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} -Wno-free-nonheap-object -Wjump-misses-init @gol
248 -Wignored-qualifiers @gol
249 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
250 -Winit-self -Winline -Wmaybe-uninitialized @gol
251 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
252 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
253 -Wlogical-op -Wlong-long @gol
254 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
255 -Wmissing-format-attribute -Wmissing-include-dirs @gol
257 -Wno-multichar -Wnonnull -Wno-overflow @gol
258 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
259 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
260 -Wpointer-arith -Wno-pointer-to-int-cast @gol
261 -Wredundant-decls @gol
262 -Wreturn-type -Wsequence-point -Wshadow @gol
263 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
264 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
265 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
266 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
267 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
268 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
269 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
270 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
271 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
272 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
273 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
274 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
276 @item C and Objective-C-only Warning Options
277 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
278 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
279 -Wold-style-declaration -Wold-style-definition @gol
280 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
281 -Wdeclaration-after-statement -Wpointer-sign}
283 @item Debugging Options
284 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
285 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
286 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
287 -fdisable-ipa-@var{pass_name} @gol
288 -fdisable-rtl-@var{pass_name} @gol
289 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
290 -fdisable-tree-@var{pass_name} @gol
291 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
292 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
293 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
294 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
295 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
297 -fdump-statistics @gol
299 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
303 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-nrv -fdump-tree-vect @gol
313 -fdump-tree-sink @gol
314 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
318 -ftree-vectorizer-verbose=@var{n} @gol
319 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
320 -fdump-final-insns=@var{file} @gol
321 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
322 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
323 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
324 -fenable-@var{kind}-@var{pass} @gol
325 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
326 -fdebug-types-section @gol
327 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
328 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
329 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
330 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
331 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
332 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
333 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
334 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
335 -gvms -gxcoff -gxcoff+ @gol
336 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
337 -fdebug-prefix-map=@var{old}=@var{new} @gol
338 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
339 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
340 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
341 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
342 -print-prog-name=@var{program} -print-search-dirs -Q @gol
343 -print-sysroot -print-sysroot-headers-suffix @gol
344 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
346 @item Optimization Options
347 @xref{Optimize Options,,Options that Control Optimization}.
348 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
349 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
350 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
351 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
352 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
353 -fcompare-elim -fcprop-registers -fcrossjumping @gol
354 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
355 -fcx-limited-range @gol
356 -fdata-sections -fdce -fdce -fdelayed-branch @gol
357 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
358 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
359 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
360 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
361 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
362 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
363 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
364 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
365 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
366 -fira-algorithm=@var{algorithm} @gol
367 -fira-region=@var{region} @gol
368 -fira-loop-pressure -fno-ira-share-save-slots @gol
369 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
370 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
371 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
372 -floop-parallelize-all -flto -flto-compression-level @gol
373 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
374 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
375 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
376 -fno-default-inline @gol
377 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
378 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
379 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
380 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
381 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
382 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
383 -fprefetch-loop-arrays @gol
384 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
385 -fprofile-generate=@var{path} @gol
386 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
387 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
388 -freorder-blocks-and-partition -freorder-functions @gol
389 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
390 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
391 -fsched-spec-load -fsched-spec-load-dangerous @gol
392 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
393 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
394 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
395 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
396 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
397 -fselective-scheduling -fselective-scheduling2 @gol
398 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
399 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
400 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
401 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
403 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
404 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
405 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
406 -ftree-loop-if-convert-stores -ftree-loop-im @gol
407 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
408 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
409 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
410 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
411 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
412 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
413 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
414 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
415 -fwhole-program -fwpa -fuse-linker-plugin @gol
416 --param @var{name}=@var{value}
417 -O -O0 -O1 -O2 -O3 -Os -Ofast}
419 @item Preprocessor Options
420 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
421 @gccoptlist{-A@var{question}=@var{answer} @gol
422 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
423 -C -dD -dI -dM -dN @gol
424 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
425 -idirafter @var{dir} @gol
426 -include @var{file} -imacros @var{file} @gol
427 -iprefix @var{file} -iwithprefix @var{dir} @gol
428 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
429 -imultilib @var{dir} -isysroot @var{dir} @gol
430 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
431 -P -fworking-directory -remap @gol
432 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
433 -Xpreprocessor @var{option}}
435 @item Assembler Option
436 @xref{Assembler Options,,Passing Options to the Assembler}.
437 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
440 @xref{Link Options,,Options for Linking}.
441 @gccoptlist{@var{object-file-name} -l@var{library} @gol
442 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
443 -s -static -static-libgcc -static-libstdc++ -shared @gol
444 -shared-libgcc -symbolic @gol
445 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
448 @item Directory Options
449 @xref{Directory Options,,Options for Directory Search}.
450 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
451 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
454 @item Machine Dependent Options
455 @xref{Submodel Options,,Hardware Models and Configurations}.
456 @c This list is ordered alphanumerically by subsection name.
457 @c Try and put the significant identifier (CPU or system) first,
458 @c so users have a clue at guessing where the ones they want will be.
461 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
462 -mabi=@var{name} @gol
463 -mapcs-stack-check -mno-apcs-stack-check @gol
464 -mapcs-float -mno-apcs-float @gol
465 -mapcs-reentrant -mno-apcs-reentrant @gol
466 -msched-prolog -mno-sched-prolog @gol
467 -mlittle-endian -mbig-endian -mwords-little-endian @gol
468 -mfloat-abi=@var{name} -mfpe @gol
469 -mfp16-format=@var{name}
470 -mthumb-interwork -mno-thumb-interwork @gol
471 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
472 -mstructure-size-boundary=@var{n} @gol
473 -mabort-on-noreturn @gol
474 -mlong-calls -mno-long-calls @gol
475 -msingle-pic-base -mno-single-pic-base @gol
476 -mpic-register=@var{reg} @gol
477 -mnop-fun-dllimport @gol
478 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
479 -mpoke-function-name @gol
481 -mtpcs-frame -mtpcs-leaf-frame @gol
482 -mcaller-super-interworking -mcallee-super-interworking @gol
483 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
484 -mword-relocations @gol
485 -mfix-cortex-m3-ldrd}
488 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
489 -mcall-prologues -mtiny-stack -mint8}
491 @emph{Blackfin Options}
492 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
493 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
494 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
495 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
496 -mno-id-shared-library -mshared-library-id=@var{n} @gol
497 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
498 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
499 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
503 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
504 -msim -msdata=@var{sdata-type}}
507 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
508 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
509 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
510 -mstack-align -mdata-align -mconst-align @gol
511 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
512 -melf -maout -melinux -mlinux -sim -sim2 @gol
513 -mmul-bug-workaround -mno-mul-bug-workaround}
515 @emph{Darwin Options}
516 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
517 -arch_only -bind_at_load -bundle -bundle_loader @gol
518 -client_name -compatibility_version -current_version @gol
520 -dependency-file -dylib_file -dylinker_install_name @gol
521 -dynamic -dynamiclib -exported_symbols_list @gol
522 -filelist -flat_namespace -force_cpusubtype_ALL @gol
523 -force_flat_namespace -headerpad_max_install_names @gol
525 -image_base -init -install_name -keep_private_externs @gol
526 -multi_module -multiply_defined -multiply_defined_unused @gol
527 -noall_load -no_dead_strip_inits_and_terms @gol
528 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
529 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
530 -private_bundle -read_only_relocs -sectalign @gol
531 -sectobjectsymbols -whyload -seg1addr @gol
532 -sectcreate -sectobjectsymbols -sectorder @gol
533 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
534 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
535 -segprot -segs_read_only_addr -segs_read_write_addr @gol
536 -single_module -static -sub_library -sub_umbrella @gol
537 -twolevel_namespace -umbrella -undefined @gol
538 -unexported_symbols_list -weak_reference_mismatches @gol
539 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
540 -mkernel -mone-byte-bool}
542 @emph{DEC Alpha Options}
543 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
544 -mieee -mieee-with-inexact -mieee-conformant @gol
545 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
546 -mtrap-precision=@var{mode} -mbuild-constants @gol
547 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
548 -mbwx -mmax -mfix -mcix @gol
549 -mfloat-vax -mfloat-ieee @gol
550 -mexplicit-relocs -msmall-data -mlarge-data @gol
551 -msmall-text -mlarge-text @gol
552 -mmemory-latency=@var{time}}
554 @emph{DEC Alpha/VMS Options}
555 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
558 @gccoptlist{-msmall-model -mno-lsim}
561 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
562 -mhard-float -msoft-float @gol
563 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
564 -mdouble -mno-double @gol
565 -mmedia -mno-media -mmuladd -mno-muladd @gol
566 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
567 -mlinked-fp -mlong-calls -malign-labels @gol
568 -mlibrary-pic -macc-4 -macc-8 @gol
569 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
570 -moptimize-membar -mno-optimize-membar @gol
571 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
572 -mvliw-branch -mno-vliw-branch @gol
573 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
574 -mno-nested-cond-exec -mtomcat-stats @gol
578 @emph{GNU/Linux Options}
579 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
580 -tno-android-cc -tno-android-ld}
582 @emph{H8/300 Options}
583 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
586 @gccoptlist{-march=@var{architecture-type} @gol
587 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
588 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
589 -mfixed-range=@var{register-range} @gol
590 -mjump-in-delay -mlinker-opt -mlong-calls @gol
591 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
592 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
593 -mno-jump-in-delay -mno-long-load-store @gol
594 -mno-portable-runtime -mno-soft-float @gol
595 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
596 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
597 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
598 -munix=@var{unix-std} -nolibdld -static -threads}
600 @emph{i386 and x86-64 Options}
601 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
602 -mfpmath=@var{unit} @gol
603 -masm=@var{dialect} -mno-fancy-math-387 @gol
604 -mno-fp-ret-in-387 -msoft-float @gol
605 -mno-wide-multiply -mrtd -malign-double @gol
606 -mpreferred-stack-boundary=@var{num} @gol
607 -mincoming-stack-boundary=@var{num} @gol
608 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
609 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
610 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
611 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
612 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
613 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
614 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
615 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
616 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
617 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
618 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
619 -mcmodel=@var{code-model} -mabi=@var{name} @gol
620 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
621 -msse2avx -mfentry -m8bit-idiv @gol
622 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
624 @emph{i386 and x86-64 Windows Options}
625 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
626 -mnop-fun-dllimport -mthread @gol
627 -municode -mwin32 -mwindows -fno-set-stack-executable}
630 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
631 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
632 -mconstant-gp -mauto-pic -mfused-madd @gol
633 -minline-float-divide-min-latency @gol
634 -minline-float-divide-max-throughput @gol
635 -mno-inline-float-divide @gol
636 -minline-int-divide-min-latency @gol
637 -minline-int-divide-max-throughput @gol
638 -mno-inline-int-divide @gol
639 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
640 -mno-inline-sqrt @gol
641 -mdwarf2-asm -mearly-stop-bits @gol
642 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
643 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
644 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
645 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
646 -msched-spec-ldc -msched-spec-control-ldc @gol
647 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
648 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
649 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
650 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
652 @emph{IA-64/VMS Options}
653 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
656 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
657 -msign-extend-enabled -muser-enabled}
659 @emph{M32R/D Options}
660 @gccoptlist{-m32r2 -m32rx -m32r @gol
662 -malign-loops -mno-align-loops @gol
663 -missue-rate=@var{number} @gol
664 -mbranch-cost=@var{number} @gol
665 -mmodel=@var{code-size-model-type} @gol
666 -msdata=@var{sdata-type} @gol
667 -mno-flush-func -mflush-func=@var{name} @gol
668 -mno-flush-trap -mflush-trap=@var{number} @gol
672 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
674 @emph{M680x0 Options}
675 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
676 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
677 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
678 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
679 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
680 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
681 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
682 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
686 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
687 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
688 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
689 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
690 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
693 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
694 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
695 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
696 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
699 @emph{MicroBlaze Options}
700 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
701 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
702 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
703 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
704 -mxl-mode-@var{app-model}}
707 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
708 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
709 -mips64 -mips64r2 @gol
710 -mips16 -mno-mips16 -mflip-mips16 @gol
711 -minterlink-mips16 -mno-interlink-mips16 @gol
712 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
713 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
714 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
715 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
716 -mfpu=@var{fpu-type} @gol
717 -msmartmips -mno-smartmips @gol
718 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
719 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
720 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
721 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
722 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
723 -membedded-data -mno-embedded-data @gol
724 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
725 -mcode-readable=@var{setting} @gol
726 -msplit-addresses -mno-split-addresses @gol
727 -mexplicit-relocs -mno-explicit-relocs @gol
728 -mcheck-zero-division -mno-check-zero-division @gol
729 -mdivide-traps -mdivide-breaks @gol
730 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
731 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
732 -mfix-24k -mno-fix-24k @gol
733 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
734 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
735 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
736 -mflush-func=@var{func} -mno-flush-func @gol
737 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
738 -mfp-exceptions -mno-fp-exceptions @gol
739 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
740 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
743 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
744 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
745 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
746 -mno-base-addresses -msingle-exit -mno-single-exit}
748 @emph{MN10300 Options}
749 @gccoptlist{-mmult-bug -mno-mult-bug @gol
750 -mno-am33 -mam33 -mam33-2 -mam34 @gol
751 -mtune=@var{cpu-type} @gol
752 -mreturn-pointer-on-d0 @gol
753 -mno-crt0 -mrelax -mliw -msetlb}
755 @emph{PDP-11 Options}
756 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
757 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
758 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
759 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
760 -mbranch-expensive -mbranch-cheap @gol
761 -munix-asm -mdec-asm}
763 @emph{picoChip Options}
764 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
765 -msymbol-as-address -mno-inefficient-warnings}
767 @emph{PowerPC Options}
768 See RS/6000 and PowerPC Options.
770 @emph{RS/6000 and PowerPC Options}
771 @gccoptlist{-mcpu=@var{cpu-type} @gol
772 -mtune=@var{cpu-type} @gol
773 -mcmodel=@var{code-model} @gol
774 -mpower -mno-power -mpower2 -mno-power2 @gol
775 -mpowerpc -mpowerpc64 -mno-powerpc @gol
776 -maltivec -mno-altivec @gol
777 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
778 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
779 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
780 -mfprnd -mno-fprnd @gol
781 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
782 -mnew-mnemonics -mold-mnemonics @gol
783 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
784 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
785 -malign-power -malign-natural @gol
786 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
787 -msingle-float -mdouble-float -msimple-fpu @gol
788 -mstring -mno-string -mupdate -mno-update @gol
789 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
790 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
791 -mstrict-align -mno-strict-align -mrelocatable @gol
792 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
793 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
794 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
795 -mprioritize-restricted-insns=@var{priority} @gol
796 -msched-costly-dep=@var{dependence_type} @gol
797 -minsert-sched-nops=@var{scheme} @gol
798 -mcall-sysv -mcall-netbsd @gol
799 -maix-struct-return -msvr4-struct-return @gol
800 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
801 -mblock-move-inline-limit=@var{num} @gol
802 -misel -mno-isel @gol
803 -misel=yes -misel=no @gol
805 -mspe=yes -mspe=no @gol
807 -mgen-cell-microcode -mwarn-cell-microcode @gol
808 -mvrsave -mno-vrsave @gol
809 -mmulhw -mno-mulhw @gol
810 -mdlmzb -mno-dlmzb @gol
811 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
812 -mprototype -mno-prototype @gol
813 -msim -mmvme -mads -myellowknife -memb -msdata @gol
814 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
815 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
816 -mno-recip-precision @gol
817 -mveclibabi=@var{type} -mfriz -mno-friz @gol
818 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
819 -msave-toc-indirect -mno-save-toc-indirect}
822 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
824 -mbig-endian-data -mlittle-endian-data @gol
827 -mas100-syntax -mno-as100-syntax@gol
829 -mmax-constant-size=@gol
831 -msave-acc-in-interrupts}
833 @emph{S/390 and zSeries Options}
834 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
835 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
836 -mlong-double-64 -mlong-double-128 @gol
837 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
838 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
839 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
840 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
841 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
844 @gccoptlist{-meb -mel @gol
848 -mscore5 -mscore5u -mscore7 -mscore7d}
851 @gccoptlist{-m1 -m2 -m2e @gol
852 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
854 -m4-nofpu -m4-single-only -m4-single -m4 @gol
855 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
856 -m5-64media -m5-64media-nofpu @gol
857 -m5-32media -m5-32media-nofpu @gol
858 -m5-compact -m5-compact-nofpu @gol
859 -mb -ml -mdalign -mrelax @gol
860 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
861 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
862 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
863 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
864 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
865 -maccumulate-outgoing-args -minvalid-symbols}
867 @emph{Solaris 2 Options}
868 @gccoptlist{-mimpure-text -mno-impure-text @gol
872 @gccoptlist{-mcpu=@var{cpu-type} @gol
873 -mtune=@var{cpu-type} @gol
874 -mcmodel=@var{code-model} @gol
875 -m32 -m64 -mapp-regs -mno-app-regs @gol
876 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
877 -mfpu -mno-fpu -mhard-float -msoft-float @gol
878 -mhard-quad-float -msoft-quad-float @gol
880 -mstack-bias -mno-stack-bias @gol
881 -munaligned-doubles -mno-unaligned-doubles @gol
882 -mv8plus -mno-v8plus -mvis -mno-vis @gol
886 @gccoptlist{-mwarn-reloc -merror-reloc @gol
887 -msafe-dma -munsafe-dma @gol
889 -msmall-mem -mlarge-mem -mstdmain @gol
890 -mfixed-range=@var{register-range} @gol
892 -maddress-space-conversion -mno-address-space-conversion @gol
893 -mcache-size=@var{cache-size} @gol
894 -matomic-updates -mno-atomic-updates}
896 @emph{System V Options}
897 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
900 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
901 -mprolog-function -mno-prolog-function -mspace @gol
902 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
903 -mapp-regs -mno-app-regs @gol
904 -mdisable-callt -mno-disable-callt @gol
907 -mv850e1 -mv850es @gol
912 @gccoptlist{-mg -mgnu -munix}
914 @emph{VxWorks Options}
915 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
916 -Xbind-lazy -Xbind-now}
918 @emph{x86-64 Options}
919 See i386 and x86-64 Options.
921 @emph{Xstormy16 Options}
924 @emph{Xtensa Options}
925 @gccoptlist{-mconst16 -mno-const16 @gol
926 -mfused-madd -mno-fused-madd @gol
928 -mserialize-volatile -mno-serialize-volatile @gol
929 -mtext-section-literals -mno-text-section-literals @gol
930 -mtarget-align -mno-target-align @gol
931 -mlongcalls -mno-longcalls}
933 @emph{zSeries Options}
934 See S/390 and zSeries Options.
936 @item Code Generation Options
937 @xref{Code Gen Options,,Options for Code Generation Conventions}.
938 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
939 -ffixed-@var{reg} -fexceptions @gol
940 -fnon-call-exceptions -funwind-tables @gol
941 -fasynchronous-unwind-tables @gol
942 -finhibit-size-directive -finstrument-functions @gol
943 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
944 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
945 -fno-common -fno-ident @gol
946 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
947 -fno-jump-tables @gol
948 -frecord-gcc-switches @gol
949 -freg-struct-return -fshort-enums @gol
950 -fshort-double -fshort-wchar @gol
951 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
952 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
953 -fno-stack-limit -fsplit-stack @gol
954 -fleading-underscore -ftls-model=@var{model} @gol
955 -ftrapv -fwrapv -fbounds-check @gol
956 -fvisibility -fstrict-volatile-bitfields}
960 * Overall Options:: Controlling the kind of output:
961 an executable, object files, assembler files,
962 or preprocessed source.
963 * C Dialect Options:: Controlling the variant of C language compiled.
964 * C++ Dialect Options:: Variations on C++.
965 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
967 * Language Independent Options:: Controlling how diagnostics should be
969 * Warning Options:: How picky should the compiler be?
970 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
971 * Optimize Options:: How much optimization?
972 * Preprocessor Options:: Controlling header files and macro definitions.
973 Also, getting dependency information for Make.
974 * Assembler Options:: Passing options to the assembler.
975 * Link Options:: Specifying libraries and so on.
976 * Directory Options:: Where to find header files and libraries.
977 Where to find the compiler executable files.
978 * Spec Files:: How to pass switches to sub-processes.
979 * Target Options:: Running a cross-compiler, or an old version of GCC.
982 @node Overall Options
983 @section Options Controlling the Kind of Output
985 Compilation can involve up to four stages: preprocessing, compilation
986 proper, assembly and linking, always in that order. GCC is capable of
987 preprocessing and compiling several files either into several
988 assembler input files, or into one assembler input file; then each
989 assembler input file produces an object file, and linking combines all
990 the object files (those newly compiled, and those specified as input)
991 into an executable file.
993 @cindex file name suffix
994 For any given input file, the file name suffix determines what kind of
999 C source code which must be preprocessed.
1002 C source code which should not be preprocessed.
1005 C++ source code which should not be preprocessed.
1008 Objective-C source code. Note that you must link with the @file{libobjc}
1009 library to make an Objective-C program work.
1012 Objective-C source code which should not be preprocessed.
1016 Objective-C++ source code. Note that you must link with the @file{libobjc}
1017 library to make an Objective-C++ program work. Note that @samp{.M} refers
1018 to a literal capital M@.
1020 @item @var{file}.mii
1021 Objective-C++ source code which should not be preprocessed.
1024 C, C++, Objective-C or Objective-C++ header file to be turned into a
1025 precompiled header (default), or C, C++ header file to be turned into an
1026 Ada spec (via the @option{-fdump-ada-spec} switch).
1029 @itemx @var{file}.cp
1030 @itemx @var{file}.cxx
1031 @itemx @var{file}.cpp
1032 @itemx @var{file}.CPP
1033 @itemx @var{file}.c++
1035 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1036 the last two letters must both be literally @samp{x}. Likewise,
1037 @samp{.C} refers to a literal capital C@.
1041 Objective-C++ source code which must be preprocessed.
1043 @item @var{file}.mii
1044 Objective-C++ source code which should not be preprocessed.
1048 @itemx @var{file}.hp
1049 @itemx @var{file}.hxx
1050 @itemx @var{file}.hpp
1051 @itemx @var{file}.HPP
1052 @itemx @var{file}.h++
1053 @itemx @var{file}.tcc
1054 C++ header file to be turned into a precompiled header or Ada spec.
1057 @itemx @var{file}.for
1058 @itemx @var{file}.ftn
1059 Fixed form Fortran source code which should not be preprocessed.
1062 @itemx @var{file}.FOR
1063 @itemx @var{file}.fpp
1064 @itemx @var{file}.FPP
1065 @itemx @var{file}.FTN
1066 Fixed form Fortran source code which must be preprocessed (with the traditional
1069 @item @var{file}.f90
1070 @itemx @var{file}.f95
1071 @itemx @var{file}.f03
1072 @itemx @var{file}.f08
1073 Free form Fortran source code which should not be preprocessed.
1075 @item @var{file}.F90
1076 @itemx @var{file}.F95
1077 @itemx @var{file}.F03
1078 @itemx @var{file}.F08
1079 Free form Fortran source code which must be preprocessed (with the
1080 traditional preprocessor).
1085 @c FIXME: Descriptions of Java file types.
1091 @item @var{file}.ads
1092 Ada source code file which contains a library unit declaration (a
1093 declaration of a package, subprogram, or generic, or a generic
1094 instantiation), or a library unit renaming declaration (a package,
1095 generic, or subprogram renaming declaration). Such files are also
1098 @item @var{file}.adb
1099 Ada source code file containing a library unit body (a subprogram or
1100 package body). Such files are also called @dfn{bodies}.
1102 @c GCC also knows about some suffixes for languages not yet included:
1113 @itemx @var{file}.sx
1114 Assembler code which must be preprocessed.
1117 An object file to be fed straight into linking.
1118 Any file name with no recognized suffix is treated this way.
1122 You can specify the input language explicitly with the @option{-x} option:
1125 @item -x @var{language}
1126 Specify explicitly the @var{language} for the following input files
1127 (rather than letting the compiler choose a default based on the file
1128 name suffix). This option applies to all following input files until
1129 the next @option{-x} option. Possible values for @var{language} are:
1131 c c-header cpp-output
1132 c++ c++-header c++-cpp-output
1133 objective-c objective-c-header objective-c-cpp-output
1134 objective-c++ objective-c++-header objective-c++-cpp-output
1135 assembler assembler-with-cpp
1137 f77 f77-cpp-input f95 f95-cpp-input
1143 Turn off any specification of a language, so that subsequent files are
1144 handled according to their file name suffixes (as they are if @option{-x}
1145 has not been used at all).
1147 @item -pass-exit-codes
1148 @opindex pass-exit-codes
1149 Normally the @command{gcc} program will exit with the code of 1 if any
1150 phase of the compiler returns a non-success return code. If you specify
1151 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1152 numerically highest error produced by any phase that returned an error
1153 indication. The C, C++, and Fortran frontends return 4, if an internal
1154 compiler error is encountered.
1157 If you only want some of the stages of compilation, you can use
1158 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1159 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1160 @command{gcc} is to stop. Note that some combinations (for example,
1161 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1166 Compile or assemble the source files, but do not link. The linking
1167 stage simply is not done. The ultimate output is in the form of an
1168 object file for each source file.
1170 By default, the object file name for a source file is made by replacing
1171 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1173 Unrecognized input files, not requiring compilation or assembly, are
1178 Stop after the stage of compilation proper; do not assemble. The output
1179 is in the form of an assembler code file for each non-assembler input
1182 By default, the assembler file name for a source file is made by
1183 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1185 Input files that don't require compilation are ignored.
1189 Stop after the preprocessing stage; do not run the compiler proper. The
1190 output is in the form of preprocessed source code, which is sent to the
1193 Input files which don't require preprocessing are ignored.
1195 @cindex output file option
1198 Place output in file @var{file}. This applies regardless to whatever
1199 sort of output is being produced, whether it be an executable file,
1200 an object file, an assembler file or preprocessed C code.
1202 If @option{-o} is not specified, the default is to put an executable
1203 file in @file{a.out}, the object file for
1204 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1205 assembler file in @file{@var{source}.s}, a precompiled header file in
1206 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1211 Print (on standard error output) the commands executed to run the stages
1212 of compilation. Also print the version number of the compiler driver
1213 program and of the preprocessor and the compiler proper.
1217 Like @option{-v} except the commands are not executed and arguments
1218 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1219 This is useful for shell scripts to capture the driver-generated command lines.
1223 Use pipes rather than temporary files for communication between the
1224 various stages of compilation. This fails to work on some systems where
1225 the assembler is unable to read from a pipe; but the GNU assembler has
1230 Print (on the standard output) a description of the command line options
1231 understood by @command{gcc}. If the @option{-v} option is also specified
1232 then @option{--help} will also be passed on to the various processes
1233 invoked by @command{gcc}, so that they can display the command line options
1234 they accept. If the @option{-Wextra} option has also been specified
1235 (prior to the @option{--help} option), then command line options which
1236 have no documentation associated with them will also be displayed.
1239 @opindex target-help
1240 Print (on the standard output) a description of target-specific command
1241 line options for each tool. For some targets extra target-specific
1242 information may also be printed.
1244 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1245 Print (on the standard output) a description of the command line
1246 options understood by the compiler that fit into all specified classes
1247 and qualifiers. These are the supported classes:
1250 @item @samp{optimizers}
1251 This will display all of the optimization options supported by the
1254 @item @samp{warnings}
1255 This will display all of the options controlling warning messages
1256 produced by the compiler.
1259 This will display target-specific options. Unlike the
1260 @option{--target-help} option however, target-specific options of the
1261 linker and assembler will not be displayed. This is because those
1262 tools do not currently support the extended @option{--help=} syntax.
1265 This will display the values recognized by the @option{--param}
1268 @item @var{language}
1269 This will display the options supported for @var{language}, where
1270 @var{language} is the name of one of the languages supported in this
1274 This will display the options that are common to all languages.
1277 These are the supported qualifiers:
1280 @item @samp{undocumented}
1281 Display only those options which are undocumented.
1284 Display options which take an argument that appears after an equal
1285 sign in the same continuous piece of text, such as:
1286 @samp{--help=target}.
1288 @item @samp{separate}
1289 Display options which take an argument that appears as a separate word
1290 following the original option, such as: @samp{-o output-file}.
1293 Thus for example to display all the undocumented target-specific
1294 switches supported by the compiler the following can be used:
1297 --help=target,undocumented
1300 The sense of a qualifier can be inverted by prefixing it with the
1301 @samp{^} character, so for example to display all binary warning
1302 options (i.e., ones that are either on or off and that do not take an
1303 argument), which have a description the following can be used:
1306 --help=warnings,^joined,^undocumented
1309 The argument to @option{--help=} should not consist solely of inverted
1312 Combining several classes is possible, although this usually
1313 restricts the output by so much that there is nothing to display. One
1314 case where it does work however is when one of the classes is
1315 @var{target}. So for example to display all the target-specific
1316 optimization options the following can be used:
1319 --help=target,optimizers
1322 The @option{--help=} option can be repeated on the command line. Each
1323 successive use will display its requested class of options, skipping
1324 those that have already been displayed.
1326 If the @option{-Q} option appears on the command line before the
1327 @option{--help=} option, then the descriptive text displayed by
1328 @option{--help=} is changed. Instead of describing the displayed
1329 options, an indication is given as to whether the option is enabled,
1330 disabled or set to a specific value (assuming that the compiler
1331 knows this at the point where the @option{--help=} option is used).
1333 Here is a truncated example from the ARM port of @command{gcc}:
1336 % gcc -Q -mabi=2 --help=target -c
1337 The following options are target specific:
1339 -mabort-on-noreturn [disabled]
1343 The output is sensitive to the effects of previous command line
1344 options, so for example it is possible to find out which optimizations
1345 are enabled at @option{-O2} by using:
1348 -Q -O2 --help=optimizers
1351 Alternatively you can discover which binary optimizations are enabled
1352 by @option{-O3} by using:
1355 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1356 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1357 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1360 @item -no-canonical-prefixes
1361 @opindex no-canonical-prefixes
1362 Do not expand any symbolic links, resolve references to @samp{/../}
1363 or @samp{/./}, or make the path absolute when generating a relative
1368 Display the version number and copyrights of the invoked GCC@.
1372 Invoke all subcommands under a wrapper program. The name of the
1373 wrapper program and its parameters are passed as a comma separated
1377 gcc -c t.c -wrapper gdb,--args
1380 This will invoke all subprograms of @command{gcc} under
1381 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1382 @samp{gdb --args cc1 @dots{}}.
1384 @item -fplugin=@var{name}.so
1385 Load the plugin code in file @var{name}.so, assumed to be a
1386 shared object to be dlopen'd by the compiler. The base name of
1387 the shared object file is used to identify the plugin for the
1388 purposes of argument parsing (See
1389 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1390 Each plugin should define the callback functions specified in the
1393 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1394 Define an argument called @var{key} with a value of @var{value}
1395 for the plugin called @var{name}.
1397 @item -fdump-ada-spec@r{[}-slim@r{]}
1398 For C and C++ source and include files, generate corresponding Ada
1399 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1400 GNAT User's Guide}, which provides detailed documentation on this feature.
1402 @item -fdump-go-spec=@var{file}
1403 For input files in any language, generate corresponding Go
1404 declarations in @var{file}. This generates Go @code{const},
1405 @code{type}, @code{var}, and @code{func} declarations which may be a
1406 useful way to start writing a Go interface to code written in some
1409 @include @value{srcdir}/../libiberty/at-file.texi
1413 @section Compiling C++ Programs
1415 @cindex suffixes for C++ source
1416 @cindex C++ source file suffixes
1417 C++ source files conventionally use one of the suffixes @samp{.C},
1418 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1419 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1420 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1421 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1422 files with these names and compiles them as C++ programs even if you
1423 call the compiler the same way as for compiling C programs (usually
1424 with the name @command{gcc}).
1428 However, the use of @command{gcc} does not add the C++ library.
1429 @command{g++} is a program that calls GCC and treats @samp{.c},
1430 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1431 files unless @option{-x} is used, and automatically specifies linking
1432 against the C++ library. This program is also useful when
1433 precompiling a C header file with a @samp{.h} extension for use in C++
1434 compilations. On many systems, @command{g++} is also installed with
1435 the name @command{c++}.
1437 @cindex invoking @command{g++}
1438 When you compile C++ programs, you may specify many of the same
1439 command-line options that you use for compiling programs in any
1440 language; or command-line options meaningful for C and related
1441 languages; or options that are meaningful only for C++ programs.
1442 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1443 explanations of options for languages related to C@.
1444 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1445 explanations of options that are meaningful only for C++ programs.
1447 @node C Dialect Options
1448 @section Options Controlling C Dialect
1449 @cindex dialect options
1450 @cindex language dialect options
1451 @cindex options, dialect
1453 The following options control the dialect of C (or languages derived
1454 from C, such as C++, Objective-C and Objective-C++) that the compiler
1458 @cindex ANSI support
1462 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1463 equivalent to @samp{-std=c++98}.
1465 This turns off certain features of GCC that are incompatible with ISO
1466 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1467 such as the @code{asm} and @code{typeof} keywords, and
1468 predefined macros such as @code{unix} and @code{vax} that identify the
1469 type of system you are using. It also enables the undesirable and
1470 rarely used ISO trigraph feature. For the C compiler,
1471 it disables recognition of C++ style @samp{//} comments as well as
1472 the @code{inline} keyword.
1474 The alternate keywords @code{__asm__}, @code{__extension__},
1475 @code{__inline__} and @code{__typeof__} continue to work despite
1476 @option{-ansi}. You would not want to use them in an ISO C program, of
1477 course, but it is useful to put them in header files that might be included
1478 in compilations done with @option{-ansi}. Alternate predefined macros
1479 such as @code{__unix__} and @code{__vax__} are also available, with or
1480 without @option{-ansi}.
1482 The @option{-ansi} option does not cause non-ISO programs to be
1483 rejected gratuitously. For that, @option{-pedantic} is required in
1484 addition to @option{-ansi}. @xref{Warning Options}.
1486 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1487 option is used. Some header files may notice this macro and refrain
1488 from declaring certain functions or defining certain macros that the
1489 ISO standard doesn't call for; this is to avoid interfering with any
1490 programs that might use these names for other things.
1492 Functions that would normally be built in but do not have semantics
1493 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1494 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1495 built-in functions provided by GCC}, for details of the functions
1500 Determine the language standard. @xref{Standards,,Language Standards
1501 Supported by GCC}, for details of these standard versions. This option
1502 is currently only supported when compiling C or C++.
1504 The compiler can accept several base standards, such as @samp{c90} or
1505 @samp{c++98}, and GNU dialects of those standards, such as
1506 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1507 compiler will accept all programs following that standard and those
1508 using GNU extensions that do not contradict it. For example,
1509 @samp{-std=c90} turns off certain features of GCC that are
1510 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1511 keywords, but not other GNU extensions that do not have a meaning in
1512 ISO C90, such as omitting the middle term of a @code{?:}
1513 expression. On the other hand, by specifying a GNU dialect of a
1514 standard, all features the compiler support are enabled, even when
1515 those features change the meaning of the base standard and some
1516 strict-conforming programs may be rejected. The particular standard
1517 is used by @option{-pedantic} to identify which features are GNU
1518 extensions given that version of the standard. For example
1519 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1520 comments, while @samp{-std=gnu99 -pedantic} would not.
1522 A value for this option must be provided; possible values are
1528 Support all ISO C90 programs (certain GNU extensions that conflict
1529 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1531 @item iso9899:199409
1532 ISO C90 as modified in amendment 1.
1538 ISO C99. Note that this standard is not yet fully supported; see
1539 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1540 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1543 ISO C1X, the draft of the next revision of the ISO C standard.
1544 Support is limited and experimental and features enabled by this
1545 option may be changed or removed if changed in or removed from the
1550 GNU dialect of ISO C90 (including some C99 features). This
1551 is the default for C code.
1555 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1556 this will become the default. The name @samp{gnu9x} is deprecated.
1559 GNU dialect of ISO C1X. Support is limited and experimental and
1560 features enabled by this option may be changed or removed if changed
1561 in or removed from the standard draft.
1564 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1568 GNU dialect of @option{-std=c++98}. This is the default for
1572 The working draft of the upcoming ISO C++0x standard. This option
1573 enables experimental features that are likely to be included in
1574 C++0x. The working draft is constantly changing, and any feature that is
1575 enabled by this flag may be removed from future versions of GCC if it is
1576 not part of the C++0x standard.
1579 GNU dialect of @option{-std=c++0x}. This option enables
1580 experimental features that may be removed in future versions of GCC.
1583 @item -fgnu89-inline
1584 @opindex fgnu89-inline
1585 The option @option{-fgnu89-inline} tells GCC to use the traditional
1586 GNU semantics for @code{inline} functions when in C99 mode.
1587 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1588 is accepted and ignored by GCC versions 4.1.3 up to but not including
1589 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1590 C99 mode. Using this option is roughly equivalent to adding the
1591 @code{gnu_inline} function attribute to all inline functions
1592 (@pxref{Function Attributes}).
1594 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1595 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1596 specifies the default behavior). This option was first supported in
1597 GCC 4.3. This option is not supported in @option{-std=c90} or
1598 @option{-std=gnu90} mode.
1600 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1601 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1602 in effect for @code{inline} functions. @xref{Common Predefined
1603 Macros,,,cpp,The C Preprocessor}.
1605 @item -aux-info @var{filename}
1607 Output to the given filename prototyped declarations for all functions
1608 declared and/or defined in a translation unit, including those in header
1609 files. This option is silently ignored in any language other than C@.
1611 Besides declarations, the file indicates, in comments, the origin of
1612 each declaration (source file and line), whether the declaration was
1613 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1614 @samp{O} for old, respectively, in the first character after the line
1615 number and the colon), and whether it came from a declaration or a
1616 definition (@samp{C} or @samp{F}, respectively, in the following
1617 character). In the case of function definitions, a K&R-style list of
1618 arguments followed by their declarations is also provided, inside
1619 comments, after the declaration.
1623 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1624 keyword, so that code can use these words as identifiers. You can use
1625 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1626 instead. @option{-ansi} implies @option{-fno-asm}.
1628 In C++, this switch only affects the @code{typeof} keyword, since
1629 @code{asm} and @code{inline} are standard keywords. You may want to
1630 use the @option{-fno-gnu-keywords} flag instead, which has the same
1631 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1632 switch only affects the @code{asm} and @code{typeof} keywords, since
1633 @code{inline} is a standard keyword in ISO C99.
1636 @itemx -fno-builtin-@var{function}
1637 @opindex fno-builtin
1638 @cindex built-in functions
1639 Don't recognize built-in functions that do not begin with
1640 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1641 functions provided by GCC}, for details of the functions affected,
1642 including those which are not built-in functions when @option{-ansi} or
1643 @option{-std} options for strict ISO C conformance are used because they
1644 do not have an ISO standard meaning.
1646 GCC normally generates special code to handle certain built-in functions
1647 more efficiently; for instance, calls to @code{alloca} may become single
1648 instructions that adjust the stack directly, and calls to @code{memcpy}
1649 may become inline copy loops. The resulting code is often both smaller
1650 and faster, but since the function calls no longer appear as such, you
1651 cannot set a breakpoint on those calls, nor can you change the behavior
1652 of the functions by linking with a different library. In addition,
1653 when a function is recognized as a built-in function, GCC may use
1654 information about that function to warn about problems with calls to
1655 that function, or to generate more efficient code, even if the
1656 resulting code still contains calls to that function. For example,
1657 warnings are given with @option{-Wformat} for bad calls to
1658 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1659 known not to modify global memory.
1661 With the @option{-fno-builtin-@var{function}} option
1662 only the built-in function @var{function} is
1663 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1664 function is named that is not built-in in this version of GCC, this
1665 option is ignored. There is no corresponding
1666 @option{-fbuiltin-@var{function}} option; if you wish to enable
1667 built-in functions selectively when using @option{-fno-builtin} or
1668 @option{-ffreestanding}, you may define macros such as:
1671 #define abs(n) __builtin_abs ((n))
1672 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1677 @cindex hosted environment
1679 Assert that compilation takes place in a hosted environment. This implies
1680 @option{-fbuiltin}. A hosted environment is one in which the
1681 entire standard library is available, and in which @code{main} has a return
1682 type of @code{int}. Examples are nearly everything except a kernel.
1683 This is equivalent to @option{-fno-freestanding}.
1685 @item -ffreestanding
1686 @opindex ffreestanding
1687 @cindex hosted environment
1689 Assert that compilation takes place in a freestanding environment. This
1690 implies @option{-fno-builtin}. A freestanding environment
1691 is one in which the standard library may not exist, and program startup may
1692 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1693 This is equivalent to @option{-fno-hosted}.
1695 @xref{Standards,,Language Standards Supported by GCC}, for details of
1696 freestanding and hosted environments.
1700 @cindex OpenMP parallel
1701 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1702 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1703 compiler generates parallel code according to the OpenMP Application
1704 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1705 implies @option{-pthread}, and thus is only supported on targets that
1706 have support for @option{-pthread}.
1708 @item -fms-extensions
1709 @opindex fms-extensions
1710 Accept some non-standard constructs used in Microsoft header files.
1712 In C++ code, this allows member names in structures to be similar
1713 to previous types declarations.
1722 Some cases of unnamed fields in structures and unions are only
1723 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1724 fields within structs/unions}, for details.
1726 @item -fplan9-extensions
1727 Accept some non-standard constructs used in Plan 9 code.
1729 This enables @option{-fms-extensions}, permits passing pointers to
1730 structures with anonymous fields to functions which expect pointers to
1731 elements of the type of the field, and permits referring to anonymous
1732 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1733 struct/union fields within structs/unions}, for details. This is only
1734 supported for C, not C++.
1738 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1739 options for strict ISO C conformance) implies @option{-trigraphs}.
1741 @item -no-integrated-cpp
1742 @opindex no-integrated-cpp
1743 Performs a compilation in two passes: preprocessing and compiling. This
1744 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1745 @option{-B} option. The user supplied compilation step can then add in
1746 an additional preprocessing step after normal preprocessing but before
1747 compiling. The default is to use the integrated cpp (internal cpp)
1749 The semantics of this option will change if "cc1", "cc1plus", and
1750 "cc1obj" are merged.
1752 @cindex traditional C language
1753 @cindex C language, traditional
1755 @itemx -traditional-cpp
1756 @opindex traditional-cpp
1757 @opindex traditional
1758 Formerly, these options caused GCC to attempt to emulate a pre-standard
1759 C compiler. They are now only supported with the @option{-E} switch.
1760 The preprocessor continues to support a pre-standard mode. See the GNU
1761 CPP manual for details.
1763 @item -fcond-mismatch
1764 @opindex fcond-mismatch
1765 Allow conditional expressions with mismatched types in the second and
1766 third arguments. The value of such an expression is void. This option
1767 is not supported for C++.
1769 @item -flax-vector-conversions
1770 @opindex flax-vector-conversions
1771 Allow implicit conversions between vectors with differing numbers of
1772 elements and/or incompatible element types. This option should not be
1775 @item -funsigned-char
1776 @opindex funsigned-char
1777 Let the type @code{char} be unsigned, like @code{unsigned char}.
1779 Each kind of machine has a default for what @code{char} should
1780 be. It is either like @code{unsigned char} by default or like
1781 @code{signed char} by default.
1783 Ideally, a portable program should always use @code{signed char} or
1784 @code{unsigned char} when it depends on the signedness of an object.
1785 But many programs have been written to use plain @code{char} and
1786 expect it to be signed, or expect it to be unsigned, depending on the
1787 machines they were written for. This option, and its inverse, let you
1788 make such a program work with the opposite default.
1790 The type @code{char} is always a distinct type from each of
1791 @code{signed char} or @code{unsigned char}, even though its behavior
1792 is always just like one of those two.
1795 @opindex fsigned-char
1796 Let the type @code{char} be signed, like @code{signed char}.
1798 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1799 the negative form of @option{-funsigned-char}. Likewise, the option
1800 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1802 @item -fsigned-bitfields
1803 @itemx -funsigned-bitfields
1804 @itemx -fno-signed-bitfields
1805 @itemx -fno-unsigned-bitfields
1806 @opindex fsigned-bitfields
1807 @opindex funsigned-bitfields
1808 @opindex fno-signed-bitfields
1809 @opindex fno-unsigned-bitfields
1810 These options control whether a bit-field is signed or unsigned, when the
1811 declaration does not use either @code{signed} or @code{unsigned}. By
1812 default, such a bit-field is signed, because this is consistent: the
1813 basic integer types such as @code{int} are signed types.
1816 @node C++ Dialect Options
1817 @section Options Controlling C++ Dialect
1819 @cindex compiler options, C++
1820 @cindex C++ options, command line
1821 @cindex options, C++
1822 This section describes the command-line options that are only meaningful
1823 for C++ programs; but you can also use most of the GNU compiler options
1824 regardless of what language your program is in. For example, you
1825 might compile a file @code{firstClass.C} like this:
1828 g++ -g -frepo -O -c firstClass.C
1832 In this example, only @option{-frepo} is an option meant
1833 only for C++ programs; you can use the other options with any
1834 language supported by GCC@.
1836 Here is a list of options that are @emph{only} for compiling C++ programs:
1840 @item -fabi-version=@var{n}
1841 @opindex fabi-version
1842 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1843 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1844 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1845 the version that conforms most closely to the C++ ABI specification.
1846 Therefore, the ABI obtained using version 0 will change as ABI bugs
1849 The default is version 2.
1851 Version 3 corrects an error in mangling a constant address as a
1854 Version 4 implements a standard mangling for vector types.
1856 Version 5 corrects the mangling of attribute const/volatile on
1857 function pointer types, decltype of a plain decl, and use of a
1858 function parameter in the declaration of another parameter.
1860 Version 6 corrects the promotion behavior of C++0x scoped enums.
1862 See also @option{-Wabi}.
1864 @item -fno-access-control
1865 @opindex fno-access-control
1866 Turn off all access checking. This switch is mainly useful for working
1867 around bugs in the access control code.
1871 Check that the pointer returned by @code{operator new} is non-null
1872 before attempting to modify the storage allocated. This check is
1873 normally unnecessary because the C++ standard specifies that
1874 @code{operator new} will only return @code{0} if it is declared
1875 @samp{throw()}, in which case the compiler will always check the
1876 return value even without this option. In all other cases, when
1877 @code{operator new} has a non-empty exception specification, memory
1878 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1879 @samp{new (nothrow)}.
1881 @item -fconserve-space
1882 @opindex fconserve-space
1883 Put uninitialized or runtime-initialized global variables into the
1884 common segment, as C does. This saves space in the executable at the
1885 cost of not diagnosing duplicate definitions. If you compile with this
1886 flag and your program mysteriously crashes after @code{main()} has
1887 completed, you may have an object that is being destroyed twice because
1888 two definitions were merged.
1890 This option is no longer useful on most targets, now that support has
1891 been added for putting variables into BSS without making them common.
1893 @item -fconstexpr-depth=@var{n}
1894 @opindex fconstexpr-depth
1895 Set the maximum nested evaluation depth for C++0x constexpr functions
1896 to @var{n}. A limit is needed to detect endless recursion during
1897 constant expression evaluation. The minimum specified by the standard
1900 @item -fno-deduce-init-list
1901 @opindex fno-deduce-init-list
1902 Disable deduction of a template type parameter as
1903 std::initializer_list from a brace-enclosed initializer list, i.e.
1906 template <class T> auto forward(T t) -> decltype (realfn (t))
1913 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1917 This option is present because this deduction is an extension to the
1918 current specification in the C++0x working draft, and there was
1919 some concern about potential overload resolution problems.
1921 @item -ffriend-injection
1922 @opindex ffriend-injection
1923 Inject friend functions into the enclosing namespace, so that they are
1924 visible outside the scope of the class in which they are declared.
1925 Friend functions were documented to work this way in the old Annotated
1926 C++ Reference Manual, and versions of G++ before 4.1 always worked
1927 that way. However, in ISO C++ a friend function which is not declared
1928 in an enclosing scope can only be found using argument dependent
1929 lookup. This option causes friends to be injected as they were in
1932 This option is for compatibility, and may be removed in a future
1935 @item -fno-elide-constructors
1936 @opindex fno-elide-constructors
1937 The C++ standard allows an implementation to omit creating a temporary
1938 which is only used to initialize another object of the same type.
1939 Specifying this option disables that optimization, and forces G++ to
1940 call the copy constructor in all cases.
1942 @item -fno-enforce-eh-specs
1943 @opindex fno-enforce-eh-specs
1944 Don't generate code to check for violation of exception specifications
1945 at runtime. This option violates the C++ standard, but may be useful
1946 for reducing code size in production builds, much like defining
1947 @samp{NDEBUG}. This does not give user code permission to throw
1948 exceptions in violation of the exception specifications; the compiler
1949 will still optimize based on the specifications, so throwing an
1950 unexpected exception will result in undefined behavior.
1953 @itemx -fno-for-scope
1955 @opindex fno-for-scope
1956 If @option{-ffor-scope} is specified, the scope of variables declared in
1957 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1958 as specified by the C++ standard.
1959 If @option{-fno-for-scope} is specified, the scope of variables declared in
1960 a @i{for-init-statement} extends to the end of the enclosing scope,
1961 as was the case in old versions of G++, and other (traditional)
1962 implementations of C++.
1964 The default if neither flag is given to follow the standard,
1965 but to allow and give a warning for old-style code that would
1966 otherwise be invalid, or have different behavior.
1968 @item -fno-gnu-keywords
1969 @opindex fno-gnu-keywords
1970 Do not recognize @code{typeof} as a keyword, so that code can use this
1971 word as an identifier. You can use the keyword @code{__typeof__} instead.
1972 @option{-ansi} implies @option{-fno-gnu-keywords}.
1974 @item -fno-implicit-templates
1975 @opindex fno-implicit-templates
1976 Never emit code for non-inline templates which are instantiated
1977 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1978 @xref{Template Instantiation}, for more information.
1980 @item -fno-implicit-inline-templates
1981 @opindex fno-implicit-inline-templates
1982 Don't emit code for implicit instantiations of inline templates, either.
1983 The default is to handle inlines differently so that compiles with and
1984 without optimization will need the same set of explicit instantiations.
1986 @item -fno-implement-inlines
1987 @opindex fno-implement-inlines
1988 To save space, do not emit out-of-line copies of inline functions
1989 controlled by @samp{#pragma implementation}. This will cause linker
1990 errors if these functions are not inlined everywhere they are called.
1992 @item -fms-extensions
1993 @opindex fms-extensions
1994 Disable pedantic warnings about constructs used in MFC, such as implicit
1995 int and getting a pointer to member function via non-standard syntax.
1997 @item -fno-nonansi-builtins
1998 @opindex fno-nonansi-builtins
1999 Disable built-in declarations of functions that are not mandated by
2000 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2001 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2004 @opindex fnothrow-opt
2005 Treat a @code{throw()} exception specification as though it were a
2006 @code{noexcept} specification to reduce or eliminate the text size
2007 overhead relative to a function with no exception specification. If
2008 the function has local variables of types with non-trivial
2009 destructors, the exception specification will actually make the
2010 function smaller because the EH cleanups for those variables can be
2011 optimized away. The semantic effect is that an exception thrown out of
2012 a function with such an exception specification will result in a call
2013 to @code{terminate} rather than @code{unexpected}.
2015 @item -fno-operator-names
2016 @opindex fno-operator-names
2017 Do not treat the operator name keywords @code{and}, @code{bitand},
2018 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2019 synonyms as keywords.
2021 @item -fno-optional-diags
2022 @opindex fno-optional-diags
2023 Disable diagnostics that the standard says a compiler does not need to
2024 issue. Currently, the only such diagnostic issued by G++ is the one for
2025 a name having multiple meanings within a class.
2028 @opindex fpermissive
2029 Downgrade some diagnostics about nonconformant code from errors to
2030 warnings. Thus, using @option{-fpermissive} will allow some
2031 nonconforming code to compile.
2033 @item -fno-pretty-templates
2034 @opindex fno-pretty-templates
2035 When an error message refers to a specialization of a function
2036 template, the compiler will normally print the signature of the
2037 template followed by the template arguments and any typedefs or
2038 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2039 rather than @code{void f(int)}) so that it's clear which template is
2040 involved. When an error message refers to a specialization of a class
2041 template, the compiler will omit any template arguments which match
2042 the default template arguments for that template. If either of these
2043 behaviors make it harder to understand the error message rather than
2044 easier, using @option{-fno-pretty-templates} will disable them.
2048 Enable automatic template instantiation at link time. This option also
2049 implies @option{-fno-implicit-templates}. @xref{Template
2050 Instantiation}, for more information.
2054 Disable generation of information about every class with virtual
2055 functions for use by the C++ runtime type identification features
2056 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2057 of the language, you can save some space by using this flag. Note that
2058 exception handling uses the same information, but it will generate it as
2059 needed. The @samp{dynamic_cast} operator can still be used for casts that
2060 do not require runtime type information, i.e.@: casts to @code{void *} or to
2061 unambiguous base classes.
2065 Emit statistics about front-end processing at the end of the compilation.
2066 This information is generally only useful to the G++ development team.
2068 @item -fstrict-enums
2069 @opindex fstrict-enums
2070 Allow the compiler to optimize using the assumption that a value of
2071 enumeration type can only be one of the values of the enumeration (as
2072 defined in the C++ standard; basically, a value which can be
2073 represented in the minimum number of bits needed to represent all the
2074 enumerators). This assumption may not be valid if the program uses a
2075 cast to convert an arbitrary integer value to the enumeration type.
2077 @item -ftemplate-depth=@var{n}
2078 @opindex ftemplate-depth
2079 Set the maximum instantiation depth for template classes to @var{n}.
2080 A limit on the template instantiation depth is needed to detect
2081 endless recursions during template class instantiation. ANSI/ISO C++
2082 conforming programs must not rely on a maximum depth greater than 17
2083 (changed to 1024 in C++0x). The default value is 900, as the compiler
2084 can run out of stack space before hitting 1024 in some situations.
2086 @item -fno-threadsafe-statics
2087 @opindex fno-threadsafe-statics
2088 Do not emit the extra code to use the routines specified in the C++
2089 ABI for thread-safe initialization of local statics. You can use this
2090 option to reduce code size slightly in code that doesn't need to be
2093 @item -fuse-cxa-atexit
2094 @opindex fuse-cxa-atexit
2095 Register destructors for objects with static storage duration with the
2096 @code{__cxa_atexit} function rather than the @code{atexit} function.
2097 This option is required for fully standards-compliant handling of static
2098 destructors, but will only work if your C library supports
2099 @code{__cxa_atexit}.
2101 @item -fno-use-cxa-get-exception-ptr
2102 @opindex fno-use-cxa-get-exception-ptr
2103 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2104 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2105 if the runtime routine is not available.
2107 @item -fvisibility-inlines-hidden
2108 @opindex fvisibility-inlines-hidden
2109 This switch declares that the user does not attempt to compare
2110 pointers to inline methods where the addresses of the two functions
2111 were taken in different shared objects.
2113 The effect of this is that GCC may, effectively, mark inline methods with
2114 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2115 appear in the export table of a DSO and do not require a PLT indirection
2116 when used within the DSO@. Enabling this option can have a dramatic effect
2117 on load and link times of a DSO as it massively reduces the size of the
2118 dynamic export table when the library makes heavy use of templates.
2120 The behavior of this switch is not quite the same as marking the
2121 methods as hidden directly, because it does not affect static variables
2122 local to the function or cause the compiler to deduce that
2123 the function is defined in only one shared object.
2125 You may mark a method as having a visibility explicitly to negate the
2126 effect of the switch for that method. For example, if you do want to
2127 compare pointers to a particular inline method, you might mark it as
2128 having default visibility. Marking the enclosing class with explicit
2129 visibility will have no effect.
2131 Explicitly instantiated inline methods are unaffected by this option
2132 as their linkage might otherwise cross a shared library boundary.
2133 @xref{Template Instantiation}.
2135 @item -fvisibility-ms-compat
2136 @opindex fvisibility-ms-compat
2137 This flag attempts to use visibility settings to make GCC's C++
2138 linkage model compatible with that of Microsoft Visual Studio.
2140 The flag makes these changes to GCC's linkage model:
2144 It sets the default visibility to @code{hidden}, like
2145 @option{-fvisibility=hidden}.
2148 Types, but not their members, are not hidden by default.
2151 The One Definition Rule is relaxed for types without explicit
2152 visibility specifications which are defined in more than one different
2153 shared object: those declarations are permitted if they would have
2154 been permitted when this option was not used.
2157 In new code it is better to use @option{-fvisibility=hidden} and
2158 export those classes which are intended to be externally visible.
2159 Unfortunately it is possible for code to rely, perhaps accidentally,
2160 on the Visual Studio behavior.
2162 Among the consequences of these changes are that static data members
2163 of the same type with the same name but defined in different shared
2164 objects will be different, so changing one will not change the other;
2165 and that pointers to function members defined in different shared
2166 objects may not compare equal. When this flag is given, it is a
2167 violation of the ODR to define types with the same name differently.
2171 Do not use weak symbol support, even if it is provided by the linker.
2172 By default, G++ will use weak symbols if they are available. This
2173 option exists only for testing, and should not be used by end-users;
2174 it will result in inferior code and has no benefits. This option may
2175 be removed in a future release of G++.
2179 Do not search for header files in the standard directories specific to
2180 C++, but do still search the other standard directories. (This option
2181 is used when building the C++ library.)
2184 In addition, these optimization, warning, and code generation options
2185 have meanings only for C++ programs:
2188 @item -fno-default-inline
2189 @opindex fno-default-inline
2190 Do not assume @samp{inline} for functions defined inside a class scope.
2191 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2192 functions will have linkage like inline functions; they just won't be
2195 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2198 Warn when G++ generates code that is probably not compatible with the
2199 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2200 all such cases, there are probably some cases that are not warned about,
2201 even though G++ is generating incompatible code. There may also be
2202 cases where warnings are emitted even though the code that is generated
2205 You should rewrite your code to avoid these warnings if you are
2206 concerned about the fact that code generated by G++ may not be binary
2207 compatible with code generated by other compilers.
2209 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2214 A template with a non-type template parameter of reference type is
2215 mangled incorrectly:
2218 template <int &> struct S @{@};
2222 This is fixed in @option{-fabi-version=3}.
2225 SIMD vector types declared using @code{__attribute ((vector_size))} are
2226 mangled in a non-standard way that does not allow for overloading of
2227 functions taking vectors of different sizes.
2229 The mangling is changed in @option{-fabi-version=4}.
2232 The known incompatibilities in @option{-fabi-version=1} include:
2237 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2238 pack data into the same byte as a base class. For example:
2241 struct A @{ virtual void f(); int f1 : 1; @};
2242 struct B : public A @{ int f2 : 1; @};
2246 In this case, G++ will place @code{B::f2} into the same byte
2247 as@code{A::f1}; other compilers will not. You can avoid this problem
2248 by explicitly padding @code{A} so that its size is a multiple of the
2249 byte size on your platform; that will cause G++ and other compilers to
2250 layout @code{B} identically.
2253 Incorrect handling of tail-padding for virtual bases. G++ does not use
2254 tail padding when laying out virtual bases. For example:
2257 struct A @{ virtual void f(); char c1; @};
2258 struct B @{ B(); char c2; @};
2259 struct C : public A, public virtual B @{@};
2263 In this case, G++ will not place @code{B} into the tail-padding for
2264 @code{A}; other compilers will. You can avoid this problem by
2265 explicitly padding @code{A} so that its size is a multiple of its
2266 alignment (ignoring virtual base classes); that will cause G++ and other
2267 compilers to layout @code{C} identically.
2270 Incorrect handling of bit-fields with declared widths greater than that
2271 of their underlying types, when the bit-fields appear in a union. For
2275 union U @{ int i : 4096; @};
2279 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2280 union too small by the number of bits in an @code{int}.
2283 Empty classes can be placed at incorrect offsets. For example:
2293 struct C : public B, public A @{@};
2297 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2298 it should be placed at offset zero. G++ mistakenly believes that the
2299 @code{A} data member of @code{B} is already at offset zero.
2302 Names of template functions whose types involve @code{typename} or
2303 template template parameters can be mangled incorrectly.
2306 template <typename Q>
2307 void f(typename Q::X) @{@}
2309 template <template <typename> class Q>
2310 void f(typename Q<int>::X) @{@}
2314 Instantiations of these templates may be mangled incorrectly.
2318 It also warns psABI related changes. The known psABI changes at this
2324 For SYSV/x86-64, when passing union with long double, it is changed to
2325 pass in memory as specified in psABI. For example:
2335 @code{union U} will always be passed in memory.
2339 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2340 @opindex Wctor-dtor-privacy
2341 @opindex Wno-ctor-dtor-privacy
2342 Warn when a class seems unusable because all the constructors or
2343 destructors in that class are private, and it has neither friends nor
2344 public static member functions.
2346 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2347 @opindex Wdelete-non-virtual-dtor
2348 @opindex Wno-delete-non-virtual-dtor
2349 Warn when @samp{delete} is used to destroy an instance of a class which
2350 has virtual functions and non-virtual destructor. It is unsafe to delete
2351 an instance of a derived class through a pointer to a base class if the
2352 base class does not have a virtual destructor. This warning is enabled
2355 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2357 @opindex Wno-narrowing
2358 With -std=c++0x, suppress the diagnostic required by the standard for
2359 narrowing conversions within @samp{@{ @}}, e.g.
2362 int i = @{ 2.2 @}; // error: narrowing from double to int
2365 This flag can be useful for compiling valid C++98 code in C++0x mode.
2367 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2369 @opindex Wno-noexcept
2370 Warn when a noexcept-expression evaluates to false because of a call
2371 to a function that does not have a non-throwing exception
2372 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2373 the compiler to never throw an exception.
2375 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2376 @opindex Wnon-virtual-dtor
2377 @opindex Wno-non-virtual-dtor
2378 Warn when a class has virtual functions and accessible non-virtual
2379 destructor, in which case it would be possible but unsafe to delete
2380 an instance of a derived class through a pointer to the base class.
2381 This warning is also enabled if -Weffc++ is specified.
2383 @item -Wreorder @r{(C++ and Objective-C++ only)}
2385 @opindex Wno-reorder
2386 @cindex reordering, warning
2387 @cindex warning for reordering of member initializers
2388 Warn when the order of member initializers given in the code does not
2389 match the order in which they must be executed. For instance:
2395 A(): j (0), i (1) @{ @}
2399 The compiler will rearrange the member initializers for @samp{i}
2400 and @samp{j} to match the declaration order of the members, emitting
2401 a warning to that effect. This warning is enabled by @option{-Wall}.
2404 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2407 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2410 Warn about violations of the following style guidelines from Scott Meyers'
2411 @cite{Effective C++} book:
2415 Item 11: Define a copy constructor and an assignment operator for classes
2416 with dynamically allocated memory.
2419 Item 12: Prefer initialization to assignment in constructors.
2422 Item 14: Make destructors virtual in base classes.
2425 Item 15: Have @code{operator=} return a reference to @code{*this}.
2428 Item 23: Don't try to return a reference when you must return an object.
2432 Also warn about violations of the following style guidelines from
2433 Scott Meyers' @cite{More Effective C++} book:
2437 Item 6: Distinguish between prefix and postfix forms of increment and
2438 decrement operators.
2441 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2445 When selecting this option, be aware that the standard library
2446 headers do not obey all of these guidelines; use @samp{grep -v}
2447 to filter out those warnings.
2449 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2450 @opindex Wstrict-null-sentinel
2451 @opindex Wno-strict-null-sentinel
2452 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2453 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2454 to @code{__null}. Although it is a null pointer constant not a null pointer,
2455 it is guaranteed to be of the same size as a pointer. But this use is
2456 not portable across different compilers.
2458 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2459 @opindex Wno-non-template-friend
2460 @opindex Wnon-template-friend
2461 Disable warnings when non-templatized friend functions are declared
2462 within a template. Since the advent of explicit template specification
2463 support in G++, if the name of the friend is an unqualified-id (i.e.,
2464 @samp{friend foo(int)}), the C++ language specification demands that the
2465 friend declare or define an ordinary, nontemplate function. (Section
2466 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2467 could be interpreted as a particular specialization of a templatized
2468 function. Because this non-conforming behavior is no longer the default
2469 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2470 check existing code for potential trouble spots and is on by default.
2471 This new compiler behavior can be turned off with
2472 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2473 but disables the helpful warning.
2475 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2476 @opindex Wold-style-cast
2477 @opindex Wno-old-style-cast
2478 Warn if an old-style (C-style) cast to a non-void type is used within
2479 a C++ program. The new-style casts (@samp{dynamic_cast},
2480 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2481 less vulnerable to unintended effects and much easier to search for.
2483 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2484 @opindex Woverloaded-virtual
2485 @opindex Wno-overloaded-virtual
2486 @cindex overloaded virtual function, warning
2487 @cindex warning for overloaded virtual function
2488 Warn when a function declaration hides virtual functions from a
2489 base class. For example, in:
2496 struct B: public A @{
2501 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2509 will fail to compile.
2511 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2512 @opindex Wno-pmf-conversions
2513 @opindex Wpmf-conversions
2514 Disable the diagnostic for converting a bound pointer to member function
2517 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2518 @opindex Wsign-promo
2519 @opindex Wno-sign-promo
2520 Warn when overload resolution chooses a promotion from unsigned or
2521 enumerated type to a signed type, over a conversion to an unsigned type of
2522 the same size. Previous versions of G++ would try to preserve
2523 unsignedness, but the standard mandates the current behavior.
2528 A& operator = (int);
2538 In this example, G++ will synthesize a default @samp{A& operator =
2539 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2542 @node Objective-C and Objective-C++ Dialect Options
2543 @section Options Controlling Objective-C and Objective-C++ Dialects
2545 @cindex compiler options, Objective-C and Objective-C++
2546 @cindex Objective-C and Objective-C++ options, command line
2547 @cindex options, Objective-C and Objective-C++
2548 (NOTE: This manual does not describe the Objective-C and Objective-C++
2549 languages themselves. @xref{Standards,,Language Standards
2550 Supported by GCC}, for references.)
2552 This section describes the command-line options that are only meaningful
2553 for Objective-C and Objective-C++ programs, but you can also use most of
2554 the language-independent GNU compiler options.
2555 For example, you might compile a file @code{some_class.m} like this:
2558 gcc -g -fgnu-runtime -O -c some_class.m
2562 In this example, @option{-fgnu-runtime} is an option meant only for
2563 Objective-C and Objective-C++ programs; you can use the other options with
2564 any language supported by GCC@.
2566 Note that since Objective-C is an extension of the C language, Objective-C
2567 compilations may also use options specific to the C front-end (e.g.,
2568 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2569 C++-specific options (e.g., @option{-Wabi}).
2571 Here is a list of options that are @emph{only} for compiling Objective-C
2572 and Objective-C++ programs:
2575 @item -fconstant-string-class=@var{class-name}
2576 @opindex fconstant-string-class
2577 Use @var{class-name} as the name of the class to instantiate for each
2578 literal string specified with the syntax @code{@@"@dots{}"}. The default
2579 class name is @code{NXConstantString} if the GNU runtime is being used, and
2580 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2581 @option{-fconstant-cfstrings} option, if also present, will override the
2582 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2583 to be laid out as constant CoreFoundation strings.
2586 @opindex fgnu-runtime
2587 Generate object code compatible with the standard GNU Objective-C
2588 runtime. This is the default for most types of systems.
2590 @item -fnext-runtime
2591 @opindex fnext-runtime
2592 Generate output compatible with the NeXT runtime. This is the default
2593 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2594 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2597 @item -fno-nil-receivers
2598 @opindex fno-nil-receivers
2599 Assume that all Objective-C message dispatches (@code{[receiver
2600 message:arg]}) in this translation unit ensure that the receiver is
2601 not @code{nil}. This allows for more efficient entry points in the
2602 runtime to be used. This option is only available in conjunction with
2603 the NeXT runtime and ABI version 0 or 1.
2605 @item -fobjc-abi-version=@var{n}
2606 @opindex fobjc-abi-version
2607 Use version @var{n} of the Objective-C ABI for the selected runtime.
2608 This option is currently supported only for the NeXT runtime. In that
2609 case, Version 0 is the traditional (32-bit) ABI without support for
2610 properties and other Objective-C 2.0 additions. Version 1 is the
2611 traditional (32-bit) ABI with support for properties and other
2612 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2613 nothing is specified, the default is Version 0 on 32-bit target
2614 machines, and Version 2 on 64-bit target machines.
2616 @item -fobjc-call-cxx-cdtors
2617 @opindex fobjc-call-cxx-cdtors
2618 For each Objective-C class, check if any of its instance variables is a
2619 C++ object with a non-trivial default constructor. If so, synthesize a
2620 special @code{- (id) .cxx_construct} instance method that will run
2621 non-trivial default constructors on any such instance variables, in order,
2622 and then return @code{self}. Similarly, check if any instance variable
2623 is a C++ object with a non-trivial destructor, and if so, synthesize a
2624 special @code{- (void) .cxx_destruct} method that will run
2625 all such default destructors, in reverse order.
2627 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2628 methods thusly generated will only operate on instance variables
2629 declared in the current Objective-C class, and not those inherited
2630 from superclasses. It is the responsibility of the Objective-C
2631 runtime to invoke all such methods in an object's inheritance
2632 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2633 by the runtime immediately after a new object instance is allocated;
2634 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2635 before the runtime deallocates an object instance.
2637 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2638 support for invoking the @code{- (id) .cxx_construct} and
2639 @code{- (void) .cxx_destruct} methods.
2641 @item -fobjc-direct-dispatch
2642 @opindex fobjc-direct-dispatch
2643 Allow fast jumps to the message dispatcher. On Darwin this is
2644 accomplished via the comm page.
2646 @item -fobjc-exceptions
2647 @opindex fobjc-exceptions
2648 Enable syntactic support for structured exception handling in
2649 Objective-C, similar to what is offered by C++ and Java. This option
2650 is required to use the Objective-C keywords @code{@@try},
2651 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2652 @code{@@synchronized}. This option is available with both the GNU
2653 runtime and the NeXT runtime (but not available in conjunction with
2654 the NeXT runtime on Mac OS X 10.2 and earlier).
2658 Enable garbage collection (GC) in Objective-C and Objective-C++
2659 programs. This option is only available with the NeXT runtime; the
2660 GNU runtime has a different garbage collection implementation that
2661 does not require special compiler flags.
2663 @item -fobjc-nilcheck
2664 @opindex fobjc-nilcheck
2665 For the NeXT runtime with version 2 of the ABI, check for a nil
2666 receiver in method invocations before doing the actual method call.
2667 This is the default and can be disabled using
2668 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2669 checked for nil in this way no matter what this flag is set to.
2670 Currently this flag does nothing when the GNU runtime, or an older
2671 version of the NeXT runtime ABI, is used.
2673 @item -fobjc-std=objc1
2675 Conform to the language syntax of Objective-C 1.0, the language
2676 recognized by GCC 4.0. This only affects the Objective-C additions to
2677 the C/C++ language; it does not affect conformance to C/C++ standards,
2678 which is controlled by the separate C/C++ dialect option flags. When
2679 this option is used with the Objective-C or Objective-C++ compiler,
2680 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2681 This is useful if you need to make sure that your Objective-C code can
2682 be compiled with older versions of GCC.
