scripts will be used to put the application into SDRAM.
10780 Loader should initialize SDRAM before loading the application
10781 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10785 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10786 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10787 are enabled; for standalone applications the default is off.
10791 @subsection C6X Options
10792 @cindex C6X Options
10795 @item -march=@var{name}
10797 This specifies the name of the target architecture. GCC uses this
10798 name to determine what kind of instructions it can emit when generating
10799 assembly code. Permissible names are: @samp{c62x},
10800 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
10803 @opindex mbig-endian
10804 Generate code for a big endian target.
10806 @item -mlittle-endian
10807 @opindex mlittle-endian
10808 Generate code for a little endian target. This is the default.
10812 Choose startup files and linker script suitable for the simulator.
10814 @item -msdata=default
10815 @opindex msdata=default
10816 Put small global and static data in the @samp{.neardata} section,
10817 which is pointed to by register @code{B14}. Put small uninitialized
10818 global and static data in the @samp{.bss} section, which is adjacent
10819 to the @samp{.neardata} section. Put small read-only data into the
10820 @samp{.rodata} section. The corresponding sections used for large
10821 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
10824 @opindex msdata=all
10825 Put all data, not just small objets, into the sections reserved for
10826 small data, and use addressing relative to the @code{B14} register to
10830 @opindex msdata=none
10831 Make no use of the sections reserved for small data, and use absolute
10832 addresses to access all data. Put all initialized global and static
10833 data in the @samp{.fardata} section, and all uninitialized data in the
10834 @samp{.far} section. Put all constant data into the @samp{.const}
10839 @subsection CRIS Options
10840 @cindex CRIS Options
10842 These options are defined specifically for the CRIS ports.
10845 @item -march=@var{architecture-type}
10846 @itemx -mcpu=@var{architecture-type}
10849 Generate code for the specified architecture. The choices for
10850 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10851 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10852 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10855 @item -mtune=@var{architecture-type}
10857 Tune to @var{architecture-type} everything applicable about the generated
10858 code, except for the ABI and the set of available instructions. The
10859 choices for @var{architecture-type} are the same as for
10860 @option{-march=@var{architecture-type}}.
10862 @item -mmax-stack-frame=@var{n}
10863 @opindex mmax-stack-frame
10864 Warn when the stack frame of a function exceeds @var{n} bytes.
10870 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10871 @option{-march=v3} and @option{-march=v8} respectively.
10873 @item -mmul-bug-workaround
10874 @itemx -mno-mul-bug-workaround
10875 @opindex mmul-bug-workaround
10876 @opindex mno-mul-bug-workaround
10877 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10878 models where it applies. This option is active by default.
10882 Enable CRIS-specific verbose debug-related information in the assembly
10883 code. This option also has the effect to turn off the @samp{#NO_APP}
10884 formatted-code indicator to the assembler at the beginning of the
10889 Do not use condition-code results from previous instruction; always emit
10890 compare and test instructions before use of condition codes.
10892 @item -mno-side-effects
10893 @opindex mno-side-effects
10894 Do not emit instructions with side-effects in addressing modes other than
10897 @item -mstack-align
10898 @itemx -mno-stack-align
10899 @itemx -mdata-align
10900 @itemx -mno-data-align
10901 @itemx -mconst-align
10902 @itemx -mno-const-align
10903 @opindex mstack-align
10904 @opindex mno-stack-align
10905 @opindex mdata-align
10906 @opindex mno-data-align
10907 @opindex mconst-align
10908 @opindex mno-const-align
10909 These options (no-options) arranges (eliminate arrangements) for the
10910 stack-frame, individual data and constants to be aligned for the maximum
10911 single data access size for the chosen CPU model. The default is to
10912 arrange for 32-bit alignment. ABI details such as structure layout are
10913 not affected by these options.
10921 Similar to the stack- data- and const-align options above, these options
10922 arrange for stack-frame, writable data and constants to all be 32-bit,
10923 16-bit or 8-bit aligned. The default is 32-bit alignment.
10925 @item -mno-prologue-epilogue
10926 @itemx -mprologue-epilogue
10927 @opindex mno-prologue-epilogue
10928 @opindex mprologue-epilogue
10929 With @option{-mno-prologue-epilogue}, the normal function prologue and
10930 epilogue that sets up the stack-frame are omitted and no return
10931 instructions or return sequences are generated in the code. Use this
10932 option only together with visual inspection of the compiled code: no
10933 warnings or errors are generated when call-saved registers must be saved,
10934 or storage for local variable needs to be allocated.
10938 @opindex mno-gotplt
10940 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10941 instruction sequences that load addresses for functions from the PLT part
10942 of the GOT rather than (traditional on other architectures) calls to the
10943 PLT@. The default is @option{-mgotplt}.
10947 Legacy no-op option only recognized with the cris-axis-elf and
10948 cris-axis-linux-gnu targets.
10952 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10956 This option, recognized for the cris-axis-elf arranges
10957 to link with input-output functions from a simulator library. Code,
10958 initialized data and zero-initialized data are allocated consecutively.
10962 Like @option{-sim}, but pass linker options to locate initialized data at
10963 0x40000000 and zero-initialized data at 0x80000000.
10966 @node Darwin Options
10967 @subsection Darwin Options
10968 @cindex Darwin options
10970 These options are defined for all architectures running the Darwin operating
10973 FSF GCC on Darwin does not create ``fat'' object files; it will create
10974 an object file for the single architecture that it was built to
10975 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10976 @option{-arch} options are used; it does so by running the compiler or
10977 linker multiple times and joining the results together with
10980 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10981 @samp{i686}) is determined by the flags that specify the ISA
10982 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10983 @option{-force_cpusubtype_ALL} option can be used to override this.
10985 The Darwin tools vary in their behavior when presented with an ISA
10986 mismatch. The assembler, @file{as}, will only permit instructions to
10987 be used that are valid for the subtype of the file it is generating,
10988 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10989 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10990 and print an error if asked to create a shared library with a less
10991 restrictive subtype than its input files (for instance, trying to put
10992 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10993 for executables, @file{ld}, will quietly give the executable the most
10994 restrictive subtype of any of its input files.
10999 Add the framework directory @var{dir} to the head of the list of
11000 directories to be searched for header files. These directories are
11001 interleaved with those specified by @option{-I} options and are
11002 scanned in a left-to-right order.
11004 A framework directory is a directory with frameworks in it. A
11005 framework is a directory with a @samp{"Headers"} and/or
11006 @samp{"PrivateHeaders"} directory contained directly in it that ends
11007 in @samp{".framework"}. The name of a framework is the name of this
11008 directory excluding the @samp{".framework"}. Headers associated with
11009 the framework are found in one of those two directories, with
11010 @samp{"Headers"} being searched first. A subframework is a framework
11011 directory that is in a framework's @samp{"Frameworks"} directory.
11012 Includes of subframework headers can only appear in a header of a
11013 framework that contains the subframework, or in a sibling subframework
11014 header. Two subframeworks are siblings if they occur in the same
11015 framework. A subframework should not have the same name as a
11016 framework, a warning will be issued if this is violated. Currently a
11017 subframework cannot have subframeworks, in the future, the mechanism
11018 may be extended to support this. The standard frameworks can be found
11019 in @samp{"/System/Library/Frameworks"} and
11020 @samp{"/Library/Frameworks"}. An example include looks like
11021 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11022 the name of the framework and header.h is found in the
11023 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11025 @item -iframework@var{dir}
11026 @opindex iframework
11027 Like @option{-F} except the directory is a treated as a system
11028 directory. The main difference between this @option{-iframework} and
11029 @option{-F} is that with @option{-iframework} the compiler does not
11030 warn about constructs contained within header files found via
11031 @var{dir}. This option is valid only for the C family of languages.
11035 Emit debugging information for symbols that are used. For STABS
11036 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11037 This is by default ON@.
11041 Emit debugging information for all symbols and types.
11043 @item -mmacosx-version-min=@var{version}
11044 The earliest version of MacOS X that this executable will run on
11045 is @var{version}. Typical values of @var{version} include @code{10.1},
11046 @code{10.2}, and @code{10.3.9}.
11048 If the compiler was built to use the system's headers by default,
11049 then the default for this option is the system version on which the
11050 compiler is running, otherwise the default is to make choices which
11051 are compatible with as many systems and code bases as possible.
11055 Enable kernel development mode. The @option{-mkernel} option sets
11056 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11057 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11058 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11059 applicable. This mode also sets @option{-mno-altivec},
11060 @option{-msoft-float}, @option{-fno-builtin} and
11061 @option{-mlong-branch} for PowerPC targets.
11063 @item -mone-byte-bool
11064 @opindex mone-byte-bool
11065 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11066 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11067 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11068 option has no effect on x86.
11070 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11071 to generate code that is not binary compatible with code generated
11072 without that switch. Using this switch may require recompiling all
11073 other modules in a program, including system libraries. Use this
11074 switch to conform to a non-default data model.
11076 @item -mfix-and-continue
11077 @itemx -ffix-and-continue
11078 @itemx -findirect-data
11079 @opindex mfix-and-continue
11080 @opindex ffix-and-continue
11081 @opindex findirect-data
11082 Generate code suitable for fast turn around development. Needed to
11083 enable gdb to dynamically load @code{.o} files into already running
11084 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11085 are provided for backwards compatibility.
11089 Loads all members of static archive libraries.
11090 See man ld(1) for more information.
11092 @item -arch_errors_fatal
11093 @opindex arch_errors_fatal
11094 Cause the errors having to do with files that have the wrong architecture
11097 @item -bind_at_load
11098 @opindex bind_at_load
11099 Causes the output file to be marked such that the dynamic linker will
11100 bind all undefined references when the file is loaded or launched.
11104 Produce a Mach-o bundle format file.
11105 See man ld(1) for more information.
11107 @item -bundle_loader @var{executable}
11108 @opindex bundle_loader
11109 This option specifies the @var{executable} that will be loading the build
11110 output file being linked. See man ld(1) for more information.
11113 @opindex dynamiclib
11114 When passed this option, GCC will produce a dynamic library instead of
11115 an executable when linking, using the Darwin @file{libtool} command.
11117 @item -force_cpusubtype_ALL
11118 @opindex force_cpusubtype_ALL
11119 This causes GCC's output file to have the @var{ALL} subtype, instead of
11120 one controlled by the @option{-mcpu} or @option{-march} option.
11122 @item -allowable_client @var{client_name}
11123 @itemx -client_name
11124 @itemx -compatibility_version
11125 @itemx -current_version
11127 @itemx -dependency-file
11129 @itemx -dylinker_install_name
11131 @itemx -exported_symbols_list
11134 @itemx -flat_namespace
11135 @itemx -force_flat_namespace
11136 @itemx -headerpad_max_install_names
11139 @itemx -install_name
11140 @itemx -keep_private_externs
11141 @itemx -multi_module
11142 @itemx -multiply_defined
11143 @itemx -multiply_defined_unused
11146 @itemx -no_dead_strip_inits_and_terms
11147 @itemx -nofixprebinding
11148 @itemx -nomultidefs
11150 @itemx -noseglinkedit
11151 @itemx -pagezero_size
11153 @itemx -prebind_all_twolevel_modules
11154 @itemx -private_bundle
11156 @itemx -read_only_relocs
11158 @itemx -sectobjectsymbols
11162 @itemx -sectobjectsymbols
11165 @itemx -segs_read_only_addr
11167 @itemx -segs_read_write_addr
11168 @itemx -seg_addr_table
11169 @itemx -seg_addr_table_filename
11170 @itemx -seglinkedit
11172 @itemx -segs_read_only_addr
11173 @itemx -segs_read_write_addr
11174 @itemx -single_module
11176 @itemx -sub_library
11178 @itemx -sub_umbrella
11179 @itemx -twolevel_namespace
11182 @itemx -unexported_symbols_list
11183 @itemx -weak_reference_mismatches
11184 @itemx -whatsloaded
11185 @opindex allowable_client
11186 @opindex client_name
11187 @opindex compatibility_version
11188 @opindex current_version
11189 @opindex dead_strip
11190 @opindex dependency-file
11191 @opindex dylib_file
11192 @opindex dylinker_install_name
11194 @opindex exported_symbols_list
11196 @opindex flat_namespace
11197 @opindex force_flat_namespace
11198 @opindex headerpad_max_install_names
11199 @opindex image_base
11201 @opindex install_name
11202 @opindex keep_private_externs
11203 @opindex multi_module
11204 @opindex multiply_defined
11205 @opindex multiply_defined_unused
11206 @opindex noall_load
11207 @opindex no_dead_strip_inits_and_terms
11208 @opindex nofixprebinding
11209 @opindex nomultidefs
11211 @opindex noseglinkedit
11212 @opindex pagezero_size
11214 @opindex prebind_all_twolevel_modules
11215 @opindex private_bundle
11216 @opindex read_only_relocs
11218 @opindex sectobjectsymbols
11221 @opindex sectcreate
11222 @opindex sectobjectsymbols
11225 @opindex segs_read_only_addr
11226 @opindex segs_read_write_addr
11227 @opindex seg_addr_table
11228 @opindex seg_addr_table_filename
11229 @opindex seglinkedit
11231 @opindex segs_read_only_addr
11232 @opindex segs_read_write_addr
11233 @opindex single_module
11235 @opindex sub_library
11236 @opindex sub_umbrella
11237 @opindex twolevel_namespace
11240 @opindex unexported_symbols_list
11241 @opindex weak_reference_mismatches
11242 @opindex whatsloaded
11243 These options are passed to the Darwin linker. The Darwin linker man page
11244 describes them in detail.
11247 @node DEC Alpha Options
11248 @subsection DEC Alpha Options
11250 These @samp{-m} options are defined for the DEC Alpha implementations:
11253 @item -mno-soft-float
11254 @itemx -msoft-float
11255 @opindex mno-soft-float
11256 @opindex msoft-float
11257 Use (do not use) the hardware floating-point instructions for
11258 floating-point operations. When @option{-msoft-float} is specified,
11259 functions in @file{libgcc.a} will be used to perform floating-point
11260 operations. Unless they are replaced by routines that emulate the
11261 floating-point operations, or compiled in such a way as to call such
11262 emulations routines, these routines will issue floating-point
11263 operations. If you are compiling for an Alpha without floating-point
11264 operations, you must ensure that the library is built so as not to call
11267 Note that Alpha implementations without floating-point operations are
11268 required to have floating-point registers.
11271 @itemx -mno-fp-regs
11273 @opindex mno-fp-regs
11274 Generate code that uses (does not use) the floating-point register set.
11275 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11276 register set is not used, floating point operands are passed in integer
11277 registers as if they were integers and floating-point results are passed
11278 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11279 so any function with a floating-point argument or return value called by code
11280 compiled with @option{-mno-fp-regs} must also be compiled with that
11283 A typical use of this option is building a kernel that does not use,
11284 and hence need not save and restore, any floating-point registers.
11288 The Alpha architecture implements floating-point hardware optimized for
11289 maximum performance. It is mostly compliant with the IEEE floating
11290 point standard. However, for full compliance, software assistance is
11291 required. This option generates code fully IEEE compliant code
11292 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11293 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11294 defined during compilation. The resulting code is less efficient but is
11295 able to correctly support denormalized numbers and exceptional IEEE
11296 values such as not-a-number and plus/minus infinity. Other Alpha
11297 compilers call this option @option{-ieee_with_no_inexact}.
11299 @item -mieee-with-inexact
11300 @opindex mieee-with-inexact
11301 This is like @option{-mieee} except the generated code also maintains
11302 the IEEE @var{inexact-flag}. Turning on this option causes the
11303 generated code to implement fully-compliant IEEE math. In addition to
11304 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11305 macro. On some Alpha implementations the resulting code may execute
11306 significantly slower than the code generated by default. Since there is
11307 very little code that depends on the @var{inexact-flag}, you should
11308 normally not specify this option. Other Alpha compilers call this
11309 option @option{-ieee_with_inexact}.
11311 @item -mfp-trap-mode=@var{trap-mode}
11312 @opindex mfp-trap-mode
11313 This option controls what floating-point related traps are enabled.
11314 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11315 The trap mode can be set to one of four values:
11319 This is the default (normal) setting. The only traps that are enabled
11320 are the ones that cannot be disabled in software (e.g., division by zero
11324 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11328 Like @samp{u}, but the instructions are marked to be safe for software
11329 completion (see Alpha architecture manual for details).
11332 Like @samp{su}, but inexact traps are enabled as well.
11335 @item -mfp-rounding-mode=@var{rounding-mode}
11336 @opindex mfp-rounding-mode
11337 Selects the IEEE rounding mode. Other Alpha compilers call this option
11338 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11343 Normal IEEE rounding mode. Floating point numbers are rounded towards
11344 the nearest machine number or towards the even machine number in case
11348 Round towards minus infinity.
11351 Chopped rounding mode. Floating point numbers are rounded towards zero.
11354 Dynamic rounding mode. A field in the floating point control register
11355 (@var{fpcr}, see Alpha architecture reference manual) controls the
11356 rounding mode in effect. The C library initializes this register for
11357 rounding towards plus infinity. Thus, unless your program modifies the
11358 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11361 @item -mtrap-precision=@var{trap-precision}
11362 @opindex mtrap-precision
11363 In the Alpha architecture, floating point traps are imprecise. This
11364 means without software assistance it is impossible to recover from a
11365 floating trap and program execution normally needs to be terminated.
11366 GCC can generate code that can assist operating system trap handlers
11367 in determining the exact location that caused a floating point trap.
11368 Depending on the requirements of an application, different levels of
11369 precisions can be selected:
11373 Program precision. This option is the default and means a trap handler
11374 can only identify which program caused a floating point exception.
11377 Function precision. The trap handler can determine the function that
11378 caused a floating point exception.
11381 Instruction precision. The trap handler can determine the exact
11382 instruction that caused a floating point exception.
11385 Other Alpha compilers provide the equivalent options called
11386 @option{-scope_safe} and @option{-resumption_safe}.
11388 @item -mieee-conformant
11389 @opindex mieee-conformant
11390 This option marks the generated code as IEEE conformant. You must not
11391 use this option unless you also specify @option{-mtrap-precision=i} and either
11392 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11393 is to emit the line @samp{.eflag 48} in the function prologue of the
11394 generated assembly file. Under DEC Unix, this has the effect that
11395 IEEE-conformant math library routines will be linked in.
11397 @item -mbuild-constants
11398 @opindex mbuild-constants
11399 Normally GCC examines a 32- or 64-bit integer constant to
11400 see if it can construct it from smaller constants in two or three
11401 instructions. If it cannot, it will output the constant as a literal and
11402 generate code to load it from the data segment at runtime.
11404 Use this option to require GCC to construct @emph{all} integer constants
11405 using code, even if it takes more instructions (the maximum is six).
11407 You would typically use this option to build a shared library dynamic
11408 loader. Itself a shared library, it must relocate itself in memory
11409 before it can find the variables and constants in its own data segment.
11415 Select whether to generate code to be assembled by the vendor-supplied
11416 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11434 Indicate whether GCC should generate code to use the optional BWX,
11435 CIX, FIX and MAX instruction sets. The default is to use the instruction
11436 sets supported by the CPU type specified via @option{-mcpu=} option or that
11437 of the CPU on which GCC was built if none was specified.
11440 @itemx -mfloat-ieee
11441 @opindex mfloat-vax
11442 @opindex mfloat-ieee
11443 Generate code that uses (does not use) VAX F and G floating point
11444 arithmetic instead of IEEE single and double precision.
11446 @item -mexplicit-relocs
11447 @itemx -mno-explicit-relocs
11448 @opindex mexplicit-relocs
11449 @opindex mno-explicit-relocs
11450 Older Alpha assemblers provided no way to generate symbol relocations
11451 except via assembler macros. Use of these macros does not allow
11452 optimal instruction scheduling. GNU binutils as of version 2.12
11453 supports a new syntax that allows the compiler to explicitly mark
11454 which relocations should apply to which instructions. This option
11455 is mostly useful for debugging, as GCC detects the capabilities of
11456 the assembler when it is built and sets the default accordingly.
11459 @itemx -mlarge-data
11460 @opindex msmall-data
11461 @opindex mlarge-data
11462 When @option{-mexplicit-relocs} is in effect, static data is
11463 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11464 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11465 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11466 16-bit relocations off of the @code{$gp} register. This limits the
11467 size of the small data area to 64KB, but allows the variables to be
11468 directly accessed via a single instruction.
11470 The default is @option{-mlarge-data}. With this option the data area
11471 is limited to just below 2GB@. Programs that require more than 2GB of
11472 data must use @code{malloc} or @code{mmap} to allocate the data in the
11473 heap instead of in the program's data segment.
11475 When generating code for shared libraries, @option{-fpic} implies
11476 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11479 @itemx -mlarge-text
11480 @opindex msmall-text
11481 @opindex mlarge-text
11482 When @option{-msmall-text} is used, the compiler assumes that the
11483 code of the entire program (or shared library) fits in 4MB, and is
11484 thus reachable with a branch instruction. When @option{-msmall-data}
11485 is used, the compiler can assume that all local symbols share the
11486 same @code{$gp} value, and thus reduce the number of instructions
11487 required for a function call from 4 to 1.
11489 The default is @option{-mlarge-text}.
11491 @item -mcpu=@var{cpu_type}
11493 Set the instruction set and instruction scheduling parameters for
11494 machine type @var{cpu_type}. You can specify either the @samp{EV}
11495 style name or the corresponding chip number. GCC supports scheduling
11496 parameters for the EV4, EV5 and EV6 family of processors and will
11497 choose the default values for the instruction set from the processor
11498 you specify. If you do not specify a processor type, GCC will default
11499 to the processor on which the compiler was built.
11501 Supported values for @var{cpu_type} are
11507 Schedules as an EV4 and has no instruction set extensions.
11511 Schedules as an EV5 and has no instruction set extensions.
11515 Schedules as an EV5 and supports the BWX extension.
11520 Schedules as an EV5 and supports the BWX and MAX extensions.
11524 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11528 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11531 Native Linux/GNU toolchains also support the value @samp{native},
11532 which selects the best architecture option for the host processor.
11533 @option{-mcpu=native} has no effect if GCC does not recognize
11536 @item -mtune=@var{cpu_type}
11538 Set only the instruction scheduling parameters for machine type
11539 @var{cpu_type}. The instruction set is not changed.
11541 Native Linux/GNU toolchains also support the value @samp{native},
11542 which selects the best architecture option for the host processor.
11543 @option{-mtune=native} has no effect if GCC does not recognize
11546 @item -mmemory-latency=@var{time}
11547 @opindex mmemory-latency
11548 Sets the latency the scheduler should assume for typical memory
11549 references as seen by the application. This number is highly
11550 dependent on the memory access patterns used by the application
11551 and the size of the external cache on the machine.
11553 Valid options for @var{time} are
11557 A decimal number representing clock cycles.
11563 The compiler contains estimates of the number of clock cycles for
11564 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11565 (also called Dcache, Scache, and Bcache), as well as to main memory.
11566 Note that L3 is only valid for EV5.
11571 @node DEC Alpha/VMS Options
11572 @subsection DEC Alpha/VMS Options
11574 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11577 @item -mvms-return-codes
11578 @opindex mvms-return-codes
11579 Return VMS condition codes from main. The default is to return POSIX
11580 style condition (e.g.@: error) codes.
11582 @item -mdebug-main=@var{prefix}
11583 @opindex mdebug-main=@var{prefix}
11584 Flag the first routine whose name starts with @var{prefix} as the main
11585 routine for the debugger.
11589 Default to 64bit memory allocation routines.
11593 @subsection FR30 Options
11594 @cindex FR30 Options
11596 These options are defined specifically for the FR30 port.
11600 @item -msmall-model
11601 @opindex msmall-model
11602 Use the small address space model. This can produce smaller code, but
11603 it does assume that all symbolic values and addresses will fit into a
11608 Assume that run-time support has been provided and so there is no need
11609 to include the simulator library (@file{libsim.a}) on the linker
11615 @subsection FRV Options
11616 @cindex FRV Options
11622 Only use the first 32 general purpose registers.
11627 Use all 64 general purpose registers.
11632 Use only the first 32 floating point registers.
11637 Use all 64 floating point registers
11640 @opindex mhard-float
11642 Use hardware instructions for floating point operations.
11645 @opindex msoft-float
11647 Use library routines for floating point operations.
11652 Dynamically allocate condition code registers.
11657 Do not try to dynamically allocate condition code registers, only
11658 use @code{icc0} and @code{fcc0}.
11663 Change ABI to use double word insns.
11668 Do not use double word instructions.
11673 Use floating point double instructions.
11676 @opindex mno-double
11678 Do not use floating point double instructions.
11683 Use media instructions.
11688 Do not use media instructions.
11693 Use multiply and add/subtract instructions.
11696 @opindex mno-muladd
11698 Do not use multiply and add/subtract instructions.
11703 Select the FDPIC ABI, that uses function descriptors to represent
11704 pointers to functions. Without any PIC/PIE-related options, it
11705 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11706 assumes GOT entries and small data are within a 12-bit range from the
11707 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11708 are computed with 32 bits.
11709 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11712 @opindex minline-plt
11714 Enable inlining of PLT entries in function calls to functions that are
11715 not known to bind locally. It has no effect without @option{-mfdpic}.
11716 It's enabled by default if optimizing for speed and compiling for
11717 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11718 optimization option such as @option{-O3} or above is present in the
11724 Assume a large TLS segment when generating thread-local code.
11729 Do not assume a large TLS segment when generating thread-local code.
11734 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11735 that is known to be in read-only sections. It's enabled by default,
11736 except for @option{-fpic} or @option{-fpie}: even though it may help
11737 make the global offset table smaller, it trades 1 instruction for 4.
11738 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11739 one of which may be shared by multiple symbols, and it avoids the need
11740 for a GOT entry for the referenced symbol, so it's more likely to be a
11741 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11743 @item -multilib-library-pic
11744 @opindex multilib-library-pic
11746 Link with the (library, not FD) pic libraries. It's implied by
11747 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11748 @option{-fpic} without @option{-mfdpic}. You should never have to use
11752 @opindex mlinked-fp
11754 Follow the EABI requirement of always creating a frame pointer whenever
11755 a stack frame is allocated. This option is enabled by default and can
11756 be disabled with @option{-mno-linked-fp}.
11759 @opindex mlong-calls
11761 Use indirect addressing to call functions outside the current
11762 compilation unit. This allows the functions to be placed anywhere
11763 within the 32-bit address space.
11765 @item -malign-labels
11766 @opindex malign-labels
11768 Try to align labels to an 8-byte boundary by inserting nops into the
11769 previous packet. This option only has an effect when VLIW packing
11770 is enabled. It doesn't create new packets; it merely adds nops to
11773 @item -mlibrary-pic
11774 @opindex mlibrary-pic
11776 Generate position-independent EABI code.
11781 Use only the first four media accumulator registers.
11786 Use all eight media accumulator registers.
11791 Pack VLIW instructions.
11796 Do not pack VLIW instructions.
11799 @opindex mno-eflags
11801 Do not mark ABI switches in e_flags.
11804 @opindex mcond-move
11806 Enable the use of conditional-move instructions (default).
11808 This switch is mainly for debugging the compiler and will likely be removed
11809 in a future version.
11811 @item -mno-cond-move
11812 @opindex mno-cond-move
11814 Disable the use of conditional-move instructions.
11816 This switch is mainly for debugging the compiler and will likely be removed
11817 in a future version.
11822 Enable the use of conditional set instructions (default).
11824 This switch is mainly for debugging the compiler and will likely be removed
11825 in a future version.
11830 Disable the use of conditional set instructions.
11832 This switch is mainly for debugging the compiler and will likely be removed
11833 in a future version.
11836 @opindex mcond-exec
11838 Enable the use of conditional execution (default).
11840 This switch is mainly for debugging the compiler and will likely be removed
11841 in a future version.
11843 @item -mno-cond-exec
11844 @opindex mno-cond-exec
11846 Disable the use of conditional execution.
11848 This switch is mainly for debugging the compiler and will likely be removed
11849 in a future version.
11851 @item -mvliw-branch
11852 @opindex mvliw-branch
11854 Run a pass to pack branches into VLIW instructions (default).
11856 This switch is mainly for debugging the compiler and will likely be removed
11857 in a future version.
11859 @item -mno-vliw-branch
11860 @opindex mno-vliw-branch
11862 Do not run a pass to pack branches into VLIW instructions.
11864 This switch is mainly for debugging the compiler and will likely be removed
11865 in a future version.
11867 @item -mmulti-cond-exec
11868 @opindex mmulti-cond-exec
11870 Enable optimization of @code{&&} and @code{||} in conditional execution
11873 This switch is mainly for debugging the compiler and will likely be removed
11874 in a future version.
11876 @item -mno-multi-cond-exec
11877 @opindex mno-multi-cond-exec
11879 Disable optimization of @code{&&} and @code{||} in conditional execution.
11881 This switch is mainly for debugging the compiler and will likely be removed
11882 in a future version.
11884 @item -mnested-cond-exec
11885 @opindex mnested-cond-exec
11887 Enable nested conditional execution optimizations (default).
11889 This switch is mainly for debugging the compiler and will likely be removed
11890 in a future version.
