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} -fallow-parameterless-variadic-functions @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
203 -Wsign-promo -Wzero-as-null-pointer-constant}
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++11-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 -Wvector-operation-performance -Wvla
275 -Wvolatile-register-var -Wwrite-strings}
277 @item C and Objective-C-only Warning Options
278 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
279 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
280 -Wold-style-declaration -Wold-style-definition @gol
281 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
282 -Wdeclaration-after-statement -Wpointer-sign}
284 @item Debugging Options
285 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
286 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
287 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
288 -fdisable-ipa-@var{pass_name} @gol
289 -fdisable-rtl-@var{pass_name} @gol
290 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
291 -fdisable-tree-@var{pass_name} @gol
292 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
293 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
294 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
295 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
296 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
298 -fdump-statistics @gol
300 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
304 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-nrv -fdump-tree-vect @gol
314 -fdump-tree-sink @gol
315 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
318 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
319 -ftree-vectorizer-verbose=@var{n} @gol
320 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
321 -fdump-final-insns=@var{file} @gol
322 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
323 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
324 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
325 -fenable-@var{kind}-@var{pass} @gol
326 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
327 -fdebug-types-section @gol
328 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
329 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
330 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
331 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
332 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
333 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
334 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
335 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
336 -gvms -gxcoff -gxcoff+ @gol
337 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
338 -fdebug-prefix-map=@var{old}=@var{new} @gol
339 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
340 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
341 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
342 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
343 -print-prog-name=@var{program} -print-search-dirs -Q @gol
344 -print-sysroot -print-sysroot-headers-suffix @gol
345 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
347 @item Optimization Options
348 @xref{Optimize Options,,Options that Control Optimization}.
349 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
350 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
351 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
352 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
353 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
354 -fcompare-elim -fcprop-registers -fcrossjumping @gol
355 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
356 -fcx-limited-range @gol
357 -fdata-sections -fdce -fdce -fdelayed-branch @gol
358 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
359 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
360 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
361 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
362 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
363 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
364 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
365 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
366 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
367 -fira-algorithm=@var{algorithm} @gol
368 -fira-region=@var{region} @gol
369 -fira-loop-pressure -fno-ira-share-save-slots @gol
370 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
371 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
372 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
373 -floop-parallelize-all -flto -flto-compression-level @gol
374 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
375 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
376 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
377 -fno-default-inline @gol
378 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
379 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
380 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
381 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
382 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
383 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
384 -fprefetch-loop-arrays @gol
385 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
386 -fprofile-generate=@var{path} @gol
387 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
388 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
389 -freorder-blocks-and-partition -freorder-functions @gol
390 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
391 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
392 -fsched-spec-load -fsched-spec-load-dangerous @gol
393 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
394 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
395 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
396 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
397 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
398 -fselective-scheduling -fselective-scheduling2 @gol
399 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
400 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
401 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
402 -fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol
403 -fthread-jumps -ftracer -ftree-bit-ccp @gol
404 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
405 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
406 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
407 -ftree-loop-if-convert-stores -ftree-loop-im @gol
408 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
409 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
410 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
411 -ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol
412 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
413 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
414 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
415 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
416 -fwhole-program -fwpa -fuse-linker-plugin @gol
417 --param @var{name}=@var{value}
418 -O -O0 -O1 -O2 -O3 -Os -Ofast}
420 @item Preprocessor Options
421 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
422 @gccoptlist{-A@var{question}=@var{answer} @gol
423 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
424 -C -dD -dI -dM -dN @gol
425 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
426 -idirafter @var{dir} @gol
427 -include @var{file} -imacros @var{file} @gol
428 -iprefix @var{file} -iwithprefix @var{dir} @gol
429 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
430 -imultilib @var{dir} -isysroot @var{dir} @gol
431 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
432 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
433 -remap -trigraphs -undef -U@var{macro} @gol
434 -Wp,@var{option} -Xpreprocessor @var{option}}
436 @item Assembler Option
437 @xref{Assembler Options,,Passing Options to the Assembler}.
438 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
441 @xref{Link Options,,Options for Linking}.
442 @gccoptlist{@var{object-file-name} -l@var{library} @gol
443 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
444 -s -static -static-libgcc -static-libstdc++ -shared @gol
445 -shared-libgcc -symbolic @gol
446 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
449 @item Directory Options
450 @xref{Directory Options,,Options for Directory Search}.