2684 @item -freplace-objc-classes
2685 @opindex freplace-objc-classes
2686 Emit a special marker instructing @command{ld(1)} not to statically link in
2687 the resulting object file, and allow @command{dyld(1)} to load it in at
2688 run time instead. This is used in conjunction with the Fix-and-Continue
2689 debugging mode, where the object file in question may be recompiled and
2690 dynamically reloaded in the course of program execution, without the need
2691 to restart the program itself. Currently, Fix-and-Continue functionality
2692 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2697 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2698 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2699 compile time) with static class references that get initialized at load time,
2700 which improves run-time performance. Specifying the @option{-fzero-link} flag
2701 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2702 to be retained. This is useful in Zero-Link debugging mode, since it allows
2703 for individual class implementations to be modified during program execution.
2704 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2705 regardless of command line options.
2709 Dump interface declarations for all classes seen in the source file to a
2710 file named @file{@var{sourcename}.decl}.
2712 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2713 @opindex Wassign-intercept
2714 @opindex Wno-assign-intercept
2715 Warn whenever an Objective-C assignment is being intercepted by the
2718 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2719 @opindex Wno-protocol
2721 If a class is declared to implement a protocol, a warning is issued for
2722 every method in the protocol that is not implemented by the class. The
2723 default behavior is to issue a warning for every method not explicitly
2724 implemented in the class, even if a method implementation is inherited
2725 from the superclass. If you use the @option{-Wno-protocol} option, then
2726 methods inherited from the superclass are considered to be implemented,
2727 and no warning is issued for them.
2729 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2731 @opindex Wno-selector
2732 Warn if multiple methods of different types for the same selector are
2733 found during compilation. The check is performed on the list of methods
2734 in the final stage of compilation. Additionally, a check is performed
2735 for each selector appearing in a @code{@@selector(@dots{})}
2736 expression, and a corresponding method for that selector has been found
2737 during compilation. Because these checks scan the method table only at
2738 the end of compilation, these warnings are not produced if the final
2739 stage of compilation is not reached, for example because an error is
2740 found during compilation, or because the @option{-fsyntax-only} option is
2743 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2744 @opindex Wstrict-selector-match
2745 @opindex Wno-strict-selector-match
2746 Warn if multiple methods with differing argument and/or return types are
2747 found for a given selector when attempting to send a message using this
2748 selector to a receiver of type @code{id} or @code{Class}. When this flag
2749 is off (which is the default behavior), the compiler will omit such warnings
2750 if any differences found are confined to types which share the same size
2753 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2754 @opindex Wundeclared-selector
2755 @opindex Wno-undeclared-selector
2756 Warn if a @code{@@selector(@dots{})} expression referring to an
2757 undeclared selector is found. A selector is considered undeclared if no
2758 method with that name has been declared before the
2759 @code{@@selector(@dots{})} expression, either explicitly in an
2760 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2761 an @code{@@implementation} section. This option always performs its
2762 checks as soon as a @code{@@selector(@dots{})} expression is found,
2763 while @option{-Wselector} only performs its checks in the final stage of
2764 compilation. This also enforces the coding style convention
2765 that methods and selectors must be declared before being used.
2767 @item -print-objc-runtime-info
2768 @opindex print-objc-runtime-info
2769 Generate C header describing the largest structure that is passed by
2774 @node Language Independent Options
2775 @section Options to Control Diagnostic Messages Formatting
2776 @cindex options to control diagnostics formatting
2777 @cindex diagnostic messages
2778 @cindex message formatting
2780 Traditionally, diagnostic messages have been formatted irrespective of
2781 the output device's aspect (e.g.@: its width, @dots{}). The options described
2782 below can be used to control the diagnostic messages formatting
2783 algorithm, e.g.@: how many characters per line, how often source location
2784 information should be reported. Right now, only the C++ front end can
2785 honor these options. However it is expected, in the near future, that
2786 the remaining front ends would be able to digest them correctly.
2789 @item -fmessage-length=@var{n}
2790 @opindex fmessage-length
2791 Try to format error messages so that they fit on lines of about @var{n}
2792 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2793 the front ends supported by GCC@. If @var{n} is zero, then no
2794 line-wrapping will be done; each error message will appear on a single
2797 @opindex fdiagnostics-show-location
2798 @item -fdiagnostics-show-location=once
2799 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2800 reporter to emit @emph{once} source location information; that is, in
2801 case the message is too long to fit on a single physical line and has to
2802 be wrapped, the source location won't be emitted (as prefix) again,
2803 over and over, in subsequent continuation lines. This is the default
2806 @item -fdiagnostics-show-location=every-line
2807 Only meaningful in line-wrapping mode. Instructs the diagnostic
2808 messages reporter to emit the same source location information (as
2809 prefix) for physical lines that result from the process of breaking
2810 a message which is too long to fit on a single line.
2812 @item -fno-diagnostics-show-option
2813 @opindex fno-diagnostics-show-option
2814 @opindex fdiagnostics-show-option
2815 By default, each diagnostic emitted includes text which indicates the
2816 command line option that directly controls the diagnostic (if such an
2817 option is known to the diagnostic machinery). Specifying the
2818 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2820 @item -Wcoverage-mismatch
2821 @opindex Wcoverage-mismatch
2822 Warn if feedback profiles do not match when using the
2823 @option{-fprofile-use} option.
2824 If a source file was changed between @option{-fprofile-gen} and
2825 @option{-fprofile-use}, the files with the profile feedback can fail
2826 to match the source file and GCC can not use the profile feedback
2827 information. By default, this warning is enabled and is treated as an
2828 error. @option{-Wno-coverage-mismatch} can be used to disable the
2829 warning or @option{-Wno-error=coverage-mismatch} can be used to
2830 disable the error. Disable the error for this warning can result in
2831 poorly optimized code, so disabling the error is useful only in the
2832 case of very minor changes such as bug fixes to an existing code-base.
2833 Completely disabling the warning is not recommended.
2837 @node Warning Options
2838 @section Options to Request or Suppress Warnings
2839 @cindex options to control warnings
2840 @cindex warning messages
2841 @cindex messages, warning
2842 @cindex suppressing warnings
2844 Warnings are diagnostic messages that report constructions which
2845 are not inherently erroneous but which are risky or suggest there
2846 may have been an error.
2848 The following language-independent options do not enable specific
2849 warnings but control the kinds of diagnostics produced by GCC.
2852 @cindex syntax checking
2854 @opindex fsyntax-only
2855 Check the code for syntax errors, but don't do anything beyond that.
2857 @item -fmax-errors=@var{n}
2858 @opindex fmax-errors
2859 Limits the maximum number of error messages to @var{n}, at which point
2860 GCC bails out rather than attempting to continue processing the source
2861 code. If @var{n} is 0 (the default), there is no limit on the number
2862 of error messages produced. If @option{-Wfatal-errors} is also
2863 specified, then @option{-Wfatal-errors} takes precedence over this
2868 Inhibit all warning messages.
2873 Make all warnings into errors.
2878 Make the specified warning into an error. The specifier for a warning
2879 is appended, for example @option{-Werror=switch} turns the warnings
2880 controlled by @option{-Wswitch} into errors. This switch takes a
2881 negative form, to be used to negate @option{-Werror} for specific
2882 warnings, for example @option{-Wno-error=switch} makes
2883 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2886 The warning message for each controllable warning includes the
2887 option which controls the warning. That option can then be used with
2888 @option{-Werror=} and @option{-Wno-error=} as described above.
2889 (Printing of the option in the warning message can be disabled using the
2890 @option{-fno-diagnostics-show-option} flag.)
2892 Note that specifying @option{-Werror=}@var{foo} automatically implies
2893 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2896 @item -Wfatal-errors
2897 @opindex Wfatal-errors
2898 @opindex Wno-fatal-errors
2899 This option causes the compiler to abort compilation on the first error
2900 occurred rather than trying to keep going and printing further error
2905 You can request many specific warnings with options beginning
2906 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2907 implicit declarations. Each of these specific warning options also
2908 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2909 example, @option{-Wno-implicit}. This manual lists only one of the
2910 two forms, whichever is not the default. For further,
2911 language-specific options also refer to @ref{C++ Dialect Options} and
2912 @ref{Objective-C and Objective-C++ Dialect Options}.
2914 When an unrecognized warning option is requested (e.g.,
2915 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2916 that the option is not recognized. However, if the @option{-Wno-} form
2917 is used, the behavior is slightly different: No diagnostic will be
2918 produced for @option{-Wno-unknown-warning} unless other diagnostics
2919 are being produced. This allows the use of new @option{-Wno-} options
2920 with old compilers, but if something goes wrong, the compiler will
2921 warn that an unrecognized option was used.
2926 Issue all the warnings demanded by strict ISO C and ISO C++;
2927 reject all programs that use forbidden extensions, and some other
2928 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2929 version of the ISO C standard specified by any @option{-std} option used.
2931 Valid ISO C and ISO C++ programs should compile properly with or without
2932 this option (though a rare few will require @option{-ansi} or a
2933 @option{-std} option specifying the required version of ISO C)@. However,
2934 without this option, certain GNU extensions and traditional C and C++
2935 features are supported as well. With this option, they are rejected.
2937 @option{-pedantic} does not cause warning messages for use of the
2938 alternate keywords whose names begin and end with @samp{__}. Pedantic
2939 warnings are also disabled in the expression that follows
2940 @code{__extension__}. However, only system header files should use
2941 these escape routes; application programs should avoid them.
2942 @xref{Alternate Keywords}.
2944 Some users try to use @option{-pedantic} to check programs for strict ISO
2945 C conformance. They soon find that it does not do quite what they want:
2946 it finds some non-ISO practices, but not all---only those for which
2947 ISO C @emph{requires} a diagnostic, and some others for which
2948 diagnostics have been added.
2950 A feature to report any failure to conform to ISO C might be useful in
2951 some instances, but would require considerable additional work and would
2952 be quite different from @option{-pedantic}. We don't have plans to
2953 support such a feature in the near future.
2955 Where the standard specified with @option{-std} represents a GNU
2956 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2957 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2958 extended dialect is based. Warnings from @option{-pedantic} are given
2959 where they are required by the base standard. (It would not make sense
2960 for such warnings to be given only for features not in the specified GNU
2961 C dialect, since by definition the GNU dialects of C include all
2962 features the compiler supports with the given option, and there would be
2963 nothing to warn about.)
2965 @item -pedantic-errors
2966 @opindex pedantic-errors
2967 Like @option{-pedantic}, except that errors are produced rather than
2973 This enables all the warnings about constructions that some users
2974 consider questionable, and that are easy to avoid (or modify to
2975 prevent the warning), even in conjunction with macros. This also
2976 enables some language-specific warnings described in @ref{C++ Dialect
2977 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2979 @option{-Wall} turns on the following warning flags:
2981 @gccoptlist{-Waddress @gol
2982 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2984 -Wchar-subscripts @gol
2985 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2986 -Wimplicit-int @r{(C and Objective-C only)} @gol
2987 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2990 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2991 -Wmaybe-uninitialized @gol
2992 -Wmissing-braces @gol
2998 -Wsequence-point @gol
2999 -Wsign-compare @r{(only in C++)} @gol
3000 -Wstrict-aliasing @gol
3001 -Wstrict-overflow=1 @gol
3004 -Wuninitialized @gol
3005 -Wunknown-pragmas @gol
3006 -Wunused-function @gol
3009 -Wunused-variable @gol
3010 -Wvolatile-register-var @gol
3013 Note that some warning flags are not implied by @option{-Wall}. Some of
3014 them warn about constructions that users generally do not consider
3015 questionable, but which occasionally you might wish to check for;
3016 others warn about constructions that are necessary or hard to avoid in
3017 some cases, and there is no simple way to modify the code to suppress
3018 the warning. Some of them are enabled by @option{-Wextra} but many of
3019 them must be enabled individually.
3025 This enables some extra warning flags that are not enabled by
3026 @option{-Wall}. (This option used to be called @option{-W}. The older
3027 name is still supported, but the newer name is more descriptive.)
3029 @gccoptlist{-Wclobbered @gol
3031 -Wignored-qualifiers @gol
3032 -Wmissing-field-initializers @gol
3033 -Wmissing-parameter-type @r{(C only)} @gol
3034 -Wold-style-declaration @r{(C only)} @gol
3035 -Woverride-init @gol
3038 -Wuninitialized @gol
3039 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3040 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3043 The option @option{-Wextra} also prints warning messages for the
3049 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3050 @samp{>}, or @samp{>=}.
3053 (C++ only) An enumerator and a non-enumerator both appear in a
3054 conditional expression.
3057 (C++ only) Ambiguous virtual bases.
3060 (C++ only) Subscripting an array which has been declared @samp{register}.
3063 (C++ only) Taking the address of a variable which has been declared
3067 (C++ only) A base class is not initialized in a derived class' copy
3072 @item -Wchar-subscripts
3073 @opindex Wchar-subscripts
3074 @opindex Wno-char-subscripts
3075 Warn if an array subscript has type @code{char}. This is a common cause
3076 of error, as programmers often forget that this type is signed on some
3078 This warning is enabled by @option{-Wall}.
3082 @opindex Wno-comment
3083 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3084 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3085 This warning is enabled by @option{-Wall}.
3088 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3090 Suppress warning messages emitted by @code{#warning} directives.
3092 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3093 @opindex Wdouble-promotion
3094 @opindex Wno-double-promotion
3095 Give a warning when a value of type @code{float} is implicitly
3096 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3097 floating-point unit implement @code{float} in hardware, but emulate
3098 @code{double} in software. On such a machine, doing computations
3099 using @code{double} values is much more expensive because of the
3100 overhead required for software emulation.
3102 It is easy to accidentally do computations with @code{double} because
3103 floating-point literals are implicitly of type @code{double}. For
3107 float area(float radius)
3109 return 3.14159 * radius * radius;
3113 the compiler will perform the entire computation with @code{double}
3114 because the floating-point literal is a @code{double}.
3119 @opindex ffreestanding
3120 @opindex fno-builtin
3121 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3122 the arguments supplied have types appropriate to the format string
3123 specified, and that the conversions specified in the format string make
3124 sense. This includes standard functions, and others specified by format
3125 attributes (@pxref{Function Attributes}), in the @code{printf},
3126 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3127 not in the C standard) families (or other target-specific families).
3128 Which functions are checked without format attributes having been
3129 specified depends on the standard version selected, and such checks of
3130 functions without the attribute specified are disabled by
3131 @option{-ffreestanding} or @option{-fno-builtin}.
3133 The formats are checked against the format features supported by GNU
3134 libc version 2.2. These include all ISO C90 and C99 features, as well
3135 as features from the Single Unix Specification and some BSD and GNU
3136 extensions. Other library implementations may not support all these
3137 features; GCC does not support warning about features that go beyond a
3138 particular library's limitations. However, if @option{-pedantic} is used
3139 with @option{-Wformat}, warnings will be given about format features not
3140 in the selected standard version (but not for @code{strfmon} formats,
3141 since those are not in any version of the C standard). @xref{C Dialect
3142 Options,,Options Controlling C Dialect}.
3144 Since @option{-Wformat} also checks for null format arguments for
3145 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3147 @option{-Wformat} is included in @option{-Wall}. For more control over some
3148 aspects of format checking, the options @option{-Wformat-y2k},
3149 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3150 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3151 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3154 @opindex Wformat-y2k
3155 @opindex Wno-format-y2k
3156 If @option{-Wformat} is specified, also warn about @code{strftime}
3157 formats which may yield only a two-digit year.
3159 @item -Wno-format-contains-nul
3160 @opindex Wno-format-contains-nul
3161 @opindex Wformat-contains-nul
3162 If @option{-Wformat} is specified, do not warn about format strings that
3165 @item -Wno-format-extra-args
3166 @opindex Wno-format-extra-args
3167 @opindex Wformat-extra-args
3168 If @option{-Wformat} is specified, do not warn about excess arguments to a
3169 @code{printf} or @code{scanf} format function. The C standard specifies
3170 that such arguments are ignored.
3172 Where the unused arguments lie between used arguments that are
3173 specified with @samp{$} operand number specifications, normally
3174 warnings are still given, since the implementation could not know what
3175 type to pass to @code{va_arg} to skip the unused arguments. However,
3176 in the case of @code{scanf} formats, this option will suppress the
3177 warning if the unused arguments are all pointers, since the Single
3178 Unix Specification says that such unused arguments are allowed.
3180 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3181 @opindex Wno-format-zero-length
3182 @opindex Wformat-zero-length
3183 If @option{-Wformat} is specified, do not warn about zero-length formats.
3184 The C standard specifies that zero-length formats are allowed.
3186 @item -Wformat-nonliteral
3187 @opindex Wformat-nonliteral
3188 @opindex Wno-format-nonliteral
3189 If @option{-Wformat} is specified, also warn if the format string is not a
3190 string literal and so cannot be checked, unless the format function
3191 takes its format arguments as a @code{va_list}.
3193 @item -Wformat-security
3194 @opindex Wformat-security
3195 @opindex Wno-format-security
3196 If @option{-Wformat} is specified, also warn about uses of format
3197 functions that represent possible security problems. At present, this
3198 warns about calls to @code{printf} and @code{scanf} functions where the
3199 format string is not a string literal and there are no format arguments,
3200 as in @code{printf (foo);}. This may be a security hole if the format
3201 string came from untrusted input and contains @samp{%n}. (This is
3202 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3203 in future warnings may be added to @option{-Wformat-security} that are not
3204 included in @option{-Wformat-nonliteral}.)
3208 @opindex Wno-format=2
3209 Enable @option{-Wformat} plus format checks not included in
3210 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3211 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3213 @item -Wnonnull @r{(C and Objective-C only)}
3215 @opindex Wno-nonnull
3216 Warn about passing a null pointer for arguments marked as
3217 requiring a non-null value by the @code{nonnull} function attribute.
3219 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3220 can be disabled with the @option{-Wno-nonnull} option.
3222 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3224 @opindex Wno-init-self
3225 Warn about uninitialized variables which are initialized with themselves.
3226 Note this option can only be used with the @option{-Wuninitialized} option.
3228 For example, GCC will warn about @code{i} being uninitialized in the
3229 following snippet only when @option{-Winit-self} has been specified:
3240 @item -Wimplicit-int @r{(C and Objective-C only)}
3241 @opindex Wimplicit-int
3242 @opindex Wno-implicit-int
3243 Warn when a declaration does not specify a type.
3244 This warning is enabled by @option{-Wall}.
3246 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3247 @opindex Wimplicit-function-declaration
3248 @opindex Wno-implicit-function-declaration
3249 Give a warning whenever a function is used before being declared. In
3250 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3251 enabled by default and it is made into an error by
3252 @option{-pedantic-errors}. This warning is also enabled by
3255 @item -Wimplicit @r{(C and Objective-C only)}
3257 @opindex Wno-implicit
3258 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3259 This warning is enabled by @option{-Wall}.
3261 @item -Wignored-qualifiers @r{(C and C++ only)}
3262 @opindex Wignored-qualifiers
3263 @opindex Wno-ignored-qualifiers
3264 Warn if the return type of a function has a type qualifier
3265 such as @code{const}. For ISO C such a type qualifier has no effect,
3266 since the value returned by a function is not an lvalue.
3267 For C++, the warning is only emitted for scalar types or @code{void}.
3268 ISO C prohibits qualified @code{void} return types on function
3269 definitions, so such return types always receive a warning
3270 even without this option.
3272 This warning is also enabled by @option{-Wextra}.
3277 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3278 a function with external linkage, returning int, taking either zero
3279 arguments, two, or three arguments of appropriate types. This warning
3280 is enabled by default in C++ and is enabled by either @option{-Wall}
3281 or @option{-pedantic}.
3283 @item -Wmissing-braces
3284 @opindex Wmissing-braces
3285 @opindex Wno-missing-braces
3286 Warn if an aggregate or union initializer is not fully bracketed. In
3287 the following example, the initializer for @samp{a} is not fully
3288 bracketed, but that for @samp{b} is fully bracketed.
3291 int a[2][2] = @{ 0, 1, 2, 3 @};
3292 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3295 This warning is enabled by @option{-Wall}.
3297 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3298 @opindex Wmissing-include-dirs
3299 @opindex Wno-missing-include-dirs
3300 Warn if a user-supplied include directory does not exist.
3303 @opindex Wparentheses
3304 @opindex Wno-parentheses
3305 Warn if parentheses are omitted in certain contexts, such
3306 as when there is an assignment in a context where a truth value
3307 is expected, or when operators are nested whose precedence people
3308 often get confused about.
3310 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3311 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3312 interpretation from that of ordinary mathematical notation.
3314 Also warn about constructions where there may be confusion to which
3315 @code{if} statement an @code{else} branch belongs. Here is an example of
3330 In C/C++, every @code{else} branch belongs to the innermost possible
3331 @code{if} statement, which in this example is @code{if (b)}. This is
3332 often not what the programmer expected, as illustrated in the above
3333 example by indentation the programmer chose. When there is the
3334 potential for this confusion, GCC will issue a warning when this flag
3335 is specified. To eliminate the warning, add explicit braces around
3336 the innermost @code{if} statement so there is no way the @code{else}
3337 could belong to the enclosing @code{if}. The resulting code would
3354 Also warn for dangerous uses of the
3355 ?: with omitted middle operand GNU extension. When the condition
3356 in the ?: operator is a boolean expression the omitted value will
3357 be always 1. Often the user expects it to be a value computed
3358 inside the conditional expression instead.
3360 This warning is enabled by @option{-Wall}.
3362 @item -Wsequence-point
3363 @opindex Wsequence-point
3364 @opindex Wno-sequence-point
3365 Warn about code that may have undefined semantics because of violations
3366 of sequence point rules in the C and C++ standards.
3368 The C and C++ standards defines the order in which expressions in a C/C++
3369 program are evaluated in terms of @dfn{sequence points}, which represent
3370 a partial ordering between the execution of parts of the program: those
3371 executed before the sequence point, and those executed after it. These
3372 occur after the evaluation of a full expression (one which is not part
3373 of a larger expression), after the evaluation of the first operand of a
3374 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3375 function is called (but after the evaluation of its arguments and the
3376 expression denoting the called function), and in certain other places.
3377 Other than as expressed by the sequence point rules, the order of
3378 evaluation of subexpressions of an expression is not specified. All
3379 these rules describe only a partial order rather than a total order,
3380 since, for example, if two functions are called within one expression
3381 with no sequence point between them, the order in which the functions
3382 are called is not specified. However, the standards committee have
3383 ruled that function calls do not overlap.
3385 It is not specified when between sequence points modifications to the
3386 values of objects take effect. Programs whose behavior depends on this
3387 have undefined behavior; the C and C++ standards specify that ``Between
3388 the previous and next sequence point an object shall have its stored
3389 value modified at most once by the evaluation of an expression.
3390 Furthermore, the prior value shall be read only to determine the value
3391 to be stored.''. If a program breaks these rules, the results on any
3392 particular implementation are entirely unpredictable.
3394 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3395 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3396 diagnosed by this option, and it may give an occasional false positive
3397 result, but in general it has been found fairly effective at detecting
3398 this sort of problem in programs.
3400 The standard is worded confusingly, therefore there is some debate
3401 over the precise meaning of the sequence point rules in subtle cases.
3402 Links to discussions of the problem, including proposed formal
3403 definitions, may be found on the GCC readings page, at
3404 @uref{http://gcc.gnu.org/@/readings.html}.
3406 This warning is enabled by @option{-Wall} for C and C++.
3409 @opindex Wreturn-type
3410 @opindex Wno-return-type
3411 Warn whenever a function is defined with a return-type that defaults
3412 to @code{int}. Also warn about any @code{return} statement with no
3413 return-value in a function whose return-type is not @code{void}
3414 (falling off the end of the function body is considered returning
3415 without a value), and about a @code{return} statement with an
3416 expression in a function whose return-type is @code{void}.
3418 For C++, a function without return type always produces a diagnostic
3419 message, even when @option{-Wno-return-type} is specified. The only
3420 exceptions are @samp{main} and functions defined in system headers.
3422 This warning is enabled by @option{-Wall}.
3427 Warn whenever a @code{switch} statement has an index of enumerated type
3428 and lacks a @code{case} for one or more of the named codes of that
3429 enumeration. (The presence of a @code{default} label prevents this
3430 warning.) @code{case} labels outside the enumeration range also
3431 provoke warnings when this option is used (even if there is a
3432 @code{default} label).
3433 This warning is enabled by @option{-Wall}.
3435 @item -Wswitch-default
3436 @opindex Wswitch-default
3437 @opindex Wno-switch-default
3438 Warn whenever a @code{switch} statement does not have a @code{default}
3442 @opindex Wswitch-enum
3443 @opindex Wno-switch-enum
3444 Warn whenever a @code{switch} statement has an index of enumerated type
3445 and lacks a @code{case} for one or more of the named codes of that
3446 enumeration. @code{case} labels outside the enumeration range also
3447 provoke warnings when this option is used. The only difference
3448 between @option{-Wswitch} and this option is that this option gives a
3449 warning about an omitted enumeration code even if there is a
3450 @code{default} label.
3452 @item -Wsync-nand @r{(C and C++ only)}
3454 @opindex Wno-sync-nand
3455 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3456 built-in functions are used. These functions changed semantics in GCC 4.4.
3460 @opindex Wno-trigraphs
3461 Warn if any trigraphs are encountered that might change the meaning of
3462 the program (trigraphs within comments are not warned about).
3463 This warning is enabled by @option{-Wall}.
3465 @item -Wunused-but-set-parameter
3466 @opindex Wunused-but-set-parameter
3467 @opindex Wno-unused-but-set-parameter
3468 Warn whenever a function parameter is assigned to, but otherwise unused
3469 (aside from its declaration).
3471 To suppress this warning use the @samp{unused} attribute
3472 (@pxref{Variable Attributes}).
3474 This warning is also enabled by @option{-Wunused} together with
3477 @item -Wunused-but-set-variable
3478 @opindex Wunused-but-set-variable
3479 @opindex Wno-unused-but-set-variable
3480 Warn whenever a local variable is assigned to, but otherwise unused
3481 (aside from its declaration).
3482 This warning is enabled by @option{-Wall}.
3484 To suppress this warning use the @samp{unused} attribute
3485 (@pxref{Variable Attributes}).
3487 This warning is also enabled by @option{-Wunused}, which is enabled
3490 @item -Wunused-function
3491 @opindex Wunused-function
3492 @opindex Wno-unused-function
3493 Warn whenever a static function is declared but not defined or a
3494 non-inline static function is unused.
3495 This warning is enabled by @option{-Wall}.
3497 @item -Wunused-label
3498 @opindex Wunused-label
3499 @opindex Wno-unused-label
3500 Warn whenever a label is declared but not used.
3501 This warning is enabled by @option{-Wall}.
3503 To suppress this warning use the @samp{unused} attribute
3504 (@pxref{Variable Attributes}).
3506 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3507 @opindex Wunused-local-typedefs
3508 Warn when a typedef locally defined in a function is not used.
3510 @item -Wunused-parameter
3511 @opindex Wunused-parameter
3512 @opindex Wno-unused-parameter
3513 Warn whenever a function parameter is unused aside from its declaration.
3515 To suppress this warning use the @samp{unused} attribute
3516 (@pxref{Variable Attributes}).
3518 @item -Wno-unused-result
3519 @opindex Wunused-result
3520 @opindex Wno-unused-result
3521 Do not warn if a caller of a function marked with attribute
3522 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3523 its return value. The default is @option{-Wunused-result}.
3525 @item -Wunused-variable
3526 @opindex Wunused-variable
3527 @opindex Wno-unused-variable
3528 Warn whenever a local variable or non-constant static variable is unused
3529 aside from its declaration.
3530 This warning is enabled by @option{-Wall}.
3532 To suppress this warning use the @samp{unused} attribute
3533 (@pxref{Variable Attributes}).
3535 @item -Wunused-value
3536 @opindex Wunused-value
3537 @opindex Wno-unused-value
3538 Warn whenever a statement computes a result that is explicitly not
3539 used. To suppress this warning cast the unused expression to
3540 @samp{void}. This includes an expression-statement or the left-hand
3541 side of a comma expression that contains no side effects. For example,
3542 an expression such as @samp{x[i,j]} will cause a warning, while
3543 @samp{x[(void)i,j]} will not.
3545 This warning is enabled by @option{-Wall}.
3550 All the above @option{-Wunused} options combined.
3552 In order to get a warning about an unused function parameter, you must
3553 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3554 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3556 @item -Wuninitialized
3557 @opindex Wuninitialized
3558 @opindex Wno-uninitialized
3559 Warn if an automatic variable is used without first being initialized
3560 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3561 warn if a non-static reference or non-static @samp{const} member
3562 appears in a class without constructors.
3564 If you want to warn about code which uses the uninitialized value of the
3565 variable in its own initializer, use the @option{-Winit-self} option.
3567 These warnings occur for individual uninitialized or clobbered
3568 elements of structure, union or array variables as well as for
3569 variables which are uninitialized or clobbered as a whole. They do
3570 not occur for variables or elements declared @code{volatile}. Because
3571 these warnings depend on optimization, the exact variables or elements
3572 for which there are warnings will depend on the precise optimization
3573 options and version of GCC used.
3575 Note that there may be no warning about a variable that is used only
3576 to compute a value that itself is never used, because such
3577 computations may be deleted by data flow analysis before the warnings
3580 @item -Wmaybe-uninitialized
3581 @opindex Wmaybe-uninitialized
3582 @opindex Wno-maybe-uninitialized
3583 For an automatic variable, if there exists a path from the function
3584 entry to a use of the variable that is initialized, but there exist
3585 some other paths the variable is not initialized, the compiler will
3586 emit a warning if it can not prove the uninitialized paths do not
3587 happen at runtime. These warnings are made optional because GCC is
3588 not smart enough to see all the reasons why the code might be correct
3589 despite appearing to have an error. Here is one example of how
3610 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3611 always initialized, but GCC doesn't know this. To suppress the
3612 warning, the user needs to provide a default case with assert(0) or
3615 @cindex @code{longjmp} warnings
3616 This option also warns when a non-volatile automatic variable might be
3617 changed by a call to @code{longjmp}. These warnings as well are possible
3618 only in optimizing compilation.
3620 The compiler sees only the calls to @code{setjmp}. It cannot know
3621 where @code{longjmp} will be called; in fact, a signal handler could
3622 call it at any point in the code. As a result, you may get a warning
3623 even when there is in fact no problem because @code{longjmp} cannot
3624 in fact be called at the place which would cause a problem.
3626 Some spurious warnings can be avoided if you declare all the functions
3627 you use that never return as @code{noreturn}. @xref{Function
3630 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3632 @item -Wunknown-pragmas
3633 @opindex Wunknown-pragmas
3634 @opindex Wno-unknown-pragmas
3635 @cindex warning for unknown pragmas
3636 @cindex unknown pragmas, warning
3637 @cindex pragmas, warning of unknown
3638 Warn when a #pragma directive is encountered which is not understood by
3639 GCC@. If this command line option is used, warnings will even be issued
3640 for unknown pragmas in system header files. This is not the case if
3641 the warnings were only enabled by the @option{-Wall} command line option.
3644 @opindex Wno-pragmas
3646 Do not warn about misuses of pragmas, such as incorrect parameters,
3647 invalid syntax, or conflicts between pragmas. See also
3648 @samp{-Wunknown-pragmas}.
3650 @item -Wstrict-aliasing
3651 @opindex Wstrict-aliasing
3652 @opindex Wno-strict-aliasing
3653 This option is only active when @option{-fstrict-aliasing} is active.
3654 It warns about code which might break the strict aliasing rules that the
3655 compiler is using for optimization. The warning does not catch all
3656 cases, but does attempt to catch the more common pitfalls. It is
3657 included in @option{-Wall}.
3658 It is equivalent to @option{-Wstrict-aliasing=3}
3660 @item -Wstrict-aliasing=n
3661 @opindex Wstrict-aliasing=n
3662 @opindex Wno-strict-aliasing=n
3663 This option is only active when @option{-fstrict-aliasing} is active.
3664 It warns about code which might break the strict aliasing rules that the
3665 compiler is using for optimization.
3666 Higher levels correspond to higher accuracy (fewer false positives).
3667 Higher levels also correspond to more effort, similar to the way -O works.
3668 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3671 Level 1: Most aggressive, quick, least accurate.
3672 Possibly useful when higher levels
3673 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3674 false negatives. However, it has many false positives.
3675 Warns for all pointer conversions between possibly incompatible types,
3676 even if never dereferenced. Runs in the frontend only.
3678 Level 2: Aggressive, quick, not too precise.
3679 May still have many false positives (not as many as level 1 though),
3680 and few false negatives (but possibly more than level 1).
3681 Unlike level 1, it only warns when an address is taken. Warns about
3682 incomplete types. Runs in the frontend only.
3684 Level 3 (default for @option{-Wstrict-aliasing}):
3685 Should have very few false positives and few false
3686 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3687 Takes care of the common pun+dereference pattern in the frontend:
3688 @code{*(int*)&some_float}.
3689 If optimization is enabled, it also runs in the backend, where it deals
3690 with multiple statement cases using flow-sensitive points-to information.
3691 Only warns when the converted pointer is dereferenced.
3692 Does not warn about incomplete types.
3694 @item -Wstrict-overflow
3695 @itemx -Wstrict-overflow=@var{n}
3696 @opindex Wstrict-overflow
3697 @opindex Wno-strict-overflow
3698 This option is only active when @option{-fstrict-overflow} is active.
3699 It warns about cases where the compiler optimizes based on the
3700 assumption that signed overflow does not occur. Note that it does not
3701 warn about all cases where the code might overflow: it only warns
3702 about cases where the compiler implements some optimization. Thus
3703 this warning depends on the optimization level.
3705 An optimization which assumes that signed overflow does not occur is
3706 perfectly safe if the values of the variables involved are such that
3707 overflow never does, in fact, occur. Therefore this warning can
3708 easily give a false positive: a warning about code which is not
3709 actually a problem. To help focus on important issues, several
3710 warning levels are defined. No warnings are issued for the use of
3711 undefined signed overflow when estimating how many iterations a loop
3712 will require, in particular when determining whether a loop will be
3716 @item -Wstrict-overflow=1
3717 Warn about cases which are both questionable and easy to avoid. For
3718 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3719 compiler will simplify this to @code{1}. This level of
3720 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3721 are not, and must be explicitly requested.
3723 @item -Wstrict-overflow=2
3724 Also warn about other cases where a comparison is simplified to a
3725 constant. For example: @code{abs (x) >= 0}. This can only be
3726 simplified when @option{-fstrict-overflow} is in effect, because
3727 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3728 zero. @option{-Wstrict-overflow} (with no level) is the same as
3729 @option{-Wstrict-overflow=2}.
3731 @item -Wstrict-overflow=3
3732 Also warn about other cases where a comparison is simplified. For
3733 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3735 @item -Wstrict-overflow=4
3736 Also warn about other simplifications not covered by the above cases.
3737 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3739 @item -Wstrict-overflow=5
3740 Also warn about cases where the compiler reduces the magnitude of a
3741 constant involved in a comparison. For example: @code{x + 2 > y} will
3742 be simplified to @code{x + 1 >= y}. This is reported only at the
3743 highest warning level because this simplification applies to many
3744 comparisons, so this warning level will give a very large number of
3748 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3749 @opindex Wsuggest-attribute=
3750 @opindex Wno-suggest-attribute=
3751 Warn for cases where adding an attribute may be beneficial. The
3752 attributes currently supported are listed below.
3755 @item -Wsuggest-attribute=pure
3756 @itemx -Wsuggest-attribute=const
3757 @itemx -Wsuggest-attribute=noreturn
3758 @opindex Wsuggest-attribute=pure
3759 @opindex Wno-suggest-attribute=pure
3760 @opindex Wsuggest-attribute=const
3761 @opindex Wno-suggest-attribute=const
3762 @opindex Wsuggest-attribute=noreturn
3763 @opindex Wno-suggest-attribute=noreturn
3765 Warn about functions which might be candidates for attributes
3766 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3767 functions visible in other compilation units or (in the case of @code{pure} and
3768 @code{const}) if it cannot prove that the function returns normally. A function
3769 returns normally if it doesn't contain an infinite loop nor returns abnormally
3770 by throwing, calling @code{abort()} or trapping. This analysis requires option
3771 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3772 higher. Higher optimization levels improve the accuracy of the analysis.
3775 @item -Warray-bounds
3776 @opindex Wno-array-bounds
3777 @opindex Warray-bounds
3778 This option is only active when @option{-ftree-vrp} is active
3779 (default for @option{-O2} and above). It warns about subscripts to arrays
3780 that are always out of bounds. This warning is enabled by @option{-Wall}.
3782 @item -Wno-div-by-zero
3783 @opindex Wno-div-by-zero
3784 @opindex Wdiv-by-zero
3785 Do not warn about compile-time integer division by zero. Floating point
3786 division by zero is not warned about, as it can be a legitimate way of
3787 obtaining infinities and NaNs.
3789 @item -Wsystem-headers
3790 @opindex Wsystem-headers
3791 @opindex Wno-system-headers
3792 @cindex warnings from system headers
3793 @cindex system headers, warnings from
3794 Print warning messages for constructs found in system header files.
3795 Warnings from system headers are normally suppressed, on the assumption
3796 that they usually do not indicate real problems and would only make the
3797 compiler output harder to read. Using this command line option tells
3798 GCC to emit warnings from system headers as if they occurred in user
3799 code. However, note that using @option{-Wall} in conjunction with this
3800 option will @emph{not} warn about unknown pragmas in system
3801 headers---for that, @option{-Wunknown-pragmas} must also be used.
3804 @opindex Wtrampolines
3805 @opindex Wno-trampolines
3806 Warn about trampolines generated for pointers to nested functions.
3808 A trampoline is a small piece of data or code that is created at run
3809 time on the stack when the address of a nested function is taken, and
3810 is used to call the nested function indirectly. For some targets, it
3811 is made up of data only and thus requires no special treatment. But,
3812 for most targets, it is made up of code and thus requires the stack
3813 to be made executable in order for the program to work properly.
3816 @opindex Wfloat-equal
3817 @opindex Wno-float-equal
3818 Warn if floating point values are used in equality comparisons.
3820 The idea behind this is that sometimes it is convenient (for the
3821 programmer) to consider floating-point values as approximations to
3822 infinitely precise real numbers. If you are doing this, then you need
3823 to compute (by analyzing the code, or in some other way) the maximum or
3824 likely maximum error that the computation introduces, and allow for it
3825 when performing comparisons (and when producing output, but that's a
3826 different problem). In particular, instead of testing for equality, you
3827 would check to see whether the two values have ranges that overlap; and
3828 this is done with the relational operators, so equality comparisons are
3831 @item -Wtraditional @r{(C and Objective-C only)}
3832 @opindex Wtraditional
3833 @opindex Wno-traditional
3834 Warn about certain constructs that behave differently in traditional and
3835 ISO C@. Also warn about ISO C constructs that have no traditional C
3836 equivalent, and/or problematic constructs which should be avoided.
3840 Macro parameters that appear within string literals in the macro body.
3841 In traditional C macro replacement takes place within string literals,
3842 but does not in ISO C@.
3845 In traditional C, some preprocessor directives did not exist.
3846 Traditional preprocessors would only consider a line to be a directive
3847 if the @samp{#} appeared in column 1 on the line. Therefore
3848 @option{-Wtraditional} warns about directives that traditional C
3849 understands but would ignore because the @samp{#} does not appear as the
3850 first character on the line. It also suggests you hide directives like
3851 @samp{#pragma} not understood by traditional C by indenting them. Some
3852 traditional implementations would not recognize @samp{#elif}, so it
3853 suggests avoiding it altogether.
3856 A function-like macro that appears without arguments.
3859 The unary plus operator.
3862 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3863 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3864 constants.) Note, these suffixes appear in macros defined in the system
3865 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3866 Use of these macros in user code might normally lead to spurious
3867 warnings, however GCC's integrated preprocessor has enough context to
3868 avoid warning in these cases.
3871 A function declared external in one block and then used after the end of
3875 A @code{switch} statement has an operand of type @code{long}.
3878 A non-@code{static} function declaration follows a @code{static} one.
3879 This construct is not accepted by some traditional C compilers.
3882 The ISO type of an integer constant has a different width or
3883 signedness from its traditional type. This warning is only issued if
3884 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3885 typically represent bit patterns, are not warned about.
3888 Usage of ISO string concatenation is detected.
3891 Initialization of automatic aggregates.
3894 Identifier conflicts with labels. Traditional C lacks a separate
3895 namespace for labels.
3898 Initialization of unions. If the initializer is zero, the warning is
3899 omitted. This is done under the assumption that the zero initializer in
3900 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3901 initializer warnings and relies on default initialization to zero in the
3905 Conversions by prototypes between fixed/floating point values and vice
3906 versa. The absence of these prototypes when compiling with traditional
3907 C would cause serious problems. This is a subset of the possible
3908 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3911 Use of ISO C style function definitions. This warning intentionally is
3912 @emph{not} issued for prototype declarations or variadic functions
3913 because these ISO C features will appear in your code when using
3914 libiberty's traditional C compatibility macros, @code{PARAMS} and
3915 @code{VPARAMS}. This warning is also bypassed for nested functions
3916 because that feature is already a GCC extension and thus not relevant to
3917 traditional C compatibility.
3920 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3921 @opindex Wtraditional-conversion
3922 @opindex Wno-traditional-conversion
3923 Warn if a prototype causes a type conversion that is different from what
3924 would happen to the same argument in the absence of a prototype. This
3925 includes conversions of fixed point to floating and vice versa, and
3926 conversions changing the width or signedness of a fixed point argument
3927 except when the same as the default promotion.
3929 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3930 @opindex Wdeclaration-after-statement
3931 @opindex Wno-declaration-after-statement
3932 Warn when a declaration is found after a statement in a block. This
3933 construct, known from C++, was introduced with ISO C99 and is by default
3934 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3935 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3940 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3942 @item -Wno-endif-labels
3943 @opindex Wno-endif-labels
3944 @opindex Wendif-labels
3945 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3950 Warn whenever a local variable or type declaration shadows another variable,
3951 parameter, type, or class member (in C++), or whenever a built-in function
3952 is shadowed. Note that in C++, the compiler will not warn if a local variable
3953 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3955 @item -Wlarger-than=@var{len}
3956 @opindex Wlarger-than=@var{len}
3957 @opindex Wlarger-than-@var{len}
3958 Warn whenever an object of larger than @var{len} bytes is defined.
3960 @item -Wframe-larger-than=@var{len}
3961 @opindex Wframe-larger-than
3962 Warn if the size of a function frame is larger than @var{len} bytes.
3963 The computation done to determine the stack frame size is approximate
3964 and not conservative.
3965 The actual requirements may be somewhat greater than @var{len}
3966 even if you do not get a warning. In addition, any space allocated
3967 via @code{alloca}, variable-length arrays, or related constructs
3968 is not included by the compiler when determining
3969 whether or not to issue a warning.
3971 @item -Wno-free-nonheap-object
3972 @opindex Wno-free-nonheap-object
3973 @opindex Wfree-nonheap-object
3974 Do not warn when attempting to free an object which was not allocated
3977 @item -Wstack-usage=@var{len}
3978 @opindex Wstack-usage
3979 Warn if the stack usage of a function might be larger than @var{len} bytes.
3980 The computation done to determine the stack usage is conservative.
3981 Any space allocated via @code{alloca}, variable-length arrays, or related
3982 constructs is included by the compiler when determining whether or not to
3985 The message is in keeping with the output of @option{-fstack-usage}.
3989 If the stack usage is fully static but exceeds the specified amount, it's:
3992 Â warning: stack usage is 1120 bytes
3995 If the stack usage is (partly) dynamic but bounded, it's:
3998 Â warning: stack usage might be 1648 bytes
4001 If the stack usage is (partly) dynamic and not bounded, it's:
4004 Â warning: stack usage might be unbounded
4008 @item -Wunsafe-loop-optimizations
4009 @opindex Wunsafe-loop-optimizations
4010 @opindex Wno-unsafe-loop-optimizations
4011 Warn if the loop cannot be optimized because the compiler could not
4012 assume anything on the bounds of the loop indices. With
4013 @option{-funsafe-loop-optimizations} warn if the compiler made
4016 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4017 @opindex Wno-pedantic-ms-format
4018 @opindex Wpedantic-ms-format
4019 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4020 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4021 depending on the MS runtime, when you are using the options @option{-Wformat}
4022 and @option{-pedantic} without gnu-extensions.
4024 @item -Wpointer-arith
4025 @opindex Wpointer-arith
4026 @opindex Wno-pointer-arith
4027 Warn about anything that depends on the ``size of'' a function type or
4028 of @code{void}. GNU C assigns these types a size of 1, for
4029 convenience in calculations with @code{void *} pointers and pointers
4030 to functions. In C++, warn also when an arithmetic operation involves
4031 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4034 @opindex Wtype-limits
4035 @opindex Wno-type-limits
4036 Warn if a comparison is always true or always false due to the limited
4037 range of the data type, but do not warn for constant expressions. For
4038 example, warn if an unsigned variable is compared against zero with
4039 @samp{<} or @samp{>=}. This warning is also enabled by
4042 @item -Wbad-function-cast @r{(C and Objective-C only)}
4043 @opindex Wbad-function-cast
4044 @opindex Wno-bad-function-cast
4045 Warn whenever a function call is cast to a non-matching type.
4046 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4048 @item -Wc++-compat @r{(C and Objective-C only)}
4049 Warn about ISO C constructs that are outside of the common subset of
4050 ISO C and ISO C++, e.g.@: request for implicit conversion from
4051 @code{void *} to a pointer to non-@code{void} type.
4053 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
4054 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
4055 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
4056 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
4060 @opindex Wno-cast-qual
4061 Warn whenever a pointer is cast so as to remove a type qualifier from
4062 the target type. For example, warn if a @code{const char *} is cast
4063 to an ordinary @code{char *}.
4065 Also warn when making a cast which introduces a type qualifier in an
4066 unsafe way. For example, casting @code{char **} to @code{const char **}
4067 is unsafe, as in this example:
4070 /* p is char ** value. */
4071 const char **q = (const char **) p;
4072 /* Assignment of readonly string to const char * is OK. */
4074 /* Now char** pointer points to read-only memory. */
4079 @opindex Wcast-align
4080 @opindex Wno-cast-align
4081 Warn whenever a pointer is cast such that the required alignment of the
4082 target is increased. For example, warn if a @code{char *} is cast to
4083 an @code{int *} on machines where integers can only be accessed at
4084 two- or four-byte boundaries.
4086 @item -Wwrite-strings
4087 @opindex Wwrite-strings
4088 @opindex Wno-write-strings
4089 When compiling C, give string constants the type @code{const
4090 char[@var{length}]} so that copying the address of one into a
4091 non-@code{const} @code{char *} pointer will get a warning. These
4092 warnings will help you find at compile time code that can try to write
4093 into a string constant, but only if you have been very careful about
4094 using @code{const} in declarations and prototypes. Otherwise, it will
4095 just be a nuisance. This is why we did not make @option{-Wall} request
4098 When compiling C++, warn about the deprecated conversion from string
4099 literals to @code{char *}. This warning is enabled by default for C++
4104 @opindex Wno-clobbered
4105 Warn for variables that might be changed by @samp{longjmp} or
4106 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4109 @opindex Wconversion
4110 @opindex Wno-conversion
4111 Warn for implicit conversions that may alter a value. This includes
4112 conversions between real and integer, like @code{abs (x)} when
4113 @code{x} is @code{double}; conversions between signed and unsigned,
4114 like @code{unsigned ui = -1}; and conversions to smaller types, like
4115 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4116 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4117 changed by the conversion like in @code{abs (2.0)}. Warnings about
4118 conversions between signed and unsigned integers can be disabled by
4119 using @option{-Wno-sign-conversion}.
4121 For C++, also warn for confusing overload resolution for user-defined
4122 conversions; and conversions that will never use a type conversion
4123 operator: conversions to @code{void}, the same type, a base class or a
4124 reference to them. Warnings about conversions between signed and
4125 unsigned integers are disabled by default in C++ unless
4126 @option{-Wsign-conversion} is explicitly enabled.
4128 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4129 @opindex Wconversion-null
4130 @opindex Wno-conversion-null
4131 Do not warn for conversions between @code{NULL} and non-pointer
4132 types. @option{-Wconversion-null} is enabled by default.
4135 @opindex Wempty-body
4136 @opindex Wno-empty-body
4137 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4138 while} statement. This warning is also enabled by @option{-Wextra}.
4140 @item -Wenum-compare
4141 @opindex Wenum-compare
4142 @opindex Wno-enum-compare
4143 Warn about a comparison between values of different enum types. In C++
4144 this warning is enabled by default. In C this warning is enabled by
4147 @item -Wjump-misses-init @r{(C, Objective-C only)}
4148 @opindex Wjump-misses-init
4149 @opindex Wno-jump-misses-init
4150 Warn if a @code{goto} statement or a @code{switch} statement jumps
4151 forward across the initialization of a variable, or jumps backward to a
4152 label after the variable has been initialized. This only warns about
4153 variables which are initialized when they are declared. This warning is
4154 only supported for C and Objective C; in C++ this sort of branch is an
4157 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4158 can be disabled with the @option{-Wno-jump-misses-init} option.
4160 @item -Wsign-compare
4161 @opindex Wsign-compare
4162 @opindex Wno-sign-compare
4163 @cindex warning for comparison of signed and unsigned values
4164 @cindex comparison of signed and unsigned values, warning
4165 @cindex signed and unsigned values, comparison warning
4166 Warn when a comparison between signed and unsigned values could produce
4167 an incorrect result when the signed value is converted to unsigned.
4168 This warning is also enabled by @option{-Wextra}; to get the other warnings
4169 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4171 @item -Wsign-conversion
4172 @opindex Wsign-conversion
4173 @opindex Wno-sign-conversion
4174 Warn for implicit conversions that may change the sign of an integer
4175 value, like assigning a signed integer expression to an unsigned
4176 integer variable. An explicit cast silences the warning. In C, this
4177 option is enabled also by @option{-Wconversion}.
4181 @opindex Wno-address
4182 Warn about suspicious uses of memory addresses. These include using
4183 the address of a function in a conditional expression, such as
4184 @code{void func(void); if (func)}, and comparisons against the memory
4185 address of a string literal, such as @code{if (x == "abc")}. Such
4186 uses typically indicate a programmer error: the address of a function
4187 always evaluates to true, so their use in a conditional usually
4188 indicate that the programmer forgot the parentheses in a function
4189 call; and comparisons against string literals result in unspecified
4190 behavior and are not portable in C, so they usually indicate that the
4191 programmer intended to use @code{strcmp}. This warning is enabled by
4195 @opindex Wlogical-op
4196 @opindex Wno-logical-op
4197 Warn about suspicious uses of logical operators in expressions.
4198 This includes using logical operators in contexts where a
4199 bit-wise operator is likely to be expected.
4201 @item -Waggregate-return
4202 @opindex Waggregate-return
4203 @opindex Wno-aggregate-return
4204 Warn if any functions that return structures or unions are defined or
4205 called. (In languages where you can return an array, this also elicits
4208 @item -Wno-attributes
4209 @opindex Wno-attributes
4210 @opindex Wattributes
4211 Do not warn if an unexpected @code{__attribute__} is used, such as
4212 unrecognized attributes, function attributes applied to variables,
4213 etc. This will not stop errors for incorrect use of supported
4216 @item -Wno-builtin-macro-redefined
4217 @opindex Wno-builtin-macro-redefined
4218 @opindex Wbuiltin-macro-redefined
4219 Do not warn if certain built-in macros are redefined. This suppresses
4220 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4221 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4223 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4224 @opindex Wstrict-prototypes
4225 @opindex Wno-strict-prototypes
4226 Warn if a function is declared or defined without specifying the
4227 argument types. (An old-style function definition is permitted without
4228 a warning if preceded by a declaration which specifies the argument
4231 @item -Wold-style-declaration @r{(C and Objective-C only)}
4232 @opindex Wold-style-declaration
4233 @opindex Wno-old-style-declaration
4234 Warn for obsolescent usages, according to the C Standard, in a
4235 declaration. For example, warn if storage-class specifiers like
4236 @code{static} are not the first things in a declaration. This warning
4237 is also enabled by @option{-Wextra}.
4239 @item -Wold-style-definition @r{(C and Objective-C only)}
4240 @opindex Wold-style-definition
4241 @opindex Wno-old-style-definition
4242 Warn if an old-style function definition is used. A warning is given
4243 even if there is a previous prototype.
4245 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4246 @opindex Wmissing-parameter-type
4247 @opindex Wno-missing-parameter-type
4248 A function parameter is declared without a type specifier in K&R-style
4255 This warning is also enabled by @option{-Wextra}.
4257 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4258 @opindex Wmissing-prototypes
4259 @opindex Wno-missing-prototypes
4260 Warn if a global function is defined without a previous prototype
4261 declaration. This warning is issued even if the definition itself
4262 provides a prototype. The aim is to detect global functions that fail
4263 to be declared in header files.
4265 @item -Wmissing-declarations
4266 @opindex Wmissing-declarations
4267 @opindex Wno-missing-declarations
4268 Warn if a global function is defined without a previous declaration.
4269 Do so even if the definition itself provides a prototype.
4270 Use this option to detect global functions that are not declared in
4271 header files. In C++, no warnings are issued for function templates,
4272 or for inline functions, or for functions in anonymous namespaces.
4274 @item -Wmissing-field-initializers
4275 @opindex Wmissing-field-initializers
4276 @opindex Wno-missing-field-initializers
4280 Warn if a structure's initializer has some fields missing. For
4281 example, the following code would cause such a warning, because
4282 @code{x.h} is implicitly zero:
4285 struct s @{ int f, g, h; @};
4286 struct s x = @{ 3, 4 @};
4289 This option does not warn about designated initializers, so the following
4290 modification would not trigger a warning:
4293 struct s @{ int f, g, h; @};
4294 struct s x = @{ .f = 3, .g = 4 @};
4297 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4298 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4300 @item -Wmissing-format-attribute
4301 @opindex Wmissing-format-attribute
4302 @opindex Wno-missing-format-attribute
4305 Warn about function pointers which might be candidates for @code{format}
4306 attributes. Note these are only possible candidates, not absolute ones.
4307 GCC will guess that function pointers with @code{format} attributes that
4308 are used in assignment, initialization, parameter passing or return
4309 statements should have a corresponding @code{format} attribute in the
4310 resulting type. I.e.@: the left-hand side of the assignment or
4311 initialization, the type of the parameter variable, or the return type
4312 of the containing function respectively should also have a @code{format}
4313 attribute to avoid the warning.
4315 GCC will also warn about function definitions which might be
4316 candidates for @code{format} attributes. Again, these are only
4317 possible candidates. GCC will guess that @code{format} attributes
4318 might be appropriate for any function that calls a function like
4319 @code{vprintf} or @code{vscanf}, but this might not always be the
4320 case, and some functions for which @code{format} attributes are
4321 appropriate may not be detected.
4323 @item -Wno-multichar
4324 @opindex Wno-multichar
4326 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4327 Usually they indicate a typo in the user's code, as they have
4328 implementation-defined values, and should not be used in portable code.
4330 @item -Wnormalized=<none|id|nfc|nfkc>
4331 @opindex Wnormalized=
4334 @cindex character set, input normalization
4335 In ISO C and ISO C++, two identifiers are different if they are
4336 different sequences of characters. However, sometimes when characters
4337 outside the basic ASCII character set are used, you can have two
4338 different character sequences that look the same. To avoid confusion,
4339 the ISO 10646 standard sets out some @dfn{normalization rules} which
4340 when applied ensure that two sequences that look the same are turned into
4341 the same sequence. GCC can warn you if you are using identifiers which
4342 have not been normalized; this option controls that warning.
4344 There are four levels of warning that GCC supports. The default is
4345 @option{-Wnormalized=nfc}, which warns about any identifier which is
4346 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4347 recommended form for most uses.
4349 Unfortunately, there are some characters which ISO C and ISO C++ allow
4350 in identifiers that when turned into NFC aren't allowable as
4351 identifiers. That is, there's no way to use these symbols in portable
4352 ISO C or C++ and have all your identifiers in NFC@.
4353 @option{-Wnormalized=id} suppresses the warning for these characters.
4354 It is hoped that future versions of the standards involved will correct
4355 this, which is why this option is not the default.
4357 You can switch the warning off for all characters by writing
4358 @option{-Wnormalized=none}. You would only want to do this if you
4359 were using some other normalization scheme (like ``D''), because
4360 otherwise you can easily create bugs that are literally impossible to see.
4362 Some characters in ISO 10646 have distinct meanings but look identical
4363 in some fonts or display methodologies, especially once formatting has
4364 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4365 LETTER N'', will display just like a regular @code{n} which has been
4366 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4367 normalization scheme to convert all these into a standard form as
4368 well, and GCC will warn if your code is not in NFKC if you use
4369 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4370 about every identifier that contains the letter O because it might be
4371 confused with the digit 0, and so is not the default, but may be
4372 useful as a local coding convention if the programming environment is
4373 unable to be fixed to display these characters distinctly.
4375 @item -Wno-deprecated
4376 @opindex Wno-deprecated
4377 @opindex Wdeprecated
4378 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4380 @item -Wno-deprecated-declarations
4381 @opindex Wno-deprecated-declarations
4382 @opindex Wdeprecated-declarations
4383 Do not warn about uses of functions (@pxref{Function Attributes}),
4384 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4385 Attributes}) marked as deprecated by using the @code{deprecated}
4389 @opindex Wno-overflow
4391 Do not warn about compile-time overflow in constant expressions.
4393 @item -Woverride-init @r{(C and Objective-C only)}
4394 @opindex Woverride-init
4395 @opindex Wno-override-init
4399 Warn if an initialized field without side effects is overridden when
4400 using designated initializers (@pxref{Designated Inits, , Designated
4403 This warning is included in @option{-Wextra}. To get other
4404 @option{-Wextra} warnings without this one, use @samp{-Wextra
4405 -Wno-override-init}.
4410 Warn if a structure is given the packed attribute, but the packed
4411 attribute has no effect on the layout or size of the structure.
4412 Such structures may be mis-aligned for little benefit. For
4413 instance, in this code, the variable @code{f.x} in @code{struct bar}
4414 will be misaligned even though @code{struct bar} does not itself
4415 have the packed attribute:
4422 @} __attribute__((packed));
4430 @item -Wpacked-bitfield-compat
4431 @opindex Wpacked-bitfield-compat
4432 @opindex Wno-packed-bitfield-compat
4433 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4434 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4435 the change can lead to differences in the structure layout. GCC
4436 informs you when the offset of such a field has changed in GCC 4.4.
4437 For example there is no longer a 4-bit padding between field @code{a}
4438 and @code{b} in this structure:
4445 @} __attribute__ ((packed));
4448 This warning is enabled by default. Use
4449 @option{-Wno-packed-bitfield-compat} to disable this warning.
4454 Warn if padding is included in a structure, either to align an element
4455 of the structure or to align the whole structure. Sometimes when this
4456 happens it is possible to rearrange the fields of the structure to
4457 reduce the padding and so make the structure smaller.
4459 @item -Wredundant-decls
4460 @opindex Wredundant-decls
4461 @opindex Wno-redundant-decls
4462 Warn if anything is declared more than once in the same scope, even in
4463 cases where multiple declaration is valid and changes nothing.
4465 @item -Wnested-externs @r{(C and Objective-C only)}
4466 @opindex Wnested-externs
4467 @opindex Wno-nested-externs
4468 Warn if an @code{extern} declaration is encountered within a function.
4473 Warn if a function can not be inlined and it was declared as inline.
4474 Even with this option, the compiler will not warn about failures to
4475 inline functions declared in system headers.
4477 The compiler uses a variety of heuristics to determine whether or not
4478 to inline a function. For example, the compiler takes into account
4479 the size of the function being inlined and the amount of inlining
4480 that has already been done in the current function. Therefore,
4481 seemingly insignificant changes in the source program can cause the
4482 warnings produced by @option{-Winline} to appear or disappear.
4484 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4485 @opindex Wno-invalid-offsetof
4486 @opindex Winvalid-offsetof
4487 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4488 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4489 to a non-POD type is undefined. In existing C++ implementations,
4490 however, @samp{offsetof} typically gives meaningful results even when
4491 applied to certain kinds of non-POD types. (Such as a simple
4492 @samp{struct} that fails to be a POD type only by virtue of having a
4493 constructor.) This flag is for users who are aware that they are
4494 writing nonportable code and who have deliberately chosen to ignore the
4497 The restrictions on @samp{offsetof} may be relaxed in a future version
4498 of the C++ standard.
4500 @item -Wno-int-to-pointer-cast
4501 @opindex Wno-int-to-pointer-cast
4502 @opindex Wint-to-pointer-cast
4503 Suppress warnings from casts to pointer type of an integer of a
4504 different size. In C++, casting to a pointer type of smaller size is
4505 an error. @option{Wint-to-pointer-cast} is enabled by default.
4508 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4509 @opindex Wno-pointer-to-int-cast
4510 @opindex Wpointer-to-int-cast
4511 Suppress warnings from casts from a pointer to an integer type of a
4515 @opindex Winvalid-pch
4516 @opindex Wno-invalid-pch
4517 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4518 the search path but can't be used.
4522 @opindex Wno-long-long
4523 Warn if @samp{long long} type is used. This is enabled by either
4524 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4525 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4527 @item -Wvariadic-macros
4528 @opindex Wvariadic-macros
4529 @opindex Wno-variadic-macros
4530 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4531 alternate syntax when in pedantic ISO C99 mode. This is default.
4532 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4537 Warn if variable length array is used in the code.
4538 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4539 the variable length array.
4541 @item -Wvolatile-register-var
4542 @opindex Wvolatile-register-var
4543 @opindex Wno-volatile-register-var
4544 Warn if a register variable is declared volatile. The volatile
4545 modifier does not inhibit all optimizations that may eliminate reads
4546 and/or writes to register variables. This warning is enabled by
4549 @item -Wdisabled-optimization
4550 @opindex Wdisabled-optimization
4551 @opindex Wno-disabled-optimization
4552 Warn if a requested optimization pass is disabled. This warning does
4553 not generally indicate that there is anything wrong with your code; it
4554 merely indicates that GCC's optimizers were unable to handle the code
4555 effectively. Often, the problem is that your code is too big or too
4556 complex; GCC will refuse to optimize programs when the optimization
4557 itself is likely to take inordinate amounts of time.
4559 @item -Wpointer-sign @r{(C and Objective-C only)}
4560 @opindex Wpointer-sign
4561 @opindex Wno-pointer-sign
4562 Warn for pointer argument passing or assignment with different signedness.
4563 This option is only supported for C and Objective-C@. It is implied by
4564 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4565 @option{-Wno-pointer-sign}.
4567 @item -Wstack-protector
4568 @opindex Wstack-protector
4569 @opindex Wno-stack-protector
4570 This option is only active when @option{-fstack-protector} is active. It
4571 warns about functions that will not be protected against stack smashing.
4574 @opindex Wno-mudflap
4575 Suppress warnings about constructs that cannot be instrumented by
4578 @item -Woverlength-strings
4579 @opindex Woverlength-strings
4580 @opindex Wno-overlength-strings
4581 Warn about string constants which are longer than the ``minimum
4582 maximum'' length specified in the C standard. Modern compilers
4583 generally allow string constants which are much longer than the
4584 standard's minimum limit, but very portable programs should avoid
4585 using longer strings.
4587 The limit applies @emph{after} string constant concatenation, and does
4588 not count the trailing NUL@. In C90, the limit was 509 characters; in
4589 C99, it was raised to 4095. C++98 does not specify a normative
4590 minimum maximum, so we do not diagnose overlength strings in C++@.
4592 This option is implied by @option{-pedantic}, and can be disabled with
4593 @option{-Wno-overlength-strings}.
4595 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4596 @opindex Wunsuffixed-float-constants
4598 GCC will issue a warning for any floating constant that does not have
4599 a suffix. When used together with @option{-Wsystem-headers} it will
4600 warn about such constants in system header files. This can be useful
4601 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4602 from the decimal floating-point extension to C99.
4605 @node Debugging Options
4606 @section Options for Debugging Your Program or GCC
4607 @cindex options, debugging
4608 @cindex debugging information options
4610 GCC has various special options that are used for debugging
4611 either your program or GCC:
4616 Produce debugging information in the operating system's native format
4617 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4620 On most systems that use stabs format, @option{-g} enables use of extra
4621 debugging information that only GDB can use; this extra information
4622 makes debugging work better in GDB but will probably make other debuggers
4624 refuse to read the program. If you want to control for certain whether
4625 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4626 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4628 GCC allows you to use @option{-g} with
4629 @option{-O}. The shortcuts taken by optimized code may occasionally
4630 produce surprising results: some variables you declared may not exist
4631 at all; flow of control may briefly move where you did not expect it;
4632 some statements may not be executed because they compute constant
4633 results or their values were already at hand; some statements may
4634 execute in different places because they were moved out of loops.
4636 Nevertheless it proves possible to debug optimized output. This makes
4637 it reasonable to use the optimizer for programs that might have bugs.
4639 The following options are useful when GCC is generated with the
4640 capability for more than one debugging format.
4644 Produce debugging information for use by GDB@. This means to use the
4645 most expressive format available (DWARF 2, stabs, or the native format
4646 if neither of those are supported), including GDB extensions if at all
4651 Produce debugging information in stabs format (if that is supported),
4652 without GDB extensions. This is the format used by DBX on most BSD
4653 systems. On MIPS, Alpha and System V Release 4 systems this option
4654 produces stabs debugging output which is not understood by DBX or SDB@.
4655 On System V Release 4 systems this option requires the GNU assembler.
4657 @item -feliminate-unused-debug-symbols
4658 @opindex feliminate-unused-debug-symbols
4659 Produce debugging information in stabs format (if that is supported),
4660 for only symbols that are actually used.
4662 @item -femit-class-debug-always
4663 Instead of emitting debugging information for a C++ class in only one
4664 object file, emit it in all object files using the class. This option
4665 should be used only with debuggers that are unable to handle the way GCC
4666 normally emits debugging information for classes because using this
4667 option will increase the size of debugging information by as much as a
4670 @item -fno-debug-types-section
4671 @opindex fno-debug-types-section
4672 @opindex fdebug-types-section
4673 By default when using DWARF v4 or higher type DIEs will be put into
4674 their own .debug_types section instead of making them part of the
4675 .debug_info section. It is more efficient to put them in a separate
4676 comdat sections since the linker will then be able to remove duplicates.
4677 But not all DWARF consumers support .debug_types sections yet.
4681 Produce debugging information in stabs format (if that is supported),
4682 using GNU extensions understood only by the GNU debugger (GDB)@. The
4683 use of these extensions is likely to make other debuggers crash or
4684 refuse to read the program.
4688 Produce debugging information in COFF format (if that is supported).
4689 This is the format used by SDB on most System V systems prior to
4694 Produce debugging information in XCOFF format (if that is supported).
4695 This is the format used by the DBX debugger on IBM RS/6000 systems.
4699 Produce debugging information in XCOFF format (if that is supported),
4700 using GNU extensions understood only by the GNU debugger (GDB)@. The
4701 use of these extensions is likely to make other debuggers crash or
4702 refuse to read the program, and may cause assemblers other than the GNU
4703 assembler (GAS) to fail with an error.
4705 @item -gdwarf-@var{version}
4706 @opindex gdwarf-@var{version}
4707 Produce debugging information in DWARF format (if that is
4708 supported). This is the format used by DBX on IRIX 6. The value
4709 of @var{version} may be either 2, 3 or 4; the default version is 2.
4711 Note that with DWARF version 2 some ports require, and will always
4712 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4714 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4715 for maximum benefit.
4717 @item -grecord-gcc-switches
4718 @opindex grecord-gcc-switches
4719 This switch causes the command line options, that were used to invoke the
4720 compiler and may affect code generation, to be appended to the
4721 DW_AT_producer attribute in DWARF debugging information. The options
4722 are concatenated with spaces separating them from each other and from
4723 the compiler version. See also @option{-frecord-gcc-switches} for another
4724 way of storing compiler options into the object file.
4726 @item -gno-record-gcc-switches
4727 @opindex gno-record-gcc-switches
4728 Disallow appending command line options to the DW_AT_producer attribute
4729 in DWARF debugging information. This is the default.
4731 @item -gstrict-dwarf
4732 @opindex gstrict-dwarf
4733 Disallow using extensions of later DWARF standard version than selected
4734 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4735 DWARF extensions from later standard versions is allowed.
4737 @item -gno-strict-dwarf
4738 @opindex gno-strict-dwarf
4739 Allow using extensions of later DWARF standard version than selected with
4740 @option{-gdwarf-@var{version}}.
4744 Produce debugging information in VMS debug format (if that is
4745 supported). This is the format used by DEBUG on VMS systems.
4748 @itemx -ggdb@var{level}
4749 @itemx -gstabs@var{level}
4750 @itemx -gcoff@var{level}
4751 @itemx -gxcoff@var{level}
4752 @itemx -gvms@var{level}
4753 Request debugging information and also use @var{level} to specify how
4754 much information. The default level is 2.
4756 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4759 Level 1 produces minimal information, enough for making backtraces in
4760 parts of the program that you don't plan to debug. This includes
4761 descriptions of functions and external variables, but no information
4762 about local variables and no line numbers.
4764 Level 3 includes extra information, such as all the macro definitions
4765 present in the program. Some debuggers support macro expansion when
4766 you use @option{-g3}.
4768 @option{-gdwarf-2} does not accept a concatenated debug level, because
4769 GCC used to support an option @option{-gdwarf} that meant to generate
4770 debug information in version 1 of the DWARF format (which is very
4771 different from version 2), and it would have been too confusing. That
4772 debug format is long obsolete, but the option cannot be changed now.
4773 Instead use an additional @option{-g@var{level}} option to change the
4774 debug level for DWARF.
4778 Turn off generation of debug info, if leaving out this option would have
4779 generated it, or turn it on at level 2 otherwise. The position of this
4780 argument in the command line does not matter, it takes effect after all
4781 other options are processed, and it does so only once, no matter how
4782 many times it is given. This is mainly intended to be used with
4783 @option{-fcompare-debug}.
4785 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4786 @opindex fdump-final-insns
4787 Dump the final internal representation (RTL) to @var{file}. If the
4788 optional argument is omitted (or if @var{file} is @code{.}), the name
4789 of the dump file will be determined by appending @code{.gkd} to the
4790 compilation output file name.
4792 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4793 @opindex fcompare-debug
4794 @opindex fno-compare-debug
4795 If no error occurs during compilation, run the compiler a second time,
4796 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4797 passed to the second compilation. Dump the final internal
4798 representation in both compilations, and print an error if they differ.
4800 If the equal sign is omitted, the default @option{-gtoggle} is used.
4802 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4803 and nonzero, implicitly enables @option{-fcompare-debug}. If
4804 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4805 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4808 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4809 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4810 of the final representation and the second compilation, preventing even
4811 @env{GCC_COMPARE_DEBUG} from taking effect.
4813 To verify full coverage during @option{-fcompare-debug} testing, set
4814 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4815 which GCC will reject as an invalid option in any actual compilation
4816 (rather than preprocessing, assembly or linking). To get just a
4817 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4818 not overridden} will do.
4820 @item -fcompare-debug-second
4821 @opindex fcompare-debug-second
4822 This option is implicitly passed to the compiler for the second
4823 compilation requested by @option{-fcompare-debug}, along with options to
4824 silence warnings, and omitting other options that would cause
4825 side-effect compiler outputs to files or to the standard output. Dump
4826 files and preserved temporary files are renamed so as to contain the
4827 @code{.gk} additional extension during the second compilation, to avoid
4828 overwriting those generated by the first.
4830 When this option is passed to the compiler driver, it causes the
4831 @emph{first} compilation to be skipped, which makes it useful for little
4832 other than debugging the compiler proper.