11892 @item -mno-nested-cond-exec
11893 @opindex mno-nested-cond-exec
11895 Disable nested conditional execution optimizations.
11897 This switch is mainly for debugging the compiler and will likely be removed
11898 in a future version.
11900 @item -moptimize-membar
11901 @opindex moptimize-membar
11903 This switch removes redundant @code{membar} instructions from the
11904 compiler generated code. It is enabled by default.
11906 @item -mno-optimize-membar
11907 @opindex mno-optimize-membar
11909 This switch disables the automatic removal of redundant @code{membar}
11910 instructions from the generated code.
11912 @item -mtomcat-stats
11913 @opindex mtomcat-stats
11915 Cause gas to print out tomcat statistics.
11917 @item -mcpu=@var{cpu}
11920 Select the processor type for which to generate code. Possible values are
11921 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11922 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11926 @node GNU/Linux Options
11927 @subsection GNU/Linux Options
11929 These @samp{-m} options are defined for GNU/Linux targets:
11934 Use the GNU C library. This is the default except
11935 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11939 Use uClibc C library. This is the default on
11940 @samp{*-*-linux-*uclibc*} targets.
11944 Use Bionic C library. This is the default on
11945 @samp{*-*-linux-*android*} targets.
11949 Compile code compatible with Android platform. This is the default on
11950 @samp{*-*-linux-*android*} targets.
11952 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11953 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11954 this option makes the GCC driver pass Android-specific options to the linker.
11955 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11958 @item -tno-android-cc
11959 @opindex tno-android-cc
11960 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11961 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11962 @option{-fno-rtti} by default.
11964 @item -tno-android-ld
11965 @opindex tno-android-ld
11966 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11967 linking options to the linker.
11971 @node H8/300 Options
11972 @subsection H8/300 Options
11974 These @samp{-m} options are defined for the H8/300 implementations:
11979 Shorten some address references at link time, when possible; uses the
11980 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11981 ld, Using ld}, for a fuller description.
11985 Generate code for the H8/300H@.
11989 Generate code for the H8S@.
11993 Generate code for the H8S and H8/300H in the normal mode. This switch
11994 must be used either with @option{-mh} or @option{-ms}.
11998 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12002 Make @code{int} data 32 bits by default.
12005 @opindex malign-300
12006 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12007 The default for the H8/300H and H8S is to align longs and floats on 4
12009 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12010 This option has no effect on the H8/300.
12014 @subsection HPPA Options
12015 @cindex HPPA Options
12017 These @samp{-m} options are defined for the HPPA family of computers:
12020 @item -march=@var{architecture-type}
12022 Generate code for the specified architecture. The choices for
12023 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12024 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12025 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12026 architecture option for your machine. Code compiled for lower numbered
12027 architectures will run on higher numbered architectures, but not the
12030 @item -mpa-risc-1-0
12031 @itemx -mpa-risc-1-1
12032 @itemx -mpa-risc-2-0
12033 @opindex mpa-risc-1-0
12034 @opindex mpa-risc-1-1
12035 @opindex mpa-risc-2-0
12036 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12039 @opindex mbig-switch
12040 Generate code suitable for big switch tables. Use this option only if
12041 the assembler/linker complain about out of range branches within a switch
12044 @item -mjump-in-delay
12045 @opindex mjump-in-delay
12046 Fill delay slots of function calls with unconditional jump instructions
12047 by modifying the return pointer for the function call to be the target
12048 of the conditional jump.
12050 @item -mdisable-fpregs
12051 @opindex mdisable-fpregs
12052 Prevent floating point registers from being used in any manner. This is
12053 necessary for compiling kernels which perform lazy context switching of
12054 floating point registers. If you use this option and attempt to perform
12055 floating point operations, the compiler will abort.
12057 @item -mdisable-indexing
12058 @opindex mdisable-indexing
12059 Prevent the compiler from using indexing address modes. This avoids some
12060 rather obscure problems when compiling MIG generated code under MACH@.
12062 @item -mno-space-regs
12063 @opindex mno-space-regs
12064 Generate code that assumes the target has no space registers. This allows
12065 GCC to generate faster indirect calls and use unscaled index address modes.
12067 Such code is suitable for level 0 PA systems and kernels.
12069 @item -mfast-indirect-calls
12070 @opindex mfast-indirect-calls
12071 Generate code that assumes calls never cross space boundaries. This
12072 allows GCC to emit code which performs faster indirect calls.
12074 This option will not work in the presence of shared libraries or nested
12077 @item -mfixed-range=@var{register-range}
12078 @opindex mfixed-range
12079 Generate code treating the given register range as fixed registers.
12080 A fixed register is one that the register allocator can not use. This is
12081 useful when compiling kernel code. A register range is specified as
12082 two registers separated by a dash. Multiple register ranges can be
12083 specified separated by a comma.
12085 @item -mlong-load-store
12086 @opindex mlong-load-store
12087 Generate 3-instruction load and store sequences as sometimes required by
12088 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12091 @item -mportable-runtime
12092 @opindex mportable-runtime
12093 Use the portable calling conventions proposed by HP for ELF systems.
12097 Enable the use of assembler directives only GAS understands.
12099 @item -mschedule=@var{cpu-type}
12101 Schedule code according to the constraints for the machine type
12102 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12103 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12104 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12105 proper scheduling option for your machine. The default scheduling is
12109 @opindex mlinker-opt
12110 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12111 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12112 linkers in which they give bogus error messages when linking some programs.
12115 @opindex msoft-float
12116 Generate output containing library calls for floating point.
12117 @strong{Warning:} the requisite libraries are not available for all HPPA
12118 targets. Normally the facilities of the machine's usual C compiler are
12119 used, but this cannot be done directly in cross-compilation. You must make
12120 your own arrangements to provide suitable library functions for
12123 @option{-msoft-float} changes the calling convention in the output file;
12124 therefore, it is only useful if you compile @emph{all} of a program with
12125 this option. In particular, you need to compile @file{libgcc.a}, the
12126 library that comes with GCC, with @option{-msoft-float} in order for
12131 Generate the predefine, @code{_SIO}, for server IO@. The default is
12132 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12133 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12134 options are available under HP-UX and HI-UX@.
12138 Use GNU ld specific options. This passes @option{-shared} to ld when
12139 building a shared library. It is the default when GCC is configured,
12140 explicitly or implicitly, with the GNU linker. This option does not
12141 have any affect on which ld is called, it only changes what parameters
12142 are passed to that ld. The ld that is called is determined by the
12143 @option{--with-ld} configure option, GCC's program search path, and
12144 finally by the user's @env{PATH}. The linker used by GCC can be printed
12145 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12146 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12150 Use HP ld specific options. This passes @option{-b} to ld when building
12151 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12152 links. It is the default when GCC is configured, explicitly or
12153 implicitly, with the HP linker. This option does not have any affect on
12154 which ld is called, it only changes what parameters are passed to that
12155 ld. The ld that is called is determined by the @option{--with-ld}
12156 configure option, GCC's program search path, and finally by the user's
12157 @env{PATH}. The linker used by GCC can be printed using @samp{which
12158 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12159 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12162 @opindex mno-long-calls
12163 Generate code that uses long call sequences. This ensures that a call
12164 is always able to reach linker generated stubs. The default is to generate
12165 long calls only when the distance from the call site to the beginning
12166 of the function or translation unit, as the case may be, exceeds a
12167 predefined limit set by the branch type being used. The limits for
12168 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12169 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12172 Distances are measured from the beginning of functions when using the
12173 @option{-ffunction-sections} option, or when using the @option{-mgas}
12174 and @option{-mno-portable-runtime} options together under HP-UX with
12177 It is normally not desirable to use this option as it will degrade
12178 performance. However, it may be useful in large applications,
12179 particularly when partial linking is used to build the application.
12181 The types of long calls used depends on the capabilities of the
12182 assembler and linker, and the type of code being generated. The
12183 impact on systems that support long absolute calls, and long pic
12184 symbol-difference or pc-relative calls should be relatively small.
12185 However, an indirect call is used on 32-bit ELF systems in pic code
12186 and it is quite long.
12188 @item -munix=@var{unix-std}
12190 Generate compiler predefines and select a startfile for the specified
12191 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12192 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12193 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12194 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12195 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12198 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12199 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12200 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12201 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12202 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12203 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12205 It is @emph{important} to note that this option changes the interfaces
12206 for various library routines. It also affects the operational behavior
12207 of the C library. Thus, @emph{extreme} care is needed in using this
12210 Library code that is intended to operate with more than one UNIX
12211 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12212 as appropriate. Most GNU software doesn't provide this capability.
12216 Suppress the generation of link options to search libdld.sl when the
12217 @option{-static} option is specified on HP-UX 10 and later.
12221 The HP-UX implementation of setlocale in libc has a dependency on
12222 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12223 when the @option{-static} option is specified, special link options
12224 are needed to resolve this dependency.
12226 On HP-UX 10 and later, the GCC driver adds the necessary options to
12227 link with libdld.sl when the @option{-static} option is specified.
12228 This causes the resulting binary to be dynamic. On the 64-bit port,
12229 the linkers generate dynamic binaries by default in any case. The
12230 @option{-nolibdld} option can be used to prevent the GCC driver from
12231 adding these link options.
12235 Add support for multithreading with the @dfn{dce thread} library
12236 under HP-UX@. This option sets flags for both the preprocessor and
12240 @node i386 and x86-64 Options
12241 @subsection Intel 386 and AMD x86-64 Options
12242 @cindex i386 Options
12243 @cindex x86-64 Options
12244 @cindex Intel 386 Options
12245 @cindex AMD x86-64 Options
12247 These @samp{-m} options are defined for the i386 and x86-64 family of
12251 @item -mtune=@var{cpu-type}
12253 Tune to @var{cpu-type} everything applicable about the generated code, except
12254 for the ABI and the set of available instructions. The choices for
12255 @var{cpu-type} are:
12258 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12259 If you know the CPU on which your code will run, then you should use
12260 the corresponding @option{-mtune} option instead of
12261 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12262 of your application will have, then you should use this option.
12264 As new processors are deployed in the marketplace, the behavior of this
12265 option will change. Therefore, if you upgrade to a newer version of
12266 GCC, the code generated option will change to reflect the processors
12267 that were most common when that version of GCC was released.
12269 There is no @option{-march=generic} option because @option{-march}
12270 indicates the instruction set the compiler can use, and there is no
12271 generic instruction set applicable to all processors. In contrast,
12272 @option{-mtune} indicates the processor (or, in this case, collection of
12273 processors) for which the code is optimized.
12275 This selects the CPU to tune for at compilation time by determining
12276 the processor type of the compiling machine. Using @option{-mtune=native}
12277 will produce code optimized for the local machine under the constraints
12278 of the selected instruction set. Using @option{-march=native} will
12279 enable all instruction subsets supported by the local machine (hence
12280 the result might not run on different machines).
12282 Original Intel's i386 CPU@.
12284 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12285 @item i586, pentium
12286 Intel Pentium CPU with no MMX support.
12288 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12290 Intel PentiumPro CPU@.
12292 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12293 instruction set will be used, so the code will run on all i686 family chips.
12295 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12296 @item pentium3, pentium3m
12297 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12300 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12301 support. Used by Centrino notebooks.
12302 @item pentium4, pentium4m
12303 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12305 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12308 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12309 SSE2 and SSE3 instruction set support.
12311 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12312 instruction set support.
12314 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12315 and SSE4.2 instruction set support.
12317 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12318 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12320 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12321 instruction set support.
12323 AMD K6 CPU with MMX instruction set support.
12325 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12326 @item athlon, athlon-tbird
12327 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12329 @item athlon-4, athlon-xp, athlon-mp
12330 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12331 instruction set support.
12332 @item k8, opteron, athlon64, athlon-fx
12333 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12334 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12335 @item k8-sse3, opteron-sse3, athlon64-sse3
12336 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12337 @item amdfam10, barcelona
12338 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12339 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12340 instruction set extensions.)
12342 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12345 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12346 instruction set support.
12348 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12349 implemented for this chip.)
12351 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12352 implemented for this chip.)
12354 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12357 While picking a specific @var{cpu-type} will schedule things appropriately
12358 for that particular chip, the compiler will not generate any code that
12359 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12362 @item -march=@var{cpu-type}
12364 Generate instructions for the machine type @var{cpu-type}. The choices
12365 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12366 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12368 @item -mcpu=@var{cpu-type}
12370 A deprecated synonym for @option{-mtune}.
12372 @item -mfpmath=@var{unit}
12374 Generate floating point arithmetics for selected unit @var{unit}. The choices
12375 for @var{unit} are:
12379 Use the standard 387 floating point coprocessor present majority of chips and
12380 emulated otherwise. Code compiled with this option will run almost everywhere.
12381 The temporary results are computed in 80bit precision instead of precision
12382 specified by the type resulting in slightly different results compared to most
12383 of other chips. See @option{-ffloat-store} for more detailed description.
12385 This is the default choice for i386 compiler.
12388 Use scalar floating point instructions present in the SSE instruction set.
12389 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12390 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12391 instruction set supports only single precision arithmetics, thus the double and
12392 extended precision arithmetics is still done using 387. Later version, present
12393 only in Pentium4 and the future AMD x86-64 chips supports double precision
12396 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12397 or @option{-msse2} switches to enable SSE extensions and make this option
12398 effective. For the x86-64 compiler, these extensions are enabled by default.
12400 The resulting code should be considerably faster in the majority of cases and avoid
12401 the numerical instability problems of 387 code, but may break some existing
12402 code that expects temporaries to be 80bit.
12404 This is the default choice for the x86-64 compiler.
12409 Attempt to utilize both instruction sets at once. This effectively double the
12410 amount of available registers and on chips with separate execution units for
12411 387 and SSE the execution resources too. Use this option with care, as it is
12412 still experimental, because the GCC register allocator does not model separate
12413 functional units well resulting in instable performance.
12416 @item -masm=@var{dialect}
12417 @opindex masm=@var{dialect}
12418 Output asm instructions using selected @var{dialect}. Supported
12419 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12420 not support @samp{intel}.
12423 @itemx -mno-ieee-fp
12425 @opindex mno-ieee-fp
12426 Control whether or not the compiler uses IEEE floating point
12427 comparisons. These handle correctly the case where the result of a
12428 comparison is unordered.
12431 @opindex msoft-float
12432 Generate output containing library calls for floating point.
12433 @strong{Warning:} the requisite libraries are not part of GCC@.
12434 Normally the facilities of the machine's usual C compiler are used, but
12435 this can't be done directly in cross-compilation. You must make your
12436 own arrangements to provide suitable library functions for
12439 On machines where a function returns floating point results in the 80387
12440 register stack, some floating point opcodes may be emitted even if
12441 @option{-msoft-float} is used.
12443 @item -mno-fp-ret-in-387
12444 @opindex mno-fp-ret-in-387
12445 Do not use the FPU registers for return values of functions.
12447 The usual calling convention has functions return values of types
12448 @code{float} and @code{double} in an FPU register, even if there
12449 is no FPU@. The idea is that the operating system should emulate
12452 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12453 in ordinary CPU registers instead.
12455 @item -mno-fancy-math-387
12456 @opindex mno-fancy-math-387
12457 Some 387 emulators do not support the @code{sin}, @code{cos} and
12458 @code{sqrt} instructions for the 387. Specify this option to avoid
12459 generating those instructions. This option is the default on FreeBSD,
12460 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12461 indicates that the target CPU will always have an FPU and so the
12462 instruction will not need emulation. As of revision 2.6.1, these
12463 instructions are not generated unless you also use the
12464 @option{-funsafe-math-optimizations} switch.
12466 @item -malign-double
12467 @itemx -mno-align-double
12468 @opindex malign-double
12469 @opindex mno-align-double
12470 Control whether GCC aligns @code{double}, @code{long double}, and
12471 @code{long long} variables on a two word boundary or a one word
12472 boundary. Aligning @code{double} variables on a two word boundary will
12473 produce code that runs somewhat faster on a @samp{Pentium} at the
12474 expense of more memory.
12476 On x86-64, @option{-malign-double} is enabled by default.
12478 @strong{Warning:} if you use the @option{-malign-double} switch,
12479 structures containing the above types will be aligned differently than
12480 the published application binary interface specifications for the 386
12481 and will not be binary compatible with structures in code compiled
12482 without that switch.
12484 @item -m96bit-long-double
12485 @itemx -m128bit-long-double
12486 @opindex m96bit-long-double
12487 @opindex m128bit-long-double
12488 These switches control the size of @code{long double} type. The i386
12489 application binary interface specifies the size to be 96 bits,
12490 so @option{-m96bit-long-double} is the default in 32 bit mode.
12492 Modern architectures (Pentium and newer) would prefer @code{long double}
12493 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12494 conforming to the ABI, this would not be possible. So specifying a
12495 @option{-m128bit-long-double} will align @code{long double}
12496 to a 16 byte boundary by padding the @code{long double} with an additional
12499 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12500 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12502 Notice that neither of these options enable any extra precision over the x87
12503 standard of 80 bits for a @code{long double}.
12505 @strong{Warning:} if you override the default value for your target ABI, the
12506 structures and arrays containing @code{long double} variables will change
12507 their size as well as function calling convention for function taking
12508 @code{long double} will be modified. Hence they will not be binary
12509 compatible with arrays or structures in code compiled without that switch.
12511 @item -mlarge-data-threshold=@var{number}
12512 @opindex mlarge-data-threshold=@var{number}
12513 When @option{-mcmodel=medium} is specified, the data greater than
12514 @var{threshold} are placed in large data section. This value must be the
12515 same across all object linked into the binary and defaults to 65535.
12519 Use a different function-calling convention, in which functions that
12520 take a fixed number of arguments return with the @code{ret} @var{num}
12521 instruction, which pops their arguments while returning. This saves one
12522 instruction in the caller since there is no need to pop the arguments
12525 You can specify that an individual function is called with this calling
12526 sequence with the function attribute @samp{stdcall}. You can also
12527 override the @option{-mrtd} option by using the function attribute
12528 @samp{cdecl}. @xref{Function Attributes}.
12530 @strong{Warning:} this calling convention is incompatible with the one
12531 normally used on Unix, so you cannot use it if you need to call
12532 libraries compiled with the Unix compiler.
12534 Also, you must provide function prototypes for all functions that
12535 take variable numbers of arguments (including @code{printf});
12536 otherwise incorrect code will be generated for calls to those
12539 In addition, seriously incorrect code will result if you call a
12540 function with too many arguments. (Normally, extra arguments are
12541 harmlessly ignored.)
12543 @item -mregparm=@var{num}
12545 Control how many registers are used to pass integer arguments. By
12546 default, no registers are used to pass arguments, and at most 3
12547 registers can be used. You can control this behavior for a specific
12548 function by using the function attribute @samp{regparm}.
12549 @xref{Function Attributes}.
12551 @strong{Warning:} if you use this switch, and
12552 @var{num} is nonzero, then you must build all modules with the same
12553 value, including any libraries. This includes the system libraries and
12557 @opindex msseregparm
12558 Use SSE register passing conventions for float and double arguments
12559 and return values. You can control this behavior for a specific
12560 function by using the function attribute @samp{sseregparm}.
12561 @xref{Function Attributes}.
12563 @strong{Warning:} if you use this switch then you must build all
12564 modules with the same value, including any libraries. This includes
12565 the system libraries and startup modules.
12567 @item -mvect8-ret-in-mem
12568 @opindex mvect8-ret-in-mem
12569 Return 8-byte vectors in memory instead of MMX registers. This is the
12570 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12571 Studio compilers until version 12. Later compiler versions (starting
12572 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12573 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12574 you need to remain compatible with existing code produced by those
12575 previous compiler versions or older versions of GCC.
12584 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12585 is specified, the significands of results of floating-point operations are
12586 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12587 significands of results of floating-point operations to 53 bits (double
12588 precision) and @option{-mpc80} rounds the significands of results of
12589 floating-point operations to 64 bits (extended double precision), which is
12590 the default. When this option is used, floating-point operations in higher
12591 precisions are not available to the programmer without setting the FPU
12592 control word explicitly.
12594 Setting the rounding of floating-point operations to less than the default
12595 80 bits can speed some programs by 2% or more. Note that some mathematical
12596 libraries assume that extended precision (80 bit) floating-point operations
12597 are enabled by default; routines in such libraries could suffer significant
12598 loss of accuracy, typically through so-called "catastrophic cancellation",
12599 when this option is used to set the precision to less than extended precision.
12601 @item -mstackrealign
12602 @opindex mstackrealign
12603 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12604 option will generate an alternate prologue and epilogue that realigns the
12605 runtime stack if necessary. This supports mixing legacy codes that keep
12606 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12607 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12608 applicable to individual functions.
12610 @item -mpreferred-stack-boundary=@var{num}
12611 @opindex mpreferred-stack-boundary
12612 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12613 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12614 the default is 4 (16 bytes or 128 bits).
12616 @item -mincoming-stack-boundary=@var{num}
12617 @opindex mincoming-stack-boundary
12618 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12619 boundary. If @option{-mincoming-stack-boundary} is not specified,
12620 the one specified by @option{-mpreferred-stack-boundary} will be used.
12622 On Pentium and PentiumPro, @code{double} and @code{long double} values
12623 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12624 suffer significant run time performance penalties. On Pentium III, the
12625 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12626 properly if it is not 16 byte aligned.
12628 To ensure proper alignment of this values on the stack, the stack boundary
12629 must be as aligned as that required by any value stored on the stack.
12630 Further, every function must be generated such that it keeps the stack
12631 aligned. Thus calling a function compiled with a higher preferred
12632 stack boundary from a function compiled with a lower preferred stack
12633 boundary will most likely misalign the stack. It is recommended that
12634 libraries that use callbacks always use the default setting.
12636 This extra alignment does consume extra stack space, and generally
12637 increases code size. Code that is sensitive to stack space usage, such
12638 as embedded systems and operating system kernels, may want to reduce the
12639 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12666 @itemx -mno-fsgsbase
12698 These switches enable or disable the use of instructions in the MMX, SSE,
12699 SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA,
12700 SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12701 These extensions are also available as built-in functions: see
12702 @ref{X86 Built-in Functions}, for details of the functions enabled and
12703 disabled by these switches.
12705 To have SSE/SSE2 instructions generated automatically from floating-point
12706 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12708 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12709 generates new AVX instructions or AVX equivalence for all SSEx instructions
12712 These options will enable GCC to use these extended instructions in
12713 generated code, even without @option{-mfpmath=sse}. Applications which
12714 perform runtime CPU detection must compile separate files for each
12715 supported architecture, using the appropriate flags. In particular,
12716 the file containing the CPU detection code should be compiled without
12721 This option instructs GCC to emit a @code{cld} instruction in the prologue
12722 of functions that use string instructions. String instructions depend on
12723 the DF flag to select between autoincrement or autodecrement mode. While the
12724 ABI specifies the DF flag to be cleared on function entry, some operating
12725 systems violate this specification by not clearing the DF flag in their
12726 exception dispatchers. The exception handler can be invoked with the DF flag
12727 set which leads to wrong direction mode, when string instructions are used.
12728 This option can be enabled by default on 32-bit x86 targets by configuring
12729 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12730 instructions can be suppressed with the @option{-mno-cld} compiler option
12734 @opindex mvzeroupper
12735 This option instructs GCC to emit a @code{vzeroupper} instruction
12736 before a transfer of control flow out of the function to minimize
12737 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12742 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12743 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12744 data types. This is useful for high resolution counters that could be updated
12745 by multiple processors (or cores). This instruction is generated as part of
12746 atomic built-in functions: see @ref{Atomic Builtins} for details.
12750 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12751 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12752 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12753 SAHF are load and store instructions, respectively, for certain status flags.
12754 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12755 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12759 This option will enable GCC to use movbe instruction to implement
12760 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12764 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12765 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12766 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12770 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12771 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12772 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12773 variants) for single precision floating point arguments. These instructions
12774 are generated only when @option{-funsafe-math-optimizations} is enabled
12775 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12776 Note that while the throughput of the sequence is higher than the throughput
12777 of the non-reciprocal instruction, the precision of the sequence can be
12778 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12780 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12781 already with @option{-ffast-math} (or the above option combination), and
12782 doesn't need @option{-mrecip}.
12784 @item -mveclibabi=@var{type}
12785 @opindex mveclibabi
12786 Specifies the ABI type to use for vectorizing intrinsics using an
12787 external library. Supported types are @code{svml} for the Intel short
12788 vector math library and @code{acml} for the AMD math core library style
12789 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12790 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12791 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12792 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12793 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12794 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12795 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12796 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12797 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12798 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12799 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12800 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12801 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12802 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12803 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12804 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12805 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12806 compatible library will have to be specified at link time.
12808 @item -mabi=@var{name}
12810 Generate code for the specified calling convention. Permissible values
12811 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12812 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12813 ABI when targeting Windows. On all other systems, the default is the
12814 SYSV ABI. You can control this behavior for a specific function by
12815 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12816 @xref{Function Attributes}.
12818 @item -mtls-dialect=@var{type}
12819 @opindex mtls-dialect
12820 Generate code to access thread-local storage using the @samp{gnu} or
12821 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
12822 @samp{gnu2} is more efficient, but it may add compile- and run-time
12823 requirements that cannot be satisfied on all systems.
12826 @itemx -mno-push-args
12827 @opindex mpush-args
12828 @opindex mno-push-args
12829 Use PUSH operations to store outgoing parameters. This method is shorter
12830 and usually equally fast as method using SUB/MOV operations and is enabled
12831 by default. In some cases disabling it may improve performance because of
12832 improved scheduling and reduced dependencies.
12834 @item -maccumulate-outgoing-args
12835 @opindex maccumulate-outgoing-args
12836 If enabled, the maximum amount of space required for outgoing arguments will be
12837 computed in the function prologue. This is faster on most modern CPUs
12838 because of reduced dependencies, improved scheduling and reduced stack usage
12839 when preferred stack boundary is not equal to 2. The drawback is a notable
12840 increase in code size. This switch implies @option{-mno-push-args}.
12844 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12845 on thread-safe exception handling must compile and link all code with the
12846 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12847 @option{-D_MT}; when linking, it links in a special thread helper library
12848 @option{-lmingwthrd} which cleans up per thread exception handling data.
12850 @item -mno-align-stringops
12851 @opindex mno-align-stringops
12852 Do not align destination of inlined string operations. This switch reduces
12853 code size and improves performance in case the destination is already aligned,
12854 but GCC doesn't know about it.
12856 @item -minline-all-stringops
12857 @opindex minline-all-stringops
12858 By default GCC inlines string operations only when destination is known to be
12859 aligned at least to 4 byte boundary. This enables more inlining, increase code
12860 size, but may improve performance of code that depends on fast memcpy, strlen
12861 and memset for short lengths.
12863 @item -minline-stringops-dynamically
12864 @opindex minline-stringops-dynamically
12865 For string operation of unknown size, inline runtime checks so for small
12866 blocks inline code is used, while for large blocks library call is used.
12868 @item -mstringop-strategy=@var{alg}
12869 @opindex mstringop-strategy=@var{alg}
12870 Overwrite internal decision heuristic about particular algorithm to inline
12871 string operation with. The allowed values are @code{rep_byte},
12872 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12873 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12874 expanding inline loop, @code{libcall} for always expanding library call.
12876 @item -momit-leaf-frame-pointer
12877 @opindex momit-leaf-frame-pointer
12878 Don't keep the frame pointer in a register for leaf functions. This
12879 avoids the instructions to save, set up and restore frame pointers and
12880 makes an extra register available in leaf functions. The option
12881 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12882 which might make debugging harder.
12884 @item -mtls-direct-seg-refs
12885 @itemx -mno-tls-direct-seg-refs
12886 @opindex mtls-direct-seg-refs
12887 Controls whether TLS variables may be accessed with offsets from the
12888 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12889 or whether the thread base pointer must be added. Whether or not this
12890 is legal depends on the operating system, and whether it maps the
12891 segment to cover the entire TLS area.
12893 For systems that use GNU libc, the default is on.
12896 @itemx -mno-sse2avx
12898 Specify that the assembler should encode SSE instructions with VEX
12899 prefix. The option @option{-mavx} turns this on by default.
12904 If profiling is active @option{-pg} put the profiling
12905 counter call before prologue.
12906 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12907 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12910 @itemx -mno-8bit-idiv
12912 On some processors, like Intel Atom, 8bit unsigned integer divide is
12913 much faster than 32bit/64bit integer divide. This option will generate a
12914 runt-time check. If both dividend and divisor are within range of 0
12915 to 255, 8bit unsigned integer divide will be used instead of
12916 32bit/64bit integer divide.
12918 @item -mavx256-split-unaligned-load
12919 @item -mavx256-split-unaligned-store
12920 @opindex avx256-split-unaligned-load
12921 @opindex avx256-split-unaligned-store
12922 Split 32-byte AVX unaligned load and store.
12926 These @samp{-m} switches are supported in addition to the above
12927 on AMD x86-64 processors in 64-bit environments.
12936 Generate code for a 32-bit or 64-bit environment.
12937 The @option{-m32} option sets int, long and pointer to 32 bits and
12938 generates code that runs on any i386 system.