4834 @item -feliminate-dwarf2-dups
4835 @opindex feliminate-dwarf2-dups
4836 Compress DWARF2 debugging information by eliminating duplicated
4837 information about each symbol. This option only makes sense when
4838 generating DWARF2 debugging information with @option{-gdwarf-2}.
4840 @item -femit-struct-debug-baseonly
4841 Emit debug information for struct-like types
4842 only when the base name of the compilation source file
4843 matches the base name of file in which the struct was defined.
4845 This option substantially reduces the size of debugging information,
4846 but at significant potential loss in type information to the debugger.
4847 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4848 See @option{-femit-struct-debug-detailed} for more detailed control.
4850 This option works only with DWARF 2.
4852 @item -femit-struct-debug-reduced
4853 Emit debug information for struct-like types
4854 only when the base name of the compilation source file
4855 matches the base name of file in which the type was defined,
4856 unless the struct is a template or defined in a system header.
4858 This option significantly reduces the size of debugging information,
4859 with some potential loss in type information to the debugger.
4860 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4861 See @option{-femit-struct-debug-detailed} for more detailed control.
4863 This option works only with DWARF 2.
4865 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4866 Specify the struct-like types
4867 for which the compiler will generate debug information.
4868 The intent is to reduce duplicate struct debug information
4869 between different object files within the same program.
4871 This option is a detailed version of
4872 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4873 which will serve for most needs.
4875 A specification has the syntax@*
4876 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4878 The optional first word limits the specification to
4879 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4880 A struct type is used directly when it is the type of a variable, member.
4881 Indirect uses arise through pointers to structs.
4882 That is, when use of an incomplete struct would be legal, the use is indirect.
4884 @samp{struct one direct; struct two * indirect;}.
4886 The optional second word limits the specification to
4887 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4888 Generic structs are a bit complicated to explain.
4889 For C++, these are non-explicit specializations of template classes,
4890 or non-template classes within the above.
4891 Other programming languages have generics,
4892 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4894 The third word specifies the source files for those
4895 structs for which the compiler will emit debug information.
4896 The values @samp{none} and @samp{any} have the normal meaning.
4897 The value @samp{base} means that
4898 the base of name of the file in which the type declaration appears
4899 must match the base of the name of the main compilation file.
4900 In practice, this means that
4901 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4902 but types declared in other header will not.
4903 The value @samp{sys} means those types satisfying @samp{base}
4904 or declared in system or compiler headers.
4906 You may need to experiment to determine the best settings for your application.
4908 The default is @samp{-femit-struct-debug-detailed=all}.
4910 This option works only with DWARF 2.
4912 @item -fno-merge-debug-strings
4913 @opindex fmerge-debug-strings
4914 @opindex fno-merge-debug-strings
4915 Direct the linker to not merge together strings in the debugging
4916 information which are identical in different object files. Merging is
4917 not supported by all assemblers or linkers. Merging decreases the size
4918 of the debug information in the output file at the cost of increasing
4919 link processing time. Merging is enabled by default.
4921 @item -fdebug-prefix-map=@var{old}=@var{new}
4922 @opindex fdebug-prefix-map
4923 When compiling files in directory @file{@var{old}}, record debugging
4924 information describing them as in @file{@var{new}} instead.
4926 @item -fno-dwarf2-cfi-asm
4927 @opindex fdwarf2-cfi-asm
4928 @opindex fno-dwarf2-cfi-asm
4929 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4930 instead of using GAS @code{.cfi_*} directives.
4932 @cindex @command{prof}
4935 Generate extra code to write profile information suitable for the
4936 analysis program @command{prof}. You must use this option when compiling
4937 the source files you want data about, and you must also use it when
4940 @cindex @command{gprof}
4943 Generate extra code to write profile information suitable for the
4944 analysis program @command{gprof}. You must use this option when compiling
4945 the source files you want data about, and you must also use it when
4950 Makes the compiler print out each function name as it is compiled, and
4951 print some statistics about each pass when it finishes.
4954 @opindex ftime-report
4955 Makes the compiler print some statistics about the time consumed by each
4956 pass when it finishes.
4959 @opindex fmem-report
4960 Makes the compiler print some statistics about permanent memory
4961 allocation when it finishes.
4963 @item -fpre-ipa-mem-report
4964 @opindex fpre-ipa-mem-report
4965 @item -fpost-ipa-mem-report
4966 @opindex fpost-ipa-mem-report
4967 Makes the compiler print some statistics about permanent memory
4968 allocation before or after interprocedural optimization.
4971 @opindex fstack-usage
4972 Makes the compiler output stack usage information for the program, on a
4973 per-function basis. The filename for the dump is made by appending
4974 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4975 the output file, if explicitly specified and it is not an executable,
4976 otherwise it is the basename of the source file. An entry is made up
4981 The name of the function.
4985 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4988 The qualifier @code{static} means that the function manipulates the stack
4989 statically: a fixed number of bytes are allocated for the frame on function
4990 entry and released on function exit; no stack adjustments are otherwise made
4991 in the function. The second field is this fixed number of bytes.
4993 The qualifier @code{dynamic} means that the function manipulates the stack
4994 dynamically: in addition to the static allocation described above, stack
4995 adjustments are made in the body of the function, for example to push/pop
4996 arguments around function calls. If the qualifier @code{bounded} is also
4997 present, the amount of these adjustments is bounded at compile-time and
4998 the second field is an upper bound of the total amount of stack used by
4999 the function. If it is not present, the amount of these adjustments is
5000 not bounded at compile-time and the second field only represents the
5003 @item -fprofile-arcs
5004 @opindex fprofile-arcs
5005 Add code so that program flow @dfn{arcs} are instrumented. During
5006 execution the program records how many times each branch and call is
5007 executed and how many times it is taken or returns. When the compiled
5008 program exits it saves this data to a file called
5009 @file{@var{auxname}.gcda} for each source file. The data may be used for
5010 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5011 test coverage analysis (@option{-ftest-coverage}). Each object file's
5012 @var{auxname} is generated from the name of the output file, if
5013 explicitly specified and it is not the final executable, otherwise it is
5014 the basename of the source file. In both cases any suffix is removed
5015 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5016 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5017 @xref{Cross-profiling}.
5019 @cindex @command{gcov}
5023 This option is used to compile and link code instrumented for coverage
5024 analysis. The option is a synonym for @option{-fprofile-arcs}
5025 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5026 linking). See the documentation for those options for more details.
5031 Compile the source files with @option{-fprofile-arcs} plus optimization
5032 and code generation options. For test coverage analysis, use the
5033 additional @option{-ftest-coverage} option. You do not need to profile
5034 every source file in a program.
5037 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5038 (the latter implies the former).
5041 Run the program on a representative workload to generate the arc profile
5042 information. This may be repeated any number of times. You can run
5043 concurrent instances of your program, and provided that the file system
5044 supports locking, the data files will be correctly updated. Also
5045 @code{fork} calls are detected and correctly handled (double counting
5049 For profile-directed optimizations, compile the source files again with
5050 the same optimization and code generation options plus
5051 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5052 Control Optimization}).
5055 For test coverage analysis, use @command{gcov} to produce human readable
5056 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5057 @command{gcov} documentation for further information.
5061 With @option{-fprofile-arcs}, for each function of your program GCC
5062 creates a program flow graph, then finds a spanning tree for the graph.
5063 Only arcs that are not on the spanning tree have to be instrumented: the
5064 compiler adds code to count the number of times that these arcs are
5065 executed. When an arc is the only exit or only entrance to a block, the
5066 instrumentation code can be added to the block; otherwise, a new basic
5067 block must be created to hold the instrumentation code.
5070 @item -ftest-coverage
5071 @opindex ftest-coverage
5072 Produce a notes file that the @command{gcov} code-coverage utility
5073 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5074 show program coverage. Each source file's note file is called
5075 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5076 above for a description of @var{auxname} and instructions on how to
5077 generate test coverage data. Coverage data will match the source files
5078 more closely, if you do not optimize.
5080 @item -fdbg-cnt-list
5081 @opindex fdbg-cnt-list
5082 Print the name and the counter upper bound for all debug counters.
5085 @item -fdbg-cnt=@var{counter-value-list}
5087 Set the internal debug counter upper bound. @var{counter-value-list}
5088 is a comma-separated list of @var{name}:@var{value} pairs
5089 which sets the upper bound of each debug counter @var{name} to @var{value}.
5090 All debug counters have the initial upper bound of @var{UINT_MAX},
5091 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5092 e.g. With -fdbg-cnt=dce:10,tail_call:0
5093 dbg_cnt(dce) will return true only for first 10 invocations
5095 @itemx -fenable-@var{kind}-@var{pass}
5096 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5100 This is a set of debugging options that are used to explicitly disable/enable
5101 optimization passes. For compiler users, regular options for enabling/disabling
5102 passes should be used instead.
5106 @item -fdisable-ipa-@var{pass}
5107 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5108 statically invoked in the compiler multiple times, the pass name should be
5109 appended with a sequential number starting from 1.
5111 @item -fdisable-rtl-@var{pass}
5112 @item -fdisable-rtl-@var{pass}=@var{range-list}
5113 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5114 statically invoked in the compiler multiple times, the pass name should be
5115 appended with a sequential number starting from 1. @var{range-list} is a comma
5116 seperated list of function ranges or assembler names. Each range is a number
5117 pair seperated by a colon. The range is inclusive in both ends. If the range
5118 is trivial, the number pair can be simplified as a single number. If the
5119 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5120 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5121 function header of a dump file, and the pass names can be dumped by using
5122 option @option{-fdump-passes}.
5124 @item -fdisable-tree-@var{pass}
5125 @item -fdisable-tree-@var{pass}=@var{range-list}
5126 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5129 @item -fenable-ipa-@var{pass}
5130 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5131 statically invoked in the compiler multiple times, the pass name should be
5132 appended with a sequential number starting from 1.
5134 @item -fenable-rtl-@var{pass}
5135 @item -fenable-rtl-@var{pass}=@var{range-list}
5136 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5137 description and examples.
5139 @item -fenable-tree-@var{pass}
5140 @item -fenable-tree-@var{pass}=@var{range-list}
5141 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5142 of option arguments.
5146 # disable ccp1 for all functions
5148 # disable complete unroll for function whose cgraph node uid is 1
5149 -fenable-tree-cunroll=1
5150 # disable gcse2 for functions at the following ranges [1,1],
5151 # [300,400], and [400,1000]
5152 # disable gcse2 for functions foo and foo2
5153 -fdisable-rtl-gcse2=foo,foo2
5154 # disable early inlining
5155 -fdisable-tree-einline
5156 # disable ipa inlining
5157 -fdisable-ipa-inline
5158 # enable tree full unroll
5159 -fenable-tree-unroll
5165 @item -d@var{letters}
5166 @itemx -fdump-rtl-@var{pass}
5168 Says to make debugging dumps during compilation at times specified by
5169 @var{letters}. This is used for debugging the RTL-based passes of the
5170 compiler. The file names for most of the dumps are made by appending
5171 a pass number and a word to the @var{dumpname}, and the files are
5172 created in the directory of the output file. Note that the pass
5173 number is computed statically as passes get registered into the pass
5174 manager. Thus the numbering is not related to the dynamic order of
5175 execution of passes. In particular, a pass installed by a plugin
5176 could have a number over 200 even if it executed quite early.
5177 @var{dumpname} is generated from the name of the output file, if
5178 explicitly specified and it is not an executable, otherwise it is the
5179 basename of the source file. These switches may have different effects
5180 when @option{-E} is used for preprocessing.
5182 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5183 @option{-d} option @var{letters}. Here are the possible
5184 letters for use in @var{pass} and @var{letters}, and their meanings:
5188 @item -fdump-rtl-alignments
5189 @opindex fdump-rtl-alignments
5190 Dump after branch alignments have been computed.
5192 @item -fdump-rtl-asmcons
5193 @opindex fdump-rtl-asmcons
5194 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5196 @item -fdump-rtl-auto_inc_dec
5197 @opindex fdump-rtl-auto_inc_dec
5198 Dump after auto-inc-dec discovery. This pass is only run on
5199 architectures that have auto inc or auto dec instructions.
5201 @item -fdump-rtl-barriers
5202 @opindex fdump-rtl-barriers
5203 Dump after cleaning up the barrier instructions.
5205 @item -fdump-rtl-bbpart
5206 @opindex fdump-rtl-bbpart
5207 Dump after partitioning hot and cold basic blocks.
5209 @item -fdump-rtl-bbro
5210 @opindex fdump-rtl-bbro
5211 Dump after block reordering.
5213 @item -fdump-rtl-btl1
5214 @itemx -fdump-rtl-btl2
5215 @opindex fdump-rtl-btl2
5216 @opindex fdump-rtl-btl2
5217 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5218 after the two branch
5219 target load optimization passes.
5221 @item -fdump-rtl-bypass
5222 @opindex fdump-rtl-bypass
5223 Dump after jump bypassing and control flow optimizations.
5225 @item -fdump-rtl-combine
5226 @opindex fdump-rtl-combine
5227 Dump after the RTL instruction combination pass.
5229 @item -fdump-rtl-compgotos
5230 @opindex fdump-rtl-compgotos
5231 Dump after duplicating the computed gotos.
5233 @item -fdump-rtl-ce1
5234 @itemx -fdump-rtl-ce2
5235 @itemx -fdump-rtl-ce3
5236 @opindex fdump-rtl-ce1
5237 @opindex fdump-rtl-ce2
5238 @opindex fdump-rtl-ce3
5239 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5240 @option{-fdump-rtl-ce3} enable dumping after the three
5241 if conversion passes.
5243 @itemx -fdump-rtl-cprop_hardreg
5244 @opindex fdump-rtl-cprop_hardreg
5245 Dump after hard register copy propagation.
5247 @itemx -fdump-rtl-csa
5248 @opindex fdump-rtl-csa
5249 Dump after combining stack adjustments.
5251 @item -fdump-rtl-cse1
5252 @itemx -fdump-rtl-cse2
5253 @opindex fdump-rtl-cse1
5254 @opindex fdump-rtl-cse2
5255 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5256 the two common sub-expression elimination passes.
5258 @itemx -fdump-rtl-dce
5259 @opindex fdump-rtl-dce
5260 Dump after the standalone dead code elimination passes.
5262 @itemx -fdump-rtl-dbr
5263 @opindex fdump-rtl-dbr
5264 Dump after delayed branch scheduling.
5266 @item -fdump-rtl-dce1
5267 @itemx -fdump-rtl-dce2
5268 @opindex fdump-rtl-dce1
5269 @opindex fdump-rtl-dce2
5270 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5271 the two dead store elimination passes.
5274 @opindex fdump-rtl-eh
5275 Dump after finalization of EH handling code.
5277 @item -fdump-rtl-eh_ranges
5278 @opindex fdump-rtl-eh_ranges
5279 Dump after conversion of EH handling range regions.
5281 @item -fdump-rtl-expand
5282 @opindex fdump-rtl-expand
5283 Dump after RTL generation.
5285 @item -fdump-rtl-fwprop1
5286 @itemx -fdump-rtl-fwprop2
5287 @opindex fdump-rtl-fwprop1
5288 @opindex fdump-rtl-fwprop2
5289 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5290 dumping after the two forward propagation passes.
5292 @item -fdump-rtl-gcse1
5293 @itemx -fdump-rtl-gcse2
5294 @opindex fdump-rtl-gcse1
5295 @opindex fdump-rtl-gcse2
5296 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5297 after global common subexpression elimination.
5299 @item -fdump-rtl-init-regs
5300 @opindex fdump-rtl-init-regs
5301 Dump after the initialization of the registers.
5303 @item -fdump-rtl-initvals
5304 @opindex fdump-rtl-initvals
5305 Dump after the computation of the initial value sets.
5307 @itemx -fdump-rtl-into_cfglayout
5308 @opindex fdump-rtl-into_cfglayout
5309 Dump after converting to cfglayout mode.
5311 @item -fdump-rtl-ira
5312 @opindex fdump-rtl-ira
5313 Dump after iterated register allocation.
5315 @item -fdump-rtl-jump
5316 @opindex fdump-rtl-jump
5317 Dump after the second jump optimization.
5319 @item -fdump-rtl-loop2
5320 @opindex fdump-rtl-loop2
5321 @option{-fdump-rtl-loop2} enables dumping after the rtl
5322 loop optimization passes.
5324 @item -fdump-rtl-mach
5325 @opindex fdump-rtl-mach
5326 Dump after performing the machine dependent reorganization pass, if that
5329 @item -fdump-rtl-mode_sw
5330 @opindex fdump-rtl-mode_sw
5331 Dump after removing redundant mode switches.
5333 @item -fdump-rtl-rnreg
5334 @opindex fdump-rtl-rnreg
5335 Dump after register renumbering.
5337 @itemx -fdump-rtl-outof_cfglayout
5338 @opindex fdump-rtl-outof_cfglayout
5339 Dump after converting from cfglayout mode.
5341 @item -fdump-rtl-peephole2
5342 @opindex fdump-rtl-peephole2
5343 Dump after the peephole pass.
5345 @item -fdump-rtl-postreload
5346 @opindex fdump-rtl-postreload
5347 Dump after post-reload optimizations.
5349 @itemx -fdump-rtl-pro_and_epilogue
5350 @opindex fdump-rtl-pro_and_epilogue
5351 Dump after generating the function pro and epilogues.
5353 @item -fdump-rtl-regmove
5354 @opindex fdump-rtl-regmove
5355 Dump after the register move pass.
5357 @item -fdump-rtl-sched1
5358 @itemx -fdump-rtl-sched2
5359 @opindex fdump-rtl-sched1
5360 @opindex fdump-rtl-sched2
5361 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5362 after the basic block scheduling passes.
5364 @item -fdump-rtl-see
5365 @opindex fdump-rtl-see
5366 Dump after sign extension elimination.
5368 @item -fdump-rtl-seqabstr
5369 @opindex fdump-rtl-seqabstr
5370 Dump after common sequence discovery.
5372 @item -fdump-rtl-shorten
5373 @opindex fdump-rtl-shorten
5374 Dump after shortening branches.
5376 @item -fdump-rtl-sibling
5377 @opindex fdump-rtl-sibling
5378 Dump after sibling call optimizations.
5380 @item -fdump-rtl-split1
5381 @itemx -fdump-rtl-split2
5382 @itemx -fdump-rtl-split3
5383 @itemx -fdump-rtl-split4
5384 @itemx -fdump-rtl-split5
5385 @opindex fdump-rtl-split1
5386 @opindex fdump-rtl-split2
5387 @opindex fdump-rtl-split3
5388 @opindex fdump-rtl-split4
5389 @opindex fdump-rtl-split5
5390 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5391 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5392 @option{-fdump-rtl-split5} enable dumping after five rounds of
5393 instruction splitting.
5395 @item -fdump-rtl-sms
5396 @opindex fdump-rtl-sms
5397 Dump after modulo scheduling. This pass is only run on some
5400 @item -fdump-rtl-stack
5401 @opindex fdump-rtl-stack
5402 Dump after conversion from GCC's "flat register file" registers to the
5403 x87's stack-like registers. This pass is only run on x86 variants.
5405 @item -fdump-rtl-subreg1
5406 @itemx -fdump-rtl-subreg2
5407 @opindex fdump-rtl-subreg1
5408 @opindex fdump-rtl-subreg2
5409 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5410 the two subreg expansion passes.
5412 @item -fdump-rtl-unshare
5413 @opindex fdump-rtl-unshare
5414 Dump after all rtl has been unshared.
5416 @item -fdump-rtl-vartrack
5417 @opindex fdump-rtl-vartrack
5418 Dump after variable tracking.
5420 @item -fdump-rtl-vregs
5421 @opindex fdump-rtl-vregs
5422 Dump after converting virtual registers to hard registers.
5424 @item -fdump-rtl-web
5425 @opindex fdump-rtl-web
5426 Dump after live range splitting.
5428 @item -fdump-rtl-regclass
5429 @itemx -fdump-rtl-subregs_of_mode_init
5430 @itemx -fdump-rtl-subregs_of_mode_finish
5431 @itemx -fdump-rtl-dfinit
5432 @itemx -fdump-rtl-dfinish
5433 @opindex fdump-rtl-regclass
5434 @opindex fdump-rtl-subregs_of_mode_init
5435 @opindex fdump-rtl-subregs_of_mode_finish
5436 @opindex fdump-rtl-dfinit
5437 @opindex fdump-rtl-dfinish
5438 These dumps are defined but always produce empty files.
5440 @item -fdump-rtl-all
5441 @opindex fdump-rtl-all
5442 Produce all the dumps listed above.
5446 Annotate the assembler output with miscellaneous debugging information.
5450 Dump all macro definitions, at the end of preprocessing, in addition to
5455 Produce a core dump whenever an error occurs.
5459 Print statistics on memory usage, at the end of the run, to
5464 Annotate the assembler output with a comment indicating which
5465 pattern and alternative was used. The length of each instruction is
5470 Dump the RTL in the assembler output as a comment before each instruction.
5471 Also turns on @option{-dp} annotation.
5475 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5476 dump a representation of the control flow graph suitable for viewing with VCG
5477 to @file{@var{file}.@var{pass}.vcg}.
5481 Just generate RTL for a function instead of compiling it. Usually used
5482 with @option{-fdump-rtl-expand}.
5486 @opindex fdump-noaddr
5487 When doing debugging dumps, suppress address output. This makes it more
5488 feasible to use diff on debugging dumps for compiler invocations with
5489 different compiler binaries and/or different
5490 text / bss / data / heap / stack / dso start locations.
5492 @item -fdump-unnumbered
5493 @opindex fdump-unnumbered
5494 When doing debugging dumps, suppress instruction numbers and address output.
5495 This makes it more feasible to use diff on debugging dumps for compiler
5496 invocations with different options, in particular with and without
5499 @item -fdump-unnumbered-links
5500 @opindex fdump-unnumbered-links
5501 When doing debugging dumps (see @option{-d} option above), suppress
5502 instruction numbers for the links to the previous and next instructions
5505 @item -fdump-translation-unit @r{(C++ only)}
5506 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5507 @opindex fdump-translation-unit
5508 Dump a representation of the tree structure for the entire translation
5509 unit to a file. The file name is made by appending @file{.tu} to the
5510 source file name, and the file is created in the same directory as the
5511 output file. If the @samp{-@var{options}} form is used, @var{options}
5512 controls the details of the dump as described for the
5513 @option{-fdump-tree} options.
5515 @item -fdump-class-hierarchy @r{(C++ only)}
5516 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5517 @opindex fdump-class-hierarchy
5518 Dump a representation of each class's hierarchy and virtual function
5519 table layout to a file. The file name is made by appending
5520 @file{.class} to the source file name, and the file is created in the
5521 same directory as the output file. If the @samp{-@var{options}} form
5522 is used, @var{options} controls the details of the dump as described
5523 for the @option{-fdump-tree} options.
5525 @item -fdump-ipa-@var{switch}
5527 Control the dumping at various stages of inter-procedural analysis
5528 language tree to a file. The file name is generated by appending a
5529 switch specific suffix to the source file name, and the file is created
5530 in the same directory as the output file. The following dumps are
5535 Enables all inter-procedural analysis dumps.
5538 Dumps information about call-graph optimization, unused function removal,
5539 and inlining decisions.
5542 Dump after function inlining.
5547 @opindex fdump-passes
5548 Dump the list of optimization passes that are turned on and off by
5549 the current command line options.
5551 @item -fdump-statistics-@var{option}
5552 @opindex fdump-statistics
5553 Enable and control dumping of pass statistics in a separate file. The
5554 file name is generated by appending a suffix ending in
5555 @samp{.statistics} to the source file name, and the file is created in
5556 the same directory as the output file. If the @samp{-@var{option}}
5557 form is used, @samp{-stats} will cause counters to be summed over the
5558 whole compilation unit while @samp{-details} will dump every event as
5559 the passes generate them. The default with no option is to sum
5560 counters for each function compiled.
5562 @item -fdump-tree-@var{switch}
5563 @itemx -fdump-tree-@var{switch}-@var{options}
5565 Control the dumping at various stages of processing the intermediate
5566 language tree to a file. The file name is generated by appending a
5567 switch specific suffix to the source file name, and the file is
5568 created in the same directory as the output file. If the
5569 @samp{-@var{options}} form is used, @var{options} is a list of
5570 @samp{-} separated options that control the details of the dump. Not
5571 all options are applicable to all dumps, those which are not
5572 meaningful will be ignored. The following options are available
5576 Print the address of each node. Usually this is not meaningful as it
5577 changes according to the environment and source file. Its primary use
5578 is for tying up a dump file with a debug environment.
5580 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5581 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5582 use working backward from mangled names in the assembly file.
5584 Inhibit dumping of members of a scope or body of a function merely
5585 because that scope has been reached. Only dump such items when they
5586 are directly reachable by some other path. When dumping pretty-printed
5587 trees, this option inhibits dumping the bodies of control structures.
5589 Print a raw representation of the tree. By default, trees are
5590 pretty-printed into a C-like representation.
5592 Enable more detailed dumps (not honored by every dump option).
5594 Enable dumping various statistics about the pass (not honored by every dump
5597 Enable showing basic block boundaries (disabled in raw dumps).
5599 Enable showing virtual operands for every statement.
5601 Enable showing line numbers for statements.
5603 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5605 Enable showing the tree dump for each statement.
5607 Enable showing the EH region number holding each statement.
5609 Enable showing scalar evolution analysis details.
5611 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5612 and @option{lineno}.
5615 The following tree dumps are possible:
5619 @opindex fdump-tree-original
5620 Dump before any tree based optimization, to @file{@var{file}.original}.
5623 @opindex fdump-tree-optimized
5624 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5627 @opindex fdump-tree-gimple
5628 Dump each function before and after the gimplification pass to a file. The
5629 file name is made by appending @file{.gimple} to the source file name.
5632 @opindex fdump-tree-cfg
5633 Dump the control flow graph of each function to a file. The file name is
5634 made by appending @file{.cfg} to the source file name.
5637 @opindex fdump-tree-vcg
5638 Dump the control flow graph of each function to a file in VCG format. The
5639 file name is made by appending @file{.vcg} to the source file name. Note
5640 that if the file contains more than one function, the generated file cannot
5641 be used directly by VCG@. You will need to cut and paste each function's
5642 graph into its own separate file first.
5645 @opindex fdump-tree-ch
5646 Dump each function after copying loop headers. The file name is made by
5647 appending @file{.ch} to the source file name.
5650 @opindex fdump-tree-ssa
5651 Dump SSA related information to a file. The file name is made by appending
5652 @file{.ssa} to the source file name.
5655 @opindex fdump-tree-alias
5656 Dump aliasing information for each function. The file name is made by
5657 appending @file{.alias} to the source file name.
5660 @opindex fdump-tree-ccp
5661 Dump each function after CCP@. The file name is made by appending
5662 @file{.ccp} to the source file name.
5665 @opindex fdump-tree-storeccp
5666 Dump each function after STORE-CCP@. The file name is made by appending
5667 @file{.storeccp} to the source file name.
5670 @opindex fdump-tree-pre
5671 Dump trees after partial redundancy elimination. The file name is made
5672 by appending @file{.pre} to the source file name.
5675 @opindex fdump-tree-fre
5676 Dump trees after full redundancy elimination. The file name is made
5677 by appending @file{.fre} to the source file name.
5680 @opindex fdump-tree-copyprop
5681 Dump trees after copy propagation. The file name is made
5682 by appending @file{.copyprop} to the source file name.
5684 @item store_copyprop
5685 @opindex fdump-tree-store_copyprop
5686 Dump trees after store copy-propagation. The file name is made
5687 by appending @file{.store_copyprop} to the source file name.
5690 @opindex fdump-tree-dce
5691 Dump each function after dead code elimination. The file name is made by
5692 appending @file{.dce} to the source file name.
5695 @opindex fdump-tree-mudflap
5696 Dump each function after adding mudflap instrumentation. The file name is
5697 made by appending @file{.mudflap} to the source file name.
5700 @opindex fdump-tree-sra
5701 Dump each function after performing scalar replacement of aggregates. The
5702 file name is made by appending @file{.sra} to the source file name.
5705 @opindex fdump-tree-sink
5706 Dump each function after performing code sinking. The file name is made
5707 by appending @file{.sink} to the source file name.
5710 @opindex fdump-tree-dom
5711 Dump each function after applying dominator tree optimizations. The file
5712 name is made by appending @file{.dom} to the source file name.
5715 @opindex fdump-tree-dse
5716 Dump each function after applying dead store elimination. The file
5717 name is made by appending @file{.dse} to the source file name.
5720 @opindex fdump-tree-phiopt
5721 Dump each function after optimizing PHI nodes into straightline code. The file
5722 name is made by appending @file{.phiopt} to the source file name.
5725 @opindex fdump-tree-forwprop
5726 Dump each function after forward propagating single use variables. The file
5727 name is made by appending @file{.forwprop} to the source file name.
5730 @opindex fdump-tree-copyrename
5731 Dump each function after applying the copy rename optimization. The file
5732 name is made by appending @file{.copyrename} to the source file name.
5735 @opindex fdump-tree-nrv
5736 Dump each function after applying the named return value optimization on
5737 generic trees. The file name is made by appending @file{.nrv} to the source
5741 @opindex fdump-tree-vect
5742 Dump each function after applying vectorization of loops. The file name is
5743 made by appending @file{.vect} to the source file name.
5746 @opindex fdump-tree-slp
5747 Dump each function after applying vectorization of basic blocks. The file name
5748 is made by appending @file{.slp} to the source file name.
5751 @opindex fdump-tree-vrp
5752 Dump each function after Value Range Propagation (VRP). The file name
5753 is made by appending @file{.vrp} to the source file name.
5756 @opindex fdump-tree-all
5757 Enable all the available tree dumps with the flags provided in this option.
5760 @item -ftree-vectorizer-verbose=@var{n}
5761 @opindex ftree-vectorizer-verbose
5762 This option controls the amount of debugging output the vectorizer prints.
5763 This information is written to standard error, unless
5764 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5765 in which case it is output to the usual dump listing file, @file{.vect}.
5766 For @var{n}=0 no diagnostic information is reported.
5767 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5768 and the total number of loops that got vectorized.
5769 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5770 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5771 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5772 level that @option{-fdump-tree-vect-stats} uses.
5773 Higher verbosity levels mean either more information dumped for each
5774 reported loop, or same amount of information reported for more loops:
5775 if @var{n}=3, vectorizer cost model information is reported.
5776 If @var{n}=4, alignment related information is added to the reports.
5777 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5778 memory access-patterns) is added to the reports.
5779 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5780 that did not pass the first analysis phase (i.e., may not be countable, or
5781 may have complicated control-flow).
5782 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5783 If @var{n}=8, SLP related information is added to the reports.
5784 For @var{n}=9, all the information the vectorizer generates during its
5785 analysis and transformation is reported. This is the same verbosity level
5786 that @option{-fdump-tree-vect-details} uses.
5788 @item -frandom-seed=@var{string}
5789 @opindex frandom-seed
5790 This option provides a seed that GCC uses when it would otherwise use
5791 random numbers. It is used to generate certain symbol names
5792 that have to be different in every compiled file. It is also used to
5793 place unique stamps in coverage data files and the object files that
5794 produce them. You can use the @option{-frandom-seed} option to produce
5795 reproducibly identical object files.
5797 The @var{string} should be different for every file you compile.
5799 @item -fsched-verbose=@var{n}
5800 @opindex fsched-verbose
5801 On targets that use instruction scheduling, this option controls the
5802 amount of debugging output the scheduler prints. This information is
5803 written to standard error, unless @option{-fdump-rtl-sched1} or
5804 @option{-fdump-rtl-sched2} is specified, in which case it is output
5805 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5806 respectively. However for @var{n} greater than nine, the output is
5807 always printed to standard error.
5809 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5810 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5811 For @var{n} greater than one, it also output basic block probabilities,
5812 detailed ready list information and unit/insn info. For @var{n} greater
5813 than two, it includes RTL at abort point, control-flow and regions info.
5814 And for @var{n} over four, @option{-fsched-verbose} also includes
5818 @itemx -save-temps=cwd
5820 Store the usual ``temporary'' intermediate files permanently; place them
5821 in the current directory and name them based on the source file. Thus,
5822 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5823 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5824 preprocessed @file{foo.i} output file even though the compiler now
5825 normally uses an integrated preprocessor.
5827 When used in combination with the @option{-x} command line option,
5828 @option{-save-temps} is sensible enough to avoid over writing an
5829 input source file with the same extension as an intermediate file.
5830 The corresponding intermediate file may be obtained by renaming the
5831 source file before using @option{-save-temps}.
5833 If you invoke GCC in parallel, compiling several different source
5834 files that share a common base name in different subdirectories or the
5835 same source file compiled for multiple output destinations, it is
5836 likely that the different parallel compilers will interfere with each
5837 other, and overwrite the temporary files. For instance:
5840 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5841 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5844 may result in @file{foo.i} and @file{foo.o} being written to
5845 simultaneously by both compilers.
5847 @item -save-temps=obj
5848 @opindex save-temps=obj
5849 Store the usual ``temporary'' intermediate files permanently. If the
5850 @option{-o} option is used, the temporary files are based on the
5851 object file. If the @option{-o} option is not used, the
5852 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5857 gcc -save-temps=obj -c foo.c
5858 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5859 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5862 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5863 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5864 @file{dir2/yfoobar.o}.
5866 @item -time@r{[}=@var{file}@r{]}
5868 Report the CPU time taken by each subprocess in the compilation
5869 sequence. For C source files, this is the compiler proper and assembler
5870 (plus the linker if linking is done).
5872 Without the specification of an output file, the output looks like this:
5879 The first number on each line is the ``user time'', that is time spent
5880 executing the program itself. The second number is ``system time'',
5881 time spent executing operating system routines on behalf of the program.
5882 Both numbers are in seconds.
5884 With the specification of an output file, the output is appended to the
5885 named file, and it looks like this:
5888 0.12 0.01 cc1 @var{options}
5889 0.00 0.01 as @var{options}
5892 The ``user time'' and the ``system time'' are moved before the program
5893 name, and the options passed to the program are displayed, so that one
5894 can later tell what file was being compiled, and with which options.
5896 @item -fvar-tracking
5897 @opindex fvar-tracking
5898 Run variable tracking pass. It computes where variables are stored at each
5899 position in code. Better debugging information is then generated
5900 (if the debugging information format supports this information).
5902 It is enabled by default when compiling with optimization (@option{-Os},
5903 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5904 the debug info format supports it.
5906 @item -fvar-tracking-assignments
5907 @opindex fvar-tracking-assignments
5908 @opindex fno-var-tracking-assignments
5909 Annotate assignments to user variables early in the compilation and
5910 attempt to carry the annotations over throughout the compilation all the
5911 way to the end, in an attempt to improve debug information while
5912 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5914 It can be enabled even if var-tracking is disabled, in which case
5915 annotations will be created and maintained, but discarded at the end.
5917 @item -fvar-tracking-assignments-toggle
5918 @opindex fvar-tracking-assignments-toggle
5919 @opindex fno-var-tracking-assignments-toggle
5920 Toggle @option{-fvar-tracking-assignments}, in the same way that
5921 @option{-gtoggle} toggles @option{-g}.
5923 @item -print-file-name=@var{library}
5924 @opindex print-file-name
5925 Print the full absolute name of the library file @var{library} that
5926 would be used when linking---and don't do anything else. With this
5927 option, GCC does not compile or link anything; it just prints the
5930 @item -print-multi-directory
5931 @opindex print-multi-directory
5932 Print the directory name corresponding to the multilib selected by any
5933 other switches present in the command line. This directory is supposed
5934 to exist in @env{GCC_EXEC_PREFIX}.
5936 @item -print-multi-lib
5937 @opindex print-multi-lib
5938 Print the mapping from multilib directory names to compiler switches
5939 that enable them. The directory name is separated from the switches by
5940 @samp{;}, and each switch starts with an @samp{@@} instead of the
5941 @samp{-}, without spaces between multiple switches. This is supposed to
5942 ease shell-processing.
5944 @item -print-multi-os-directory
5945 @opindex print-multi-os-directory
5946 Print the path to OS libraries for the selected
5947 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5948 present in the @file{lib} subdirectory and no multilibs are used, this is
5949 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5950 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5951 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5952 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5954 @item -print-prog-name=@var{program}
5955 @opindex print-prog-name
5956 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5958 @item -print-libgcc-file-name
5959 @opindex print-libgcc-file-name
5960 Same as @option{-print-file-name=libgcc.a}.
5962 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5963 but you do want to link with @file{libgcc.a}. You can do
5966 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5969 @item -print-search-dirs
5970 @opindex print-search-dirs
5971 Print the name of the configured installation directory and a list of
5972 program and library directories @command{gcc} will search---and don't do anything else.
5974 This is useful when @command{gcc} prints the error message
5975 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5976 To resolve this you either need to put @file{cpp0} and the other compiler
5977 components where @command{gcc} expects to find them, or you can set the environment
5978 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5979 Don't forget the trailing @samp{/}.
5980 @xref{Environment Variables}.
5982 @item -print-sysroot
5983 @opindex print-sysroot
5984 Print the target sysroot directory that will be used during
5985 compilation. This is the target sysroot specified either at configure
5986 time or using the @option{--sysroot} option, possibly with an extra
5987 suffix that depends on compilation options. If no target sysroot is
5988 specified, the option prints nothing.
5990 @item -print-sysroot-headers-suffix
5991 @opindex print-sysroot-headers-suffix
5992 Print the suffix added to the target sysroot when searching for
5993 headers, or give an error if the compiler is not configured with such
5994 a suffix---and don't do anything else.
5997 @opindex dumpmachine
5998 Print the compiler's target machine (for example,
5999 @samp{i686-pc-linux-gnu})---and don't do anything else.
6002 @opindex dumpversion
6003 Print the compiler version (for example, @samp{3.0})---and don't do
6008 Print the compiler's built-in specs---and don't do anything else. (This
6009 is used when GCC itself is being built.) @xref{Spec Files}.
6011 @item -feliminate-unused-debug-types
6012 @opindex feliminate-unused-debug-types
6013 Normally, when producing DWARF2 output, GCC will emit debugging
6014 information for all types declared in a compilation
6015 unit, regardless of whether or not they are actually used
6016 in that compilation unit. Sometimes this is useful, such as
6017 if, in the debugger, you want to cast a value to a type that is
6018 not actually used in your program (but is declared). More often,
6019 however, this results in a significant amount of wasted space.
6020 With this option, GCC will avoid producing debug symbol output
6021 for types that are nowhere used in the source file being compiled.
6024 @node Optimize Options
6025 @section Options That Control Optimization
6026 @cindex optimize options
6027 @cindex options, optimization
6029 These options control various sorts of optimizations.
6031 Without any optimization option, the compiler's goal is to reduce the
6032 cost of compilation and to make debugging produce the expected
6033 results. Statements are independent: if you stop the program with a
6034 breakpoint between statements, you can then assign a new value to any
6035 variable or change the program counter to any other statement in the
6036 function and get exactly the results you would expect from the source
6039 Turning on optimization flags makes the compiler attempt to improve
6040 the performance and/or code size at the expense of compilation time
6041 and possibly the ability to debug the program.
6043 The compiler performs optimization based on the knowledge it has of the
6044 program. Compiling multiple files at once to a single output file mode allows
6045 the compiler to use information gained from all of the files when compiling
6048 Not all optimizations are controlled directly by a flag. Only
6049 optimizations that have a flag are listed in this section.
6051 Most optimizations are only enabled if an @option{-O} level is set on
6052 the command line. Otherwise they are disabled, even if individual
6053 optimization flags are specified.
6055 Depending on the target and how GCC was configured, a slightly different
6056 set of optimizations may be enabled at each @option{-O} level than
6057 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6058 to find out the exact set of optimizations that are enabled at each level.
6059 @xref{Overall Options}, for examples.
6066 Optimize. Optimizing compilation takes somewhat more time, and a lot
6067 more memory for a large function.
6069 With @option{-O}, the compiler tries to reduce code size and execution
6070 time, without performing any optimizations that take a great deal of
6073 @option{-O} turns on the following optimization flags:
6077 -fcprop-registers @gol
6080 -fdelayed-branch @gol
6082 -fguess-branch-probability @gol
6083 -fif-conversion2 @gol
6084 -fif-conversion @gol
6085 -fipa-pure-const @gol
6087 -fipa-reference @gol
6089 -fsplit-wide-types @gol
6091 -ftree-builtin-call-dce @gol
6094 -ftree-copyrename @gol
6096 -ftree-dominator-opts @gol
6098 -ftree-forwprop @gol
6106 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6107 where doing so does not interfere with debugging.
6111 Optimize even more. GCC performs nearly all supported optimizations
6112 that do not involve a space-speed tradeoff.
6113 As compared to @option{-O}, this option increases both compilation time
6114 and the performance of the generated code.
6116 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6117 also turns on the following optimization flags:
6118 @gccoptlist{-fthread-jumps @gol
6119 -falign-functions -falign-jumps @gol
6120 -falign-loops -falign-labels @gol
6123 -fcse-follow-jumps -fcse-skip-blocks @gol
6124 -fdelete-null-pointer-checks @gol
6126 -fexpensive-optimizations @gol
6127 -fgcse -fgcse-lm @gol
6128 -finline-small-functions @gol
6129 -findirect-inlining @gol
6131 -foptimize-sibling-calls @gol
6132 -fpartial-inlining @gol
6135 -freorder-blocks -freorder-functions @gol
6136 -frerun-cse-after-loop @gol
6137 -fsched-interblock -fsched-spec @gol
6138 -fschedule-insns -fschedule-insns2 @gol
6139 -fstrict-aliasing -fstrict-overflow @gol
6140 -ftree-switch-conversion @gol
6144 Please note the warning under @option{-fgcse} about
6145 invoking @option{-O2} on programs that use computed gotos.
6149 Optimize yet more. @option{-O3} turns on all optimizations specified
6150 by @option{-O2} and also turns on the @option{-finline-functions},
6151 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6152 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6153 @option{-fipa-cp-clone} options.
6157 Reduce compilation time and make debugging produce the expected
6158 results. This is the default.
6162 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6163 do not typically increase code size. It also performs further
6164 optimizations designed to reduce code size.
6166 @option{-Os} disables the following optimization flags:
6167 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6168 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6169 -fprefetch-loop-arrays -ftree-vect-loop-version}
6173 Disregard strict standards compliance. @option{-Ofast} enables all
6174 @option{-O3} optimizations. It also enables optimizations that are not
6175 valid for all standard compliant programs.
6176 It turns on @option{-ffast-math} and the Fortran-specific
6177 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6179 If you use multiple @option{-O} options, with or without level numbers,
6180 the last such option is the one that is effective.
6183 Options of the form @option{-f@var{flag}} specify machine-independent
6184 flags. Most flags have both positive and negative forms; the negative
6185 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6186 below, only one of the forms is listed---the one you typically will
6187 use. You can figure out the other form by either removing @samp{no-}
6190 The following options control specific optimizations. They are either
6191 activated by @option{-O} options or are related to ones that are. You
6192 can use the following flags in the rare cases when ``fine-tuning'' of
6193 optimizations to be performed is desired.
6196 @item -fno-default-inline
6197 @opindex fno-default-inline
6198 Do not make member functions inline by default merely because they are
6199 defined inside the class scope (C++ only). Otherwise, when you specify
6200 @w{@option{-O}}, member functions defined inside class scope are compiled
6201 inline by default; i.e., you don't need to add @samp{inline} in front of
6202 the member function name.
6204 @item -fno-defer-pop
6205 @opindex fno-defer-pop
6206 Always pop the arguments to each function call as soon as that function
6207 returns. For machines which must pop arguments after a function call,
6208 the compiler normally lets arguments accumulate on the stack for several
6209 function calls and pops them all at once.
6211 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6213 @item -fforward-propagate
6214 @opindex fforward-propagate
6215 Perform a forward propagation pass on RTL@. The pass tries to combine two
6216 instructions and checks if the result can be simplified. If loop unrolling
6217 is active, two passes are performed and the second is scheduled after
6220 This option is enabled by default at optimization levels @option{-O},
6221 @option{-O2}, @option{-O3}, @option{-Os}.
6223 @item -ffp-contract=@var{style}
6224 @opindex ffp-contract
6225 @option{-ffp-contract=off} disables floating-point expression contraction.
6226 @option{-ffp-contract=fast} enables floating-point expression contraction
6227 such as forming of fused multiply-add operations if the target has
6228 native support for them.
6229 @option{-ffp-contract=on} enables floating-point expression contraction
6230 if allowed by the language standard. This is currently not implemented
6231 and treated equal to @option{-ffp-contract=off}.
6233 The default is @option{-ffp-contract=fast}.
6235 @item -fomit-frame-pointer
6236 @opindex fomit-frame-pointer
6237 Don't keep the frame pointer in a register for functions that
6238 don't need one. This avoids the instructions to save, set up and
6239 restore frame pointers; it also makes an extra register available
6240 in many functions. @strong{It also makes debugging impossible on
6243 On some machines, such as the VAX, this flag has no effect, because
6244 the standard calling sequence automatically handles the frame pointer
6245 and nothing is saved by pretending it doesn't exist. The
6246 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6247 whether a target machine supports this flag. @xref{Registers,,Register
6248 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6250 Starting with GCC version 4.6, the default setting (when not optimizing for
6251 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6252 @option{-fomit-frame-pointer}. The default can be reverted to
6253 @option{-fno-omit-frame-pointer} by configuring GCC with the
6254 @option{--enable-frame-pointer} configure option.
6256 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6258 @item -foptimize-sibling-calls
6259 @opindex foptimize-sibling-calls
6260 Optimize sibling and tail recursive calls.
6262 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6266 Don't pay attention to the @code{inline} keyword. Normally this option
6267 is used to keep the compiler from expanding any functions inline.
6268 Note that if you are not optimizing, no functions can be expanded inline.
6270 @item -finline-small-functions
6271 @opindex finline-small-functions
6272 Integrate functions into their callers when their body is smaller than expected
6273 function call code (so overall size of program gets smaller). The compiler
6274 heuristically decides which functions are simple enough to be worth integrating
6277 Enabled at level @option{-O2}.
6279 @item -findirect-inlining
6280 @opindex findirect-inlining
6281 Inline also indirect calls that are discovered to be known at compile
6282 time thanks to previous inlining. This option has any effect only
6283 when inlining itself is turned on by the @option{-finline-functions}
6284 or @option{-finline-small-functions} options.
6286 Enabled at level @option{-O2}.
6288 @item -finline-functions
6289 @opindex finline-functions
6290 Integrate all simple functions into their callers. The compiler
6291 heuristically decides which functions are simple enough to be worth
6292 integrating in this way.
6294 If all calls to a given function are integrated, and the function is
6295 declared @code{static}, then the function is normally not output as
6296 assembler code in its own right.
6298 Enabled at level @option{-O3}.
6300 @item -finline-functions-called-once
6301 @opindex finline-functions-called-once
6302 Consider all @code{static} functions called once for inlining into their
6303 caller even if they are not marked @code{inline}. If a call to a given
6304 function is integrated, then the function is not output as assembler code
6307 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6309 @item -fearly-inlining
6310 @opindex fearly-inlining
6311 Inline functions marked by @code{always_inline} and functions whose body seems
6312 smaller than the function call overhead early before doing
6313 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6314 makes profiling significantly cheaper and usually inlining faster on programs
6315 having large chains of nested wrapper functions.
6321 Perform interprocedural scalar replacement of aggregates, removal of
6322 unused parameters and replacement of parameters passed by reference
6323 by parameters passed by value.
6325 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6327 @item -finline-limit=@var{n}
6328 @opindex finline-limit
6329 By default, GCC limits the size of functions that can be inlined. This flag
6330 allows coarse control of this limit. @var{n} is the size of functions that
6331 can be inlined in number of pseudo instructions.
6333 Inlining is actually controlled by a number of parameters, which may be
6334 specified individually by using @option{--param @var{name}=@var{value}}.
6335 The @option{-finline-limit=@var{n}} option sets some of these parameters
6339 @item max-inline-insns-single
6340 is set to @var{n}/2.
6341 @item max-inline-insns-auto
6342 is set to @var{n}/2.
6345 See below for a documentation of the individual
6346 parameters controlling inlining and for the defaults of these parameters.
6348 @emph{Note:} there may be no value to @option{-finline-limit} that results
6349 in default behavior.
6351 @emph{Note:} pseudo instruction represents, in this particular context, an
6352 abstract measurement of function's size. In no way does it represent a count
6353 of assembly instructions and as such its exact meaning might change from one
6354 release to an another.
6356 @item -fno-keep-inline-dllexport
6357 @opindex -fno-keep-inline-dllexport
6358 This is a more fine-grained version of @option{-fkeep-inline-functions},
6359 which applies only to functions that are declared using the @code{dllexport}
6360 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6363 @item -fkeep-inline-functions
6364 @opindex fkeep-inline-functions
6365 In C, emit @code{static} functions that are declared @code{inline}
6366 into the object file, even if the function has been inlined into all
6367 of its callers. This switch does not affect functions using the
6368 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6369 inline functions into the object file.
6371 @item -fkeep-static-consts
6372 @opindex fkeep-static-consts
6373 Emit variables declared @code{static const} when optimization isn't turned
6374 on, even if the variables aren't referenced.
6376 GCC enables this option by default. If you want to force the compiler to
6377 check if the variable was referenced, regardless of whether or not
6378 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6380 @item -fmerge-constants
6381 @opindex fmerge-constants
6382 Attempt to merge identical constants (string constants and floating point
6383 constants) across compilation units.
6385 This option is the default for optimized compilation if the assembler and
6386 linker support it. Use @option{-fno-merge-constants} to inhibit this
6389 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6391 @item -fmerge-all-constants
6392 @opindex fmerge-all-constants
6393 Attempt to merge identical constants and identical variables.
6395 This option implies @option{-fmerge-constants}. In addition to
6396 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6397 arrays or initialized constant variables with integral or floating point
6398 types. Languages like C or C++ require each variable, including multiple
6399 instances of the same variable in recursive calls, to have distinct locations,
6400 so using this option will result in non-conforming
6403 @item -fmodulo-sched
6404 @opindex fmodulo-sched
6405 Perform swing modulo scheduling immediately before the first scheduling
6406 pass. This pass looks at innermost loops and reorders their
6407 instructions by overlapping different iterations.
6409 @item -fmodulo-sched-allow-regmoves
6410 @opindex fmodulo-sched-allow-regmoves
6411 Perform more aggressive SMS based modulo scheduling with register moves
6412 allowed. By setting this flag certain anti-dependences edges will be
6413 deleted which will trigger the generation of reg-moves based on the
6414 life-range analysis. This option is effective only with
6415 @option{-fmodulo-sched} enabled.
6417 @item -fno-branch-count-reg
6418 @opindex fno-branch-count-reg
6419 Do not use ``decrement and branch'' instructions on a count register,
6420 but instead generate a sequence of instructions that decrement a
6421 register, compare it against zero, then branch based upon the result.
6422 This option is only meaningful on architectures that support such
6423 instructions, which include x86, PowerPC, IA-64 and S/390.
6425 The default is @option{-fbranch-count-reg}.
6427 @item -fno-function-cse
6428 @opindex fno-function-cse
6429 Do not put function addresses in registers; make each instruction that
6430 calls a constant function contain the function's address explicitly.
6432 This option results in less efficient code, but some strange hacks
6433 that alter the assembler output may be confused by the optimizations
6434 performed when this option is not used.
6436 The default is @option{-ffunction-cse}
6438 @item -fno-zero-initialized-in-bss
6439 @opindex fno-zero-initialized-in-bss
6440 If the target supports a BSS section, GCC by default puts variables that
6441 are initialized to zero into BSS@. This can save space in the resulting
6444 This option turns off this behavior because some programs explicitly
6445 rely on variables going to the data section. E.g., so that the
6446 resulting executable can find the beginning of that section and/or make
6447 assumptions based on that.
6449 The default is @option{-fzero-initialized-in-bss}.
6451 @item -fmudflap -fmudflapth -fmudflapir
6455 @cindex bounds checking
6457 For front-ends that support it (C and C++), instrument all risky
6458 pointer/array dereferencing operations, some standard library
6459 string/heap functions, and some other associated constructs with
6460 range/validity tests. Modules so instrumented should be immune to
6461 buffer overflows, invalid heap use, and some other classes of C/C++
6462 programming errors. The instrumentation relies on a separate runtime
6463 library (@file{libmudflap}), which will be linked into a program if
6464 @option{-fmudflap} is given at link time. Run-time behavior of the
6465 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6466 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6469 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6470 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6471 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6472 instrumentation should ignore pointer reads. This produces less
6473 instrumentation (and therefore faster execution) and still provides
6474 some protection against outright memory corrupting writes, but allows
6475 erroneously read data to propagate within a program.
6477 @item -fthread-jumps
6478 @opindex fthread-jumps
6479 Perform optimizations where we check to see if a jump branches to a
6480 location where another comparison subsumed by the first is found. If
6481 so, the first branch is redirected to either the destination of the
6482 second branch or a point immediately following it, depending on whether
6483 the condition is known to be true or false.
6485 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6487 @item -fsplit-wide-types
6488 @opindex fsplit-wide-types
6489 When using a type that occupies multiple registers, such as @code{long
6490 long} on a 32-bit system, split the registers apart and allocate them
6491 independently. This normally generates better code for those types,
6492 but may make debugging more difficult.
6494 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6497 @item -fcse-follow-jumps
6498 @opindex fcse-follow-jumps
6499 In common subexpression elimination (CSE), scan through jump instructions
6500 when the target of the jump is not reached by any other path. For
6501 example, when CSE encounters an @code{if} statement with an
6502 @code{else} clause, CSE will follow the jump when the condition
6505 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6507 @item -fcse-skip-blocks
6508 @opindex fcse-skip-blocks
6509 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6510 follow jumps which conditionally skip over blocks. When CSE
6511 encounters a simple @code{if} statement with no else clause,
6512 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6513 body of the @code{if}.
6515 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6517 @item -frerun-cse-after-loop
6518 @opindex frerun-cse-after-loop
6519 Re-run common subexpression elimination after loop optimizations has been
6522 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6526 Perform a global common subexpression elimination pass.
6527 This pass also performs global constant and copy propagation.
6529 @emph{Note:} When compiling a program using computed gotos, a GCC
6530 extension, you may get better runtime performance if you disable
6531 the global common subexpression elimination pass by adding
6532 @option{-fno-gcse} to the command line.
6534 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6538 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6539 attempt to move loads which are only killed by stores into themselves. This
6540 allows a loop containing a load/store sequence to be changed to a load outside
6541 the loop, and a copy/store within the loop.
6543 Enabled by default when gcse is enabled.
6547 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6548 global common subexpression elimination. This pass will attempt to move
6549 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6550 loops containing a load/store sequence can be changed to a load before
6551 the loop and a store after the loop.
6553 Not enabled at any optimization level.
6557 When @option{-fgcse-las} is enabled, the global common subexpression
6558 elimination pass eliminates redundant loads that come after stores to the
6559 same memory location (both partial and full redundancies).
6561 Not enabled at any optimization level.
6563 @item -fgcse-after-reload
6564 @opindex fgcse-after-reload
6565 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6566 pass is performed after reload. The purpose of this pass is to cleanup
6569 @item -funsafe-loop-optimizations
6570 @opindex funsafe-loop-optimizations
6571 If given, the loop optimizer will assume that loop indices do not
6572 overflow, and that the loops with nontrivial exit condition are not
6573 infinite. This enables a wider range of loop optimizations even if
6574 the loop optimizer itself cannot prove that these assumptions are valid.
6575 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6576 if it finds this kind of loop.
6578 @item -fcrossjumping
6579 @opindex fcrossjumping
6580 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6581 resulting code may or may not perform better than without cross-jumping.
6583 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6585 @item -fauto-inc-dec
6586 @opindex fauto-inc-dec
6587 Combine increments or decrements of addresses with memory accesses.
6588 This pass is always skipped on architectures that do not have
6589 instructions to support this. Enabled by default at @option{-O} and
6590 higher on architectures that support this.
6594 Perform dead code elimination (DCE) on RTL@.
6595 Enabled by default at @option{-O} and higher.
6599 Perform dead store elimination (DSE) on RTL@.
6600 Enabled by default at @option{-O} and higher.
6602 @item -fif-conversion
6603 @opindex fif-conversion
6604 Attempt to transform conditional jumps into branch-less equivalents. This
6605 include use of conditional moves, min, max, set flags and abs instructions, and
6606 some tricks doable by standard arithmetics. The use of conditional execution
6607 on chips where it is available is controlled by @code{if-conversion2}.
6609 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6611 @item -fif-conversion2
6612 @opindex fif-conversion2
6613 Use conditional execution (where available) to transform conditional jumps into
6614 branch-less equivalents.
6616 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6618 @item -fdelete-null-pointer-checks
6619 @opindex fdelete-null-pointer-checks
6620 Assume that programs cannot safely dereference null pointers, and that
6621 no code or data element resides there. This enables simple constant
6622 folding optimizations at all optimization levels. In addition, other
6623 optimization passes in GCC use this flag to control global dataflow
6624 analyses that eliminate useless checks for null pointers; these assume
6625 that if a pointer is checked after it has already been dereferenced,
6628 Note however that in some environments this assumption is not true.
6629 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6630 for programs which depend on that behavior.
6632 Some targets, especially embedded ones, disable this option at all levels.
6633 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6634 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6635 are enabled independently at different optimization levels.
6637 @item -fdevirtualize
6638 @opindex fdevirtualize
6639 Attempt to convert calls to virtual functions to direct calls. This
6640 is done both within a procedure and interprocedurally as part of
6641 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6642 propagation (@option{-fipa-cp}).
6643 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6645 @item -fexpensive-optimizations
6646 @opindex fexpensive-optimizations
6647 Perform a number of minor optimizations that are relatively expensive.
6649 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6651 @item -foptimize-register-move
6653 @opindex foptimize-register-move
6655 Attempt to reassign register numbers in move instructions and as
6656 operands of other simple instructions in order to maximize the amount of
6657 register tying. This is especially helpful on machines with two-operand
6660 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6663 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6665 @item -fira-algorithm=@var{algorithm}
6666 Use specified coloring algorithm for the integrated register
6667 allocator. The @var{algorithm} argument should be @code{priority} or
6668 @code{CB}. The first algorithm specifies Chow's priority coloring,
6669 the second one specifies Chaitin-Briggs coloring. The second
6670 algorithm can be unimplemented for some architectures. If it is
6671 implemented, it is the default because Chaitin-Briggs coloring as a
6672 rule generates a better code.
6674 @item -fira-region=@var{region}
6675 Use specified regions for the integrated register allocator. The
6676 @var{region} argument should be one of @code{all}, @code{mixed}, or
6677 @code{one}. The first value means using all loops as register
6678 allocation regions, the second value which is the default means using
6679 all loops except for loops with small register pressure as the
6680 regions, and third one means using all function as a single region.
6681 The first value can give best result for machines with small size and
6682 irregular register set, the third one results in faster and generates
6683 decent code and the smallest size code, and the default value usually
6684 give the best results in most cases and for most architectures.
6686 @item -fira-loop-pressure
6687 @opindex fira-loop-pressure
6688 Use IRA to evaluate register pressure in loops for decision to move
6689 loop invariants. Usage of this option usually results in generation
6690 of faster and smaller code on machines with big register files (>= 32
6691 registers) but it can slow compiler down.
6693 This option is enabled at level @option{-O3} for some targets.
6695 @item -fno-ira-share-save-slots
6696 @opindex fno-ira-share-save-slots
6697 Switch off sharing stack slots used for saving call used hard
6698 registers living through a call. Each hard register will get a
6699 separate stack slot and as a result function stack frame will be
6702 @item -fno-ira-share-spill-slots
6703 @opindex fno-ira-share-spill-slots
6704 Switch off sharing stack slots allocated for pseudo-registers. Each
6705 pseudo-register which did not get a hard register will get a separate
6706 stack slot and as a result function stack frame will be bigger.
6708 @item -fira-verbose=@var{n}
6709 @opindex fira-verbose
6710 Set up how verbose dump file for the integrated register allocator
6711 will be. Default value is 5. If the value is greater or equal to 10,
6712 the dump file will be stderr as if the value were @var{n} minus 10.
6714 @item -fdelayed-branch
6715 @opindex fdelayed-branch
6716 If supported for the target machine, attempt to reorder instructions
6717 to exploit instruction slots available after delayed branch
6720 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6722 @item -fschedule-insns
6723 @opindex fschedule-insns
6724 If supported for the target machine, attempt to reorder instructions to
6725 eliminate execution stalls due to required data being unavailable. This
6726 helps machines that have slow floating point or memory load instructions
6727 by allowing other instructions to be issued until the result of the load
6728 or floating point instruction is required.
6730 Enabled at levels @option{-O2}, @option{-O3}.
6732 @item -fschedule-insns2
6733 @opindex fschedule-insns2
6734 Similar to @option{-fschedule-insns}, but requests an additional pass of
6735 instruction scheduling after register allocation has been done. This is
6736 especially useful on machines with a relatively small number of
6737 registers and where memory load instructions take more than one cycle.
6739 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6741 @item -fno-sched-interblock
6742 @opindex fno-sched-interblock
6743 Don't schedule instructions across basic blocks. This is normally
6744 enabled by default when scheduling before register allocation, i.e.@:
6745 with @option{-fschedule-insns} or at @option{-O2} or higher.
6747 @item -fno-sched-spec
6748 @opindex fno-sched-spec
6749 Don't allow speculative motion of non-load instructions. This is normally
6750 enabled by default when scheduling before register allocation, i.e.@:
6751 with @option{-fschedule-insns} or at @option{-O2} or higher.
6753 @item -fsched-pressure
6754 @opindex fsched-pressure
6755 Enable register pressure sensitive insn scheduling before the register
6756 allocation. This only makes sense when scheduling before register
6757 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6758 @option{-O2} or higher. Usage of this option can improve the
6759 generated code and decrease its size by preventing register pressure
6760 increase above the number of available hard registers and as a
6761 consequence register spills in the register allocation.
6763 @item -fsched-spec-load
6764 @opindex fsched-spec-load
6765 Allow speculative motion of some load instructions. This only makes
6766 sense when scheduling before register allocation, i.e.@: with
6767 @option{-fschedule-insns} or at @option{-O2} or higher.
6769 @item -fsched-spec-load-dangerous
6770 @opindex fsched-spec-load-dangerous
6771 Allow speculative motion of more load instructions. This only makes
6772 sense when scheduling before register allocation, i.e.@: with
6773 @option{-fschedule-insns} or at @option{-O2} or higher.
6775 @item -fsched-stalled-insns
6776 @itemx -fsched-stalled-insns=@var{n}
6777 @opindex fsched-stalled-insns
6778 Define how many insns (if any) can be moved prematurely from the queue
6779 of stalled insns into the ready list, during the second scheduling pass.
6780 @option{-fno-sched-stalled-insns} means that no insns will be moved
6781 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6782 on how many queued insns can be moved prematurely.
6783 @option{-fsched-stalled-insns} without a value is equivalent to
6784 @option{-fsched-stalled-insns=1}.
6786 @item -fsched-stalled-insns-dep
6787 @itemx -fsched-stalled-insns-dep=@var{n}
6788 @opindex fsched-stalled-insns-dep
6789 Define how many insn groups (cycles) will be examined for a dependency
6790 on a stalled insn that is candidate for premature removal from the queue
6791 of stalled insns. This has an effect only during the second scheduling pass,
6792 and only if @option{-fsched-stalled-insns} is used.
6793 @option{-fno-sched-stalled-insns-dep} is equivalent to
6794 @option{-fsched-stalled-insns-dep=0}.
6795 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6796 @option{-fsched-stalled-insns-dep=1}.
6798 @item -fsched2-use-superblocks
6799 @opindex fsched2-use-superblocks
6800 When scheduling after register allocation, do use superblock scheduling
6801 algorithm. Superblock scheduling allows motion across basic block boundaries
6802 resulting on faster schedules. This option is experimental, as not all machine
6803 descriptions used by GCC model the CPU closely enough to avoid unreliable
6804 results from the algorithm.
6806 This only makes sense when scheduling after register allocation, i.e.@: with
6807 @option{-fschedule-insns2} or at @option{-O2} or higher.
6809 @item -fsched-group-heuristic
6810 @opindex fsched-group-heuristic
6811 Enable the group heuristic in the scheduler. This heuristic favors
6812 the instruction that belongs to a schedule group. This is enabled
6813 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6814 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6816 @item -fsched-critical-path-heuristic
6817 @opindex fsched-critical-path-heuristic
6818 Enable the critical-path heuristic in the scheduler. This heuristic favors
6819 instructions on the critical path. This is enabled by default when
6820 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6821 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6823 @item -fsched-spec-insn-heuristic
6824 @opindex fsched-spec-insn-heuristic
6825 Enable the speculative instruction heuristic in the scheduler. This
6826 heuristic favors speculative instructions with greater dependency weakness.
6827 This is enabled by default when scheduling is enabled, i.e.@:
6828 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6829 or at @option{-O2} or higher.
6831 @item -fsched-rank-heuristic
6832 @opindex fsched-rank-heuristic
6833 Enable the rank heuristic in the scheduler. This heuristic favors
6834 the instruction belonging to a basic block with greater size or frequency.
6835 This is enabled by default when scheduling is enabled, i.e.@:
6836 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6837 at @option{-O2} or higher.
6839 @item -fsched-last-insn-heuristic
6840 @opindex fsched-last-insn-heuristic
6841 Enable the last-instruction heuristic in the scheduler. This heuristic
6842 favors the instruction that is less dependent on the last instruction
6843 scheduled. This is enabled by default when scheduling is enabled,
6844 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6845 at @option{-O2} or higher.
6847 @item -fsched-dep-count-heuristic
6848 @opindex fsched-dep-count-heuristic
6849 Enable the dependent-count heuristic in the scheduler. This heuristic
6850 favors the instruction that has more instructions depending on it.
6851 This is enabled by default when scheduling is enabled, i.e.@:
6852 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6853 at @option{-O2} or higher.
6855 @item -freschedule-modulo-scheduled-loops
6856 @opindex freschedule-modulo-scheduled-loops
6857 The modulo scheduling comes before the traditional scheduling, if a loop
6858 was modulo scheduled we may want to prevent the later scheduling passes
6859 from changing its schedule, we use this option to control that.
6861 @item -fselective-scheduling
6862 @opindex fselective-scheduling
6863 Schedule instructions using selective scheduling algorithm. Selective
6864 scheduling runs instead of the first scheduler pass.
6866 @item -fselective-scheduling2
6867 @opindex fselective-scheduling2
6868 Schedule instructions using selective scheduling algorithm. Selective
6869 scheduling runs instead of the second scheduler pass.
6871 @item -fsel-sched-pipelining
6872 @opindex fsel-sched-pipelining
6873 Enable software pipelining of innermost loops during selective scheduling.
6874 This option has no effect until one of @option{-fselective-scheduling} or
6875 @option{-fselective-scheduling2} is turned on.
6877 @item -fsel-sched-pipelining-outer-loops
6878 @opindex fsel-sched-pipelining-outer-loops
6879 When pipelining loops during selective scheduling, also pipeline outer loops.
6880 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6882 @item -fcaller-saves
6883 @opindex fcaller-saves
6884 Enable values to be allocated in registers that will be clobbered by
6885 function calls, by emitting extra instructions to save and restore the
6886 registers around such calls. Such allocation is done only when it
6887 seems to result in better code than would otherwise be produced.
6889 This option is always enabled by default on certain machines, usually
6890 those which have no call-preserved registers to use instead.
6892 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6894 @item -fcombine-stack-adjustments
6895 @opindex fcombine-stack-adjustments
6896 Tracks stack adjustments (pushes and pops) and stack memory references
6897 and then tries to find ways to combine them.
6899 Enabled by default at @option{-O1} and higher.
6901 @item -fconserve-stack
6902 @opindex fconserve-stack
6903 Attempt to minimize stack usage. The compiler will attempt to use less
6904 stack space, even if that makes the program slower. This option
6905 implies setting the @option{large-stack-frame} parameter to 100
6906 and the @option{large-stack-frame-growth} parameter to 400.
6908 @item -ftree-reassoc
6909 @opindex ftree-reassoc
6910 Perform reassociation on trees. This flag is enabled by default
6911 at @option{-O} and higher.
6915 Perform partial redundancy elimination (PRE) on trees. This flag is
6916 enabled by default at @option{-O2} and @option{-O3}.
6918 @item -ftree-forwprop
6919 @opindex ftree-forwprop
6920 Perform forward propagation on trees. This flag is enabled by default
6921 at @option{-O} and higher.
6925 Perform full redundancy elimination (FRE) on trees. The difference
6926 between FRE and PRE is that FRE only considers expressions
6927 that are computed on all paths leading to the redundant computation.
6928 This analysis is faster than PRE, though it exposes fewer redundancies.
6929 This flag is enabled by default at @option{-O} and higher.
6931 @item -ftree-phiprop
6932 @opindex ftree-phiprop
6933 Perform hoisting of loads from conditional pointers on trees. This
6934 pass is enabled by default at @option{-O} and higher.
6936 @item -ftree-copy-prop
6937 @opindex ftree-copy-prop
6938 Perform copy propagation on trees. This pass eliminates unnecessary
6939 copy operations. This flag is enabled by default at @option{-O} and
6942 @item -fipa-pure-const
6943 @opindex fipa-pure-const
6944 Discover which functions are pure or constant.
6945 Enabled by default at @option{-O} and higher.
6947 @item -fipa-reference
6948 @opindex fipa-reference
6949 Discover which static variables do not escape cannot escape the
6951 Enabled by default at @option{-O} and higher.
6955 Perform interprocedural pointer analysis and interprocedural modification
6956 and reference analysis. This option can cause excessive memory and
6957 compile-time usage on large compilation units. It is not enabled by
6958 default at any optimization level.
6961 @opindex fipa-profile
6962 Perform interprocedural profile propagation. The functions called only from
6963 cold functions are marked as cold. Also functions executed once (such as
6964 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6965 functions and loop less parts of functions executed once are then optimized for
6967 Enabled by default at @option{-O} and higher.
6971 Perform interprocedural constant propagation.
6972 This optimization analyzes the program to determine when values passed
6973 to functions are constants and then optimizes accordingly.
6974 This optimization can substantially increase performance
6975 if the application has constants passed to functions.
6976 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6978 @item -fipa-cp-clone
6979 @opindex fipa-cp-clone
6980 Perform function cloning to make interprocedural constant propagation stronger.
6981 When enabled, interprocedural constant propagation will perform function cloning
6982 when externally visible function can be called with constant arguments.
6983 Because this optimization can create multiple copies of functions,
6984 it may significantly increase code size
6985 (see @option{--param ipcp-unit-growth=@var{value}}).
6986 This flag is enabled by default at @option{-O3}.
6988 @item -fipa-matrix-reorg
6989 @opindex fipa-matrix-reorg
6990 Perform matrix flattening and transposing.
6991 Matrix flattening tries to replace an @math{m}-dimensional matrix
6992 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6993 This reduces the level of indirection needed for accessing the elements
6994 of the matrix. The second optimization is matrix transposing that
6995 attempts to change the order of the matrix's dimensions in order to
6996 improve cache locality.
6997 Both optimizations need the @option{-fwhole-program} flag.
6998 Transposing is enabled only if profiling information is available.
7002 Perform forward store motion on trees. This flag is
7003 enabled by default at @option{-O} and higher.
7005 @item -ftree-bit-ccp
7006 @opindex ftree-bit-ccp
7007 Perform sparse conditional bit constant propagation on trees and propagate
7008 pointer alignment information.
7009 This pass only operates on local scalar variables and is enabled by default
7010 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7014 Perform sparse conditional constant propagation (CCP) on trees. This
7015 pass only operates on local scalar variables and is enabled by default
7016 at @option{-O} and higher.
7018 @item -ftree-switch-conversion
7019 Perform conversion of simple initializations in a switch to
7020 initializations from a scalar array. This flag is enabled by default
7021 at @option{-O2} and higher.
7025 Perform dead code elimination (DCE) on trees. This flag is enabled by
7026 default at @option{-O} and higher.
7028 @item -ftree-builtin-call-dce
7029 @opindex ftree-builtin-call-dce
7030 Perform conditional dead code elimination (DCE) for calls to builtin functions
7031 that may set @code{errno} but are otherwise side-effect free. This flag is
7032 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7035 @item -ftree-dominator-opts
7036 @opindex ftree-dominator-opts
7037 Perform a variety of simple scalar cleanups (constant/copy
7038 propagation, redundancy elimination, range propagation and expression
7039 simplification) based on a dominator tree traversal. This also
7040 performs jump threading (to reduce jumps to jumps). This flag is
7041 enabled by default at @option{-O} and higher.
7045 Perform dead store elimination (DSE) on trees. A dead store is a store into
7046 a memory location which will later be overwritten by another store without
7047 any intervening loads. In this case the earlier store can be deleted. This
7048 flag is enabled by default at @option{-O} and higher.
7052 Perform loop header copying on trees. This is beneficial since it increases
7053 effectiveness of code motion optimizations. It also saves one jump. This flag
7054 is enabled by default at @option{-O} and higher. It is not enabled
7055 for @option{-Os}, since it usually increases code size.
7057 @item -ftree-loop-optimize
7058 @opindex ftree-loop-optimize
7059 Perform loop optimizations on trees. This flag is enabled by default
7060 at @option{-O} and higher.
7062 @item -ftree-loop-linear
7063 @opindex ftree-loop-linear
7064 Perform loop interchange transformations on tree. Same as
7065 @option{-floop-interchange}. To use this code transformation, GCC has
7066 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7067 enable the Graphite loop transformation infrastructure.
7069 @item -floop-interchange
7070 @opindex floop-interchange
7071 Perform loop interchange transformations on loops. Interchanging two
7072 nested loops switches the inner and outer loops. For example, given a
7077 A(J, I) = A(J, I) * C
7081 loop interchange will transform the loop as if the user had written:
7085 A(J, I) = A(J, I) * C
7089 which can be beneficial when @code{N} is larger than the caches,
7090 because in Fortran, the elements of an array are stored in memory
7091 contiguously by column, and the original loop iterates over rows,
7092 potentially creating at each access a cache miss. This optimization
7093 applies to all the languages supported by GCC and is not limited to
7094 Fortran. To use this code transformation, GCC has to be configured
7095 with @option{--with-ppl} and @option{--with-cloog} to enable the
7096 Graphite loop transformation infrastructure.
7098 @item -floop-strip-mine
7099 @opindex floop-strip-mine
7100 Perform loop strip mining transformations on loops. Strip mining
7101 splits a loop into two nested loops. The outer loop has strides
7102 equal to the strip size and the inner loop has strides of the
7103 original loop within a strip. The strip length can be changed
7104 using the @option{loop-block-tile-size} parameter. For example,
7111 loop strip mining will transform the loop as if the user had written:
7114 DO I = II, min (II + 50, N)
7119 This optimization applies to all the languages supported by GCC and is
7120 not limited to Fortran. To use this code transformation, GCC has to
7121 be configured with @option{--with-ppl} and @option{--with-cloog} to
7122 enable the Graphite loop transformation infrastructure.
7125 @opindex floop-block
7126 Perform loop blocking transformations on loops. Blocking strip mines
7127 each loop in the loop nest such that the memory accesses of the
7128 element loops fit inside caches. The strip length can be changed
7129 using the @option{loop-block-tile-size} parameter. For example, given
7134 A(J, I) = B(I) + C(J)
7138 loop blocking will transform the loop as if the user had written:
7142 DO I = II, min (II + 50, N)
7143 DO J = JJ, min (JJ + 50, M)
7144 A(J, I) = B(I) + C(J)
7150 which can be beneficial when @code{M} is larger than the caches,
7151 because the innermost loop will iterate over a smaller amount of data
7152 that can be kept in the caches. This optimization applies to all the
7153 languages supported by GCC and is not limited to Fortran. To use this
7154 code transformation, GCC has to be configured with @option{--with-ppl}
7155 and @option{--with-cloog} to enable the Graphite loop transformation
7158 @item -fgraphite-identity
7159 @opindex fgraphite-identity
7160 Enable the identity transformation for graphite. For every SCoP we generate
7161 the polyhedral representation and transform it back to gimple. Using
7162 @option{-fgraphite-identity} we can check the costs or benefits of the
7163 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7164 are also performed by the code generator CLooG, like index splitting and
7165 dead code elimination in loops.
7167 @item -floop-flatten
7168 @opindex floop-flatten
7169 Removes the loop nesting structure: transforms the loop nest into a
7170 single loop. This transformation can be useful to vectorize all the
7171 levels of the loop nest.
7173 @item -floop-parallelize-all
7174 @opindex floop-parallelize-all
7175 Use the Graphite data dependence analysis to identify loops that can
7176 be parallelized. Parallelize all the loops that can be analyzed to
7177 not contain loop carried dependences without checking that it is
7178 profitable to parallelize the loops.
7180 @item -fcheck-data-deps
7181 @opindex fcheck-data-deps
7182 Compare the results of several data dependence analyzers. This option
7183 is used for debugging the data dependence analyzers.
7185 @item -ftree-loop-if-convert
7186 Attempt to transform conditional jumps in the innermost loops to
7187 branch-less equivalents. The intent is to remove control-flow from
7188 the innermost loops in order to improve the ability of the
7189 vectorization pass to handle these loops. This is enabled by default
7190 if vectorization is enabled.
7192 @item -ftree-loop-if-convert-stores
7193 Attempt to also if-convert conditional jumps containing memory writes.
7194 This transformation can be unsafe for multi-threaded programs as it
7195 transforms conditional memory writes into unconditional memory writes.
7198 for (i = 0; i < N; i++)
7202 would be transformed to
7204 for (i = 0; i < N; i++)
7205 A[i] = cond ? expr : A[i];
7207 potentially producing data races.
7209 @item -ftree-loop-distribution
7210 Perform loop distribution. This flag can improve cache performance on
7211 big loop bodies and allow further loop optimizations, like
7212 parallelization or vectorization, to take place. For example, the loop
7229 @item -ftree-loop-distribute-patterns
7230 Perform loop distribution of patterns that can be code generated with
7231 calls to a library. This flag is enabled by default at @option{-O3}.
7233 This pass distributes the initialization loops and generates a call to
7234 memset zero. For example, the loop
7250 and the initialization loop is transformed into a call to memset zero.
7252 @item -ftree-loop-im
7253 @opindex ftree-loop-im
7254 Perform loop invariant motion on trees. This pass moves only invariants that
7255 would be hard to handle at RTL level (function calls, operations that expand to
7256 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7257 operands of conditions that are invariant out of the loop, so that we can use
7258 just trivial invariantness analysis in loop unswitching. The pass also includes
7261 @item -ftree-loop-ivcanon
7262 @opindex ftree-loop-ivcanon
7263 Create a canonical counter for number of iterations in the loop for that
7264 determining number of iterations requires complicated analysis. Later
7265 optimizations then may determine the number easily. Useful especially
7266 in connection with unrolling.
7270 Perform induction variable optimizations (strength reduction, induction
7271 variable merging and induction variable elimination) on trees.
7273 @item -ftree-parallelize-loops=n
7274 @opindex ftree-parallelize-loops
7275 Parallelize loops, i.e., split their iteration space to run in n threads.
7276 This is only possible for loops whose iterations are independent
7277 and can be arbitrarily reordered. The optimization is only
7278 profitable on multiprocessor machines, for loops that are CPU-intensive,
7279 rather than constrained e.g.@: by memory bandwidth. This option
7280 implies @option{-pthread}, and thus is only supported on targets
7281 that have support for @option{-pthread}.
7285 Perform function-local points-to analysis on trees. This flag is
7286 enabled by default at @option{-O} and higher.
7290 Perform scalar replacement of aggregates. This pass replaces structure
7291 references with scalars to prevent committing structures to memory too
7292 early. This flag is enabled by default at @option{-O} and higher.
7294 @item -ftree-copyrename
7295 @opindex ftree-copyrename
7296 Perform copy renaming on trees. This pass attempts to rename compiler
7297 temporaries to other variables at copy locations, usually resulting in
7298 variable names which more closely resemble the original variables. This flag
7299 is enabled by default at @option{-O} and higher.
7303 Perform temporary expression replacement during the SSA->normal phase. Single
7304 use/single def temporaries are replaced at their use location with their
7305 defining expression. This results in non-GIMPLE code, but gives the expanders
7306 much more complex trees to work on resulting in better RTL generation. This is
7307 enabled by default at @option{-O} and higher.
7309 @item -ftree-vectorize
7310 @opindex ftree-vectorize
7311 Perform loop vectorization on trees. This flag is enabled by default at
7314 @item -ftree-slp-vectorize
7315 @opindex ftree-slp-vectorize
7316 Perform basic block vectorization on trees. This flag is enabled by default at
7317 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7319 @item -ftree-vect-loop-version
7320 @opindex ftree-vect-loop-version
7321 Perform loop versioning when doing loop vectorization on trees. When a loop
7322 appears to be vectorizable except that data alignment or data dependence cannot
7323 be determined at compile time then vectorized and non-vectorized versions of
7324 the loop are generated along with runtime checks for alignment or dependence
7325 to control which version is executed. This option is enabled by default
7326 except at level @option{-Os} where it is disabled.
7328 @item -fvect-cost-model
7329 @opindex fvect-cost-model
7330 Enable cost model for vectorization.
7334 Perform Value Range Propagation on trees. This is similar to the
7335 constant propagation pass, but instead of values, ranges of values are
7336 propagated. This allows the optimizers to remove unnecessary range
7337 checks like array bound checks and null pointer checks. This is
7338 enabled by default at @option{-O2} and higher. Null pointer check
7339 elimination is only done if @option{-fdelete-null-pointer-checks} is
7344 Perform tail duplication to enlarge superblock size. This transformation
7345 simplifies the control flow of the function allowing other optimizations to do
7348 @item -funroll-loops
7349 @opindex funroll-loops
7350 Unroll loops whose number of iterations can be determined at compile
7351 time or upon entry to the loop. @option{-funroll-loops} implies
7352 @option{-frerun-cse-after-loop}. This option makes code larger,
7353 and may or may not make it run faster.
7355 @item -funroll-all-loops
7356 @opindex funroll-all-loops
7357 Unroll all loops, even if their number of iterations is uncertain when
7358 the loop is entered. This usually makes programs run more slowly.
7359 @option{-funroll-all-loops} implies the same options as
7360 @option{-funroll-loops},
7362 @item -fsplit-ivs-in-unroller
7363 @opindex fsplit-ivs-in-unroller
7364 Enables expressing of values of induction variables in later iterations
7365 of the unrolled loop using the value in the first iteration. This breaks
7366 long dependency chains, thus improving efficiency of the scheduling passes.
7368 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7369 same effect. However in cases the loop body is more complicated than
7370 a single basic block, this is not reliable. It also does not work at all
7371 on some of the architectures due to restrictions in the CSE pass.
7373 This optimization is enabled by default.
7375 @item -fvariable-expansion-in-unroller
7376 @opindex fvariable-expansion-in-unroller
7377 With this option, the compiler will create multiple copies of some
7378 local variables when unrolling a loop which can result in superior code.
7380 @item -fpartial-inlining
7381 @opindex fpartial-inlining
7382 Inline parts of functions. This option has any effect only
7383 when inlining itself is turned on by the @option{-finline-functions}
7384 or @option{-finline-small-functions} options.
7386 Enabled at level @option{-O2}.
7388 @item -fpredictive-commoning
7389 @opindex fpredictive-commoning
7390 Perform predictive commoning optimization, i.e., reusing computations
7391 (especially memory loads and stores) performed in previous
7392 iterations of loops.
7394 This option is enabled at level @option{-O3}.
7396 @item -fprefetch-loop-arrays
7397 @opindex fprefetch-loop-arrays
7398 If supported by the target machine, generate instructions to prefetch
7399 memory to improve the performance of loops that access large arrays.
7401 This option may generate better or worse code; results are highly
7402 dependent on the structure of loops within the source code.
7404 Disabled at level @option{-Os}.
7407 @itemx -fno-peephole2
7408 @opindex fno-peephole
7409 @opindex fno-peephole2
7410 Disable any machine-specific peephole optimizations. The difference
7411 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7412 are implemented in the compiler; some targets use one, some use the
7413 other, a few use both.
7415 @option{-fpeephole} is enabled by default.
7416 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7418 @item -fno-guess-branch-probability
7419 @opindex fno-guess-branch-probability
7420 Do not guess branch probabilities using heuristics.
7422 GCC will use heuristics to guess branch probabilities if they are
7423 not provided by profiling feedback (@option{-fprofile-arcs}). These
7424 heuristics are based on the control flow graph. If some branch probabilities
7425 are specified by @samp{__builtin_expect}, then the heuristics will be
7426 used to guess branch probabilities for the rest of the control flow graph,
7427 taking the @samp{__builtin_expect} info into account. The interactions
7428 between the heuristics and @samp{__builtin_expect} can be complex, and in
7429 some cases, it may be useful to disable the heuristics so that the effects
7430 of @samp{__builtin_expect} are easier to understand.
7432 The default is @option{-fguess-branch-probability} at levels
7433 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7435 @item -freorder-blocks
7436 @opindex freorder-blocks
7437 Reorder basic blocks in the compiled function in order to reduce number of
7438 taken branches and improve code locality.
7440 Enabled at levels @option{-O2}, @option{-O3}.
7442 @item -freorder-blocks-and-partition
7443 @opindex freorder-blocks-and-partition
7444 In addition to reordering basic blocks in the compiled function, in order
7445 to reduce number of taken branches, partitions hot and cold basic blocks
7446 into separate sections of the assembly and .o files, to improve
7447 paging and cache locality performance.
7449 This optimization is automatically turned off in the presence of
7450 exception handling, for linkonce sections, for functions with a user-defined
7451 section attribute and on any architecture that does not support named
7454 @item -freorder-functions
7455 @opindex freorder-functions
7456 Reorder functions in the object file in order to
7457 improve code locality. This is implemented by using special
7458 subsections @code{.text.hot} for most frequently executed functions and
7459 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7460 the linker so object file format must support named sections and linker must
7461 place them in a reasonable way.
7463 Also profile feedback must be available in to make this option effective. See
7464 @option{-fprofile-arcs} for details.
7466 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7468 @item -fstrict-aliasing
7469 @opindex fstrict-aliasing
7470 Allow the compiler to assume the strictest aliasing rules applicable to
7471 the language being compiled. For C (and C++), this activates
7472 optimizations based on the type of expressions. In particular, an
7473 object of one type is assumed never to reside at the same address as an
7474 object of a different type, unless the types are almost the same. For
7475 example, an @code{unsigned int} can alias an @code{int}, but not a
7476 @code{void*} or a @code{double}. A character type may alias any other
7479 @anchor{Type-punning}Pay special attention to code like this:
7492 The practice of reading from a different union member than the one most
7493 recently written to (called ``type-punning'') is common. Even with
7494 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7495 is accessed through the union type. So, the code above will work as
7496 expected. @xref{Structures unions enumerations and bit-fields
7497 implementation}. However, this code might not:
7508 Similarly, access by taking the address, casting the resulting pointer
7509 and dereferencing the result has undefined behavior, even if the cast
7510 uses a union type, e.g.:
7514 return ((union a_union *) &d)->i;
7518 The @option{-fstrict-aliasing} option is enabled at levels
7519 @option{-O2}, @option{-O3}, @option{-Os}.
7521 @item -fstrict-overflow
7522 @opindex fstrict-overflow
7523 Allow the compiler to assume strict signed overflow rules, depending
7524 on the language being compiled. For C (and C++) this means that
7525 overflow when doing arithmetic with signed numbers is undefined, which
7526 means that the compiler may assume that it will not happen. This
7527 permits various optimizations. For example, the compiler will assume
7528 that an expression like @code{i + 10 > i} will always be true for
7529 signed @code{i}. This assumption is only valid if signed overflow is
7530 undefined, as the expression is false if @code{i + 10} overflows when
7531 using twos complement arithmetic. When this option is in effect any
7532 attempt to determine whether an operation on signed numbers will
7533 overflow must be written carefully to not actually involve overflow.
7535 This option also allows the compiler to assume strict pointer
7536 semantics: given a pointer to an object, if adding an offset to that
7537 pointer does not produce a pointer to the same object, the addition is
7538 undefined. This permits the compiler to conclude that @code{p + u >
7539 p} is always true for a pointer @code{p} and unsigned integer
7540 @code{u}. This assumption is only valid because pointer wraparound is
7541 undefined, as the expression is false if @code{p + u} overflows using
7542 twos complement arithmetic.
7544 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7545 that integer signed overflow is fully defined: it wraps. When
7546 @option{-fwrapv} is used, there is no difference between
7547 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7548 integers. With @option{-fwrapv} certain types of overflow are
7549 permitted. For example, if the compiler gets an overflow when doing
7550 arithmetic on constants, the overflowed value can still be used with
7551 @option{-fwrapv}, but not otherwise.
7553 The @option{-fstrict-overflow} option is enabled at levels
7554 @option{-O2}, @option{-O3}, @option{-Os}.
7556 @item -falign-functions
7557 @itemx -falign-functions=@var{n}
7558 @opindex falign-functions
7559 Align the start of functions to the next power-of-two greater than
7560 @var{n}, skipping up to @var{n} bytes. For instance,
7561 @option{-falign-functions=32} aligns functions to the next 32-byte
7562 boundary, but @option{-falign-functions=24} would align to the next
7563 32-byte boundary only if this can be done by skipping 23 bytes or less.
7565 @option{-fno-align-functions} and @option{-falign-functions=1} are
7566 equivalent and mean that functions will not be aligned.
7568 Some assemblers only support this flag when @var{n} is a power of two;
7569 in that case, it is rounded up.
7571 If @var{n} is not specified or is zero, use a machine-dependent default.
7573 Enabled at levels @option{-O2}, @option{-O3}.
7575 @item -falign-labels
7576 @itemx -falign-labels=@var{n}
7577 @opindex falign-labels
7578 Align all branch targets to a power-of-two boundary, skipping up to
7579 @var{n} bytes like @option{-falign-functions}. This option can easily
7580 make code slower, because it must insert dummy operations for when the
7581 branch target is reached in the usual flow of the code.
7583 @option{-fno-align-labels} and @option{-falign-labels=1} are
7584 equivalent and mean that labels will not be aligned.
7586 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7587 are greater than this value, then their values are used instead.
7589 If @var{n} is not specified or is zero, use a machine-dependent default
7590 which is very likely to be @samp{1}, meaning no alignment.
7592 Enabled at levels @option{-O2}, @option{-O3}.
7595 @itemx -falign-loops=@var{n}
7596 @opindex falign-loops
7597 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7598 like @option{-falign-functions}. The hope is that the loop will be
7599 executed many times, which will make up for any execution of the dummy
7602 @option{-fno-align-loops} and @option{-falign-loops=1} are
7603 equivalent and mean that loops will not be aligned.
7605 If @var{n} is not specified or is zero, use a machine-dependent default.
7607 Enabled at levels @option{-O2}, @option{-O3}.
7610 @itemx -falign-jumps=@var{n}
7611 @opindex falign-jumps
7612 Align branch targets to a power-of-two boundary, for branch targets
7613 where the targets can only be reached by jumping, skipping up to @var{n}
7614 bytes like @option{-falign-functions}. In this case, no dummy operations
7617 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7618 equivalent and mean that loops will not be aligned.
7620 If @var{n} is not specified or is zero, use a machine-dependent default.
7622 Enabled at levels @option{-O2}, @option{-O3}.
7624 @item -funit-at-a-time
7625 @opindex funit-at-a-time
7626 This option is left for compatibility reasons. @option{-funit-at-a-time}
7627 has no effect, while @option{-fno-unit-at-a-time} implies
7628 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7632 @item -fno-toplevel-reorder
7633 @opindex fno-toplevel-reorder
7634 Do not reorder top-level functions, variables, and @code{asm}
7635 statements. Output them in the same order that they appear in the
7636 input file. When this option is used, unreferenced static variables
7637 will not be removed. This option is intended to support existing code
7638 which relies on a particular ordering. For new code, it is better to
7641 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7642 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7647 Constructs webs as commonly used for register allocation purposes and assign
7648 each web individual pseudo register. This allows the register allocation pass
7649 to operate on pseudos directly, but also strengthens several other optimization
7650 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7651 however, make debugging impossible, since variables will no longer stay in a
7654 Enabled by default with @option{-funroll-loops}.
7656 @item -fwhole-program
7657 @opindex fwhole-program
7658 Assume that the current compilation unit represents the whole program being
7659 compiled. All public functions and variables with the exception of @code{main}
7660 and those merged by attribute @code{externally_visible} become static functions
7661 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.
7662 While this option is equivalent to proper use of the @code{static} keyword for
7663 programs consisting of a single file, in combination with option
7664 @option{-flto} this flag can be used to
7665 compile many smaller scale programs since the functions and variables become
7666 local for the whole combined compilation unit, not for the single source file
7669 This option implies @option{-fwhole-file} for Fortran programs.
7671 @item -flto[=@var{n}]
7673 This option runs the standard link-time optimizer. When invoked
7674 with source code, it generates GIMPLE (one of GCC's internal
7675 representations) and writes it to special ELF sections in the object
7676 file. When the object files are linked together, all the function
7677 bodies are read from these ELF sections and instantiated as if they
7678 had been part of the same translation unit.
7680 To use the link-timer optimizer, @option{-flto} needs to be specified at
7681 compile time and during the final link. For example,
7684 gcc -c -O2 -flto foo.c
7685 gcc -c -O2 -flto bar.c
7686 gcc -o myprog -flto -O2 foo.o bar.o
7689 The first two invocations to GCC will save a bytecode representation
7690 of GIMPLE into special ELF sections inside @file{foo.o} and
7691 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7692 @file{foo.o} and @file{bar.o}, merge the two files into a single
7693 internal image, and compile the result as usual. Since both
7694 @file{foo.o} and @file{bar.o} are merged into a single image, this
7695 causes all the inter-procedural analyses and optimizations in GCC to
7696 work across the two files as if they were a single one. This means,
7697 for example, that the inliner will be able to inline functions in
7698 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7700 Another (simpler) way to enable link-time optimization is,
7703 gcc -o myprog -flto -O2 foo.c bar.c
7706 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7707 merge them together into a single GIMPLE representation and optimize
7708 them as usual to produce @file{myprog}.
7710 The only important thing to keep in mind is that to enable link-time
7711 optimizations the @option{-flto} flag needs to be passed to both the
7712 compile and the link commands.
7714 To make whole program optimization effective, it is necessary to make
7715 certain whole program assumptions. The compiler needs to know
7716 what functions and variables can be accessed by libraries and runtime
7717 outside of the link time optimized unit. When supported by the linker,
7718 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7719 compiler information about used and externally visible symbols. When
7720 the linker plugin is not available, @option{-fwhole-program} should be
7721 used to allow the compiler to make these assumptions, which will lead
7722 to more aggressive optimization decisions.
7724 Note that when a file is compiled with @option{-flto}, the generated
7725 object file will be larger than a regular object file because it will
7726 contain GIMPLE bytecodes and the usual final code. This means that
7727 object files with LTO information can be linked as a normal object
7728 file. So, in the previous example, if the final link is done with
7731 gcc -o myprog foo.o bar.o
7734 The only difference will be that no inter-procedural optimizations
7735 will be applied to produce @file{myprog}. The two object files
7736 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7739 Additionally, the optimization flags used to compile individual files
7740 are not necessarily related to those used at link-time. For instance,
7743 gcc -c -O0 -flto foo.c
7744 gcc -c -O0 -flto bar.c
7745 gcc -o myprog -flto -O3 foo.o bar.o
7748 This will produce individual object files with unoptimized assembler
7749 code, but the resulting binary @file{myprog} will be optimized at
7750 @option{-O3}. Now, if the final binary is generated without
7751 @option{-flto}, then @file{myprog} will not be optimized.
7753 When producing the final binary with @option{-flto}, GCC will only
7754 apply link-time optimizations to those files that contain bytecode.
7755 Therefore, you can mix and match object files and libraries with
7756 GIMPLE bytecodes and final object code. GCC will automatically select
7757 which files to optimize in LTO mode and which files to link without
7760 There are some code generation flags that GCC will preserve when
7761 generating bytecodes, as they need to be used during the final link
7762 stage. Currently, the following options are saved into the GIMPLE
7763 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7764 @option{-m} target flags.
7766 At link time, these options are read-in and reapplied. Note that the
7767 current implementation makes no attempt at recognizing conflicting
7768 values for these options. If two or more files have a conflicting
7769 value (e.g., one file is compiled with @option{-fPIC} and another
7770 isn't), the compiler will simply use the last value read from the
7771 bytecode files. It is recommended, then, that all the files
7772 participating in the same link be compiled with the same options.
7774 Another feature of LTO is that it is possible to apply interprocedural
7775 optimizations on files written in different languages. This requires
7776 some support in the language front end. Currently, the C, C++ and
7777 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7778 something like this should work
7783 gfortran -c -flto baz.f90
7784 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7787 Notice that the final link is done with @command{g++} to get the C++
7788 runtime libraries and @option{-lgfortran} is added to get the Fortran
7789 runtime libraries. In general, when mixing languages in LTO mode, you
7790 should use the same link command used when mixing languages in a
7791 regular (non-LTO) compilation. This means that if your build process
7792 was mixing languages before, all you need to add is @option{-flto} to
7793 all the compile and link commands.
7795 If LTO encounters objects with C linkage declared with incompatible
7796 types in separate translation units to be linked together (undefined
7797 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7798 issued. The behavior is still undefined at runtime.
7800 If object files containing GIMPLE bytecode are stored in a library archive, say
7801 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7802 are using a linker with linker plugin support. To enable this feature, use
7803 the flag @option{-fuse-linker-plugin} at link-time:
7806 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7809 With the linker plugin enabled, the linker will extract the needed
7810 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7811 to make them part of the aggregated GIMPLE image to be optimized.
7813 If you are not using a linker with linker plugin support and/or do not
7814 enable linker plugin then the objects inside @file{libfoo.a}
7815 will be extracted and linked as usual, but they will not participate
7816 in the LTO optimization process.
7818 Link time optimizations do not require the presence of the whole program to
7819 operate. If the program does not require any symbols to be exported, it is
7820 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7821 the interprocedural optimizers to use more aggressive assumptions which may
7822 lead to improved optimization opportunities.
7823 Use of @option{-fwhole-program} is not needed when linker plugin is
7824 active (see @option{-fuse-linker-plugin}).
7826 Regarding portability: the current implementation of LTO makes no
7827 attempt at generating bytecode that can be ported between different
7828 types of hosts. The bytecode files are versioned and there is a
7829 strict version check, so bytecode files generated in one version of
7830 GCC will not work with an older/newer version of GCC.
7832 Link time optimization does not play well with generating debugging
7833 information. Combining @option{-flto} with
7834 @option{-g} is currently experimental and expected to produce wrong
7837 If you specify the optional @var{n}, the optimization and code
7838 generation done at link time is executed in parallel using @var{n}
7839 parallel jobs by utilizing an installed @command{make} program. The
7840 environment variable @env{MAKE} may be used to override the program
7841 used. The default value for @var{n} is 1.
7843 You can also specify @option{-flto=jobserver} to use GNU make's
7844 job server mode to determine the number of parallel jobs. This
7845 is useful when the Makefile calling GCC is already executing in parallel.
7846 The parent Makefile will need a @samp{+} prepended to the command recipe
7847 for this to work. This will likely only work if @env{MAKE} is
7850 This option is disabled by default.
7852 @item -flto-partition=@var{alg}
7853 @opindex flto-partition
7854 Specify the partitioning algorithm used by the link time optimizer.
7855 The value is either @code{1to1} to specify a partitioning mirroring
7856 the original source files or @code{balanced} to specify partitioning
7857 into equally sized chunks (whenever possible). Specifying @code{none}
7858 as an algorithm disables partitioning and streaming completely. The
7859 default value is @code{balanced}.
7861 @item -flto-compression-level=@var{n}
7862 This option specifies the level of compression used for intermediate
7863 language written to LTO object files, and is only meaningful in
7864 conjunction with LTO mode (@option{-flto}). Valid
7865 values are 0 (no compression) to 9 (maximum compression). Values
7866 outside this range are clamped to either 0 or 9. If the option is not
7867 given, a default balanced compression setting is used.
7870 Prints a report with internal details on the workings of the link-time
7871 optimizer. The contents of this report vary from version to version,
7872 it is meant to be useful to GCC developers when processing object
7873 files in LTO mode (via @option{-flto}).
7875 Disabled by default.
7877 @item -fuse-linker-plugin
7878 Enables the use of a linker plugin during link time optimization. This
7879 option relies on plugin support in the linker, which is available in gold
7880 or in GNU ld 2.21 or newer.
7882 This option enables the extraction of object files with GIMPLE bytecode out
7883 of library archives. This improves the quality of optimization by exposing
7884 more code to the link time optimizer. This information specifies what
7885 symbols can be accessed externally (by non-LTO object or during dynamic
7886 linking). Resulting code quality improvements on binaries (and shared
7887 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7888 See @option{-flto} for a description of the effect of this flag and how to
7891 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7892 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7894 @item -fcompare-elim
7895 @opindex fcompare-elim
7896 After register allocation and post-register allocation instruction splitting,
7897 identify arithmetic instructions that compute processor flags similar to a
7898 comparison operation based on that arithmetic. If possible, eliminate the
7899 explicit comparison operation.
7901 This pass only applies to certain targets that cannot explicitly represent
7902 the comparison operation before register allocation is complete.
7904 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7906 @item -fcprop-registers
7907 @opindex fcprop-registers
7908 After register allocation and post-register allocation instruction splitting,
7909 we perform a copy-propagation pass to try to reduce scheduling dependencies
7910 and occasionally eliminate the copy.
7912 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7914 @item -fprofile-correction
7915 @opindex fprofile-correction
7916 Profiles collected using an instrumented binary for multi-threaded programs may
7917 be inconsistent due to missed counter updates. When this option is specified,
7918 GCC will use heuristics to correct or smooth out such inconsistencies. By
7919 default, GCC will emit an error message when an inconsistent profile is detected.
7921 @item -fprofile-dir=@var{path}
7922 @opindex fprofile-dir
7924 Set the directory to search for the profile data files in to @var{path}.
7925 This option affects only the profile data generated by
7926 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7927 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7928 and its related options. Both absolute and relative paths can be used.
7929 By default, GCC will use the current directory as @var{path}, thus the
7930 profile data file will appear in the same directory as the object file.
7932 @item -fprofile-generate
7933 @itemx -fprofile-generate=@var{path}
7934 @opindex fprofile-generate
7936 Enable options usually used for instrumenting application to produce
7937 profile useful for later recompilation with profile feedback based
7938 optimization. You must use @option{-fprofile-generate} both when
7939 compiling and when linking your program.
7941 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7943 If @var{path} is specified, GCC will look at the @var{path} to find
7944 the profile feedback data files. See @option{-fprofile-dir}.
7947 @itemx -fprofile-use=@var{path}
7948 @opindex fprofile-use
7949 Enable profile feedback directed optimizations, and optimizations
7950 generally profitable only with profile feedback available.
7952 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7953 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7955 By default, GCC emits an error message if the feedback profiles do not
7956 match the source code. This error can be turned into a warning by using
7957 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7960 If @var{path} is specified, GCC will look at the @var{path} to find
7961 the profile feedback data files. See @option{-fprofile-dir}.
7964 The following options control compiler behavior regarding floating
7965 point arithmetic. These options trade off between speed and
7966 correctness. All must be specifically enabled.
7970 @opindex ffloat-store
7971 Do not store floating point variables in registers, and inhibit other
7972 options that might change whether a floating point value is taken from a
7975 @cindex floating point precision
7976 This option prevents undesirable excess precision on machines such as
7977 the 68000 where the floating registers (of the 68881) keep more
7978 precision than a @code{double} is supposed to have. Similarly for the
7979 x86 architecture. For most programs, the excess precision does only
7980 good, but a few programs rely on the precise definition of IEEE floating
7981 point. Use @option{-ffloat-store} for such programs, after modifying
7982 them to store all pertinent intermediate computations into variables.
7984 @item -fexcess-precision=@var{style}
7985 @opindex fexcess-precision
7986 This option allows further control over excess precision on machines
7987 where floating-point registers have more precision than the IEEE
7988 @code{float} and @code{double} types and the processor does not
7989 support operations rounding to those types. By default,
7990 @option{-fexcess-precision=fast} is in effect; this means that
7991 operations are carried out in the precision of the registers and that
7992 it is unpredictable when rounding to the types specified in the source
7993 code takes place. When compiling C, if
7994 @option{-fexcess-precision=standard} is specified then excess
7995 precision will follow the rules specified in ISO C99; in particular,
7996 both casts and assignments cause values to be rounded to their
7997 semantic types (whereas @option{-ffloat-store} only affects
7998 assignments). This option is enabled by default for C if a strict
7999 conformance option such as @option{-std=c99} is used.
8002 @option{-fexcess-precision=standard} is not implemented for languages
8003 other than C, and has no effect if
8004 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8005 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8006 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8007 semantics apply without excess precision, and in the latter, rounding
8012 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8013 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8014 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8016 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8018 This option is not turned on by any @option{-O} option besides
8019 @option{-Ofast} since it can result in incorrect output for programs
8020 which depend on an exact implementation of IEEE or ISO rules/specifications
8021 for math functions. It may, however, yield faster code for programs
8022 that do not require the guarantees of these specifications.
8024 @item -fno-math-errno
8025 @opindex fno-math-errno
8026 Do not set ERRNO after calling math functions that are executed
8027 with a single instruction, e.g., sqrt. A program that relies on
8028 IEEE exceptions for math error handling may want to use this flag
8029 for speed while maintaining IEEE arithmetic compatibility.
8031 This option is not turned on by any @option{-O} option since
8032 it can result in incorrect output for programs which depend on
8033 an exact implementation of IEEE or ISO rules/specifications for
8034 math functions. It may, however, yield faster code for programs
8035 that do not require the guarantees of these specifications.
8037 The default is @option{-fmath-errno}.
8039 On Darwin systems, the math library never sets @code{errno}. There is
8040 therefore no reason for the compiler to consider the possibility that
8041 it might, and @option{-fno-math-errno} is the default.
8043 @item -funsafe-math-optimizations
8044 @opindex funsafe-math-optimizations
8046 Allow optimizations for floating-point arithmetic that (a) assume
8047 that arguments and results are valid and (b) may violate IEEE or
8048 ANSI standards. When used at link-time, it may include libraries
8049 or startup files that change the default FPU control word or other
8050 similar optimizations.
8052 This option is not turned on by any @option{-O} option since
8053 it can result in incorrect output for programs which depend on
8054 an exact implementation of IEEE or ISO rules/specifications for
8055 math functions. It may, however, yield faster code for programs
8056 that do not require the guarantees of these specifications.
8057 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8058 @option{-fassociative-math} and @option{-freciprocal-math}.
8060 The default is @option{-fno-unsafe-math-optimizations}.
8062 @item -fassociative-math
8063 @opindex fassociative-math
8065 Allow re-association of operands in series of floating-point operations.
8066 This violates the ISO C and C++ language standard by possibly changing
8067 computation result. NOTE: re-ordering may change the sign of zero as
8068 well as ignore NaNs and inhibit or create underflow or overflow (and
8069 thus cannot be used on a code which relies on rounding behavior like
8070 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
8071 and thus may not be used when ordered comparisons are required.
8072 This option requires that both @option{-fno-signed-zeros} and
8073 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8074 much sense with @option{-frounding-math}. For Fortran the option
8075 is automatically enabled when both @option{-fno-signed-zeros} and
8076 @option{-fno-trapping-math} are in effect.
8078 The default is @option{-fno-associative-math}.
8080 @item -freciprocal-math
8081 @opindex freciprocal-math
8083 Allow the reciprocal of a value to be used instead of dividing by
8084 the value if this enables optimizations. For example @code{x / y}
8085 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8086 is subject to common subexpression elimination. Note that this loses
8087 precision and increases the number of flops operating on the value.
8089 The default is @option{-fno-reciprocal-math}.
8091 @item -ffinite-math-only
8092 @opindex ffinite-math-only
8093 Allow optimizations for floating-point arithmetic that assume
8094 that arguments and results are not NaNs or +-Infs.
8096 This option is not turned on by any @option{-O} option since
8097 it can result in incorrect output for programs which depend on
8098 an exact implementation of IEEE or ISO rules/specifications for
8099 math functions. It may, however, yield faster code for programs
8100 that do not require the guarantees of these specifications.
8102 The default is @option{-fno-finite-math-only}.
8104 @item -fno-signed-zeros
8105 @opindex fno-signed-zeros
8106 Allow optimizations for floating point arithmetic that ignore the
8107 signedness of zero. IEEE arithmetic specifies the behavior of
8108 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8109 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8110 This option implies that the sign of a zero result isn't significant.
8112 The default is @option{-fsigned-zeros}.
8114 @item -fno-trapping-math
8115 @opindex fno-trapping-math
8116 Compile code assuming that floating-point operations cannot generate
8117 user-visible traps. These traps include division by zero, overflow,
8118 underflow, inexact result and invalid operation. This option requires
8119 that @option{-fno-signaling-nans} be in effect. Setting this option may
8120 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8122 This option should never be turned on by any @option{-O} option since
8123 it can result in incorrect output for programs which depend on
8124 an exact implementation of IEEE or ISO rules/specifications for
8127 The default is @option{-ftrapping-math}.
8129 @item -frounding-math
8130 @opindex frounding-math
8131 Disable transformations and optimizations that assume default floating
8132 point rounding behavior. This is round-to-zero for all floating point
8133 to integer conversions, and round-to-nearest for all other arithmetic
8134 truncations. This option should be specified for programs that change
8135 the FP rounding mode dynamically, or that may be executed with a
8136 non-default rounding mode. This option disables constant folding of
8137 floating point expressions at compile-time (which may be affected by
8138 rounding mode) and arithmetic transformations that are unsafe in the
8139 presence of sign-dependent rounding modes.
8141 The default is @option{-fno-rounding-math}.
8143 This option is experimental and does not currently guarantee to
8144 disable all GCC optimizations that are affected by rounding mode.
8145 Future versions of GCC may provide finer control of this setting
8146 using C99's @code{FENV_ACCESS} pragma. This command line option
8147 will be used to specify the default state for @code{FENV_ACCESS}.
8149 @item -fsignaling-nans
8150 @opindex fsignaling-nans
8151 Compile code assuming that IEEE signaling NaNs may generate user-visible
8152 traps during floating-point operations. Setting this option disables
8153 optimizations that may change the number of exceptions visible with
8154 signaling NaNs. This option implies @option{-ftrapping-math}.
8156 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8159 The default is @option{-fno-signaling-nans}.
8161 This option is experimental and does not currently guarantee to
8162 disable all GCC optimizations that affect signaling NaN behavior.
8164 @item -fsingle-precision-constant
8165 @opindex fsingle-precision-constant
8166 Treat floating point constant as single precision constant instead of
8167 implicitly converting it to double precision constant.
8169 @item -fcx-limited-range
8170 @opindex fcx-limited-range
8171 When enabled, this option states that a range reduction step is not
8172 needed when performing complex division. Also, there is no checking
8173 whether the result of a complex multiplication or division is @code{NaN
8174 + I*NaN}, with an attempt to rescue the situation in that case. The
8175 default is @option{-fno-cx-limited-range}, but is enabled by
8176 @option{-ffast-math}.
8178 This option controls the default setting of the ISO C99
8179 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8182 @item -fcx-fortran-rules
8183 @opindex fcx-fortran-rules
8184 Complex multiplication and division follow Fortran rules. Range
8185 reduction is done as part of complex division, but there is no checking
8186 whether the result of a complex multiplication or division is @code{NaN
8187 + I*NaN}, with an attempt to rescue the situation in that case.
8189 The default is @option{-fno-cx-fortran-rules}.
8193 The following options control optimizations that may improve
8194 performance, but are not enabled by any @option{-O} options. This
8195 section includes experimental options that may produce broken code.
8198 @item -fbranch-probabilities
8199 @opindex fbranch-probabilities
8200 After running a program compiled with @option{-fprofile-arcs}
8201 (@pxref{Debugging Options,, Options for Debugging Your Program or
8202 @command{gcc}}), you can compile it a second time using
8203 @option{-fbranch-probabilities}, to improve optimizations based on
8204 the number of times each branch was taken. When the program
8205 compiled with @option{-fprofile-arcs} exits it saves arc execution
8206 counts to a file called @file{@var{sourcename}.gcda} for each source
8207 file. The information in this data file is very dependent on the
8208 structure of the generated code, so you must use the same source code
8209 and the same optimization options for both compilations.
8211 With @option{-fbranch-probabilities}, GCC puts a
8212 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8213 These can be used to improve optimization. Currently, they are only
8214 used in one place: in @file{reorg.c}, instead of guessing which path a
8215 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8216 exactly determine which path is taken more often.
8218 @item -fprofile-values
8219 @opindex fprofile-values
8220 If combined with @option{-fprofile-arcs}, it adds code so that some
8221 data about values of expressions in the program is gathered.
8223 With @option{-fbranch-probabilities}, it reads back the data gathered
8224 from profiling values of expressions for usage in optimizations.
8226 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8230 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8231 a code to gather information about values of expressions.
8233 With @option{-fbranch-probabilities}, it reads back the data gathered
8234 and actually performs the optimizations based on them.
8235 Currently the optimizations include specialization of division operation
8236 using the knowledge about the value of the denominator.
8238 @item -frename-registers
8239 @opindex frename-registers
8240 Attempt to avoid false dependencies in scheduled code by making use
8241 of registers left over after register allocation. This optimization
8242 will most benefit processors with lots of registers. Depending on the
8243 debug information format adopted by the target, however, it can
8244 make debugging impossible, since variables will no longer stay in
8245 a ``home register''.
8247 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8251 Perform tail duplication to enlarge superblock size. This transformation
8252 simplifies the control flow of the function allowing other optimizations to do
8255 Enabled with @option{-fprofile-use}.
8257 @item -funroll-loops
8258 @opindex funroll-loops
8259 Unroll loops whose number of iterations can be determined at compile time or
8260 upon entry to the loop. @option{-funroll-loops} implies
8261 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8262 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8263 small constant number of iterations). This option makes code larger, and may
8264 or may not make it run faster.
8266 Enabled with @option{-fprofile-use}.
8268 @item -funroll-all-loops
8269 @opindex funroll-all-loops
8270 Unroll all loops, even if their number of iterations is uncertain when
8271 the loop is entered. This usually makes programs run more slowly.
8272 @option{-funroll-all-loops} implies the same options as
8273 @option{-funroll-loops}.
8276 @opindex fpeel-loops
8277 Peels the loops for that there is enough information that they do not
8278 roll much (from profile feedback). It also turns on complete loop peeling
8279 (i.e.@: complete removal of loops with small constant number of iterations).
8281 Enabled with @option{-fprofile-use}.
8283 @item -fmove-loop-invariants
8284 @opindex fmove-loop-invariants
8285 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8286 at level @option{-O1}
8288 @item -funswitch-loops
8289 @opindex funswitch-loops
8290 Move branches with loop invariant conditions out of the loop, with duplicates
8291 of the loop on both branches (modified according to result of the condition).
8293 @item -ffunction-sections
8294 @itemx -fdata-sections
8295 @opindex ffunction-sections
8296 @opindex fdata-sections
8297 Place each function or data item into its own section in the output
8298 file if the target supports arbitrary sections. The name of the
8299 function or the name of the data item determines the section's name
8302 Use these options on systems where the linker can perform optimizations
8303 to improve locality of reference in the instruction space. Most systems
8304 using the ELF object format and SPARC processors running Solaris 2 have
8305 linkers with such optimizations. AIX may have these optimizations in
8308 Only use these options when there are significant benefits from doing
8309 so. When you specify these options, the assembler and linker will
8310 create larger object and executable files and will also be slower.
8311 You will not be able to use @code{gprof} on all systems if you
8312 specify this option and you may have problems with debugging if
8313 you specify both this option and @option{-g}.
8315 @item -fbranch-target-load-optimize
8316 @opindex fbranch-target-load-optimize
8317 Perform branch target register load optimization before prologue / epilogue
8319 The use of target registers can typically be exposed only during reload,
8320 thus hoisting loads out of loops and doing inter-block scheduling needs
8321 a separate optimization pass.
8323 @item -fbranch-target-load-optimize2
8324 @opindex fbranch-target-load-optimize2
8325 Perform branch target register load optimization after prologue / epilogue
8328 @item -fbtr-bb-exclusive
8329 @opindex fbtr-bb-exclusive
8330 When performing branch target register load optimization, don't reuse
8331 branch target registers in within any basic block.
8333 @item -fstack-protector
8334 @opindex fstack-protector
8335 Emit extra code to check for buffer overflows, such as stack smashing
8336 attacks. This is done by adding a guard variable to functions with
8337 vulnerable objects. This includes functions that call alloca, and
8338 functions with buffers larger than 8 bytes. The guards are initialized
8339 when a function is entered and then checked when the function exits.
8340 If a guard check fails, an error message is printed and the program exits.
8342 @item -fstack-protector-all
8343 @opindex fstack-protector-all
8344 Like @option{-fstack-protector} except that all functions are protected.
8346 @item -fsection-anchors
8347 @opindex fsection-anchors
8348 Try to reduce the number of symbolic address calculations by using
8349 shared ``anchor'' symbols to address nearby objects. This transformation
8350 can help to reduce the number of GOT entries and GOT accesses on some
8353 For example, the implementation of the following function @code{foo}:
8357 int foo (void) @{ return a + b + c; @}
8360 would usually calculate the addresses of all three variables, but if you
8361 compile it with @option{-fsection-anchors}, it will access the variables
8362 from a common anchor point instead. The effect is similar to the
8363 following pseudocode (which isn't valid C):
8368 register int *xr = &x;
8369 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8373 Not all targets support this option.
8375 @item --param @var{name}=@var{value}
8377 In some places, GCC uses various constants to control the amount of
8378 optimization that is done. For example, GCC will not inline functions
8379 that contain more that a certain number of instructions. You can
8380 control some of these constants on the command-line using the
8381 @option{--param} option.
8383 The names of specific parameters, and the meaning of the values, are
8384 tied to the internals of the compiler, and are subject to change
8385 without notice in future releases.
8387 In each case, the @var{value} is an integer. The allowable choices for
8388 @var{name} are given in the following table:
8391 @item predictable-branch-outcome
8392 When branch is predicted to be taken with probability lower than this threshold
8393 (in percent), then it is considered well predictable. The default is 10.
8395 @item max-crossjump-edges
8396 The maximum number of incoming edges to consider for crossjumping.
8397 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8398 the number of edges incoming to each block. Increasing values mean
8399 more aggressive optimization, making the compile time increase with
8400 probably small improvement in executable size.
8402 @item min-crossjump-insns
8403 The minimum number of instructions which must be matched at the end
8404 of two blocks before crossjumping will be performed on them. This
8405 value is ignored in the case where all instructions in the block being
8406 crossjumped from are matched. The default value is 5.
8408 @item max-grow-copy-bb-insns
8409 The maximum code size expansion factor when copying basic blocks
8410 instead of jumping. The expansion is relative to a jump instruction.
8411 The default value is 8.
8413 @item max-goto-duplication-insns
8414 The maximum number of instructions to duplicate to a block that jumps
8415 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8416 passes, GCC factors computed gotos early in the compilation process,
8417 and unfactors them as late as possible. Only computed jumps at the
8418 end of a basic blocks with no more than max-goto-duplication-insns are
8419 unfactored. The default value is 8.
8421 @item max-delay-slot-insn-search
8422 The maximum number of instructions to consider when looking for an
8423 instruction to fill a delay slot. If more than this arbitrary number of
8424 instructions is searched, the time savings from filling the delay slot
8425 will be minimal so stop searching. Increasing values mean more
8426 aggressive optimization, making the compile time increase with probably
8427 small improvement in executable run time.
8429 @item max-delay-slot-live-search
8430 When trying to fill delay slots, the maximum number of instructions to
8431 consider when searching for a block with valid live register
8432 information. Increasing this arbitrarily chosen value means more
8433 aggressive optimization, increasing the compile time. This parameter
8434 should be removed when the delay slot code is rewritten to maintain the
8437 @item max-gcse-memory
8438 The approximate maximum amount of memory that will be allocated in
8439 order to perform the global common subexpression elimination
8440 optimization. If more memory than specified is required, the
8441 optimization will not be done.
8443 @item max-gcse-insertion-ratio
8444 If the ratio of expression insertions to deletions is larger than this value
8445 for any expression, then RTL PRE will insert or remove the expression and thus
8446 leave partially redundant computations in the instruction stream. The default value is 20.
8448 @item max-pending-list-length
8449 The maximum number of pending dependencies scheduling will allow
8450 before flushing the current state and starting over. Large functions
8451 with few branches or calls can create excessively large lists which
8452 needlessly consume memory and resources.
8454 @item max-inline-insns-single
8455 Several parameters control the tree inliner used in gcc.
8456 This number sets the maximum number of instructions (counted in GCC's
8457 internal representation) in a single function that the tree inliner
8458 will consider for inlining. This only affects functions declared
8459 inline and methods implemented in a class declaration (C++).
8460 The default value is 400.
8462 @item max-inline-insns-auto
8463 When you use @option{-finline-functions} (included in @option{-O3}),
8464 a lot of functions that would otherwise not be considered for inlining
8465 by the compiler will be investigated. To those functions, a different
8466 (more restrictive) limit compared to functions declared inline can
8468 The default value is 40.
8470 @item large-function-insns
8471 The limit specifying really large functions. For functions larger than this
8472 limit after inlining, inlining is constrained by
8473 @option{--param large-function-growth}. This parameter is useful primarily
8474 to avoid extreme compilation time caused by non-linear algorithms used by the
8476 The default value is 2700.
8478 @item large-function-growth
8479 Specifies maximal growth of large function caused by inlining in percents.
8480 The default value is 100 which limits large function growth to 2.0 times
8483 @item large-unit-insns
8484 The limit specifying large translation unit. Growth caused by inlining of
8485 units larger than this limit is limited by @option{--param inline-unit-growth}.
8486 For small units this might be too tight (consider unit consisting of function A
8487 that is inline and B that just calls A three time. If B is small relative to
8488 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8489 large units consisting of small inlineable functions however the overall unit
8490 growth limit is needed to avoid exponential explosion of code size. Thus for
8491 smaller units, the size is increased to @option{--param large-unit-insns}
8492 before applying @option{--param inline-unit-growth}. The default is 10000
8494 @item inline-unit-growth
8495 Specifies maximal overall growth of the compilation unit caused by inlining.
8496 The default value is 30 which limits unit growth to 1.3 times the original
8499 @item ipcp-unit-growth
8500 Specifies maximal overall growth of the compilation unit caused by
8501 interprocedural constant propagation. The default value is 10 which limits
8502 unit growth to 1.1 times the original size.
8504 @item large-stack-frame
8505 The limit specifying large stack frames. While inlining the algorithm is trying
8506 to not grow past this limit too much. Default value is 256 bytes.
8508 @item large-stack-frame-growth
8509 Specifies maximal growth of large stack frames caused by inlining in percents.
8510 The default value is 1000 which limits large stack frame growth to 11 times
8513 @item max-inline-insns-recursive
8514 @itemx max-inline-insns-recursive-auto
8515 Specifies maximum number of instructions out-of-line copy of self recursive inline
8516 function can grow into by performing recursive inlining.
8518 For functions declared inline @option{--param max-inline-insns-recursive} is
8519 taken into account. For function not declared inline, recursive inlining
8520 happens only when @option{-finline-functions} (included in @option{-O3}) is
8521 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8522 default value is 450.
8524 @item max-inline-recursive-depth
8525 @itemx max-inline-recursive-depth-auto
8526 Specifies maximum recursion depth used by the recursive inlining.
8528 For functions declared inline @option{--param max-inline-recursive-depth} is
8529 taken into account. For function not declared inline, recursive inlining
8530 happens only when @option{-finline-functions} (included in @option{-O3}) is
8531 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8534 @item min-inline-recursive-probability
8535 Recursive inlining is profitable only for function having deep recursion
8536 in average and can hurt for function having little recursion depth by
8537 increasing the prologue size or complexity of function body to other
8540 When profile feedback is available (see @option{-fprofile-generate}) the actual
8541 recursion depth can be guessed from probability that function will recurse via
8542 given call expression. This parameter limits inlining only to call expression
8543 whose probability exceeds given threshold (in percents). The default value is
8546 @item early-inlining-insns
8547 Specify growth that early inliner can make. In effect it increases amount of
8548 inlining for code having large abstraction penalty. The default value is 10.
8550 @item max-early-inliner-iterations
8551 @itemx max-early-inliner-iterations
8552 Limit of iterations of early inliner. This basically bounds number of nested
8553 indirect calls early inliner can resolve. Deeper chains are still handled by
8556 @item comdat-sharing-probability
8557 @itemx comdat-sharing-probability
8558 Probability (in percent) that C++ inline function with comdat visibility
8559 will be shared across multiple compilation units. The default value is 20.
8561 @item min-vect-loop-bound
8562 The minimum number of iterations under which a loop will not get vectorized
8563 when @option{-ftree-vectorize} is used. The number of iterations after
8564 vectorization needs to be greater than the value specified by this option
8565 to allow vectorization. The default value is 0.
8567 @item gcse-cost-distance-ratio
8568 Scaling factor in calculation of maximum distance an expression
8569 can be moved by GCSE optimizations. This is currently supported only in the
8570 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8571 will be with simple expressions, i.e., the expressions which have cost
8572 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8573 hoisting of simple expressions. The default value is 10.
8575 @item gcse-unrestricted-cost
8576 Cost, roughly measured as the cost of a single typical machine
8577 instruction, at which GCSE optimizations will not constrain
8578 the distance an expression can travel. This is currently
8579 supported only in the code hoisting pass. The lesser the cost,
8580 the more aggressive code hoisting will be. Specifying 0 will
8581 allow all expressions to travel unrestricted distances.
8582 The default value is 3.
8584 @item max-hoist-depth
8585 The depth of search in the dominator tree for expressions to hoist.
8586 This is used to avoid quadratic behavior in hoisting algorithm.
8587 The value of 0 will avoid limiting the search, but may slow down compilation
8588 of huge functions. The default value is 30.
8590 @item max-unrolled-insns
8591 The maximum number of instructions that a loop should have if that loop
8592 is unrolled, and if the loop is unrolled, it determines how many times
8593 the loop code is unrolled.
8595 @item max-average-unrolled-insns
8596 The maximum number of instructions biased by probabilities of their execution
8597 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8598 it determines how many times the loop code is unrolled.
8600 @item max-unroll-times
8601 The maximum number of unrollings of a single loop.
8603 @item max-peeled-insns
8604 The maximum number of instructions that a loop should have if that loop
8605 is peeled, and if the loop is peeled, it determines how many times
8606 the loop code is peeled.
8608 @item max-peel-times
8609 The maximum number of peelings of a single loop.
8611 @item max-completely-peeled-insns
8612 The maximum number of insns of a completely peeled loop.
8614 @item max-completely-peel-times
8615 The maximum number of iterations of a loop to be suitable for complete peeling.
8617 @item max-completely-peel-loop-nest-depth
8618 The maximum depth of a loop nest suitable for complete peeling.
8620 @item max-unswitch-insns
8621 The maximum number of insns of an unswitched loop.
8623 @item max-unswitch-level
8624 The maximum number of branches unswitched in a single loop.
8627 The minimum cost of an expensive expression in the loop invariant motion.
8629 @item iv-consider-all-candidates-bound
8630 Bound on number of candidates for induction variables below that
8631 all candidates are considered for each use in induction variable
8632 optimizations. Only the most relevant candidates are considered
8633 if there are more candidates, to avoid quadratic time complexity.
8635 @item iv-max-considered-uses
8636 The induction variable optimizations give up on loops that contain more
8637 induction variable uses.
8639 @item iv-always-prune-cand-set-bound
8640 If number of candidates in the set is smaller than this value,
8641 we always try to remove unnecessary ivs from the set during its
8642 optimization when a new iv is added to the set.
8644 @item scev-max-expr-size
8645 Bound on size of expressions used in the scalar evolutions analyzer.
8646 Large expressions slow the analyzer.
8648 @item scev-max-expr-complexity
8649 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8650 Complex expressions slow the analyzer.
8652 @item omega-max-vars
8653 The maximum number of variables in an Omega constraint system.
8654 The default value is 128.
8656 @item omega-max-geqs
8657 The maximum number of inequalities in an Omega constraint system.
8658 The default value is 256.
8661 The maximum number of equalities in an Omega constraint system.
8662 The default value is 128.
8664 @item omega-max-wild-cards
8665 The maximum number of wildcard variables that the Omega solver will
8666 be able to insert. The default value is 18.
8668 @item omega-hash-table-size
8669 The size of the hash table in the Omega solver. The default value is
8672 @item omega-max-keys
8673 The maximal number of keys used by the Omega solver. The default
8676 @item omega-eliminate-redundant-constraints
8677 When set to 1, use expensive methods to eliminate all redundant
8678 constraints. The default value is 0.
8680 @item vect-max-version-for-alignment-checks
8681 The maximum number of runtime checks that can be performed when
8682 doing loop versioning for alignment in the vectorizer. See option
8683 ftree-vect-loop-version for more information.
8685 @item vect-max-version-for-alias-checks
8686 The maximum number of runtime checks that can be performed when
8687 doing loop versioning for alias in the vectorizer. See option
8688 ftree-vect-loop-version for more information.
8690 @item max-iterations-to-track
8692 The maximum number of iterations of a loop the brute force algorithm
8693 for analysis of # of iterations of the loop tries to evaluate.
8695 @item hot-bb-count-fraction
8696 Select fraction of the maximal count of repetitions of basic block in program
8697 given basic block needs to have to be considered hot.
8699 @item hot-bb-frequency-fraction
8700 Select fraction of the entry block frequency of executions of basic block in
8701 function given basic block needs to have to be considered hot.
8703 @item max-predicted-iterations
8704 The maximum number of loop iterations we predict statically. This is useful
8705 in cases where function contain single loop with known bound and other loop
8706 with unknown. We predict the known number of iterations correctly, while
8707 the unknown number of iterations average to roughly 10. This means that the
8708 loop without bounds would appear artificially cold relative to the other one.
8710 @item align-threshold
8712 Select fraction of the maximal frequency of executions of basic block in
8713 function given basic block will get aligned.
8715 @item align-loop-iterations
8717 A loop expected to iterate at lest the selected number of iterations will get
8720 @item tracer-dynamic-coverage
8721 @itemx tracer-dynamic-coverage-feedback
8723 This value is used to limit superblock formation once the given percentage of
8724 executed instructions is covered. This limits unnecessary code size
8727 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8728 feedback is available. The real profiles (as opposed to statically estimated
8729 ones) are much less balanced allowing the threshold to be larger value.
8731 @item tracer-max-code-growth
8732 Stop tail duplication once code growth has reached given percentage. This is
8733 rather hokey argument, as most of the duplicates will be eliminated later in
8734 cross jumping, so it may be set to much higher values than is the desired code
8737 @item tracer-min-branch-ratio
8739 Stop reverse growth when the reverse probability of best edge is less than this
8740 threshold (in percent).
8742 @item tracer-min-branch-ratio
8743 @itemx tracer-min-branch-ratio-feedback
8745 Stop forward growth if the best edge do have probability lower than this
8748 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8749 compilation for profile feedback and one for compilation without. The value
8750 for compilation with profile feedback needs to be more conservative (higher) in
8751 order to make tracer effective.
8753 @item max-cse-path-length
8755 Maximum number of basic blocks on path that cse considers. The default is 10.
8758 The maximum instructions CSE process before flushing. The default is 1000.
8760 @item ggc-min-expand
8762 GCC uses a garbage collector to manage its own memory allocation. This
8763 parameter specifies the minimum percentage by which the garbage
8764 collector's heap should be allowed to expand between collections.
8765 Tuning this may improve compilation speed; it has no effect on code
8768 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8769 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8770 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8771 GCC is not able to calculate RAM on a particular platform, the lower
8772 bound of 30% is used. Setting this parameter and
8773 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8774 every opportunity. This is extremely slow, but can be useful for
8777 @item ggc-min-heapsize
8779 Minimum size of the garbage collector's heap before it begins bothering
8780 to collect garbage. The first collection occurs after the heap expands
8781 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8782 tuning this may improve compilation speed, and has no effect on code
8785 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8786 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8787 with a lower bound of 4096 (four megabytes) and an upper bound of
8788 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8789 particular platform, the lower bound is used. Setting this parameter
8790 very large effectively disables garbage collection. Setting this
8791 parameter and @option{ggc-min-expand} to zero causes a full collection
8792 to occur at every opportunity.
8794 @item max-reload-search-insns
8795 The maximum number of instruction reload should look backward for equivalent
8796 register. Increasing values mean more aggressive optimization, making the
8797 compile time increase with probably slightly better performance. The default
8800 @item max-cselib-memory-locations
8801 The maximum number of memory locations cselib should take into account.
8802 Increasing values mean more aggressive optimization, making the compile time
8803 increase with probably slightly better performance. The default value is 500.
8805 @item reorder-blocks-duplicate
8806 @itemx reorder-blocks-duplicate-feedback
8808 Used by basic block reordering pass to decide whether to use unconditional
8809 branch or duplicate the code on its destination. Code is duplicated when its
8810 estimated size is smaller than this value multiplied by the estimated size of
8811 unconditional jump in the hot spots of the program.
8813 The @option{reorder-block-duplicate-feedback} is used only when profile
8814 feedback is available and may be set to higher values than
8815 @option{reorder-block-duplicate} since information about the hot spots is more
8818 @item max-sched-ready-insns
8819 The maximum number of instructions ready to be issued the scheduler should
8820 consider at any given time during the first scheduling pass. Increasing
8821 values mean more thorough searches, making the compilation time increase
8822 with probably little benefit. The default value is 100.
8824 @item max-sched-region-blocks
8825 The maximum number of blocks in a region to be considered for
8826 interblock scheduling. The default value is 10.
8828 @item max-pipeline-region-blocks
8829 The maximum number of blocks in a region to be considered for
8830 pipelining in the selective scheduler. The default value is 15.
8832 @item max-sched-region-insns
8833 The maximum number of insns in a region to be considered for
8834 interblock scheduling. The default value is 100.
8836 @item max-pipeline-region-insns
8837 The maximum number of insns in a region to be considered for
8838 pipelining in the selective scheduler. The default value is 200.
8841 The minimum probability (in percents) of reaching a source block
8842 for interblock speculative scheduling. The default value is 40.
8844 @item max-sched-extend-regions-iters
8845 The maximum number of iterations through CFG to extend regions.
8846 0 - disable region extension,
8847 N - do at most N iterations.
8848 The default value is 0.
8850 @item max-sched-insn-conflict-delay
8851 The maximum conflict delay for an insn to be considered for speculative motion.
8852 The default value is 3.
8854 @item sched-spec-prob-cutoff
8855 The minimal probability of speculation success (in percents), so that
8856 speculative insn will be scheduled.
8857 The default value is 40.
8859 @item sched-mem-true-dep-cost
8860 Minimal distance (in CPU cycles) between store and load targeting same
8861 memory locations. The default value is 1.
8863 @item selsched-max-lookahead
8864 The maximum size of the lookahead window of selective scheduling. It is a
8865 depth of search for available instructions.
8866 The default value is 50.
8868 @item selsched-max-sched-times
8869 The maximum number of times that an instruction will be scheduled during
8870 selective scheduling. This is the limit on the number of iterations
8871 through which the instruction may be pipelined. The default value is 2.
8873 @item selsched-max-insns-to-rename
8874 The maximum number of best instructions in the ready list that are considered
8875 for renaming in the selective scheduler. The default value is 2.
8878 The minimum value of stage count that swing modulo scheduler will
8879 generate. The default value is 2.
8881 @item max-last-value-rtl
8882 The maximum size measured as number of RTLs that can be recorded in an expression
8883 in combiner for a pseudo register as last known value of that register. The default
8886 @item integer-share-limit
8887 Small integer constants can use a shared data structure, reducing the
8888 compiler's memory usage and increasing its speed. This sets the maximum
8889 value of a shared integer constant. The default value is 256.
8891 @item min-virtual-mappings
8892 Specifies the minimum number of virtual mappings in the incremental
8893 SSA updater that should be registered to trigger the virtual mappings
8894 heuristic defined by virtual-mappings-ratio. The default value is
8897 @item virtual-mappings-ratio
8898 If the number of virtual mappings is virtual-mappings-ratio bigger
8899 than the number of virtual symbols to be updated, then the incremental
8900 SSA updater switches to a full update for those symbols. The default
8903 @item ssp-buffer-size
8904 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8905 protection when @option{-fstack-protection} is used.
8907 @item max-jump-thread-duplication-stmts
8908 Maximum number of statements allowed in a block that needs to be
8909 duplicated when threading jumps.
8911 @item max-fields-for-field-sensitive
8912 Maximum number of fields in a structure we will treat in
8913 a field sensitive manner during pointer analysis. The default is zero
8914 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8916 @item prefetch-latency
8917 Estimate on average number of instructions that are executed before
8918 prefetch finishes. The distance we prefetch ahead is proportional
8919 to this constant. Increasing this number may also lead to less
8920 streams being prefetched (see @option{simultaneous-prefetches}).
8922 @item simultaneous-prefetches
8923 Maximum number of prefetches that can run at the same time.
8925 @item l1-cache-line-size
8926 The size of cache line in L1 cache, in bytes.
8929 The size of L1 cache, in kilobytes.
8932 The size of L2 cache, in kilobytes.
8934 @item min-insn-to-prefetch-ratio
8935 The minimum ratio between the number of instructions and the
8936 number of prefetches to enable prefetching in a loop.
8938 @item prefetch-min-insn-to-mem-ratio
8939 The minimum ratio between the number of instructions and the
8940 number of memory references to enable prefetching in a loop.
8942 @item use-canonical-types
8943 Whether the compiler should use the ``canonical'' type system. By
8944 default, this should always be 1, which uses a more efficient internal
8945 mechanism for comparing types in C++ and Objective-C++. However, if
8946 bugs in the canonical type system are causing compilation failures,
8947 set this value to 0 to disable canonical types.
8949 @item switch-conversion-max-branch-ratio
8950 Switch initialization conversion will refuse to create arrays that are
8951 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8952 branches in the switch.
8954 @item max-partial-antic-length
8955 Maximum length of the partial antic set computed during the tree
8956 partial redundancy elimination optimization (@option{-ftree-pre}) when
8957 optimizing at @option{-O3} and above. For some sorts of source code
8958 the enhanced partial redundancy elimination optimization can run away,
8959 consuming all of the memory available on the host machine. This
8960 parameter sets a limit on the length of the sets that are computed,
8961 which prevents the runaway behavior. Setting a value of 0 for
8962 this parameter will allow an unlimited set length.
8964 @item sccvn-max-scc-size
8965 Maximum size of a strongly connected component (SCC) during SCCVN
8966 processing. If this limit is hit, SCCVN processing for the whole
8967 function will not be done and optimizations depending on it will
8968 be disabled. The default maximum SCC size is 10000.
8970 @item ira-max-loops-num
8971 IRA uses a regional register allocation by default. If a function
8972 contains loops more than number given by the parameter, only at most
8973 given number of the most frequently executed loops will form regions
8974 for the regional register allocation. The default value of the
8977 @item ira-max-conflict-table-size
8978 Although IRA uses a sophisticated algorithm of compression conflict
8979 table, the table can be still big for huge functions. If the conflict
8980 table for a function could be more than size in MB given by the
8981 parameter, the conflict table is not built and faster, simpler, and
8982 lower quality register allocation algorithm will be used. The
8983 algorithm do not use pseudo-register conflicts. The default value of
8984 the parameter is 2000.
8986 @item ira-loop-reserved-regs
8987 IRA can be used to evaluate more accurate register pressure in loops
8988 for decision to move loop invariants (see @option{-O3}). The number
8989 of available registers reserved for some other purposes is described
8990 by this parameter. The default value of the parameter is 2 which is
8991 minimal number of registers needed for execution of typical
8992 instruction. This value is the best found from numerous experiments.
8994 @item loop-invariant-max-bbs-in-loop
8995 Loop invariant motion can be very expensive, both in compile time and
8996 in amount of needed compile time memory, with very large loops. Loops
8997 with more basic blocks than this parameter won't have loop invariant
8998 motion optimization performed on them. The default value of the
8999 parameter is 1000 for -O1 and 10000 for -O2 and above.
9001 @item max-vartrack-size
9002 Sets a maximum number of hash table slots to use during variable
9003 tracking dataflow analysis of any function. If this limit is exceeded
9004 with variable tracking at assignments enabled, analysis for that
9005 function is retried without it, after removing all debug insns from
9006 the function. If the limit is exceeded even without debug insns, var
9007 tracking analysis is completely disabled for the function. Setting
9008 the parameter to zero makes it unlimited.
9010 @item max-vartrack-expr-depth
9011 Sets a maximum number of recursion levels when attempting to map
9012 variable names or debug temporaries to value expressions. This trades
9013 compile time for more complete debug information. If this is set too
9014 low, value expressions that are available and could be represented in
9015 debug information may end up not being used; setting this higher may
9016 enable the compiler to find more complex debug expressions, but compile
9017 time may grow exponentially, and even then, it may fail to find more
9018 usable expressions. The default is 10.
9020 @item min-nondebug-insn-uid
9021 Use uids starting at this parameter for nondebug insns. The range below
9022 the parameter is reserved exclusively for debug insns created by
9023 @option{-fvar-tracking-assignments}, but debug insns may get
9024 (non-overlapping) uids above it if the reserved range is exhausted.
9026 @item ipa-sra-ptr-growth-factor
9027 IPA-SRA will replace a pointer to an aggregate with one or more new
9028 parameters only when their cumulative size is less or equal to
9029 @option{ipa-sra-ptr-growth-factor} times the size of the original
9032 @item graphite-max-nb-scop-params
9033 To avoid exponential effects in the Graphite loop transforms, the
9034 number of parameters in a Static Control Part (SCoP) is bounded. The
9035 default value is 10 parameters. A variable whose value is unknown at
9036 compile time and defined outside a SCoP is a parameter of the SCoP.
9038 @item graphite-max-bbs-per-function
9039 To avoid exponential effects in the detection of SCoPs, the size of
9040 the functions analyzed by Graphite is bounded. The default value is
9043 @item loop-block-tile-size
9044 Loop blocking or strip mining transforms, enabled with
9045 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9046 loop in the loop nest by a given number of iterations. The strip
9047 length can be changed using the @option{loop-block-tile-size}
9048 parameter. The default value is 51 iterations.
9050 @item ipa-cp-value-list-size
9051 IPA-CP attempts to track all possible values and types passed to a function's
9052 parameter in order to propagate them and perform devirtualization.
9053 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9054 stores per one formal parameter of a function.