12939 The @option{-m64} option sets int to 32 bits and long and pointer
12940 to 64 bits and generates code for AMD's x86-64 architecture.
12941 The @option{-mx32} option sets int, long and pointer to 32 bits and
12942 generates code for AMD's x86-64 architecture.
12943 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
12944 and @option{-mdynamic-no-pic} options.
12946 @item -mno-red-zone
12947 @opindex mno-red-zone
12948 Do not use a so called red zone for x86-64 code. The red zone is mandated
12949 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12950 stack pointer that will not be modified by signal or interrupt handlers
12951 and therefore can be used for temporary data without adjusting the stack
12952 pointer. The flag @option{-mno-red-zone} disables this red zone.
12954 @item -mcmodel=small
12955 @opindex mcmodel=small
12956 Generate code for the small code model: the program and its symbols must
12957 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12958 Programs can be statically or dynamically linked. This is the default
12961 @item -mcmodel=kernel
12962 @opindex mcmodel=kernel
12963 Generate code for the kernel code model. The kernel runs in the
12964 negative 2 GB of the address space.
12965 This model has to be used for Linux kernel code.
12967 @item -mcmodel=medium
12968 @opindex mcmodel=medium
12969 Generate code for the medium model: The program is linked in the lower 2
12970 GB of the address space. Small symbols are also placed there. Symbols
12971 with sizes larger than @option{-mlarge-data-threshold} are put into
12972 large data or bss sections and can be located above 2GB. Programs can
12973 be statically or dynamically linked.
12975 @item -mcmodel=large
12976 @opindex mcmodel=large
12977 Generate code for the large model: This model makes no assumptions
12978 about addresses and sizes of sections.
12981 @node i386 and x86-64 Windows Options
12982 @subsection i386 and x86-64 Windows Options
12983 @cindex i386 and x86-64 Windows Options
12985 These additional options are available for Windows targets:
12990 This option is available for Cygwin and MinGW targets. It
12991 specifies that a console application is to be generated, by
12992 instructing the linker to set the PE header subsystem type
12993 required for console applications.
12994 This is the default behavior for Cygwin and MinGW targets.
12998 This option is available for Cygwin and MinGW targets. It
12999 specifies that a DLL - a dynamic link library - is to be
13000 generated, enabling the selection of the required runtime
13001 startup object and entry point.
13003 @item -mnop-fun-dllimport
13004 @opindex mnop-fun-dllimport
13005 This option is available for Cygwin and MinGW targets. It
13006 specifies that the dllimport attribute should be ignored.
13010 This option is available for MinGW targets. It specifies
13011 that MinGW-specific thread support is to be used.
13015 This option is available for mingw-w64 targets. It specifies
13016 that the UNICODE macro is getting pre-defined and that the
13017 unicode capable runtime startup code is chosen.
13021 This option is available for Cygwin and MinGW targets. It
13022 specifies that the typical Windows pre-defined macros are to
13023 be set in the pre-processor, but does not influence the choice
13024 of runtime library/startup code.
13028 This option is available for Cygwin and MinGW targets. It
13029 specifies that a GUI application is to be generated by
13030 instructing the linker to set the PE header subsystem type
13033 @item -fno-set-stack-executable
13034 @opindex fno-set-stack-executable
13035 This option is available for MinGW targets. It specifies that
13036 the executable flag for stack used by nested functions isn't
13037 set. This is necessary for binaries running in kernel mode of
13038 Windows, as there the user32 API, which is used to set executable
13039 privileges, isn't available.
13041 @item -mpe-aligned-commons
13042 @opindex mpe-aligned-commons
13043 This option is available for Cygwin and MinGW targets. It
13044 specifies that the GNU extension to the PE file format that
13045 permits the correct alignment of COMMON variables should be
13046 used when generating code. It will be enabled by default if
13047 GCC detects that the target assembler found during configuration
13048 supports the feature.
13051 See also under @ref{i386 and x86-64 Options} for standard options.
13053 @node IA-64 Options
13054 @subsection IA-64 Options
13055 @cindex IA-64 Options
13057 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13061 @opindex mbig-endian
13062 Generate code for a big endian target. This is the default for HP-UX@.
13064 @item -mlittle-endian
13065 @opindex mlittle-endian
13066 Generate code for a little endian target. This is the default for AIX5
13072 @opindex mno-gnu-as
13073 Generate (or don't) code for the GNU assembler. This is the default.
13074 @c Also, this is the default if the configure option @option{--with-gnu-as}
13080 @opindex mno-gnu-ld
13081 Generate (or don't) code for the GNU linker. This is the default.
13082 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13087 Generate code that does not use a global pointer register. The result
13088 is not position independent code, and violates the IA-64 ABI@.
13090 @item -mvolatile-asm-stop
13091 @itemx -mno-volatile-asm-stop
13092 @opindex mvolatile-asm-stop
13093 @opindex mno-volatile-asm-stop
13094 Generate (or don't) a stop bit immediately before and after volatile asm
13097 @item -mregister-names
13098 @itemx -mno-register-names
13099 @opindex mregister-names
13100 @opindex mno-register-names
13101 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13102 the stacked registers. This may make assembler output more readable.
13108 Disable (or enable) optimizations that use the small data section. This may
13109 be useful for working around optimizer bugs.
13111 @item -mconstant-gp
13112 @opindex mconstant-gp
13113 Generate code that uses a single constant global pointer value. This is
13114 useful when compiling kernel code.
13118 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13119 This is useful when compiling firmware code.
13121 @item -minline-float-divide-min-latency
13122 @opindex minline-float-divide-min-latency
13123 Generate code for inline divides of floating point values
13124 using the minimum latency algorithm.
13126 @item -minline-float-divide-max-throughput
13127 @opindex minline-float-divide-max-throughput
13128 Generate code for inline divides of floating point values
13129 using the maximum throughput algorithm.
13131 @item -mno-inline-float-divide
13132 @opindex mno-inline-float-divide
13133 Do not generate inline code for divides of floating point values.
13135 @item -minline-int-divide-min-latency
13136 @opindex minline-int-divide-min-latency
13137 Generate code for inline divides of integer values
13138 using the minimum latency algorithm.
13140 @item -minline-int-divide-max-throughput
13141 @opindex minline-int-divide-max-throughput
13142 Generate code for inline divides of integer values
13143 using the maximum throughput algorithm.
13145 @item -mno-inline-int-divide
13146 @opindex mno-inline-int-divide
13147 Do not generate inline code for divides of integer values.
13149 @item -minline-sqrt-min-latency
13150 @opindex minline-sqrt-min-latency
13151 Generate code for inline square roots
13152 using the minimum latency algorithm.
13154 @item -minline-sqrt-max-throughput
13155 @opindex minline-sqrt-max-throughput
13156 Generate code for inline square roots
13157 using the maximum throughput algorithm.
13159 @item -mno-inline-sqrt
13160 @opindex mno-inline-sqrt
13161 Do not generate inline code for sqrt.
13164 @itemx -mno-fused-madd
13165 @opindex mfused-madd
13166 @opindex mno-fused-madd
13167 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13168 instructions. The default is to use these instructions.
13170 @item -mno-dwarf2-asm
13171 @itemx -mdwarf2-asm
13172 @opindex mno-dwarf2-asm
13173 @opindex mdwarf2-asm
13174 Don't (or do) generate assembler code for the DWARF2 line number debugging
13175 info. This may be useful when not using the GNU assembler.
13177 @item -mearly-stop-bits
13178 @itemx -mno-early-stop-bits
13179 @opindex mearly-stop-bits
13180 @opindex mno-early-stop-bits
13181 Allow stop bits to be placed earlier than immediately preceding the
13182 instruction that triggered the stop bit. This can improve instruction
13183 scheduling, but does not always do so.
13185 @item -mfixed-range=@var{register-range}
13186 @opindex mfixed-range
13187 Generate code treating the given register range as fixed registers.
13188 A fixed register is one that the register allocator can not use. This is
13189 useful when compiling kernel code. A register range is specified as
13190 two registers separated by a dash. Multiple register ranges can be
13191 specified separated by a comma.
13193 @item -mtls-size=@var{tls-size}
13195 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13198 @item -mtune=@var{cpu-type}
13200 Tune the instruction scheduling for a particular CPU, Valid values are
13201 itanium, itanium1, merced, itanium2, and mckinley.
13207 Generate code for a 32-bit or 64-bit environment.
13208 The 32-bit environment sets int, long and pointer to 32 bits.
13209 The 64-bit environment sets int to 32 bits and long and pointer
13210 to 64 bits. These are HP-UX specific flags.
13212 @item -mno-sched-br-data-spec
13213 @itemx -msched-br-data-spec
13214 @opindex mno-sched-br-data-spec
13215 @opindex msched-br-data-spec
13216 (Dis/En)able data speculative scheduling before reload.
13217 This will result in generation of the ld.a instructions and
13218 the corresponding check instructions (ld.c / chk.a).
13219 The default is 'disable'.
13221 @item -msched-ar-data-spec
13222 @itemx -mno-sched-ar-data-spec
13223 @opindex msched-ar-data-spec
13224 @opindex mno-sched-ar-data-spec
13225 (En/Dis)able data speculative scheduling after reload.
13226 This will result in generation of the ld.a instructions and
13227 the corresponding check instructions (ld.c / chk.a).
13228 The default is 'enable'.
13230 @item -mno-sched-control-spec
13231 @itemx -msched-control-spec
13232 @opindex mno-sched-control-spec
13233 @opindex msched-control-spec
13234 (Dis/En)able control speculative scheduling. This feature is
13235 available only during region scheduling (i.e.@: before reload).
13236 This will result in generation of the ld.s instructions and
13237 the corresponding check instructions chk.s .
13238 The default is 'disable'.
13240 @item -msched-br-in-data-spec
13241 @itemx -mno-sched-br-in-data-spec
13242 @opindex msched-br-in-data-spec
13243 @opindex mno-sched-br-in-data-spec
13244 (En/Dis)able speculative scheduling of the instructions that
13245 are dependent on the data speculative loads before reload.
13246 This is effective only with @option{-msched-br-data-spec} enabled.
13247 The default is 'enable'.
13249 @item -msched-ar-in-data-spec
13250 @itemx -mno-sched-ar-in-data-spec
13251 @opindex msched-ar-in-data-spec
13252 @opindex mno-sched-ar-in-data-spec
13253 (En/Dis)able speculative scheduling of the instructions that
13254 are dependent on the data speculative loads after reload.
13255 This is effective only with @option{-msched-ar-data-spec} enabled.
13256 The default is 'enable'.
13258 @item -msched-in-control-spec
13259 @itemx -mno-sched-in-control-spec
13260 @opindex msched-in-control-spec
13261 @opindex mno-sched-in-control-spec
13262 (En/Dis)able speculative scheduling of the instructions that
13263 are dependent on the control speculative loads.
13264 This is effective only with @option{-msched-control-spec} enabled.
13265 The default is 'enable'.
13267 @item -mno-sched-prefer-non-data-spec-insns
13268 @itemx -msched-prefer-non-data-spec-insns
13269 @opindex mno-sched-prefer-non-data-spec-insns
13270 @opindex msched-prefer-non-data-spec-insns
13271 If enabled, data speculative instructions will be chosen for schedule
13272 only if there are no other choices at the moment. This will make
13273 the use of the data speculation much more conservative.
13274 The default is 'disable'.
13276 @item -mno-sched-prefer-non-control-spec-insns
13277 @itemx -msched-prefer-non-control-spec-insns
13278 @opindex mno-sched-prefer-non-control-spec-insns
13279 @opindex msched-prefer-non-control-spec-insns
13280 If enabled, control speculative instructions will be chosen for schedule
13281 only if there are no other choices at the moment. This will make
13282 the use of the control speculation much more conservative.
13283 The default is 'disable'.
13285 @item -mno-sched-count-spec-in-critical-path
13286 @itemx -msched-count-spec-in-critical-path
13287 @opindex mno-sched-count-spec-in-critical-path
13288 @opindex msched-count-spec-in-critical-path
13289 If enabled, speculative dependencies will be considered during
13290 computation of the instructions priorities. This will make the use of the
13291 speculation a bit more conservative.
13292 The default is 'disable'.
13294 @item -msched-spec-ldc
13295 @opindex msched-spec-ldc
13296 Use a simple data speculation check. This option is on by default.
13298 @item -msched-control-spec-ldc
13299 @opindex msched-spec-ldc
13300 Use a simple check for control speculation. This option is on by default.
13302 @item -msched-stop-bits-after-every-cycle
13303 @opindex msched-stop-bits-after-every-cycle
13304 Place a stop bit after every cycle when scheduling. This option is on
13307 @item -msched-fp-mem-deps-zero-cost
13308 @opindex msched-fp-mem-deps-zero-cost
13309 Assume that floating-point stores and loads are not likely to cause a conflict
13310 when placed into the same instruction group. This option is disabled by
13313 @item -msel-sched-dont-check-control-spec
13314 @opindex msel-sched-dont-check-control-spec
13315 Generate checks for control speculation in selective scheduling.
13316 This flag is disabled by default.
13318 @item -msched-max-memory-insns=@var{max-insns}
13319 @opindex msched-max-memory-insns
13320 Limit on the number of memory insns per instruction group, giving lower
13321 priority to subsequent memory insns attempting to schedule in the same
13322 instruction group. Frequently useful to prevent cache bank conflicts.
13323 The default value is 1.
13325 @item -msched-max-memory-insns-hard-limit
13326 @opindex msched-max-memory-insns-hard-limit
13327 Disallow more than `msched-max-memory-insns' in instruction group.
13328 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13329 when limit is reached but may still schedule memory operations.
13333 @node IA-64/VMS Options
13334 @subsection IA-64/VMS Options
13336 These @samp{-m} options are defined for the IA-64/VMS implementations:
13339 @item -mvms-return-codes
13340 @opindex mvms-return-codes
13341 Return VMS condition codes from main. The default is to return POSIX
13342 style condition (e.g.@ error) codes.
13344 @item -mdebug-main=@var{prefix}
13345 @opindex mdebug-main=@var{prefix}
13346 Flag the first routine whose name starts with @var{prefix} as the main
13347 routine for the debugger.
13351 Default to 64bit memory allocation routines.
13355 @subsection LM32 Options
13356 @cindex LM32 options
13358 These @option{-m} options are defined for the Lattice Mico32 architecture:
13361 @item -mbarrel-shift-enabled
13362 @opindex mbarrel-shift-enabled
13363 Enable barrel-shift instructions.
13365 @item -mdivide-enabled
13366 @opindex mdivide-enabled
13367 Enable divide and modulus instructions.
13369 @item -mmultiply-enabled
13370 @opindex multiply-enabled
13371 Enable multiply instructions.
13373 @item -msign-extend-enabled
13374 @opindex msign-extend-enabled
13375 Enable sign extend instructions.
13377 @item -muser-enabled
13378 @opindex muser-enabled
13379 Enable user-defined instructions.
13384 @subsection M32C Options
13385 @cindex M32C options
13388 @item -mcpu=@var{name}
13390 Select the CPU for which code is generated. @var{name} may be one of
13391 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13392 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13393 the M32C/80 series.
13397 Specifies that the program will be run on the simulator. This causes
13398 an alternate runtime library to be linked in which supports, for
13399 example, file I/O@. You must not use this option when generating
13400 programs that will run on real hardware; you must provide your own
13401 runtime library for whatever I/O functions are needed.
13403 @item -memregs=@var{number}
13405 Specifies the number of memory-based pseudo-registers GCC will use
13406 during code generation. These pseudo-registers will be used like real
13407 registers, so there is a tradeoff between GCC's ability to fit the
13408 code into available registers, and the performance penalty of using
13409 memory instead of registers. Note that all modules in a program must
13410 be compiled with the same value for this option. Because of that, you
13411 must not use this option with the default runtime libraries gcc
13416 @node M32R/D Options
13417 @subsection M32R/D Options
13418 @cindex M32R/D options
13420 These @option{-m} options are defined for Renesas M32R/D architectures:
13425 Generate code for the M32R/2@.
13429 Generate code for the M32R/X@.
13433 Generate code for the M32R@. This is the default.
13435 @item -mmodel=small
13436 @opindex mmodel=small
13437 Assume all objects live in the lower 16MB of memory (so that their addresses
13438 can be loaded with the @code{ld24} instruction), and assume all subroutines
13439 are reachable with the @code{bl} instruction.
13440 This is the default.
13442 The addressability of a particular object can be set with the
13443 @code{model} attribute.
13445 @item -mmodel=medium
13446 @opindex mmodel=medium
13447 Assume objects may be anywhere in the 32-bit address space (the compiler
13448 will generate @code{seth/add3} instructions to load their addresses), and
13449 assume all subroutines are reachable with the @code{bl} instruction.
13451 @item -mmodel=large
13452 @opindex mmodel=large
13453 Assume objects may be anywhere in the 32-bit address space (the compiler
13454 will generate @code{seth/add3} instructions to load their addresses), and
13455 assume subroutines may not be reachable with the @code{bl} instruction
13456 (the compiler will generate the much slower @code{seth/add3/jl}
13457 instruction sequence).
13460 @opindex msdata=none
13461 Disable use of the small data area. Variables will be put into
13462 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13463 @code{section} attribute has been specified).
13464 This is the default.
13466 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13467 Objects may be explicitly put in the small data area with the
13468 @code{section} attribute using one of these sections.
13470 @item -msdata=sdata
13471 @opindex msdata=sdata
13472 Put small global and static data in the small data area, but do not
13473 generate special code to reference them.
13476 @opindex msdata=use
13477 Put small global and static data in the small data area, and generate
13478 special instructions to reference them.
13482 @cindex smaller data references
13483 Put global and static objects less than or equal to @var{num} bytes
13484 into the small data or bss sections instead of the normal data or bss
13485 sections. The default value of @var{num} is 8.
13486 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13487 for this option to have any effect.
13489 All modules should be compiled with the same @option{-G @var{num}} value.
13490 Compiling with different values of @var{num} may or may not work; if it
13491 doesn't the linker will give an error message---incorrect code will not be
13496 Makes the M32R specific code in the compiler display some statistics
13497 that might help in debugging programs.
13499 @item -malign-loops
13500 @opindex malign-loops
13501 Align all loops to a 32-byte boundary.
13503 @item -mno-align-loops
13504 @opindex mno-align-loops
13505 Do not enforce a 32-byte alignment for loops. This is the default.
13507 @item -missue-rate=@var{number}
13508 @opindex missue-rate=@var{number}
13509 Issue @var{number} instructions per cycle. @var{number} can only be 1
13512 @item -mbranch-cost=@var{number}
13513 @opindex mbranch-cost=@var{number}
13514 @var{number} can only be 1 or 2. If it is 1 then branches will be
13515 preferred over conditional code, if it is 2, then the opposite will
13518 @item -mflush-trap=@var{number}
13519 @opindex mflush-trap=@var{number}
13520 Specifies the trap number to use to flush the cache. The default is
13521 12. Valid numbers are between 0 and 15 inclusive.
13523 @item -mno-flush-trap
13524 @opindex mno-flush-trap
13525 Specifies that the cache cannot be flushed by using a trap.
13527 @item -mflush-func=@var{name}
13528 @opindex mflush-func=@var{name}
13529 Specifies the name of the operating system function to call to flush
13530 the cache. The default is @emph{_flush_cache}, but a function call
13531 will only be used if a trap is not available.
13533 @item -mno-flush-func
13534 @opindex mno-flush-func
13535 Indicates that there is no OS function for flushing the cache.
13539 @node M680x0 Options
13540 @subsection M680x0 Options
13541 @cindex M680x0 options
13543 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13544 The default settings depend on which architecture was selected when
13545 the compiler was configured; the defaults for the most common choices
13549 @item -march=@var{arch}
13551 Generate code for a specific M680x0 or ColdFire instruction set
13552 architecture. Permissible values of @var{arch} for M680x0
13553 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13554 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13555 architectures are selected according to Freescale's ISA classification
13556 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13557 @samp{isab} and @samp{isac}.
13559 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13560 code for a ColdFire target. The @var{arch} in this macro is one of the
13561 @option{-march} arguments given above.
13563 When used together, @option{-march} and @option{-mtune} select code
13564 that runs on a family of similar processors but that is optimized
13565 for a particular microarchitecture.
13567 @item -mcpu=@var{cpu}
13569 Generate code for a specific M680x0 or ColdFire processor.
13570 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13571 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13572 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13573 below, which also classifies the CPUs into families:
13575 @multitable @columnfractions 0.20 0.80
13576 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13577 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13578 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13579 @item @samp{5206e} @tab @samp{5206e}
13580 @item @samp{5208} @tab @samp{5207} @samp{5208}
13581 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13582 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13583 @item @samp{5216} @tab @samp{5214} @samp{5216}
13584 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13585 @item @samp{5225} @tab @samp{5224} @samp{5225}
13586 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13587 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13588 @item @samp{5249} @tab @samp{5249}
13589 @item @samp{5250} @tab @samp{5250}
13590 @item @samp{5271} @tab @samp{5270} @samp{5271}
13591 @item @samp{5272} @tab @samp{5272}
13592 @item @samp{5275} @tab @samp{5274} @samp{5275}
13593 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13594 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13595 @item @samp{5307} @tab @samp{5307}
13596 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13597 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13598 @item @samp{5407} @tab @samp{5407}
13599 @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}
13602 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13603 @var{arch} is compatible with @var{cpu}. Other combinations of
13604 @option{-mcpu} and @option{-march} are rejected.
13606 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13607 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13608 where the value of @var{family} is given by the table above.
13610 @item -mtune=@var{tune}
13612 Tune the code for a particular microarchitecture, within the
13613 constraints set by @option{-march} and @option{-mcpu}.
13614 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13615 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13616 and @samp{cpu32}. The ColdFire microarchitectures
13617 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13619 You can also use @option{-mtune=68020-40} for code that needs
13620 to run relatively well on 68020, 68030 and 68040 targets.
13621 @option{-mtune=68020-60} is similar but includes 68060 targets
13622 as well. These two options select the same tuning decisions as
13623 @option{-m68020-40} and @option{-m68020-60} respectively.
13625 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13626 when tuning for 680x0 architecture @var{arch}. It also defines
13627 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13628 option is used. If gcc is tuning for a range of architectures,
13629 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13630 it defines the macros for every architecture in the range.
13632 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13633 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13634 of the arguments given above.
13640 Generate output for a 68000. This is the default
13641 when the compiler is configured for 68000-based systems.
13642 It is equivalent to @option{-march=68000}.
13644 Use this option for microcontrollers with a 68000 or EC000 core,
13645 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13649 Generate output for a 68010. This is the default
13650 when the compiler is configured for 68010-based systems.
13651 It is equivalent to @option{-march=68010}.
13657 Generate output for a 68020. This is the default
13658 when the compiler is configured for 68020-based systems.
13659 It is equivalent to @option{-march=68020}.
13663 Generate output for a 68030. This is the default when the compiler is
13664 configured for 68030-based systems. It is equivalent to
13665 @option{-march=68030}.
13669 Generate output for a 68040. This is the default when the compiler is
13670 configured for 68040-based systems. It is equivalent to
13671 @option{-march=68040}.
13673 This option inhibits the use of 68881/68882 instructions that have to be
13674 emulated by software on the 68040. Use this option if your 68040 does not
13675 have code to emulate those instructions.
13679 Generate output for a 68060. This is the default when the compiler is
13680 configured for 68060-based systems. It is equivalent to
13681 @option{-march=68060}.
13683 This option inhibits the use of 68020 and 68881/68882 instructions that
13684 have to be emulated by software on the 68060. Use this option if your 68060
13685 does not have code to emulate those instructions.
13689 Generate output for a CPU32. This is the default
13690 when the compiler is configured for CPU32-based systems.
13691 It is equivalent to @option{-march=cpu32}.
13693 Use this option for microcontrollers with a
13694 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13695 68336, 68340, 68341, 68349 and 68360.
13699 Generate output for a 520X ColdFire CPU@. This is the default
13700 when the compiler is configured for 520X-based systems.
13701 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13702 in favor of that option.
13704 Use this option for microcontroller with a 5200 core, including
13705 the MCF5202, MCF5203, MCF5204 and MCF5206.
13709 Generate output for a 5206e ColdFire CPU@. The option is now
13710 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13714 Generate output for a member of the ColdFire 528X family.
13715 The option is now deprecated in favor of the equivalent
13716 @option{-mcpu=528x}.
13720 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13721 in favor of the equivalent @option{-mcpu=5307}.
13725 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13726 in favor of the equivalent @option{-mcpu=5407}.
13730 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13731 This includes use of hardware floating point instructions.
13732 The option is equivalent to @option{-mcpu=547x}, and is now
13733 deprecated in favor of that option.
13737 Generate output for a 68040, without using any of the new instructions.
13738 This results in code which can run relatively efficiently on either a
13739 68020/68881 or a 68030 or a 68040. The generated code does use the
13740 68881 instructions that are emulated on the 68040.
13742 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13746 Generate output for a 68060, without using any of the new instructions.
13747 This results in code which can run relatively efficiently on either a
13748 68020/68881 or a 68030 or a 68040. The generated code does use the
13749 68881 instructions that are emulated on the 68060.
13751 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13755 @opindex mhard-float
13757 Generate floating-point instructions. This is the default for 68020
13758 and above, and for ColdFire devices that have an FPU@. It defines the
13759 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13760 on ColdFire targets.
13763 @opindex msoft-float
13764 Do not generate floating-point instructions; use library calls instead.
13765 This is the default for 68000, 68010, and 68832 targets. It is also
13766 the default for ColdFire devices that have no FPU.
13772 Generate (do not generate) ColdFire hardware divide and remainder
13773 instructions. If @option{-march} is used without @option{-mcpu},
13774 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13775 architectures. Otherwise, the default is taken from the target CPU
13776 (either the default CPU, or the one specified by @option{-mcpu}). For
13777 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13778 @option{-mcpu=5206e}.
13780 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13784 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13785 Additionally, parameters passed on the stack are also aligned to a
13786 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13790 Do not consider type @code{int} to be 16 bits wide. This is the default.
13793 @itemx -mno-bitfield
13794 @opindex mnobitfield
13795 @opindex mno-bitfield
13796 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13797 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13801 Do use the bit-field instructions. The @option{-m68020} option implies
13802 @option{-mbitfield}. This is the default if you use a configuration
13803 designed for a 68020.
13807 Use a different function-calling convention, in which functions
13808 that take a fixed number of arguments return with the @code{rtd}
13809 instruction, which pops their arguments while returning. This
13810 saves one instruction in the caller since there is no need to pop
13811 the arguments there.
13813 This calling convention is incompatible with the one normally
13814 used on Unix, so you cannot use it if you need to call libraries
13815 compiled with the Unix compiler.
13817 Also, you must provide function prototypes for all functions that
13818 take variable numbers of arguments (including @code{printf});
13819 otherwise incorrect code will be generated for calls to those
13822 In addition, seriously incorrect code will result if you call a
13823 function with too many arguments. (Normally, extra arguments are
13824 harmlessly ignored.)
13826 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13827 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13831 Do not use the calling conventions selected by @option{-mrtd}.
13832 This is the default.
13835 @itemx -mno-align-int
13836 @opindex malign-int
13837 @opindex mno-align-int
13838 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13839 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13840 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13841 Aligning variables on 32-bit boundaries produces code that runs somewhat
13842 faster on processors with 32-bit busses at the expense of more memory.
13844 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13845 align structures containing the above types differently than
13846 most published application binary interface specifications for the m68k.
13850 Use the pc-relative addressing mode of the 68000 directly, instead of
13851 using a global offset table. At present, this option implies @option{-fpic},
13852 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13853 not presently supported with @option{-mpcrel}, though this could be supported for
13854 68020 and higher processors.
13856 @item -mno-strict-align
13857 @itemx -mstrict-align
13858 @opindex mno-strict-align
13859 @opindex mstrict-align
13860 Do not (do) assume that unaligned memory references will be handled by
13864 Generate code that allows the data segment to be located in a different
13865 area of memory from the text segment. This allows for execute in place in
13866 an environment without virtual memory management. This option implies
13869 @item -mno-sep-data
13870 Generate code that assumes that the data segment follows the text segment.
13871 This is the default.
13873 @item -mid-shared-library
13874 Generate code that supports shared libraries via the library ID method.
13875 This allows for execute in place and shared libraries in an environment
13876 without virtual memory management. This option implies @option{-fPIC}.
13878 @item -mno-id-shared-library
13879 Generate code that doesn't assume ID based shared libraries are being used.
13880 This is the default.
13882 @item -mshared-library-id=n
13883 Specified the identification number of the ID based shared library being
13884 compiled. Specifying a value of 0 will generate more compact code, specifying
13885 other values will force the allocation of that number to the current
13886 library but is no more space or time efficient than omitting this option.
13892 When generating position-independent code for ColdFire, generate code
13893 that works if the GOT has more than 8192 entries. This code is
13894 larger and slower than code generated without this option. On M680x0
13895 processors, this option is not needed; @option{-fPIC} suffices.