9056 @item lto-partitions
9057 Specify desired number of partitions produced during WHOPR compilation.
9058 The number of partitions should exceed the number of CPUs used for compilation.
9059 The default value is 32.
9061 @item lto-minpartition
9062 Size of minimal partition for WHOPR (in estimated instructions).
9063 This prevents expenses of splitting very small programs into too many
9066 @item cxx-max-namespaces-for-diagnostic-help
9067 The maximum number of namespaces to consult for suggestions when C++
9068 name lookup fails for an identifier. The default is 1000.
9070 @item max-stores-to-sink
9071 The maximum number of conditional stores paires that can be sunk. Set to 0
9072 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9073 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9075 @item allow-store-data-races
9076 Allow optimizers to introduce new data races on stores.
9077 Set to 1 to allow, otherwise to 0. This option is enabled by default
9078 unless implicitly set by the @option{-fmemory-model=} option.
9080 @item case-values-threshold
9081 The smallest number of different values for which it is best to use a
9082 jump-table instead of a tree of conditional branches. If the value is
9083 0, use the default for the machine. The default is 0.
9085 @item tree-reassoc-width
9086 Set the maximum number of instructions executed in parallel in
9087 reassociated tree. This parameter overrides target dependent
9088 heuristics used by default if has non zero value.
9093 @node Preprocessor Options
9094 @section Options Controlling the Preprocessor
9095 @cindex preprocessor options
9096 @cindex options, preprocessor
9098 These options control the C preprocessor, which is run on each C source
9099 file before actual compilation.
9101 If you use the @option{-E} option, nothing is done except preprocessing.
9102 Some of these options make sense only together with @option{-E} because
9103 they cause the preprocessor output to be unsuitable for actual
9107 @item -Wp,@var{option}
9109 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9110 and pass @var{option} directly through to the preprocessor. If
9111 @var{option} contains commas, it is split into multiple options at the
9112 commas. However, many options are modified, translated or interpreted
9113 by the compiler driver before being passed to the preprocessor, and
9114 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9115 interface is undocumented and subject to change, so whenever possible
9116 you should avoid using @option{-Wp} and let the driver handle the
9119 @item -Xpreprocessor @var{option}
9120 @opindex Xpreprocessor
9121 Pass @var{option} as an option to the preprocessor. You can use this to
9122 supply system-specific preprocessor 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{-Xpreprocessor} twice, once for the option and once for the argument.
9129 @include cppopts.texi
9131 @node Assembler Options
9132 @section Passing Options to the Assembler
9134 @c prevent bad page break with this line
9135 You can pass options to the assembler.
9138 @item -Wa,@var{option}
9140 Pass @var{option} as an option to the assembler. If @var{option}
9141 contains commas, it is split into multiple options at the commas.
9143 @item -Xassembler @var{option}
9145 Pass @var{option} as an option to the assembler. You can use this to
9146 supply system-specific assembler options which GCC does not know how to
9149 If you want to pass an option that takes an argument, you must use
9150 @option{-Xassembler} twice, once for the option and once for the argument.
9155 @section Options for Linking
9156 @cindex link options
9157 @cindex options, linking
9159 These options come into play when the compiler links object files into
9160 an executable output file. They are meaningless if the compiler is
9161 not doing a link step.
9165 @item @var{object-file-name}
9166 A file name that does not end in a special recognized suffix is
9167 considered to name an object file or library. (Object files are
9168 distinguished from libraries by the linker according to the file
9169 contents.) If linking is done, these object files are used as input
9178 If any of these options is used, then the linker is not run, and
9179 object file names should not be used as arguments. @xref{Overall
9183 @item -l@var{library}
9184 @itemx -l @var{library}
9186 Search the library named @var{library} when linking. (The second
9187 alternative with the library as a separate argument is only for
9188 POSIX compliance and is not recommended.)
9190 It makes a difference where in the command you write this option; the
9191 linker searches and processes libraries and object files in the order they
9192 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9193 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9194 to functions in @samp{z}, those functions may not be loaded.
9196 The linker searches a standard list of directories for the library,
9197 which is actually a file named @file{lib@var{library}.a}. The linker
9198 then uses this file as if it had been specified precisely by name.
9200 The directories searched include several standard system directories
9201 plus any that you specify with @option{-L}.
9203 Normally the files found this way are library files---archive files
9204 whose members are object files. The linker handles an archive file by
9205 scanning through it for members which define symbols that have so far
9206 been referenced but not defined. But if the file that is found is an
9207 ordinary object file, it is linked in the usual fashion. The only
9208 difference between using an @option{-l} option and specifying a file name
9209 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9210 and searches several directories.
9214 You need this special case of the @option{-l} option in order to
9215 link an Objective-C or Objective-C++ program.
9218 @opindex nostartfiles
9219 Do not use the standard system startup files when linking.
9220 The standard system libraries are used normally, unless @option{-nostdlib}
9221 or @option{-nodefaultlibs} is used.
9223 @item -nodefaultlibs
9224 @opindex nodefaultlibs
9225 Do not use the standard system libraries when linking.
9226 Only the libraries you specify will be passed to the linker, options
9227 specifying linkage of the system libraries, such as @code{-static-libgcc}
9228 or @code{-shared-libgcc}, will be ignored.
9229 The standard startup files are used normally, unless @option{-nostartfiles}
9230 is used. The compiler may generate calls to @code{memcmp},
9231 @code{memset}, @code{memcpy} and @code{memmove}.
9232 These entries are usually resolved by entries in
9233 libc. These entry points should be supplied through some other
9234 mechanism when this option is specified.
9238 Do not use the standard system startup files or libraries when linking.
9239 No startup files and only the libraries you specify will be passed to
9240 the linker, options specifying linkage of the system libraries, such as
9241 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9242 The compiler may generate calls to @code{memcmp}, @code{memset},
9243 @code{memcpy} and @code{memmove}.
9244 These entries are usually resolved by entries in
9245 libc. These entry points should be supplied through some other
9246 mechanism when this option is specified.
9248 @cindex @option{-lgcc}, use with @option{-nostdlib}
9249 @cindex @option{-nostdlib} and unresolved references
9250 @cindex unresolved references and @option{-nostdlib}
9251 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9252 @cindex @option{-nodefaultlibs} and unresolved references
9253 @cindex unresolved references and @option{-nodefaultlibs}
9254 One of the standard libraries bypassed by @option{-nostdlib} and
9255 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9256 that GCC uses to overcome shortcomings of particular machines, or special
9257 needs for some languages.
9258 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9259 Collection (GCC) Internals},
9260 for more discussion of @file{libgcc.a}.)
9261 In most cases, you need @file{libgcc.a} even when you want to avoid
9262 other standard libraries. In other words, when you specify @option{-nostdlib}
9263 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9264 This ensures that you have no unresolved references to internal GCC
9265 library subroutines. (For example, @samp{__main}, used to ensure C++
9266 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9267 GNU Compiler Collection (GCC) Internals}.)
9271 Produce a position independent executable on targets which support it.
9272 For predictable results, you must also specify the same set of options
9273 that were used to generate code (@option{-fpie}, @option{-fPIE},
9274 or model suboptions) when you specify this option.
9278 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9279 that support it. This instructs the linker to add all symbols, not
9280 only used ones, to the dynamic symbol table. This option is needed
9281 for some uses of @code{dlopen} or to allow obtaining backtraces
9282 from within a program.
9286 Remove all symbol table and relocation information from the executable.
9290 On systems that support dynamic linking, this prevents linking with the shared
9291 libraries. On other systems, this option has no effect.
9295 Produce a shared object which can then be linked with other objects to
9296 form an executable. Not all systems support this option. For predictable
9297 results, you must also specify the same set of options that were used to
9298 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9299 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9300 needs to build supplementary stub code for constructors to work. On
9301 multi-libbed systems, @samp{gcc -shared} must select the correct support
9302 libraries to link against. Failing to supply the correct flags may lead
9303 to subtle defects. Supplying them in cases where they are not necessary
9306 @item -shared-libgcc
9307 @itemx -static-libgcc
9308 @opindex shared-libgcc
9309 @opindex static-libgcc
9310 On systems that provide @file{libgcc} as a shared library, these options
9311 force the use of either the shared or static version respectively.
9312 If no shared version of @file{libgcc} was built when the compiler was
9313 configured, these options have no effect.
9315 There are several situations in which an application should use the
9316 shared @file{libgcc} instead of the static version. The most common
9317 of these is when the application wishes to throw and catch exceptions
9318 across different shared libraries. In that case, each of the libraries
9319 as well as the application itself should use the shared @file{libgcc}.
9321 Therefore, the G++ and GCJ drivers automatically add
9322 @option{-shared-libgcc} whenever you build a shared library or a main
9323 executable, because C++ and Java programs typically use exceptions, so
9324 this is the right thing to do.
9326 If, instead, you use the GCC driver to create shared libraries, you may
9327 find that they will not always be linked with the shared @file{libgcc}.
9328 If GCC finds, at its configuration time, that you have a non-GNU linker
9329 or a GNU linker that does not support option @option{--eh-frame-hdr},
9330 it will link the shared version of @file{libgcc} into shared libraries
9331 by default. Otherwise, it will take advantage of the linker and optimize
9332 away the linking with the shared version of @file{libgcc}, linking with
9333 the static version of libgcc by default. This allows exceptions to
9334 propagate through such shared libraries, without incurring relocation
9335 costs at library load time.
9337 However, if a library or main executable is supposed to throw or catch
9338 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9339 for the languages used in the program, or using the option
9340 @option{-shared-libgcc}, such that it is linked with the shared
9343 @item -static-libstdc++
9344 When the @command{g++} program is used to link a C++ program, it will
9345 normally automatically link against @option{libstdc++}. If
9346 @file{libstdc++} is available as a shared library, and the
9347 @option{-static} option is not used, then this will link against the
9348 shared version of @file{libstdc++}. That is normally fine. However, it
9349 is sometimes useful to freeze the version of @file{libstdc++} used by
9350 the program without going all the way to a fully static link. The
9351 @option{-static-libstdc++} option directs the @command{g++} driver to
9352 link @file{libstdc++} statically, without necessarily linking other
9353 libraries statically.
9357 Bind references to global symbols when building a shared object. Warn
9358 about any unresolved references (unless overridden by the link editor
9359 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9362 @item -T @var{script}
9364 @cindex linker script
9365 Use @var{script} as the linker script. This option is supported by most
9366 systems using the GNU linker. On some targets, such as bare-board
9367 targets without an operating system, the @option{-T} option may be required
9368 when linking to avoid references to undefined symbols.
9370 @item -Xlinker @var{option}
9372 Pass @var{option} as an option to the linker. You can use this to
9373 supply system-specific linker options which GCC does not know how to
9376 If you want to pass an option that takes a separate argument, you must use
9377 @option{-Xlinker} twice, once for the option and once for the argument.
9378 For example, to pass @option{-assert definitions}, you must write
9379 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9380 @option{-Xlinker "-assert definitions"}, because this passes the entire
9381 string as a single argument, which is not what the linker expects.
9383 When using the GNU linker, it is usually more convenient to pass
9384 arguments to linker options using the @option{@var{option}=@var{value}}
9385 syntax than as separate arguments. For example, you can specify
9386 @samp{-Xlinker -Map=output.map} rather than
9387 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9388 this syntax for command-line options.
9390 @item -Wl,@var{option}
9392 Pass @var{option} as an option to the linker. If @var{option} contains
9393 commas, it is split into multiple options at the commas. You can use this
9394 syntax to pass an argument to the option.
9395 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9396 linker. When using the GNU linker, you can also get the same effect with
9397 @samp{-Wl,-Map=output.map}.
9399 @item -u @var{symbol}
9401 Pretend the symbol @var{symbol} is undefined, to force linking of
9402 library modules to define it. You can use @option{-u} multiple times with
9403 different symbols to force loading of additional library modules.
9406 @node Directory Options
9407 @section Options for Directory Search
9408 @cindex directory options
9409 @cindex options, directory search
9412 These options specify directories to search for header files, for
9413 libraries and for parts of the compiler:
9418 Add the directory @var{dir} to the head of the list of directories to be
9419 searched for header files. This can be used to override a system header
9420 file, substituting your own version, since these directories are
9421 searched before the system header file directories. However, you should
9422 not use this option to add directories that contain vendor-supplied
9423 system header files (use @option{-isystem} for that). If you use more than
9424 one @option{-I} option, the directories are scanned in left-to-right
9425 order; the standard system directories come after.
9427 If a standard system include directory, or a directory specified with
9428 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9429 option will be ignored. The directory will still be searched but as a
9430 system directory at its normal position in the system include chain.
9431 This is to ensure that GCC's procedure to fix buggy system headers and
9432 the ordering for the include_next directive are not inadvertently changed.
9433 If you really need to change the search order for system directories,
9434 use the @option{-nostdinc} and/or @option{-isystem} options.
9436 @item -iplugindir=@var{dir}
9437 Set the directory to search for plugins which are passed
9438 by @option{-fplugin=@var{name}} instead of
9439 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9440 to be used by the user, but only passed by the driver.
9442 @item -iquote@var{dir}
9444 Add the directory @var{dir} to the head of the list of directories to
9445 be searched for header files only for the case of @samp{#include
9446 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9447 otherwise just like @option{-I}.
9451 Add directory @var{dir} to the list of directories to be searched
9454 @item -B@var{prefix}
9456 This option specifies where to find the executables, libraries,
9457 include files, and data files of the compiler itself.
9459 The compiler driver program runs one or more of the subprograms
9460 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9461 @var{prefix} as a prefix for each program it tries to run, both with and
9462 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9464 For each subprogram to be run, the compiler driver first tries the
9465 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9466 was not specified, the driver tries two standard prefixes, which are
9467 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9468 those results in a file name that is found, the unmodified program
9469 name is searched for using the directories specified in your
9470 @env{PATH} environment variable.
9472 The compiler will check to see if the path provided by the @option{-B}
9473 refers to a directory, and if necessary it will add a directory
9474 separator character at the end of the path.
9476 @option{-B} prefixes that effectively specify directory names also apply
9477 to libraries in the linker, because the compiler translates these
9478 options into @option{-L} options for the linker. They also apply to
9479 includes files in the preprocessor, because the compiler translates these
9480 options into @option{-isystem} options for the preprocessor. In this case,
9481 the compiler appends @samp{include} to the prefix.
9483 The run-time support file @file{libgcc.a} can also be searched for using
9484 the @option{-B} prefix, if needed. If it is not found there, the two
9485 standard prefixes above are tried, and that is all. The file is left
9486 out of the link if it is not found by those means.
9488 Another way to specify a prefix much like the @option{-B} prefix is to use
9489 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9492 As a special kludge, if the path provided by @option{-B} is
9493 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9494 9, then it will be replaced by @file{[dir/]include}. This is to help
9495 with boot-strapping the compiler.
9497 @item -specs=@var{file}
9499 Process @var{file} after the compiler reads in the standard @file{specs}
9500 file, in order to override the defaults that the @file{gcc} driver
9501 program uses when determining what switches to pass to @file{cc1},
9502 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9503 @option{-specs=@var{file}} can be specified on the command line, and they
9504 are processed in order, from left to right.
9506 @item --sysroot=@var{dir}
9508 Use @var{dir} as the logical root directory for headers and libraries.
9509 For example, if the compiler would normally search for headers in
9510 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9511 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9513 If you use both this option and the @option{-isysroot} option, then
9514 the @option{--sysroot} option will apply to libraries, but the
9515 @option{-isysroot} option will apply to header files.
9517 The GNU linker (beginning with version 2.16) has the necessary support
9518 for this option. If your linker does not support this option, the
9519 header file aspect of @option{--sysroot} will still work, but the
9520 library aspect will not.
9524 This option has been deprecated. Please use @option{-iquote} instead for
9525 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9526 Any directories you specify with @option{-I} options before the @option{-I-}
9527 option are searched only for the case of @samp{#include "@var{file}"};
9528 they are not searched for @samp{#include <@var{file}>}.
9530 If additional directories are specified with @option{-I} options after
9531 the @option{-I-}, these directories are searched for all @samp{#include}
9532 directives. (Ordinarily @emph{all} @option{-I} directories are used
9535 In addition, the @option{-I-} option inhibits the use of the current
9536 directory (where the current input file came from) as the first search
9537 directory for @samp{#include "@var{file}"}. There is no way to
9538 override this effect of @option{-I-}. With @option{-I.} you can specify
9539 searching the directory which was current when the compiler was
9540 invoked. That is not exactly the same as what the preprocessor does
9541 by default, but it is often satisfactory.
9543 @option{-I-} does not inhibit the use of the standard system directories
9544 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9551 @section Specifying subprocesses and the switches to pass to them
9554 @command{gcc} is a driver program. It performs its job by invoking a
9555 sequence of other programs to do the work of compiling, assembling and
9556 linking. GCC interprets its command-line parameters and uses these to
9557 deduce which programs it should invoke, and which command-line options
9558 it ought to place on their command lines. This behavior is controlled
9559 by @dfn{spec strings}. In most cases there is one spec string for each
9560 program that GCC can invoke, but a few programs have multiple spec
9561 strings to control their behavior. The spec strings built into GCC can
9562 be overridden by using the @option{-specs=} command-line switch to specify
9565 @dfn{Spec files} are plaintext files that are used to construct spec
9566 strings. They consist of a sequence of directives separated by blank
9567 lines. The type of directive is determined by the first non-whitespace
9568 character on the line and it can be one of the following:
9571 @item %@var{command}
9572 Issues a @var{command} to the spec file processor. The commands that can
9576 @item %include <@var{file}>
9577 @cindex @code{%include}
9578 Search for @var{file} and insert its text at the current point in the
9581 @item %include_noerr <@var{file}>
9582 @cindex @code{%include_noerr}
9583 Just like @samp{%include}, but do not generate an error message if the include
9584 file cannot be found.
9586 @item %rename @var{old_name} @var{new_name}
9587 @cindex @code{%rename}
9588 Rename the spec string @var{old_name} to @var{new_name}.
9592 @item *[@var{spec_name}]:
9593 This tells the compiler to create, override or delete the named spec
9594 string. All lines after this directive up to the next directive or
9595 blank line are considered to be the text for the spec string. If this
9596 results in an empty string then the spec will be deleted. (Or, if the
9597 spec did not exist, then nothing will happen.) Otherwise, if the spec
9598 does not currently exist a new spec will be created. If the spec does
9599 exist then its contents will be overridden by the text of this
9600 directive, unless the first character of that text is the @samp{+}
9601 character, in which case the text will be appended to the spec.
9603 @item [@var{suffix}]:
9604 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9605 and up to the next directive or blank line are considered to make up the
9606 spec string for the indicated suffix. When the compiler encounters an
9607 input file with the named suffix, it will processes the spec string in
9608 order to work out how to compile that file. For example:
9615 This says that any input file whose name ends in @samp{.ZZ} should be
9616 passed to the program @samp{z-compile}, which should be invoked with the
9617 command-line switch @option{-input} and with the result of performing the
9618 @samp{%i} substitution. (See below.)
9620 As an alternative to providing a spec string, the text that follows a
9621 suffix directive can be one of the following:
9624 @item @@@var{language}
9625 This says that the suffix is an alias for a known @var{language}. This is
9626 similar to using the @option{-x} command-line switch to GCC to specify a
9627 language explicitly. For example:
9634 Says that .ZZ files are, in fact, C++ source files.
9637 This causes an error messages saying:
9640 @var{name} compiler not installed on this system.
9644 GCC already has an extensive list of suffixes built into it.
9645 This directive will add an entry to the end of the list of suffixes, but
9646 since the list is searched from the end backwards, it is effectively
9647 possible to override earlier entries using this technique.
9651 GCC has the following spec strings built into it. Spec files can
9652 override these strings or create their own. Note that individual
9653 targets can also add their own spec strings to this list.
9656 asm Options to pass to the assembler
9657 asm_final Options to pass to the assembler post-processor
9658 cpp Options to pass to the C preprocessor
9659 cc1 Options to pass to the C compiler
9660 cc1plus Options to pass to the C++ compiler
9661 endfile Object files to include at the end of the link
9662 link Options to pass to the linker
9663 lib Libraries to include on the command line to the linker
9664 libgcc Decides which GCC support library to pass to the linker
9665 linker Sets the name of the linker
9666 predefines Defines to be passed to the C preprocessor
9667 signed_char Defines to pass to CPP to say whether @code{char} is signed
9669 startfile Object files to include at the start of the link
9672 Here is a small example of a spec file:
9678 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9681 This example renames the spec called @samp{lib} to @samp{old_lib} and
9682 then overrides the previous definition of @samp{lib} with a new one.
9683 The new definition adds in some extra command-line options before
9684 including the text of the old definition.
9686 @dfn{Spec strings} are a list of command-line options to be passed to their
9687 corresponding program. In addition, the spec strings can contain
9688 @samp{%}-prefixed sequences to substitute variable text or to
9689 conditionally insert text into the command line. Using these constructs
9690 it is possible to generate quite complex command lines.
9692 Here is a table of all defined @samp{%}-sequences for spec
9693 strings. Note that spaces are not generated automatically around the
9694 results of expanding these sequences. Therefore you can concatenate them
9695 together or combine them with constant text in a single argument.
9699 Substitute one @samp{%} into the program name or argument.
9702 Substitute the name of the input file being processed.
9705 Substitute the basename of the input file being processed.
9706 This is the substring up to (and not including) the last period
9707 and not including the directory.
9710 This is the same as @samp{%b}, but include the file suffix (text after
9714 Marks the argument containing or following the @samp{%d} as a
9715 temporary file name, so that that file will be deleted if GCC exits
9716 successfully. Unlike @samp{%g}, this contributes no text to the
9719 @item %g@var{suffix}
9720 Substitute a file name that has suffix @var{suffix} and is chosen
9721 once per compilation, and mark the argument in the same way as
9722 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9723 name is now chosen in a way that is hard to predict even when previously
9724 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9725 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9726 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9727 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9728 was simply substituted with a file name chosen once per compilation,
9729 without regard to any appended suffix (which was therefore treated
9730 just like ordinary text), making such attacks more likely to succeed.
9732 @item %u@var{suffix}
9733 Like @samp{%g}, but generates a new temporary file name even if
9734 @samp{%u@var{suffix}} was already seen.
9736 @item %U@var{suffix}
9737 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9738 new one if there is no such last file name. In the absence of any
9739 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9740 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9741 would involve the generation of two distinct file names, one
9742 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9743 simply substituted with a file name chosen for the previous @samp{%u},
9744 without regard to any appended suffix.
9746 @item %j@var{suffix}
9747 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9748 writable, and if save-temps is off; otherwise, substitute the name
9749 of a temporary file, just like @samp{%u}. This temporary file is not
9750 meant for communication between processes, but rather as a junk
9753 @item %|@var{suffix}
9754 @itemx %m@var{suffix}
9755 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9756 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9757 all. These are the two most common ways to instruct a program that it
9758 should read from standard input or write to standard output. If you
9759 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9760 construct: see for example @file{f/lang-specs.h}.
9762 @item %.@var{SUFFIX}
9763 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9764 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9765 terminated by the next space or %.
9768 Marks the argument containing or following the @samp{%w} as the
9769 designated output file of this compilation. This puts the argument
9770 into the sequence of arguments that @samp{%o} will substitute later.
9773 Substitutes the names of all the output files, with spaces
9774 automatically placed around them. You should write spaces
9775 around the @samp{%o} as well or the results are undefined.
9776 @samp{%o} is for use in the specs for running the linker.
9777 Input files whose names have no recognized suffix are not compiled
9778 at all, but they are included among the output files, so they will
9782 Substitutes the suffix for object files. Note that this is
9783 handled specially when it immediately follows @samp{%g, %u, or %U},
9784 because of the need for those to form complete file names. The
9785 handling is such that @samp{%O} is treated exactly as if it had already
9786 been substituted, except that @samp{%g, %u, and %U} do not currently
9787 support additional @var{suffix} characters following @samp{%O} as they would
9788 following, for example, @samp{.o}.
9791 Substitutes the standard macro predefinitions for the
9792 current target machine. Use this when running @code{cpp}.
9795 Like @samp{%p}, but puts @samp{__} before and after the name of each
9796 predefined macro, except for macros that start with @samp{__} or with
9797 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9801 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9802 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9803 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9804 and @option{-imultilib} as necessary.
9807 Current argument is the name of a library or startup file of some sort.
9808 Search for that file in a standard list of directories and substitute
9809 the full name found. The current working directory is included in the
9810 list of directories scanned.
9813 Current argument is the name of a linker script. Search for that file
9814 in the current list of directories to scan for libraries. If the file
9815 is located insert a @option{--script} option into the command line
9816 followed by the full path name found. If the file is not found then
9817 generate an error message. Note: the current working directory is not
9821 Print @var{str} as an error message. @var{str} is terminated by a newline.
9822 Use this when inconsistent options are detected.
9825 Substitute the contents of spec string @var{name} at this point.
9827 @item %x@{@var{option}@}
9828 Accumulate an option for @samp{%X}.
9831 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9835 Output the accumulated assembler options specified by @option{-Wa}.
9838 Output the accumulated preprocessor options specified by @option{-Wp}.
9841 Process the @code{asm} spec. This is used to compute the
9842 switches to be passed to the assembler.
9845 Process the @code{asm_final} spec. This is a spec string for
9846 passing switches to an assembler post-processor, if such a program is
9850 Process the @code{link} spec. This is the spec for computing the
9851 command line passed to the linker. Typically it will make use of the
9852 @samp{%L %G %S %D and %E} sequences.
9855 Dump out a @option{-L} option for each directory that GCC believes might
9856 contain startup files. If the target supports multilibs then the
9857 current multilib directory will be prepended to each of these paths.
9860 Process the @code{lib} spec. This is a spec string for deciding which
9861 libraries should be included on the command line to the linker.
9864 Process the @code{libgcc} spec. This is a spec string for deciding
9865 which GCC support library should be included on the command line to the linker.
9868 Process the @code{startfile} spec. This is a spec for deciding which
9869 object files should be the first ones passed to the linker. Typically
9870 this might be a file named @file{crt0.o}.
9873 Process the @code{endfile} spec. This is a spec string that specifies
9874 the last object files that will be passed to the linker.
9877 Process the @code{cpp} spec. This is used to construct the arguments
9878 to be passed to the C preprocessor.
9881 Process the @code{cc1} spec. This is used to construct the options to be
9882 passed to the actual C compiler (@samp{cc1}).
9885 Process the @code{cc1plus} spec. This is used to construct the options to be
9886 passed to the actual C++ compiler (@samp{cc1plus}).
9889 Substitute the variable part of a matched option. See below.
9890 Note that each comma in the substituted string is replaced by
9894 Remove all occurrences of @code{-S} from the command line. Note---this
9895 command is position dependent. @samp{%} commands in the spec string
9896 before this one will see @code{-S}, @samp{%} commands in the spec string
9897 after this one will not.
9899 @item %:@var{function}(@var{args})
9900 Call the named function @var{function}, passing it @var{args}.
9901 @var{args} is first processed as a nested spec string, then split
9902 into an argument vector in the usual fashion. The function returns
9903 a string which is processed as if it had appeared literally as part
9904 of the current spec.
9906 The following built-in spec functions are provided:
9910 The @code{getenv} spec function takes two arguments: an environment
9911 variable name and a string. If the environment variable is not
9912 defined, a fatal error is issued. Otherwise, the return value is the
9913 value of the environment variable concatenated with the string. For
9914 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9917 %:getenv(TOPDIR /include)
9920 expands to @file{/path/to/top/include}.
9922 @item @code{if-exists}
9923 The @code{if-exists} spec function takes one argument, an absolute
9924 pathname to a file. If the file exists, @code{if-exists} returns the
9925 pathname. Here is a small example of its usage:
9929 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9932 @item @code{if-exists-else}
9933 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9934 spec function, except that it takes two arguments. The first argument is
9935 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9936 returns the pathname. If it does not exist, it returns the second argument.
9937 This way, @code{if-exists-else} can be used to select one file or another,
9938 based on the existence of the first. Here is a small example of its usage:
9942 crt0%O%s %:if-exists(crti%O%s) \
9943 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9946 @item @code{replace-outfile}
9947 The @code{replace-outfile} spec function takes two arguments. It looks for the
9948 first argument in the outfiles array and replaces it with the second argument. Here
9949 is a small example of its usage:
9952 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9955 @item @code{remove-outfile}
9956 The @code{remove-outfile} spec function takes one argument. It looks for the
9957 first argument in the outfiles array and removes it. Here is a small example
9961 %:remove-outfile(-lm)
9964 @item @code{pass-through-libs}
9965 The @code{pass-through-libs} spec function takes any number of arguments. It
9966 finds any @option{-l} options and any non-options ending in ".a" (which it
9967 assumes are the names of linker input library archive files) and returns a
9968 result containing all the found arguments each prepended by
9969 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9970 intended to be passed to the LTO linker plugin.
9973 %:pass-through-libs(%G %L %G)
9976 @item @code{print-asm-header}
9977 The @code{print-asm-header} function takes no arguments and simply
9978 prints a banner like:
9984 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9987 It is used to separate compiler options from assembler options
9988 in the @option{--target-help} output.
9992 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9993 If that switch was not specified, this substitutes nothing. Note that
9994 the leading dash is omitted when specifying this option, and it is
9995 automatically inserted if the substitution is performed. Thus the spec
9996 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9997 and would output the command line option @option{-foo}.
9999 @item %W@{@code{S}@}
10000 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10001 deleted on failure.
10003 @item %@{@code{S}*@}
10004 Substitutes all the switches specified to GCC whose names start
10005 with @code{-S}, but which also take an argument. This is used for
10006 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10007 GCC considers @option{-o foo} as being
10008 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10009 text, including the space. Thus two arguments would be generated.
10011 @item %@{@code{S}*&@code{T}*@}
10012 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10013 (the order of @code{S} and @code{T} in the spec is not significant).
10014 There can be any number of ampersand-separated variables; for each the
10015 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10017 @item %@{@code{S}:@code{X}@}
10018 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10020 @item %@{!@code{S}:@code{X}@}
10021 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10023 @item %@{@code{S}*:@code{X}@}
10024 Substitutes @code{X} if one or more switches whose names start with
10025 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10026 once, no matter how many such switches appeared. However, if @code{%*}
10027 appears somewhere in @code{X}, then @code{X} will be substituted once
10028 for each matching switch, with the @code{%*} replaced by the part of
10029 that switch that matched the @code{*}.
10031 @item %@{.@code{S}:@code{X}@}
10032 Substitutes @code{X}, if processing a file with suffix @code{S}.
10034 @item %@{!.@code{S}:@code{X}@}
10035 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10037 @item %@{,@code{S}:@code{X}@}
10038 Substitutes @code{X}, if processing a file for language @code{S}.
10040 @item %@{!,@code{S}:@code{X}@}
10041 Substitutes @code{X}, if not processing a file for language @code{S}.
10043 @item %@{@code{S}|@code{P}:@code{X}@}
10044 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10045 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10046 @code{*} sequences as well, although they have a stronger binding than
10047 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10048 alternatives must be starred, and only the first matching alternative
10051 For example, a spec string like this:
10054 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10057 will output the following command-line options from the following input
10058 command-line options:
10063 -d fred.c -foo -baz -boggle
10064 -d jim.d -bar -baz -boggle
10067 @item %@{S:X; T:Y; :D@}
10069 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10070 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10071 be as many clauses as you need. This may be combined with @code{.},
10072 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10077 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10078 construct may contain other nested @samp{%} constructs or spaces, or
10079 even newlines. They are processed as usual, as described above.
10080 Trailing white space in @code{X} is ignored. White space may also
10081 appear anywhere on the left side of the colon in these constructs,
10082 except between @code{.} or @code{*} and the corresponding word.
10084 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10085 handled specifically in these constructs. If another value of
10086 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10087 @option{-W} switch is found later in the command line, the earlier
10088 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10089 just one letter, which passes all matching options.
10091 The character @samp{|} at the beginning of the predicate text is used to
10092 indicate that a command should be piped to the following command, but
10093 only if @option{-pipe} is specified.
10095 It is built into GCC which switches take arguments and which do not.
10096 (You might think it would be useful to generalize this to allow each
10097 compiler's spec to say which switches take arguments. But this cannot
10098 be done in a consistent fashion. GCC cannot even decide which input
10099 files have been specified without knowing which switches take arguments,
10100 and it must know which input files to compile in order to tell which
10103 GCC also knows implicitly that arguments starting in @option{-l} are to be
10104 treated as compiler output files, and passed to the linker in their
10105 proper position among the other output files.
10107 @c man begin OPTIONS
10109 @node Target Options
10110 @section Specifying Target Machine and Compiler Version
10111 @cindex target options
10112 @cindex cross compiling
10113 @cindex specifying machine version
10114 @cindex specifying compiler version and target machine
10115 @cindex compiler version, specifying
10116 @cindex target machine, specifying
10118 The usual way to run GCC is to run the executable called @command{gcc}, or
10119 @command{@var{machine}-gcc} when cross-compiling, or
10120 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10121 one that was installed last.
10123 @node Submodel Options
10124 @section Hardware Models and Configurations
10125 @cindex submodel options
10126 @cindex specifying hardware config
10127 @cindex hardware models and configurations, specifying
10128 @cindex machine dependent options
10130 Each target machine types can have its own
10131 special options, starting with @samp{-m}, to choose among various
10132 hardware models or configurations---for example, 68010 vs 68020,
10133 floating coprocessor or none. A single installed version of the
10134 compiler can compile for any model or configuration, according to the
10137 Some configurations of the compiler also support additional special
10138 options, usually for compatibility with other compilers on the same
10141 @c This list is ordered alphanumerically by subsection name.
10142 @c It should be the same order and spelling as these options are listed
10143 @c in Machine Dependent Options
10148 * Blackfin Options::
10152 * DEC Alpha Options::
10153 * DEC Alpha/VMS Options::
10156 * GNU/Linux Options::
10159 * i386 and x86-64 Options::
10160 * i386 and x86-64 Windows Options::
10162 * IA-64/VMS Options::
10169 * MicroBlaze Options::
10172 * MN10300 Options::
10174 * picoChip Options::
10175 * PowerPC Options::
10176 * RS/6000 and PowerPC Options::
10178 * S/390 and zSeries Options::
10181 * Solaris 2 Options::
10184 * System V Options::
10187 * VxWorks Options::
10189 * Xstormy16 Options::
10191 * zSeries Options::
10195 @subsection ARM Options
10196 @cindex ARM options
10198 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10202 @item -mabi=@var{name}
10204 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10205 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10208 @opindex mapcs-frame
10209 Generate a stack frame that is compliant with the ARM Procedure Call
10210 Standard for all functions, even if this is not strictly necessary for
10211 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10212 with this option will cause the stack frames not to be generated for
10213 leaf functions. The default is @option{-mno-apcs-frame}.
10217 This is a synonym for @option{-mapcs-frame}.
10220 @c not currently implemented
10221 @item -mapcs-stack-check
10222 @opindex mapcs-stack-check
10223 Generate code to check the amount of stack space available upon entry to
10224 every function (that actually uses some stack space). If there is
10225 insufficient space available then either the function
10226 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10227 called, depending upon the amount of stack space required. The run time
10228 system is required to provide these functions. The default is
10229 @option{-mno-apcs-stack-check}, since this produces smaller code.
10231 @c not currently implemented
10233 @opindex mapcs-float
10234 Pass floating point arguments using the float point registers. This is
10235 one of the variants of the APCS@. This option is recommended if the
10236 target hardware has a floating point unit or if a lot of floating point
10237 arithmetic is going to be performed by the code. The default is
10238 @option{-mno-apcs-float}, since integer only code is slightly increased in
10239 size if @option{-mapcs-float} is used.
10241 @c not currently implemented
10242 @item -mapcs-reentrant
10243 @opindex mapcs-reentrant
10244 Generate reentrant, position independent code. The default is
10245 @option{-mno-apcs-reentrant}.
10248 @item -mthumb-interwork
10249 @opindex mthumb-interwork
10250 Generate code which supports calling between the ARM and Thumb
10251 instruction sets. Without this option, on pre-v5 architectures, the
10252 two instruction sets cannot be reliably used inside one program. The
10253 default is @option{-mno-thumb-interwork}, since slightly larger code
10254 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10255 configurations this option is meaningless.
10257 @item -mno-sched-prolog
10258 @opindex mno-sched-prolog
10259 Prevent the reordering of instructions in the function prolog, or the
10260 merging of those instruction with the instructions in the function's
10261 body. This means that all functions will start with a recognizable set
10262 of instructions (or in fact one of a choice from a small set of
10263 different function prologues), and this information can be used to
10264 locate the start if functions inside an executable piece of code. The
10265 default is @option{-msched-prolog}.
10267 @item -mfloat-abi=@var{name}
10268 @opindex mfloat-abi
10269 Specifies which floating-point ABI to use. Permissible values
10270 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10272 Specifying @samp{soft} causes GCC to generate output containing
10273 library calls for floating-point operations.
10274 @samp{softfp} allows the generation of code using hardware floating-point
10275 instructions, but still uses the soft-float calling conventions.
10276 @samp{hard} allows generation of floating-point instructions
10277 and uses FPU-specific calling conventions.
10279 The default depends on the specific target configuration. Note that
10280 the hard-float and soft-float ABIs are not link-compatible; you must
10281 compile your entire program with the same ABI, and link with a
10282 compatible set of libraries.
10284 @item -mlittle-endian
10285 @opindex mlittle-endian
10286 Generate code for a processor running in little-endian mode. This is
10287 the default for all standard configurations.
10290 @opindex mbig-endian
10291 Generate code for a processor running in big-endian mode; the default is
10292 to compile code for a little-endian processor.
10294 @item -mwords-little-endian
10295 @opindex mwords-little-endian
10296 This option only applies when generating code for big-endian processors.
10297 Generate code for a little-endian word order but a big-endian byte
10298 order. That is, a byte order of the form @samp{32107654}. Note: this
10299 option should only be used if you require compatibility with code for
10300 big-endian ARM processors generated by versions of the compiler prior to
10301 2.8. This option is now deprecated.
10303 @item -mcpu=@var{name}
10305 This specifies the name of the target ARM processor. GCC uses this name
10306 to determine what kind of instructions it can emit when generating
10307 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10308 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10309 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10310 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10311 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10313 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10314 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10315 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10316 @samp{strongarm1110},
10317 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10318 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10319 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10320 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10321 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10322 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10323 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10324 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10325 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10326 @samp{cortex-m4}, @samp{cortex-m3},
10329 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10330 @samp{fa526}, @samp{fa626},
10331 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10333 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10334 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10335 See @option{-mtune} for more information.
10337 @item -mtune=@var{name}
10339 This option is very similar to the @option{-mcpu=} option, except that
10340 instead of specifying the actual target processor type, and hence
10341 restricting which instructions can be used, it specifies that GCC should
10342 tune the performance of the code as if the target were of the type
10343 specified in this option, but still choosing the instructions that it
10344 will generate based on the CPU specified by a @option{-mcpu=} option.
10345 For some ARM implementations better performance can be obtained by using
10348 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10349 performance for a blend of processors within architecture @var{arch}.
10350 The aim is to generate code that run well on the current most popular
10351 processors, balancing between optimizations that benefit some CPUs in the
10352 range, and avoiding performance pitfalls of other CPUs. The effects of
10353 this option may change in future GCC versions as CPU models come and go.
10355 @item -march=@var{name}
10357 This specifies the name of the target ARM architecture. GCC uses this
10358 name to determine what kind of instructions it can emit when generating
10359 assembly code. This option can be used in conjunction with or instead
10360 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10361 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10362 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10363 @samp{armv6}, @samp{armv6j},
10364 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10365 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10366 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10368 @item -mfpu=@var{name}
10369 @itemx -mfpe=@var{number}
10370 @itemx -mfp=@var{number}
10374 This specifies what floating point hardware (or hardware emulation) is
10375 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10376 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10377 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10378 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10379 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10380 @option{-mfp} and @option{-mfpe} are synonyms for
10381 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10384 If @option{-msoft-float} is specified this specifies the format of
10385 floating point values.
10387 If the selected floating-point hardware includes the NEON extension
10388 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10389 operations will not be used by GCC's auto-vectorization pass unless
10390 @option{-funsafe-math-optimizations} is also specified. This is
10391 because NEON hardware does not fully implement the IEEE 754 standard for
10392 floating-point arithmetic (in particular denormal values are treated as
10393 zero), so the use of NEON instructions may lead to a loss of precision.
10395 @item -mfp16-format=@var{name}
10396 @opindex mfp16-format
10397 Specify the format of the @code{__fp16} half-precision floating-point type.
10398 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10399 the default is @samp{none}, in which case the @code{__fp16} type is not
10400 defined. @xref{Half-Precision}, for more information.
10402 @item -mstructure-size-boundary=@var{n}
10403 @opindex mstructure-size-boundary
10404 The size of all structures and unions will be rounded up to a multiple
10405 of the number of bits set by this option. Permissible values are 8, 32
10406 and 64. The default value varies for different toolchains. For the COFF
10407 targeted toolchain the default value is 8. A value of 64 is only allowed
10408 if the underlying ABI supports it.
10410 Specifying the larger number can produce faster, more efficient code, but
10411 can also increase the size of the program. Different values are potentially
10412 incompatible. Code compiled with one value cannot necessarily expect to
10413 work with code or libraries compiled with another value, if they exchange
10414 information using structures or unions.
10416 @item -mabort-on-noreturn
10417 @opindex mabort-on-noreturn
10418 Generate a call to the function @code{abort} at the end of a
10419 @code{noreturn} function. It will be executed if the function tries to
10423 @itemx -mno-long-calls
10424 @opindex mlong-calls
10425 @opindex mno-long-calls
10426 Tells the compiler to perform function calls by first loading the
10427 address of the function into a register and then performing a subroutine
10428 call on this register. This switch is needed if the target function
10429 will lie outside of the 64 megabyte addressing range of the offset based
10430 version of subroutine call instruction.
10432 Even if this switch is enabled, not all function calls will be turned
10433 into long calls. The heuristic is that static functions, functions
10434 which have the @samp{short-call} attribute, functions that are inside
10435 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10436 definitions have already been compiled within the current compilation
10437 unit, will not be turned into long calls. The exception to this rule is
10438 that weak function definitions, functions with the @samp{long-call}
10439 attribute or the @samp{section} attribute, and functions that are within
10440 the scope of a @samp{#pragma long_calls} directive, will always be
10441 turned into long calls.
10443 This feature is not enabled by default. Specifying
10444 @option{-mno-long-calls} will restore the default behavior, as will
10445 placing the function calls within the scope of a @samp{#pragma
10446 long_calls_off} directive. Note these switches have no effect on how
10447 the compiler generates code to handle function calls via function
10450 @item -msingle-pic-base
10451 @opindex msingle-pic-base
10452 Treat the register used for PIC addressing as read-only, rather than
10453 loading it in the prologue for each function. The run-time system is
10454 responsible for initializing this register with an appropriate value
10455 before execution begins.
10457 @item -mpic-register=@var{reg}
10458 @opindex mpic-register
10459 Specify the register to be used for PIC addressing. The default is R10
10460 unless stack-checking is enabled, when R9 is used.
10462 @item -mcirrus-fix-invalid-insns
10463 @opindex mcirrus-fix-invalid-insns
10464 @opindex mno-cirrus-fix-invalid-insns
10465 Insert NOPs into the instruction stream to in order to work around
10466 problems with invalid Maverick instruction combinations. This option
10467 is only valid if the @option{-mcpu=ep9312} option has been used to
10468 enable generation of instructions for the Cirrus Maverick floating
10469 point co-processor. This option is not enabled by default, since the
10470 problem is only present in older Maverick implementations. The default
10471 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10474 @item -mpoke-function-name
10475 @opindex mpoke-function-name
10476 Write the name of each function into the text section, directly
10477 preceding the function prologue. The generated code is similar to this:
10481 .ascii "arm_poke_function_name", 0
10484 .word 0xff000000 + (t1 - t0)
10485 arm_poke_function_name
10487 stmfd sp!, @{fp, ip, lr, pc@}
10491 When performing a stack backtrace, code can inspect the value of
10492 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10493 location @code{pc - 12} and the top 8 bits are set, then we know that
10494 there is a function name embedded immediately preceding this location
10495 and has length @code{((pc[-3]) & 0xff000000)}.
10502 Select between generating code that executes in ARM and Thumb
10503 states. The default for most configurations is to generate code
10504 that executes in ARM state, but the default can be changed by
10505 configuring GCC with the @option{--with-mode=}@var{state}
10509 @opindex mtpcs-frame
10510 Generate a stack frame that is compliant with the Thumb Procedure Call
10511 Standard for all non-leaf functions. (A leaf function is one that does
10512 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10514 @item -mtpcs-leaf-frame
10515 @opindex mtpcs-leaf-frame
10516 Generate a stack frame that is compliant with the Thumb Procedure Call
10517 Standard for all leaf functions. (A leaf function is one that does
10518 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10520 @item -mcallee-super-interworking
10521 @opindex mcallee-super-interworking
10522 Gives all externally visible functions in the file being compiled an ARM
10523 instruction set header which switches to Thumb mode before executing the
10524 rest of the function. This allows these functions to be called from
10525 non-interworking code. This option is not valid in AAPCS configurations
10526 because interworking is enabled by default.
10528 @item -mcaller-super-interworking
10529 @opindex mcaller-super-interworking
10530 Allows calls via function pointers (including virtual functions) to
10531 execute correctly regardless of whether the target code has been
10532 compiled for interworking or not. There is a small overhead in the cost
10533 of executing a function pointer if this option is enabled. This option
10534 is not valid in AAPCS configurations because interworking is enabled
10537 @item -mtp=@var{name}
10539 Specify the access model for the thread local storage pointer. The valid
10540 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10541 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10542 (supported in the arm6k architecture), and @option{auto}, which uses the
10543 best available method for the selected processor. The default setting is
10546 @item -mtls-dialect=@var{dialect}
10547 @opindex mtls-dialect
10548 Specify the dialect to use for accessing thread local storage. Two
10549 dialects are supported --- @option{gnu} and @option{gnu2}. The
10550 @option{gnu} dialect selects the original GNU scheme for supporting
10551 local and global dynamic TLS models. The @option{gnu2} dialect
10552 selects the GNU descriptor scheme, which provides better performance
10553 for shared libraries. The GNU descriptor scheme is compatible with
10554 the original scheme, but does require new assembler, linker and
10555 library support. Initial and local exec TLS models are unaffected by
10556 this option and always use the original scheme.
10558 @item -mword-relocations
10559 @opindex mword-relocations
10560 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10561 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10562 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10565 @item -mfix-cortex-m3-ldrd
10566 @opindex mfix-cortex-m3-ldrd
10567 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10568 with overlapping destination and base registers are used. This option avoids
10569 generating these instructions. This option is enabled by default when
10570 @option{-mcpu=cortex-m3} is specified.
10575 @subsection AVR Options
10576 @cindex AVR Options
10578 These options are defined for AVR implementations:
10581 @item -mmcu=@var{mcu}
10583 Specify ATMEL AVR instruction set or MCU type.
10585 Instruction set avr1 is for the minimal AVR core, not supported by the C
10586 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10587 attiny11, attiny12, attiny15, attiny28).
10589 Instruction set avr2 (default) is for the classic AVR core with up to
10590 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10591 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10592 at90c8534, at90s8535).
10594 Instruction set avr3 is for the classic AVR core with up to 128K program
10595 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10597 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10598 memory space (MCU types: atmega8, atmega83, atmega85).
10600 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10601 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10602 atmega64, atmega128, at43usb355, at94k).
10604 @item -mno-interrupts
10605 @opindex mno-interrupts
10606 Generated code is not compatible with hardware interrupts.
10607 Code size will be smaller.
10609 @item -mcall-prologues
10610 @opindex mcall-prologues
10611 Functions prologues/epilogues expanded as call to appropriate
10612 subroutines. Code size will be smaller.
10615 @opindex mtiny-stack
10616 Change only the low 8 bits of the stack pointer.
10620 Assume int to be 8 bit integer. This affects the sizes of all types: A
10621 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10622 and long long will be 4 bytes. Please note that this option does not
10623 comply to the C standards, but it will provide you with smaller code
10627 @node Blackfin Options
10628 @subsection Blackfin Options
10629 @cindex Blackfin Options
10632 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10634 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10635 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10636 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10637 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10638 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10639 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10640 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10641 @samp{bf561}, @samp{bf592}.
10642 The optional @var{sirevision} specifies the silicon revision of the target
10643 Blackfin processor. Any workarounds available for the targeted silicon revision
10644 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10645 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10646 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10647 hexadecimal digits representing the major and minor numbers in the silicon
10648 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10649 is not defined. If @var{sirevision} is @samp{any}, the
10650 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10651 If this optional @var{sirevision} is not used, GCC assumes the latest known
10652 silicon revision of the targeted Blackfin processor.
10654 Support for @samp{bf561} is incomplete. For @samp{bf561},
10655 Only the processor macro is defined.
10656 Without this option, @samp{bf532} is used as the processor by default.
10657 The corresponding predefined processor macros for @var{cpu} is to
10658 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10659 provided by libgloss to be linked in if @option{-msim} is not given.
10663 Specifies that the program will be run on the simulator. This causes
10664 the simulator BSP provided by libgloss to be linked in. This option
10665 has effect only for @samp{bfin-elf} toolchain.
10666 Certain other options, such as @option{-mid-shared-library} and
10667 @option{-mfdpic}, imply @option{-msim}.
10669 @item -momit-leaf-frame-pointer
10670 @opindex momit-leaf-frame-pointer
10671 Don't keep the frame pointer in a register for leaf functions. This
10672 avoids the instructions to save, set up and restore frame pointers and
10673 makes an extra register available in leaf functions. The option
10674 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10675 which might make debugging harder.
10677 @item -mspecld-anomaly
10678 @opindex mspecld-anomaly
10679 When enabled, the compiler will ensure that the generated code does not
10680 contain speculative loads after jump instructions. If this option is used,
10681 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10683 @item -mno-specld-anomaly
10684 @opindex mno-specld-anomaly
10685 Don't generate extra code to prevent speculative loads from occurring.
10687 @item -mcsync-anomaly
10688 @opindex mcsync-anomaly
10689 When enabled, the compiler will ensure that the generated code does not
10690 contain CSYNC or SSYNC instructions too soon after conditional branches.
10691 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10693 @item -mno-csync-anomaly
10694 @opindex mno-csync-anomaly
10695 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10696 occurring too soon after a conditional branch.
10700 When enabled, the compiler is free to take advantage of the knowledge that
10701 the entire program fits into the low 64k of memory.
10704 @opindex mno-low-64k
10705 Assume that the program is arbitrarily large. This is the default.
10707 @item -mstack-check-l1
10708 @opindex mstack-check-l1
10709 Do stack checking using information placed into L1 scratchpad memory by the
10712 @item -mid-shared-library
10713 @opindex mid-shared-library
10714 Generate code that supports shared libraries via the library ID method.
10715 This allows for execute in place and shared libraries in an environment
10716 without virtual memory management. This option implies @option{-fPIC}.
10717 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10719 @item -mno-id-shared-library
10720 @opindex mno-id-shared-library
10721 Generate code that doesn't assume ID based shared libraries are being used.
10722 This is the default.
10724 @item -mleaf-id-shared-library
10725 @opindex mleaf-id-shared-library
10726 Generate code that supports shared libraries via the library ID method,
10727 but assumes that this library or executable won't link against any other
10728 ID shared libraries. That allows the compiler to use faster code for jumps
10731 @item -mno-leaf-id-shared-library
10732 @opindex mno-leaf-id-shared-library
10733 Do not assume that the code being compiled won't link against any ID shared
10734 libraries. Slower code will be generated for jump and call insns.
10736 @item -mshared-library-id=n
10737 @opindex mshared-library-id
10738 Specified the identification number of the ID based shared library being
10739 compiled. Specifying a value of 0 will generate more compact code, specifying
10740 other values will force the allocation of that number to the current
10741 library but is no more space or time efficient than omitting this option.
10745 Generate code that allows the data segment to be located in a different
10746 area of memory from the text segment. This allows for execute in place in
10747 an environment without virtual memory management by eliminating relocations
10748 against the text section.
10750 @item -mno-sep-data
10751 @opindex mno-sep-data
10752 Generate code that assumes that the data segment follows the text segment.
10753 This is the default.
10756 @itemx -mno-long-calls
10757 @opindex mlong-calls
10758 @opindex mno-long-calls
10759 Tells the compiler to perform function calls by first loading the
10760 address of the function into a register and then performing a subroutine
10761 call on this register. This switch is needed if the target function
10762 will lie outside of the 24 bit addressing range of the offset based
10763 version of subroutine call instruction.
10765 This feature is not enabled by default. Specifying
10766 @option{-mno-long-calls} will restore the default behavior. Note these
10767 switches have no effect on how the compiler generates code to handle
10768 function calls via function pointers.
10772 Link with the fast floating-point library. This library relaxes some of
10773 the IEEE floating-point standard's rules for checking inputs against
10774 Not-a-Number (NAN), in the interest of performance.
10777 @opindex minline-plt
10778 Enable inlining of PLT entries in function calls to functions that are
10779 not known to bind locally. It has no effect without @option{-mfdpic}.
10782 @opindex mmulticore
10783 Build standalone application for multicore Blackfin processor. Proper
10784 start files and link scripts will be used to support multicore.
10785 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10786 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10787 @option{-mcorea} or @option{-mcoreb}. If it's used without
10788 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10789 programming model is used. In this model, the main function of Core B
10790 should be named as coreb_main. If it's used with @option{-mcorea} or
10791 @option{-mcoreb}, one application per core programming model is used.
10792 If this option is not used, single core application programming
10797 Build standalone application for Core A of BF561 when using
10798 one application per core programming model. Proper start files
10799 and link scripts will be used to support Core A. This option
10800 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10804 Build standalone application for Core B of BF561 when using
10805 one application per core programming model. Proper start files
10806 and link scripts will be used to support Core B. This option
10807 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10808 should be used instead of main. It must be used with
10809 @option{-mmulticore}.
10813 Build standalone application for SDRAM. Proper start files and
10814 link scripts will be used to put the application into SDRAM.
10815 Loader should initialize SDRAM before loading the application
10816 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10820 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10821 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10822 are enabled; for standalone applications the default is off.
10826 @subsection C6X Options
10827 @cindex C6X Options
10830 @item -march=@var{name}
10832 This specifies the name of the target architecture. GCC uses this
10833 name to determine what kind of instructions it can emit when generating
10834 assembly code. Permissible names are: @samp{c62x},
10835 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
10838 @opindex mbig-endian
10839 Generate code for a big endian target.
10841 @item -mlittle-endian
10842 @opindex mlittle-endian
10843 Generate code for a little endian target. This is the default.
10847 Choose startup files and linker script suitable for the simulator.
10849 @item -msdata=default
10850 @opindex msdata=default
10851 Put small global and static data in the @samp{.neardata} section,
10852 which is pointed to by register @code{B14}. Put small uninitialized
10853 global and static data in the @samp{.bss} section, which is adjacent
10854 to the @samp{.neardata} section. Put small read-only data into the
10855 @samp{.rodata} section. The corresponding sections used for large
10856 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
10859 @opindex msdata=all
10860 Put all data, not just small objets, into the sections reserved for
10861 small data, and use addressing relative to the @code{B14} register to
10865 @opindex msdata=none
10866 Make no use of the sections reserved for small data, and use absolute
10867 addresses to access all data. Put all initialized global and static
10868 data in the @samp{.fardata} section, and all uninitialized data in the
10869 @samp{.far} section. Put all constant data into the @samp{.const}
10874 @subsection CRIS Options
10875 @cindex CRIS Options
10877 These options are defined specifically for the CRIS ports.
10880 @item -march=@var{architecture-type}
10881 @itemx -mcpu=@var{architecture-type}
10884 Generate code for the specified architecture. The choices for
10885 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10886 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10887 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10890 @item -mtune=@var{architecture-type}
10892 Tune to @var{architecture-type} everything applicable about the generated
10893 code, except for the ABI and the set of available instructions. The
10894 choices for @var{architecture-type} are the same as for
10895 @option{-march=@var{architecture-type}}.
10897 @item -mmax-stack-frame=@var{n}
10898 @opindex mmax-stack-frame
10899 Warn when the stack frame of a function exceeds @var{n} bytes.
10905 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10906 @option{-march=v3} and @option{-march=v8} respectively.
10908 @item -mmul-bug-workaround
10909 @itemx -mno-mul-bug-workaround
10910 @opindex mmul-bug-workaround
10911 @opindex mno-mul-bug-workaround
10912 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10913 models where it applies. This option is active by default.
10917 Enable CRIS-specific verbose debug-related information in the assembly
10918 code. This option also has the effect to turn off the @samp{#NO_APP}
10919 formatted-code indicator to the assembler at the beginning of the
10924 Do not use condition-code results from previous instruction; always emit
10925 compare and test instructions before use of condition codes.
10927 @item -mno-side-effects
10928 @opindex mno-side-effects
10929 Do not emit instructions with side-effects in addressing modes other than
10932 @item -mstack-align
10933 @itemx -mno-stack-align
10934 @itemx -mdata-align
10935 @itemx -mno-data-align
10936 @itemx -mconst-align
10937 @itemx -mno-const-align
10938 @opindex mstack-align
10939 @opindex mno-stack-align
10940 @opindex mdata-align
10941 @opindex mno-data-align
10942 @opindex mconst-align
10943 @opindex mno-const-align
10944 These options (no-options) arranges (eliminate arrangements) for the
10945 stack-frame, individual data and constants to be aligned for the maximum
10946 single data access size for the chosen CPU model. The default is to
10947 arrange for 32-bit alignment. ABI details such as structure layout are
10948 not affected by these options.
10956 Similar to the stack- data- and const-align options above, these options
10957 arrange for stack-frame, writable data and constants to all be 32-bit,
10958 16-bit or 8-bit aligned. The default is 32-bit alignment.
10960 @item -mno-prologue-epilogue
10961 @itemx -mprologue-epilogue
10962 @opindex mno-prologue-epilogue
10963 @opindex mprologue-epilogue
10964 With @option{-mno-prologue-epilogue}, the normal function prologue and
10965 epilogue that sets up the stack-frame are omitted and no return
10966 instructions or return sequences are generated in the code. Use this
10967 option only together with visual inspection of the compiled code: no
10968 warnings or errors are generated when call-saved registers must be saved,
10969 or storage for local variable needs to be allocated.
10973 @opindex mno-gotplt
10975 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10976 instruction sequences that load addresses for functions from the PLT part
10977 of the GOT rather than (traditional on other architectures) calls to the
10978 PLT@. The default is @option{-mgotplt}.
10982 Legacy no-op option only recognized with the cris-axis-elf and
10983 cris-axis-linux-gnu targets.
10987 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10991 This option, recognized for the cris-axis-elf arranges
10992 to link with input-output functions from a simulator library. Code,
10993 initialized data and zero-initialized data are allocated consecutively.
10997 Like @option{-sim}, but pass linker options to locate initialized data at
10998 0x40000000 and zero-initialized data at 0x80000000.
11001 @node Darwin Options
11002 @subsection Darwin Options
11003 @cindex Darwin options
11005 These options are defined for all architectures running the Darwin operating
11008 FSF GCC on Darwin does not create ``fat'' object files; it will create
11009 an object file for the single architecture that it was built to
11010 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11011 @option{-arch} options are used; it does so by running the compiler or
11012 linker multiple times and joining the results together with
11015 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11016 @samp{i686}) is determined by the flags that specify the ISA
11017 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11018 @option{-force_cpusubtype_ALL} option can be used to override this.
11020 The Darwin tools vary in their behavior when presented with an ISA
11021 mismatch. The assembler, @file{as}, will only permit instructions to
11022 be used that are valid for the subtype of the file it is generating,
11023 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11024 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11025 and print an error if asked to create a shared library with a less
11026 restrictive subtype than its input files (for instance, trying to put
11027 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11028 for executables, @file{ld}, will quietly give the executable the most
11029 restrictive subtype of any of its input files.
11034 Add the framework directory @var{dir} to the head of the list of
11035 directories to be searched for header files. These directories are
11036 interleaved with those specified by @option{-I} options and are
11037 scanned in a left-to-right order.
11039 A framework directory is a directory with frameworks in it. A
11040 framework is a directory with a @samp{"Headers"} and/or
11041 @samp{"PrivateHeaders"} directory contained directly in it that ends
11042 in @samp{".framework"}. The name of a framework is the name of this
11043 directory excluding the @samp{".framework"}. Headers associated with
11044 the framework are found in one of those two directories, with
11045 @samp{"Headers"} being searched first. A subframework is a framework
11046 directory that is in a framework's @samp{"Frameworks"} directory.
11047 Includes of subframework headers can only appear in a header of a
11048 framework that contains the subframework, or in a sibling subframework
11049 header. Two subframeworks are siblings if they occur in the same
11050 framework. A subframework should not have the same name as a
11051 framework, a warning will be issued if this is violated. Currently a
11052 subframework cannot have subframeworks, in the future, the mechanism
11053 may be extended to support this. The standard frameworks can be found
11054 in @samp{"/System/Library/Frameworks"} and
11055 @samp{"/Library/Frameworks"}. An example include looks like
11056 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11057 the name of the framework and header.h is found in the
11058 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11060 @item -iframework@var{dir}
11061 @opindex iframework
11062 Like @option{-F} except the directory is a treated as a system
11063 directory. The main difference between this @option{-iframework} and
11064 @option{-F} is that with @option{-iframework} the compiler does not
11065 warn about constructs contained within header files found via
11066 @var{dir}. This option is valid only for the C family of languages.
11070 Emit debugging information for symbols that are used. For STABS
11071 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11072 This is by default ON@.
11076 Emit debugging information for all symbols and types.
11078 @item -mmacosx-version-min=@var{version}
11079 The earliest version of MacOS X that this executable will run on
11080 is @var{version}. Typical values of @var{version} include @code{10.1},
11081 @code{10.2}, and @code{10.3.9}.
11083 If the compiler was built to use the system's headers by default,
11084 then the default for this option is the system version on which the
11085 compiler is running, otherwise the default is to make choices which
11086 are compatible with as many systems and code bases as possible.
11090 Enable kernel development mode. The @option{-mkernel} option sets
11091 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11092 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11093 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11094 applicable. This mode also sets @option{-mno-altivec},
11095 @option{-msoft-float}, @option{-fno-builtin} and
11096 @option{-mlong-branch} for PowerPC targets.
11098 @item -mone-byte-bool
11099 @opindex mone-byte-bool
11100 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11101 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11102 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11103 option has no effect on x86.
11105 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11106 to generate code that is not binary compatible with code generated
11107 without that switch. Using this switch may require recompiling all
11108 other modules in a program, including system libraries. Use this
11109 switch to conform to a non-default data model.
11111 @item -mfix-and-continue
11112 @itemx -ffix-and-continue
11113 @itemx -findirect-data
11114 @opindex mfix-and-continue
11115 @opindex ffix-and-continue
11116 @opindex findirect-data
11117 Generate code suitable for fast turn around development. Needed to
11118 enable gdb to dynamically load @code{.o} files into already running
11119 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11120 are provided for backwards compatibility.
11124 Loads all members of static archive libraries.
11125 See man ld(1) for more information.
11127 @item -arch_errors_fatal
11128 @opindex arch_errors_fatal
11129 Cause the errors having to do with files that have the wrong architecture
11132 @item -bind_at_load
11133 @opindex bind_at_load
11134 Causes the output file to be marked such that the dynamic linker will
11135 bind all undefined references when the file is loaded or launched.
11139 Produce a Mach-o bundle format file.
11140 See man ld(1) for more information.
11142 @item -bundle_loader @var{executable}
11143 @opindex bundle_loader
11144 This option specifies the @var{executable} that will be loading the build
11145 output file being linked. See man ld(1) for more information.
11148 @opindex dynamiclib
11149 When passed this option, GCC will produce a dynamic library instead of
11150 an executable when linking, using the Darwin @file{libtool} command.
11152 @item -force_cpusubtype_ALL
11153 @opindex force_cpusubtype_ALL
11154 This causes GCC's output file to have the @var{ALL} subtype, instead of
11155 one controlled by the @option{-mcpu} or @option{-march} option.
11157 @item -allowable_client @var{client_name}
11158 @itemx -client_name
11159 @itemx -compatibility_version
11160 @itemx -current_version
11162 @itemx -dependency-file
11164 @itemx -dylinker_install_name
11166 @itemx -exported_symbols_list
11169 @itemx -flat_namespace
11170 @itemx -force_flat_namespace
11171 @itemx -headerpad_max_install_names
11174 @itemx -install_name
11175 @itemx -keep_private_externs
11176 @itemx -multi_module
11177 @itemx -multiply_defined
11178 @itemx -multiply_defined_unused
11181 @itemx -no_dead_strip_inits_and_terms
11182 @itemx -nofixprebinding
11183 @itemx -nomultidefs
11185 @itemx -noseglinkedit
11186 @itemx -pagezero_size
11188 @itemx -prebind_all_twolevel_modules
11189 @itemx -private_bundle
11191 @itemx -read_only_relocs
11193 @itemx -sectobjectsymbols
11197 @itemx -sectobjectsymbols
11200 @itemx -segs_read_only_addr
11202 @itemx -segs_read_write_addr
11203 @itemx -seg_addr_table
11204 @itemx -seg_addr_table_filename
11205 @itemx -seglinkedit
11207 @itemx -segs_read_only_addr
11208 @itemx -segs_read_write_addr
11209 @itemx -single_module
11211 @itemx -sub_library
11213 @itemx -sub_umbrella
11214 @itemx -twolevel_namespace
11217 @itemx -unexported_symbols_list
11218 @itemx -weak_reference_mismatches
11219 @itemx -whatsloaded
11220 @opindex allowable_client
11221 @opindex client_name
11222 @opindex compatibility_version
11223 @opindex current_version
11224 @opindex dead_strip
11225 @opindex dependency-file
11226 @opindex dylib_file
11227 @opindex dylinker_install_name
11229 @opindex exported_symbols_list
11231 @opindex flat_namespace
11232 @opindex force_flat_namespace
11233 @opindex headerpad_max_install_names
11234 @opindex image_base
11236 @opindex install_name
11237 @opindex keep_private_externs
11238 @opindex multi_module
11239 @opindex multiply_defined
11240 @opindex multiply_defined_unused
11241 @opindex noall_load
11242 @opindex no_dead_strip_inits_and_terms
11243 @opindex nofixprebinding
11244 @opindex nomultidefs
11246 @opindex noseglinkedit
11247 @opindex pagezero_size
11249 @opindex prebind_all_twolevel_modules
11250 @opindex private_bundle
11251 @opindex read_only_relocs
11253 @opindex sectobjectsymbols
11256 @opindex sectcreate
11257 @opindex sectobjectsymbols
11260 @opindex segs_read_only_addr
11261 @opindex segs_read_write_addr
11262 @opindex seg_addr_table
11263 @opindex seg_addr_table_filename
11264 @opindex seglinkedit
11266 @opindex segs_read_only_addr
11267 @opindex segs_read_write_addr
11268 @opindex single_module
11270 @opindex sub_library
11271 @opindex sub_umbrella
11272 @opindex twolevel_namespace
11275 @opindex unexported_symbols_list
11276 @opindex weak_reference_mismatches
11277 @opindex whatsloaded
11278 These options are passed to the Darwin linker. The Darwin linker man page
11279 describes them in detail.
11282 @node DEC Alpha Options
11283 @subsection DEC Alpha Options
11285 These @samp{-m} options are defined for the DEC Alpha implementations:
11288 @item -mno-soft-float
11289 @itemx -msoft-float
11290 @opindex mno-soft-float
11291 @opindex msoft-float
11292 Use (do not use) the hardware floating-point instructions for
11293 floating-point operations. When @option{-msoft-float} is specified,
11294 functions in @file{libgcc.a} will be used to perform floating-point
11295 operations. Unless they are replaced by routines that emulate the
11296 floating-point operations, or compiled in such a way as to call such
11297 emulations routines, these routines will issue floating-point
11298 operations. If you are compiling for an Alpha without floating-point
11299 operations, you must ensure that the library is built so as not to call
11302 Note that Alpha implementations without floating-point operations are
11303 required to have floating-point registers.
11306 @itemx -mno-fp-regs
11308 @opindex mno-fp-regs
11309 Generate code that uses (does not use) the floating-point register set.
11310 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11311 register set is not used, floating point operands are passed in integer
11312 registers as if they were integers and floating-point results are passed
11313 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11314 so any function with a floating-point argument or return value called by code
11315 compiled with @option{-mno-fp-regs} must also be compiled with that
11318 A typical use of this option is building a kernel that does not use,
11319 and hence need not save and restore, any floating-point registers.
11323 The Alpha architecture implements floating-point hardware optimized for
11324 maximum performance. It is mostly compliant with the IEEE floating
11325 point standard. However, for full compliance, software assistance is
11326 required. This option generates code fully IEEE compliant code
11327 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11328 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11329 defined during compilation. The resulting code is less efficient but is
11330 able to correctly support denormalized numbers and exceptional IEEE
11331 values such as not-a-number and plus/minus infinity. Other Alpha
11332 compilers call this option @option{-ieee_with_no_inexact}.
11334 @item -mieee-with-inexact
11335 @opindex mieee-with-inexact
11336 This is like @option{-mieee} except the generated code also maintains
11337 the IEEE @var{inexact-flag}. Turning on this option causes the
11338 generated code to implement fully-compliant IEEE math. In addition to
11339 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11340 macro. On some Alpha implementations the resulting code may execute
11341 significantly slower than the code generated by default. Since there is
11342 very little code that depends on the @var{inexact-flag}, you should
11343 normally not specify this option. Other Alpha compilers call this
11344 option @option{-ieee_with_inexact}.
11346 @item -mfp-trap-mode=@var{trap-mode}
11347 @opindex mfp-trap-mode
11348 This option controls what floating-point related traps are enabled.
11349 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11350 The trap mode can be set to one of four values:
11354 This is the default (normal) setting. The only traps that are enabled
11355 are the ones that cannot be disabled in software (e.g., division by zero
11359 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11363 Like @samp{u}, but the instructions are marked to be safe for software
11364 completion (see Alpha architecture manual for details).
11367 Like @samp{su}, but inexact traps are enabled as well.
11370 @item -mfp-rounding-mode=@var{rounding-mode}
11371 @opindex mfp-rounding-mode
11372 Selects the IEEE rounding mode. Other Alpha compilers call this option
11373 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11378 Normal IEEE rounding mode. Floating point numbers are rounded towards
11379 the nearest machine number or towards the even machine number in case
11383 Round towards minus infinity.
11386 Chopped rounding mode. Floating point numbers are rounded towards zero.
11389 Dynamic rounding mode. A field in the floating point control register
11390 (@var{fpcr}, see Alpha architecture reference manual) controls the
11391 rounding mode in effect. The C library initializes this register for
11392 rounding towards plus infinity. Thus, unless your program modifies the
11393 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11396 @item -mtrap-precision=@var{trap-precision}
11397 @opindex mtrap-precision
11398 In the Alpha architecture, floating point traps are imprecise. This
11399 means without software assistance it is impossible to recover from a
11400 floating trap and program execution normally needs to be terminated.
11401 GCC can generate code that can assist operating system trap handlers
11402 in determining the exact location that caused a floating point trap.
11403 Depending on the requirements of an application, different levels of
11404 precisions can be selected:
11408 Program precision. This option is the default and means a trap handler
11409 can only identify which program caused a floating point exception.
11412 Function precision. The trap handler can determine the function that
11413 caused a floating point exception.
11416 Instruction precision. The trap handler can determine the exact
11417 instruction that caused a floating point exception.
11420 Other Alpha compilers provide the equivalent options called
11421 @option{-scope_safe} and @option{-resumption_safe}.