13897 GCC normally uses a single instruction to load values from the GOT@.
13898 While this is relatively efficient, it only works if the GOT
13899 is smaller than about 64k. Anything larger causes the linker
13900 to report an error such as:
13902 @cindex relocation truncated to fit (ColdFire)
13904 relocation truncated to fit: R_68K_GOT16O foobar
13907 If this happens, you should recompile your code with @option{-mxgot}.
13908 It should then work with very large GOTs. However, code generated with
13909 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13910 the value of a global symbol.
13912 Note that some linkers, including newer versions of the GNU linker,
13913 can create multiple GOTs and sort GOT entries. If you have such a linker,
13914 you should only need to use @option{-mxgot} when compiling a single
13915 object file that accesses more than 8192 GOT entries. Very few do.
13917 These options have no effect unless GCC is generating
13918 position-independent code.
13922 @node MCore Options
13923 @subsection MCore Options
13924 @cindex MCore options
13926 These are the @samp{-m} options defined for the Motorola M*Core
13932 @itemx -mno-hardlit
13934 @opindex mno-hardlit
13935 Inline constants into the code stream if it can be done in two
13936 instructions or less.
13942 Use the divide instruction. (Enabled by default).
13944 @item -mrelax-immediate
13945 @itemx -mno-relax-immediate
13946 @opindex mrelax-immediate
13947 @opindex mno-relax-immediate
13948 Allow arbitrary sized immediates in bit operations.
13950 @item -mwide-bitfields
13951 @itemx -mno-wide-bitfields
13952 @opindex mwide-bitfields
13953 @opindex mno-wide-bitfields
13954 Always treat bit-fields as int-sized.
13956 @item -m4byte-functions
13957 @itemx -mno-4byte-functions
13958 @opindex m4byte-functions
13959 @opindex mno-4byte-functions
13960 Force all functions to be aligned to a four byte boundary.
13962 @item -mcallgraph-data
13963 @itemx -mno-callgraph-data
13964 @opindex mcallgraph-data
13965 @opindex mno-callgraph-data
13966 Emit callgraph information.
13969 @itemx -mno-slow-bytes
13970 @opindex mslow-bytes
13971 @opindex mno-slow-bytes
13972 Prefer word access when reading byte quantities.
13974 @item -mlittle-endian
13975 @itemx -mbig-endian
13976 @opindex mlittle-endian
13977 @opindex mbig-endian
13978 Generate code for a little endian target.
13984 Generate code for the 210 processor.
13988 Assume that run-time support has been provided and so omit the
13989 simulator library (@file{libsim.a)} from the linker command line.
13991 @item -mstack-increment=@var{size}
13992 @opindex mstack-increment
13993 Set the maximum amount for a single stack increment operation. Large
13994 values can increase the speed of programs which contain functions
13995 that need a large amount of stack space, but they can also trigger a
13996 segmentation fault if the stack is extended too much. The default
14002 @subsection MeP Options
14003 @cindex MeP options
14009 Enables the @code{abs} instruction, which is the absolute difference
14010 between two registers.
14014 Enables all the optional instructions - average, multiply, divide, bit
14015 operations, leading zero, absolute difference, min/max, clip, and
14021 Enables the @code{ave} instruction, which computes the average of two
14024 @item -mbased=@var{n}
14026 Variables of size @var{n} bytes or smaller will be placed in the
14027 @code{.based} section by default. Based variables use the @code{$tp}
14028 register as a base register, and there is a 128 byte limit to the
14029 @code{.based} section.
14033 Enables the bit operation instructions - bit test (@code{btstm}), set
14034 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14035 test-and-set (@code{tas}).
14037 @item -mc=@var{name}
14039 Selects which section constant data will be placed in. @var{name} may
14040 be @code{tiny}, @code{near}, or @code{far}.
14044 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14045 useful unless you also provide @code{-mminmax}.
14047 @item -mconfig=@var{name}
14049 Selects one of the build-in core configurations. Each MeP chip has
14050 one or more modules in it; each module has a core CPU and a variety of
14051 coprocessors, optional instructions, and peripherals. The
14052 @code{MeP-Integrator} tool, not part of GCC, provides these
14053 configurations through this option; using this option is the same as
14054 using all the corresponding command line options. The default
14055 configuration is @code{default}.
14059 Enables the coprocessor instructions. By default, this is a 32-bit
14060 coprocessor. Note that the coprocessor is normally enabled via the
14061 @code{-mconfig=} option.
14065 Enables the 32-bit coprocessor's instructions.
14069 Enables the 64-bit coprocessor's instructions.
14073 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14077 Causes constant variables to be placed in the @code{.near} section.
14081 Enables the @code{div} and @code{divu} instructions.
14085 Generate big-endian code.
14089 Generate little-endian code.
14091 @item -mio-volatile
14092 @opindex mio-volatile
14093 Tells the compiler that any variable marked with the @code{io}
14094 attribute is to be considered volatile.
14098 Causes variables to be assigned to the @code{.far} section by default.
14102 Enables the @code{leadz} (leading zero) instruction.
14106 Causes variables to be assigned to the @code{.near} section by default.
14110 Enables the @code{min} and @code{max} instructions.
14114 Enables the multiplication and multiply-accumulate instructions.
14118 Disables all the optional instructions enabled by @code{-mall-opts}.
14122 Enables the @code{repeat} and @code{erepeat} instructions, used for
14123 low-overhead looping.
14127 Causes all variables to default to the @code{.tiny} section. Note
14128 that there is a 65536 byte limit to this section. Accesses to these
14129 variables use the @code{%gp} base register.
14133 Enables the saturation instructions. Note that the compiler does not
14134 currently generate these itself, but this option is included for
14135 compatibility with other tools, like @code{as}.
14139 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14143 Link the simulator runtime libraries.
14147 Link the simulator runtime libraries, excluding built-in support
14148 for reset and exception vectors and tables.
14152 Causes all functions to default to the @code{.far} section. Without
14153 this option, functions default to the @code{.near} section.
14155 @item -mtiny=@var{n}
14157 Variables that are @var{n} bytes or smaller will be allocated to the
14158 @code{.tiny} section. These variables use the @code{$gp} base
14159 register. The default for this option is 4, but note that there's a
14160 65536 byte limit to the @code{.tiny} section.
14164 @node MicroBlaze Options
14165 @subsection MicroBlaze Options
14166 @cindex MicroBlaze Options
14171 @opindex msoft-float
14172 Use software emulation for floating point (default).
14175 @opindex mhard-float
14176 Use hardware floating point instructions.
14180 Do not optimize block moves, use @code{memcpy}.
14182 @item -mno-clearbss
14183 @opindex mno-clearbss
14184 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14186 @item -mcpu=@var{cpu-type}
14188 Use features of and schedule code for given CPU.
14189 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14190 where @var{X} is a major version, @var{YY} is the minor version, and
14191 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14192 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14194 @item -mxl-soft-mul
14195 @opindex mxl-soft-mul
14196 Use software multiply emulation (default).
14198 @item -mxl-soft-div
14199 @opindex mxl-soft-div
14200 Use software emulation for divides (default).
14202 @item -mxl-barrel-shift
14203 @opindex mxl-barrel-shift
14204 Use the hardware barrel shifter.
14206 @item -mxl-pattern-compare
14207 @opindex mxl-pattern-compare
14208 Use pattern compare instructions.
14210 @item -msmall-divides
14211 @opindex msmall-divides
14212 Use table lookup optimization for small signed integer divisions.
14214 @item -mxl-stack-check
14215 @opindex mxl-stack-check
14216 This option is deprecated. Use -fstack-check instead.
14219 @opindex mxl-gp-opt
14220 Use GP relative sdata/sbss sections.
14222 @item -mxl-multiply-high
14223 @opindex mxl-multiply-high
14224 Use multiply high instructions for high part of 32x32 multiply.
14226 @item -mxl-float-convert
14227 @opindex mxl-float-convert
14228 Use hardware floating point conversion instructions.
14230 @item -mxl-float-sqrt
14231 @opindex mxl-float-sqrt
14232 Use hardware floating point square root instruction.
14234 @item -mxl-mode-@var{app-model}
14235 Select application model @var{app-model}. Valid models are
14238 normal executable (default), uses startup code @file{crt0.o}.
14241 for use with Xilinx Microprocessor Debugger (XMD) based
14242 software intrusive debug agent called xmdstub. This uses startup file
14243 @file{crt1.o} and sets the start address of the program to be 0x800.
14246 for applications that are loaded using a bootloader.
14247 This model uses startup file @file{crt2.o} which does not contain a processor
14248 reset vector handler. This is suitable for transferring control on a
14249 processor reset to the bootloader rather than the application.
14252 for applications that do not require any of the
14253 MicroBlaze vectors. This option may be useful for applications running
14254 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14257 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14258 @option{-mxl-mode-@var{app-model}}.
14263 @subsection MIPS Options
14264 @cindex MIPS options
14270 Generate big-endian code.
14274 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14277 @item -march=@var{arch}
14279 Generate code that will run on @var{arch}, which can be the name of a
14280 generic MIPS ISA, or the name of a particular processor.
14282 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14283 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14284 The processor names are:
14285 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14286 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14287 @samp{5kc}, @samp{5kf},
14289 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14290 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14291 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14292 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14293 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14294 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14298 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14299 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14300 @samp{rm7000}, @samp{rm9000},
14301 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14304 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14305 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14307 The special value @samp{from-abi} selects the
14308 most compatible architecture for the selected ABI (that is,
14309 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14311 Native Linux/GNU toolchains also support the value @samp{native},
14312 which selects the best architecture option for the host processor.
14313 @option{-march=native} has no effect if GCC does not recognize
14316 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14317 (for example, @samp{-march=r2k}). Prefixes are optional, and
14318 @samp{vr} may be written @samp{r}.
14320 Names of the form @samp{@var{n}f2_1} refer to processors with
14321 FPUs clocked at half the rate of the core, names of the form
14322 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14323 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14324 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14325 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14326 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14327 accepted as synonyms for @samp{@var{n}f1_1}.
14329 GCC defines two macros based on the value of this option. The first
14330 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14331 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14332 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14333 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14334 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14336 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14337 above. In other words, it will have the full prefix and will not
14338 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14339 the macro names the resolved architecture (either @samp{"mips1"} or
14340 @samp{"mips3"}). It names the default architecture when no
14341 @option{-march} option is given.
14343 @item -mtune=@var{arch}
14345 Optimize for @var{arch}. Among other things, this option controls
14346 the way instructions are scheduled, and the perceived cost of arithmetic
14347 operations. The list of @var{arch} values is the same as for
14350 When this option is not used, GCC will optimize for the processor
14351 specified by @option{-march}. By using @option{-march} and
14352 @option{-mtune} together, it is possible to generate code that will
14353 run on a family of processors, but optimize the code for one
14354 particular member of that family.
14356 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14357 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14358 @samp{-march} ones described above.
14362 Equivalent to @samp{-march=mips1}.
14366 Equivalent to @samp{-march=mips2}.
14370 Equivalent to @samp{-march=mips3}.
14374 Equivalent to @samp{-march=mips4}.
14378 Equivalent to @samp{-march=mips32}.
14382 Equivalent to @samp{-march=mips32r2}.
14386 Equivalent to @samp{-march=mips64}.
14390 Equivalent to @samp{-march=mips64r2}.
14395 @opindex mno-mips16
14396 Generate (do not generate) MIPS16 code. If GCC is targetting a
14397 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14399 MIPS16 code generation can also be controlled on a per-function basis
14400 by means of @code{mips16} and @code{nomips16} attributes.
14401 @xref{Function Attributes}, for more information.
14403 @item -mflip-mips16
14404 @opindex mflip-mips16
14405 Generate MIPS16 code on alternating functions. This option is provided
14406 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14407 not intended for ordinary use in compiling user code.
14409 @item -minterlink-mips16
14410 @itemx -mno-interlink-mips16
14411 @opindex minterlink-mips16
14412 @opindex mno-interlink-mips16
14413 Require (do not require) that non-MIPS16 code be link-compatible with
14416 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14417 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14418 therefore disables direct jumps unless GCC knows that the target of the
14419 jump is not MIPS16.
14431 Generate code for the given ABI@.
14433 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14434 generates 64-bit code when you select a 64-bit architecture, but you
14435 can use @option{-mgp32} to get 32-bit code instead.
14437 For information about the O64 ABI, see
14438 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14440 GCC supports a variant of the o32 ABI in which floating-point registers
14441 are 64 rather than 32 bits wide. You can select this combination with
14442 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14443 and @samp{mfhc1} instructions and is therefore only supported for
14444 MIPS32R2 processors.
14446 The register assignments for arguments and return values remain the
14447 same, but each scalar value is passed in a single 64-bit register
14448 rather than a pair of 32-bit registers. For example, scalar
14449 floating-point values are returned in @samp{$f0} only, not a
14450 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14451 remains the same, but all 64 bits are saved.
14454 @itemx -mno-abicalls
14456 @opindex mno-abicalls
14457 Generate (do not generate) code that is suitable for SVR4-style
14458 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14463 Generate (do not generate) code that is fully position-independent,
14464 and that can therefore be linked into shared libraries. This option
14465 only affects @option{-mabicalls}.
14467 All @option{-mabicalls} code has traditionally been position-independent,
14468 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14469 as an extension, the GNU toolchain allows executables to use absolute
14470 accesses for locally-binding symbols. It can also use shorter GP
14471 initialization sequences and generate direct calls to locally-defined
14472 functions. This mode is selected by @option{-mno-shared}.
14474 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14475 objects that can only be linked by the GNU linker. However, the option
14476 does not affect the ABI of the final executable; it only affects the ABI
14477 of relocatable objects. Using @option{-mno-shared} will generally make
14478 executables both smaller and quicker.
14480 @option{-mshared} is the default.
14486 Assume (do not assume) that the static and dynamic linkers
14487 support PLTs and copy relocations. This option only affects
14488 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14489 has no effect without @samp{-msym32}.
14491 You can make @option{-mplt} the default by configuring
14492 GCC with @option{--with-mips-plt}. The default is
14493 @option{-mno-plt} otherwise.
14499 Lift (do not lift) the usual restrictions on the size of the global
14502 GCC normally uses a single instruction to load values from the GOT@.
14503 While this is relatively efficient, it will only work if the GOT
14504 is smaller than about 64k. Anything larger will cause the linker
14505 to report an error such as:
14507 @cindex relocation truncated to fit (MIPS)
14509 relocation truncated to fit: R_MIPS_GOT16 foobar
14512 If this happens, you should recompile your code with @option{-mxgot}.
14513 It should then work with very large GOTs, although it will also be
14514 less efficient, since it will take three instructions to fetch the
14515 value of a global symbol.
14517 Note that some linkers can create multiple GOTs. If you have such a
14518 linker, you should only need to use @option{-mxgot} when a single object
14519 file accesses more than 64k's worth of GOT entries. Very few do.
14521 These options have no effect unless GCC is generating position
14526 Assume that general-purpose registers are 32 bits wide.
14530 Assume that general-purpose registers are 64 bits wide.
14534 Assume that floating-point registers are 32 bits wide.
14538 Assume that floating-point registers are 64 bits wide.
14541 @opindex mhard-float
14542 Use floating-point coprocessor instructions.
14545 @opindex msoft-float
14546 Do not use floating-point coprocessor instructions. Implement
14547 floating-point calculations using library calls instead.
14549 @item -msingle-float
14550 @opindex msingle-float
14551 Assume that the floating-point coprocessor only supports single-precision
14554 @item -mdouble-float
14555 @opindex mdouble-float
14556 Assume that the floating-point coprocessor supports double-precision
14557 operations. This is the default.
14563 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14564 implement atomic memory built-in functions. When neither option is
14565 specified, GCC will use the instructions if the target architecture
14568 @option{-mllsc} is useful if the runtime environment can emulate the
14569 instructions and @option{-mno-llsc} can be useful when compiling for
14570 nonstandard ISAs. You can make either option the default by
14571 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14572 respectively. @option{--with-llsc} is the default for some
14573 configurations; see the installation documentation for details.
14579 Use (do not use) revision 1 of the MIPS DSP ASE@.
14580 @xref{MIPS DSP Built-in Functions}. This option defines the
14581 preprocessor macro @samp{__mips_dsp}. It also defines
14582 @samp{__mips_dsp_rev} to 1.
14588 Use (do not use) revision 2 of the MIPS DSP ASE@.
14589 @xref{MIPS DSP Built-in Functions}. This option defines the
14590 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14591 It also defines @samp{__mips_dsp_rev} to 2.
14594 @itemx -mno-smartmips
14595 @opindex msmartmips
14596 @opindex mno-smartmips
14597 Use (do not use) the MIPS SmartMIPS ASE.
14599 @item -mpaired-single
14600 @itemx -mno-paired-single
14601 @opindex mpaired-single
14602 @opindex mno-paired-single
14603 Use (do not use) paired-single floating-point instructions.
14604 @xref{MIPS Paired-Single Support}. This option requires
14605 hardware floating-point support to be enabled.
14611 Use (do not use) MIPS Digital Media Extension instructions.
14612 This option can only be used when generating 64-bit code and requires
14613 hardware floating-point support to be enabled.
14618 @opindex mno-mips3d
14619 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14620 The option @option{-mips3d} implies @option{-mpaired-single}.
14626 Use (do not use) MT Multithreading instructions.
14630 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14631 an explanation of the default and the way that the pointer size is
14636 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14638 The default size of @code{int}s, @code{long}s and pointers depends on
14639 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14640 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14641 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14642 or the same size as integer registers, whichever is smaller.
14648 Assume (do not assume) that all symbols have 32-bit values, regardless
14649 of the selected ABI@. This option is useful in combination with
14650 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14651 to generate shorter and faster references to symbolic addresses.
14655 Put definitions of externally-visible data in a small data section
14656 if that data is no bigger than @var{num} bytes. GCC can then access
14657 the data more efficiently; see @option{-mgpopt} for details.
14659 The default @option{-G} option depends on the configuration.
14661 @item -mlocal-sdata
14662 @itemx -mno-local-sdata
14663 @opindex mlocal-sdata
14664 @opindex mno-local-sdata
14665 Extend (do not extend) the @option{-G} behavior to local data too,
14666 such as to static variables in C@. @option{-mlocal-sdata} is the
14667 default for all configurations.
14669 If the linker complains that an application is using too much small data,
14670 you might want to try rebuilding the less performance-critical parts with
14671 @option{-mno-local-sdata}. You might also want to build large
14672 libraries with @option{-mno-local-sdata}, so that the libraries leave
14673 more room for the main program.
14675 @item -mextern-sdata
14676 @itemx -mno-extern-sdata
14677 @opindex mextern-sdata
14678 @opindex mno-extern-sdata
14679 Assume (do not assume) that externally-defined data will be in
14680 a small data section if that data is within the @option{-G} limit.
14681 @option{-mextern-sdata} is the default for all configurations.
14683 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14684 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14685 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14686 is placed in a small data section. If @var{Var} is defined by another
14687 module, you must either compile that module with a high-enough
14688 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14689 definition. If @var{Var} is common, you must link the application
14690 with a high-enough @option{-G} setting.
14692 The easiest way of satisfying these restrictions is to compile
14693 and link every module with the same @option{-G} option. However,
14694 you may wish to build a library that supports several different
14695 small data limits. You can do this by compiling the library with
14696 the highest supported @option{-G} setting and additionally using
14697 @option{-mno-extern-sdata} to stop the library from making assumptions
14698 about externally-defined data.
14704 Use (do not use) GP-relative accesses for symbols that are known to be
14705 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14706 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14709 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14710 might not hold the value of @code{_gp}. For example, if the code is
14711 part of a library that might be used in a boot monitor, programs that
14712 call boot monitor routines will pass an unknown value in @code{$gp}.
14713 (In such situations, the boot monitor itself would usually be compiled
14714 with @option{-G0}.)
14716 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14717 @option{-mno-extern-sdata}.
14719 @item -membedded-data
14720 @itemx -mno-embedded-data
14721 @opindex membedded-data
14722 @opindex mno-embedded-data
14723 Allocate variables to the read-only data section first if possible, then
14724 next in the small data section if possible, otherwise in data. This gives
14725 slightly slower code than the default, but reduces the amount of RAM required
14726 when executing, and thus may be preferred for some embedded systems.
14728 @item -muninit-const-in-rodata
14729 @itemx -mno-uninit-const-in-rodata
14730 @opindex muninit-const-in-rodata
14731 @opindex mno-uninit-const-in-rodata
14732 Put uninitialized @code{const} variables in the read-only data section.
14733 This option is only meaningful in conjunction with @option{-membedded-data}.
14735 @item -mcode-readable=@var{setting}
14736 @opindex mcode-readable
14737 Specify whether GCC may generate code that reads from executable sections.
14738 There are three possible settings:
14741 @item -mcode-readable=yes
14742 Instructions may freely access executable sections. This is the
14745 @item -mcode-readable=pcrel
14746 MIPS16 PC-relative load instructions can access executable sections,
14747 but other instructions must not do so. This option is useful on 4KSc
14748 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14749 It is also useful on processors that can be configured to have a dual
14750 instruction/data SRAM interface and that, like the M4K, automatically
14751 redirect PC-relative loads to the instruction RAM.
14753 @item -mcode-readable=no
14754 Instructions must not access executable sections. This option can be
14755 useful on targets that are configured to have a dual instruction/data
14756 SRAM interface but that (unlike the M4K) do not automatically redirect
14757 PC-relative loads to the instruction RAM.
14760 @item -msplit-addresses
14761 @itemx -mno-split-addresses
14762 @opindex msplit-addresses
14763 @opindex mno-split-addresses
14764 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14765 relocation operators. This option has been superseded by
14766 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14768 @item -mexplicit-relocs
14769 @itemx -mno-explicit-relocs
14770 @opindex mexplicit-relocs
14771 @opindex mno-explicit-relocs
14772 Use (do not use) assembler relocation operators when dealing with symbolic
14773 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14774 is to use assembler macros instead.
14776 @option{-mexplicit-relocs} is the default if GCC was configured
14777 to use an assembler that supports relocation operators.
14779 @item -mcheck-zero-division
14780 @itemx -mno-check-zero-division
14781 @opindex mcheck-zero-division
14782 @opindex mno-check-zero-division
14783 Trap (do not trap) on integer division by zero.
14785 The default is @option{-mcheck-zero-division}.
14787 @item -mdivide-traps
14788 @itemx -mdivide-breaks
14789 @opindex mdivide-traps
14790 @opindex mdivide-breaks
14791 MIPS systems check for division by zero by generating either a
14792 conditional trap or a break instruction. Using traps results in
14793 smaller code, but is only supported on MIPS II and later. Also, some
14794 versions of the Linux kernel have a bug that prevents trap from
14795 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14796 allow conditional traps on architectures that support them and
14797 @option{-mdivide-breaks} to force the use of breaks.
14799 The default is usually @option{-mdivide-traps}, but this can be
14800 overridden at configure time using @option{--with-divide=breaks}.
14801 Divide-by-zero checks can be completely disabled using
14802 @option{-mno-check-zero-division}.
14807 @opindex mno-memcpy
14808 Force (do not force) the use of @code{memcpy()} for non-trivial block
14809 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14810 most constant-sized copies.
14813 @itemx -mno-long-calls
14814 @opindex mlong-calls
14815 @opindex mno-long-calls
14816 Disable (do not disable) use of the @code{jal} instruction. Calling
14817 functions using @code{jal} is more efficient but requires the caller
14818 and callee to be in the same 256 megabyte segment.
14820 This option has no effect on abicalls code. The default is
14821 @option{-mno-long-calls}.
14827 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14828 instructions, as provided by the R4650 ISA@.
14831 @itemx -mno-fused-madd
14832 @opindex mfused-madd
14833 @opindex mno-fused-madd
14834 Enable (disable) use of the floating point multiply-accumulate
14835 instructions, when they are available. The default is
14836 @option{-mfused-madd}.
14838 When multiply-accumulate instructions are used, the intermediate
14839 product is calculated to infinite precision and is not subject to
14840 the FCSR Flush to Zero bit. This may be undesirable in some
14845 Tell the MIPS assembler to not run its preprocessor over user
14846 assembler files (with a @samp{.s} suffix) when assembling them.
14851 @opindex mno-fix-24k
14852 Work around the 24K E48 (lost data on stores during refill) errata.
14853 The workarounds are implemented by the assembler rather than by GCC.
14856 @itemx -mno-fix-r4000
14857 @opindex mfix-r4000
14858 @opindex mno-fix-r4000
14859 Work around certain R4000 CPU errata:
14862 A double-word or a variable shift may give an incorrect result if executed
14863 immediately after starting an integer division.
14865 A double-word or a variable shift may give an incorrect result if executed
14866 while an integer multiplication is in progress.
14868 An integer division may give an incorrect result if started in a delay slot
14869 of a taken branch or a jump.
14873 @itemx -mno-fix-r4400
14874 @opindex mfix-r4400
14875 @opindex mno-fix-r4400
14876 Work around certain R4400 CPU errata:
14879 A double-word or a variable shift may give an incorrect result if executed
14880 immediately after starting an integer division.
14884 @itemx -mno-fix-r10000
14885 @opindex mfix-r10000
14886 @opindex mno-fix-r10000
14887 Work around certain R10000 errata:
14890 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14891 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14894 This option can only be used if the target architecture supports
14895 branch-likely instructions. @option{-mfix-r10000} is the default when
14896 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14900 @itemx -mno-fix-vr4120
14901 @opindex mfix-vr4120
14902 Work around certain VR4120 errata:
14905 @code{dmultu} does not always produce the correct result.
14907 @code{div} and @code{ddiv} do not always produce the correct result if one
14908 of the operands is negative.
14910 The workarounds for the division errata rely on special functions in
14911 @file{libgcc.a}. At present, these functions are only provided by
14912 the @code{mips64vr*-elf} configurations.
14914 Other VR4120 errata require a nop to be inserted between certain pairs of
14915 instructions. These errata are handled by the assembler, not by GCC itself.
14918 @opindex mfix-vr4130
14919 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14920 workarounds are implemented by the assembler rather than by GCC,
14921 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14922 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14923 instructions are available instead.
14926 @itemx -mno-fix-sb1
14928 Work around certain SB-1 CPU core errata.
14929 (This flag currently works around the SB-1 revision 2
14930 ``F1'' and ``F2'' floating point errata.)
14932 @item -mr10k-cache-barrier=@var{setting}
14933 @opindex mr10k-cache-barrier
14934 Specify whether GCC should insert cache barriers to avoid the
14935 side-effects of speculation on R10K processors.
14937 In common with many processors, the R10K tries to predict the outcome
14938 of a conditional branch and speculatively executes instructions from
14939 the ``taken'' branch. It later aborts these instructions if the
14940 predicted outcome was wrong. However, on the R10K, even aborted
14941 instructions can have side effects.
14943 This problem only affects kernel stores and, depending on the system,
14944 kernel loads. As an example, a speculatively-executed store may load
14945 the target memory into cache and mark the cache line as dirty, even if
14946 the store itself is later aborted. If a DMA operation writes to the
14947 same area of memory before the ``dirty'' line is flushed, the cached
14948 data will overwrite the DMA-ed data. See the R10K processor manual
14949 for a full description, including other potential problems.
14951 One workaround is to insert cache barrier instructions before every memory
14952 access that might be speculatively executed and that might have side
14953 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14954 controls GCC's implementation of this workaround. It assumes that
14955 aborted accesses to any byte in the following regions will not have
14960 the memory occupied by the current function's stack frame;
14963 the memory occupied by an incoming stack argument;
14966 the memory occupied by an object with a link-time-constant address.
14969 It is the kernel's responsibility to ensure that speculative
14970 accesses to these regions are indeed safe.
14972 If the input program contains a function declaration such as:
14978 then the implementation of @code{foo} must allow @code{j foo} and
14979 @code{jal foo} to be executed speculatively. GCC honors this
14980 restriction for functions it compiles itself. It expects non-GCC
14981 functions (such as hand-written assembly code) to do the same.
14983 The option has three forms:
14986 @item -mr10k-cache-barrier=load-store
14987 Insert a cache barrier before a load or store that might be
14988 speculatively executed and that might have side effects even
14991 @item -mr10k-cache-barrier=store
14992 Insert a cache barrier before a store that might be speculatively
14993 executed and that might have side effects even if aborted.
14995 @item -mr10k-cache-barrier=none
14996 Disable the insertion of cache barriers. This is the default setting.
14999 @item -mflush-func=@var{func}
15000 @itemx -mno-flush-func
15001 @opindex mflush-func
15002 Specifies the function to call to flush the I and D caches, or to not
15003 call any such function. If called, the function must take the same
15004 arguments as the common @code{_flush_func()}, that is, the address of the
15005 memory range for which the cache is being flushed, the size of the
15006 memory range, and the number 3 (to flush both caches). The default
15007 depends on the target GCC was configured for, but commonly is either
15008 @samp{_flush_func} or @samp{__cpu_flush}.
15010 @item mbranch-cost=@var{num}
15011 @opindex mbranch-cost
15012 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15013 This cost is only a heuristic and is not guaranteed to produce
15014 consistent results across releases. A zero cost redundantly selects
15015 the default, which is based on the @option{-mtune} setting.
15017 @item -mbranch-likely
15018 @itemx -mno-branch-likely
15019 @opindex mbranch-likely
15020 @opindex mno-branch-likely
15021 Enable or disable use of Branch Likely instructions, regardless of the
15022 default for the selected architecture. By default, Branch Likely
15023 instructions may be generated if they are supported by the selected
15024 architecture. An exception is for the MIPS32 and MIPS64 architectures
15025 and processors which implement those architectures; for those, Branch
15026 Likely instructions will not be generated by default because the MIPS32
15027 and MIPS64 architectures specifically deprecate their use.
15029 @item -mfp-exceptions
15030 @itemx -mno-fp-exceptions
15031 @opindex mfp-exceptions
15032 Specifies whether FP exceptions are enabled. This affects how we schedule
15033 FP instructions for some processors. The default is that FP exceptions are
15036 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15037 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15040 @item -mvr4130-align
15041 @itemx -mno-vr4130-align
15042 @opindex mvr4130-align
15043 The VR4130 pipeline is two-way superscalar, but can only issue two
15044 instructions together if the first one is 8-byte aligned. When this
15045 option is enabled, GCC will align pairs of instructions that it
15046 thinks should execute in parallel.
15048 This option only has an effect when optimizing for the VR4130.
15049 It normally makes code faster, but at the expense of making it bigger.
15050 It is enabled by default at optimization level @option{-O3}.
15055 Enable (disable) generation of @code{synci} instructions on
15056 architectures that support it. The @code{synci} instructions (if
15057 enabled) will be generated when @code{__builtin___clear_cache()} is
15060 This option defaults to @code{-mno-synci}, but the default can be
15061 overridden by configuring with @code{--with-synci}.
15063 When compiling code for single processor systems, it is generally safe
15064 to use @code{synci}. However, on many multi-core (SMP) systems, it
15065 will not invalidate the instruction caches on all cores and may lead
15066 to undefined behavior.
15068 @item -mrelax-pic-calls
15069 @itemx -mno-relax-pic-calls
15070 @opindex mrelax-pic-calls
15071 Try to turn PIC calls that are normally dispatched via register
15072 @code{$25} into direct calls. This is only possible if the linker can
15073 resolve the destination at link-time and if the destination is within
15074 range for a direct call.
15076 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15077 an assembler and a linker that supports the @code{.reloc} assembly
15078 directive and @code{-mexplicit-relocs} is in effect. With
15079 @code{-mno-explicit-relocs}, this optimization can be performed by the
15080 assembler and the linker alone without help from the compiler.
15082 @item -mmcount-ra-address
15083 @itemx -mno-mcount-ra-address
15084 @opindex mmcount-ra-address
15085 @opindex mno-mcount-ra-address
15086 Emit (do not emit) code that allows @code{_mcount} to modify the
15087 calling function's return address. When enabled, this option extends
15088 the usual @code{_mcount} interface with a new @var{ra-address}
15089 parameter, which has type @code{intptr_t *} and is passed in register
15090 @code{$12}. @code{_mcount} can then modify the return address by
15091 doing both of the following:
15094 Returning the new address in register @code{$31}.
15096 Storing the new address in @code{*@var{ra-address}},
15097 if @var{ra-address} is nonnull.
15100 The default is @option{-mno-mcount-ra-address}.
15105 @subsection MMIX Options
15106 @cindex MMIX Options
15108 These options are defined for the MMIX:
15112 @itemx -mno-libfuncs
15114 @opindex mno-libfuncs
15115 Specify that intrinsic library functions are being compiled, passing all
15116 values in registers, no matter the size.
15119 @itemx -mno-epsilon
15121 @opindex mno-epsilon
15122 Generate floating-point comparison instructions that compare with respect
15123 to the @code{rE} epsilon register.
15125 @item -mabi=mmixware
15127 @opindex mabi=mmixware
15129 Generate code that passes function parameters and return values that (in
15130 the called function) are seen as registers @code{$0} and up, as opposed to
15131 the GNU ABI which uses global registers @code{$231} and up.
15133 @item -mzero-extend
15134 @itemx -mno-zero-extend
15135 @opindex mzero-extend
15136 @opindex mno-zero-extend
15137 When reading data from memory in sizes shorter than 64 bits, use (do not
15138 use) zero-extending load instructions by default, rather than
15139 sign-extending ones.
15142 @itemx -mno-knuthdiv
15144 @opindex mno-knuthdiv
15145 Make the result of a division yielding a remainder have the same sign as
15146 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15147 remainder follows the sign of the dividend. Both methods are
15148 arithmetically valid, the latter being almost exclusively used.
15150 @item -mtoplevel-symbols
15151 @itemx -mno-toplevel-symbols
15152 @opindex mtoplevel-symbols
15153 @opindex mno-toplevel-symbols
15154 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15155 code can be used with the @code{PREFIX} assembly directive.
15159 Generate an executable in the ELF format, rather than the default
15160 @samp{mmo} format used by the @command{mmix} simulator.
15162 @item -mbranch-predict
15163 @itemx -mno-branch-predict
15164 @opindex mbranch-predict
15165 @opindex mno-branch-predict
15166 Use (do not use) the probable-branch instructions, when static branch
15167 prediction indicates a probable branch.
15169 @item -mbase-addresses
15170 @itemx -mno-base-addresses
15171 @opindex mbase-addresses
15172 @opindex mno-base-addresses
15173 Generate (do not generate) code that uses @emph{base addresses}. Using a
15174 base address automatically generates a request (handled by the assembler
15175 and the linker) for a constant to be set up in a global register. The
15176 register is used for one or more base address requests within the range 0
15177 to 255 from the value held in the register. The generally leads to short
15178 and fast code, but the number of different data items that can be
15179 addressed is limited. This means that a program that uses lots of static
15180 data may require @option{-mno-base-addresses}.
15182 @item -msingle-exit
15183 @itemx -mno-single-exit
15184 @opindex msingle-exit
15185 @opindex mno-single-exit
15186 Force (do not force) generated code to have a single exit point in each
15190 @node MN10300 Options
15191 @subsection MN10300 Options
15192 @cindex MN10300 options
15194 These @option{-m} options are defined for Matsushita MN10300 architectures:
15199 Generate code to avoid bugs in the multiply instructions for the MN10300
15200 processors. This is the default.
15202 @item -mno-mult-bug
15203 @opindex mno-mult-bug
15204 Do not generate code to avoid bugs in the multiply instructions for the
15205 MN10300 processors.
15209 Generate code which uses features specific to the AM33 processor.
15213 Do not generate code which uses features specific to the AM33 processor. This
15218 Generate code which uses features specific to the AM33/2.0 processor.
15222 Generate code which uses features specific to the AM34 processor.
15224 @item -mtune=@var{cpu-type}
15226 Use the timing characteristics of the indicated CPU type when
15227 scheduling instructions. This does not change the targeted processor
15228 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15229 @samp{am33-2} or @samp{am34}.
15231 @item -mreturn-pointer-on-d0
15232 @opindex mreturn-pointer-on-d0
15233 When generating a function which returns a pointer, return the pointer
15234 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15235 only in a0, and attempts to call such functions without a prototype
15236 would result in errors. Note that this option is on by default; use
15237 @option{-mno-return-pointer-on-d0} to disable it.
15241 Do not link in the C run-time initialization object file.
15245 Indicate to the linker that it should perform a relaxation optimization pass
15246 to shorten branches, calls and absolute memory addresses. This option only
15247 has an effect when used on the command line for the final link step.
15249 This option makes symbolic debugging impossible.
15253 Allow the compiler to generate @emph{Long Instruction Word}
15254 instructions if the target is the @samp{AM33} or later. This is the
15255 default. This option defines the preprocessor macro @samp{__LIW__}.
15259 Do not allow the compiler to generate @emph{Long Instruction Word}
15260 instructions. This option defines the preprocessor macro
15265 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15266 instructions if the target is the @samp{AM33} or later. This is the
15267 default. This option defines the preprocessor macro @samp{__SETLB__}.
15271 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15272 instructions. This option defines the preprocessor macro
15273 @samp{__NO_SETLB__}.
15277 @node PDP-11 Options
15278 @subsection PDP-11 Options
15279 @cindex PDP-11 Options
15281 These options are defined for the PDP-11:
15286 Use hardware FPP floating point. This is the default. (FIS floating
15287 point on the PDP-11/40 is not supported.)
15290 @opindex msoft-float
15291 Do not use hardware floating point.
15295 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15299 Return floating-point results in memory. This is the default.
15303 Generate code for a PDP-11/40.
15307 Generate code for a PDP-11/45. This is the default.
15311 Generate code for a PDP-11/10.
15313 @item -mbcopy-builtin
15314 @opindex mbcopy-builtin
15315 Use inline @code{movmemhi} patterns for copying memory. This is the
15320 Do not use inline @code{movmemhi} patterns for copying memory.
15326 Use 16-bit @code{int}. This is the default.
15332 Use 32-bit @code{int}.
15335 @itemx -mno-float32
15337 @opindex mno-float32
15338 Use 64-bit @code{float}. This is the default.
15341 @itemx -mno-float64
15343 @opindex mno-float64
15344 Use 32-bit @code{float}.
15348 Use @code{abshi2} pattern. This is the default.
15352 Do not use @code{abshi2} pattern.
15354 @item -mbranch-expensive
15355 @opindex mbranch-expensive
15356 Pretend that branches are expensive. This is for experimenting with
15357 code generation only.
15359 @item -mbranch-cheap
15360 @opindex mbranch-cheap
15361 Do not pretend that branches are expensive. This is the default.
15365 Use Unix assembler syntax. This is the default when configured for
15366 @samp{pdp11-*-bsd}.
15370 Use DEC assembler syntax. This is the default when configured for any
15371 PDP-11 target other than @samp{pdp11-*-bsd}.
15374 @node picoChip Options
15375 @subsection picoChip Options
15376 @cindex picoChip options
15378 These @samp{-m} options are defined for picoChip implementations:
15382 @item -mae=@var{ae_type}
15384 Set the instruction set, register set, and instruction scheduling
15385 parameters for array element type @var{ae_type}. Supported values
15386 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15388 @option{-mae=ANY} selects a completely generic AE type. Code
15389 generated with this option will run on any of the other AE types. The
15390 code will not be as efficient as it would be if compiled for a specific
15391 AE type, and some types of operation (e.g., multiplication) will not
15392 work properly on all types of AE.
15394 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15395 for compiled code, and is the default.
15397 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15398 option may suffer from poor performance of byte (char) manipulation,
15399 since the DSP AE does not provide hardware support for byte load/stores.
15401 @item -msymbol-as-address
15402 Enable the compiler to directly use a symbol name as an address in a
15403 load/store instruction, without first loading it into a
15404 register. Typically, the use of this option will generate larger
15405 programs, which run faster than when the option isn't used. However, the
15406 results vary from program to program, so it is left as a user option,
15407 rather than being permanently enabled.
15409 @item -mno-inefficient-warnings
15410 Disables warnings about the generation of inefficient code. These
15411 warnings can be generated, for example, when compiling code which
15412 performs byte-level memory operations on the MAC AE type. The MAC AE has
15413 no hardware support for byte-level memory operations, so all byte
15414 load/stores must be synthesized from word load/store operations. This is
15415 inefficient and a warning will be generated indicating to the programmer
15416 that they should rewrite the code to avoid byte operations, or to target
15417 an AE type which has the necessary hardware support. This option enables
15418 the warning to be turned off.
15422 @node PowerPC Options
15423 @subsection PowerPC Options
15424 @cindex PowerPC options
15426 These are listed under @xref{RS/6000 and PowerPC Options}.
15428 @node RS/6000 and PowerPC Options
15429 @subsection IBM RS/6000 and PowerPC Options
15430 @cindex RS/6000 and PowerPC Options
15431 @cindex IBM RS/6000 and PowerPC Options
15433 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15440 @itemx -mno-powerpc
15441 @itemx -mpowerpc-gpopt
15442 @itemx -mno-powerpc-gpopt
15443 @itemx -mpowerpc-gfxopt
15444 @itemx -mno-powerpc-gfxopt
15447 @itemx -mno-powerpc64
15451 @itemx -mno-popcntb
15453 @itemx -mno-popcntd
15462 @itemx -mno-hard-dfp
15466 @opindex mno-power2
15468 @opindex mno-powerpc
15469 @opindex mpowerpc-gpopt
15470 @opindex mno-powerpc-gpopt
15471 @opindex mpowerpc-gfxopt
15472 @opindex mno-powerpc-gfxopt
15473 @opindex mpowerpc64
15474 @opindex mno-powerpc64
15478 @opindex mno-popcntb
15480 @opindex mno-popcntd
15486 @opindex mno-mfpgpr
15488 @opindex mno-hard-dfp
15489 GCC supports two related instruction set architectures for the
15490 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15491 instructions supported by the @samp{rios} chip set used in the original
15492 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15493 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15494 the IBM 4xx, 6xx, and follow-on microprocessors.
15496 Neither architecture is a subset of the other. However there is a
15497 large common subset of instructions supported by both. An MQ
15498 register is included in processors supporting the POWER architecture.
15500 You use these options to specify which instructions are available on the
15501 processor you are using. The default value of these options is
15502 determined when configuring GCC@. Specifying the
15503 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15504 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15505 rather than the options listed above.
15507 The @option{-mpower} option allows GCC to generate instructions that
15508 are found only in the POWER architecture and to use the MQ register.
15509 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15510 to generate instructions that are present in the POWER2 architecture but
15511 not the original POWER architecture.
15513 The @option{-mpowerpc} option allows GCC to generate instructions that
15514 are found only in the 32-bit subset of the PowerPC architecture.
15515 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15516 GCC to use the optional PowerPC architecture instructions in the
15517 General Purpose group, including floating-point square root. Specifying
15518 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15519 use the optional PowerPC architecture instructions in the Graphics
15520 group, including floating-point select.
15522 The @option{-mmfcrf} option allows GCC to generate the move from
15523 condition register field instruction implemented on the POWER4
15524 processor and other processors that support the PowerPC V2.01
15526 The @option{-mpopcntb} option allows GCC to generate the popcount and
15527 double precision FP reciprocal estimate instruction implemented on the
15528 POWER5 processor and other processors that support the PowerPC V2.02
15530 The @option{-mpopcntd} option allows GCC to generate the popcount
15531 instruction implemented on the POWER7 processor and other processors
15532 that support the PowerPC V2.06 architecture.
15533 The @option{-mfprnd} option allows GCC to generate the FP round to
15534 integer instructions implemented on the POWER5+ processor and other
15535 processors that support the PowerPC V2.03 architecture.
15536 The @option{-mcmpb} option allows GCC to generate the compare bytes
15537 instruction implemented on the POWER6 processor and other processors
15538 that support the PowerPC V2.05 architecture.
15539 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15540 general purpose register instructions implemented on the POWER6X
15541 processor and other processors that support the extended PowerPC V2.05
15543 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15544 point instructions implemented on some POWER processors.
15546 The @option{-mpowerpc64} option allows GCC to generate the additional
15547 64-bit instructions that are found in the full PowerPC64 architecture
15548 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15549 @option{-mno-powerpc64}.
15551 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15552 will use only the instructions in the common subset of both
15553 architectures plus some special AIX common-mode calls, and will not use
15554 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15555 permits GCC to use any instruction from either architecture and to
15556 allow use of the MQ register; specify this for the Motorola MPC601.
15558 @item -mnew-mnemonics
15559 @itemx -mold-mnemonics
15560 @opindex mnew-mnemonics
15561 @opindex mold-mnemonics
15562 Select which mnemonics to use in the generated assembler code. With
15563 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15564 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15565 assembler mnemonics defined for the POWER architecture. Instructions
15566 defined in only one architecture have only one mnemonic; GCC uses that
15567 mnemonic irrespective of which of these options is specified.
15569 GCC defaults to the mnemonics appropriate for the architecture in
15570 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15571 value of these option. Unless you are building a cross-compiler, you
15572 should normally not specify either @option{-mnew-mnemonics} or
15573 @option{-mold-mnemonics}, but should instead accept the default.
15575 @item -mcpu=@var{cpu_type}
15577 Set architecture type, register usage, choice of mnemonics, and
15578 instruction scheduling parameters for machine type @var{cpu_type}.
15579 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15580 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15581 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15582 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15583 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15584 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15585 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15586 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15587 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15588 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15589 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15591 @option{-mcpu=common} selects a completely generic processor. Code
15592 generated under this option will run on any POWER or PowerPC processor.
15593 GCC will use only the instructions in the common subset of both
15594 architectures, and will not use the MQ register. GCC assumes a generic
15595 processor model for scheduling purposes.
15597 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15598 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15599 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15600 types, with an appropriate, generic processor model assumed for
15601 scheduling purposes.
15603 The other options specify a specific processor. Code generated under
15604 those options will run best on that processor, and may not run at all on
15607 The @option{-mcpu} options automatically enable or disable the
15610 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15611 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15612 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15613 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15615 The particular options set for any particular CPU will vary between
15616 compiler versions, depending on what setting seems to produce optimal
15617 code for that CPU; it doesn't necessarily reflect the actual hardware's
15618 capabilities. If you wish to set an individual option to a particular
15619 value, you may specify it after the @option{-mcpu} option, like
15620 @samp{-mcpu=970 -mno-altivec}.
15622 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15623 not enabled or disabled by the @option{-mcpu} option at present because
15624 AIX does not have full support for these options. You may still
15625 enable or disable them individually if you're sure it'll work in your
15628 @item -mtune=@var{cpu_type}
15630 Set the instruction scheduling parameters for machine type
15631 @var{cpu_type}, but do not set the architecture type, register usage, or
15632 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15633 values for @var{cpu_type} are used for @option{-mtune} as for
15634 @option{-mcpu}. If both are specified, the code generated will use the
15635 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15636 scheduling parameters set by @option{-mtune}.
15638 @item -mcmodel=small
15639 @opindex mcmodel=small
15640 Generate PowerPC64 code for the small model: The TOC is limited to
15643 @item -mcmodel=medium
15644 @opindex mcmodel=medium
15645 Generate PowerPC64 code for the medium model: The TOC and other static
15646 data may be up to a total of 4G in size.
15648 @item -mcmodel=large
15649 @opindex mcmodel=large
15650 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15651 in size. Other data and code is only limited by the 64-bit address
15655 @itemx -mno-altivec
15657 @opindex mno-altivec
15658 Generate code that uses (does not use) AltiVec instructions, and also
15659 enable the use of built-in functions that allow more direct access to
15660 the AltiVec instruction set. You may also need to set
15661 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15667 @opindex mno-vrsave
15668 Generate VRSAVE instructions when generating AltiVec code.
15670 @item -mgen-cell-microcode
15671 @opindex mgen-cell-microcode
15672 Generate Cell microcode instructions
15674 @item -mwarn-cell-microcode
15675 @opindex mwarn-cell-microcode
15676 Warning when a Cell microcode instruction is going to emitted. An example
15677 of a Cell microcode instruction is a variable shift.
15680 @opindex msecure-plt
15681 Generate code that allows ld and ld.so to build executables and shared
15682 libraries with non-exec .plt and .got sections. This is a PowerPC
15683 32-bit SYSV ABI option.
15687 Generate code that uses a BSS .plt section that ld.so fills in, and
15688 requires .plt and .got sections that are both writable and executable.
15689 This is a PowerPC 32-bit SYSV ABI option.
15695 This switch enables or disables the generation of ISEL instructions.
15697 @item -misel=@var{yes/no}
15698 This switch has been deprecated. Use @option{-misel} and
15699 @option{-mno-isel} instead.
15705 This switch enables or disables the generation of SPE simd
15711 @opindex mno-paired
15712 This switch enables or disables the generation of PAIRED simd
15715 @item -mspe=@var{yes/no}
15716 This option has been deprecated. Use @option{-mspe} and
15717 @option{-mno-spe} instead.
15723 Generate code that uses (does not use) vector/scalar (VSX)
15724 instructions, and also enable the use of built-in functions that allow
15725 more direct access to the VSX instruction set.
15727 @item -mfloat-gprs=@var{yes/single/double/no}
15728 @itemx -mfloat-gprs
15729 @opindex mfloat-gprs
15730 This switch enables or disables the generation of floating point
15731 operations on the general purpose registers for architectures that
15734 The argument @var{yes} or @var{single} enables the use of
15735 single-precision floating point operations.
15737 The argument @var{double} enables the use of single and
15738 double-precision floating point operations.
15740 The argument @var{no} disables floating point operations on the
15741 general purpose registers.
15743 This option is currently only available on the MPC854x.
15749 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15750 targets (including GNU/Linux). The 32-bit environment sets int, long
15751 and pointer to 32 bits and generates code that runs on any PowerPC
15752 variant. The 64-bit environment sets int to 32 bits and long and
15753 pointer to 64 bits, and generates code for PowerPC64, as for
15754 @option{-mpowerpc64}.
15757 @itemx -mno-fp-in-toc
15758 @itemx -mno-sum-in-toc
15759 @itemx -mminimal-toc
15761 @opindex mno-fp-in-toc
15762 @opindex mno-sum-in-toc
15763 @opindex mminimal-toc
15764 Modify generation of the TOC (Table Of Contents), which is created for
15765 every executable file. The @option{-mfull-toc} option is selected by
15766 default. In that case, GCC will allocate at least one TOC entry for
15767 each unique non-automatic variable reference in your program. GCC
15768 will also place floating-point constants in the TOC@. However, only
15769 16,384 entries are available in the TOC@.
15771 If you receive a linker error message that saying you have overflowed
15772 the available TOC space, you can reduce the amount of TOC space used
15773 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15774 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15775 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15776 generate code to calculate the sum of an address and a constant at
15777 run-time instead of putting that sum into the TOC@. You may specify one
15778 or both of these options. Each causes GCC to produce very slightly
15779 slower and larger code at the expense of conserving TOC space.
15781 If you still run out of space in the TOC even when you specify both of
15782 these options, specify @option{-mminimal-toc} instead. This option causes
15783 GCC to make only one TOC entry for every file. When you specify this
15784 option, GCC will produce code that is slower and larger but which
15785 uses extremely little TOC space. You may wish to use this option
15786 only on files that contain less frequently executed code.
15792 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15793 @code{long} type, and the infrastructure needed to support them.
15794 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15795 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15796 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15799 @itemx -mno-xl-compat
15800 @opindex mxl-compat
15801 @opindex mno-xl-compat
15802 Produce code that conforms more closely to IBM XL compiler semantics
15803 when using AIX-compatible ABI@. Pass floating-point arguments to
15804 prototyped functions beyond the register save area (RSA) on the stack
15805 in addition to argument FPRs. Do not assume that most significant
15806 double in 128-bit long double value is properly rounded when comparing
15807 values and converting to double. Use XL symbol names for long double
15810 The AIX calling convention was extended but not initially documented to
15811 handle an obscure K&R C case of calling a function that takes the
15812 address of its arguments with fewer arguments than declared. IBM XL
15813 compilers access floating point arguments which do not fit in the
15814 RSA from the stack when a subroutine is compiled without
15815 optimization. Because always storing floating-point arguments on the
15816 stack is inefficient and rarely needed, this option is not enabled by
15817 default and only is necessary when calling subroutines compiled by IBM
15818 XL compilers without optimization.
15822 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15823 application written to use message passing with special startup code to
15824 enable the application to run. The system must have PE installed in the
15825 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15826 must be overridden with the @option{-specs=} option to specify the
15827 appropriate directory location. The Parallel Environment does not
15828 support threads, so the @option{-mpe} option and the @option{-pthread}
15829 option are incompatible.
15831 @item -malign-natural
15832 @itemx -malign-power
15833 @opindex malign-natural
15834 @opindex malign-power
15835 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15836 @option{-malign-natural} overrides the ABI-defined alignment of larger
15837 types, such as floating-point doubles, on their natural size-based boundary.
15838 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15839 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15841 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15845 @itemx -mhard-float
15846 @opindex msoft-float
15847 @opindex mhard-float
15848 Generate code that does not use (uses) the floating-point register set.
15849 Software floating point emulation is provided if you use the
15850 @option{-msoft-float} option, and pass the option to GCC when linking.
15852 @item -msingle-float
15853 @itemx -mdouble-float
15854 @opindex msingle-float
15855 @opindex mdouble-float
15856 Generate code for single or double-precision floating point operations.
15857 @option{-mdouble-float} implies @option{-msingle-float}.
15860 @opindex msimple-fpu
15861 Do not generate sqrt and div instructions for hardware floating point unit.
15865 Specify type of floating point unit. Valid values are @var{sp_lite}
15866 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15867 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15868 and @var{dp_full} (equivalent to -mdouble-float).
15871 @opindex mxilinx-fpu
15872 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15875 @itemx -mno-multiple
15877 @opindex mno-multiple
15878 Generate code that uses (does not use) the load multiple word
15879 instructions and the store multiple word instructions. These
15880 instructions are generated by default on POWER systems, and not
15881 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15882 endian PowerPC systems, since those instructions do not work when the
15883 processor is in little endian mode. The exceptions are PPC740 and
15884 PPC750 which permit the instructions usage in little endian mode.
15889 @opindex mno-string
15890 Generate code that uses (does not use) the load string instructions
15891 and the store string word instructions to save multiple registers and
15892 do small block moves. These instructions are generated by default on
15893 POWER systems, and not generated on PowerPC systems. Do not use
15894 @option{-mstring} on little endian PowerPC systems, since those
15895 instructions do not work when the processor is in little endian mode.
15896 The exceptions are PPC740 and PPC750 which permit the instructions
15897 usage in little endian mode.
15902 @opindex mno-update
15903 Generate code that uses (does not use) the load or store instructions
15904 that update the base register to the address of the calculated memory
15905 location. These instructions are generated by default. If you use
15906 @option{-mno-update}, there is a small window between the time that the
15907 stack pointer is updated and the address of the previous frame is
15908 stored, which means code that walks the stack frame across interrupts or
15909 signals may get corrupted data.
15911 @item -mavoid-indexed-addresses
15912 @itemx -mno-avoid-indexed-addresses
15913 @opindex mavoid-indexed-addresses
15914 @opindex mno-avoid-indexed-addresses
15915 Generate code that tries to avoid (not avoid) the use of indexed load
15916 or store instructions. These instructions can incur a performance
15917 penalty on Power6 processors in certain situations, such as when
15918 stepping through large arrays that cross a 16M boundary. This option
15919 is enabled by default when targetting Power6 and disabled otherwise.
15922 @itemx -mno-fused-madd
15923 @opindex mfused-madd
15924 @opindex mno-fused-madd
15925 Generate code that uses (does not use) the floating point multiply and
15926 accumulate instructions. These instructions are generated by default
15927 if hardware floating point is used. The machine dependent
15928 @option{-mfused-madd} option is now mapped to the machine independent
15929 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15930 mapped to @option{-ffp-contract=off}.
15936 Generate code that uses (does not use) the half-word multiply and
15937 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15938 These instructions are generated by default when targetting those
15945 Generate code that uses (does not use) the string-search @samp{dlmzb}
15946 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15947 generated by default when targetting those processors.
15949 @item -mno-bit-align
15951 @opindex mno-bit-align
15952 @opindex mbit-align
15953 On System V.4 and embedded PowerPC systems do not (do) force structures
15954 and unions that contain bit-fields to be aligned to the base type of the
15957 For example, by default a structure containing nothing but 8
15958 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15959 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15960 the structure would be aligned to a 1 byte boundary and be one byte in
15963 @item -mno-strict-align
15964 @itemx -mstrict-align
15965 @opindex mno-strict-align
15966 @opindex mstrict-align
15967 On System V.4 and embedded PowerPC systems do not (do) assume that
15968 unaligned memory references will be handled by the system.
15970 @item -mrelocatable
15971 @itemx -mno-relocatable
15972 @opindex mrelocatable
15973 @opindex mno-relocatable
15974 Generate code that allows (does not allow) a static executable to be
15975 relocated to a different address at runtime. A simple embedded
15976 PowerPC system loader should relocate the entire contents of
15977 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15978 a table of 32-bit addresses generated by this option. For this to
15979 work, all objects linked together must be compiled with
15980 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15981 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15983 @item -mrelocatable-lib
15984 @itemx -mno-relocatable-lib
15985 @opindex mrelocatable-lib
15986 @opindex mno-relocatable-lib
15987 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15988 @code{.fixup} section to allow static executables to be relocated at
15989 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15990 alignment of @option{-mrelocatable}. Objects compiled with
15991 @option{-mrelocatable-lib} may be linked with objects compiled with
15992 any combination of the @option{-mrelocatable} options.
15998 On System V.4 and embedded PowerPC systems do not (do) assume that
15999 register 2 contains a pointer to a global area pointing to the addresses
16000 used in the program.
16003 @itemx -mlittle-endian
16005 @opindex mlittle-endian
16006 On System V.4 and embedded PowerPC systems compile code for the
16007 processor in little endian mode. The @option{-mlittle-endian} option is
16008 the same as @option{-mlittle}.
16011 @itemx -mbig-endian
16013 @opindex mbig-endian
16014 On System V.4 and embedded PowerPC systems compile code for the
16015 processor in big endian mode. The @option{-mbig-endian} option is
16016 the same as @option{-mbig}.