11423 @item -mieee-conformant
11424 @opindex mieee-conformant
11425 This option marks the generated code as IEEE conformant. You must not
11426 use this option unless you also specify @option{-mtrap-precision=i} and either
11427 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11428 is to emit the line @samp{.eflag 48} in the function prologue of the
11429 generated assembly file. Under DEC Unix, this has the effect that
11430 IEEE-conformant math library routines will be linked in.
11432 @item -mbuild-constants
11433 @opindex mbuild-constants
11434 Normally GCC examines a 32- or 64-bit integer constant to
11435 see if it can construct it from smaller constants in two or three
11436 instructions. If it cannot, it will output the constant as a literal and
11437 generate code to load it from the data segment at runtime.
11439 Use this option to require GCC to construct @emph{all} integer constants
11440 using code, even if it takes more instructions (the maximum is six).
11442 You would typically use this option to build a shared library dynamic
11443 loader. Itself a shared library, it must relocate itself in memory
11444 before it can find the variables and constants in its own data segment.
11450 Select whether to generate code to be assembled by the vendor-supplied
11451 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11469 Indicate whether GCC should generate code to use the optional BWX,
11470 CIX, FIX and MAX instruction sets. The default is to use the instruction
11471 sets supported by the CPU type specified via @option{-mcpu=} option or that
11472 of the CPU on which GCC was built if none was specified.
11475 @itemx -mfloat-ieee
11476 @opindex mfloat-vax
11477 @opindex mfloat-ieee
11478 Generate code that uses (does not use) VAX F and G floating point
11479 arithmetic instead of IEEE single and double precision.
11481 @item -mexplicit-relocs
11482 @itemx -mno-explicit-relocs
11483 @opindex mexplicit-relocs
11484 @opindex mno-explicit-relocs
11485 Older Alpha assemblers provided no way to generate symbol relocations
11486 except via assembler macros. Use of these macros does not allow
11487 optimal instruction scheduling. GNU binutils as of version 2.12
11488 supports a new syntax that allows the compiler to explicitly mark
11489 which relocations should apply to which instructions. This option
11490 is mostly useful for debugging, as GCC detects the capabilities of
11491 the assembler when it is built and sets the default accordingly.
11494 @itemx -mlarge-data
11495 @opindex msmall-data
11496 @opindex mlarge-data
11497 When @option{-mexplicit-relocs} is in effect, static data is
11498 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11499 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11500 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11501 16-bit relocations off of the @code{$gp} register. This limits the
11502 size of the small data area to 64KB, but allows the variables to be
11503 directly accessed via a single instruction.
11505 The default is @option{-mlarge-data}. With this option the data area
11506 is limited to just below 2GB@. Programs that require more than 2GB of
11507 data must use @code{malloc} or @code{mmap} to allocate the data in the
11508 heap instead of in the program's data segment.
11510 When generating code for shared libraries, @option{-fpic} implies
11511 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11514 @itemx -mlarge-text
11515 @opindex msmall-text
11516 @opindex mlarge-text
11517 When @option{-msmall-text} is used, the compiler assumes that the
11518 code of the entire program (or shared library) fits in 4MB, and is
11519 thus reachable with a branch instruction. When @option{-msmall-data}
11520 is used, the compiler can assume that all local symbols share the
11521 same @code{$gp} value, and thus reduce the number of instructions
11522 required for a function call from 4 to 1.
11524 The default is @option{-mlarge-text}.
11526 @item -mcpu=@var{cpu_type}
11528 Set the instruction set and instruction scheduling parameters for
11529 machine type @var{cpu_type}. You can specify either the @samp{EV}
11530 style name or the corresponding chip number. GCC supports scheduling
11531 parameters for the EV4, EV5 and EV6 family of processors and will
11532 choose the default values for the instruction set from the processor
11533 you specify. If you do not specify a processor type, GCC will default
11534 to the processor on which the compiler was built.
11536 Supported values for @var{cpu_type} are
11542 Schedules as an EV4 and has no instruction set extensions.
11546 Schedules as an EV5 and has no instruction set extensions.
11550 Schedules as an EV5 and supports the BWX extension.
11555 Schedules as an EV5 and supports the BWX and MAX extensions.
11559 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11563 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11566 Native toolchains also support the value @samp{native},
11567 which selects the best architecture option for the host processor.
11568 @option{-mcpu=native} has no effect if GCC does not recognize
11571 @item -mtune=@var{cpu_type}
11573 Set only the instruction scheduling parameters for machine type
11574 @var{cpu_type}. The instruction set is not changed.
11576 Native toolchains also support the value @samp{native},
11577 which selects the best architecture option for the host processor.
11578 @option{-mtune=native} has no effect if GCC does not recognize
11581 @item -mmemory-latency=@var{time}
11582 @opindex mmemory-latency
11583 Sets the latency the scheduler should assume for typical memory
11584 references as seen by the application. This number is highly
11585 dependent on the memory access patterns used by the application
11586 and the size of the external cache on the machine.
11588 Valid options for @var{time} are
11592 A decimal number representing clock cycles.
11598 The compiler contains estimates of the number of clock cycles for
11599 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11600 (also called Dcache, Scache, and Bcache), as well as to main memory.
11601 Note that L3 is only valid for EV5.
11606 @node DEC Alpha/VMS Options
11607 @subsection DEC Alpha/VMS Options
11609 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11612 @item -mvms-return-codes
11613 @opindex mvms-return-codes
11614 Return VMS condition codes from main. The default is to return POSIX
11615 style condition (e.g.@: error) codes.
11617 @item -mdebug-main=@var{prefix}
11618 @opindex mdebug-main=@var{prefix}
11619 Flag the first routine whose name starts with @var{prefix} as the main
11620 routine for the debugger.
11624 Default to 64bit memory allocation routines.
11628 @subsection FR30 Options
11629 @cindex FR30 Options
11631 These options are defined specifically for the FR30 port.
11635 @item -msmall-model
11636 @opindex msmall-model
11637 Use the small address space model. This can produce smaller code, but
11638 it does assume that all symbolic values and addresses will fit into a
11643 Assume that run-time support has been provided and so there is no need
11644 to include the simulator library (@file{libsim.a}) on the linker
11650 @subsection FRV Options
11651 @cindex FRV Options
11657 Only use the first 32 general purpose registers.
11662 Use all 64 general purpose registers.
11667 Use only the first 32 floating point registers.
11672 Use all 64 floating point registers
11675 @opindex mhard-float
11677 Use hardware instructions for floating point operations.
11680 @opindex msoft-float
11682 Use library routines for floating point operations.
11687 Dynamically allocate condition code registers.
11692 Do not try to dynamically allocate condition code registers, only
11693 use @code{icc0} and @code{fcc0}.
11698 Change ABI to use double word insns.
11703 Do not use double word instructions.
11708 Use floating point double instructions.
11711 @opindex mno-double
11713 Do not use floating point double instructions.
11718 Use media instructions.
11723 Do not use media instructions.
11728 Use multiply and add/subtract instructions.
11731 @opindex mno-muladd
11733 Do not use multiply and add/subtract instructions.
11738 Select the FDPIC ABI, that uses function descriptors to represent
11739 pointers to functions. Without any PIC/PIE-related options, it
11740 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11741 assumes GOT entries and small data are within a 12-bit range from the
11742 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11743 are computed with 32 bits.
11744 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11747 @opindex minline-plt
11749 Enable inlining of PLT entries in function calls to functions that are
11750 not known to bind locally. It has no effect without @option{-mfdpic}.
11751 It's enabled by default if optimizing for speed and compiling for
11752 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11753 optimization option such as @option{-O3} or above is present in the
11759 Assume a large TLS segment when generating thread-local code.
11764 Do not assume a large TLS segment when generating thread-local code.
11769 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11770 that is known to be in read-only sections. It's enabled by default,
11771 except for @option{-fpic} or @option{-fpie}: even though it may help
11772 make the global offset table smaller, it trades 1 instruction for 4.
11773 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11774 one of which may be shared by multiple symbols, and it avoids the need
11775 for a GOT entry for the referenced symbol, so it's more likely to be a
11776 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11778 @item -multilib-library-pic
11779 @opindex multilib-library-pic
11781 Link with the (library, not FD) pic libraries. It's implied by
11782 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11783 @option{-fpic} without @option{-mfdpic}. You should never have to use
11787 @opindex mlinked-fp
11789 Follow the EABI requirement of always creating a frame pointer whenever
11790 a stack frame is allocated. This option is enabled by default and can
11791 be disabled with @option{-mno-linked-fp}.
11794 @opindex mlong-calls
11796 Use indirect addressing to call functions outside the current
11797 compilation unit. This allows the functions to be placed anywhere
11798 within the 32-bit address space.
11800 @item -malign-labels
11801 @opindex malign-labels
11803 Try to align labels to an 8-byte boundary by inserting nops into the
11804 previous packet. This option only has an effect when VLIW packing
11805 is enabled. It doesn't create new packets; it merely adds nops to
11808 @item -mlibrary-pic
11809 @opindex mlibrary-pic
11811 Generate position-independent EABI code.
11816 Use only the first four media accumulator registers.
11821 Use all eight media accumulator registers.
11826 Pack VLIW instructions.
11831 Do not pack VLIW instructions.
11834 @opindex mno-eflags
11836 Do not mark ABI switches in e_flags.
11839 @opindex mcond-move
11841 Enable the use of conditional-move instructions (default).
11843 This switch is mainly for debugging the compiler and will likely be removed
11844 in a future version.
11846 @item -mno-cond-move
11847 @opindex mno-cond-move
11849 Disable the use of conditional-move instructions.
11851 This switch is mainly for debugging the compiler and will likely be removed
11852 in a future version.
11857 Enable the use of conditional set instructions (default).
11859 This switch is mainly for debugging the compiler and will likely be removed
11860 in a future version.
11865 Disable the use of conditional set instructions.
11867 This switch is mainly for debugging the compiler and will likely be removed
11868 in a future version.
11871 @opindex mcond-exec
11873 Enable the use of conditional execution (default).
11875 This switch is mainly for debugging the compiler and will likely be removed
11876 in a future version.
11878 @item -mno-cond-exec
11879 @opindex mno-cond-exec
11881 Disable the use of conditional execution.
11883 This switch is mainly for debugging the compiler and will likely be removed
11884 in a future version.
11886 @item -mvliw-branch
11887 @opindex mvliw-branch
11889 Run a pass to pack branches into VLIW instructions (default).
11891 This switch is mainly for debugging the compiler and will likely be removed
11892 in a future version.
11894 @item -mno-vliw-branch
11895 @opindex mno-vliw-branch
11897 Do not run a pass to pack branches into VLIW instructions.
11899 This switch is mainly for debugging the compiler and will likely be removed
11900 in a future version.
11902 @item -mmulti-cond-exec
11903 @opindex mmulti-cond-exec
11905 Enable optimization of @code{&&} and @code{||} in conditional execution
11908 This switch is mainly for debugging the compiler and will likely be removed
11909 in a future version.
11911 @item -mno-multi-cond-exec
11912 @opindex mno-multi-cond-exec
11914 Disable optimization of @code{&&} and @code{||} in conditional execution.
11916 This switch is mainly for debugging the compiler and will likely be removed
11917 in a future version.
11919 @item -mnested-cond-exec
11920 @opindex mnested-cond-exec
11922 Enable nested conditional execution optimizations (default).
11924 This switch is mainly for debugging the compiler and will likely be removed
11925 in a future version.
11927 @item -mno-nested-cond-exec
11928 @opindex mno-nested-cond-exec
11930 Disable nested conditional execution optimizations.
11932 This switch is mainly for debugging the compiler and will likely be removed
11933 in a future version.
11935 @item -moptimize-membar
11936 @opindex moptimize-membar
11938 This switch removes redundant @code{membar} instructions from the
11939 compiler generated code. It is enabled by default.
11941 @item -mno-optimize-membar
11942 @opindex mno-optimize-membar
11944 This switch disables the automatic removal of redundant @code{membar}
11945 instructions from the generated code.
11947 @item -mtomcat-stats
11948 @opindex mtomcat-stats
11950 Cause gas to print out tomcat statistics.
11952 @item -mcpu=@var{cpu}
11955 Select the processor type for which to generate code. Possible values are
11956 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11957 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11961 @node GNU/Linux Options
11962 @subsection GNU/Linux Options
11964 These @samp{-m} options are defined for GNU/Linux targets:
11969 Use the GNU C library. This is the default except
11970 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11974 Use uClibc C library. This is the default on
11975 @samp{*-*-linux-*uclibc*} targets.
11979 Use Bionic C library. This is the default on
11980 @samp{*-*-linux-*android*} targets.
11984 Compile code compatible with Android platform. This is the default on
11985 @samp{*-*-linux-*android*} targets.
11987 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11988 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11989 this option makes the GCC driver pass Android-specific options to the linker.
11990 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11993 @item -tno-android-cc
11994 @opindex tno-android-cc
11995 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11996 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11997 @option{-fno-rtti} by default.
11999 @item -tno-android-ld
12000 @opindex tno-android-ld
12001 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12002 linking options to the linker.
12006 @node H8/300 Options
12007 @subsection H8/300 Options
12009 These @samp{-m} options are defined for the H8/300 implementations:
12014 Shorten some address references at link time, when possible; uses the
12015 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12016 ld, Using ld}, for a fuller description.
12020 Generate code for the H8/300H@.
12024 Generate code for the H8S@.
12028 Generate code for the H8S and H8/300H in the normal mode. This switch
12029 must be used either with @option{-mh} or @option{-ms}.
12033 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12037 Make @code{int} data 32 bits by default.
12040 @opindex malign-300
12041 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12042 The default for the H8/300H and H8S is to align longs and floats on 4
12044 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12045 This option has no effect on the H8/300.
12049 @subsection HPPA Options
12050 @cindex HPPA Options
12052 These @samp{-m} options are defined for the HPPA family of computers:
12055 @item -march=@var{architecture-type}
12057 Generate code for the specified architecture. The choices for
12058 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12059 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12060 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12061 architecture option for your machine. Code compiled for lower numbered
12062 architectures will run on higher numbered architectures, but not the
12065 @item -mpa-risc-1-0
12066 @itemx -mpa-risc-1-1
12067 @itemx -mpa-risc-2-0
12068 @opindex mpa-risc-1-0
12069 @opindex mpa-risc-1-1
12070 @opindex mpa-risc-2-0
12071 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12074 @opindex mbig-switch
12075 Generate code suitable for big switch tables. Use this option only if
12076 the assembler/linker complain about out of range branches within a switch
12079 @item -mjump-in-delay
12080 @opindex mjump-in-delay
12081 Fill delay slots of function calls with unconditional jump instructions
12082 by modifying the return pointer for the function call to be the target
12083 of the conditional jump.
12085 @item -mdisable-fpregs
12086 @opindex mdisable-fpregs
12087 Prevent floating point registers from being used in any manner. This is
12088 necessary for compiling kernels which perform lazy context switching of
12089 floating point registers. If you use this option and attempt to perform
12090 floating point operations, the compiler will abort.
12092 @item -mdisable-indexing
12093 @opindex mdisable-indexing
12094 Prevent the compiler from using indexing address modes. This avoids some
12095 rather obscure problems when compiling MIG generated code under MACH@.
12097 @item -mno-space-regs
12098 @opindex mno-space-regs
12099 Generate code that assumes the target has no space registers. This allows
12100 GCC to generate faster indirect calls and use unscaled index address modes.
12102 Such code is suitable for level 0 PA systems and kernels.
12104 @item -mfast-indirect-calls
12105 @opindex mfast-indirect-calls
12106 Generate code that assumes calls never cross space boundaries. This
12107 allows GCC to emit code which performs faster indirect calls.
12109 This option will not work in the presence of shared libraries or nested
12112 @item -mfixed-range=@var{register-range}
12113 @opindex mfixed-range
12114 Generate code treating the given register range as fixed registers.
12115 A fixed register is one that the register allocator can not use. This is
12116 useful when compiling kernel code. A register range is specified as
12117 two registers separated by a dash. Multiple register ranges can be
12118 specified separated by a comma.
12120 @item -mlong-load-store
12121 @opindex mlong-load-store
12122 Generate 3-instruction load and store sequences as sometimes required by
12123 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12126 @item -mportable-runtime
12127 @opindex mportable-runtime
12128 Use the portable calling conventions proposed by HP for ELF systems.
12132 Enable the use of assembler directives only GAS understands.
12134 @item -mschedule=@var{cpu-type}
12136 Schedule code according to the constraints for the machine type
12137 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12138 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12139 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12140 proper scheduling option for your machine. The default scheduling is
12144 @opindex mlinker-opt
12145 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12146 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12147 linkers in which they give bogus error messages when linking some programs.
12150 @opindex msoft-float
12151 Generate output containing library calls for floating point.
12152 @strong{Warning:} the requisite libraries are not available for all HPPA
12153 targets. Normally the facilities of the machine's usual C compiler are
12154 used, but this cannot be done directly in cross-compilation. You must make
12155 your own arrangements to provide suitable library functions for
12158 @option{-msoft-float} changes the calling convention in the output file;
12159 therefore, it is only useful if you compile @emph{all} of a program with
12160 this option. In particular, you need to compile @file{libgcc.a}, the
12161 library that comes with GCC, with @option{-msoft-float} in order for
12166 Generate the predefine, @code{_SIO}, for server IO@. The default is
12167 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12168 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12169 options are available under HP-UX and HI-UX@.
12173 Use GNU ld specific options. This passes @option{-shared} to ld when
12174 building a shared library. It is the default when GCC is configured,
12175 explicitly or implicitly, with the GNU linker. This option does not
12176 have any affect on which ld is called, it only changes what parameters
12177 are passed to that ld. The ld that is called is determined by the
12178 @option{--with-ld} configure option, GCC's program search path, and
12179 finally by the user's @env{PATH}. The linker used by GCC can be printed
12180 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12181 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12185 Use HP ld specific options. This passes @option{-b} to ld when building
12186 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12187 links. It is the default when GCC is configured, explicitly or
12188 implicitly, with the HP linker. This option does not have any affect on
12189 which ld is called, it only changes what parameters are passed to that
12190 ld. The ld that is called is determined by the @option{--with-ld}
12191 configure option, GCC's program search path, and finally by the user's
12192 @env{PATH}. The linker used by GCC can be printed using @samp{which
12193 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12194 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12197 @opindex mno-long-calls
12198 Generate code that uses long call sequences. This ensures that a call
12199 is always able to reach linker generated stubs. The default is to generate
12200 long calls only when the distance from the call site to the beginning
12201 of the function or translation unit, as the case may be, exceeds a
12202 predefined limit set by the branch type being used. The limits for
12203 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12204 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12207 Distances are measured from the beginning of functions when using the
12208 @option{-ffunction-sections} option, or when using the @option{-mgas}
12209 and @option{-mno-portable-runtime} options together under HP-UX with
12212 It is normally not desirable to use this option as it will degrade
12213 performance. However, it may be useful in large applications,
12214 particularly when partial linking is used to build the application.
12216 The types of long calls used depends on the capabilities of the
12217 assembler and linker, and the type of code being generated. The
12218 impact on systems that support long absolute calls, and long pic
12219 symbol-difference or pc-relative calls should be relatively small.
12220 However, an indirect call is used on 32-bit ELF systems in pic code
12221 and it is quite long.
12223 @item -munix=@var{unix-std}
12225 Generate compiler predefines and select a startfile for the specified
12226 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12227 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12228 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12229 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12230 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12233 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12234 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12235 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12236 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12237 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12238 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12240 It is @emph{important} to note that this option changes the interfaces
12241 for various library routines. It also affects the operational behavior
12242 of the C library. Thus, @emph{extreme} care is needed in using this
12245 Library code that is intended to operate with more than one UNIX
12246 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12247 as appropriate. Most GNU software doesn't provide this capability.
12251 Suppress the generation of link options to search libdld.sl when the
12252 @option{-static} option is specified on HP-UX 10 and later.
12256 The HP-UX implementation of setlocale in libc has a dependency on
12257 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12258 when the @option{-static} option is specified, special link options
12259 are needed to resolve this dependency.
12261 On HP-UX 10 and later, the GCC driver adds the necessary options to
12262 link with libdld.sl when the @option{-static} option is specified.
12263 This causes the resulting binary to be dynamic. On the 64-bit port,
12264 the linkers generate dynamic binaries by default in any case. The
12265 @option{-nolibdld} option can be used to prevent the GCC driver from
12266 adding these link options.
12270 Add support for multithreading with the @dfn{dce thread} library
12271 under HP-UX@. This option sets flags for both the preprocessor and
12275 @node i386 and x86-64 Options
12276 @subsection Intel 386 and AMD x86-64 Options
12277 @cindex i386 Options
12278 @cindex x86-64 Options
12279 @cindex Intel 386 Options
12280 @cindex AMD x86-64 Options
12282 These @samp{-m} options are defined for the i386 and x86-64 family of
12286 @item -mtune=@var{cpu-type}
12288 Tune to @var{cpu-type} everything applicable about the generated code, except
12289 for the ABI and the set of available instructions. The choices for
12290 @var{cpu-type} are:
12293 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12294 If you know the CPU on which your code will run, then you should use
12295 the corresponding @option{-mtune} option instead of
12296 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12297 of your application will have, then you should use this option.
12299 As new processors are deployed in the marketplace, the behavior of this
12300 option will change. Therefore, if you upgrade to a newer version of
12301 GCC, the code generated option will change to reflect the processors
12302 that were most common when that version of GCC was released.
12304 There is no @option{-march=generic} option because @option{-march}
12305 indicates the instruction set the compiler can use, and there is no
12306 generic instruction set applicable to all processors. In contrast,
12307 @option{-mtune} indicates the processor (or, in this case, collection of
12308 processors) for which the code is optimized.
12310 This selects the CPU to tune for at compilation time by determining
12311 the processor type of the compiling machine. Using @option{-mtune=native}
12312 will produce code optimized for the local machine under the constraints
12313 of the selected instruction set. Using @option{-march=native} will
12314 enable all instruction subsets supported by the local machine (hence
12315 the result might not run on different machines).
12317 Original Intel's i386 CPU@.
12319 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12320 @item i586, pentium
12321 Intel Pentium CPU with no MMX support.
12323 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12325 Intel PentiumPro CPU@.
12327 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12328 instruction set will be used, so the code will run on all i686 family chips.
12330 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12331 @item pentium3, pentium3m
12332 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12335 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12336 support. Used by Centrino notebooks.
12337 @item pentium4, pentium4m
12338 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12340 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12343 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12344 SSE2 and SSE3 instruction set support.
12346 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12347 instruction set support.
12349 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12350 and SSE4.2 instruction set support.
12352 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12353 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12355 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12356 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
12359 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12360 instruction set support.
12362 AMD K6 CPU with MMX instruction set support.
12364 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12365 @item athlon, athlon-tbird
12366 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12368 @item athlon-4, athlon-xp, athlon-mp
12369 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12370 instruction set support.
12371 @item k8, opteron, athlon64, athlon-fx
12372 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12373 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12374 @item k8-sse3, opteron-sse3, athlon64-sse3
12375 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12376 @item amdfam10, barcelona
12377 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12378 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12379 instruction set extensions.)
12381 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12384 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12385 instruction set support.
12387 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12388 implemented for this chip.)
12390 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12391 implemented for this chip.)
12393 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12396 While picking a specific @var{cpu-type} will schedule things appropriately
12397 for that particular chip, the compiler will not generate any code that
12398 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12401 @item -march=@var{cpu-type}
12403 Generate instructions for the machine type @var{cpu-type}. The choices
12404 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12405 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12407 @item -mcpu=@var{cpu-type}
12409 A deprecated synonym for @option{-mtune}.
12411 @item -mfpmath=@var{unit}
12413 Generate floating point arithmetics for selected unit @var{unit}. The choices
12414 for @var{unit} are:
12418 Use the standard 387 floating point coprocessor present majority of chips and
12419 emulated otherwise. Code compiled with this option will run almost everywhere.
12420 The temporary results are computed in 80bit precision instead of precision
12421 specified by the type resulting in slightly different results compared to most
12422 of other chips. See @option{-ffloat-store} for more detailed description.
12424 This is the default choice for i386 compiler.
12427 Use scalar floating point instructions present in the SSE instruction set.
12428 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12429 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12430 instruction set supports only single precision arithmetics, thus the double and
12431 extended precision arithmetics is still done using 387. Later version, present
12432 only in Pentium4 and the future AMD x86-64 chips supports double precision
12435 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12436 or @option{-msse2} switches to enable SSE extensions and make this option
12437 effective. For the x86-64 compiler, these extensions are enabled by default.
12439 The resulting code should be considerably faster in the majority of cases and avoid
12440 the numerical instability problems of 387 code, but may break some existing
12441 code that expects temporaries to be 80bit.
12443 This is the default choice for the x86-64 compiler.
12448 Attempt to utilize both instruction sets at once. This effectively double the
12449 amount of available registers and on chips with separate execution units for
12450 387 and SSE the execution resources too. Use this option with care, as it is
12451 still experimental, because the GCC register allocator does not model separate
12452 functional units well resulting in instable performance.
12455 @item -masm=@var{dialect}
12456 @opindex masm=@var{dialect}
12457 Output asm instructions using selected @var{dialect}. Supported
12458 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12459 not support @samp{intel}.
12462 @itemx -mno-ieee-fp
12464 @opindex mno-ieee-fp
12465 Control whether or not the compiler uses IEEE floating point
12466 comparisons. These handle correctly the case where the result of a
12467 comparison is unordered.
12470 @opindex msoft-float
12471 Generate output containing library calls for floating point.
12472 @strong{Warning:} the requisite libraries are not part of GCC@.
12473 Normally the facilities of the machine's usual C compiler are used, but
12474 this can't be done directly in cross-compilation. You must make your
12475 own arrangements to provide suitable library functions for
12478 On machines where a function returns floating point results in the 80387
12479 register stack, some floating point opcodes may be emitted even if
12480 @option{-msoft-float} is used.
12482 @item -mno-fp-ret-in-387
12483 @opindex mno-fp-ret-in-387
12484 Do not use the FPU registers for return values of functions.
12486 The usual calling convention has functions return values of types
12487 @code{float} and @code{double} in an FPU register, even if there
12488 is no FPU@. The idea is that the operating system should emulate
12491 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12492 in ordinary CPU registers instead.
12494 @item -mno-fancy-math-387
12495 @opindex mno-fancy-math-387
12496 Some 387 emulators do not support the @code{sin}, @code{cos} and
12497 @code{sqrt} instructions for the 387. Specify this option to avoid
12498 generating those instructions. This option is the default on FreeBSD,
12499 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12500 indicates that the target CPU will always have an FPU and so the
12501 instruction will not need emulation. As of revision 2.6.1, these
12502 instructions are not generated unless you also use the
12503 @option{-funsafe-math-optimizations} switch.
12505 @item -malign-double
12506 @itemx -mno-align-double
12507 @opindex malign-double
12508 @opindex mno-align-double
12509 Control whether GCC aligns @code{double}, @code{long double}, and
12510 @code{long long} variables on a two word boundary or a one word
12511 boundary. Aligning @code{double} variables on a two word boundary will
12512 produce code that runs somewhat faster on a @samp{Pentium} at the
12513 expense of more memory.
12515 On x86-64, @option{-malign-double} is enabled by default.
12517 @strong{Warning:} if you use the @option{-malign-double} switch,
12518 structures containing the above types will be aligned differently than
12519 the published application binary interface specifications for the 386
12520 and will not be binary compatible with structures in code compiled
12521 without that switch.
12523 @item -m96bit-long-double
12524 @itemx -m128bit-long-double
12525 @opindex m96bit-long-double
12526 @opindex m128bit-long-double
12527 These switches control the size of @code{long double} type. The i386
12528 application binary interface specifies the size to be 96 bits,
12529 so @option{-m96bit-long-double} is the default in 32 bit mode.
12531 Modern architectures (Pentium and newer) would prefer @code{long double}
12532 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12533 conforming to the ABI, this would not be possible. So specifying a
12534 @option{-m128bit-long-double} will align @code{long double}
12535 to a 16 byte boundary by padding the @code{long double} with an additional
12538 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12539 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12541 Notice that neither of these options enable any extra precision over the x87
12542 standard of 80 bits for a @code{long double}.
12544 @strong{Warning:} if you override the default value for your target ABI, the
12545 structures and arrays containing @code{long double} variables will change
12546 their size as well as function calling convention for function taking
12547 @code{long double} will be modified. Hence they will not be binary
12548 compatible with arrays or structures in code compiled without that switch.
12550 @item -mlarge-data-threshold=@var{number}
12551 @opindex mlarge-data-threshold=@var{number}
12552 When @option{-mcmodel=medium} is specified, the data greater than
12553 @var{threshold} are placed in large data section. This value must be the
12554 same across all object linked into the binary and defaults to 65535.
12558 Use a different function-calling convention, in which functions that
12559 take a fixed number of arguments return with the @code{ret} @var{num}
12560 instruction, which pops their arguments while returning. This saves one
12561 instruction in the caller since there is no need to pop the arguments
12564 You can specify that an individual function is called with this calling
12565 sequence with the function attribute @samp{stdcall}. You can also
12566 override the @option{-mrtd} option by using the function attribute
12567 @samp{cdecl}. @xref{Function Attributes}.
12569 @strong{Warning:} this calling convention is incompatible with the one
12570 normally used on Unix, so you cannot use it if you need to call
12571 libraries compiled with the Unix compiler.
12573 Also, you must provide function prototypes for all functions that
12574 take variable numbers of arguments (including @code{printf});
12575 otherwise incorrect code will be generated for calls to those
12578 In addition, seriously incorrect code will result if you call a
12579 function with too many arguments. (Normally, extra arguments are
12580 harmlessly ignored.)
12582 @item -mregparm=@var{num}
12584 Control how many registers are used to pass integer arguments. By
12585 default, no registers are used to pass arguments, and at most 3
12586 registers can be used. You can control this behavior for a specific
12587 function by using the function attribute @samp{regparm}.
12588 @xref{Function Attributes}.
12590 @strong{Warning:} if you use this switch, and
12591 @var{num} is nonzero, then you must build all modules with the same
12592 value, including any libraries. This includes the system libraries and
12596 @opindex msseregparm
12597 Use SSE register passing conventions for float and double arguments
12598 and return values. You can control this behavior for a specific
12599 function by using the function attribute @samp{sseregparm}.
12600 @xref{Function Attributes}.
12602 @strong{Warning:} if you use this switch then you must build all
12603 modules with the same value, including any libraries. This includes
12604 the system libraries and startup modules.
12606 @item -mvect8-ret-in-mem
12607 @opindex mvect8-ret-in-mem
12608 Return 8-byte vectors in memory instead of MMX registers. This is the
12609 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12610 Studio compilers until version 12. Later compiler versions (starting
12611 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12612 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12613 you need to remain compatible with existing code produced by those
12614 previous compiler versions or older versions of GCC.
12623 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12624 is specified, the significands of results of floating-point operations are
12625 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12626 significands of results of floating-point operations to 53 bits (double
12627 precision) and @option{-mpc80} rounds the significands of results of
12628 floating-point operations to 64 bits (extended double precision), which is
12629 the default. When this option is used, floating-point operations in higher
12630 precisions are not available to the programmer without setting the FPU
12631 control word explicitly.
12633 Setting the rounding of floating-point operations to less than the default
12634 80 bits can speed some programs by 2% or more. Note that some mathematical
12635 libraries assume that extended precision (80 bit) floating-point operations
12636 are enabled by default; routines in such libraries could suffer significant
12637 loss of accuracy, typically through so-called "catastrophic cancellation",
12638 when this option is used to set the precision to less than extended precision.
12640 @item -mstackrealign
12641 @opindex mstackrealign
12642 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12643 option will generate an alternate prologue and epilogue that realigns the
12644 runtime stack if necessary. This supports mixing legacy codes that keep
12645 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12646 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12647 applicable to individual functions.
12649 @item -mpreferred-stack-boundary=@var{num}
12650 @opindex mpreferred-stack-boundary
12651 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12652 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12653 the default is 4 (16 bytes or 128 bits).
12655 @item -mincoming-stack-boundary=@var{num}
12656 @opindex mincoming-stack-boundary
12657 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12658 boundary. If @option{-mincoming-stack-boundary} is not specified,
12659 the one specified by @option{-mpreferred-stack-boundary} will be used.
12661 On Pentium and PentiumPro, @code{double} and @code{long double} values
12662 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12663 suffer significant run time performance penalties. On Pentium III, the
12664 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12665 properly if it is not 16 byte aligned.
12667 To ensure proper alignment of this values on the stack, the stack boundary
12668 must be as aligned as that required by any value stored on the stack.
12669 Further, every function must be generated such that it keeps the stack
12670 aligned. Thus calling a function compiled with a higher preferred
12671 stack boundary from a function compiled with a lower preferred stack
12672 boundary will most likely misalign the stack. It is recommended that
12673 libraries that use callbacks always use the default setting.
12675 This extra alignment does consume extra stack space, and generally
12676 increases code size. Code that is sensitive to stack space usage, such
12677 as embedded systems and operating system kernels, may want to reduce the
12678 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12707 @itemx -mno-fsgsbase
12743 These switches enable or disable the use of instructions in the MMX, SSE,
12744 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
12745 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
12746 @: extended instruction sets.
12747 These extensions are also available as built-in functions: see
12748 @ref{X86 Built-in Functions}, for details of the functions enabled and
12749 disabled by these switches.
12751 To have SSE/SSE2 instructions generated automatically from floating-point
12752 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12754 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12755 generates new AVX instructions or AVX equivalence for all SSEx instructions
12758 These options will enable GCC to use these extended instructions in
12759 generated code, even without @option{-mfpmath=sse}. Applications which
12760 perform runtime CPU detection must compile separate files for each
12761 supported architecture, using the appropriate flags. In particular,
12762 the file containing the CPU detection code should be compiled without
12767 This option instructs GCC to emit a @code{cld} instruction in the prologue
12768 of functions that use string instructions. String instructions depend on
12769 the DF flag to select between autoincrement or autodecrement mode. While the
12770 ABI specifies the DF flag to be cleared on function entry, some operating
12771 systems violate this specification by not clearing the DF flag in their
12772 exception dispatchers. The exception handler can be invoked with the DF flag
12773 set which leads to wrong direction mode, when string instructions are used.
12774 This option can be enabled by default on 32-bit x86 targets by configuring
12775 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12776 instructions can be suppressed with the @option{-mno-cld} compiler option
12780 @opindex mvzeroupper
12781 This option instructs GCC to emit a @code{vzeroupper} instruction
12782 before a transfer of control flow out of the function to minimize
12783 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12788 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12789 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12790 data types. This is useful for high resolution counters that could be updated
12791 by multiple processors (or cores). This instruction is generated as part of
12792 atomic built-in functions: see @ref{Atomic Builtins} for details.
12796 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12797 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12798 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12799 SAHF are load and store instructions, respectively, for certain status flags.
12800 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12801 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12805 This option will enable GCC to use movbe instruction to implement
12806 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12810 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12811 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12812 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12816 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12817 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12818 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12819 variants) for single precision floating point arguments. These instructions
12820 are generated only when @option{-funsafe-math-optimizations} is enabled
12821 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12822 Note that while the throughput of the sequence is higher than the throughput
12823 of the non-reciprocal instruction, the precision of the sequence can be
12824 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12826 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12827 already with @option{-ffast-math} (or the above option combination), and
12828 doesn't need @option{-mrecip}.
12830 @item -mveclibabi=@var{type}
12831 @opindex mveclibabi
12832 Specifies the ABI type to use for vectorizing intrinsics using an
12833 external library. Supported types are @code{svml} for the Intel short
12834 vector math library and @code{acml} for the AMD math core library style
12835 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12836 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12837 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12838 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12839 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12840 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12841 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12842 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12843 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12844 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12845 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12846 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12847 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12848 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12849 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12850 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12851 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12852 compatible library will have to be specified at link time.
12854 @item -mabi=@var{name}
12856 Generate code for the specified calling convention. Permissible values
12857 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12858 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12859 ABI when targeting Windows. On all other systems, the default is the
12860 SYSV ABI. You can control this behavior for a specific function by
12861 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12862 @xref{Function Attributes}.
12864 @item -mtls-dialect=@var{type}
12865 @opindex mtls-dialect
12866 Generate code to access thread-local storage using the @samp{gnu} or
12867 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
12868 @samp{gnu2} is more efficient, but it may add compile- and run-time
12869 requirements that cannot be satisfied on all systems.
12872 @itemx -mno-push-args
12873 @opindex mpush-args
12874 @opindex mno-push-args
12875 Use PUSH operations to store outgoing parameters. This method is shorter
12876 and usually equally fast as method using SUB/MOV operations and is enabled
12877 by default. In some cases disabling it may improve performance because of
12878 improved scheduling and reduced dependencies.
12880 @item -maccumulate-outgoing-args
12881 @opindex maccumulate-outgoing-args
12882 If enabled, the maximum amount of space required for outgoing arguments will be
12883 computed in the function prologue. This is faster on most modern CPUs
12884 because of reduced dependencies, improved scheduling and reduced stack usage
12885 when preferred stack boundary is not equal to 2. The drawback is a notable
12886 increase in code size. This switch implies @option{-mno-push-args}.
12890 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12891 on thread-safe exception handling must compile and link all code with the
12892 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12893 @option{-D_MT}; when linking, it links in a special thread helper library
12894 @option{-lmingwthrd} which cleans up per thread exception handling data.
12896 @item -mno-align-stringops
12897 @opindex mno-align-stringops
12898 Do not align destination of inlined string operations. This switch reduces
12899 code size and improves performance in case the destination is already aligned,
12900 but GCC doesn't know about it.
12902 @item -minline-all-stringops
12903 @opindex minline-all-stringops
12904 By default GCC inlines string operations only when destination is known to be
12905 aligned at least to 4 byte boundary. This enables more inlining, increase code
12906 size, but may improve performance of code that depends on fast memcpy, strlen
12907 and memset for short lengths.
12909 @item -minline-stringops-dynamically
12910 @opindex minline-stringops-dynamically
12911 For string operation of unknown size, inline runtime checks so for small
12912 blocks inline code is used, while for large blocks library call is used.
12914 @item -mstringop-strategy=@var{alg}
12915 @opindex mstringop-strategy=@var{alg}
12916 Overwrite internal decision heuristic about particular algorithm to inline
12917 string operation with. The allowed values are @code{rep_byte},
12918 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12919 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12920 expanding inline loop, @code{libcall} for always expanding library call.
12922 @item -momit-leaf-frame-pointer
12923 @opindex momit-leaf-frame-pointer
12924 Don't keep the frame pointer in a register for leaf functions. This
12925 avoids the instructions to save, set up and restore frame pointers and
12926 makes an extra register available in leaf functions. The option
12927 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12928 which might make debugging harder.
12930 @item -mtls-direct-seg-refs
12931 @itemx -mno-tls-direct-seg-refs
12932 @opindex mtls-direct-seg-refs
12933 Controls whether TLS variables may be accessed with offsets from the
12934 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12935 or whether the thread base pointer must be added. Whether or not this
12936 is legal depends on the operating system, and whether it maps the
12937 segment to cover the entire TLS area.
12939 For systems that use GNU libc, the default is on.
12942 @itemx -mno-sse2avx
12944 Specify that the assembler should encode SSE instructions with VEX
12945 prefix. The option @option{-mavx} turns this on by default.
12950 If profiling is active @option{-pg} put the profiling
12951 counter call before prologue.
12952 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12953 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12956 @itemx -mno-8bit-idiv
12958 On some processors, like Intel Atom, 8bit unsigned integer divide is
12959 much faster than 32bit/64bit integer divide. This option will generate a
12960 runt-time check. If both dividend and divisor are within range of 0
12961 to 255, 8bit unsigned integer divide will be used instead of
12962 32bit/64bit integer divide.
12964 @item -mavx256-split-unaligned-load
12965 @item -mavx256-split-unaligned-store
12966 @opindex avx256-split-unaligned-load
12967 @opindex avx256-split-unaligned-store
12968 Split 32-byte AVX unaligned load and store.
12972 These @samp{-m} switches are supported in addition to the above
12973 on AMD x86-64 processors in 64-bit environments.
12982 Generate code for a 32-bit or 64-bit environment.
12983 The @option{-m32} option sets int, long and pointer to 32 bits and
12984 generates code that runs on any i386 system.
12985 The @option{-m64} option sets int to 32 bits and long and pointer
12986 to 64 bits and generates code for AMD's x86-64 architecture.
12987 The @option{-mx32} option sets int, long and pointer to 32 bits and
12988 generates code for AMD's x86-64 architecture.
12989 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
12990 and @option{-mdynamic-no-pic} options.
12992 @item -mno-red-zone
12993 @opindex mno-red-zone
12994 Do not use a so called red zone for x86-64 code. The red zone is mandated
12995 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12996 stack pointer that will not be modified by signal or interrupt handlers
12997 and therefore can be used for temporary data without adjusting the stack
12998 pointer. The flag @option{-mno-red-zone} disables this red zone.
13000 @item -mcmodel=small
13001 @opindex mcmodel=small
13002 Generate code for the small code model: the program and its symbols must
13003 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13004 Programs can be statically or dynamically linked. This is the default
13007 @item -mcmodel=kernel
13008 @opindex mcmodel=kernel
13009 Generate code for the kernel code model. The kernel runs in the
13010 negative 2 GB of the address space.
13011 This model has to be used for Linux kernel code.
13013 @item -mcmodel=medium
13014 @opindex mcmodel=medium
13015 Generate code for the medium model: The program is linked in the lower 2
13016 GB of the address space. Small symbols are also placed there. Symbols
13017 with sizes larger than @option{-mlarge-data-threshold} are put into
13018 large data or bss sections and can be located above 2GB. Programs can
13019 be statically or dynamically linked.
13021 @item -mcmodel=large
13022 @opindex mcmodel=large
13023 Generate code for the large model: This model makes no assumptions
13024 about addresses and sizes of sections.
13027 @node i386 and x86-64 Windows Options
13028 @subsection i386 and x86-64 Windows Options
13029 @cindex i386 and x86-64 Windows Options
13031 These additional options are available for Windows targets:
13036 This option is available for Cygwin and MinGW targets. It
13037 specifies that a console application is to be generated, by
13038 instructing the linker to set the PE header subsystem type
13039 required for console applications.
13040 This is the default behavior for Cygwin and MinGW targets.
13044 This option is available for Cygwin and MinGW targets. It
13045 specifies that a DLL - a dynamic link library - is to be
13046 generated, enabling the selection of the required runtime
13047 startup object and entry point.
13049 @item -mnop-fun-dllimport
13050 @opindex mnop-fun-dllimport
13051 This option is available for Cygwin and MinGW targets. It
13052 specifies that the dllimport attribute should be ignored.
13056 This option is available for MinGW targets. It specifies
13057 that MinGW-specific thread support is to be used.
13061 This option is available for mingw-w64 targets. It specifies
13062 that the UNICODE macro is getting pre-defined and that the
13063 unicode capable runtime startup code is chosen.
13067 This option is available for Cygwin and MinGW targets. It
13068 specifies that the typical Windows pre-defined macros are to
13069 be set in the pre-processor, but does not influence the choice
13070 of runtime library/startup code.
13074 This option is available for Cygwin and MinGW targets. It
13075 specifies that a GUI application is to be generated by
13076 instructing the linker to set the PE header subsystem type
13079 @item -fno-set-stack-executable
13080 @opindex fno-set-stack-executable
13081 This option is available for MinGW targets. It specifies that
13082 the executable flag for stack used by nested functions isn't
13083 set. This is necessary for binaries running in kernel mode of
13084 Windows, as there the user32 API, which is used to set executable
13085 privileges, isn't available.
13087 @item -mpe-aligned-commons
13088 @opindex mpe-aligned-commons
13089 This option is available for Cygwin and MinGW targets. It
13090 specifies that the GNU extension to the PE file format that
13091 permits the correct alignment of COMMON variables should be
13092 used when generating code. It will be enabled by default if
13093 GCC detects that the target assembler found during configuration
13094 supports the feature.
13097 See also under @ref{i386 and x86-64 Options} for standard options.
13099 @node IA-64 Options
13100 @subsection IA-64 Options
13101 @cindex IA-64 Options
13103 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13107 @opindex mbig-endian
13108 Generate code for a big endian target. This is the default for HP-UX@.
13110 @item -mlittle-endian
13111 @opindex mlittle-endian
13112 Generate code for a little endian target. This is the default for AIX5
13118 @opindex mno-gnu-as
13119 Generate (or don't) code for the GNU assembler. This is the default.
13120 @c Also, this is the default if the configure option @option{--with-gnu-as}
13126 @opindex mno-gnu-ld
13127 Generate (or don't) code for the GNU linker. This is the default.
13128 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13133 Generate code that does not use a global pointer register. The result
13134 is not position independent code, and violates the IA-64 ABI@.
13136 @item -mvolatile-asm-stop
13137 @itemx -mno-volatile-asm-stop
13138 @opindex mvolatile-asm-stop
13139 @opindex mno-volatile-asm-stop
13140 Generate (or don't) a stop bit immediately before and after volatile asm
13143 @item -mregister-names
13144 @itemx -mno-register-names
13145 @opindex mregister-names
13146 @opindex mno-register-names
13147 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13148 the stacked registers. This may make assembler output more readable.
13154 Disable (or enable) optimizations that use the small data section. This may
13155 be useful for working around optimizer bugs.
13157 @item -mconstant-gp
13158 @opindex mconstant-gp
13159 Generate code that uses a single constant global pointer value. This is
13160 useful when compiling kernel code.
13164 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13165 This is useful when compiling firmware code.
13167 @item -minline-float-divide-min-latency
13168 @opindex minline-float-divide-min-latency
13169 Generate code for inline divides of floating point values
13170 using the minimum latency algorithm.
13172 @item -minline-float-divide-max-throughput
13173 @opindex minline-float-divide-max-throughput
13174 Generate code for inline divides of floating point values
13175 using the maximum throughput algorithm.
13177 @item -mno-inline-float-divide
13178 @opindex mno-inline-float-divide
13179 Do not generate inline code for divides of floating point values.
13181 @item -minline-int-divide-min-latency
13182 @opindex minline-int-divide-min-latency
13183 Generate code for inline divides of integer values
13184 using the minimum latency algorithm.
13186 @item -minline-int-divide-max-throughput
13187 @opindex minline-int-divide-max-throughput
13188 Generate code for inline divides of integer values
13189 using the maximum throughput algorithm.
13191 @item -mno-inline-int-divide
13192 @opindex mno-inline-int-divide
13193 Do not generate inline code for divides of integer values.
13195 @item -minline-sqrt-min-latency
13196 @opindex minline-sqrt-min-latency
13197 Generate code for inline square roots
13198 using the minimum latency algorithm.
13200 @item -minline-sqrt-max-throughput
13201 @opindex minline-sqrt-max-throughput
13202 Generate code for inline square roots
13203 using the maximum throughput algorithm.
13205 @item -mno-inline-sqrt
13206 @opindex mno-inline-sqrt
13207 Do not generate inline code for sqrt.
13210 @itemx -mno-fused-madd
13211 @opindex mfused-madd
13212 @opindex mno-fused-madd
13213 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13214 instructions. The default is to use these instructions.
13216 @item -mno-dwarf2-asm
13217 @itemx -mdwarf2-asm
13218 @opindex mno-dwarf2-asm
13219 @opindex mdwarf2-asm
13220 Don't (or do) generate assembler code for the DWARF2 line number debugging
13221 info. This may be useful when not using the GNU assembler.
13223 @item -mearly-stop-bits
13224 @itemx -mno-early-stop-bits
13225 @opindex mearly-stop-bits
13226 @opindex mno-early-stop-bits
13227 Allow stop bits to be placed earlier than immediately preceding the
13228 instruction that triggered the stop bit. This can improve instruction
13229 scheduling, but does not always do so.
13231 @item -mfixed-range=@var{register-range}
13232 @opindex mfixed-range
13233 Generate code treating the given register range as fixed registers.
13234 A fixed register is one that the register allocator can not use. This is
13235 useful when compiling kernel code. A register range is specified as
13236 two registers separated by a dash. Multiple register ranges can be
13237 specified separated by a comma.
13239 @item -mtls-size=@var{tls-size}
13241 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13244 @item -mtune=@var{cpu-type}
13246 Tune the instruction scheduling for a particular CPU, Valid values are
13247 itanium, itanium1, merced, itanium2, and mckinley.
13253 Generate code for a 32-bit or 64-bit environment.
13254 The 32-bit environment sets int, long and pointer to 32 bits.
13255 The 64-bit environment sets int to 32 bits and long and pointer
13256 to 64 bits. These are HP-UX specific flags.
13258 @item -mno-sched-br-data-spec
13259 @itemx -msched-br-data-spec
13260 @opindex mno-sched-br-data-spec
13261 @opindex msched-br-data-spec
13262 (Dis/En)able data speculative scheduling before reload.
13263 This will result in generation of the ld.a instructions and
13264 the corresponding check instructions (ld.c / chk.a).
13265 The default is 'disable'.
13267 @item -msched-ar-data-spec
13268 @itemx -mno-sched-ar-data-spec
13269 @opindex msched-ar-data-spec
13270 @opindex mno-sched-ar-data-spec
13271 (En/Dis)able data speculative scheduling after reload.
13272 This will result in generation of the ld.a instructions and
13273 the corresponding check instructions (ld.c / chk.a).
13274 The default is 'enable'.
13276 @item -mno-sched-control-spec
13277 @itemx -msched-control-spec
13278 @opindex mno-sched-control-spec
13279 @opindex msched-control-spec
13280 (Dis/En)able control speculative scheduling. This feature is
13281 available only during region scheduling (i.e.@: before reload).
13282 This will result in generation of the ld.s instructions and
13283 the corresponding check instructions chk.s .
13284 The default is 'disable'.
13286 @item -msched-br-in-data-spec
13287 @itemx -mno-sched-br-in-data-spec
13288 @opindex msched-br-in-data-spec
13289 @opindex mno-sched-br-in-data-spec
13290 (En/Dis)able speculative scheduling of the instructions that
13291 are dependent on the data speculative loads before reload.
13292 This is effective only with @option{-msched-br-data-spec} enabled.
13293 The default is 'enable'.
13295 @item -msched-ar-in-data-spec
13296 @itemx -mno-sched-ar-in-data-spec
13297 @opindex msched-ar-in-data-spec
13298 @opindex mno-sched-ar-in-data-spec
13299 (En/Dis)able speculative scheduling of the instructions that
13300 are dependent on the data speculative loads after reload.
13301 This is effective only with @option{-msched-ar-data-spec} enabled.
13302 The default is 'enable'.
13304 @item -msched-in-control-spec
13305 @itemx -mno-sched-in-control-spec
13306 @opindex msched-in-control-spec
13307 @opindex mno-sched-in-control-spec
13308 (En/Dis)able speculative scheduling of the instructions that
13309 are dependent on the control speculative loads.
13310 This is effective only with @option{-msched-control-spec} enabled.
13311 The default is 'enable'.
13313 @item -mno-sched-prefer-non-data-spec-insns
13314 @itemx -msched-prefer-non-data-spec-insns
13315 @opindex mno-sched-prefer-non-data-spec-insns
13316 @opindex msched-prefer-non-data-spec-insns
13317 If enabled, data speculative instructions will be chosen for schedule
13318 only if there are no other choices at the moment. This will make
13319 the use of the data speculation much more conservative.
13320 The default is 'disable'.
13322 @item -mno-sched-prefer-non-control-spec-insns
13323 @itemx -msched-prefer-non-control-spec-insns
13324 @opindex mno-sched-prefer-non-control-spec-insns
13325 @opindex msched-prefer-non-control-spec-insns
13326 If enabled, control speculative instructions will be chosen for schedule
13327 only if there are no other choices at the moment. This will make
13328 the use of the control speculation much more conservative.
13329 The default is 'disable'.
13331 @item -mno-sched-count-spec-in-critical-path
13332 @itemx -msched-count-spec-in-critical-path
13333 @opindex mno-sched-count-spec-in-critical-path
13334 @opindex msched-count-spec-in-critical-path
13335 If enabled, speculative dependencies will be considered during
13336 computation of the instructions priorities. This will make the use of the
13337 speculation a bit more conservative.
13338 The default is 'disable'.
13340 @item -msched-spec-ldc
13341 @opindex msched-spec-ldc
13342 Use a simple data speculation check. This option is on by default.
13344 @item -msched-control-spec-ldc
13345 @opindex msched-spec-ldc
13346 Use a simple check for control speculation. This option is on by default.
13348 @item -msched-stop-bits-after-every-cycle
13349 @opindex msched-stop-bits-after-every-cycle
13350 Place a stop bit after every cycle when scheduling. This option is on
13353 @item -msched-fp-mem-deps-zero-cost
13354 @opindex msched-fp-mem-deps-zero-cost
13355 Assume that floating-point stores and loads are not likely to cause a conflict
13356 when placed into the same instruction group. This option is disabled by
13359 @item -msel-sched-dont-check-control-spec
13360 @opindex msel-sched-dont-check-control-spec
13361 Generate checks for control speculation in selective scheduling.
13362 This flag is disabled by default.
13364 @item -msched-max-memory-insns=@var{max-insns}
13365 @opindex msched-max-memory-insns
13366 Limit on the number of memory insns per instruction group, giving lower
13367 priority to subsequent memory insns attempting to schedule in the same
13368 instruction group. Frequently useful to prevent cache bank conflicts.
13369 The default value is 1.
13371 @item -msched-max-memory-insns-hard-limit
13372 @opindex msched-max-memory-insns-hard-limit
13373 Disallow more than `msched-max-memory-insns' in instruction group.
13374 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13375 when limit is reached but may still schedule memory operations.
13379 @node IA-64/VMS Options
13380 @subsection IA-64/VMS Options
13382 These @samp{-m} options are defined for the IA-64/VMS implementations:
13385 @item -mvms-return-codes
13386 @opindex mvms-return-codes
13387 Return VMS condition codes from main. The default is to return POSIX
13388 style condition (e.g.@ error) codes.
13390 @item -mdebug-main=@var{prefix}
13391 @opindex mdebug-main=@var{prefix}
13392 Flag the first routine whose name starts with @var{prefix} as the main
13393 routine for the debugger.
13397 Default to 64bit memory allocation routines.
13401 @subsection LM32 Options
13402 @cindex LM32 options
13404 These @option{-m} options are defined for the Lattice Mico32 architecture:
13407 @item -mbarrel-shift-enabled
13408 @opindex mbarrel-shift-enabled
13409 Enable barrel-shift instructions.
13411 @item -mdivide-enabled
13412 @opindex mdivide-enabled
13413 Enable divide and modulus instructions.
13415 @item -mmultiply-enabled
13416 @opindex multiply-enabled
13417 Enable multiply instructions.
13419 @item -msign-extend-enabled
13420 @opindex msign-extend-enabled
13421 Enable sign extend instructions.
13423 @item -muser-enabled
13424 @opindex muser-enabled
13425 Enable user-defined instructions.
13430 @subsection M32C Options
13431 @cindex M32C options
13434 @item -mcpu=@var{name}
13436 Select the CPU for which code is generated. @var{name} may be one of
13437 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13438 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13439 the M32C/80 series.
13443 Specifies that the program will be run on the simulator. This causes
13444 an alternate runtime library to be linked in which supports, for
13445 example, file I/O@. You must not use this option when generating
13446 programs that will run on real hardware; you must provide your own
13447 runtime library for whatever I/O functions are needed.
13449 @item -memregs=@var{number}
13451 Specifies the number of memory-based pseudo-registers GCC will use
13452 during code generation. These pseudo-registers will be used like real
13453 registers, so there is a tradeoff between GCC's ability to fit the
13454 code into available registers, and the performance penalty of using
13455 memory instead of registers. Note that all modules in a program must
13456 be compiled with the same value for this option. Because of that, you
13457 must not use this option with the default runtime libraries gcc
13462 @node M32R/D Options
13463 @subsection M32R/D Options
13464 @cindex M32R/D options
13466 These @option{-m} options are defined for Renesas M32R/D architectures:
13471 Generate code for the M32R/2@.
13475 Generate code for the M32R/X@.
13479 Generate code for the M32R@. This is the default.
13481 @item -mmodel=small
13482 @opindex mmodel=small
13483 Assume all objects live in the lower 16MB of memory (so that their addresses
13484 can be loaded with the @code{ld24} instruction), and assume all subroutines
13485 are reachable with the @code{bl} instruction.
13486 This is the default.
13488 The addressability of a particular object can be set with the
13489 @code{model} attribute.
13491 @item -mmodel=medium
13492 @opindex mmodel=medium
13493 Assume objects may be anywhere in the 32-bit address space (the compiler
13494 will generate @code{seth/add3} instructions to load their addresses), and
13495 assume all subroutines are reachable with the @code{bl} instruction.
13497 @item -mmodel=large
13498 @opindex mmodel=large
13499 Assume objects may be anywhere in the 32-bit address space (the compiler
13500 will generate @code{seth/add3} instructions to load their addresses), and
13501 assume subroutines may not be reachable with the @code{bl} instruction
13502 (the compiler will generate the much slower @code{seth/add3/jl}
13503 instruction sequence).
13506 @opindex msdata=none
13507 Disable use of the small data area. Variables will be put into
13508 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13509 @code{section} attribute has been specified).
13510 This is the default.
13512 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13513 Objects may be explicitly put in the small data area with the
13514 @code{section} attribute using one of these sections.
13516 @item -msdata=sdata
13517 @opindex msdata=sdata
13518 Put small global and static data in the small data area, but do not
13519 generate special code to reference them.
13522 @opindex msdata=use
13523 Put small global and static data in the small data area, and generate
13524 special instructions to reference them.
13528 @cindex smaller data references
13529 Put global and static objects less than or equal to @var{num} bytes
13530 into the small data or bss sections instead of the normal data or bss
13531 sections. The default value of @var{num} is 8.
13532 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13533 for this option to have any effect.
13535 All modules should be compiled with the same @option{-G @var{num}} value.
13536 Compiling with different values of @var{num} may or may not work; if it
13537 doesn't the linker will give an error message---incorrect code will not be
13542 Makes the M32R specific code in the compiler display some statistics
13543 that might help in debugging programs.
13545 @item -malign-loops
13546 @opindex malign-loops
13547 Align all loops to a 32-byte boundary.
13549 @item -mno-align-loops
13550 @opindex mno-align-loops
13551 Do not enforce a 32-byte alignment for loops. This is the default.
13553 @item -missue-rate=@var{number}
13554 @opindex missue-rate=@var{number}
13555 Issue @var{number} instructions per cycle. @var{number} can only be 1
13558 @item -mbranch-cost=@var{number}
13559 @opindex mbranch-cost=@var{number}
13560 @var{number} can only be 1 or 2. If it is 1 then branches will be
13561 preferred over conditional code, if it is 2, then the opposite will
13564 @item -mflush-trap=@var{number}
13565 @opindex mflush-trap=@var{number}
13566 Specifies the trap number to use to flush the cache. The default is
13567 12. Valid numbers are between 0 and 15 inclusive.
13569 @item -mno-flush-trap
13570 @opindex mno-flush-trap
13571 Specifies that the cache cannot be flushed by using a trap.
13573 @item -mflush-func=@var{name}
13574 @opindex mflush-func=@var{name}
13575 Specifies the name of the operating system function to call to flush
13576 the cache. The default is @emph{_flush_cache}, but a function call
13577 will only be used if a trap is not available.
13579 @item -mno-flush-func
13580 @opindex mno-flush-func
13581 Indicates that there is no OS function for flushing the cache.
13585 @node M680x0 Options
13586 @subsection M680x0 Options
13587 @cindex M680x0 options
13589 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13590 The default settings depend on which architecture was selected when
13591 the compiler was configured; the defaults for the most common choices
13595 @item -march=@var{arch}
13597 Generate code for a specific M680x0 or ColdFire instruction set
13598 architecture. Permissible values of @var{arch} for M680x0
13599 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13600 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13601 architectures are selected according to Freescale's ISA classification
13602 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13603 @samp{isab} and @samp{isac}.
13605 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13606 code for a ColdFire target. The @var{arch} in this macro is one of the
13607 @option{-march} arguments given above.
13609 When used together, @option{-march} and @option{-mtune} select code
13610 that runs on a family of similar processors but that is optimized
13611 for a particular microarchitecture.
13613 @item -mcpu=@var{cpu}
13615 Generate code for a specific M680x0 or ColdFire processor.
13616 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13617 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13618 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13619 below, which also classifies the CPUs into families:
13621 @multitable @columnfractions 0.20 0.80
13622 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13623 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13624 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13625 @item @samp{5206e} @tab @samp{5206e}
13626 @item @samp{5208} @tab @samp{5207} @samp{5208}
13627 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13628 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13629 @item @samp{5216} @tab @samp{5214} @samp{5216}
13630 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13631 @item @samp{5225} @tab @samp{5224} @samp{5225}
13632 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13633 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13634 @item @samp{5249} @tab @samp{5249}
13635 @item @samp{5250} @tab @samp{5250}
13636 @item @samp{5271} @tab @samp{5270} @samp{5271}
13637 @item @samp{5272} @tab @samp{5272}
13638 @item @samp{5275} @tab @samp{5274} @samp{5275}
13639 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13640 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13641 @item @samp{5307} @tab @samp{5307}
13642 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13643 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13644 @item @samp{5407} @tab @samp{5407}
13645 @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}
13648 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13649 @var{arch} is compatible with @var{cpu}. Other combinations of
13650 @option{-mcpu} and @option{-march} are rejected.
13652 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13653 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13654 where the value of @var{family} is given by the table above.
13656 @item -mtune=@var{tune}
13658 Tune the code for a particular microarchitecture, within the
13659 constraints set by @option{-march} and @option{-mcpu}.
13660 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13661 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13662 and @samp{cpu32}. The ColdFire microarchitectures
13663 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13665 You can also use @option{-mtune=68020-40} for code that needs
13666 to run relatively well on 68020, 68030 and 68040 targets.
13667 @option{-mtune=68020-60} is similar but includes 68060 targets
13668 as well. These two options select the same tuning decisions as
13669 @option{-m68020-40} and @option{-m68020-60} respectively.
13671 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13672 when tuning for 680x0 architecture @var{arch}. It also defines
13673 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13674 option is used. If gcc is tuning for a range of architectures,
13675 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13676 it defines the macros for every architecture in the range.
13678 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13679 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13680 of the arguments given above.
13686 Generate output for a 68000. This is the default
13687 when the compiler is configured for 68000-based systems.
13688 It is equivalent to @option{-march=68000}.
13690 Use this option for microcontrollers with a 68000 or EC000 core,
13691 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13695 Generate output for a 68010. This is the default
13696 when the compiler is configured for 68010-based systems.
13697 It is equivalent to @option{-march=68010}.
13703 Generate output for a 68020. This is the default
13704 when the compiler is configured for 68020-based systems.
13705 It is equivalent to @option{-march=68020}.
13709 Generate output for a 68030. This is the default when the compiler is
13710 configured for 68030-based systems. It is equivalent to
13711 @option{-march=68030}.
13715 Generate output for a 68040. This is the default when the compiler is
13716 configured for 68040-based systems. It is equivalent to
13717 @option{-march=68040}.
13719 This option inhibits the use of 68881/68882 instructions that have to be
13720 emulated by software on the 68040. Use this option if your 68040 does not
13721 have code to emulate those instructions.
13725 Generate output for a 68060. This is the default when the compiler is
13726 configured for 68060-based systems. It is equivalent to
13727 @option{-march=68060}.
13729 This option inhibits the use of 68020 and 68881/68882 instructions that
13730 have to be emulated by software on the 68060. Use this option if your 68060
13731 does not have code to emulate those instructions.
13735 Generate output for a CPU32. This is the default
13736 when the compiler is configured for CPU32-based systems.
13737 It is equivalent to @option{-march=cpu32}.
13739 Use this option for microcontrollers with a
13740 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13741 68336, 68340, 68341, 68349 and 68360.
13745 Generate output for a 520X ColdFire CPU@. This is the default
13746 when the compiler is configured for 520X-based systems.
13747 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13748 in favor of that option.
13750 Use this option for microcontroller with a 5200 core, including
13751 the MCF5202, MCF5203, MCF5204 and MCF5206.
13755 Generate output for a 5206e ColdFire CPU@. The option is now
13756 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13760 Generate output for a member of the ColdFire 528X family.
13761 The option is now deprecated in favor of the equivalent
13762 @option{-mcpu=528x}.
13766 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13767 in favor of the equivalent @option{-mcpu=5307}.
13771 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13772 in favor of the equivalent @option{-mcpu=5407}.
13776 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13777 This includes use of hardware floating point instructions.
13778 The option is equivalent to @option{-mcpu=547x}, and is now
13779 deprecated in favor of that option.
13783 Generate output for a 68040, without using any of the new instructions.
13784 This results in code which can run relatively efficiently on either a
13785 68020/68881 or a 68030 or a 68040. The generated code does use the
13786 68881 instructions that are emulated on the 68040.
13788 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13792 Generate output for a 68060, without using any of the new instructions.
13793 This results in code which can run relatively efficiently on either a
13794 68020/68881 or a 68030 or a 68040. The generated code does use the
13795 68881 instructions that are emulated on the 68060.
13797 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13801 @opindex mhard-float
13803 Generate floating-point instructions. This is the default for 68020
13804 and above, and for ColdFire devices that have an FPU@. It defines the
13805 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13806 on ColdFire targets.
13809 @opindex msoft-float
13810 Do not generate floating-point instructions; use library calls instead.
13811 This is the default for 68000, 68010, and 68832 targets. It is also
13812 the default for ColdFire devices that have no FPU.
13818 Generate (do not generate) ColdFire hardware divide and remainder
13819 instructions. If @option{-march} is used without @option{-mcpu},
13820 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13821 architectures. Otherwise, the default is taken from the target CPU
13822 (either the default CPU, or the one specified by @option{-mcpu}). For
13823 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13824 @option{-mcpu=5206e}.
13826 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13830 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13831 Additionally, parameters passed on the stack are also aligned to a
13832 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13836 Do not consider type @code{int} to be 16 bits wide. This is the default.
13839 @itemx -mno-bitfield
13840 @opindex mnobitfield
13841 @opindex mno-bitfield
13842 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13843 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13847 Do use the bit-field instructions. The @option{-m68020} option implies
13848 @option{-mbitfield}. This is the default if you use a configuration
13849 designed for a 68020.
13853 Use a different function-calling convention, in which functions
13854 that take a fixed number of arguments return with the @code{rtd}
13855 instruction, which pops their arguments while returning. This
13856 saves one instruction in the caller since there is no need to pop
13857 the arguments there.
13859 This calling convention is incompatible with the one normally
13860 used on Unix, so you cannot use it if you need to call libraries
13861 compiled with the Unix compiler.
13863 Also, you must provide function prototypes for all functions that
13864 take variable numbers of arguments (including @code{printf});
13865 otherwise incorrect code will be generated for calls to those
13868 In addition, seriously incorrect code will result if you call a
13869 function with too many arguments. (Normally, extra arguments are
13870 harmlessly ignored.)
13872 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13873 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13877 Do not use the calling conventions selected by @option{-mrtd}.
13878 This is the default.
13881 @itemx -mno-align-int
13882 @opindex malign-int
13883 @opindex mno-align-int
13884 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13885 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13886 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13887 Aligning variables on 32-bit boundaries produces code that runs somewhat
13888 faster on processors with 32-bit busses at the expense of more memory.
13890 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13891 align structures containing the above types differently than
13892 most published application binary interface specifications for the m68k.
13896 Use the pc-relative addressing mode of the 68000 directly, instead of
13897 using a global offset table. At present, this option implies @option{-fpic},
13898 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13899 not presently supported with @option{-mpcrel}, though this could be supported for
13900 68020 and higher processors.
13902 @item -mno-strict-align
13903 @itemx -mstrict-align
13904 @opindex mno-strict-align
13905 @opindex mstrict-align
13906 Do not (do) assume that unaligned memory references will be handled by
13910 Generate code that allows the data segment to be located in a different
13911 area of memory from the text segment. This allows for execute in place in
13912 an environment without virtual memory management. This option implies
13915 @item -mno-sep-data
13916 Generate code that assumes that the data segment follows the text segment.
13917 This is the default.
13919 @item -mid-shared-library
13920 Generate code that supports shared libraries via the library ID method.
13921 This allows for execute in place and shared libraries in an environment
13922 without virtual memory management. This option implies @option{-fPIC}.
13924 @item -mno-id-shared-library
13925 Generate code that doesn't assume ID based shared libraries are being used.
13926 This is the default.
13928 @item -mshared-library-id=n
13929 Specified the identification number of the ID based shared library being
13930 compiled. Specifying a value of 0 will generate more compact code, specifying
13931 other values will force the allocation of that number to the current
13932 library but is no more space or time efficient than omitting this option.
13938 When generating position-independent code for ColdFire, generate code
13939 that works if the GOT has more than 8192 entries. This code is
13940 larger and slower than code generated without this option. On M680x0
13941 processors, this option is not needed; @option{-fPIC} suffices.
13943 GCC normally uses a single instruction to load values from the GOT@.
13944 While this is relatively efficient, it only works if the GOT
13945 is smaller than about 64k. Anything larger causes the linker
13946 to report an error such as:
13948 @cindex relocation truncated to fit (ColdFire)
13950 relocation truncated to fit: R_68K_GOT16O foobar
13953 If this happens, you should recompile your code with @option{-mxgot}.
13954 It should then work with very large GOTs. However, code generated with
13955 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13956 the value of a global symbol.
13958 Note that some linkers, including newer versions of the GNU linker,
13959 can create multiple GOTs and sort GOT entries. If you have such a linker,
13960 you should only need to use @option{-mxgot} when compiling a single
13961 object file that accesses more than 8192 GOT entries. Very few do.
13963 These options have no effect unless GCC is generating
13964 position-independent code.
13968 @node MCore Options
13969 @subsection MCore Options
13970 @cindex MCore options
13972 These are the @samp{-m} options defined for the Motorola M*Core
13978 @itemx -mno-hardlit
13980 @opindex mno-hardlit
13981 Inline constants into the code stream if it can be done in two
13982 instructions or less.
13988 Use the divide instruction. (Enabled by default).
13990 @item -mrelax-immediate
13991 @itemx -mno-relax-immediate
13992 @opindex mrelax-immediate
13993 @opindex mno-relax-immediate
13994 Allow arbitrary sized immediates in bit operations.
13996 @item -mwide-bitfields
13997 @itemx -mno-wide-bitfields
13998 @opindex mwide-bitfields
13999 @opindex mno-wide-bitfields
14000 Always treat bit-fields as int-sized.
14002 @item -m4byte-functions
14003 @itemx -mno-4byte-functions
14004 @opindex m4byte-functions
14005 @opindex mno-4byte-functions
14006 Force all functions to be aligned to a four byte boundary.
14008 @item -mcallgraph-data
14009 @itemx -mno-callgraph-data
14010 @opindex mcallgraph-data
14011 @opindex mno-callgraph-data
14012 Emit callgraph information.
14015 @itemx -mno-slow-bytes
14016 @opindex mslow-bytes
14017 @opindex mno-slow-bytes
14018 Prefer word access when reading byte quantities.
14020 @item -mlittle-endian
14021 @itemx -mbig-endian
14022 @opindex mlittle-endian
14023 @opindex mbig-endian
14024 Generate code for a little endian target.
14030 Generate code for the 210 processor.
14034 Assume that run-time support has been provided and so omit the
14035 simulator library (@file{libsim.a)} from the linker command line.
14037 @item -mstack-increment=@var{size}
14038 @opindex mstack-increment
14039 Set the maximum amount for a single stack increment operation. Large
14040 values can increase the speed of programs which contain functions
14041 that need a large amount of stack space, but they can also trigger a
14042 segmentation fault if the stack is extended too much. The default
14048 @subsection MeP Options
14049 @cindex MeP options
14055 Enables the @code{abs} instruction, which is the absolute difference
14056 between two registers.