16018 @item -mdynamic-no-pic
16019 @opindex mdynamic-no-pic
16020 On Darwin and Mac OS X systems, compile code so that it is not
16021 relocatable, but that its external references are relocatable. The
16022 resulting code is suitable for applications, but not shared
16025 @item -msingle-pic-base
16026 @opindex msingle-pic-base
16027 Treat the register used for PIC addressing as read-only, rather than
16028 loading it in the prologue for each function. The run-time system is
16029 responsible for initializing this register with an appropriate value
16030 before execution begins.
16032 @item -mprioritize-restricted-insns=@var{priority}
16033 @opindex mprioritize-restricted-insns
16034 This option controls the priority that is assigned to
16035 dispatch-slot restricted instructions during the second scheduling
16036 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16037 @var{no/highest/second-highest} priority to dispatch slot restricted
16040 @item -msched-costly-dep=@var{dependence_type}
16041 @opindex msched-costly-dep
16042 This option controls which dependences are considered costly
16043 by the target during instruction scheduling. The argument
16044 @var{dependence_type} takes one of the following values:
16045 @var{no}: no dependence is costly,
16046 @var{all}: all dependences are costly,
16047 @var{true_store_to_load}: a true dependence from store to load is costly,
16048 @var{store_to_load}: any dependence from store to load is costly,
16049 @var{number}: any dependence which latency >= @var{number} is costly.
16051 @item -minsert-sched-nops=@var{scheme}
16052 @opindex minsert-sched-nops
16053 This option controls which nop insertion scheme will be used during
16054 the second scheduling pass. The argument @var{scheme} takes one of the
16056 @var{no}: Don't insert nops.
16057 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16058 according to the scheduler's grouping.
16059 @var{regroup_exact}: Insert nops to force costly dependent insns into
16060 separate groups. Insert exactly as many nops as needed to force an insn
16061 to a new group, according to the estimated processor grouping.
16062 @var{number}: Insert nops to force costly dependent insns into
16063 separate groups. Insert @var{number} nops to force an insn to a new group.
16066 @opindex mcall-sysv
16067 On System V.4 and embedded PowerPC systems compile code using calling
16068 conventions that adheres to the March 1995 draft of the System V
16069 Application Binary Interface, PowerPC processor supplement. This is the
16070 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16072 @item -mcall-sysv-eabi
16074 @opindex mcall-sysv-eabi
16075 @opindex mcall-eabi
16076 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16078 @item -mcall-sysv-noeabi
16079 @opindex mcall-sysv-noeabi
16080 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16082 @item -mcall-aixdesc
16084 On System V.4 and embedded PowerPC systems compile code for the AIX
16088 @opindex mcall-linux
16089 On System V.4 and embedded PowerPC systems compile code for the
16090 Linux-based GNU system.
16092 @item -mcall-freebsd
16093 @opindex mcall-freebsd
16094 On System V.4 and embedded PowerPC systems compile code for the
16095 FreeBSD operating system.
16097 @item -mcall-netbsd
16098 @opindex mcall-netbsd
16099 On System V.4 and embedded PowerPC systems compile code for the
16100 NetBSD operating system.
16102 @item -mcall-openbsd
16103 @opindex mcall-netbsd
16104 On System V.4 and embedded PowerPC systems compile code for the
16105 OpenBSD operating system.
16107 @item -maix-struct-return
16108 @opindex maix-struct-return
16109 Return all structures in memory (as specified by the AIX ABI)@.
16111 @item -msvr4-struct-return
16112 @opindex msvr4-struct-return
16113 Return structures smaller than 8 bytes in registers (as specified by the
16116 @item -mabi=@var{abi-type}
16118 Extend the current ABI with a particular extension, or remove such extension.
16119 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16120 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16124 Extend the current ABI with SPE ABI extensions. This does not change
16125 the default ABI, instead it adds the SPE ABI extensions to the current
16129 @opindex mabi=no-spe
16130 Disable Booke SPE ABI extensions for the current ABI@.
16132 @item -mabi=ibmlongdouble
16133 @opindex mabi=ibmlongdouble
16134 Change the current ABI to use IBM extended precision long double.
16135 This is a PowerPC 32-bit SYSV ABI option.
16137 @item -mabi=ieeelongdouble
16138 @opindex mabi=ieeelongdouble
16139 Change the current ABI to use IEEE extended precision long double.
16140 This is a PowerPC 32-bit Linux ABI option.
16143 @itemx -mno-prototype
16144 @opindex mprototype
16145 @opindex mno-prototype
16146 On System V.4 and embedded PowerPC systems assume that all calls to
16147 variable argument functions are properly prototyped. Otherwise, the
16148 compiler must insert an instruction before every non prototyped call to
16149 set or clear bit 6 of the condition code register (@var{CR}) to
16150 indicate whether floating point values were passed in the floating point
16151 registers in case the function takes a variable arguments. With
16152 @option{-mprototype}, only calls to prototyped variable argument functions
16153 will set or clear the bit.
16157 On embedded PowerPC systems, assume that the startup module is called
16158 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16159 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16164 On embedded PowerPC systems, assume that the startup module is called
16165 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16170 On embedded PowerPC systems, assume that the startup module is called
16171 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16174 @item -myellowknife
16175 @opindex myellowknife
16176 On embedded PowerPC systems, assume that the startup module is called
16177 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16182 On System V.4 and embedded PowerPC systems, specify that you are
16183 compiling for a VxWorks system.
16187 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16188 header to indicate that @samp{eabi} extended relocations are used.
16194 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16195 Embedded Applications Binary Interface (eabi) which is a set of
16196 modifications to the System V.4 specifications. Selecting @option{-meabi}
16197 means that the stack is aligned to an 8 byte boundary, a function
16198 @code{__eabi} is called to from @code{main} to set up the eabi
16199 environment, and the @option{-msdata} option can use both @code{r2} and
16200 @code{r13} to point to two separate small data areas. Selecting
16201 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16202 do not call an initialization function from @code{main}, and the
16203 @option{-msdata} option will only use @code{r13} to point to a single
16204 small data area. The @option{-meabi} option is on by default if you
16205 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16208 @opindex msdata=eabi
16209 On System V.4 and embedded PowerPC systems, put small initialized
16210 @code{const} global and static data in the @samp{.sdata2} section, which
16211 is pointed to by register @code{r2}. Put small initialized
16212 non-@code{const} global and static data in the @samp{.sdata} section,
16213 which is pointed to by register @code{r13}. Put small uninitialized
16214 global and static data in the @samp{.sbss} section, which is adjacent to
16215 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16216 incompatible with the @option{-mrelocatable} option. The
16217 @option{-msdata=eabi} option also sets the @option{-memb} option.
16220 @opindex msdata=sysv
16221 On System V.4 and embedded PowerPC systems, put small global and static
16222 data in the @samp{.sdata} section, which is pointed to by register
16223 @code{r13}. Put small uninitialized global and static data in the
16224 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16225 The @option{-msdata=sysv} option is incompatible with the
16226 @option{-mrelocatable} option.
16228 @item -msdata=default
16230 @opindex msdata=default
16232 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16233 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16234 same as @option{-msdata=sysv}.
16237 @opindex msdata=data
16238 On System V.4 and embedded PowerPC systems, put small global
16239 data in the @samp{.sdata} section. Put small uninitialized global
16240 data in the @samp{.sbss} section. Do not use register @code{r13}
16241 to address small data however. This is the default behavior unless
16242 other @option{-msdata} options are used.
16246 @opindex msdata=none
16248 On embedded PowerPC systems, put all initialized global and static data
16249 in the @samp{.data} section, and all uninitialized data in the
16250 @samp{.bss} section.
16252 @item -mblock-move-inline-limit=@var{num}
16253 @opindex mblock-move-inline-limit
16254 Inline all block moves (such as calls to @code{memcpy} or structure
16255 copies) less than or equal to @var{num} bytes. The minimum value for
16256 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16257 targets. The default value is target-specific.
16261 @cindex smaller data references (PowerPC)
16262 @cindex .sdata/.sdata2 references (PowerPC)
16263 On embedded PowerPC systems, put global and static items less than or
16264 equal to @var{num} bytes into the small data or bss sections instead of
16265 the normal data or bss section. By default, @var{num} is 8. The
16266 @option{-G @var{num}} switch is also passed to the linker.
16267 All modules should be compiled with the same @option{-G @var{num}} value.
16270 @itemx -mno-regnames
16272 @opindex mno-regnames
16273 On System V.4 and embedded PowerPC systems do (do not) emit register
16274 names in the assembly language output using symbolic forms.
16277 @itemx -mno-longcall
16279 @opindex mno-longcall
16280 By default assume that all calls are far away so that a longer more
16281 expensive calling sequence is required. This is required for calls
16282 further than 32 megabytes (33,554,432 bytes) from the current location.
16283 A short call will be generated if the compiler knows
16284 the call cannot be that far away. This setting can be overridden by
16285 the @code{shortcall} function attribute, or by @code{#pragma
16288 Some linkers are capable of detecting out-of-range calls and generating
16289 glue code on the fly. On these systems, long calls are unnecessary and
16290 generate slower code. As of this writing, the AIX linker can do this,
16291 as can the GNU linker for PowerPC/64. It is planned to add this feature
16292 to the GNU linker for 32-bit PowerPC systems as well.
16294 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16295 callee, L42'', plus a ``branch island'' (glue code). The two target
16296 addresses represent the callee and the ``branch island''. The
16297 Darwin/PPC linker will prefer the first address and generate a ``bl
16298 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16299 otherwise, the linker will generate ``bl L42'' to call the ``branch
16300 island''. The ``branch island'' is appended to the body of the
16301 calling function; it computes the full 32-bit address of the callee
16304 On Mach-O (Darwin) systems, this option directs the compiler emit to
16305 the glue for every direct call, and the Darwin linker decides whether
16306 to use or discard it.
16308 In the future, we may cause GCC to ignore all longcall specifications
16309 when the linker is known to generate glue.
16311 @item -mtls-markers
16312 @itemx -mno-tls-markers
16313 @opindex mtls-markers
16314 @opindex mno-tls-markers
16315 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16316 specifying the function argument. The relocation allows ld to
16317 reliably associate function call with argument setup instructions for
16318 TLS optimization, which in turn allows gcc to better schedule the
16323 Adds support for multithreading with the @dfn{pthreads} library.
16324 This option sets flags for both the preprocessor and linker.
16329 This option will enable GCC to use the reciprocal estimate and
16330 reciprocal square root estimate instructions with additional
16331 Newton-Raphson steps to increase precision instead of doing a divide or
16332 square root and divide for floating point arguments. You should use
16333 the @option{-ffast-math} option when using @option{-mrecip} (or at
16334 least @option{-funsafe-math-optimizations},
16335 @option{-finite-math-only}, @option{-freciprocal-math} and
16336 @option{-fno-trapping-math}). Note that while the throughput of the
16337 sequence is generally higher than the throughput of the non-reciprocal
16338 instruction, the precision of the sequence can be decreased by up to 2
16339 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16342 @item -mrecip=@var{opt}
16343 @opindex mrecip=opt
16344 This option allows to control which reciprocal estimate instructions
16345 may be used. @var{opt} is a comma separated list of options, that may
16346 be preceded by a @code{!} to invert the option:
16347 @code{all}: enable all estimate instructions,
16348 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16349 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16350 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16351 @code{divf}: enable the single precision reciprocal approximation instructions;
16352 @code{divd}: enable the double precision reciprocal approximation instructions;
16353 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16354 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16355 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16357 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16358 all of the reciprocal estimate instructions, except for the
16359 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16360 which handle the double precision reciprocal square root calculations.
16362 @item -mrecip-precision
16363 @itemx -mno-recip-precision
16364 @opindex mrecip-precision
16365 Assume (do not assume) that the reciprocal estimate instructions
16366 provide higher precision estimates than is mandated by the powerpc
16367 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16368 automatically selects @option{-mrecip-precision}. The double
16369 precision square root estimate instructions are not generated by
16370 default on low precision machines, since they do not provide an
16371 estimate that converges after three steps.
16373 @item -mveclibabi=@var{type}
16374 @opindex mveclibabi
16375 Specifies the ABI type to use for vectorizing intrinsics using an
16376 external library. The only type supported at present is @code{mass},
16377 which specifies to use IBM's Mathematical Acceleration Subsystem
16378 (MASS) libraries for vectorizing intrinsics using external libraries.
16379 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16380 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16381 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16382 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16383 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16384 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16385 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16386 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16387 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16388 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16389 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16390 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16391 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16392 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16393 for power7. Both @option{-ftree-vectorize} and
16394 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16395 libraries will have to be specified at link time.
16400 Generate (do not generate) the @code{friz} instruction when the
16401 @option{-funsafe-math-optimizations} option is used to optimize
16402 rounding a floating point value to 64-bit integer and back to floating
16403 point. The @code{friz} instruction does not return the same value if
16404 the floating point number is too large to fit in an integer.
16406 @item -mpointers-to-nested-functions
16407 @itemx -mno-pointers-to-nested-functions
16408 @opindex mpointers-to-nested-functions
16409 Generate (do not generate) code to load up the static chain register
16410 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16411 systems where a function pointer points to a 3 word descriptor giving
16412 the function address, TOC value to be loaded in register @var{r2}, and
16413 static chain value to be loaded in register @var{r11}. The
16414 @option{-mpointers-to-nested-functions} is on by default. You will
16415 not be able to call through pointers to nested functions or pointers
16416 to functions compiled in other languages that use the static chain if
16417 you use the @option{-mno-pointers-to-nested-functions}.
16421 @subsection RX Options
16424 These command line options are defined for RX targets:
16427 @item -m64bit-doubles
16428 @itemx -m32bit-doubles
16429 @opindex m64bit-doubles
16430 @opindex m32bit-doubles
16431 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16432 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16433 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16434 works on 32-bit values, which is why the default is
16435 @option{-m32bit-doubles}.
16441 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16442 floating point hardware. The default is enabled for the @var{RX600}
16443 series and disabled for the @var{RX200} series.
16445 Floating point instructions will only be generated for 32-bit floating
16446 point values however, so if the @option{-m64bit-doubles} option is in
16447 use then the FPU hardware will not be used for doubles.
16449 @emph{Note} If the @option{-fpu} option is enabled then
16450 @option{-funsafe-math-optimizations} is also enabled automatically.
16451 This is because the RX FPU instructions are themselves unsafe.
16453 @item -mcpu=@var{name}
16455 Selects the type of RX CPU to be targeted. Currently three types are
16456 supported, the generic @var{RX600} and @var{RX200} series hardware and
16457 the specific @var{RX610} CPU. The default is @var{RX600}.
16459 The only difference between @var{RX600} and @var{RX610} is that the
16460 @var{RX610} does not support the @code{MVTIPL} instruction.
16462 The @var{RX200} series does not have a hardware floating point unit
16463 and so @option{-nofpu} is enabled by default when this type is
16466 @item -mbig-endian-data
16467 @itemx -mlittle-endian-data
16468 @opindex mbig-endian-data
16469 @opindex mlittle-endian-data
16470 Store data (but not code) in the big-endian format. The default is
16471 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16474 @item -msmall-data-limit=@var{N}
16475 @opindex msmall-data-limit
16476 Specifies the maximum size in bytes of global and static variables
16477 which can be placed into the small data area. Using the small data
16478 area can lead to smaller and faster code, but the size of area is
16479 limited and it is up to the programmer to ensure that the area does
16480 not overflow. Also when the small data area is used one of the RX's
16481 registers (@code{r13}) is reserved for use pointing to this area, so
16482 it is no longer available for use by the compiler. This could result
16483 in slower and/or larger code if variables which once could have been
16484 held in @code{r13} are now pushed onto the stack.
16486 Note, common variables (variables which have not been initialised) and
16487 constants are not placed into the small data area as they are assigned
16488 to other sections in the output executable.
16490 The default value is zero, which disables this feature. Note, this
16491 feature is not enabled by default with higher optimization levels
16492 (@option{-O2} etc) because of the potentially detrimental effects of
16493 reserving register @code{r13}. It is up to the programmer to
16494 experiment and discover whether this feature is of benefit to their
16501 Use the simulator runtime. The default is to use the libgloss board
16504 @item -mas100-syntax
16505 @itemx -mno-as100-syntax
16506 @opindex mas100-syntax
16507 @opindex mno-as100-syntax
16508 When generating assembler output use a syntax that is compatible with
16509 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16510 assembler but it has some restrictions so generating it is not the
16513 @item -mmax-constant-size=@var{N}
16514 @opindex mmax-constant-size
16515 Specifies the maximum size, in bytes, of a constant that can be used as
16516 an operand in a RX instruction. Although the RX instruction set does
16517 allow constants of up to 4 bytes in length to be used in instructions,
16518 a longer value equates to a longer instruction. Thus in some
16519 circumstances it can be beneficial to restrict the size of constants
16520 that are used in instructions. Constants that are too big are instead
16521 placed into a constant pool and referenced via register indirection.
16523 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16524 or 4 means that constants of any size are allowed.
16528 Enable linker relaxation. Linker relaxation is a process whereby the
16529 linker will attempt to reduce the size of a program by finding shorter
16530 versions of various instructions. Disabled by default.
16532 @item -mint-register=@var{N}
16533 @opindex mint-register
16534 Specify the number of registers to reserve for fast interrupt handler
16535 functions. The value @var{N} can be between 0 and 4. A value of 1
16536 means that register @code{r13} will be reserved for the exclusive use
16537 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16538 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16539 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16540 A value of 0, the default, does not reserve any registers.
16542 @item -msave-acc-in-interrupts
16543 @opindex msave-acc-in-interrupts
16544 Specifies that interrupt handler functions should preserve the
16545 accumulator register. This is only necessary if normal code might use
16546 the accumulator register, for example because it performs 64-bit
16547 multiplications. The default is to ignore the accumulator as this
16548 makes the interrupt handlers faster.
16552 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16553 has special significance to the RX port when used with the
16554 @code{interrupt} function attribute. This attribute indicates a
16555 function intended to process fast interrupts. GCC will will ensure
16556 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16557 and/or @code{r13} and only provided that the normal use of the
16558 corresponding registers have been restricted via the
16559 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16562 @node S/390 and zSeries Options
16563 @subsection S/390 and zSeries Options
16564 @cindex S/390 and zSeries Options
16566 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16570 @itemx -msoft-float
16571 @opindex mhard-float
16572 @opindex msoft-float
16573 Use (do not use) the hardware floating-point instructions and registers
16574 for floating-point operations. When @option{-msoft-float} is specified,
16575 functions in @file{libgcc.a} will be used to perform floating-point
16576 operations. When @option{-mhard-float} is specified, the compiler
16577 generates IEEE floating-point instructions. This is the default.
16580 @itemx -mno-hard-dfp
16582 @opindex mno-hard-dfp
16583 Use (do not use) the hardware decimal-floating-point instructions for
16584 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16585 specified, functions in @file{libgcc.a} will be used to perform
16586 decimal-floating-point operations. When @option{-mhard-dfp} is
16587 specified, the compiler generates decimal-floating-point hardware
16588 instructions. This is the default for @option{-march=z9-ec} or higher.
16590 @item -mlong-double-64
16591 @itemx -mlong-double-128
16592 @opindex mlong-double-64
16593 @opindex mlong-double-128
16594 These switches control the size of @code{long double} type. A size
16595 of 64bit makes the @code{long double} type equivalent to the @code{double}
16596 type. This is the default.
16599 @itemx -mno-backchain
16600 @opindex mbackchain
16601 @opindex mno-backchain
16602 Store (do not store) the address of the caller's frame as backchain pointer
16603 into the callee's stack frame.
16604 A backchain may be needed to allow debugging using tools that do not understand
16605 DWARF-2 call frame information.
16606 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16607 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16608 the backchain is placed into the topmost word of the 96/160 byte register
16611 In general, code compiled with @option{-mbackchain} is call-compatible with
16612 code compiled with @option{-mmo-backchain}; however, use of the backchain
16613 for debugging purposes usually requires that the whole binary is built with
16614 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16615 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16616 to build a linux kernel use @option{-msoft-float}.
16618 The default is to not maintain the backchain.
16620 @item -mpacked-stack
16621 @itemx -mno-packed-stack
16622 @opindex mpacked-stack
16623 @opindex mno-packed-stack
16624 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16625 specified, the compiler uses the all fields of the 96/160 byte register save
16626 area only for their default purpose; unused fields still take up stack space.
16627 When @option{-mpacked-stack} is specified, register save slots are densely
16628 packed at the top of the register save area; unused space is reused for other
16629 purposes, allowing for more efficient use of the available stack space.
16630 However, when @option{-mbackchain} is also in effect, the topmost word of
16631 the save area is always used to store the backchain, and the return address
16632 register is always saved two words below the backchain.
16634 As long as the stack frame backchain is not used, code generated with
16635 @option{-mpacked-stack} is call-compatible with code generated with
16636 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16637 S/390 or zSeries generated code that uses the stack frame backchain at run
16638 time, not just for debugging purposes. Such code is not call-compatible
16639 with code compiled with @option{-mpacked-stack}. Also, note that the
16640 combination of @option{-mbackchain},
16641 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16642 to build a linux kernel use @option{-msoft-float}.
16644 The default is to not use the packed stack layout.
16647 @itemx -mno-small-exec
16648 @opindex msmall-exec
16649 @opindex mno-small-exec
16650 Generate (or do not generate) code using the @code{bras} instruction
16651 to do subroutine calls.
16652 This only works reliably if the total executable size does not
16653 exceed 64k. The default is to use the @code{basr} instruction instead,
16654 which does not have this limitation.
16660 When @option{-m31} is specified, generate code compliant to the
16661 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16662 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16663 particular to generate 64-bit instructions. For the @samp{s390}
16664 targets, the default is @option{-m31}, while the @samp{s390x}
16665 targets default to @option{-m64}.
16671 When @option{-mzarch} is specified, generate code using the
16672 instructions available on z/Architecture.
16673 When @option{-mesa} is specified, generate code using the
16674 instructions available on ESA/390. Note that @option{-mesa} is
16675 not possible with @option{-m64}.
16676 When generating code compliant to the GNU/Linux for S/390 ABI,
16677 the default is @option{-mesa}. When generating code compliant
16678 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16684 Generate (or do not generate) code using the @code{mvcle} instruction
16685 to perform block moves. When @option{-mno-mvcle} is specified,
16686 use a @code{mvc} loop instead. This is the default unless optimizing for
16693 Print (or do not print) additional debug information when compiling.
16694 The default is to not print debug information.
16696 @item -march=@var{cpu-type}
16698 Generate code that will run on @var{cpu-type}, which is the name of a system
16699 representing a certain processor type. Possible values for
16700 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16701 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16702 When generating code using the instructions available on z/Architecture,
16703 the default is @option{-march=z900}. Otherwise, the default is
16704 @option{-march=g5}.
16706 @item -mtune=@var{cpu-type}
16708 Tune to @var{cpu-type} everything applicable about the generated code,
16709 except for the ABI and the set of available instructions.
16710 The list of @var{cpu-type} values is the same as for @option{-march}.
16711 The default is the value used for @option{-march}.
16714 @itemx -mno-tpf-trace
16715 @opindex mtpf-trace
16716 @opindex mno-tpf-trace
16717 Generate code that adds (does not add) in TPF OS specific branches to trace
16718 routines in the operating system. This option is off by default, even
16719 when compiling for the TPF OS@.
16722 @itemx -mno-fused-madd
16723 @opindex mfused-madd
16724 @opindex mno-fused-madd
16725 Generate code that uses (does not use) the floating point multiply and
16726 accumulate instructions. These instructions are generated by default if
16727 hardware floating point is used.
16729 @item -mwarn-framesize=@var{framesize}
16730 @opindex mwarn-framesize
16731 Emit a warning if the current function exceeds the given frame size. Because
16732 this is a compile time check it doesn't need to be a real problem when the program
16733 runs. It is intended to identify functions which most probably cause
16734 a stack overflow. It is useful to be used in an environment with limited stack
16735 size e.g.@: the linux kernel.
16737 @item -mwarn-dynamicstack
16738 @opindex mwarn-dynamicstack
16739 Emit a warning if the function calls alloca or uses dynamically
16740 sized arrays. This is generally a bad idea with a limited stack size.
16742 @item -mstack-guard=@var{stack-guard}
16743 @itemx -mstack-size=@var{stack-size}
16744 @opindex mstack-guard
16745 @opindex mstack-size
16746 If these options are provided the s390 back end emits additional instructions in
16747 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16748 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16749 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16750 the frame size of the compiled function is chosen.
16751 These options are intended to be used to help debugging stack overflow problems.
16752 The additionally emitted code causes only little overhead and hence can also be
16753 used in production like systems without greater performance degradation. The given
16754 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16755 @var{stack-guard} without exceeding 64k.
16756 In order to be efficient the extra code makes the assumption that the stack starts
16757 at an address aligned to the value given by @var{stack-size}.
16758 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16761 @node Score Options
16762 @subsection Score Options
16763 @cindex Score Options
16765 These options are defined for Score implementations:
16770 Compile code for big endian mode. This is the default.
16774 Compile code for little endian mode.
16778 Disable generate bcnz instruction.
16782 Enable generate unaligned load and store instruction.
16786 Enable the use of multiply-accumulate instructions. Disabled by default.
16790 Specify the SCORE5 as the target architecture.
16794 Specify the SCORE5U of the target architecture.
16798 Specify the SCORE7 as the target architecture. This is the default.
16802 Specify the SCORE7D as the target architecture.
16806 @subsection SH Options
16808 These @samp{-m} options are defined for the SH implementations:
16813 Generate code for the SH1.
16817 Generate code for the SH2.
16820 Generate code for the SH2e.
16824 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16825 that the floating-point unit is not used.
16827 @item -m2a-single-only
16828 @opindex m2a-single-only
16829 Generate code for the SH2a-FPU, in such a way that no double-precision
16830 floating point operations are used.
16833 @opindex m2a-single
16834 Generate code for the SH2a-FPU assuming the floating-point unit is in
16835 single-precision mode by default.
16839 Generate code for the SH2a-FPU assuming the floating-point unit is in
16840 double-precision mode by default.
16844 Generate code for the SH3.
16848 Generate code for the SH3e.
16852 Generate code for the SH4 without a floating-point unit.
16854 @item -m4-single-only
16855 @opindex m4-single-only
16856 Generate code for the SH4 with a floating-point unit that only
16857 supports single-precision arithmetic.
16861 Generate code for the SH4 assuming the floating-point unit is in
16862 single-precision mode by default.
16866 Generate code for the SH4.
16870 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16871 floating-point unit is not used.
16873 @item -m4a-single-only
16874 @opindex m4a-single-only
16875 Generate code for the SH4a, in such a way that no double-precision
16876 floating point operations are used.
16879 @opindex m4a-single
16880 Generate code for the SH4a assuming the floating-point unit is in
16881 single-precision mode by default.
16885 Generate code for the SH4a.
16889 Same as @option{-m4a-nofpu}, except that it implicitly passes
16890 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16891 instructions at the moment.
16895 Compile code for the processor in big endian mode.
16899 Compile code for the processor in little endian mode.
16903 Align doubles at 64-bit boundaries. Note that this changes the calling
16904 conventions, and thus some functions from the standard C library will
16905 not work unless you recompile it first with @option{-mdalign}.
16909 Shorten some address references at link time, when possible; uses the
16910 linker option @option{-relax}.
16914 Use 32-bit offsets in @code{switch} tables. The default is to use
16919 Enable the use of bit manipulation instructions on SH2A.
16923 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16924 alignment constraints.
16928 Comply with the calling conventions defined by Renesas.
16932 Comply with the calling conventions defined by Renesas.
16936 Comply with the calling conventions defined for GCC before the Renesas
16937 conventions were available. This option is the default for all
16938 targets of the SH toolchain.
16941 @opindex mnomacsave
16942 Mark the @code{MAC} register as call-clobbered, even if
16943 @option{-mhitachi} is given.
16947 Increase IEEE-compliance of floating-point code.
16948 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16949 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16950 comparisons of NANs / infinities incurs extra overhead in every
16951 floating point comparison, therefore the default is set to
16952 @option{-ffinite-math-only}.
16954 @item -minline-ic_invalidate
16955 @opindex minline-ic_invalidate
16956 Inline code to invalidate instruction cache entries after setting up
16957 nested function trampolines.
16958 This option has no effect if -musermode is in effect and the selected
16959 code generation option (e.g. -m4) does not allow the use of the icbi
16961 If the selected code generation option does not allow the use of the icbi
16962 instruction, and -musermode is not in effect, the inlined code will
16963 manipulate the instruction cache address array directly with an associative
16964 write. This not only requires privileged mode, but it will also
16965 fail if the cache line had been mapped via the TLB and has become unmapped.
16969 Dump instruction size and location in the assembly code.
16972 @opindex mpadstruct
16973 This option is deprecated. It pads structures to multiple of 4 bytes,
16974 which is incompatible with the SH ABI@.
16978 Optimize for space instead of speed. Implied by @option{-Os}.
16981 @opindex mprefergot
16982 When generating position-independent code, emit function calls using
16983 the Global Offset Table instead of the Procedure Linkage Table.
16987 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16988 if the inlined code would not work in user mode.
16989 This is the default when the target is @code{sh-*-linux*}.
16991 @item -multcost=@var{number}
16992 @opindex multcost=@var{number}
16993 Set the cost to assume for a multiply insn.
16995 @item -mdiv=@var{strategy}
16996 @opindex mdiv=@var{strategy}
16997 Set the division strategy to use for SHmedia code. @var{strategy} must be
16998 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16999 inv:call2, inv:fp .
17000 "fp" performs the operation in floating point. This has a very high latency,
17001 but needs only a few instructions, so it might be a good choice if
17002 your code has enough easily exploitable ILP to allow the compiler to
17003 schedule the floating point instructions together with other instructions.
17004 Division by zero causes a floating point exception.
17005 "inv" uses integer operations to calculate the inverse of the divisor,
17006 and then multiplies the dividend with the inverse. This strategy allows
17007 cse and hoisting of the inverse calculation. Division by zero calculates
17008 an unspecified result, but does not trap.
17009 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17010 have been found, or if the entire operation has been hoisted to the same
17011 place, the last stages of the inverse calculation are intertwined with the
17012 final multiply to reduce the overall latency, at the expense of using a few
17013 more instructions, and thus offering fewer scheduling opportunities with
17015 "call" calls a library function that usually implements the inv:minlat
17017 This gives high code density for m5-*media-nofpu compilations.
17018 "call2" uses a different entry point of the same library function, where it
17019 assumes that a pointer to a lookup table has already been set up, which
17020 exposes the pointer load to cse / code hoisting optimizations.
17021 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17022 code generation, but if the code stays unoptimized, revert to the "call",
17023 "call2", or "fp" strategies, respectively. Note that the
17024 potentially-trapping side effect of division by zero is carried by a
17025 separate instruction, so it is possible that all the integer instructions
17026 are hoisted out, but the marker for the side effect stays where it is.
17027 A recombination to fp operations or a call is not possible in that case.
17028 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17029 that the inverse calculation was nor separated from the multiply, they speed
17030 up division where the dividend fits into 20 bits (plus sign where applicable),
17031 by inserting a test to skip a number of operations in this case; this test
17032 slows down the case of larger dividends. inv20u assumes the case of a such
17033 a small dividend to be unlikely, and inv20l assumes it to be likely.
17035 @item -maccumulate-outgoing-args
17036 @opindex maccumulate-outgoing-args
17037 Reserve space once for outgoing arguments in the function prologue rather
17038 than around each call. Generally beneficial for performance and size. Also
17039 needed for unwinding to avoid changing the stack frame around conditional code.
17041 @item -mdivsi3_libfunc=@var{name}
17042 @opindex mdivsi3_libfunc=@var{name}
17043 Set the name of the library function used for 32 bit signed division to
17044 @var{name}. This only affect the name used in the call and inv:call
17045 division strategies, and the compiler will still expect the same
17046 sets of input/output/clobbered registers as if this option was not present.
17048 @item -mfixed-range=@var{register-range}
17049 @opindex mfixed-range
17050 Generate code treating the given register range as fixed registers.
17051 A fixed register is one that the register allocator can not use. This is
17052 useful when compiling kernel code. A register range is specified as
17053 two registers separated by a dash. Multiple register ranges can be
17054 specified separated by a comma.
17056 @item -madjust-unroll
17057 @opindex madjust-unroll
17058 Throttle unrolling to avoid thrashing target registers.
17059 This option only has an effect if the gcc code base supports the
17060 TARGET_ADJUST_UNROLL_MAX target hook.
17062 @item -mindexed-addressing
17063 @opindex mindexed-addressing
17064 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17065 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17066 semantics for the indexed addressing mode. The architecture allows the
17067 implementation of processors with 64 bit MMU, which the OS could use to
17068 get 32 bit addressing, but since no current hardware implementation supports
17069 this or any other way to make the indexed addressing mode safe to use in
17070 the 32 bit ABI, the default is -mno-indexed-addressing.
17072 @item -mgettrcost=@var{number}
17073 @opindex mgettrcost=@var{number}
17074 Set the cost assumed for the gettr instruction to @var{number}.
17075 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17079 Assume pt* instructions won't trap. This will generally generate better
17080 scheduled code, but is unsafe on current hardware. The current architecture
17081 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17082 This has the unintentional effect of making it unsafe to schedule ptabs /
17083 ptrel before a branch, or hoist it out of a loop. For example,
17084 __do_global_ctors, a part of libgcc that runs constructors at program
17085 startup, calls functions in a list which is delimited by @minus{}1. With the
17086 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17087 That means that all the constructors will be run a bit quicker, but when
17088 the loop comes to the end of the list, the program crashes because ptabs
17089 loads @minus{}1 into a target register. Since this option is unsafe for any
17090 hardware implementing the current architecture specification, the default
17091 is -mno-pt-fixed. Unless the user specifies a specific cost with
17092 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17093 this deters register allocation using target registers for storing
17096 @item -minvalid-symbols
17097 @opindex minvalid-symbols
17098 Assume symbols might be invalid. Ordinary function symbols generated by
17099 the compiler will always be valid to load with movi/shori/ptabs or
17100 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17101 to generate symbols that will cause ptabs / ptrel to trap.
17102 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17103 It will then prevent cross-basic-block cse, hoisting and most scheduling
17104 of symbol loads. The default is @option{-mno-invalid-symbols}.
17107 @node Solaris 2 Options
17108 @subsection Solaris 2 Options
17109 @cindex Solaris 2 options
17111 These @samp{-m} options are supported on Solaris 2:
17114 @item -mimpure-text
17115 @opindex mimpure-text
17116 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17117 the compiler to not pass @option{-z text} to the linker when linking a
17118 shared object. Using this option, you can link position-dependent
17119 code into a shared object.
17121 @option{-mimpure-text} suppresses the ``relocations remain against
17122 allocatable but non-writable sections'' linker error message.
17123 However, the necessary relocations will trigger copy-on-write, and the
17124 shared object is not actually shared across processes. Instead of
17125 using @option{-mimpure-text}, you should compile all source code with
17126 @option{-fpic} or @option{-fPIC}.
17130 These switches are supported in addition to the above on Solaris 2:
17135 Add support for multithreading using the POSIX threads library. This
17136 option sets flags for both the preprocessor and linker. This option does
17137 not affect the thread safety of object code produced by the compiler or
17138 that of libraries supplied with it.
17142 This is a synonym for @option{-pthreads}.
17145 @node SPARC Options
17146 @subsection SPARC Options
17147 @cindex SPARC options
17149 These @samp{-m} options are supported on the SPARC:
17152 @item -mno-app-regs
17154 @opindex mno-app-regs
17156 Specify @option{-mapp-regs} to generate output using the global registers
17157 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17160 To be fully SVR4 ABI compliant at the cost of some performance loss,
17161 specify @option{-mno-app-regs}. You should compile libraries and system
17162 software with this option.
17168 With @option{-mflat}, the compiler does not generate save/restore instructions
17169 and uses a ``flat'' or single register window model. This model is compatible
17170 with the regular register window model. The local registers and the input
17171 registers (0--5) are still treated as ``call-saved'' registers and will be
17172 saved on the stack as needed.
17174 With @option{-mno-flat} (the default), the compiler generates save/restore
17175 instructions (except for leaf functions). This is the normal operating mode.
17178 @itemx -mhard-float
17180 @opindex mhard-float
17181 Generate output containing floating point instructions. This is the
17185 @itemx -msoft-float
17187 @opindex msoft-float
17188 Generate output containing library calls for floating point.
17189 @strong{Warning:} the requisite libraries are not available for all SPARC
17190 targets. Normally the facilities of the machine's usual C compiler are
17191 used, but this cannot be done directly in cross-compilation. You must make
17192 your own arrangements to provide suitable library functions for
17193 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17194 @samp{sparclite-*-*} do provide software floating point support.
17196 @option{-msoft-float} changes the calling convention in the output file;
17197 therefore, it is only useful if you compile @emph{all} of a program with
17198 this option. In particular, you need to compile @file{libgcc.a}, the
17199 library that comes with GCC, with @option{-msoft-float} in order for
17202 @item -mhard-quad-float
17203 @opindex mhard-quad-float
17204 Generate output containing quad-word (long double) floating point
17207 @item -msoft-quad-float
17208 @opindex msoft-quad-float
17209 Generate output containing library calls for quad-word (long double)
17210 floating point instructions. The functions called are those specified
17211 in the SPARC ABI@. This is the default.
17213 As of this writing, there are no SPARC implementations that have hardware
17214 support for the quad-word floating point instructions. They all invoke
17215 a trap handler for one of these instructions, and then the trap handler
17216 emulates the effect of the instruction. Because of the trap handler overhead,
17217 this is much slower than calling the ABI library routines. Thus the
17218 @option{-msoft-quad-float} option is the default.
17220 @item -mno-unaligned-doubles
17221 @itemx -munaligned-doubles
17222 @opindex mno-unaligned-doubles
17223 @opindex munaligned-doubles
17224 Assume that doubles have 8 byte alignment. This is the default.
17226 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17227 alignment only if they are contained in another type, or if they have an
17228 absolute address. Otherwise, it assumes they have 4 byte alignment.
17229 Specifying this option avoids some rare compatibility problems with code
17230 generated by other compilers. It is not the default because it results
17231 in a performance loss, especially for floating point code.
17233 @item -mno-faster-structs
17234 @itemx -mfaster-structs
17235 @opindex mno-faster-structs
17236 @opindex mfaster-structs
17237 With @option{-mfaster-structs}, the compiler assumes that structures
17238 should have 8 byte alignment. This enables the use of pairs of
17239 @code{ldd} and @code{std} instructions for copies in structure
17240 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17241 However, the use of this changed alignment directly violates the SPARC
17242 ABI@. Thus, it's intended only for use on targets where the developer
17243 acknowledges that their resulting code will not be directly in line with
17244 the rules of the ABI@.
17246 @item -mcpu=@var{cpu_type}
17248 Set the instruction set, register set, and instruction scheduling parameters
17249 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17250 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17251 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17252 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17253 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
17255 Default instruction scheduling parameters are used for values that select
17256 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17257 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17259 Here is a list of each supported architecture and their supported
17264 v8: supersparc, hypersparc, leon
17265 sparclite: f930, f934, sparclite86x
17267 v9: ultrasparc, ultrasparc3, niagara, niagara2
17270 By default (unless configured otherwise), GCC generates code for the V7
17271 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17272 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17273 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17274 SPARCStation 1, 2, IPX etc.
17276 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17277 architecture. The only difference from V7 code is that the compiler emits
17278 the integer multiply and integer divide instructions which exist in SPARC-V8
17279 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17280 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17283 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17284 the SPARC architecture. This adds the integer multiply, integer divide step
17285 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17286 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17287 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17288 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17289 MB86934 chip, which is the more recent SPARClite with FPU@.
17291 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17292 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17293 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17294 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17295 optimizes it for the TEMIC SPARClet chip.
17297 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17298 architecture. This adds 64-bit integer and floating-point move instructions,
17299 3 additional floating-point condition code registers and conditional move
17300 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17301 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17302 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17303 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17304 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17305 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17306 additionally optimizes it for Sun UltraSPARC T2 chips.
17308 @item -mtune=@var{cpu_type}
17310 Set the instruction scheduling parameters for machine type
17311 @var{cpu_type}, but do not set the instruction set or register set that the
17312 option @option{-mcpu=@var{cpu_type}} would.
17314 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17315 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17316 that select a particular CPU implementation. Those are @samp{cypress},
17317 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17318 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17319 @samp{niagara}, and @samp{niagara2}.
17324 @opindex mno-v8plus
17325 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17326 difference from the V8 ABI is that the global and out registers are
17327 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17328 mode for all SPARC-V9 processors.
17334 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17335 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17338 These @samp{-m} options are supported in addition to the above
17339 on SPARC-V9 processors in 64-bit environments:
17342 @item -mlittle-endian
17343 @opindex mlittle-endian
17344 Generate code for a processor running in little-endian mode. It is only
17345 available for a few configurations and most notably not on Solaris and Linux.
17351 Generate code for a 32-bit or 64-bit environment.
17352 The 32-bit environment sets int, long and pointer to 32 bits.
17353 The 64-bit environment sets int to 32 bits and long and pointer
17356 @item -mcmodel=medlow
17357 @opindex mcmodel=medlow
17358 Generate code for the Medium/Low code model: 64-bit addresses, programs
17359 must be linked in the low 32 bits of memory. Programs can be statically
17360 or dynamically linked.
17362 @item -mcmodel=medmid
17363 @opindex mcmodel=medmid
17364 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17365 must be linked in the low 44 bits of memory, the text and data segments must
17366 be less than 2GB in size and the data segment must be located within 2GB of
17369 @item -mcmodel=medany
17370 @opindex mcmodel=medany
17371 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17372 may be linked anywhere in memory, the text and data segments must be less
17373 than 2GB in size and the data segment must be located within 2GB of the
17376 @item -mcmodel=embmedany
17377 @opindex mcmodel=embmedany
17378 Generate code for the Medium/Anywhere code model for embedded systems:
17379 64-bit addresses, the text and data segments must be less than 2GB in
17380 size, both starting anywhere in memory (determined at link time). The
17381 global register %g4 points to the base of the data segment. Programs
17382 are statically linked and PIC is not supported.
17385 @itemx -mno-stack-bias
17386 @opindex mstack-bias
17387 @opindex mno-stack-bias
17388 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17389 frame pointer if present, are offset by @minus{}2047 which must be added back
17390 when making stack frame references. This is the default in 64-bit mode.
17391 Otherwise, assume no such offset is present.
17395 @subsection SPU Options
17396 @cindex SPU options
17398 These @samp{-m} options are supported on the SPU:
17402 @itemx -merror-reloc
17403 @opindex mwarn-reloc
17404 @opindex merror-reloc
17406 The loader for SPU does not handle dynamic relocations. By default, GCC
17407 will give an error when it generates code that requires a dynamic
17408 relocation. @option{-mno-error-reloc} disables the error,
17409 @option{-mwarn-reloc} will generate a warning instead.
17412 @itemx -munsafe-dma
17414 @opindex munsafe-dma
17416 Instructions which initiate or test completion of DMA must not be
17417 reordered with respect to loads and stores of the memory which is being
17418 accessed. Users typically address this problem using the volatile
17419 keyword, but that can lead to inefficient code in places where the
17420 memory is known to not change. Rather than mark the memory as volatile
17421 we treat the DMA instructions as potentially effecting all memory. With
17422 @option{-munsafe-dma} users must use the volatile keyword to protect
17425 @item -mbranch-hints
17426 @opindex mbranch-hints
17428 By default, GCC will generate a branch hint instruction to avoid
17429 pipeline stalls for always taken or probably taken branches. A hint
17430 will not be generated closer than 8 instructions away from its branch.
17431 There is little reason to disable them, except for debugging purposes,
17432 or to make an object a little bit smaller.
17436 @opindex msmall-mem
17437 @opindex mlarge-mem
17439 By default, GCC generates code assuming that addresses are never larger
17440 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17441 a full 32 bit address.
17446 By default, GCC links against startup code that assumes the SPU-style
17447 main function interface (which has an unconventional parameter list).
17448 With @option{-mstdmain}, GCC will link your program against startup
17449 code that assumes a C99-style interface to @code{main}, including a
17450 local copy of @code{argv} strings.
17452 @item -mfixed-range=@var{register-range}
17453 @opindex mfixed-range
17454 Generate code treating the given register range as fixed registers.
17455 A fixed register is one that the register allocator can not use. This is
17456 useful when compiling kernel code. A register range is specified as
17457 two registers separated by a dash. Multiple register ranges can be
17458 specified separated by a comma.
17464 Compile code assuming that pointers to the PPU address space accessed
17465 via the @code{__ea} named address space qualifier are either 32 or 64
17466 bits wide. The default is 32 bits. As this is an ABI changing option,
17467 all object code in an executable must be compiled with the same setting.
17469 @item -maddress-space-conversion
17470 @itemx -mno-address-space-conversion
17471 @opindex maddress-space-conversion
17472 @opindex mno-address-space-conversion
17473 Allow/disallow treating the @code{__ea} address space as superset
17474 of the generic address space. This enables explicit type casts
17475 between @code{__ea} and generic pointer as well as implicit
17476 conversions of generic pointers to @code{__ea} pointers. The
17477 default is to allow address space pointer conversions.
17479 @item -mcache-size=@var{cache-size}
17480 @opindex mcache-size
17481 This option controls the version of libgcc that the compiler links to an
17482 executable and selects a software-managed cache for accessing variables
17483 in the @code{__ea} address space with a particular cache size. Possible
17484 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17485 and @samp{128}. The default cache size is 64KB.
17487 @item -matomic-updates
17488 @itemx -mno-atomic-updates
17489 @opindex matomic-updates
17490 @opindex mno-atomic-updates
17491 This option controls the version of libgcc that the compiler links to an
17492 executable and selects whether atomic updates to the software-managed
17493 cache of PPU-side variables are used. If you use atomic updates, changes
17494 to a PPU variable from SPU code using the @code{__ea} named address space
17495 qualifier will not interfere with changes to other PPU variables residing
17496 in the same cache line from PPU code. If you do not use atomic updates,
17497 such interference may occur; however, writing back cache lines will be
17498 more efficient. The default behavior is to use atomic updates.
17501 @itemx -mdual-nops=@var{n}
17502 @opindex mdual-nops
17503 By default, GCC will insert nops to increase dual issue when it expects
17504 it to increase performance. @var{n} can be a value from 0 to 10. A
17505 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17506 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17508 @item -mhint-max-nops=@var{n}
17509 @opindex mhint-max-nops
17510 Maximum number of nops to insert for a branch hint. A branch hint must
17511 be at least 8 instructions away from the branch it is effecting. GCC
17512 will insert up to @var{n} nops to enforce this, otherwise it will not
17513 generate the branch hint.
17515 @item -mhint-max-distance=@var{n}
17516 @opindex mhint-max-distance
17517 The encoding of the branch hint instruction limits the hint to be within
17518 256 instructions of the branch it is effecting. By default, GCC makes
17519 sure it is within 125.
17522 @opindex msafe-hints
17523 Work around a hardware bug which causes the SPU to stall indefinitely.
17524 By default, GCC will insert the @code{hbrp} instruction to make sure
17525 this stall won't happen.
17529 @node System V Options
17530 @subsection Options for System V
17532 These additional options are available on System V Release 4 for
17533 compatibility with other compilers on those systems:
17538 Create a shared object.
17539 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17543 Identify the versions of each tool used by the compiler, in a
17544 @code{.ident} assembler directive in the output.
17548 Refrain from adding @code{.ident} directives to the output file (this is
17551 @item -YP,@var{dirs}
17553 Search the directories @var{dirs}, and no others, for libraries
17554 specified with @option{-l}.
17556 @item -Ym,@var{dir}
17558 Look in the directory @var{dir} to find the M4 preprocessor.
17559 The assembler uses this option.
17560 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17561 @c the generic assembler that comes with Solaris takes just -Ym.
17565 @subsection V850 Options
17566 @cindex V850 Options
17568 These @samp{-m} options are defined for V850 implementations:
17572 @itemx -mno-long-calls
17573 @opindex mlong-calls
17574 @opindex mno-long-calls
17575 Treat all calls as being far away (near). If calls are assumed to be
17576 far away, the compiler will always load the functions address up into a
17577 register, and call indirect through the pointer.
17583 Do not optimize (do optimize) basic blocks that use the same index
17584 pointer 4 or more times to copy pointer into the @code{ep} register, and
17585 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17586 option is on by default if you optimize.
17588 @item -mno-prolog-function
17589 @itemx -mprolog-function
17590 @opindex mno-prolog-function
17591 @opindex mprolog-function
17592 Do not use (do use) external functions to save and restore registers
17593 at the prologue and epilogue of a function. The external functions
17594 are slower, but use less code space if more than one function saves
17595 the same number of registers. The @option{-mprolog-function} option
17596 is on by default if you optimize.
17600 Try to make the code as small as possible. At present, this just turns
17601 on the @option{-mep} and @option{-mprolog-function} options.
17603 @item -mtda=@var{n}
17605 Put static or global variables whose size is @var{n} bytes or less into
17606 the tiny data area that register @code{ep} points to. The tiny data
17607 area can hold up to 256 bytes in total (128 bytes for byte references).
17609 @item -msda=@var{n}
17611 Put static or global variables whose size is @var{n} bytes or less into
17612 the small data area that register @code{gp} points to. The small data
17613 area can hold up to 64 kilobytes.
17615 @item -mzda=@var{n}
17617 Put static or global variables whose size is @var{n} bytes or less into
17618 the first 32 kilobytes of memory.
17622 Specify that the target processor is the V850.
17625 @opindex mbig-switch
17626 Generate code suitable for big switch tables. Use this option only if
17627 the assembler/linker complain about out of range branches within a switch
17632 This option will cause r2 and r5 to be used in the code generated by
17633 the compiler. This setting is the default.
17635 @item -mno-app-regs
17636 @opindex mno-app-regs
17637 This option will cause r2 and r5 to be treated as fixed registers.
17641 Specify that the target processor is the V850E2V3. The preprocessor
17642 constants @samp{__v850e2v3__} will be defined if
17643 this option is used.
17647 Specify that the target processor is the V850E2. The preprocessor
17648 constants @samp{__v850e2__} will be defined if this option is used.
17652 Specify that the target processor is the V850E1. The preprocessor
17653 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17654 this option is used.
17658 Specify that the target processor is the V850ES. This is an alias for
17659 the @option{-mv850e1} option.
17663 Specify that the target processor is the V850E@. The preprocessor
17664 constant @samp{__v850e__} will be defined if this option is used.
17666 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17667 nor @option{-mv850e2} nor @option{-mv850e2v3}
17668 are defined then a default target processor will be chosen and the
17669 relevant @samp{__v850*__} preprocessor constant will be defined.
17671 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17672 defined, regardless of which processor variant is the target.
17674 @item -mdisable-callt
17675 @opindex mdisable-callt
17676 This option will suppress generation of the CALLT instruction for the
17677 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17678 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17683 @subsection VAX Options
17684 @cindex VAX options
17686 These @samp{-m} options are defined for the VAX:
17691 Do not output certain jump instructions (@code{aobleq} and so on)
17692 that the Unix assembler for the VAX cannot handle across long
17697 Do output those jump instructions, on the assumption that you
17698 will assemble with the GNU assembler.
17702 Output code for g-format floating point numbers instead of d-format.
17705 @node VxWorks Options
17706 @subsection VxWorks Options
17707 @cindex VxWorks Options
17709 The options in this section are defined for all VxWorks targets.
17710 Options specific to the target hardware are listed with the other
17711 options for that target.
17716 GCC can generate code for both VxWorks kernels and real time processes
17717 (RTPs). This option switches from the former to the latter. It also
17718 defines the preprocessor macro @code{__RTP__}.
17721 @opindex non-static
17722 Link an RTP executable against shared libraries rather than static
17723 libraries. The options @option{-static} and @option{-shared} can
17724 also be used for RTPs (@pxref{Link Options}); @option{-static}
17731 These options are passed down to the linker. They are defined for
17732 compatibility with Diab.
17735 @opindex Xbind-lazy
17736 Enable lazy binding of function calls. This option is equivalent to
17737 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17741 Disable lazy binding of function calls. This option is the default and
17742 is defined for compatibility with Diab.
17745 @node x86-64 Options
17746 @subsection x86-64 Options
17747 @cindex x86-64 options
17749 These are listed under @xref{i386 and x86-64 Options}.
17751 @node Xstormy16 Options
17752 @subsection Xstormy16 Options
17753 @cindex Xstormy16 Options
17755 These options are defined for Xstormy16:
17760 Choose startup files and linker script suitable for the simulator.
17763 @node Xtensa Options
17764 @subsection Xtensa Options
17765 @cindex Xtensa Options
17767 These options are supported for Xtensa targets:
17771 @itemx -mno-const16
17773 @opindex mno-const16
17774 Enable or disable use of @code{CONST16} instructions for loading
17775 constant values. The @code{CONST16} instruction is currently not a
17776 standard option from Tensilica. When enabled, @code{CONST16}
17777 instructions are always used in place of the standard @code{L32R}
17778 instructions. The use of @code{CONST16} is enabled by default only if
17779 the @code{L32R} instruction is not available.
17782 @itemx -mno-fused-madd
17783 @opindex mfused-madd
17784 @opindex mno-fused-madd
17785 Enable or disable use of fused multiply/add and multiply/subtract
17786 instructions in the floating-point option. This has no effect if the
17787 floating-point option is not also enabled. Disabling fused multiply/add
17788 and multiply/subtract instructions forces the compiler to use separate
17789 instructions for the multiply and add/subtract operations. This may be
17790 desirable in some cases where strict IEEE 754-compliant results are
17791 required: the fused multiply add/subtract instructions do not round the
17792 intermediate result, thereby producing results with @emph{more} bits of
17793 precision than specified by the IEEE standard. Disabling fused multiply
17794 add/subtract instructions also ensures that the program output is not
17795 sensitive to the compiler's ability to combine multiply and add/subtract
17798 @item -mserialize-volatile
17799 @itemx -mno-serialize-volatile
17800 @opindex mserialize-volatile
17801 @opindex mno-serialize-volatile
17802 When this option is enabled, GCC inserts @code{MEMW} instructions before
17803 @code{volatile} memory references to guarantee sequential consistency.
17804 The default is @option{-mserialize-volatile}. Use
17805 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17807 @item -mforce-no-pic
17808 @opindex mforce-no-pic
17809 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17810 position-independent code (PIC), this option disables PIC for compiling
17813 @item -mtext-section-literals
17814 @itemx -mno-text-section-literals
17815 @opindex mtext-section-literals
17816 @opindex mno-text-section-literals
17817 Control the treatment of literal pools. The default is
17818 @option{-mno-text-section-literals}, which places literals in a separate
17819 section in the output file. This allows the literal pool to be placed
17820 in a data RAM/ROM, and it also allows the linker to combine literal
17821 pools from separate object files to remove redundant literals and
17822 improve code size. With @option{-mtext-section-literals}, the literals
17823 are interspersed in the text section in order to keep them as close as
17824 possible to their references. This may be necessary for large assembly
17827 @item -mtarget-align
17828 @itemx -mno-target-align
17829 @opindex mtarget-align
17830 @opindex mno-target-align
17831 When this option is enabled, GCC instructs the assembler to
17832 automatically align instructions to reduce branch penalties at the
17833 expense of some code density. The assembler attempts to widen density
17834 instructions to align branch targets and the instructions following call
17835 instructions. If there are not enough preceding safe density
17836 instructions to align a target, no widening will be performed. The
17837 default is @option{-mtarget-align}. These options do not affect the
17838 treatment of auto-aligned instructions like @code{LOOP}, which the
17839 assembler will always align, either by widening density instructions or
17840 by inserting no-op instructions.
17843 @itemx -mno-longcalls
17844 @opindex mlongcalls
17845 @opindex mno-longcalls
17846 When this option is enabled, GCC instructs the assembler to translate
17847 direct calls to indirect calls unless it can determine that the target
17848 of a direct call is in the range allowed by the call instruction. This
17849 translation typically occurs for calls to functions in other source
17850 files. Specifically, the assembler translates a direct @code{CALL}
17851 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17852 The default is @option{-mno-longcalls}. This option should be used in
17853 programs where the call target can potentially be out of range. This
17854 option is implemented in the assembler, not the compiler, so the
17855 assembly code generated by GCC will still show direct call
17856 instructions---look at the disassembled object code to see the actual
17857 instructions. Note that the assembler will use an indirect call for
17858 every cross-file call, not just those that really will be out of range.
17861 @node zSeries Options
17862 @subsection zSeries Options
17863 @cindex zSeries options
17865 These are listed under @xref{S/390 and zSeries Options}.
17867 @node Code Gen Options
17868 @section Options for Code Generation Conventions
17869 @cindex code generation conventions
17870 @cindex options, code generation
17871 @cindex run-time options
17873 These machine-independent options control the interface conventions
17874 used in code generation.
17876 Most of them have both positive and negative forms; the negative form
17877 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17878 one of the forms is listed---the one which is not the default. You
17879 can figure out the other form by either removing @samp{no-} or adding
17883 @item -fbounds-check
17884 @opindex fbounds-check
17885 For front-ends that support it, generate additional code to check that
17886 indices used to access arrays are within the declared range. This is
17887 currently only supported by the Java and Fortran front-ends, where
17888 this option defaults to true and false respectively.
17892 This option generates traps for signed overflow on addition, subtraction,
17893 multiplication operations.
17897 This option instructs the compiler to assume that signed arithmetic
17898 overflow of addition, subtraction and multiplication wraps around
17899 using twos-complement representation. This flag enables some optimizations
17900 and disables others. This option is enabled by default for the Java
17901 front-end, as required by the Java language specification.
17904 @opindex fexceptions
17905 Enable exception handling. Generates extra code needed to propagate
17906 exceptions. For some targets, this implies GCC will generate frame
17907 unwind information for all functions, which can produce significant data
17908 size overhead, although it does not affect execution. If you do not
17909 specify this option, GCC will enable it by default for languages like
17910 C++ which normally require exception handling, and disable it for
17911 languages like C that do not normally require it. However, you may need
17912 to enable this option when compiling C code that needs to interoperate
17913 properly with exception handlers written in C++. You may also wish to
17914 disable this option if you are compiling older C++ programs that don't
17915 use exception handling.