14060 Enables all the optional instructions - average, multiply, divide, bit
14061 operations, leading zero, absolute difference, min/max, clip, and
14067 Enables the @code{ave} instruction, which computes the average of two
14070 @item -mbased=@var{n}
14072 Variables of size @var{n} bytes or smaller will be placed in the
14073 @code{.based} section by default. Based variables use the @code{$tp}
14074 register as a base register, and there is a 128 byte limit to the
14075 @code{.based} section.
14079 Enables the bit operation instructions - bit test (@code{btstm}), set
14080 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14081 test-and-set (@code{tas}).
14083 @item -mc=@var{name}
14085 Selects which section constant data will be placed in. @var{name} may
14086 be @code{tiny}, @code{near}, or @code{far}.
14090 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14091 useful unless you also provide @code{-mminmax}.
14093 @item -mconfig=@var{name}
14095 Selects one of the build-in core configurations. Each MeP chip has
14096 one or more modules in it; each module has a core CPU and a variety of
14097 coprocessors, optional instructions, and peripherals. The
14098 @code{MeP-Integrator} tool, not part of GCC, provides these
14099 configurations through this option; using this option is the same as
14100 using all the corresponding command line options. The default
14101 configuration is @code{default}.
14105 Enables the coprocessor instructions. By default, this is a 32-bit
14106 coprocessor. Note that the coprocessor is normally enabled via the
14107 @code{-mconfig=} option.
14111 Enables the 32-bit coprocessor's instructions.
14115 Enables the 64-bit coprocessor's instructions.
14119 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14123 Causes constant variables to be placed in the @code{.near} section.
14127 Enables the @code{div} and @code{divu} instructions.
14131 Generate big-endian code.
14135 Generate little-endian code.
14137 @item -mio-volatile
14138 @opindex mio-volatile
14139 Tells the compiler that any variable marked with the @code{io}
14140 attribute is to be considered volatile.
14144 Causes variables to be assigned to the @code{.far} section by default.
14148 Enables the @code{leadz} (leading zero) instruction.
14152 Causes variables to be assigned to the @code{.near} section by default.
14156 Enables the @code{min} and @code{max} instructions.
14160 Enables the multiplication and multiply-accumulate instructions.
14164 Disables all the optional instructions enabled by @code{-mall-opts}.
14168 Enables the @code{repeat} and @code{erepeat} instructions, used for
14169 low-overhead looping.
14173 Causes all variables to default to the @code{.tiny} section. Note
14174 that there is a 65536 byte limit to this section. Accesses to these
14175 variables use the @code{%gp} base register.
14179 Enables the saturation instructions. Note that the compiler does not
14180 currently generate these itself, but this option is included for
14181 compatibility with other tools, like @code{as}.
14185 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14189 Link the simulator runtime libraries.
14193 Link the simulator runtime libraries, excluding built-in support
14194 for reset and exception vectors and tables.
14198 Causes all functions to default to the @code{.far} section. Without
14199 this option, functions default to the @code{.near} section.
14201 @item -mtiny=@var{n}
14203 Variables that are @var{n} bytes or smaller will be allocated to the
14204 @code{.tiny} section. These variables use the @code{$gp} base
14205 register. The default for this option is 4, but note that there's a
14206 65536 byte limit to the @code{.tiny} section.
14210 @node MicroBlaze Options
14211 @subsection MicroBlaze Options
14212 @cindex MicroBlaze Options
14217 @opindex msoft-float
14218 Use software emulation for floating point (default).
14221 @opindex mhard-float
14222 Use hardware floating point instructions.
14226 Do not optimize block moves, use @code{memcpy}.
14228 @item -mno-clearbss
14229 @opindex mno-clearbss
14230 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14232 @item -mcpu=@var{cpu-type}
14234 Use features of and schedule code for given CPU.
14235 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14236 where @var{X} is a major version, @var{YY} is the minor version, and
14237 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14238 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14240 @item -mxl-soft-mul
14241 @opindex mxl-soft-mul
14242 Use software multiply emulation (default).
14244 @item -mxl-soft-div
14245 @opindex mxl-soft-div
14246 Use software emulation for divides (default).
14248 @item -mxl-barrel-shift
14249 @opindex mxl-barrel-shift
14250 Use the hardware barrel shifter.
14252 @item -mxl-pattern-compare
14253 @opindex mxl-pattern-compare
14254 Use pattern compare instructions.
14256 @item -msmall-divides
14257 @opindex msmall-divides
14258 Use table lookup optimization for small signed integer divisions.
14260 @item -mxl-stack-check
14261 @opindex mxl-stack-check
14262 This option is deprecated. Use -fstack-check instead.
14265 @opindex mxl-gp-opt
14266 Use GP relative sdata/sbss sections.
14268 @item -mxl-multiply-high
14269 @opindex mxl-multiply-high
14270 Use multiply high instructions for high part of 32x32 multiply.
14272 @item -mxl-float-convert
14273 @opindex mxl-float-convert
14274 Use hardware floating point conversion instructions.
14276 @item -mxl-float-sqrt
14277 @opindex mxl-float-sqrt
14278 Use hardware floating point square root instruction.
14280 @item -mxl-mode-@var{app-model}
14281 Select application model @var{app-model}. Valid models are
14284 normal executable (default), uses startup code @file{crt0.o}.
14287 for use with Xilinx Microprocessor Debugger (XMD) based
14288 software intrusive debug agent called xmdstub. This uses startup file
14289 @file{crt1.o} and sets the start address of the program to be 0x800.
14292 for applications that are loaded using a bootloader.
14293 This model uses startup file @file{crt2.o} which does not contain a processor
14294 reset vector handler. This is suitable for transferring control on a
14295 processor reset to the bootloader rather than the application.
14298 for applications that do not require any of the
14299 MicroBlaze vectors. This option may be useful for applications running
14300 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14303 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14304 @option{-mxl-mode-@var{app-model}}.
14309 @subsection MIPS Options
14310 @cindex MIPS options
14316 Generate big-endian code.
14320 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14323 @item -march=@var{arch}
14325 Generate code that will run on @var{arch}, which can be the name of a
14326 generic MIPS ISA, or the name of a particular processor.
14328 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14329 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14330 The processor names are:
14331 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14332 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14333 @samp{5kc}, @samp{5kf},
14335 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14336 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14337 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14338 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14339 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14340 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14344 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14345 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14346 @samp{rm7000}, @samp{rm9000},
14347 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14350 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14351 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14353 The special value @samp{from-abi} selects the
14354 most compatible architecture for the selected ABI (that is,
14355 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14357 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
14358 which selects the best architecture option for the host processor.
14359 @option{-march=native} has no effect if GCC does not recognize
14362 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14363 (for example, @samp{-march=r2k}). Prefixes are optional, and
14364 @samp{vr} may be written @samp{r}.
14366 Names of the form @samp{@var{n}f2_1} refer to processors with
14367 FPUs clocked at half the rate of the core, names of the form
14368 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14369 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14370 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14371 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14372 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14373 accepted as synonyms for @samp{@var{n}f1_1}.
14375 GCC defines two macros based on the value of this option. The first
14376 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14377 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14378 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14379 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14380 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14382 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14383 above. In other words, it will have the full prefix and will not
14384 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14385 the macro names the resolved architecture (either @samp{"mips1"} or
14386 @samp{"mips3"}). It names the default architecture when no
14387 @option{-march} option is given.
14389 @item -mtune=@var{arch}
14391 Optimize for @var{arch}. Among other things, this option controls
14392 the way instructions are scheduled, and the perceived cost of arithmetic
14393 operations. The list of @var{arch} values is the same as for
14396 When this option is not used, GCC will optimize for the processor
14397 specified by @option{-march}. By using @option{-march} and
14398 @option{-mtune} together, it is possible to generate code that will
14399 run on a family of processors, but optimize the code for one
14400 particular member of that family.
14402 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14403 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14404 @samp{-march} ones described above.
14408 Equivalent to @samp{-march=mips1}.
14412 Equivalent to @samp{-march=mips2}.
14416 Equivalent to @samp{-march=mips3}.
14420 Equivalent to @samp{-march=mips4}.
14424 Equivalent to @samp{-march=mips32}.
14428 Equivalent to @samp{-march=mips32r2}.
14432 Equivalent to @samp{-march=mips64}.
14436 Equivalent to @samp{-march=mips64r2}.
14441 @opindex mno-mips16
14442 Generate (do not generate) MIPS16 code. If GCC is targetting a
14443 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14445 MIPS16 code generation can also be controlled on a per-function basis
14446 by means of @code{mips16} and @code{nomips16} attributes.
14447 @xref{Function Attributes}, for more information.
14449 @item -mflip-mips16
14450 @opindex mflip-mips16
14451 Generate MIPS16 code on alternating functions. This option is provided
14452 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14453 not intended for ordinary use in compiling user code.
14455 @item -minterlink-mips16
14456 @itemx -mno-interlink-mips16
14457 @opindex minterlink-mips16
14458 @opindex mno-interlink-mips16
14459 Require (do not require) that non-MIPS16 code be link-compatible with
14462 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14463 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14464 therefore disables direct jumps unless GCC knows that the target of the
14465 jump is not MIPS16.
14477 Generate code for the given ABI@.
14479 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14480 generates 64-bit code when you select a 64-bit architecture, but you
14481 can use @option{-mgp32} to get 32-bit code instead.
14483 For information about the O64 ABI, see
14484 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14486 GCC supports a variant of the o32 ABI in which floating-point registers
14487 are 64 rather than 32 bits wide. You can select this combination with
14488 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14489 and @samp{mfhc1} instructions and is therefore only supported for
14490 MIPS32R2 processors.
14492 The register assignments for arguments and return values remain the
14493 same, but each scalar value is passed in a single 64-bit register
14494 rather than a pair of 32-bit registers. For example, scalar
14495 floating-point values are returned in @samp{$f0} only, not a
14496 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14497 remains the same, but all 64 bits are saved.
14500 @itemx -mno-abicalls
14502 @opindex mno-abicalls
14503 Generate (do not generate) code that is suitable for SVR4-style
14504 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14509 Generate (do not generate) code that is fully position-independent,
14510 and that can therefore be linked into shared libraries. This option
14511 only affects @option{-mabicalls}.
14513 All @option{-mabicalls} code has traditionally been position-independent,
14514 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14515 as an extension, the GNU toolchain allows executables to use absolute
14516 accesses for locally-binding symbols. It can also use shorter GP
14517 initialization sequences and generate direct calls to locally-defined
14518 functions. This mode is selected by @option{-mno-shared}.
14520 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14521 objects that can only be linked by the GNU linker. However, the option
14522 does not affect the ABI of the final executable; it only affects the ABI
14523 of relocatable objects. Using @option{-mno-shared} will generally make
14524 executables both smaller and quicker.
14526 @option{-mshared} is the default.
14532 Assume (do not assume) that the static and dynamic linkers
14533 support PLTs and copy relocations. This option only affects
14534 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14535 has no effect without @samp{-msym32}.
14537 You can make @option{-mplt} the default by configuring
14538 GCC with @option{--with-mips-plt}. The default is
14539 @option{-mno-plt} otherwise.
14545 Lift (do not lift) the usual restrictions on the size of the global
14548 GCC normally uses a single instruction to load values from the GOT@.
14549 While this is relatively efficient, it will only work if the GOT
14550 is smaller than about 64k. Anything larger will cause the linker
14551 to report an error such as:
14553 @cindex relocation truncated to fit (MIPS)
14555 relocation truncated to fit: R_MIPS_GOT16 foobar
14558 If this happens, you should recompile your code with @option{-mxgot}.
14559 It should then work with very large GOTs, although it will also be
14560 less efficient, since it will take three instructions to fetch the
14561 value of a global symbol.
14563 Note that some linkers can create multiple GOTs. If you have such a
14564 linker, you should only need to use @option{-mxgot} when a single object
14565 file accesses more than 64k's worth of GOT entries. Very few do.
14567 These options have no effect unless GCC is generating position
14572 Assume that general-purpose registers are 32 bits wide.
14576 Assume that general-purpose registers are 64 bits wide.
14580 Assume that floating-point registers are 32 bits wide.
14584 Assume that floating-point registers are 64 bits wide.
14587 @opindex mhard-float
14588 Use floating-point coprocessor instructions.
14591 @opindex msoft-float
14592 Do not use floating-point coprocessor instructions. Implement
14593 floating-point calculations using library calls instead.
14595 @item -msingle-float
14596 @opindex msingle-float
14597 Assume that the floating-point coprocessor only supports single-precision
14600 @item -mdouble-float
14601 @opindex mdouble-float
14602 Assume that the floating-point coprocessor supports double-precision
14603 operations. This is the default.
14609 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14610 implement atomic memory built-in functions. When neither option is
14611 specified, GCC will use the instructions if the target architecture
14614 @option{-mllsc} is useful if the runtime environment can emulate the
14615 instructions and @option{-mno-llsc} can be useful when compiling for
14616 nonstandard ISAs. You can make either option the default by
14617 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14618 respectively. @option{--with-llsc} is the default for some
14619 configurations; see the installation documentation for details.
14625 Use (do not use) revision 1 of the MIPS DSP ASE@.
14626 @xref{MIPS DSP Built-in Functions}. This option defines the
14627 preprocessor macro @samp{__mips_dsp}. It also defines
14628 @samp{__mips_dsp_rev} to 1.
14634 Use (do not use) revision 2 of the MIPS DSP ASE@.
14635 @xref{MIPS DSP Built-in Functions}. This option defines the
14636 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14637 It also defines @samp{__mips_dsp_rev} to 2.
14640 @itemx -mno-smartmips
14641 @opindex msmartmips
14642 @opindex mno-smartmips
14643 Use (do not use) the MIPS SmartMIPS ASE.
14645 @item -mpaired-single
14646 @itemx -mno-paired-single
14647 @opindex mpaired-single
14648 @opindex mno-paired-single
14649 Use (do not use) paired-single floating-point instructions.
14650 @xref{MIPS Paired-Single Support}. This option requires
14651 hardware floating-point support to be enabled.
14657 Use (do not use) MIPS Digital Media Extension instructions.
14658 This option can only be used when generating 64-bit code and requires
14659 hardware floating-point support to be enabled.
14664 @opindex mno-mips3d
14665 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14666 The option @option{-mips3d} implies @option{-mpaired-single}.
14672 Use (do not use) MT Multithreading instructions.
14676 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14677 an explanation of the default and the way that the pointer size is
14682 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14684 The default size of @code{int}s, @code{long}s and pointers depends on
14685 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14686 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14687 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14688 or the same size as integer registers, whichever is smaller.
14694 Assume (do not assume) that all symbols have 32-bit values, regardless
14695 of the selected ABI@. This option is useful in combination with
14696 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14697 to generate shorter and faster references to symbolic addresses.
14701 Put definitions of externally-visible data in a small data section
14702 if that data is no bigger than @var{num} bytes. GCC can then access
14703 the data more efficiently; see @option{-mgpopt} for details.
14705 The default @option{-G} option depends on the configuration.
14707 @item -mlocal-sdata
14708 @itemx -mno-local-sdata
14709 @opindex mlocal-sdata
14710 @opindex mno-local-sdata
14711 Extend (do not extend) the @option{-G} behavior to local data too,
14712 such as to static variables in C@. @option{-mlocal-sdata} is the
14713 default for all configurations.
14715 If the linker complains that an application is using too much small data,
14716 you might want to try rebuilding the less performance-critical parts with
14717 @option{-mno-local-sdata}. You might also want to build large
14718 libraries with @option{-mno-local-sdata}, so that the libraries leave
14719 more room for the main program.
14721 @item -mextern-sdata
14722 @itemx -mno-extern-sdata
14723 @opindex mextern-sdata
14724 @opindex mno-extern-sdata
14725 Assume (do not assume) that externally-defined data will be in
14726 a small data section if that data is within the @option{-G} limit.
14727 @option{-mextern-sdata} is the default for all configurations.
14729 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14730 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14731 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14732 is placed in a small data section. If @var{Var} is defined by another
14733 module, you must either compile that module with a high-enough
14734 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14735 definition. If @var{Var} is common, you must link the application
14736 with a high-enough @option{-G} setting.
14738 The easiest way of satisfying these restrictions is to compile
14739 and link every module with the same @option{-G} option. However,
14740 you may wish to build a library that supports several different
14741 small data limits. You can do this by compiling the library with
14742 the highest supported @option{-G} setting and additionally using
14743 @option{-mno-extern-sdata} to stop the library from making assumptions
14744 about externally-defined data.
14750 Use (do not use) GP-relative accesses for symbols that are known to be
14751 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14752 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14755 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14756 might not hold the value of @code{_gp}. For example, if the code is
14757 part of a library that might be used in a boot monitor, programs that
14758 call boot monitor routines will pass an unknown value in @code{$gp}.
14759 (In such situations, the boot monitor itself would usually be compiled
14760 with @option{-G0}.)
14762 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14763 @option{-mno-extern-sdata}.
14765 @item -membedded-data
14766 @itemx -mno-embedded-data
14767 @opindex membedded-data
14768 @opindex mno-embedded-data
14769 Allocate variables to the read-only data section first if possible, then
14770 next in the small data section if possible, otherwise in data. This gives
14771 slightly slower code than the default, but reduces the amount of RAM required
14772 when executing, and thus may be preferred for some embedded systems.
14774 @item -muninit-const-in-rodata
14775 @itemx -mno-uninit-const-in-rodata
14776 @opindex muninit-const-in-rodata
14777 @opindex mno-uninit-const-in-rodata
14778 Put uninitialized @code{const} variables in the read-only data section.
14779 This option is only meaningful in conjunction with @option{-membedded-data}.
14781 @item -mcode-readable=@var{setting}
14782 @opindex mcode-readable
14783 Specify whether GCC may generate code that reads from executable sections.
14784 There are three possible settings:
14787 @item -mcode-readable=yes
14788 Instructions may freely access executable sections. This is the
14791 @item -mcode-readable=pcrel
14792 MIPS16 PC-relative load instructions can access executable sections,
14793 but other instructions must not do so. This option is useful on 4KSc
14794 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14795 It is also useful on processors that can be configured to have a dual
14796 instruction/data SRAM interface and that, like the M4K, automatically
14797 redirect PC-relative loads to the instruction RAM.
14799 @item -mcode-readable=no
14800 Instructions must not access executable sections. This option can be
14801 useful on targets that are configured to have a dual instruction/data
14802 SRAM interface but that (unlike the M4K) do not automatically redirect
14803 PC-relative loads to the instruction RAM.
14806 @item -msplit-addresses
14807 @itemx -mno-split-addresses
14808 @opindex msplit-addresses
14809 @opindex mno-split-addresses
14810 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14811 relocation operators. This option has been superseded by
14812 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14814 @item -mexplicit-relocs
14815 @itemx -mno-explicit-relocs
14816 @opindex mexplicit-relocs
14817 @opindex mno-explicit-relocs
14818 Use (do not use) assembler relocation operators when dealing with symbolic
14819 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14820 is to use assembler macros instead.
14822 @option{-mexplicit-relocs} is the default if GCC was configured
14823 to use an assembler that supports relocation operators.
14825 @item -mcheck-zero-division
14826 @itemx -mno-check-zero-division
14827 @opindex mcheck-zero-division
14828 @opindex mno-check-zero-division
14829 Trap (do not trap) on integer division by zero.
14831 The default is @option{-mcheck-zero-division}.
14833 @item -mdivide-traps
14834 @itemx -mdivide-breaks
14835 @opindex mdivide-traps
14836 @opindex mdivide-breaks
14837 MIPS systems check for division by zero by generating either a
14838 conditional trap or a break instruction. Using traps results in
14839 smaller code, but is only supported on MIPS II and later. Also, some
14840 versions of the Linux kernel have a bug that prevents trap from
14841 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14842 allow conditional traps on architectures that support them and
14843 @option{-mdivide-breaks} to force the use of breaks.
14845 The default is usually @option{-mdivide-traps}, but this can be
14846 overridden at configure time using @option{--with-divide=breaks}.
14847 Divide-by-zero checks can be completely disabled using
14848 @option{-mno-check-zero-division}.
14853 @opindex mno-memcpy
14854 Force (do not force) the use of @code{memcpy()} for non-trivial block
14855 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14856 most constant-sized copies.
14859 @itemx -mno-long-calls
14860 @opindex mlong-calls
14861 @opindex mno-long-calls
14862 Disable (do not disable) use of the @code{jal} instruction. Calling
14863 functions using @code{jal} is more efficient but requires the caller
14864 and callee to be in the same 256 megabyte segment.
14866 This option has no effect on abicalls code. The default is
14867 @option{-mno-long-calls}.
14873 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14874 instructions, as provided by the R4650 ISA@.
14877 @itemx -mno-fused-madd
14878 @opindex mfused-madd
14879 @opindex mno-fused-madd
14880 Enable (disable) use of the floating point multiply-accumulate
14881 instructions, when they are available. The default is
14882 @option{-mfused-madd}.
14884 When multiply-accumulate instructions are used, the intermediate
14885 product is calculated to infinite precision and is not subject to
14886 the FCSR Flush to Zero bit. This may be undesirable in some
14891 Tell the MIPS assembler to not run its preprocessor over user
14892 assembler files (with a @samp{.s} suffix) when assembling them.
14897 @opindex mno-fix-24k
14898 Work around the 24K E48 (lost data on stores during refill) errata.
14899 The workarounds are implemented by the assembler rather than by GCC.
14902 @itemx -mno-fix-r4000
14903 @opindex mfix-r4000
14904 @opindex mno-fix-r4000
14905 Work around certain R4000 CPU errata:
14908 A double-word or a variable shift may give an incorrect result if executed
14909 immediately after starting an integer division.
14911 A double-word or a variable shift may give an incorrect result if executed
14912 while an integer multiplication is in progress.
14914 An integer division may give an incorrect result if started in a delay slot
14915 of a taken branch or a jump.
14919 @itemx -mno-fix-r4400
14920 @opindex mfix-r4400
14921 @opindex mno-fix-r4400
14922 Work around certain R4400 CPU errata:
14925 A double-word or a variable shift may give an incorrect result if executed
14926 immediately after starting an integer division.
14930 @itemx -mno-fix-r10000
14931 @opindex mfix-r10000
14932 @opindex mno-fix-r10000
14933 Work around certain R10000 errata:
14936 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14937 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14940 This option can only be used if the target architecture supports
14941 branch-likely instructions. @option{-mfix-r10000} is the default when
14942 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14946 @itemx -mno-fix-vr4120
14947 @opindex mfix-vr4120
14948 Work around certain VR4120 errata:
14951 @code{dmultu} does not always produce the correct result.
14953 @code{div} and @code{ddiv} do not always produce the correct result if one
14954 of the operands is negative.
14956 The workarounds for the division errata rely on special functions in
14957 @file{libgcc.a}. At present, these functions are only provided by
14958 the @code{mips64vr*-elf} configurations.
14960 Other VR4120 errata require a nop to be inserted between certain pairs of
14961 instructions. These errata are handled by the assembler, not by GCC itself.
14964 @opindex mfix-vr4130
14965 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14966 workarounds are implemented by the assembler rather than by GCC,
14967 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14968 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14969 instructions are available instead.
14972 @itemx -mno-fix-sb1
14974 Work around certain SB-1 CPU core errata.
14975 (This flag currently works around the SB-1 revision 2
14976 ``F1'' and ``F2'' floating point errata.)
14978 @item -mr10k-cache-barrier=@var{setting}
14979 @opindex mr10k-cache-barrier
14980 Specify whether GCC should insert cache barriers to avoid the
14981 side-effects of speculation on R10K processors.
14983 In common with many processors, the R10K tries to predict the outcome
14984 of a conditional branch and speculatively executes instructions from
14985 the ``taken'' branch. It later aborts these instructions if the
14986 predicted outcome was wrong. However, on the R10K, even aborted
14987 instructions can have side effects.
14989 This problem only affects kernel stores and, depending on the system,
14990 kernel loads. As an example, a speculatively-executed store may load
14991 the target memory into cache and mark the cache line as dirty, even if
14992 the store itself is later aborted. If a DMA operation writes to the
14993 same area of memory before the ``dirty'' line is flushed, the cached
14994 data will overwrite the DMA-ed data. See the R10K processor manual
14995 for a full description, including other potential problems.
14997 One workaround is to insert cache barrier instructions before every memory
14998 access that might be speculatively executed and that might have side
14999 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15000 controls GCC's implementation of this workaround. It assumes that
15001 aborted accesses to any byte in the following regions will not have
15006 the memory occupied by the current function's stack frame;
15009 the memory occupied by an incoming stack argument;
15012 the memory occupied by an object with a link-time-constant address.
15015 It is the kernel's responsibility to ensure that speculative
15016 accesses to these regions are indeed safe.
15018 If the input program contains a function declaration such as:
15024 then the implementation of @code{foo} must allow @code{j foo} and
15025 @code{jal foo} to be executed speculatively. GCC honors this
15026 restriction for functions it compiles itself. It expects non-GCC
15027 functions (such as hand-written assembly code) to do the same.
15029 The option has three forms:
15032 @item -mr10k-cache-barrier=load-store
15033 Insert a cache barrier before a load or store that might be
15034 speculatively executed and that might have side effects even
15037 @item -mr10k-cache-barrier=store
15038 Insert a cache barrier before a store that might be speculatively
15039 executed and that might have side effects even if aborted.
15041 @item -mr10k-cache-barrier=none
15042 Disable the insertion of cache barriers. This is the default setting.
15045 @item -mflush-func=@var{func}
15046 @itemx -mno-flush-func
15047 @opindex mflush-func
15048 Specifies the function to call to flush the I and D caches, or to not
15049 call any such function. If called, the function must take the same
15050 arguments as the common @code{_flush_func()}, that is, the address of the
15051 memory range for which the cache is being flushed, the size of the
15052 memory range, and the number 3 (to flush both caches). The default
15053 depends on the target GCC was configured for, but commonly is either
15054 @samp{_flush_func} or @samp{__cpu_flush}.
15056 @item mbranch-cost=@var{num}
15057 @opindex mbranch-cost
15058 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15059 This cost is only a heuristic and is not guaranteed to produce
15060 consistent results across releases. A zero cost redundantly selects
15061 the default, which is based on the @option{-mtune} setting.
15063 @item -mbranch-likely
15064 @itemx -mno-branch-likely
15065 @opindex mbranch-likely
15066 @opindex mno-branch-likely
15067 Enable or disable use of Branch Likely instructions, regardless of the
15068 default for the selected architecture. By default, Branch Likely
15069 instructions may be generated if they are supported by the selected
15070 architecture. An exception is for the MIPS32 and MIPS64 architectures
15071 and processors which implement those architectures; for those, Branch
15072 Likely instructions will not be generated by default because the MIPS32
15073 and MIPS64 architectures specifically deprecate their use.
15075 @item -mfp-exceptions
15076 @itemx -mno-fp-exceptions
15077 @opindex mfp-exceptions
15078 Specifies whether FP exceptions are enabled. This affects how we schedule
15079 FP instructions for some processors. The default is that FP exceptions are
15082 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15083 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15086 @item -mvr4130-align
15087 @itemx -mno-vr4130-align
15088 @opindex mvr4130-align
15089 The VR4130 pipeline is two-way superscalar, but can only issue two
15090 instructions together if the first one is 8-byte aligned. When this
15091 option is enabled, GCC will align pairs of instructions that it
15092 thinks should execute in parallel.
15094 This option only has an effect when optimizing for the VR4130.
15095 It normally makes code faster, but at the expense of making it bigger.
15096 It is enabled by default at optimization level @option{-O3}.
15101 Enable (disable) generation of @code{synci} instructions on
15102 architectures that support it. The @code{synci} instructions (if
15103 enabled) will be generated when @code{__builtin___clear_cache()} is
15106 This option defaults to @code{-mno-synci}, but the default can be
15107 overridden by configuring with @code{--with-synci}.
15109 When compiling code for single processor systems, it is generally safe
15110 to use @code{synci}. However, on many multi-core (SMP) systems, it
15111 will not invalidate the instruction caches on all cores and may lead
15112 to undefined behavior.
15114 @item -mrelax-pic-calls
15115 @itemx -mno-relax-pic-calls
15116 @opindex mrelax-pic-calls
15117 Try to turn PIC calls that are normally dispatched via register
15118 @code{$25} into direct calls. This is only possible if the linker can
15119 resolve the destination at link-time and if the destination is within
15120 range for a direct call.
15122 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15123 an assembler and a linker that supports the @code{.reloc} assembly
15124 directive and @code{-mexplicit-relocs} is in effect. With
15125 @code{-mno-explicit-relocs}, this optimization can be performed by the
15126 assembler and the linker alone without help from the compiler.
15128 @item -mmcount-ra-address
15129 @itemx -mno-mcount-ra-address
15130 @opindex mmcount-ra-address
15131 @opindex mno-mcount-ra-address
15132 Emit (do not emit) code that allows @code{_mcount} to modify the
15133 calling function's return address. When enabled, this option extends
15134 the usual @code{_mcount} interface with a new @var{ra-address}
15135 parameter, which has type @code{intptr_t *} and is passed in register
15136 @code{$12}. @code{_mcount} can then modify the return address by
15137 doing both of the following:
15140 Returning the new address in register @code{$31}.
15142 Storing the new address in @code{*@var{ra-address}},
15143 if @var{ra-address} is nonnull.
15146 The default is @option{-mno-mcount-ra-address}.
15151 @subsection MMIX Options
15152 @cindex MMIX Options
15154 These options are defined for the MMIX:
15158 @itemx -mno-libfuncs
15160 @opindex mno-libfuncs
15161 Specify that intrinsic library functions are being compiled, passing all
15162 values in registers, no matter the size.
15165 @itemx -mno-epsilon
15167 @opindex mno-epsilon
15168 Generate floating-point comparison instructions that compare with respect
15169 to the @code{rE} epsilon register.
15171 @item -mabi=mmixware
15173 @opindex mabi=mmixware
15175 Generate code that passes function parameters and return values that (in
15176 the called function) are seen as registers @code{$0} and up, as opposed to
15177 the GNU ABI which uses global registers @code{$231} and up.
15179 @item -mzero-extend
15180 @itemx -mno-zero-extend
15181 @opindex mzero-extend
15182 @opindex mno-zero-extend
15183 When reading data from memory in sizes shorter than 64 bits, use (do not
15184 use) zero-extending load instructions by default, rather than
15185 sign-extending ones.
15188 @itemx -mno-knuthdiv
15190 @opindex mno-knuthdiv
15191 Make the result of a division yielding a remainder have the same sign as
15192 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15193 remainder follows the sign of the dividend. Both methods are
15194 arithmetically valid, the latter being almost exclusively used.
15196 @item -mtoplevel-symbols
15197 @itemx -mno-toplevel-symbols
15198 @opindex mtoplevel-symbols
15199 @opindex mno-toplevel-symbols
15200 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15201 code can be used with the @code{PREFIX} assembly directive.
15205 Generate an executable in the ELF format, rather than the default
15206 @samp{mmo} format used by the @command{mmix} simulator.
15208 @item -mbranch-predict
15209 @itemx -mno-branch-predict
15210 @opindex mbranch-predict
15211 @opindex mno-branch-predict
15212 Use (do not use) the probable-branch instructions, when static branch
15213 prediction indicates a probable branch.
15215 @item -mbase-addresses
15216 @itemx -mno-base-addresses
15217 @opindex mbase-addresses
15218 @opindex mno-base-addresses
15219 Generate (do not generate) code that uses @emph{base addresses}. Using a
15220 base address automatically generates a request (handled by the assembler
15221 and the linker) for a constant to be set up in a global register. The
15222 register is used for one or more base address requests within the range 0
15223 to 255 from the value held in the register. The generally leads to short
15224 and fast code, but the number of different data items that can be
15225 addressed is limited. This means that a program that uses lots of static
15226 data may require @option{-mno-base-addresses}.
15228 @item -msingle-exit
15229 @itemx -mno-single-exit
15230 @opindex msingle-exit
15231 @opindex mno-single-exit
15232 Force (do not force) generated code to have a single exit point in each
15236 @node MN10300 Options
15237 @subsection MN10300 Options
15238 @cindex MN10300 options
15240 These @option{-m} options are defined for Matsushita MN10300 architectures:
15245 Generate code to avoid bugs in the multiply instructions for the MN10300
15246 processors. This is the default.
15248 @item -mno-mult-bug
15249 @opindex mno-mult-bug
15250 Do not generate code to avoid bugs in the multiply instructions for the
15251 MN10300 processors.
15255 Generate code which uses features specific to the AM33 processor.
15259 Do not generate code which uses features specific to the AM33 processor. This
15264 Generate code which uses features specific to the AM33/2.0 processor.
15268 Generate code which uses features specific to the AM34 processor.
15270 @item -mtune=@var{cpu-type}
15272 Use the timing characteristics of the indicated CPU type when
15273 scheduling instructions. This does not change the targeted processor
15274 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15275 @samp{am33-2} or @samp{am34}.
15277 @item -mreturn-pointer-on-d0
15278 @opindex mreturn-pointer-on-d0
15279 When generating a function which returns a pointer, return the pointer
15280 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15281 only in a0, and attempts to call such functions without a prototype
15282 would result in errors. Note that this option is on by default; use
15283 @option{-mno-return-pointer-on-d0} to disable it.
15287 Do not link in the C run-time initialization object file.
15291 Indicate to the linker that it should perform a relaxation optimization pass
15292 to shorten branches, calls and absolute memory addresses. This option only
15293 has an effect when used on the command line for the final link step.
15295 This option makes symbolic debugging impossible.
15299 Allow the compiler to generate @emph{Long Instruction Word}
15300 instructions if the target is the @samp{AM33} or later. This is the
15301 default. This option defines the preprocessor macro @samp{__LIW__}.
15305 Do not allow the compiler to generate @emph{Long Instruction Word}
15306 instructions. This option defines the preprocessor macro
15311 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15312 instructions if the target is the @samp{AM33} or later. This is the
15313 default. This option defines the preprocessor macro @samp{__SETLB__}.
15317 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15318 instructions. This option defines the preprocessor macro
15319 @samp{__NO_SETLB__}.
15323 @node PDP-11 Options
15324 @subsection PDP-11 Options
15325 @cindex PDP-11 Options
15327 These options are defined for the PDP-11:
15332 Use hardware FPP floating point. This is the default. (FIS floating
15333 point on the PDP-11/40 is not supported.)
15336 @opindex msoft-float
15337 Do not use hardware floating point.
15341 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15345 Return floating-point results in memory. This is the default.
15349 Generate code for a PDP-11/40.
15353 Generate code for a PDP-11/45. This is the default.
15357 Generate code for a PDP-11/10.
15359 @item -mbcopy-builtin
15360 @opindex mbcopy-builtin
15361 Use inline @code{movmemhi} patterns for copying memory. This is the
15366 Do not use inline @code{movmemhi} patterns for copying memory.
15372 Use 16-bit @code{int}. This is the default.
15378 Use 32-bit @code{int}.
15381 @itemx -mno-float32
15383 @opindex mno-float32
15384 Use 64-bit @code{float}. This is the default.
15387 @itemx -mno-float64
15389 @opindex mno-float64
15390 Use 32-bit @code{float}.
15394 Use @code{abshi2} pattern. This is the default.
15398 Do not use @code{abshi2} pattern.
15400 @item -mbranch-expensive
15401 @opindex mbranch-expensive
15402 Pretend that branches are expensive. This is for experimenting with
15403 code generation only.
15405 @item -mbranch-cheap
15406 @opindex mbranch-cheap
15407 Do not pretend that branches are expensive. This is the default.
15411 Use Unix assembler syntax. This is the default when configured for
15412 @samp{pdp11-*-bsd}.
15416 Use DEC assembler syntax. This is the default when configured for any
15417 PDP-11 target other than @samp{pdp11-*-bsd}.
15420 @node picoChip Options
15421 @subsection picoChip Options
15422 @cindex picoChip options
15424 These @samp{-m} options are defined for picoChip implementations:
15428 @item -mae=@var{ae_type}
15430 Set the instruction set, register set, and instruction scheduling
15431 parameters for array element type @var{ae_type}. Supported values
15432 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15434 @option{-mae=ANY} selects a completely generic AE type. Code
15435 generated with this option will run on any of the other AE types. The
15436 code will not be as efficient as it would be if compiled for a specific
15437 AE type, and some types of operation (e.g., multiplication) will not
15438 work properly on all types of AE.
15440 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15441 for compiled code, and is the default.
15443 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15444 option may suffer from poor performance of byte (char) manipulation,
15445 since the DSP AE does not provide hardware support for byte load/stores.
15447 @item -msymbol-as-address
15448 Enable the compiler to directly use a symbol name as an address in a
15449 load/store instruction, without first loading it into a
15450 register. Typically, the use of this option will generate larger
15451 programs, which run faster than when the option isn't used. However, the
15452 results vary from program to program, so it is left as a user option,
15453 rather than being permanently enabled.
15455 @item -mno-inefficient-warnings
15456 Disables warnings about the generation of inefficient code. These
15457 warnings can be generated, for example, when compiling code which
15458 performs byte-level memory operations on the MAC AE type. The MAC AE has
15459 no hardware support for byte-level memory operations, so all byte
15460 load/stores must be synthesized from word load/store operations. This is
15461 inefficient and a warning will be generated indicating to the programmer
15462 that they should rewrite the code to avoid byte operations, or to target
15463 an AE type which has the necessary hardware support. This option enables
15464 the warning to be turned off.
15468 @node PowerPC Options
15469 @subsection PowerPC Options
15470 @cindex PowerPC options
15472 These are listed under @xref{RS/6000 and PowerPC Options}.
15474 @node RS/6000 and PowerPC Options
15475 @subsection IBM RS/6000 and PowerPC Options
15476 @cindex RS/6000 and PowerPC Options
15477 @cindex IBM RS/6000 and PowerPC Options
15479 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15486 @itemx -mno-powerpc
15487 @itemx -mpowerpc-gpopt
15488 @itemx -mno-powerpc-gpopt
15489 @itemx -mpowerpc-gfxopt
15490 @itemx -mno-powerpc-gfxopt
15493 @itemx -mno-powerpc64
15497 @itemx -mno-popcntb
15499 @itemx -mno-popcntd
15508 @itemx -mno-hard-dfp
15512 @opindex mno-power2
15514 @opindex mno-powerpc
15515 @opindex mpowerpc-gpopt
15516 @opindex mno-powerpc-gpopt
15517 @opindex mpowerpc-gfxopt
15518 @opindex mno-powerpc-gfxopt
15519 @opindex mpowerpc64
15520 @opindex mno-powerpc64
15524 @opindex mno-popcntb
15526 @opindex mno-popcntd
15532 @opindex mno-mfpgpr
15534 @opindex mno-hard-dfp
15535 GCC supports two related instruction set architectures for the
15536 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15537 instructions supported by the @samp{rios} chip set used in the original
15538 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15539 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15540 the IBM 4xx, 6xx, and follow-on microprocessors.
15542 Neither architecture is a subset of the other. However there is a
15543 large common subset of instructions supported by both. An MQ
15544 register is included in processors supporting the POWER architecture.
15546 You use these options to specify which instructions are available on the
15547 processor you are using. The default value of these options is
15548 determined when configuring GCC@. Specifying the
15549 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15550 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15551 rather than the options listed above.
15553 The @option{-mpower} option allows GCC to generate instructions that
15554 are found only in the POWER architecture and to use the MQ register.
15555 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15556 to generate instructions that are present in the POWER2 architecture but
15557 not the original POWER architecture.
15559 The @option{-mpowerpc} option allows GCC to generate instructions that
15560 are found only in the 32-bit subset of the PowerPC architecture.
15561 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15562 GCC to use the optional PowerPC architecture instructions in the
15563 General Purpose group, including floating-point square root. Specifying
15564 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15565 use the optional PowerPC architecture instructions in the Graphics
15566 group, including floating-point select.
15568 The @option{-mmfcrf} option allows GCC to generate the move from
15569 condition register field instruction implemented on the POWER4
15570 processor and other processors that support the PowerPC V2.01
15572 The @option{-mpopcntb} option allows GCC to generate the popcount and
15573 double precision FP reciprocal estimate instruction implemented on the
15574 POWER5 processor and other processors that support the PowerPC V2.02
15576 The @option{-mpopcntd} option allows GCC to generate the popcount
15577 instruction implemented on the POWER7 processor and other processors
15578 that support the PowerPC V2.06 architecture.
15579 The @option{-mfprnd} option allows GCC to generate the FP round to
15580 integer instructions implemented on the POWER5+ processor and other
15581 processors that support the PowerPC V2.03 architecture.
15582 The @option{-mcmpb} option allows GCC to generate the compare bytes
15583 instruction implemented on the POWER6 processor and other processors
15584 that support the PowerPC V2.05 architecture.
15585 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15586 general purpose register instructions implemented on the POWER6X
15587 processor and other processors that support the extended PowerPC V2.05
15589 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15590 point instructions implemented on some POWER processors.
15592 The @option{-mpowerpc64} option allows GCC to generate the additional
15593 64-bit instructions that are found in the full PowerPC64 architecture
15594 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15595 @option{-mno-powerpc64}.
15597 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15598 will use only the instructions in the common subset of both
15599 architectures plus some special AIX common-mode calls, and will not use
15600 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15601 permits GCC to use any instruction from either architecture and to
15602 allow use of the MQ register; specify this for the Motorola MPC601.
15604 @item -mnew-mnemonics
15605 @itemx -mold-mnemonics
15606 @opindex mnew-mnemonics
15607 @opindex mold-mnemonics
15608 Select which mnemonics to use in the generated assembler code. With
15609 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15610 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15611 assembler mnemonics defined for the POWER architecture. Instructions
15612 defined in only one architecture have only one mnemonic; GCC uses that
15613 mnemonic irrespective of which of these options is specified.
15615 GCC defaults to the mnemonics appropriate for the architecture in
15616 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15617 value of these option. Unless you are building a cross-compiler, you
15618 should normally not specify either @option{-mnew-mnemonics} or
15619 @option{-mold-mnemonics}, but should instead accept the default.
15621 @item -mcpu=@var{cpu_type}
15623 Set architecture type, register usage, choice of mnemonics, and
15624 instruction scheduling parameters for machine type @var{cpu_type}.
15625 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15626 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15627 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15628 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15629 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15630 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15631 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15632 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15633 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15634 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15635 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15637 @option{-mcpu=common} selects a completely generic processor. Code
15638 generated under this option will run on any POWER or PowerPC processor.
15639 GCC will use only the instructions in the common subset of both
15640 architectures, and will not use the MQ register. GCC assumes a generic
15641 processor model for scheduling purposes.
15643 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15644 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15645 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15646 types, with an appropriate, generic processor model assumed for
15647 scheduling purposes.
15649 The other options specify a specific processor. Code generated under
15650 those options will run best on that processor, and may not run at all on
15653 The @option{-mcpu} options automatically enable or disable the
15656 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15657 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15658 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15659 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15661 The particular options set for any particular CPU will vary between
15662 compiler versions, depending on what setting seems to produce optimal
15663 code for that CPU; it doesn't necessarily reflect the actual hardware's
15664 capabilities. If you wish to set an individual option to a particular
15665 value, you may specify it after the @option{-mcpu} option, like
15666 @samp{-mcpu=970 -mno-altivec}.
15668 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15669 not enabled or disabled by the @option{-mcpu} option at present because
15670 AIX does not have full support for these options. You may still
15671 enable or disable them individually if you're sure it'll work in your
15674 @item -mtune=@var{cpu_type}
15676 Set the instruction scheduling parameters for machine type
15677 @var{cpu_type}, but do not set the architecture type, register usage, or
15678 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15679 values for @var{cpu_type} are used for @option{-mtune} as for
15680 @option{-mcpu}. If both are specified, the code generated will use the
15681 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15682 scheduling parameters set by @option{-mtune}.
15684 @item -mcmodel=small
15685 @opindex mcmodel=small
15686 Generate PowerPC64 code for the small model: The TOC is limited to
15689 @item -mcmodel=medium
15690 @opindex mcmodel=medium
15691 Generate PowerPC64 code for the medium model: The TOC and other static
15692 data may be up to a total of 4G in size.
15694 @item -mcmodel=large
15695 @opindex mcmodel=large
15696 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15697 in size. Other data and code is only limited by the 64-bit address
15701 @itemx -mno-altivec
15703 @opindex mno-altivec
15704 Generate code that uses (does not use) AltiVec instructions, and also
15705 enable the use of built-in functions that allow more direct access to
15706 the AltiVec instruction set. You may also need to set
15707 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15713 @opindex mno-vrsave
15714 Generate VRSAVE instructions when generating AltiVec code.
15716 @item -mgen-cell-microcode
15717 @opindex mgen-cell-microcode
15718 Generate Cell microcode instructions
15720 @item -mwarn-cell-microcode
15721 @opindex mwarn-cell-microcode
15722 Warning when a Cell microcode instruction is going to emitted. An example
15723 of a Cell microcode instruction is a variable shift.
15726 @opindex msecure-plt
15727 Generate code that allows ld and ld.so to build executables and shared
15728 libraries with non-exec .plt and .got sections. This is a PowerPC
15729 32-bit SYSV ABI option.
15733 Generate code that uses a BSS .plt section that ld.so fills in, and
15734 requires .plt and .got sections that are both writable and executable.
15735 This is a PowerPC 32-bit SYSV ABI option.
15741 This switch enables or disables the generation of ISEL instructions.
15743 @item -misel=@var{yes/no}
15744 This switch has been deprecated. Use @option{-misel} and
15745 @option{-mno-isel} instead.
15751 This switch enables or disables the generation of SPE simd
15757 @opindex mno-paired
15758 This switch enables or disables the generation of PAIRED simd
15761 @item -mspe=@var{yes/no}
15762 This option has been deprecated. Use @option{-mspe} and
15763 @option{-mno-spe} instead.
15769 Generate code that uses (does not use) vector/scalar (VSX)
15770 instructions, and also enable the use of built-in functions that allow
15771 more direct access to the VSX instruction set.
15773 @item -mfloat-gprs=@var{yes/single/double/no}
15774 @itemx -mfloat-gprs
15775 @opindex mfloat-gprs
15776 This switch enables or disables the generation of floating point
15777 operations on the general purpose registers for architectures that
15780 The argument @var{yes} or @var{single} enables the use of
15781 single-precision floating point operations.
15783 The argument @var{double} enables the use of single and
15784 double-precision floating point operations.
15786 The argument @var{no} disables floating point operations on the
15787 general purpose registers.
15789 This option is currently only available on the MPC854x.
15795 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15796 targets (including GNU/Linux). The 32-bit environment sets int, long
15797 and pointer to 32 bits and generates code that runs on any PowerPC
15798 variant. The 64-bit environment sets int to 32 bits and long and
15799 pointer to 64 bits, and generates code for PowerPC64, as for
15800 @option{-mpowerpc64}.
15803 @itemx -mno-fp-in-toc
15804 @itemx -mno-sum-in-toc
15805 @itemx -mminimal-toc
15807 @opindex mno-fp-in-toc
15808 @opindex mno-sum-in-toc
15809 @opindex mminimal-toc
15810 Modify generation of the TOC (Table Of Contents), which is created for
15811 every executable file. The @option{-mfull-toc} option is selected by
15812 default. In that case, GCC will allocate at least one TOC entry for
15813 each unique non-automatic variable reference in your program. GCC
15814 will also place floating-point constants in the TOC@. However, only
15815 16,384 entries are available in the TOC@.
15817 If you receive a linker error message that saying you have overflowed
15818 the available TOC space, you can reduce the amount of TOC space used
15819 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15820 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15821 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15822 generate code to calculate the sum of an address and a constant at
15823 run-time instead of putting that sum into the TOC@. You may specify one
15824 or both of these options. Each causes GCC to produce very slightly
15825 slower and larger code at the expense of conserving TOC space.
15827 If you still run out of space in the TOC even when you specify both of
15828 these options, specify @option{-mminimal-toc} instead. This option causes
15829 GCC to make only one TOC entry for every file. When you specify this
15830 option, GCC will produce code that is slower and larger but which
15831 uses extremely little TOC space. You may wish to use this option
15832 only on files that contain less frequently executed code.
15838 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15839 @code{long} type, and the infrastructure needed to support them.
15840 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15841 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15842 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15845 @itemx -mno-xl-compat
15846 @opindex mxl-compat
15847 @opindex mno-xl-compat
15848 Produce code that conforms more closely to IBM XL compiler semantics
15849 when using AIX-compatible ABI@. Pass floating-point arguments to
15850 prototyped functions beyond the register save area (RSA) on the stack
15851 in addition to argument FPRs. Do not assume that most significant
15852 double in 128-bit long double value is properly rounded when comparing
15853 values and converting to double. Use XL symbol names for long double
15856 The AIX calling convention was extended but not initially documented to
15857 handle an obscure K&R C case of calling a function that takes the
15858 address of its arguments with fewer arguments than declared. IBM XL
15859 compilers access floating point arguments which do not fit in the
15860 RSA from the stack when a subroutine is compiled without
15861 optimization. Because always storing floating-point arguments on the
15862 stack is inefficient and rarely needed, this option is not enabled by
15863 default and only is necessary when calling subroutines compiled by IBM
15864 XL compilers without optimization.
15868 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15869 application written to use message passing with special startup code to
15870 enable the application to run. The system must have PE installed in the
15871 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15872 must be overridden with the @option{-specs=} option to specify the
15873 appropriate directory location. The Parallel Environment does not
15874 support threads, so the @option{-mpe} option and the @option{-pthread}
15875 option are incompatible.
15877 @item -malign-natural
15878 @itemx -malign-power
15879 @opindex malign-natural
15880 @opindex malign-power
15881 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15882 @option{-malign-natural} overrides the ABI-defined alignment of larger
15883 types, such as floating-point doubles, on their natural size-based boundary.
15884 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15885 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15887 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15891 @itemx -mhard-float
15892 @opindex msoft-float
15893 @opindex mhard-float
15894 Generate code that does not use (uses) the floating-point register set.
15895 Software floating point emulation is provided if you use the
15896 @option{-msoft-float} option, and pass the option to GCC when linking.
15898 @item -msingle-float
15899 @itemx -mdouble-float
15900 @opindex msingle-float
15901 @opindex mdouble-float
15902 Generate code for single or double-precision floating point operations.
15903 @option{-mdouble-float} implies @option{-msingle-float}.
15906 @opindex msimple-fpu
15907 Do not generate sqrt and div instructions for hardware floating point unit.
15911 Specify type of floating point unit. Valid values are @var{sp_lite}
15912 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15913 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15914 and @var{dp_full} (equivalent to -mdouble-float).
15917 @opindex mxilinx-fpu
15918 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15921 @itemx -mno-multiple
15923 @opindex mno-multiple
15924 Generate code that uses (does not use) the load multiple word
15925 instructions and the store multiple word instructions. These
15926 instructions are generated by default on POWER systems, and not
15927 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15928 endian PowerPC systems, since those instructions do not work when the
15929 processor is in little endian mode. The exceptions are PPC740 and
15930 PPC750 which permit the instructions usage in little endian mode.
15935 @opindex mno-string
15936 Generate code that uses (does not use) the load string instructions
15937 and the store string word instructions to save multiple registers and
15938 do small block moves. These instructions are generated by default on
15939 POWER systems, and not generated on PowerPC systems. Do not use
15940 @option{-mstring} on little endian PowerPC systems, since those
15941 instructions do not work when the processor is in little endian mode.
15942 The exceptions are PPC740 and PPC750 which permit the instructions
15943 usage in little endian mode.
15948 @opindex mno-update
15949 Generate code that uses (does not use) the load or store instructions
15950 that update the base register to the address of the calculated memory
15951 location. These instructions are generated by default. If you use
15952 @option{-mno-update}, there is a small window between the time that the
15953 stack pointer is updated and the address of the previous frame is
15954 stored, which means code that walks the stack frame across interrupts or
15955 signals may get corrupted data.
15957 @item -mavoid-indexed-addresses
15958 @itemx -mno-avoid-indexed-addresses
15959 @opindex mavoid-indexed-addresses
15960 @opindex mno-avoid-indexed-addresses
15961 Generate code that tries to avoid (not avoid) the use of indexed load
15962 or store instructions. These instructions can incur a performance
15963 penalty on Power6 processors in certain situations, such as when
15964 stepping through large arrays that cross a 16M boundary. This option
15965 is enabled by default when targetting Power6 and disabled otherwise.
15968 @itemx -mno-fused-madd
15969 @opindex mfused-madd
15970 @opindex mno-fused-madd
15971 Generate code that uses (does not use) the floating point multiply and
15972 accumulate instructions. These instructions are generated by default
15973 if hardware floating point is used. The machine dependent
15974 @option{-mfused-madd} option is now mapped to the machine independent
15975 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15976 mapped to @option{-ffp-contract=off}.
15982 Generate code that uses (does not use) the half-word multiply and
15983 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15984 These instructions are generated by default when targetting those
15991 Generate code that uses (does not use) the string-search @samp{dlmzb}
15992 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15993 generated by default when targetting those processors.
15995 @item -mno-bit-align
15997 @opindex mno-bit-align
15998 @opindex mbit-align
15999 On System V.4 and embedded PowerPC systems do not (do) force structures
16000 and unions that contain bit-fields to be aligned to the base type of the
16003 For example, by default a structure containing nothing but 8
16004 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
16005 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
16006 the structure would be aligned to a 1 byte boundary and be one byte in
16009 @item -mno-strict-align
16010 @itemx -mstrict-align
16011 @opindex mno-strict-align
16012 @opindex mstrict-align
16013 On System V.4 and embedded PowerPC systems do not (do) assume that
16014 unaligned memory references will be handled by the system.
16016 @item -mrelocatable
16017 @itemx -mno-relocatable
16018 @opindex mrelocatable
16019 @opindex mno-relocatable
16020 Generate code that allows (does not allow) a static executable to be
16021 relocated to a different address at runtime. A simple embedded
16022 PowerPC system loader should relocate the entire contents of
16023 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16024 a table of 32-bit addresses generated by this option. For this to
16025 work, all objects linked together must be compiled with
16026 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16027 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
16029 @item -mrelocatable-lib
16030 @itemx -mno-relocatable-lib
16031 @opindex mrelocatable-lib
16032 @opindex mno-relocatable-lib
16033 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16034 @code{.fixup} section to allow static executables to be relocated at
16035 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
16036 alignment of @option{-mrelocatable}. Objects compiled with
16037 @option{-mrelocatable-lib} may be linked with objects compiled with
16038 any combination of the @option{-mrelocatable} options.
16044 On System V.4 and embedded PowerPC systems do not (do) assume that
16045 register 2 contains a pointer to a global area pointing to the addresses
16046 used in the program.
16049 @itemx -mlittle-endian
16051 @opindex mlittle-endian
16052 On System V.4 and embedded PowerPC systems compile code for the
16053 processor in little endian mode. The @option{-mlittle-endian} option is
16054 the same as @option{-mlittle}.
16057 @itemx -mbig-endian
16059 @opindex mbig-endian
16060 On System V.4 and embedded PowerPC systems compile code for the
16061 processor in big endian mode. The @option{-mbig-endian} option is
16062 the same as @option{-mbig}.
16064 @item -mdynamic-no-pic
16065 @opindex mdynamic-no-pic
16066 On Darwin and Mac OS X systems, compile code so that it is not
16067 relocatable, but that its external references are relocatable. The
16068 resulting code is suitable for applications, but not shared
16071 @item -msingle-pic-base
16072 @opindex msingle-pic-base
16073 Treat the register used for PIC addressing as read-only, rather than
16074 loading it in the prologue for each function. The run-time system is
16075 responsible for initializing this register with an appropriate value
16076 before execution begins.
16078 @item -mprioritize-restricted-insns=@var{priority}
16079 @opindex mprioritize-restricted-insns
16080 This option controls the priority that is assigned to
16081 dispatch-slot restricted instructions during the second scheduling
16082 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16083 @var{no/highest/second-highest} priority to dispatch slot restricted
16086 @item -msched-costly-dep=@var{dependence_type}
16087 @opindex msched-costly-dep
16088 This option controls which dependences are considered costly
16089 by the target during instruction scheduling. The argument
16090 @var{dependence_type} takes one of the following values:
16091 @var{no}: no dependence is costly,
16092 @var{all}: all dependences are costly,
16093 @var{true_store_to_load}: a true dependence from store to load is costly,
16094 @var{store_to_load}: any dependence from store to load is costly,
16095 @var{number}: any dependence which latency >= @var{number} is costly.
16097 @item -minsert-sched-nops=@var{scheme}
16098 @opindex minsert-sched-nops
16099 This option controls which nop insertion scheme will be used during
16100 the second scheduling pass. The argument @var{scheme} takes one of the
16102 @var{no}: Don't insert nops.
16103 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16104 according to the scheduler's grouping.
16105 @var{regroup_exact}: Insert nops to force costly dependent insns into
16106 separate groups. Insert exactly as many nops as needed to force an insn
16107 to a new group, according to the estimated processor grouping.
16108 @var{number}: Insert nops to force costly dependent insns into
16109 separate groups. Insert @var{number} nops to force an insn to a new group.
16112 @opindex mcall-sysv
16113 On System V.4 and embedded PowerPC systems compile code using calling
16114 conventions that adheres to the March 1995 draft of the System V
16115 Application Binary Interface, PowerPC processor supplement. This is the
16116 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16118 @item -mcall-sysv-eabi
16120 @opindex mcall-sysv-eabi
16121 @opindex mcall-eabi
16122 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16124 @item -mcall-sysv-noeabi
16125 @opindex mcall-sysv-noeabi
16126 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16128 @item -mcall-aixdesc
16130 On System V.4 and embedded PowerPC systems compile code for the AIX
16134 @opindex mcall-linux
16135 On System V.4 and embedded PowerPC systems compile code for the
16136 Linux-based GNU system.
16138 @item -mcall-freebsd
16139 @opindex mcall-freebsd
16140 On System V.4 and embedded PowerPC systems compile code for the
16141 FreeBSD operating system.
16143 @item -mcall-netbsd
16144 @opindex mcall-netbsd
16145 On System V.4 and embedded PowerPC systems compile code for the
16146 NetBSD operating system.
16148 @item -mcall-openbsd
16149 @opindex mcall-netbsd
16150 On System V.4 and embedded PowerPC systems compile code for the
16151 OpenBSD operating system.
16153 @item -maix-struct-return
16154 @opindex maix-struct-return
16155 Return all structures in memory (as specified by the AIX ABI)@.
16157 @item -msvr4-struct-return
16158 @opindex msvr4-struct-return
16159 Return structures smaller than 8 bytes in registers (as specified by the
16162 @item -mabi=@var{abi-type}
16164 Extend the current ABI with a particular extension, or remove such extension.
16165 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16166 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16170 Extend the current ABI with SPE ABI extensions. This does not change
16171 the default ABI, instead it adds the SPE ABI extensions to the current
16175 @opindex mabi=no-spe
16176 Disable Booke SPE ABI extensions for the current ABI@.
16178 @item -mabi=ibmlongdouble
16179 @opindex mabi=ibmlongdouble
16180 Change the current ABI to use IBM extended precision long double.
16181 This is a PowerPC 32-bit SYSV ABI option.
16183 @item -mabi=ieeelongdouble
16184 @opindex mabi=ieeelongdouble
16185 Change the current ABI to use IEEE extended precision long double.
16186 This is a PowerPC 32-bit Linux ABI option.
16189 @itemx -mno-prototype
16190 @opindex mprototype
16191 @opindex mno-prototype
16192 On System V.4 and embedded PowerPC systems assume that all calls to
16193 variable argument functions are properly prototyped. Otherwise, the
16194 compiler must insert an instruction before every non prototyped call to
16195 set or clear bit 6 of the condition code register (@var{CR}) to
16196 indicate whether floating point values were passed in the floating point
16197 registers in case the function takes a variable arguments. With
16198 @option{-mprototype}, only calls to prototyped variable argument functions
16199 will set or clear the bit.
16203 On embedded PowerPC systems, assume that the startup module is called
16204 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16205 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16210 On embedded PowerPC systems, assume that the startup module is called
16211 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16216 On embedded PowerPC systems, assume that the startup module is called
16217 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16220 @item -myellowknife
16221 @opindex myellowknife
16222 On embedded PowerPC systems, assume that the startup module is called
16223 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16228 On System V.4 and embedded PowerPC systems, specify that you are
16229 compiling for a VxWorks system.
16233 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16234 header to indicate that @samp{eabi} extended relocations are used.
16240 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16241 Embedded Applications Binary Interface (eabi) which is a set of
16242 modifications to the System V.4 specifications. Selecting @option{-meabi}
16243 means that the stack is aligned to an 8 byte boundary, a function
16244 @code{__eabi} is called to from @code{main} to set up the eabi
16245 environment, and the @option{-msdata} option can use both @code{r2} and
16246 @code{r13} to point to two separate small data areas. Selecting
16247 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16248 do not call an initialization function from @code{main}, and the
16249 @option{-msdata} option will only use @code{r13} to point to a single
16250 small data area. The @option{-meabi} option is on by default if you
16251 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16254 @opindex msdata=eabi
16255 On System V.4 and embedded PowerPC systems, put small initialized
16256 @code{const} global and static data in the @samp{.sdata2} section, which
16257 is pointed to by register @code{r2}. Put small initialized
16258 non-@code{const} global and static data in the @samp{.sdata} section,
16259 which is pointed to by register @code{r13}. Put small uninitialized
16260 global and static data in the @samp{.sbss} section, which is adjacent to
16261 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16262 incompatible with the @option{-mrelocatable} option. The
16263 @option{-msdata=eabi} option also sets the @option{-memb} option.
16266 @opindex msdata=sysv
16267 On System V.4 and embedded PowerPC systems, put small global and static
16268 data in the @samp{.sdata} section, which is pointed to by register
16269 @code{r13}. Put small uninitialized global and static data in the
16270 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16271 The @option{-msdata=sysv} option is incompatible with the
16272 @option{-mrelocatable} option.
16274 @item -msdata=default
16276 @opindex msdata=default
16278 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16279 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16280 same as @option{-msdata=sysv}.
16283 @opindex msdata=data
16284 On System V.4 and embedded PowerPC systems, put small global
16285 data in the @samp{.sdata} section. Put small uninitialized global
16286 data in the @samp{.sbss} section. Do not use register @code{r13}
16287 to address small data however. This is the default behavior unless
16288 other @option{-msdata} options are used.
16292 @opindex msdata=none
16294 On embedded PowerPC systems, put all initialized global and static data
16295 in the @samp{.data} section, and all uninitialized data in the
16296 @samp{.bss} section.
16298 @item -mblock-move-inline-limit=@var{num}
16299 @opindex mblock-move-inline-limit
16300 Inline all block moves (such as calls to @code{memcpy} or structure
16301 copies) less than or equal to @var{num} bytes. The minimum value for
16302 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16303 targets. The default value is target-specific.
16307 @cindex smaller data references (PowerPC)
16308 @cindex .sdata/.sdata2 references (PowerPC)
16309 On embedded PowerPC systems, put global and static items less than or
16310 equal to @var{num} bytes into the small data or bss sections instead of
16311 the normal data or bss section. By default, @var{num} is 8. The
16312 @option{-G @var{num}} switch is also passed to the linker.
16313 All modules should be compiled with the same @option{-G @var{num}} value.
16316 @itemx -mno-regnames
16318 @opindex mno-regnames
16319 On System V.4 and embedded PowerPC systems do (do not) emit register
16320 names in the assembly language output using symbolic forms.
16323 @itemx -mno-longcall
16325 @opindex mno-longcall
16326 By default assume that all calls are far away so that a longer more
16327 expensive calling sequence is required. This is required for calls
16328 further than 32 megabytes (33,554,432 bytes) from the current location.
16329 A short call will be generated if the compiler knows
16330 the call cannot be that far away. This setting can be overridden by
16331 the @code{shortcall} function attribute, or by @code{#pragma
16334 Some linkers are capable of detecting out-of-range calls and generating
16335 glue code on the fly. On these systems, long calls are unnecessary and
16336 generate slower code. As of this writing, the AIX linker can do this,
16337 as can the GNU linker for PowerPC/64. It is planned to add this feature
16338 to the GNU linker for 32-bit PowerPC systems as well.
16340 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16341 callee, L42'', plus a ``branch island'' (glue code). The two target
16342 addresses represent the callee and the ``branch island''. The
16343 Darwin/PPC linker will prefer the first address and generate a ``bl
16344 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16345 otherwise, the linker will generate ``bl L42'' to call the ``branch
16346 island''. The ``branch island'' is appended to the body of the
16347 calling function; it computes the full 32-bit address of the callee
16350 On Mach-O (Darwin) systems, this option directs the compiler emit to
16351 the glue for every direct call, and the Darwin linker decides whether
16352 to use or discard it.
16354 In the future, we may cause GCC to ignore all longcall specifications
16355 when the linker is known to generate glue.
16357 @item -mtls-markers
16358 @itemx -mno-tls-markers
16359 @opindex mtls-markers
16360 @opindex mno-tls-markers
16361 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16362 specifying the function argument. The relocation allows ld to
16363 reliably associate function call with argument setup instructions for
16364 TLS optimization, which in turn allows gcc to better schedule the
16369 Adds support for multithreading with the @dfn{pthreads} library.
16370 This option sets flags for both the preprocessor and linker.
16375 This option will enable GCC to use the reciprocal estimate and
16376 reciprocal square root estimate instructions with additional
16377 Newton-Raphson steps to increase precision instead of doing a divide or
16378 square root and divide for floating point arguments. You should use
16379 the @option{-ffast-math} option when using @option{-mrecip} (or at
16380 least @option{-funsafe-math-optimizations},
16381 @option{-finite-math-only}, @option{-freciprocal-math} and
16382 @option{-fno-trapping-math}). Note that while the throughput of the
16383 sequence is generally higher than the throughput of the non-reciprocal
16384 instruction, the precision of the sequence can be decreased by up to 2
16385 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16388 @item -mrecip=@var{opt}
16389 @opindex mrecip=opt
16390 This option allows to control which reciprocal estimate instructions
16391 may be used. @var{opt} is a comma separated list of options, that may
16392 be preceded by a @code{!} to invert the option:
16393 @code{all}: enable all estimate instructions,
16394 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16395 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16396 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16397 @code{divf}: enable the single precision reciprocal approximation instructions;
16398 @code{divd}: enable the double precision reciprocal approximation instructions;
16399 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16400 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16401 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16403 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16404 all of the reciprocal estimate instructions, except for the
16405 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16406 which handle the double precision reciprocal square root calculations.
16408 @item -mrecip-precision
16409 @itemx -mno-recip-precision
16410 @opindex mrecip-precision
16411 Assume (do not assume) that the reciprocal estimate instructions
16412 provide higher precision estimates than is mandated by the powerpc
16413 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16414 automatically selects @option{-mrecip-precision}. The double
16415 precision square root estimate instructions are not generated by
16416 default on low precision machines, since they do not provide an
16417 estimate that converges after three steps.
16419 @item -mveclibabi=@var{type}
16420 @opindex mveclibabi
16421 Specifies the ABI type to use for vectorizing intrinsics using an
16422 external library. The only type supported at present is @code{mass},
16423 which specifies to use IBM's Mathematical Acceleration Subsystem
16424 (MASS) libraries for vectorizing intrinsics using external libraries.
16425 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16426 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16427 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16428 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16429 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16430 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16431 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16432 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16433 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16434 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16435 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16436 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16437 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16438 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16439 for power7. Both @option{-ftree-vectorize} and
16440 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16441 libraries will have to be specified at link time.
16446 Generate (do not generate) the @code{friz} instruction when the
16447 @option{-funsafe-math-optimizations} option is used to optimize
16448 rounding a floating point value to 64-bit integer and back to floating
16449 point. The @code{friz} instruction does not return the same value if
16450 the floating point number is too large to fit in an integer.
16452 @item -mpointers-to-nested-functions
16453 @itemx -mno-pointers-to-nested-functions
16454 @opindex mpointers-to-nested-functions
16455 Generate (do not generate) code to load up the static chain register
16456 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16457 systems where a function pointer points to a 3 word descriptor giving
16458 the function address, TOC value to be loaded in register @var{r2}, and
16459 static chain value to be loaded in register @var{r11}. The
16460 @option{-mpointers-to-nested-functions} is on by default. You will
16461 not be able to call through pointers to nested functions or pointers
16462 to functions compiled in other languages that use the static chain if
16463 you use the @option{-mno-pointers-to-nested-functions}.
16465 @item -msave-toc-indirect
16466 @itemx -mno-save-toc-indirect
16467 @opindex msave-toc-indirect
16468 Generate (do not generate) code to save the TOC value in the reserved
16469 stack location in the function prologue if the function calls through
16470 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
16471 saved in the prologue, it is saved just before the call through the
16472 pointer. The @option{-mno-save-toc-indirect} option is the default.
16476 @subsection RX Options
16479 These command line options are defined for RX targets:
16482 @item -m64bit-doubles
16483 @itemx -m32bit-doubles
16484 @opindex m64bit-doubles
16485 @opindex m32bit-doubles
16486 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16487 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16488 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16489 works on 32-bit values, which is why the default is
16490 @option{-m32bit-doubles}.
16496 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16497 floating point hardware. The default is enabled for the @var{RX600}
16498 series and disabled for the @var{RX200} series.
16500 Floating point instructions will only be generated for 32-bit floating
16501 point values however, so if the @option{-m64bit-doubles} option is in
16502 use then the FPU hardware will not be used for doubles.
16504 @emph{Note} If the @option{-fpu} option is enabled then
16505 @option{-funsafe-math-optimizations} is also enabled automatically.
16506 This is because the RX FPU instructions are themselves unsafe.
16508 @item -mcpu=@var{name}
16510 Selects the type of RX CPU to be targeted. Currently three types are
16511 supported, the generic @var{RX600} and @var{RX200} series hardware and
16512 the specific @var{RX610} CPU. The default is @var{RX600}.
16514 The only difference between @var{RX600} and @var{RX610} is that the
16515 @var{RX610} does not support the @code{MVTIPL} instruction.
16517 The @var{RX200} series does not have a hardware floating point unit
16518 and so @option{-nofpu} is enabled by default when this type is
16521 @item -mbig-endian-data
16522 @itemx -mlittle-endian-data
16523 @opindex mbig-endian-data
16524 @opindex mlittle-endian-data
16525 Store data (but not code) in the big-endian format. The default is
16526 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16529 @item -msmall-data-limit=@var{N}
16530 @opindex msmall-data-limit
16531 Specifies the maximum size in bytes of global and static variables
16532 which can be placed into the small data area. Using the small data
16533 area can lead to smaller and faster code, but the size of area is
16534 limited and it is up to the programmer to ensure that the area does
16535 not overflow. Also when the small data area is used one of the RX's
16536 registers (@code{r13}) is reserved for use pointing to this area, so
16537 it is no longer available for use by the compiler. This could result
16538 in slower and/or larger code if variables which once could have been
16539 held in @code{r13} are now pushed onto the stack.
16541 Note, common variables (variables which have not been initialised) and
16542 constants are not placed into the small data area as they are assigned
16543 to other sections in the output executable.