17917 @item -fnon-call-exceptions
17918 @opindex fnon-call-exceptions
17919 Generate code that allows trapping instructions to throw exceptions.
17920 Note that this requires platform-specific runtime support that does
17921 not exist everywhere. Moreover, it only allows @emph{trapping}
17922 instructions to throw exceptions, i.e.@: memory references or floating
17923 point instructions. It does not allow exceptions to be thrown from
17924 arbitrary signal handlers such as @code{SIGALRM}.
17926 @item -funwind-tables
17927 @opindex funwind-tables
17928 Similar to @option{-fexceptions}, except that it will just generate any needed
17929 static data, but will not affect the generated code in any other way.
17930 You will normally not enable this option; instead, a language processor
17931 that needs this handling would enable it on your behalf.
17933 @item -fasynchronous-unwind-tables
17934 @opindex fasynchronous-unwind-tables
17935 Generate unwind table in dwarf2 format, if supported by target machine. The
17936 table is exact at each instruction boundary, so it can be used for stack
17937 unwinding from asynchronous events (such as debugger or garbage collector).
17939 @item -fpcc-struct-return
17940 @opindex fpcc-struct-return
17941 Return ``short'' @code{struct} and @code{union} values in memory like
17942 longer ones, rather than in registers. This convention is less
17943 efficient, but it has the advantage of allowing intercallability between
17944 GCC-compiled files and files compiled with other compilers, particularly
17945 the Portable C Compiler (pcc).
17947 The precise convention for returning structures in memory depends
17948 on the target configuration macros.
17950 Short structures and unions are those whose size and alignment match
17951 that of some integer type.
17953 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17954 switch is not binary compatible with code compiled with the
17955 @option{-freg-struct-return} switch.
17956 Use it to conform to a non-default application binary interface.
17958 @item -freg-struct-return
17959 @opindex freg-struct-return
17960 Return @code{struct} and @code{union} values in registers when possible.
17961 This is more efficient for small structures than
17962 @option{-fpcc-struct-return}.
17964 If you specify neither @option{-fpcc-struct-return} nor
17965 @option{-freg-struct-return}, GCC defaults to whichever convention is
17966 standard for the target. If there is no standard convention, GCC
17967 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17968 the principal compiler. In those cases, we can choose the standard, and
17969 we chose the more efficient register return alternative.
17971 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17972 switch is not binary compatible with code compiled with the
17973 @option{-fpcc-struct-return} switch.
17974 Use it to conform to a non-default application binary interface.
17976 @item -fshort-enums
17977 @opindex fshort-enums
17978 Allocate to an @code{enum} type only as many bytes as it needs for the
17979 declared range of possible values. Specifically, the @code{enum} type
17980 will be equivalent to the smallest integer type which has enough room.
17982 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17983 code that is not binary compatible with code generated without that switch.
17984 Use it to conform to a non-default application binary interface.
17986 @item -fshort-double
17987 @opindex fshort-double
17988 Use the same size for @code{double} as for @code{float}.
17990 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17991 code that is not binary compatible with code generated without that switch.
17992 Use it to conform to a non-default application binary interface.
17994 @item -fshort-wchar
17995 @opindex fshort-wchar
17996 Override the underlying type for @samp{wchar_t} to be @samp{short
17997 unsigned int} instead of the default for the target. This option is
17998 useful for building programs to run under WINE@.
18000 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18001 code that is not binary compatible with code generated without that switch.
18002 Use it to conform to a non-default application binary interface.
18005 @opindex fno-common
18006 In C code, controls the placement of uninitialized global variables.
18007 Unix C compilers have traditionally permitted multiple definitions of
18008 such variables in different compilation units by placing the variables
18010 This is the behavior specified by @option{-fcommon}, and is the default
18011 for GCC on most targets.
18012 On the other hand, this behavior is not required by ISO C, and on some
18013 targets may carry a speed or code size penalty on variable references.
18014 The @option{-fno-common} option specifies that the compiler should place
18015 uninitialized global variables in the data section of the object file,
18016 rather than generating them as common blocks.
18017 This has the effect that if the same variable is declared
18018 (without @code{extern}) in two different compilations,
18019 you will get a multiple-definition error when you link them.
18020 In this case, you must compile with @option{-fcommon} instead.
18021 Compiling with @option{-fno-common} is useful on targets for which
18022 it provides better performance, or if you wish to verify that the
18023 program will work on other systems which always treat uninitialized
18024 variable declarations this way.
18028 Ignore the @samp{#ident} directive.
18030 @item -finhibit-size-directive
18031 @opindex finhibit-size-directive
18032 Don't output a @code{.size} assembler directive, or anything else that
18033 would cause trouble if the function is split in the middle, and the
18034 two halves are placed at locations far apart in memory. This option is
18035 used when compiling @file{crtstuff.c}; you should not need to use it
18038 @item -fverbose-asm
18039 @opindex fverbose-asm
18040 Put extra commentary information in the generated assembly code to
18041 make it more readable. This option is generally only of use to those
18042 who actually need to read the generated assembly code (perhaps while
18043 debugging the compiler itself).
18045 @option{-fno-verbose-asm}, the default, causes the
18046 extra information to be omitted and is useful when comparing two assembler
18049 @item -frecord-gcc-switches
18050 @opindex frecord-gcc-switches
18051 This switch causes the command line that was used to invoke the
18052 compiler to be recorded into the object file that is being created.
18053 This switch is only implemented on some targets and the exact format
18054 of the recording is target and binary file format dependent, but it
18055 usually takes the form of a section containing ASCII text. This
18056 switch is related to the @option{-fverbose-asm} switch, but that
18057 switch only records information in the assembler output file as
18058 comments, so it never reaches the object file.
18059 See also @option{-grecord-gcc-switches} for another
18060 way of storing compiler options into the object file.
18064 @cindex global offset table
18066 Generate position-independent code (PIC) suitable for use in a shared
18067 library, if supported for the target machine. Such code accesses all
18068 constant addresses through a global offset table (GOT)@. The dynamic
18069 loader resolves the GOT entries when the program starts (the dynamic
18070 loader is not part of GCC; it is part of the operating system). If
18071 the GOT size for the linked executable exceeds a machine-specific
18072 maximum size, you get an error message from the linker indicating that
18073 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18074 instead. (These maximums are 8k on the SPARC and 32k
18075 on the m68k and RS/6000. The 386 has no such limit.)
18077 Position-independent code requires special support, and therefore works
18078 only on certain machines. For the 386, GCC supports PIC for System V
18079 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18080 position-independent.
18082 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18087 If supported for the target machine, emit position-independent code,
18088 suitable for dynamic linking and avoiding any limit on the size of the
18089 global offset table. This option makes a difference on the m68k,
18090 PowerPC and SPARC@.
18092 Position-independent code requires special support, and therefore works
18093 only on certain machines.
18095 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18102 These options are similar to @option{-fpic} and @option{-fPIC}, but
18103 generated position independent code can be only linked into executables.
18104 Usually these options are used when @option{-pie} GCC option will be
18105 used during linking.
18107 @option{-fpie} and @option{-fPIE} both define the macros
18108 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18109 for @option{-fpie} and 2 for @option{-fPIE}.
18111 @item -fno-jump-tables
18112 @opindex fno-jump-tables
18113 Do not use jump tables for switch statements even where it would be
18114 more efficient than other code generation strategies. This option is
18115 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18116 building code which forms part of a dynamic linker and cannot
18117 reference the address of a jump table. On some targets, jump tables
18118 do not require a GOT and this option is not needed.
18120 @item -ffixed-@var{reg}
18122 Treat the register named @var{reg} as a fixed register; generated code
18123 should never refer to it (except perhaps as a stack pointer, frame
18124 pointer or in some other fixed role).
18126 @var{reg} must be the name of a register. The register names accepted
18127 are machine-specific and are defined in the @code{REGISTER_NAMES}
18128 macro in the machine description macro file.
18130 This flag does not have a negative form, because it specifies a
18133 @item -fcall-used-@var{reg}
18134 @opindex fcall-used
18135 Treat the register named @var{reg} as an allocable register that is
18136 clobbered by function calls. It may be allocated for temporaries or
18137 variables that do not live across a call. Functions compiled this way
18138 will not save and restore the register @var{reg}.
18140 It is an error to used this flag with the frame pointer or stack pointer.
18141 Use of this flag for other registers that have fixed pervasive roles in
18142 the machine's execution model will produce disastrous results.
18144 This flag does not have a negative form, because it specifies a
18147 @item -fcall-saved-@var{reg}
18148 @opindex fcall-saved
18149 Treat the register named @var{reg} as an allocable register saved by
18150 functions. It may be allocated even for temporaries or variables that
18151 live across a call. Functions compiled this way will save and restore
18152 the register @var{reg} if they use it.
18154 It is an error to used this flag with the frame pointer or stack pointer.
18155 Use of this flag for other registers that have fixed pervasive roles in
18156 the machine's execution model will produce disastrous results.
18158 A different sort of disaster will result from the use of this flag for
18159 a register in which function values may be returned.
18161 This flag does not have a negative form, because it specifies a
18164 @item -fpack-struct[=@var{n}]
18165 @opindex fpack-struct
18166 Without a value specified, pack all structure members together without
18167 holes. When a value is specified (which must be a small power of two), pack
18168 structure members according to this value, representing the maximum
18169 alignment (that is, objects with default alignment requirements larger than
18170 this will be output potentially unaligned at the next fitting location.
18172 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18173 code that is not binary compatible with code generated without that switch.
18174 Additionally, it makes the code suboptimal.
18175 Use it to conform to a non-default application binary interface.
18177 @item -finstrument-functions
18178 @opindex finstrument-functions
18179 Generate instrumentation calls for entry and exit to functions. Just
18180 after function entry and just before function exit, the following
18181 profiling functions will be called with the address of the current
18182 function and its call site. (On some platforms,
18183 @code{__builtin_return_address} does not work beyond the current
18184 function, so the call site information may not be available to the
18185 profiling functions otherwise.)
18188 void __cyg_profile_func_enter (void *this_fn,
18190 void __cyg_profile_func_exit (void *this_fn,
18194 The first argument is the address of the start of the current function,
18195 which may be looked up exactly in the symbol table.
18197 This instrumentation is also done for functions expanded inline in other
18198 functions. The profiling calls will indicate where, conceptually, the
18199 inline function is entered and exited. This means that addressable
18200 versions of such functions must be available. If all your uses of a
18201 function are expanded inline, this may mean an additional expansion of
18202 code size. If you use @samp{extern inline} in your C code, an
18203 addressable version of such functions must be provided. (This is
18204 normally the case anyways, but if you get lucky and the optimizer always
18205 expands the functions inline, you might have gotten away without
18206 providing static copies.)
18208 A function may be given the attribute @code{no_instrument_function}, in
18209 which case this instrumentation will not be done. This can be used, for
18210 example, for the profiling functions listed above, high-priority
18211 interrupt routines, and any functions from which the profiling functions
18212 cannot safely be called (perhaps signal handlers, if the profiling
18213 routines generate output or allocate memory).
18215 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18216 @opindex finstrument-functions-exclude-file-list
18218 Set the list of functions that are excluded from instrumentation (see
18219 the description of @code{-finstrument-functions}). If the file that
18220 contains a function definition matches with one of @var{file}, then
18221 that function is not instrumented. The match is done on substrings:
18222 if the @var{file} parameter is a substring of the file name, it is
18223 considered to be a match.
18228 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18232 will exclude any inline function defined in files whose pathnames
18233 contain @code{/bits/stl} or @code{include/sys}.
18235 If, for some reason, you want to include letter @code{','} in one of
18236 @var{sym}, write @code{'\,'}. For example,
18237 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18238 (note the single quote surrounding the option).
18240 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18241 @opindex finstrument-functions-exclude-function-list
18243 This is similar to @code{-finstrument-functions-exclude-file-list},
18244 but this option sets the list of function names to be excluded from
18245 instrumentation. The function name to be matched is its user-visible
18246 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18247 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18248 match is done on substrings: if the @var{sym} parameter is a substring
18249 of the function name, it is considered to be a match. For C99 and C++
18250 extended identifiers, the function name must be given in UTF-8, not
18251 using universal character names.
18253 @item -fstack-check
18254 @opindex fstack-check
18255 Generate code to verify that you do not go beyond the boundary of the
18256 stack. You should specify this flag if you are running in an
18257 environment with multiple threads, but only rarely need to specify it in
18258 a single-threaded environment since stack overflow is automatically
18259 detected on nearly all systems if there is only one stack.
18261 Note that this switch does not actually cause checking to be done; the
18262 operating system or the language runtime must do that. The switch causes
18263 generation of code to ensure that they see the stack being extended.
18265 You can additionally specify a string parameter: @code{no} means no
18266 checking, @code{generic} means force the use of old-style checking,
18267 @code{specific} means use the best checking method and is equivalent
18268 to bare @option{-fstack-check}.
18270 Old-style checking is a generic mechanism that requires no specific
18271 target support in the compiler but comes with the following drawbacks:
18275 Modified allocation strategy for large objects: they will always be
18276 allocated dynamically if their size exceeds a fixed threshold.
18279 Fixed limit on the size of the static frame of functions: when it is
18280 topped by a particular function, stack checking is not reliable and
18281 a warning is issued by the compiler.
18284 Inefficiency: because of both the modified allocation strategy and the
18285 generic implementation, the performances of the code are hampered.
18288 Note that old-style stack checking is also the fallback method for
18289 @code{specific} if no target support has been added in the compiler.
18291 @item -fstack-limit-register=@var{reg}
18292 @itemx -fstack-limit-symbol=@var{sym}
18293 @itemx -fno-stack-limit
18294 @opindex fstack-limit-register
18295 @opindex fstack-limit-symbol
18296 @opindex fno-stack-limit
18297 Generate code to ensure that the stack does not grow beyond a certain value,
18298 either the value of a register or the address of a symbol. If the stack
18299 would grow beyond the value, a signal is raised. For most targets,
18300 the signal is raised before the stack overruns the boundary, so
18301 it is possible to catch the signal without taking special precautions.
18303 For instance, if the stack starts at absolute address @samp{0x80000000}
18304 and grows downwards, you can use the flags
18305 @option{-fstack-limit-symbol=__stack_limit} and
18306 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18307 of 128KB@. Note that this may only work with the GNU linker.
18309 @item -fsplit-stack
18310 @opindex fsplit-stack
18311 Generate code to automatically split the stack before it overflows.
18312 The resulting program has a discontiguous stack which can only
18313 overflow if the program is unable to allocate any more memory. This
18314 is most useful when running threaded programs, as it is no longer
18315 necessary to calculate a good stack size to use for each thread. This
18316 is currently only implemented for the i386 and x86_64 backends running
18319 When code compiled with @option{-fsplit-stack} calls code compiled
18320 without @option{-fsplit-stack}, there may not be much stack space
18321 available for the latter code to run. If compiling all code,
18322 including library code, with @option{-fsplit-stack} is not an option,
18323 then the linker can fix up these calls so that the code compiled
18324 without @option{-fsplit-stack} always has a large stack. Support for
18325 this is implemented in the gold linker in GNU binutils release 2.21
18328 @item -fleading-underscore
18329 @opindex fleading-underscore
18330 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18331 change the way C symbols are represented in the object file. One use
18332 is to help link with legacy assembly code.
18334 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18335 generate code that is not binary compatible with code generated without that
18336 switch. Use it to conform to a non-default application binary interface.
18337 Not all targets provide complete support for this switch.
18339 @item -ftls-model=@var{model}
18340 @opindex ftls-model
18341 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18342 The @var{model} argument should be one of @code{global-dynamic},
18343 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18345 The default without @option{-fpic} is @code{initial-exec}; with
18346 @option{-fpic} the default is @code{global-dynamic}.
18348 @item -fvisibility=@var{default|internal|hidden|protected}
18349 @opindex fvisibility
18350 Set the default ELF image symbol visibility to the specified option---all
18351 symbols will be marked with this unless overridden within the code.
18352 Using this feature can very substantially improve linking and
18353 load times of shared object libraries, produce more optimized
18354 code, provide near-perfect API export and prevent symbol clashes.
18355 It is @strong{strongly} recommended that you use this in any shared objects
18358 Despite the nomenclature, @code{default} always means public; i.e.,
18359 available to be linked against from outside the shared object.
18360 @code{protected} and @code{internal} are pretty useless in real-world
18361 usage so the only other commonly used option will be @code{hidden}.
18362 The default if @option{-fvisibility} isn't specified is
18363 @code{default}, i.e., make every
18364 symbol public---this causes the same behavior as previous versions of
18367 A good explanation of the benefits offered by ensuring ELF
18368 symbols have the correct visibility is given by ``How To Write
18369 Shared Libraries'' by Ulrich Drepper (which can be found at
18370 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18371 solution made possible by this option to marking things hidden when
18372 the default is public is to make the default hidden and mark things
18373 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18374 and @code{__attribute__ ((visibility("default")))} instead of
18375 @code{__declspec(dllexport)} you get almost identical semantics with
18376 identical syntax. This is a great boon to those working with
18377 cross-platform projects.
18379 For those adding visibility support to existing code, you may find
18380 @samp{#pragma GCC visibility} of use. This works by you enclosing
18381 the declarations you wish to set visibility for with (for example)
18382 @samp{#pragma GCC visibility push(hidden)} and
18383 @samp{#pragma GCC visibility pop}.
18384 Bear in mind that symbol visibility should be viewed @strong{as
18385 part of the API interface contract} and thus all new code should
18386 always specify visibility when it is not the default; i.e., declarations
18387 only for use within the local DSO should @strong{always} be marked explicitly
18388 as hidden as so to avoid PLT indirection overheads---making this
18389 abundantly clear also aids readability and self-documentation of the code.
18390 Note that due to ISO C++ specification requirements, operator new and
18391 operator delete must always be of default visibility.
18393 Be aware that headers from outside your project, in particular system
18394 headers and headers from any other library you use, may not be
18395 expecting to be compiled with visibility other than the default. You
18396 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18397 before including any such headers.
18399 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18400 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18401 no modifications. However, this means that calls to @samp{extern}
18402 functions with no explicit visibility will use the PLT, so it is more
18403 effective to use @samp{__attribute ((visibility))} and/or
18404 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18405 declarations should be treated as hidden.
18407 Note that @samp{-fvisibility} does affect C++ vague linkage
18408 entities. This means that, for instance, an exception class that will
18409 be thrown between DSOs must be explicitly marked with default
18410 visibility so that the @samp{type_info} nodes will be unified between
18413 An overview of these techniques, their benefits and how to use them
18414 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18416 @item -fstrict-volatile-bitfields
18417 @opindex fstrict-volatile-bitfields
18418 This option should be used if accesses to volatile bitfields (or other
18419 structure fields, although the compiler usually honors those types
18420 anyway) should use a single access of the width of the
18421 field's type, aligned to a natural alignment if possible. For
18422 example, targets with memory-mapped peripheral registers might require
18423 all such accesses to be 16 bits wide; with this flag the user could
18424 declare all peripheral bitfields as ``unsigned short'' (assuming short
18425 is 16 bits on these targets) to force GCC to use 16 bit accesses
18426 instead of, perhaps, a more efficient 32 bit access.
18428 If this option is disabled, the compiler will use the most efficient
18429 instruction. In the previous example, that might be a 32-bit load
18430 instruction, even though that will access bytes that do not contain
18431 any portion of the bitfield, or memory-mapped registers unrelated to
18432 the one being updated.
18434 If the target requires strict alignment, and honoring the field
18435 type would require violating this alignment, a warning is issued.
18436 If the field has @code{packed} attribute, the access is done without
18437 honoring the field type. If the field doesn't have @code{packed}
18438 attribute, the access is done honoring the field type. In both cases,
18439 GCC assumes that the user knows something about the target hardware
18440 that it is unaware of.
18442 The default value of this option is determined by the application binary
18443 interface for the target processor.
18449 @node Environment Variables
18450 @section Environment Variables Affecting GCC
18451 @cindex environment variables
18453 @c man begin ENVIRONMENT
18454 This section describes several environment variables that affect how GCC
18455 operates. Some of them work by specifying directories or prefixes to use
18456 when searching for various kinds of files. Some are used to specify other
18457 aspects of the compilation environment.
18459 Note that you can also specify places to search using options such as
18460 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18461 take precedence over places specified using environment variables, which
18462 in turn take precedence over those specified by the configuration of GCC@.
18463 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18464 GNU Compiler Collection (GCC) Internals}.
18469 @c @itemx LC_COLLATE
18471 @c @itemx LC_MONETARY
18472 @c @itemx LC_NUMERIC
18477 @c @findex LC_COLLATE
18478 @findex LC_MESSAGES
18479 @c @findex LC_MONETARY
18480 @c @findex LC_NUMERIC
18484 These environment variables control the way that GCC uses
18485 localization information that allow GCC to work with different
18486 national conventions. GCC inspects the locale categories
18487 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18488 so. These locale categories can be set to any value supported by your
18489 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18490 Kingdom encoded in UTF-8.
18492 The @env{LC_CTYPE} environment variable specifies character
18493 classification. GCC uses it to determine the character boundaries in
18494 a string; this is needed for some multibyte encodings that contain quote
18495 and escape characters that would otherwise be interpreted as a string
18498 The @env{LC_MESSAGES} environment variable specifies the language to
18499 use in diagnostic messages.
18501 If the @env{LC_ALL} environment variable is set, it overrides the value
18502 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18503 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18504 environment variable. If none of these variables are set, GCC
18505 defaults to traditional C English behavior.
18509 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18510 files. GCC uses temporary files to hold the output of one stage of
18511 compilation which is to be used as input to the next stage: for example,
18512 the output of the preprocessor, which is the input to the compiler
18515 @item GCC_COMPARE_DEBUG
18516 @findex GCC_COMPARE_DEBUG
18517 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
18518 @option{-fcompare-debug} to the compiler driver. See the documentation
18519 of this option for more details.
18521 @item GCC_EXEC_PREFIX
18522 @findex GCC_EXEC_PREFIX
18523 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18524 names of the subprograms executed by the compiler. No slash is added
18525 when this prefix is combined with the name of a subprogram, but you can
18526 specify a prefix that ends with a slash if you wish.
18528 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18529 an appropriate prefix to use based on the pathname it was invoked with.
18531 If GCC cannot find the subprogram using the specified prefix, it
18532 tries looking in the usual places for the subprogram.
18534 The default value of @env{GCC_EXEC_PREFIX} is
18535 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18536 the installed compiler. In many cases @var{prefix} is the value
18537 of @code{prefix} when you ran the @file{configure} script.
18539 Other prefixes specified with @option{-B} take precedence over this prefix.
18541 This prefix is also used for finding files such as @file{crt0.o} that are
18544 In addition, the prefix is used in an unusual way in finding the
18545 directories to search for header files. For each of the standard
18546 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18547 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18548 replacing that beginning with the specified prefix to produce an
18549 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18550 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18551 These alternate directories are searched first; the standard directories
18552 come next. If a standard directory begins with the configured
18553 @var{prefix} then the value of @var{prefix} is replaced by
18554 @env{GCC_EXEC_PREFIX} when looking for header files.
18556 @item COMPILER_PATH
18557 @findex COMPILER_PATH
18558 The value of @env{COMPILER_PATH} is a colon-separated list of
18559 directories, much like @env{PATH}. GCC tries the directories thus
18560 specified when searching for subprograms, if it can't find the
18561 subprograms using @env{GCC_EXEC_PREFIX}.
18564 @findex LIBRARY_PATH
18565 The value of @env{LIBRARY_PATH} is a colon-separated list of
18566 directories, much like @env{PATH}. When configured as a native compiler,
18567 GCC tries the directories thus specified when searching for special
18568 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18569 using GCC also uses these directories when searching for ordinary
18570 libraries for the @option{-l} option (but directories specified with
18571 @option{-L} come first).
18575 @cindex locale definition
18576 This variable is used to pass locale information to the compiler. One way in
18577 which this information is used is to determine the character set to be used
18578 when character literals, string literals and comments are parsed in C and C++.
18579 When the compiler is configured to allow multibyte characters,
18580 the following values for @env{LANG} are recognized:
18584 Recognize JIS characters.
18586 Recognize SJIS characters.
18588 Recognize EUCJP characters.
18591 If @env{LANG} is not defined, or if it has some other value, then the
18592 compiler will use mblen and mbtowc as defined by the default locale to
18593 recognize and translate multibyte characters.
18597 Some additional environments variables affect the behavior of the
18600 @include cppenv.texi
18604 @node Precompiled Headers
18605 @section Using Precompiled Headers
18606 @cindex precompiled headers
18607 @cindex speed of compilation
18609 Often large projects have many header files that are included in every
18610 source file. The time the compiler takes to process these header files
18611 over and over again can account for nearly all of the time required to
18612 build the project. To make builds faster, GCC allows users to
18613 `precompile' a header file; then, if builds can use the precompiled
18614 header file they will be much faster.
18616 To create a precompiled header file, simply compile it as you would any
18617 other file, if necessary using the @option{-x} option to make the driver
18618 treat it as a C or C++ header file. You will probably want to use a
18619 tool like @command{make} to keep the precompiled header up-to-date when
18620 the headers it contains change.
18622 A precompiled header file will be searched for when @code{#include} is
18623 seen in the compilation. As it searches for the included file
18624 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18625 compiler looks for a precompiled header in each directory just before it
18626 looks for the include file in that directory. The name searched for is
18627 the name specified in the @code{#include} with @samp{.gch} appended. If
18628 the precompiled header file can't be used, it is ignored.
18630 For instance, if you have @code{#include "all.h"}, and you have
18631 @file{all.h.gch} in the same directory as @file{all.h}, then the
18632 precompiled header file will be used if possible, and the original
18633 header will be used otherwise.
18635 Alternatively, you might decide to put the precompiled header file in a
18636 directory and use @option{-I} to ensure that directory is searched
18637 before (or instead of) the directory containing the original header.
18638 Then, if you want to check that the precompiled header file is always
18639 used, you can put a file of the same name as the original header in this
18640 directory containing an @code{#error} command.
18642 This also works with @option{-include}. So yet another way to use
18643 precompiled headers, good for projects not designed with precompiled
18644 header files in mind, is to simply take most of the header files used by
18645 a project, include them from another header file, precompile that header
18646 file, and @option{-include} the precompiled header. If the header files
18647 have guards against multiple inclusion, they will be skipped because
18648 they've already been included (in the precompiled header).
18650 If you need to precompile the same header file for different
18651 languages, targets, or compiler options, you can instead make a
18652 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18653 header in the directory, perhaps using @option{-o}. It doesn't matter
18654 what you call the files in the directory, every precompiled header in
18655 the directory will be considered. The first precompiled header
18656 encountered in the directory that is valid for this compilation will
18657 be used; they're searched in no particular order.
18659 There are many other possibilities, limited only by your imagination,
18660 good sense, and the constraints of your build system.
18662 A precompiled header file can be used only when these conditions apply:
18666 Only one precompiled header can be used in a particular compilation.
18669 A precompiled header can't be used once the first C token is seen. You
18670 can have preprocessor directives before a precompiled header; you can
18671 even include a precompiled header from inside another header, so long as
18672 there are no C tokens before the @code{#include}.
18675 The precompiled header file must be produced for the same language as
18676 the current compilation. You can't use a C precompiled header for a C++
18680 The precompiled header file must have been produced by the same compiler
18681 binary as the current compilation is using.
18684 Any macros defined before the precompiled header is included must
18685 either be defined in the same way as when the precompiled header was
18686 generated, or must not affect the precompiled header, which usually
18687 means that they don't appear in the precompiled header at all.
18689 The @option{-D} option is one way to define a macro before a
18690 precompiled header is included; using a @code{#define} can also do it.
18691 There are also some options that define macros implicitly, like
18692 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18695 @item If debugging information is output when using the precompiled
18696 header, using @option{-g} or similar, the same kind of debugging information
18697 must have been output when building the precompiled header. However,
18698 a precompiled header built using @option{-g} can be used in a compilation
18699 when no debugging information is being output.
18701 @item The same @option{-m} options must generally be used when building
18702 and using the precompiled header. @xref{Submodel Options},
18703 for any cases where this rule is relaxed.
18705 @item Each of the following options must be the same when building and using
18706 the precompiled header:
18708 @gccoptlist{-fexceptions}
18711 Some other command-line options starting with @option{-f},
18712 @option{-p}, or @option{-O} must be defined in the same way as when
18713 the precompiled header was generated. At present, it's not clear
18714 which options are safe to change and which are not; the safest choice
18715 is to use exactly the same options when generating and using the
18716 precompiled header. The following are known to be safe:
18718 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18719 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18720 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18725 For all of these except the last, the compiler will automatically
18726 ignore the precompiled header if the conditions aren't met. If you
18727 find an option combination that doesn't work and doesn't cause the
18728 precompiled header to be ignored, please consider filing a bug report,
18731 If you do use differing options when generating and using the
18732 precompiled header, the actual behavior will be a mixture of the
18733 behavior for the options. For instance, if you use @option{-g} to
18734 generate the precompiled header but not when using it, you may or may
18735 not get debugging information for routines in the precompiled header.