16545 The default value is zero, which disables this feature. Note, this
16546 feature is not enabled by default with higher optimization levels
16547 (@option{-O2} etc) because of the potentially detrimental effects of
16548 reserving register @code{r13}. It is up to the programmer to
16549 experiment and discover whether this feature is of benefit to their
16556 Use the simulator runtime. The default is to use the libgloss board
16559 @item -mas100-syntax
16560 @itemx -mno-as100-syntax
16561 @opindex mas100-syntax
16562 @opindex mno-as100-syntax
16563 When generating assembler output use a syntax that is compatible with
16564 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16565 assembler but it has some restrictions so generating it is not the
16568 @item -mmax-constant-size=@var{N}
16569 @opindex mmax-constant-size
16570 Specifies the maximum size, in bytes, of a constant that can be used as
16571 an operand in a RX instruction. Although the RX instruction set does
16572 allow constants of up to 4 bytes in length to be used in instructions,
16573 a longer value equates to a longer instruction. Thus in some
16574 circumstances it can be beneficial to restrict the size of constants
16575 that are used in instructions. Constants that are too big are instead
16576 placed into a constant pool and referenced via register indirection.
16578 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16579 or 4 means that constants of any size are allowed.
16583 Enable linker relaxation. Linker relaxation is a process whereby the
16584 linker will attempt to reduce the size of a program by finding shorter
16585 versions of various instructions. Disabled by default.
16587 @item -mint-register=@var{N}
16588 @opindex mint-register
16589 Specify the number of registers to reserve for fast interrupt handler
16590 functions. The value @var{N} can be between 0 and 4. A value of 1
16591 means that register @code{r13} will be reserved for the exclusive use
16592 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16593 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16594 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16595 A value of 0, the default, does not reserve any registers.
16597 @item -msave-acc-in-interrupts
16598 @opindex msave-acc-in-interrupts
16599 Specifies that interrupt handler functions should preserve the
16600 accumulator register. This is only necessary if normal code might use
16601 the accumulator register, for example because it performs 64-bit
16602 multiplications. The default is to ignore the accumulator as this
16603 makes the interrupt handlers faster.
16607 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16608 has special significance to the RX port when used with the
16609 @code{interrupt} function attribute. This attribute indicates a
16610 function intended to process fast interrupts. GCC will will ensure
16611 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16612 and/or @code{r13} and only provided that the normal use of the
16613 corresponding registers have been restricted via the
16614 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16617 @node S/390 and zSeries Options
16618 @subsection S/390 and zSeries Options
16619 @cindex S/390 and zSeries Options
16621 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16625 @itemx -msoft-float
16626 @opindex mhard-float
16627 @opindex msoft-float
16628 Use (do not use) the hardware floating-point instructions and registers
16629 for floating-point operations. When @option{-msoft-float} is specified,
16630 functions in @file{libgcc.a} will be used to perform floating-point
16631 operations. When @option{-mhard-float} is specified, the compiler
16632 generates IEEE floating-point instructions. This is the default.
16635 @itemx -mno-hard-dfp
16637 @opindex mno-hard-dfp
16638 Use (do not use) the hardware decimal-floating-point instructions for
16639 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16640 specified, functions in @file{libgcc.a} will be used to perform
16641 decimal-floating-point operations. When @option{-mhard-dfp} is
16642 specified, the compiler generates decimal-floating-point hardware
16643 instructions. This is the default for @option{-march=z9-ec} or higher.
16645 @item -mlong-double-64
16646 @itemx -mlong-double-128
16647 @opindex mlong-double-64
16648 @opindex mlong-double-128
16649 These switches control the size of @code{long double} type. A size
16650 of 64bit makes the @code{long double} type equivalent to the @code{double}
16651 type. This is the default.
16654 @itemx -mno-backchain
16655 @opindex mbackchain
16656 @opindex mno-backchain
16657 Store (do not store) the address of the caller's frame as backchain pointer
16658 into the callee's stack frame.
16659 A backchain may be needed to allow debugging using tools that do not understand
16660 DWARF-2 call frame information.
16661 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16662 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16663 the backchain is placed into the topmost word of the 96/160 byte register
16666 In general, code compiled with @option{-mbackchain} is call-compatible with
16667 code compiled with @option{-mmo-backchain}; however, use of the backchain
16668 for debugging purposes usually requires that the whole binary is built with
16669 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16670 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16671 to build a linux kernel use @option{-msoft-float}.
16673 The default is to not maintain the backchain.
16675 @item -mpacked-stack
16676 @itemx -mno-packed-stack
16677 @opindex mpacked-stack
16678 @opindex mno-packed-stack
16679 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16680 specified, the compiler uses the all fields of the 96/160 byte register save
16681 area only for their default purpose; unused fields still take up stack space.
16682 When @option{-mpacked-stack} is specified, register save slots are densely
16683 packed at the top of the register save area; unused space is reused for other
16684 purposes, allowing for more efficient use of the available stack space.
16685 However, when @option{-mbackchain} is also in effect, the topmost word of
16686 the save area is always used to store the backchain, and the return address
16687 register is always saved two words below the backchain.
16689 As long as the stack frame backchain is not used, code generated with
16690 @option{-mpacked-stack} is call-compatible with code generated with
16691 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16692 S/390 or zSeries generated code that uses the stack frame backchain at run
16693 time, not just for debugging purposes. Such code is not call-compatible
16694 with code compiled with @option{-mpacked-stack}. Also, note that the
16695 combination of @option{-mbackchain},
16696 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16697 to build a linux kernel use @option{-msoft-float}.
16699 The default is to not use the packed stack layout.
16702 @itemx -mno-small-exec
16703 @opindex msmall-exec
16704 @opindex mno-small-exec
16705 Generate (or do not generate) code using the @code{bras} instruction
16706 to do subroutine calls.
16707 This only works reliably if the total executable size does not
16708 exceed 64k. The default is to use the @code{basr} instruction instead,
16709 which does not have this limitation.
16715 When @option{-m31} is specified, generate code compliant to the
16716 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16717 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16718 particular to generate 64-bit instructions. For the @samp{s390}
16719 targets, the default is @option{-m31}, while the @samp{s390x}
16720 targets default to @option{-m64}.
16726 When @option{-mzarch} is specified, generate code using the
16727 instructions available on z/Architecture.
16728 When @option{-mesa} is specified, generate code using the
16729 instructions available on ESA/390. Note that @option{-mesa} is
16730 not possible with @option{-m64}.
16731 When generating code compliant to the GNU/Linux for S/390 ABI,
16732 the default is @option{-mesa}. When generating code compliant
16733 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16739 Generate (or do not generate) code using the @code{mvcle} instruction
16740 to perform block moves. When @option{-mno-mvcle} is specified,
16741 use a @code{mvc} loop instead. This is the default unless optimizing for
16748 Print (or do not print) additional debug information when compiling.
16749 The default is to not print debug information.
16751 @item -march=@var{cpu-type}
16753 Generate code that will run on @var{cpu-type}, which is the name of a system
16754 representing a certain processor type. Possible values for
16755 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16756 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16757 When generating code using the instructions available on z/Architecture,
16758 the default is @option{-march=z900}. Otherwise, the default is
16759 @option{-march=g5}.
16761 @item -mtune=@var{cpu-type}
16763 Tune to @var{cpu-type} everything applicable about the generated code,
16764 except for the ABI and the set of available instructions.
16765 The list of @var{cpu-type} values is the same as for @option{-march}.
16766 The default is the value used for @option{-march}.
16769 @itemx -mno-tpf-trace
16770 @opindex mtpf-trace
16771 @opindex mno-tpf-trace
16772 Generate code that adds (does not add) in TPF OS specific branches to trace
16773 routines in the operating system. This option is off by default, even
16774 when compiling for the TPF OS@.
16777 @itemx -mno-fused-madd
16778 @opindex mfused-madd
16779 @opindex mno-fused-madd
16780 Generate code that uses (does not use) the floating point multiply and
16781 accumulate instructions. These instructions are generated by default if
16782 hardware floating point is used.
16784 @item -mwarn-framesize=@var{framesize}
16785 @opindex mwarn-framesize
16786 Emit a warning if the current function exceeds the given frame size. Because
16787 this is a compile time check it doesn't need to be a real problem when the program
16788 runs. It is intended to identify functions which most probably cause
16789 a stack overflow. It is useful to be used in an environment with limited stack
16790 size e.g.@: the linux kernel.
16792 @item -mwarn-dynamicstack
16793 @opindex mwarn-dynamicstack
16794 Emit a warning if the function calls alloca or uses dynamically
16795 sized arrays. This is generally a bad idea with a limited stack size.
16797 @item -mstack-guard=@var{stack-guard}
16798 @itemx -mstack-size=@var{stack-size}
16799 @opindex mstack-guard
16800 @opindex mstack-size
16801 If these options are provided the s390 back end emits additional instructions in
16802 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16803 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16804 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16805 the frame size of the compiled function is chosen.
16806 These options are intended to be used to help debugging stack overflow problems.
16807 The additionally emitted code causes only little overhead and hence can also be
16808 used in production like systems without greater performance degradation. The given
16809 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16810 @var{stack-guard} without exceeding 64k.
16811 In order to be efficient the extra code makes the assumption that the stack starts
16812 at an address aligned to the value given by @var{stack-size}.
16813 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16816 @node Score Options
16817 @subsection Score Options
16818 @cindex Score Options
16820 These options are defined for Score implementations:
16825 Compile code for big endian mode. This is the default.
16829 Compile code for little endian mode.
16833 Disable generate bcnz instruction.
16837 Enable generate unaligned load and store instruction.
16841 Enable the use of multiply-accumulate instructions. Disabled by default.
16845 Specify the SCORE5 as the target architecture.
16849 Specify the SCORE5U of the target architecture.
16853 Specify the SCORE7 as the target architecture. This is the default.
16857 Specify the SCORE7D as the target architecture.
16861 @subsection SH Options
16863 These @samp{-m} options are defined for the SH implementations:
16868 Generate code for the SH1.
16872 Generate code for the SH2.
16875 Generate code for the SH2e.
16879 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16880 that the floating-point unit is not used.
16882 @item -m2a-single-only
16883 @opindex m2a-single-only
16884 Generate code for the SH2a-FPU, in such a way that no double-precision
16885 floating point operations are used.
16888 @opindex m2a-single
16889 Generate code for the SH2a-FPU assuming the floating-point unit is in
16890 single-precision mode by default.
16894 Generate code for the SH2a-FPU assuming the floating-point unit is in
16895 double-precision mode by default.
16899 Generate code for the SH3.
16903 Generate code for the SH3e.
16907 Generate code for the SH4 without a floating-point unit.
16909 @item -m4-single-only
16910 @opindex m4-single-only
16911 Generate code for the SH4 with a floating-point unit that only
16912 supports single-precision arithmetic.
16916 Generate code for the SH4 assuming the floating-point unit is in
16917 single-precision mode by default.
16921 Generate code for the SH4.
16925 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16926 floating-point unit is not used.
16928 @item -m4a-single-only
16929 @opindex m4a-single-only
16930 Generate code for the SH4a, in such a way that no double-precision
16931 floating point operations are used.
16934 @opindex m4a-single
16935 Generate code for the SH4a assuming the floating-point unit is in
16936 single-precision mode by default.
16940 Generate code for the SH4a.
16944 Same as @option{-m4a-nofpu}, except that it implicitly passes
16945 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16946 instructions at the moment.
16950 Compile code for the processor in big endian mode.
16954 Compile code for the processor in little endian mode.
16958 Align doubles at 64-bit boundaries. Note that this changes the calling
16959 conventions, and thus some functions from the standard C library will
16960 not work unless you recompile it first with @option{-mdalign}.
16964 Shorten some address references at link time, when possible; uses the
16965 linker option @option{-relax}.
16969 Use 32-bit offsets in @code{switch} tables. The default is to use
16974 Enable the use of bit manipulation instructions on SH2A.
16978 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16979 alignment constraints.
16983 Comply with the calling conventions defined by Renesas.
16987 Comply with the calling conventions defined by Renesas.
16991 Comply with the calling conventions defined for GCC before the Renesas
16992 conventions were available. This option is the default for all
16993 targets of the SH toolchain.
16996 @opindex mnomacsave
16997 Mark the @code{MAC} register as call-clobbered, even if
16998 @option{-mhitachi} is given.
17002 Increase IEEE-compliance of floating-point code.
17003 At the moment, this is equivalent to @option{-fno-finite-math-only}.
17004 When generating 16 bit SH opcodes, getting IEEE-conforming results for
17005 comparisons of NANs / infinities incurs extra overhead in every
17006 floating point comparison, therefore the default is set to
17007 @option{-ffinite-math-only}.
17009 @item -minline-ic_invalidate
17010 @opindex minline-ic_invalidate
17011 Inline code to invalidate instruction cache entries after setting up
17012 nested function trampolines.
17013 This option has no effect if -musermode is in effect and the selected
17014 code generation option (e.g. -m4) does not allow the use of the icbi
17016 If the selected code generation option does not allow the use of the icbi
17017 instruction, and -musermode is not in effect, the inlined code will
17018 manipulate the instruction cache address array directly with an associative
17019 write. This not only requires privileged mode, but it will also
17020 fail if the cache line had been mapped via the TLB and has become unmapped.
17024 Dump instruction size and location in the assembly code.
17027 @opindex mpadstruct
17028 This option is deprecated. It pads structures to multiple of 4 bytes,
17029 which is incompatible with the SH ABI@.
17033 Optimize for space instead of speed. Implied by @option{-Os}.
17036 @opindex mprefergot
17037 When generating position-independent code, emit function calls using
17038 the Global Offset Table instead of the Procedure Linkage Table.
17042 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17043 if the inlined code would not work in user mode.
17044 This is the default when the target is @code{sh-*-linux*}.
17046 @item -multcost=@var{number}
17047 @opindex multcost=@var{number}
17048 Set the cost to assume for a multiply insn.
17050 @item -mdiv=@var{strategy}
17051 @opindex mdiv=@var{strategy}
17052 Set the division strategy to use for SHmedia code. @var{strategy} must be
17053 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17054 inv:call2, inv:fp .
17055 "fp" performs the operation in floating point. This has a very high latency,
17056 but needs only a few instructions, so it might be a good choice if
17057 your code has enough easily exploitable ILP to allow the compiler to
17058 schedule the floating point instructions together with other instructions.
17059 Division by zero causes a floating point exception.
17060 "inv" uses integer operations to calculate the inverse of the divisor,
17061 and then multiplies the dividend with the inverse. This strategy allows
17062 cse and hoisting of the inverse calculation. Division by zero calculates
17063 an unspecified result, but does not trap.
17064 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17065 have been found, or if the entire operation has been hoisted to the same
17066 place, the last stages of the inverse calculation are intertwined with the
17067 final multiply to reduce the overall latency, at the expense of using a few
17068 more instructions, and thus offering fewer scheduling opportunities with
17070 "call" calls a library function that usually implements the inv:minlat
17072 This gives high code density for m5-*media-nofpu compilations.
17073 "call2" uses a different entry point of the same library function, where it
17074 assumes that a pointer to a lookup table has already been set up, which
17075 exposes the pointer load to cse / code hoisting optimizations.
17076 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17077 code generation, but if the code stays unoptimized, revert to the "call",
17078 "call2", or "fp" strategies, respectively. Note that the
17079 potentially-trapping side effect of division by zero is carried by a
17080 separate instruction, so it is possible that all the integer instructions
17081 are hoisted out, but the marker for the side effect stays where it is.
17082 A recombination to fp operations or a call is not possible in that case.
17083 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17084 that the inverse calculation was nor separated from the multiply, they speed
17085 up division where the dividend fits into 20 bits (plus sign where applicable),
17086 by inserting a test to skip a number of operations in this case; this test
17087 slows down the case of larger dividends. inv20u assumes the case of a such
17088 a small dividend to be unlikely, and inv20l assumes it to be likely.
17090 @item -maccumulate-outgoing-args
17091 @opindex maccumulate-outgoing-args
17092 Reserve space once for outgoing arguments in the function prologue rather
17093 than around each call. Generally beneficial for performance and size. Also
17094 needed for unwinding to avoid changing the stack frame around conditional code.
17096 @item -mdivsi3_libfunc=@var{name}
17097 @opindex mdivsi3_libfunc=@var{name}
17098 Set the name of the library function used for 32 bit signed division to
17099 @var{name}. This only affect the name used in the call and inv:call
17100 division strategies, and the compiler will still expect the same
17101 sets of input/output/clobbered registers as if this option was not present.
17103 @item -mfixed-range=@var{register-range}
17104 @opindex mfixed-range
17105 Generate code treating the given register range as fixed registers.
17106 A fixed register is one that the register allocator can not use. This is
17107 useful when compiling kernel code. A register range is specified as
17108 two registers separated by a dash. Multiple register ranges can be
17109 specified separated by a comma.
17111 @item -madjust-unroll
17112 @opindex madjust-unroll
17113 Throttle unrolling to avoid thrashing target registers.
17114 This option only has an effect if the gcc code base supports the
17115 TARGET_ADJUST_UNROLL_MAX target hook.
17117 @item -mindexed-addressing
17118 @opindex mindexed-addressing
17119 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17120 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17121 semantics for the indexed addressing mode. The architecture allows the
17122 implementation of processors with 64 bit MMU, which the OS could use to
17123 get 32 bit addressing, but since no current hardware implementation supports
17124 this or any other way to make the indexed addressing mode safe to use in
17125 the 32 bit ABI, the default is -mno-indexed-addressing.
17127 @item -mgettrcost=@var{number}
17128 @opindex mgettrcost=@var{number}
17129 Set the cost assumed for the gettr instruction to @var{number}.
17130 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17134 Assume pt* instructions won't trap. This will generally generate better
17135 scheduled code, but is unsafe on current hardware. The current architecture
17136 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17137 This has the unintentional effect of making it unsafe to schedule ptabs /
17138 ptrel before a branch, or hoist it out of a loop. For example,
17139 __do_global_ctors, a part of libgcc that runs constructors at program
17140 startup, calls functions in a list which is delimited by @minus{}1. With the
17141 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17142 That means that all the constructors will be run a bit quicker, but when
17143 the loop comes to the end of the list, the program crashes because ptabs
17144 loads @minus{}1 into a target register. Since this option is unsafe for any
17145 hardware implementing the current architecture specification, the default
17146 is -mno-pt-fixed. Unless the user specifies a specific cost with
17147 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17148 this deters register allocation using target registers for storing
17151 @item -minvalid-symbols
17152 @opindex minvalid-symbols
17153 Assume symbols might be invalid. Ordinary function symbols generated by
17154 the compiler will always be valid to load with movi/shori/ptabs or
17155 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17156 to generate symbols that will cause ptabs / ptrel to trap.
17157 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17158 It will then prevent cross-basic-block cse, hoisting and most scheduling
17159 of symbol loads. The default is @option{-mno-invalid-symbols}.
17162 @node Solaris 2 Options
17163 @subsection Solaris 2 Options
17164 @cindex Solaris 2 options
17166 These @samp{-m} options are supported on Solaris 2:
17169 @item -mimpure-text
17170 @opindex mimpure-text
17171 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17172 the compiler to not pass @option{-z text} to the linker when linking a
17173 shared object. Using this option, you can link position-dependent
17174 code into a shared object.
17176 @option{-mimpure-text} suppresses the ``relocations remain against
17177 allocatable but non-writable sections'' linker error message.
17178 However, the necessary relocations will trigger copy-on-write, and the
17179 shared object is not actually shared across processes. Instead of
17180 using @option{-mimpure-text}, you should compile all source code with
17181 @option{-fpic} or @option{-fPIC}.
17185 These switches are supported in addition to the above on Solaris 2:
17190 Add support for multithreading using the POSIX threads library. This
17191 option sets flags for both the preprocessor and linker. This option does
17192 not affect the thread safety of object code produced by the compiler or
17193 that of libraries supplied with it.
17197 This is a synonym for @option{-pthreads}.
17200 @node SPARC Options
17201 @subsection SPARC Options
17202 @cindex SPARC options
17204 These @samp{-m} options are supported on the SPARC:
17207 @item -mno-app-regs
17209 @opindex mno-app-regs
17211 Specify @option{-mapp-regs} to generate output using the global registers
17212 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17215 To be fully SVR4 ABI compliant at the cost of some performance loss,
17216 specify @option{-mno-app-regs}. You should compile libraries and system
17217 software with this option.
17223 With @option{-mflat}, the compiler does not generate save/restore instructions
17224 and uses a ``flat'' or single register window model. This model is compatible
17225 with the regular register window model. The local registers and the input
17226 registers (0--5) are still treated as ``call-saved'' registers and will be
17227 saved on the stack as needed.
17229 With @option{-mno-flat} (the default), the compiler generates save/restore
17230 instructions (except for leaf functions). This is the normal operating mode.
17233 @itemx -mhard-float
17235 @opindex mhard-float
17236 Generate output containing floating point instructions. This is the
17240 @itemx -msoft-float
17242 @opindex msoft-float
17243 Generate output containing library calls for floating point.
17244 @strong{Warning:} the requisite libraries are not available for all SPARC
17245 targets. Normally the facilities of the machine's usual C compiler are
17246 used, but this cannot be done directly in cross-compilation. You must make
17247 your own arrangements to provide suitable library functions for
17248 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17249 @samp{sparclite-*-*} do provide software floating point support.
17251 @option{-msoft-float} changes the calling convention in the output file;
17252 therefore, it is only useful if you compile @emph{all} of a program with
17253 this option. In particular, you need to compile @file{libgcc.a}, the
17254 library that comes with GCC, with @option{-msoft-float} in order for
17257 @item -mhard-quad-float
17258 @opindex mhard-quad-float
17259 Generate output containing quad-word (long double) floating point
17262 @item -msoft-quad-float
17263 @opindex msoft-quad-float
17264 Generate output containing library calls for quad-word (long double)
17265 floating point instructions. The functions called are those specified
17266 in the SPARC ABI@. This is the default.
17268 As of this writing, there are no SPARC implementations that have hardware
17269 support for the quad-word floating point instructions. They all invoke
17270 a trap handler for one of these instructions, and then the trap handler
17271 emulates the effect of the instruction. Because of the trap handler overhead,
17272 this is much slower than calling the ABI library routines. Thus the
17273 @option{-msoft-quad-float} option is the default.
17275 @item -mno-unaligned-doubles
17276 @itemx -munaligned-doubles
17277 @opindex mno-unaligned-doubles
17278 @opindex munaligned-doubles
17279 Assume that doubles have 8 byte alignment. This is the default.
17281 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17282 alignment only if they are contained in another type, or if they have an
17283 absolute address. Otherwise, it assumes they have 4 byte alignment.
17284 Specifying this option avoids some rare compatibility problems with code
17285 generated by other compilers. It is not the default because it results
17286 in a performance loss, especially for floating point code.
17288 @item -mno-faster-structs
17289 @itemx -mfaster-structs
17290 @opindex mno-faster-structs
17291 @opindex mfaster-structs
17292 With @option{-mfaster-structs}, the compiler assumes that structures
17293 should have 8 byte alignment. This enables the use of pairs of
17294 @code{ldd} and @code{std} instructions for copies in structure
17295 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17296 However, the use of this changed alignment directly violates the SPARC
17297 ABI@. Thus, it's intended only for use on targets where the developer
17298 acknowledges that their resulting code will not be directly in line with
17299 the rules of the ABI@.
17301 @item -mcpu=@var{cpu_type}
17303 Set the instruction set, register set, and instruction scheduling parameters
17304 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17305 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17306 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17307 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17308 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
17309 and @samp{niagara4}.
17311 Native Solaris and Linux toolchains also support the value @samp{native},
17312 which selects the best architecture option for the host processor.
17313 @option{-mcpu=native} has no effect if GCC does not recognize
17316 Default instruction scheduling parameters are used for values that select
17317 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17318 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17320 Here is a list of each supported architecture and their supported
17325 v8: supersparc, hypersparc, leon
17326 sparclite: f930, f934, sparclite86x
17328 v9: ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
17331 By default (unless configured otherwise), GCC generates code for the V7
17332 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17333 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17334 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17335 SPARCStation 1, 2, IPX etc.
17337 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17338 architecture. The only difference from V7 code is that the compiler emits
17339 the integer multiply and integer divide instructions which exist in SPARC-V8
17340 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17341 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17344 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17345 the SPARC architecture. This adds the integer multiply, integer divide step
17346 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17347 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17348 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17349 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17350 MB86934 chip, which is the more recent SPARClite with FPU@.
17352 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17353 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17354 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17355 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17356 optimizes it for the TEMIC SPARClet chip.
17358 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17359 architecture. This adds 64-bit integer and floating-point move instructions,
17360 3 additional floating-point condition code registers and conditional move
17361 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17362 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17363 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17364 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17365 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17366 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17367 additionally optimizes it for Sun UltraSPARC T2 chips. With
17368 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
17369 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
17370 additionally optimizes it for Sun UltraSPARC T4 chips.
17372 @item -mtune=@var{cpu_type}
17374 Set the instruction scheduling parameters for machine type
17375 @var{cpu_type}, but do not set the instruction set or register set that the
17376 option @option{-mcpu=@var{cpu_type}} would.
17378 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17379 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17380 that select a particular CPU implementation. Those are @samp{cypress},
17381 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17382 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17383 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
17384 native Solaris and Linux toolchains, @samp{native} can also be used.
17389 @opindex mno-v8plus
17390 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17391 difference from the V8 ABI is that the global and out registers are
17392 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17393 mode for all SPARC-V9 processors.
17399 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17400 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17406 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
17407 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
17408 when targetting a cpu that supports such instructions, such as Niagara-3 and
17412 These @samp{-m} options are supported in addition to the above
17413 on SPARC-V9 processors in 64-bit environments:
17416 @item -mlittle-endian
17417 @opindex mlittle-endian
17418 Generate code for a processor running in little-endian mode. It is only
17419 available for a few configurations and most notably not on Solaris and Linux.
17425 Generate code for a 32-bit or 64-bit environment.
17426 The 32-bit environment sets int, long and pointer to 32 bits.
17427 The 64-bit environment sets int to 32 bits and long and pointer
17430 @item -mcmodel=medlow
17431 @opindex mcmodel=medlow
17432 Generate code for the Medium/Low code model: 64-bit addresses, programs
17433 must be linked in the low 32 bits of memory. Programs can be statically
17434 or dynamically linked.
17436 @item -mcmodel=medmid
17437 @opindex mcmodel=medmid
17438 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17439 must be linked in the low 44 bits of memory, the text and data segments must
17440 be less than 2GB in size and the data segment must be located within 2GB of
17443 @item -mcmodel=medany
17444 @opindex mcmodel=medany
17445 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17446 may be linked anywhere in memory, the text and data segments must be less
17447 than 2GB in size and the data segment must be located within 2GB of the
17450 @item -mcmodel=embmedany
17451 @opindex mcmodel=embmedany
17452 Generate code for the Medium/Anywhere code model for embedded systems:
17453 64-bit addresses, the text and data segments must be less than 2GB in
17454 size, both starting anywhere in memory (determined at link time). The
17455 global register %g4 points to the base of the data segment. Programs
17456 are statically linked and PIC is not supported.
17459 @itemx -mno-stack-bias
17460 @opindex mstack-bias
17461 @opindex mno-stack-bias
17462 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17463 frame pointer if present, are offset by @minus{}2047 which must be added back
17464 when making stack frame references. This is the default in 64-bit mode.
17465 Otherwise, assume no such offset is present.
17469 @subsection SPU Options
17470 @cindex SPU options
17472 These @samp{-m} options are supported on the SPU:
17476 @itemx -merror-reloc
17477 @opindex mwarn-reloc
17478 @opindex merror-reloc
17480 The loader for SPU does not handle dynamic relocations. By default, GCC
17481 will give an error when it generates code that requires a dynamic
17482 relocation. @option{-mno-error-reloc} disables the error,
17483 @option{-mwarn-reloc} will generate a warning instead.
17486 @itemx -munsafe-dma
17488 @opindex munsafe-dma
17490 Instructions which initiate or test completion of DMA must not be
17491 reordered with respect to loads and stores of the memory which is being
17492 accessed. Users typically address this problem using the volatile
17493 keyword, but that can lead to inefficient code in places where the
17494 memory is known to not change. Rather than mark the memory as volatile
17495 we treat the DMA instructions as potentially effecting all memory. With
17496 @option{-munsafe-dma} users must use the volatile keyword to protect
17499 @item -mbranch-hints
17500 @opindex mbranch-hints
17502 By default, GCC will generate a branch hint instruction to avoid
17503 pipeline stalls for always taken or probably taken branches. A hint
17504 will not be generated closer than 8 instructions away from its branch.
17505 There is little reason to disable them, except for debugging purposes,
17506 or to make an object a little bit smaller.
17510 @opindex msmall-mem
17511 @opindex mlarge-mem
17513 By default, GCC generates code assuming that addresses are never larger
17514 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17515 a full 32 bit address.
17520 By default, GCC links against startup code that assumes the SPU-style
17521 main function interface (which has an unconventional parameter list).
17522 With @option{-mstdmain}, GCC will link your program against startup
17523 code that assumes a C99-style interface to @code{main}, including a
17524 local copy of @code{argv} strings.
17526 @item -mfixed-range=@var{register-range}
17527 @opindex mfixed-range
17528 Generate code treating the given register range as fixed registers.
17529 A fixed register is one that the register allocator can not use. This is
17530 useful when compiling kernel code. A register range is specified as
17531 two registers separated by a dash. Multiple register ranges can be
17532 specified separated by a comma.
17538 Compile code assuming that pointers to the PPU address space accessed
17539 via the @code{__ea} named address space qualifier are either 32 or 64
17540 bits wide. The default is 32 bits. As this is an ABI changing option,
17541 all object code in an executable must be compiled with the same setting.
17543 @item -maddress-space-conversion
17544 @itemx -mno-address-space-conversion
17545 @opindex maddress-space-conversion
17546 @opindex mno-address-space-conversion
17547 Allow/disallow treating the @code{__ea} address space as superset
17548 of the generic address space. This enables explicit type casts
17549 between @code{__ea} and generic pointer as well as implicit
17550 conversions of generic pointers to @code{__ea} pointers. The
17551 default is to allow address space pointer conversions.
17553 @item -mcache-size=@var{cache-size}
17554 @opindex mcache-size
17555 This option controls the version of libgcc that the compiler links to an
17556 executable and selects a software-managed cache for accessing variables
17557 in the @code{__ea} address space with a particular cache size. Possible
17558 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17559 and @samp{128}. The default cache size is 64KB.
17561 @item -matomic-updates
17562 @itemx -mno-atomic-updates
17563 @opindex matomic-updates
17564 @opindex mno-atomic-updates
17565 This option controls the version of libgcc that the compiler links to an
17566 executable and selects whether atomic updates to the software-managed
17567 cache of PPU-side variables are used. If you use atomic updates, changes
17568 to a PPU variable from SPU code using the @code{__ea} named address space
17569 qualifier will not interfere with changes to other PPU variables residing
17570 in the same cache line from PPU code. If you do not use atomic updates,
17571 such interference may occur; however, writing back cache lines will be
17572 more efficient. The default behavior is to use atomic updates.
17575 @itemx -mdual-nops=@var{n}
17576 @opindex mdual-nops
17577 By default, GCC will insert nops to increase dual issue when it expects
17578 it to increase performance. @var{n} can be a value from 0 to 10. A
17579 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17580 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17582 @item -mhint-max-nops=@var{n}
17583 @opindex mhint-max-nops
17584 Maximum number of nops to insert for a branch hint. A branch hint must
17585 be at least 8 instructions away from the branch it is effecting. GCC
17586 will insert up to @var{n} nops to enforce this, otherwise it will not
17587 generate the branch hint.
17589 @item -mhint-max-distance=@var{n}
17590 @opindex mhint-max-distance
17591 The encoding of the branch hint instruction limits the hint to be within
17592 256 instructions of the branch it is effecting. By default, GCC makes
17593 sure it is within 125.
17596 @opindex msafe-hints
17597 Work around a hardware bug which causes the SPU to stall indefinitely.
17598 By default, GCC will insert the @code{hbrp} instruction to make sure
17599 this stall won't happen.
17603 @node System V Options
17604 @subsection Options for System V
17606 These additional options are available on System V Release 4 for
17607 compatibility with other compilers on those systems:
17612 Create a shared object.
17613 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17617 Identify the versions of each tool used by the compiler, in a
17618 @code{.ident} assembler directive in the output.
17622 Refrain from adding @code{.ident} directives to the output file (this is
17625 @item -YP,@var{dirs}
17627 Search the directories @var{dirs}, and no others, for libraries
17628 specified with @option{-l}.
17630 @item -Ym,@var{dir}
17632 Look in the directory @var{dir} to find the M4 preprocessor.
17633 The assembler uses this option.
17634 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17635 @c the generic assembler that comes with Solaris takes just -Ym.
17639 @subsection V850 Options
17640 @cindex V850 Options
17642 These @samp{-m} options are defined for V850 implementations:
17646 @itemx -mno-long-calls
17647 @opindex mlong-calls
17648 @opindex mno-long-calls
17649 Treat all calls as being far away (near). If calls are assumed to be
17650 far away, the compiler will always load the functions address up into a
17651 register, and call indirect through the pointer.
17657 Do not optimize (do optimize) basic blocks that use the same index
17658 pointer 4 or more times to copy pointer into the @code{ep} register, and
17659 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17660 option is on by default if you optimize.
17662 @item -mno-prolog-function
17663 @itemx -mprolog-function
17664 @opindex mno-prolog-function
17665 @opindex mprolog-function
17666 Do not use (do use) external functions to save and restore registers
17667 at the prologue and epilogue of a function. The external functions
17668 are slower, but use less code space if more than one function saves
17669 the same number of registers. The @option{-mprolog-function} option
17670 is on by default if you optimize.
17674 Try to make the code as small as possible. At present, this just turns
17675 on the @option{-mep} and @option{-mprolog-function} options.
17677 @item -mtda=@var{n}
17679 Put static or global variables whose size is @var{n} bytes or less into
17680 the tiny data area that register @code{ep} points to. The tiny data
17681 area can hold up to 256 bytes in total (128 bytes for byte references).
17683 @item -msda=@var{n}
17685 Put static or global variables whose size is @var{n} bytes or less into
17686 the small data area that register @code{gp} points to. The small data
17687 area can hold up to 64 kilobytes.
17689 @item -mzda=@var{n}
17691 Put static or global variables whose size is @var{n} bytes or less into
17692 the first 32 kilobytes of memory.
17696 Specify that the target processor is the V850.
17699 @opindex mbig-switch
17700 Generate code suitable for big switch tables. Use this option only if
17701 the assembler/linker complain about out of range branches within a switch
17706 This option will cause r2 and r5 to be used in the code generated by
17707 the compiler. This setting is the default.
17709 @item -mno-app-regs
17710 @opindex mno-app-regs
17711 This option will cause r2 and r5 to be treated as fixed registers.
17715 Specify that the target processor is the V850E2V3. The preprocessor
17716 constants @samp{__v850e2v3__} will be defined if
17717 this option is used.
17721 Specify that the target processor is the V850E2. The preprocessor
17722 constants @samp{__v850e2__} will be defined if this option is used.
17726 Specify that the target processor is the V850E1. The preprocessor
17727 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17728 this option is used.
17732 Specify that the target processor is the V850ES. This is an alias for
17733 the @option{-mv850e1} option.
17737 Specify that the target processor is the V850E@. The preprocessor
17738 constant @samp{__v850e__} will be defined if this option is used.
17740 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17741 nor @option{-mv850e2} nor @option{-mv850e2v3}
17742 are defined then a default target processor will be chosen and the
17743 relevant @samp{__v850*__} preprocessor constant will be defined.
17745 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17746 defined, regardless of which processor variant is the target.
17748 @item -mdisable-callt
17749 @opindex mdisable-callt
17750 This option will suppress generation of the CALLT instruction for the
17751 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17752 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17757 @subsection VAX Options
17758 @cindex VAX options
17760 These @samp{-m} options are defined for the VAX:
17765 Do not output certain jump instructions (@code{aobleq} and so on)
17766 that the Unix assembler for the VAX cannot handle across long
17771 Do output those jump instructions, on the assumption that you
17772 will assemble with the GNU assembler.
17776 Output code for g-format floating point numbers instead of d-format.
17779 @node VxWorks Options
17780 @subsection VxWorks Options
17781 @cindex VxWorks Options
17783 The options in this section are defined for all VxWorks targets.
17784 Options specific to the target hardware are listed with the other
17785 options for that target.
17790 GCC can generate code for both VxWorks kernels and real time processes
17791 (RTPs). This option switches from the former to the latter. It also
17792 defines the preprocessor macro @code{__RTP__}.
17795 @opindex non-static
17796 Link an RTP executable against shared libraries rather than static
17797 libraries. The options @option{-static} and @option{-shared} can
17798 also be used for RTPs (@pxref{Link Options}); @option{-static}
17805 These options are passed down to the linker. They are defined for
17806 compatibility with Diab.
17809 @opindex Xbind-lazy
17810 Enable lazy binding of function calls. This option is equivalent to
17811 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17815 Disable lazy binding of function calls. This option is the default and
17816 is defined for compatibility with Diab.
17819 @node x86-64 Options
17820 @subsection x86-64 Options
17821 @cindex x86-64 options
17823 These are listed under @xref{i386 and x86-64 Options}.
17825 @node Xstormy16 Options
17826 @subsection Xstormy16 Options
17827 @cindex Xstormy16 Options
17829 These options are defined for Xstormy16:
17834 Choose startup files and linker script suitable for the simulator.
17837 @node Xtensa Options
17838 @subsection Xtensa Options
17839 @cindex Xtensa Options
17841 These options are supported for Xtensa targets:
17845 @itemx -mno-const16
17847 @opindex mno-const16
17848 Enable or disable use of @code{CONST16} instructions for loading
17849 constant values. The @code{CONST16} instruction is currently not a
17850 standard option from Tensilica. When enabled, @code{CONST16}
17851 instructions are always used in place of the standard @code{L32R}
17852 instructions. The use of @code{CONST16} is enabled by default only if
17853 the @code{L32R} instruction is not available.
17856 @itemx -mno-fused-madd
17857 @opindex mfused-madd
17858 @opindex mno-fused-madd
17859 Enable or disable use of fused multiply/add and multiply/subtract
17860 instructions in the floating-point option. This has no effect if the
17861 floating-point option is not also enabled. Disabling fused multiply/add
17862 and multiply/subtract instructions forces the compiler to use separate
17863 instructions for the multiply and add/subtract operations. This may be
17864 desirable in some cases where strict IEEE 754-compliant results are
17865 required: the fused multiply add/subtract instructions do not round the
17866 intermediate result, thereby producing results with @emph{more} bits of
17867 precision than specified by the IEEE standard. Disabling fused multiply
17868 add/subtract instructions also ensures that the program output is not
17869 sensitive to the compiler's ability to combine multiply and add/subtract
17872 @item -mserialize-volatile
17873 @itemx -mno-serialize-volatile
17874 @opindex mserialize-volatile
17875 @opindex mno-serialize-volatile
17876 When this option is enabled, GCC inserts @code{MEMW} instructions before
17877 @code{volatile} memory references to guarantee sequential consistency.
17878 The default is @option{-mserialize-volatile}. Use
17879 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17881 @item -mforce-no-pic
17882 @opindex mforce-no-pic
17883 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17884 position-independent code (PIC), this option disables PIC for compiling
17887 @item -mtext-section-literals
17888 @itemx -mno-text-section-literals
17889 @opindex mtext-section-literals
17890 @opindex mno-text-section-literals
17891 Control the treatment of literal pools. The default is
17892 @option{-mno-text-section-literals}, which places literals in a separate
17893 section in the output file. This allows the literal pool to be placed
17894 in a data RAM/ROM, and it also allows the linker to combine literal
17895 pools from separate object files to remove redundant literals and
17896 improve code size. With @option{-mtext-section-literals}, the literals
17897 are interspersed in the text section in order to keep them as close as
17898 possible to their references. This may be necessary for large assembly
17901 @item -mtarget-align
17902 @itemx -mno-target-align
17903 @opindex mtarget-align
17904 @opindex mno-target-align
17905 When this option is enabled, GCC instructs the assembler to
17906 automatically align instructions to reduce branch penalties at the
17907 expense of some code density. The assembler attempts to widen density
17908 instructions to align branch targets and the instructions following call
17909 instructions. If there are not enough preceding safe density
17910 instructions to align a target, no widening will be performed. The
17911 default is @option{-mtarget-align}. These options do not affect the
17912 treatment of auto-aligned instructions like @code{LOOP}, which the
17913 assembler will always align, either by widening density instructions or
17914 by inserting no-op instructions.
17917 @itemx -mno-longcalls
17918 @opindex mlongcalls
17919 @opindex mno-longcalls
17920 When this option is enabled, GCC instructs the assembler to translate
17921 direct calls to indirect calls unless it can determine that the target
17922 of a direct call is in the range allowed by the call instruction. This
17923 translation typically occurs for calls to functions in other source
17924 files. Specifically, the assembler translates a direct @code{CALL}
17925 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17926 The default is @option{-mno-longcalls}. This option should be used in
17927 programs where the call target can potentially be out of range. This
17928 option is implemented in the assembler, not the compiler, so the
17929 assembly code generated by GCC will still show direct call
17930 instructions---look at the disassembled object code to see the actual
17931 instructions. Note that the assembler will use an indirect call for
17932 every cross-file call, not just those that really will be out of range.
17935 @node zSeries Options
17936 @subsection zSeries Options
17937 @cindex zSeries options
17939 These are listed under @xref{S/390 and zSeries Options}.
17941 @node Code Gen Options
17942 @section Options for Code Generation Conventions
17943 @cindex code generation conventions
17944 @cindex options, code generation
17945 @cindex run-time options
17947 These machine-independent options control the interface conventions
17948 used in code generation.
17950 Most of them have both positive and negative forms; the negative form
17951 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17952 one of the forms is listed---the one which is not the default. You
17953 can figure out the other form by either removing @samp{no-} or adding
17957 @item -fbounds-check
17958 @opindex fbounds-check
17959 For front-ends that support it, generate additional code to check that
17960 indices used to access arrays are within the declared range. This is
17961 currently only supported by the Java and Fortran front-ends, where
17962 this option defaults to true and false respectively.
17966 This option generates traps for signed overflow on addition, subtraction,
17967 multiplication operations.
17971 This option instructs the compiler to assume that signed arithmetic
17972 overflow of addition, subtraction and multiplication wraps around
17973 using twos-complement representation. This flag enables some optimizations
17974 and disables others. This option is enabled by default for the Java
17975 front-end, as required by the Java language specification.
17978 @opindex fexceptions
17979 Enable exception handling. Generates extra code needed to propagate
17980 exceptions. For some targets, this implies GCC will generate frame
17981 unwind information for all functions, which can produce significant data
17982 size overhead, although it does not affect execution. If you do not
17983 specify this option, GCC will enable it by default for languages like
17984 C++ which normally require exception handling, and disable it for
17985 languages like C that do not normally require it. However, you may need
17986 to enable this option when compiling C code that needs to interoperate
17987 properly with exception handlers written in C++. You may also wish to
17988 disable this option if you are compiling older C++ programs that don't
17989 use exception handling.
17991 @item -fnon-call-exceptions
17992 @opindex fnon-call-exceptions
17993 Generate code that allows trapping instructions to throw exceptions.
17994 Note that this requires platform-specific runtime support that does
17995 not exist everywhere. Moreover, it only allows @emph{trapping}
17996 instructions to throw exceptions, i.e.@: memory references or floating
17997 point instructions. It does not allow exceptions to be thrown from
17998 arbitrary signal handlers such as @code{SIGALRM}.
18000 @item -funwind-tables
18001 @opindex funwind-tables
18002 Similar to @option{-fexceptions}, except that it will just generate any needed
18003 static data, but will not affect the generated code in any other way.
18004 You will normally not enable this option; instead, a language processor
18005 that needs this handling would enable it on your behalf.
18007 @item -fasynchronous-unwind-tables
18008 @opindex fasynchronous-unwind-tables
18009 Generate unwind table in dwarf2 format, if supported by target machine. The
18010 table is exact at each instruction boundary, so it can be used for stack
18011 unwinding from asynchronous events (such as debugger or garbage collector).
18013 @item -fpcc-struct-return
18014 @opindex fpcc-struct-return
18015 Return ``short'' @code{struct} and @code{union} values in memory like
18016 longer ones, rather than in registers. This convention is less
18017 efficient, but it has the advantage of allowing intercallability between
18018 GCC-compiled files and files compiled with other compilers, particularly
18019 the Portable C Compiler (pcc).
18021 The precise convention for returning structures in memory depends
18022 on the target configuration macros.
18024 Short structures and unions are those whose size and alignment match
18025 that of some integer type.
18027 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
18028 switch is not binary compatible with code compiled with the
18029 @option{-freg-struct-return} switch.
18030 Use it to conform to a non-default application binary interface.
18032 @item -freg-struct-return
18033 @opindex freg-struct-return
18034 Return @code{struct} and @code{union} values in registers when possible.
18035 This is more efficient for small structures than
18036 @option{-fpcc-struct-return}.
18038 If you specify neither @option{-fpcc-struct-return} nor
18039 @option{-freg-struct-return}, GCC defaults to whichever convention is
18040 standard for the target. If there is no standard convention, GCC
18041 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18042 the principal compiler. In those cases, we can choose the standard, and
18043 we chose the more efficient register return alternative.
18045 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18046 switch is not binary compatible with code compiled with the
18047 @option{-fpcc-struct-return} switch.
18048 Use it to conform to a non-default application binary interface.
18050 @item -fshort-enums
18051 @opindex fshort-enums
18052 Allocate to an @code{enum} type only as many bytes as it needs for the
18053 declared range of possible values. Specifically, the @code{enum} type
18054 will be equivalent to the smallest integer type which has enough room.
18056 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18057 code that is not binary compatible with code generated without that switch.
18058 Use it to conform to a non-default application binary interface.
18060 @item -fshort-double
18061 @opindex fshort-double
18062 Use the same size for @code{double} as for @code{float}.
18064 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18065 code that is not binary compatible with code generated without that switch.
18066 Use it to conform to a non-default application binary interface.
18068 @item -fshort-wchar
18069 @opindex fshort-wchar
18070 Override the underlying type for @samp{wchar_t} to be @samp{short
18071 unsigned int} instead of the default for the target. This option is
18072 useful for building programs to run under WINE@.
18074 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18075 code that is not binary compatible with code generated without that switch.
18076 Use it to conform to a non-default application binary interface.
18079 @opindex fno-common
18080 In C code, controls the placement of uninitialized global variables.
18081 Unix C compilers have traditionally permitted multiple definitions of
18082 such variables in different compilation units by placing the variables
18084 This is the behavior specified by @option{-fcommon}, and is the default
18085 for GCC on most targets.
18086 On the other hand, this behavior is not required by ISO C, and on some
18087 targets may carry a speed or code size penalty on variable references.
18088 The @option{-fno-common} option specifies that the compiler should place
18089 uninitialized global variables in the data section of the object file,
18090 rather than generating them as common blocks.
18091 This has the effect that if the same variable is declared
18092 (without @code{extern}) in two different compilations,
18093 you will get a multiple-definition error when you link them.
18094 In this case, you must compile with @option{-fcommon} instead.
18095 Compiling with @option{-fno-common} is useful on targets for which
18096 it provides better performance, or if you wish to verify that the
18097 program will work on other systems which always treat uninitialized
18098 variable declarations this way.
18102 Ignore the @samp{#ident} directive.
18104 @item -finhibit-size-directive
18105 @opindex finhibit-size-directive
18106 Don't output a @code{.size} assembler directive, or anything else that
18107 would cause trouble if the function is split in the middle, and the
18108 two halves are placed at locations far apart in memory. This option is
18109 used when compiling @file{crtstuff.c}; you should not need to use it
18112 @item -fverbose-asm
18113 @opindex fverbose-asm
18114 Put extra commentary information in the generated assembly code to
18115 make it more readable. This option is generally only of use to those
18116 who actually need to read the generated assembly code (perhaps while
18117 debugging the compiler itself).
18119 @option{-fno-verbose-asm}, the default, causes the
18120 extra information to be omitted and is useful when comparing two assembler
18123 @item -frecord-gcc-switches
18124 @opindex frecord-gcc-switches
18125 This switch causes the command line that was used to invoke the
18126 compiler to be recorded into the object file that is being created.
18127 This switch is only implemented on some targets and the exact format
18128 of the recording is target and binary file format dependent, but it
18129 usually takes the form of a section containing ASCII text. This
18130 switch is related to the @option{-fverbose-asm} switch, but that
18131 switch only records information in the assembler output file as
18132 comments, so it never reaches the object file.
18133 See also @option{-grecord-gcc-switches} for another
18134 way of storing compiler options into the object file.
18138 @cindex global offset table
18140 Generate position-independent code (PIC) suitable for use in a shared
18141 library, if supported for the target machine. Such code accesses all
18142 constant addresses through a global offset table (GOT)@. The dynamic
18143 loader resolves the GOT entries when the program starts (the dynamic
18144 loader is not part of GCC; it is part of the operating system). If
18145 the GOT size for the linked executable exceeds a machine-specific
18146 maximum size, you get an error message from the linker indicating that
18147 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18148 instead. (These maximums are 8k on the SPARC and 32k
18149 on the m68k and RS/6000. The 386 has no such limit.)
18151 Position-independent code requires special support, and therefore works
18152 only on certain machines. For the 386, GCC supports PIC for System V
18153 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18154 position-independent.
18156 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18161 If supported for the target machine, emit position-independent code,
18162 suitable for dynamic linking and avoiding any limit on the size of the
18163 global offset table. This option makes a difference on the m68k,
18164 PowerPC and SPARC@.
18166 Position-independent code requires special support, and therefore works
18167 only on certain machines.
18169 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18176 These options are similar to @option{-fpic} and @option{-fPIC}, but
18177 generated position independent code can be only linked into executables.
18178 Usually these options are used when @option{-pie} GCC option will be
18179 used during linking.
18181 @option{-fpie} and @option{-fPIE} both define the macros
18182 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18183 for @option{-fpie} and 2 for @option{-fPIE}.
18185 @item -fno-jump-tables
18186 @opindex fno-jump-tables
18187 Do not use jump tables for switch statements even where it would be
18188 more efficient than other code generation strategies. This option is
18189 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18190 building code which forms part of a dynamic linker and cannot
18191 reference the address of a jump table. On some targets, jump tables
18192 do not require a GOT and this option is not needed.
18194 @item -ffixed-@var{reg}
18196 Treat the register named @var{reg} as a fixed register; generated code
18197 should never refer to it (except perhaps as a stack pointer, frame
18198 pointer or in some other fixed role).
18200 @var{reg} must be the name of a register. The register names accepted
18201 are machine-specific and are defined in the @code{REGISTER_NAMES}
18202 macro in the machine description macro file.
18204 This flag does not have a negative form, because it specifies a
18207 @item -fcall-used-@var{reg}
18208 @opindex fcall-used
18209 Treat the register named @var{reg} as an allocable register that is
18210 clobbered by function calls. It may be allocated for temporaries or
18211 variables that do not live across a call. Functions compiled this way
18212 will not save and restore the register @var{reg}.
18214 It is an error to used this flag with the frame pointer or stack pointer.
18215 Use of this flag for other registers that have fixed pervasive roles in
18216 the machine's execution model will produce disastrous results.
18218 This flag does not have a negative form, because it specifies a
18221 @item -fcall-saved-@var{reg}
18222 @opindex fcall-saved
18223 Treat the register named @var{reg} as an allocable register saved by
18224 functions. It may be allocated even for temporaries or variables that
18225 live across a call. Functions compiled this way will save and restore
18226 the register @var{reg} if they use it.
18228 It is an error to used this flag with the frame pointer or stack pointer.
18229 Use of this flag for other registers that have fixed pervasive roles in
18230 the machine's execution model will produce disastrous results.
18232 A different sort of disaster will result from the use of this flag for
18233 a register in which function values may be returned.
18235 This flag does not have a negative form, because it specifies a
18238 @item -fpack-struct[=@var{n}]
18239 @opindex fpack-struct
18240 Without a value specified, pack all structure members together without
18241 holes. When a value is specified (which must be a small power of two), pack
18242 structure members according to this value, representing the maximum
18243 alignment (that is, objects with default alignment requirements larger than
18244 this will be output potentially unaligned at the next fitting location.
18246 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18247 code that is not binary compatible with code generated without that switch.
18248 Additionally, it makes the code suboptimal.
18249 Use it to conform to a non-default application binary interface.
18251 @item -finstrument-functions
18252 @opindex finstrument-functions
18253 Generate instrumentation calls for entry and exit to functions. Just
18254 after function entry and just before function exit, the following
18255 profiling functions will be called with the address of the current
18256 function and its call site. (On some platforms,
18257 @code{__builtin_return_address} does not work beyond the current
18258 function, so the call site information may not be available to the
18259 profiling functions otherwise.)
18262 void __cyg_profile_func_enter (void *this_fn,
18264 void __cyg_profile_func_exit (void *this_fn,
18268 The first argument is the address of the start of the current function,
18269 which may be looked up exactly in the symbol table.
18271 This instrumentation is also done for functions expanded inline in other
18272 functions. The profiling calls will indicate where, conceptually, the
18273 inline function is entered and exited. This means that addressable
18274 versions of such functions must be available. If all your uses of a
18275 function are expanded inline, this may mean an additional expansion of
18276 code size. If you use @samp{extern inline} in your C code, an
18277 addressable version of such functions must be provided. (This is
18278 normally the case anyways, but if you get lucky and the optimizer always
18279 expands the functions inline, you might have gotten away without
18280 providing static copies.)
18282 A function may be given the attribute @code{no_instrument_function}, in
18283 which case this instrumentation will not be done. This can be used, for
18284 example, for the profiling functions listed above, high-priority
18285 interrupt routines, and any functions from which the profiling functions
18286 cannot safely be called (perhaps signal handlers, if the profiling
18287 routines generate output or allocate memory).
18289 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18290 @opindex finstrument-functions-exclude-file-list
18292 Set the list of functions that are excluded from instrumentation (see
18293 the description of @code{-finstrument-functions}). If the file that
18294 contains a function definition matches with one of @var{file}, then
18295 that function is not instrumented. The match is done on substrings:
18296 if the @var{file} parameter is a substring of the file name, it is
18297 considered to be a match.
18302 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18306 will exclude any inline function defined in files whose pathnames
18307 contain @code{/bits/stl} or @code{include/sys}.
18309 If, for some reason, you want to include letter @code{','} in one of
18310 @var{sym}, write @code{'\,'}. For example,
18311 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18312 (note the single quote surrounding the option).
18314 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18315 @opindex finstrument-functions-exclude-function-list
18317 This is similar to @code{-finstrument-functions-exclude-file-list},
18318 but this option sets the list of function names to be excluded from
18319 instrumentation. The function name to be matched is its user-visible
18320 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18321 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18322 match is done on substrings: if the @var{sym} parameter is a substring
18323 of the function name, it is considered to be a match. For C99 and C++
18324 extended identifiers, the function name must be given in UTF-8, not
18325 using universal character names.
18327 @item -fstack-check
18328 @opindex fstack-check
18329 Generate code to verify that you do not go beyond the boundary of the
18330 stack. You should specify this flag if you are running in an
18331 environment with multiple threads, but only rarely need to specify it in
18332 a single-threaded environment since stack overflow is automatically
18333 detected on nearly all systems if there is only one stack.
18335 Note that this switch does not actually cause checking to be done; the
18336 operating system or the language runtime must do that. The switch causes
18337 generation of code to ensure that they see the stack being extended.
18339 You can additionally specify a string parameter: @code{no} means no
18340 checking, @code{generic} means force the use of old-style checking,
18341 @code{specific} means use the best checking method and is equivalent
18342 to bare @option{-fstack-check}.
18344 Old-style checking is a generic mechanism that requires no specific
18345 target support in the compiler but comes with the following drawbacks:
18349 Modified allocation strategy for large objects: they will always be
18350 allocated dynamically if their size exceeds a fixed threshold.
18353 Fixed limit on the size of the static frame of functions: when it is
18354 topped by a particular function, stack checking is not reliable and
18355 a warning is issued by the compiler.
18358 Inefficiency: because of both the modified allocation strategy and the
18359 generic implementation, the performances of the code are hampered.
18362 Note that old-style stack checking is also the fallback method for
18363 @code{specific} if no target support has been added in the compiler.
18365 @item -fstack-limit-register=@var{reg}
18366 @itemx -fstack-limit-symbol=@var{sym}
18367 @itemx -fno-stack-limit
18368 @opindex fstack-limit-register
18369 @opindex fstack-limit-symbol
18370 @opindex fno-stack-limit
18371 Generate code to ensure that the stack does not grow beyond a certain value,
18372 either the value of a register or the address of a symbol. If the stack
18373 would grow beyond the value, a signal is raised. For most targets,
18374 the signal is raised before the stack overruns the boundary, so
18375 it is possible to catch the signal without taking special precautions.
18377 For instance, if the stack starts at absolute address @samp{0x80000000}
18378 and grows downwards, you can use the flags
18379 @option{-fstack-limit-symbol=__stack_limit} and
18380 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18381 of 128KB@. Note that this may only work with the GNU linker.
18383 @item -fsplit-stack
18384 @opindex fsplit-stack
18385 Generate code to automatically split the stack before it overflows.
18386 The resulting program has a discontiguous stack which can only
18387 overflow if the program is unable to allocate any more memory. This
18388 is most useful when running threaded programs, as it is no longer
18389 necessary to calculate a good stack size to use for each thread. This
18390 is currently only implemented for the i386 and x86_64 backends running
18393 When code compiled with @option{-fsplit-stack} calls code compiled
18394 without @option{-fsplit-stack}, there may not be much stack space
18395 available for the latter code to run. If compiling all code,
18396 including library code, with @option{-fsplit-stack} is not an option,
18397 then the linker can fix up these calls so that the code compiled
18398 without @option{-fsplit-stack} always has a large stack. Support for
18399 this is implemented in the gold linker in GNU binutils release 2.21
18402 @item -fleading-underscore
18403 @opindex fleading-underscore
18404 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18405 change the way C symbols are represented in the object file. One use
18406 is to help link with legacy assembly code.
18408 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18409 generate code that is not binary compatible with code generated without that
18410 switch. Use it to conform to a non-default application binary interface.
18411 Not all targets provide complete support for this switch.
18413 @item -ftls-model=@var{model}
18414 @opindex ftls-model
18415 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18416 The @var{model} argument should be one of @code{global-dynamic},
18417 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18419 The default without @option{-fpic} is @code{initial-exec}; with
18420 @option{-fpic} the default is @code{global-dynamic}.
18422 @item -fvisibility=@var{default|internal|hidden|protected}
18423 @opindex fvisibility
18424 Set the default ELF image symbol visibility to the specified option---all
18425 symbols will be marked with this unless overridden within the code.
18426 Using this feature can very substantially improve linking and
18427 load times of shared object libraries, produce more optimized
18428 code, provide near-perfect API export and prevent symbol clashes.
18429 It is @strong{strongly} recommended that you use this in any shared objects
18432 Despite the nomenclature, @code{default} always means public; i.e.,
18433 available to be linked against from outside the shared object.
18434 @code{protected} and @code{internal} are pretty useless in real-world
18435 usage so the only other commonly used option will be @code{hidden}.
18436 The default if @option{-fvisibility} isn't specified is
18437 @code{default}, i.e., make every
18438 symbol public---this causes the same behavior as previous versions of
18441 A good explanation of the benefits offered by ensuring ELF
18442 symbols have the correct visibility is given by ``How To Write
18443 Shared Libraries'' by Ulrich Drepper (which can be found at
18444 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18445 solution made possible by this option to marking things hidden when
18446 the default is public is to make the default hidden and mark things
18447 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18448 and @code{__attribute__ ((visibility("default")))} instead of
18449 @code{__declspec(dllexport)} you get almost identical semantics with
18450 identical syntax. This is a great boon to those working with
18451 cross-platform projects.
18453 For those adding visibility support to existing code, you may find
18454 @samp{#pragma GCC visibility} of use. This works by you enclosing
18455 the declarations you wish to set visibility for with (for example)
18456 @samp{#pragma GCC visibility push(hidden)} and
18457 @samp{#pragma GCC visibility pop}.
18458 Bear in mind that symbol visibility should be viewed @strong{as
18459 part of the API interface contract} and thus all new code should
18460 always specify visibility when it is not the default; i.e., declarations
18461 only for use within the local DSO should @strong{always} be marked explicitly
18462 as hidden as so to avoid PLT indirection overheads---making this
18463 abundantly clear also aids readability and self-documentation of the code.
18464 Note that due to ISO C++ specification requirements, operator new and
18465 operator delete must always be of default visibility.
18467 Be aware that headers from outside your project, in particular system
18468 headers and headers from any other library you use, may not be
18469 expecting to be compiled with visibility other than the default. You
18470 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18471 before including any such headers.
18473 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18474 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18475 no modifications. However, this means that calls to @samp{extern}
18476 functions with no explicit visibility will use the PLT, so it is more
18477 effective to use @samp{__attribute ((visibility))} and/or
18478 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18479 declarations should be treated as hidden.
18481 Note that @samp{-fvisibility} does affect C++ vague linkage
18482 entities. This means that, for instance, an exception class that will
18483 be thrown between DSOs must be explicitly marked with default
18484 visibility so that the @samp{type_info} nodes will be unified between
18487 An overview of these techniques, their benefits and how to use them
18488 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18490 @item -fstrict-volatile-bitfields
18491 @opindex fstrict-volatile-bitfields
18492 This option should be used if accesses to volatile bitfields (or other
18493 structure fields, although the compiler usually honors those types
18494 anyway) should use a single access of the width of the
18495 field's type, aligned to a natural alignment if possible. For
18496 example, targets with memory-mapped peripheral registers might require
18497 all such accesses to be 16 bits wide; with this flag the user could
18498 declare all peripheral bitfields as ``unsigned short'' (assuming short
18499 is 16 bits on these targets) to force GCC to use 16 bit accesses
18500 instead of, perhaps, a more efficient 32 bit access.
18502 If this option is disabled, the compiler will use the most efficient
18503 instruction. In the previous example, that might be a 32-bit load
18504 instruction, even though that will access bytes that do not contain
18505 any portion of the bitfield, or memory-mapped registers unrelated to
18506 the one being updated.
18508 If the target requires strict alignment, and honoring the field
18509 type would require violating this alignment, a warning is issued.
18510 If the field has @code{packed} attribute, the access is done without
18511 honoring the field type. If the field doesn't have @code{packed}
18512 attribute, the access is done honoring the field type. In both cases,
18513 GCC assumes that the user knows something about the target hardware
18514 that it is unaware of.
18516 The default value of this option is determined by the application binary
18517 interface for the target processor.
18523 @node Environment Variables
18524 @section Environment Variables Affecting GCC
18525 @cindex environment variables
18527 @c man begin ENVIRONMENT
18528 This section describes several environment variables that affect how GCC
18529 operates. Some of them work by specifying directories or prefixes to use
18530 when searching for various kinds of files. Some are used to specify other
18531 aspects of the compilation environment.
18533 Note that you can also specify places to search using options such as
18534 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18535 take precedence over places specified using environment variables, which
18536 in turn take precedence over those specified by the configuration of GCC@.
18537 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18538 GNU Compiler Collection (GCC) Internals}.
18543 @c @itemx LC_COLLATE
18545 @c @itemx LC_MONETARY
18546 @c @itemx LC_NUMERIC
18551 @c @findex LC_COLLATE
18552 @findex LC_MESSAGES
18553 @c @findex LC_MONETARY
18554 @c @findex LC_NUMERIC
18558 These environment variables control the way that GCC uses
18559 localization information that allow GCC to work with different
18560 national conventions. GCC inspects the locale categories
18561 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18562 so. These locale categories can be set to any value supported by your
18563 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18564 Kingdom encoded in UTF-8.
18566 The @env{LC_CTYPE} environment variable specifies character
18567 classification. GCC uses it to determine the character boundaries in
18568 a string; this is needed for some multibyte encodings that contain quote
18569 and escape characters that would otherwise be interpreted as a string
18572 The @env{LC_MESSAGES} environment variable specifies the language to
18573 use in diagnostic messages.
18575 If the @env{LC_ALL} environment variable is set, it overrides the value
18576 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18577 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18578 environment variable. If none of these variables are set, GCC
18579 defaults to traditional C English behavior.
18583 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18584 files. GCC uses temporary files to hold the output of one stage of
18585 compilation which is to be used as input to the next stage: for example,
18586 the output of the preprocessor, which is the input to the compiler
18589 @item GCC_COMPARE_DEBUG
18590 @findex GCC_COMPARE_DEBUG
18591 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
18592 @option{-fcompare-debug} to the compiler driver. See the documentation
18593 of this option for more details.
18595 @item GCC_EXEC_PREFIX
18596 @findex GCC_EXEC_PREFIX
18597 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18598 names of the subprograms executed by the compiler. No slash is added
18599 when this prefix is combined with the name of a subprogram, but you can
18600 specify a prefix that ends with a slash if you wish.
18602 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18603 an appropriate prefix to use based on the pathname it was invoked with.
18605 If GCC cannot find the subprogram using the specified prefix, it
18606 tries looking in the usual places for the subprogram.
18608 The default value of @env{GCC_EXEC_PREFIX} is
18609 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18610 the installed compiler. In many cases @var{prefix} is the value
18611 of @code{prefix} when you ran the @file{configure} script.
18613 Other prefixes specified with @option{-B} take precedence over this prefix.
18615 This prefix is also used for finding files such as @file{crt0.o} that are
18618 In addition, the prefix is used in an unusual way in finding the
18619 directories to search for header files. For each of the standard
18620 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18621 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18622 replacing that beginning with the specified prefix to produce an
18623 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18624 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18625 These alternate directories are searched first; the standard directories
18626 come next. If a standard directory begins with the configured
18627 @var{prefix} then the value of @var{prefix} is replaced by
18628 @env{GCC_EXEC_PREFIX} when looking for header files.
18630 @item COMPILER_PATH
18631 @findex COMPILER_PATH
18632 The value of @env{COMPILER_PATH} is a colon-separated list of
18633 directories, much like @env{PATH}. GCC tries the directories thus
18634 specified when searching for subprograms, if it can't find the
18635 subprograms using @env{GCC_EXEC_PREFIX}.
18638 @findex LIBRARY_PATH
18639 The value of @env{LIBRARY_PATH} is a colon-separated list of
18640 directories, much like @env{PATH}. When configured as a native compiler,
18641 GCC tries the directories thus specified when searching for special
18642 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18643 using GCC also uses these directories when searching for ordinary
18644 libraries for the @option{-l} option (but directories specified with
18645 @option{-L} come first).
18649 @cindex locale definition
18650 This variable is used to pass locale information to the compiler. One way in
18651 which this information is used is to determine the character set to be used
18652 when character literals, string literals and comments are parsed in C and C++.
18653 When the compiler is configured to allow multibyte characters,
18654 the following values for @env{LANG} are recognized:
18658 Recognize JIS characters.
18660 Recognize SJIS characters.
18662 Recognize EUCJP characters.
18665 If @env{LANG} is not defined, or if it has some other value, then the
18666 compiler will use mblen and mbtowc as defined by the default locale to
18667 recognize and translate multibyte characters.
18671 Some additional environments variables affect the behavior of the
18674 @include cppenv.texi
18678 @node Precompiled Headers
18679 @section Using Precompiled Headers
18680 @cindex precompiled headers
18681 @cindex speed of compilation
18683 Often large projects have many header files that are included in every
18684 source file. The time the compiler takes to process these header files
18685 over and over again can account for nearly all of the time required to
18686 build the project. To make builds faster, GCC allows users to
18687 `precompile' a header file; then, if builds can use the precompiled
18688 header file they will be much faster.
18690 To create a precompiled header file, simply compile it as you would any
18691 other file, if necessary using the @option{-x} option to make the driver
18692 treat it as a C or C++ header file. You will probably want to use a
18693 tool like @command{make} to keep the precompiled header up-to-date when
18694 the headers it contains change.
18696 A precompiled header file will be searched for when @code{#include} is
18697 seen in the compilation. As it searches for the included file
18698 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18699 compiler looks for a precompiled header in each directory just before it
18700 looks for the include file in that directory. The name searched for is
18701 the name specified in the @code{#include} with @samp{.gch} appended. If
18702 the precompiled header file can't be used, it is ignored.
18704 For instance, if you have @code{#include "all.h"}, and you have
18705 @file{all.h.gch} in the same directory as @file{all.h}, then the
18706 precompiled header file will be used if possible, and the original
18707 header will be used otherwise.
18709 Alternatively, you might decide to put the precompiled header file in a
18710 directory and use @option{-I} to ensure that directory is searched
18711 before (or instead of) the directory containing the original header.
18712 Then, if you want to check that the precompiled header file is always
18713 used, you can put a file of the same name as the original header in this
18714 directory containing an @code{#error} command.
18716 This also works with @option{-include}. So yet another way to use
18717 precompiled headers, good for projects not designed with precompiled
18718 header files in mind, is to simply take most of the header files used by
18719 a project, include them from another header file, precompile that header
18720 file, and @option{-include} the precompiled header. If the header files
18721 have guards against multiple inclusion, they will be skipped because
18722 they've already been included (in the precompiled header).
18724 If you need to precompile the same header file for different
18725 languages, targets, or compiler options, you can instead make a
18726 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18727 header in the directory, perhaps using @option{-o}. It doesn't matter
18728 what you call the files in the directory, every precompiled header in
18729 the directory will be considered. The first precompiled header
18730 encountered in the directory that is valid for this compilation will
18731 be used; they're searched in no particular order.
18733 There are many other possibilities, limited only by your imagination,
18734 good sense, and the constraints of your build system.
18736 A precompiled header file can be used only when these conditions apply:
18740 Only one precompiled header can be used in a particular compilation.
18743 A precompiled header can't be used once the first C token is seen. You
18744 can have preprocessor directives before a precompiled header; you can
18745 even include a precompiled header from inside another header, so long as
18746 there are no C tokens before the @code{#include}.
18749 The precompiled header file must be produced for the same language as
18750 the current compilation. You can't use a C precompiled header for a C++
18754 The precompiled header file must have been produced by the same compiler
18755 binary as the current compilation is using.
18758 Any macros defined before the precompiled header is included must
18759 either be defined in the same way as when the precompiled header was
18760 generated, or must not affect the precompiled header, which usually
18761 means that they don't appear in the precompiled header at all.
18763 The @option{-D} option is one way to define a macro before a
18764 precompiled header is included; using a @code{#define} can also do it.
18765 There are also some options that define macros implicitly, like
18766 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18769 @item If debugging information is output when using the precompiled
18770 header, using @option{-g} or similar, the same kind of debugging information
18771 must have been output when building the precompiled header. However,
18772 a precompiled header built using @option{-g} can be used in a compilation
18773 when no debugging information is being output.
18775 @item The same @option{-m} options must generally be used when building
18776 and using the precompiled header. @xref{Submodel Options},
18777 for any cases where this rule is relaxed.
18779 @item Each of the following options must be the same when building and using
18780 the precompiled header:
18782 @gccoptlist{-fexceptions}
18785 Some other command-line options starting with @option{-f},
18786 @option{-p}, or @option{-O} must be defined in the same way as when
18787 the precompiled header was generated. At present, it's not clear
18788 which options are safe to change and which are not; the safest choice
18789 is to use exactly the same options when generating and using the
18790 precompiled header. The following are known to be safe:
18792 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18793 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18794 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18799 For all of these except the last, the compiler will automatically
18800 ignore the precompiled header if the conditions aren't met. If you
18801 find an option combination that doesn't work and doesn't cause the
18802 precompiled header to be ignored, please consider filing a bug report,
18805 If you do use differing options when generating and using the
18806 precompiled header, the actual behavior will be a mixture of the
18807 behavior for the options. For instance, if you use @option{-g} to
18808 generate the precompiled header but not when using it, you may or may
18809 not get debugging information for routines in the precompiled header.