451 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
452 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
455 @item Machine Dependent Options
456 @xref{Submodel Options,,Hardware Models and Configurations}.
457 @c This list is ordered alphanumerically by subsection name.
458 @c Try and put the significant identifier (CPU or system) first,
459 @c so users have a clue at guessing where the ones they want will be.
461 @emph{Adapteva Epiphany Options}
462 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
463 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
464 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
465 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
466 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
467 -msplit-vecmove-early -m1reg-@var{reg}}
470 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
471 -mabi=@var{name} @gol
472 -mapcs-stack-check -mno-apcs-stack-check @gol
473 -mapcs-float -mno-apcs-float @gol
474 -mapcs-reentrant -mno-apcs-reentrant @gol
475 -msched-prolog -mno-sched-prolog @gol
476 -mlittle-endian -mbig-endian -mwords-little-endian @gol
477 -mfloat-abi=@var{name} -mfpe @gol
478 -mfp16-format=@var{name}
479 -mthumb-interwork -mno-thumb-interwork @gol
480 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
481 -mstructure-size-boundary=@var{n} @gol
482 -mabort-on-noreturn @gol
483 -mlong-calls -mno-long-calls @gol
484 -msingle-pic-base -mno-single-pic-base @gol
485 -mpic-register=@var{reg} @gol
486 -mnop-fun-dllimport @gol
487 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
488 -mpoke-function-name @gol
490 -mtpcs-frame -mtpcs-leaf-frame @gol
491 -mcaller-super-interworking -mcallee-super-interworking @gol
492 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
493 -mword-relocations @gol
494 -mfix-cortex-m3-ldrd}
497 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
498 -mcall-prologues -mtiny-stack -mint8 -mstrict-X}
500 @emph{Blackfin Options}
501 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
502 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
503 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
504 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
505 -mno-id-shared-library -mshared-library-id=@var{n} @gol
506 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
507 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
508 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
512 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
513 -msim -msdata=@var{sdata-type}}
516 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
517 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
518 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
519 -mstack-align -mdata-align -mconst-align @gol
520 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
521 -melf -maout -melinux -mlinux -sim -sim2 @gol
522 -mmul-bug-workaround -mno-mul-bug-workaround}
524 @emph{Darwin Options}
525 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
526 -arch_only -bind_at_load -bundle -bundle_loader @gol
527 -client_name -compatibility_version -current_version @gol
529 -dependency-file -dylib_file -dylinker_install_name @gol
530 -dynamic -dynamiclib -exported_symbols_list @gol
531 -filelist -flat_namespace -force_cpusubtype_ALL @gol
532 -force_flat_namespace -headerpad_max_install_names @gol
534 -image_base -init -install_name -keep_private_externs @gol
535 -multi_module -multiply_defined -multiply_defined_unused @gol
536 -noall_load -no_dead_strip_inits_and_terms @gol
537 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
538 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
539 -private_bundle -read_only_relocs -sectalign @gol
540 -sectobjectsymbols -whyload -seg1addr @gol
541 -sectcreate -sectobjectsymbols -sectorder @gol
542 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
543 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
544 -segprot -segs_read_only_addr -segs_read_write_addr @gol
545 -single_module -static -sub_library -sub_umbrella @gol
546 -twolevel_namespace -umbrella -undefined @gol
547 -unexported_symbols_list -weak_reference_mismatches @gol
548 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
549 -mkernel -mone-byte-bool}
551 @emph{DEC Alpha Options}
552 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
553 -mieee -mieee-with-inexact -mieee-conformant @gol
554 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
555 -mtrap-precision=@var{mode} -mbuild-constants @gol
556 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
557 -mbwx -mmax -mfix -mcix @gol
558 -mfloat-vax -mfloat-ieee @gol
559 -mexplicit-relocs -msmall-data -mlarge-data @gol
560 -msmall-text -mlarge-text @gol
561 -mmemory-latency=@var{time}}
563 @emph{DEC Alpha/VMS Options}
564 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
567 @gccoptlist{-msmall-model -mno-lsim}
570 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
571 -mhard-float -msoft-float @gol
572 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
573 -mdouble -mno-double @gol
574 -mmedia -mno-media -mmuladd -mno-muladd @gol
575 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
576 -mlinked-fp -mlong-calls -malign-labels @gol
577 -mlibrary-pic -macc-4 -macc-8 @gol
578 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
579 -moptimize-membar -mno-optimize-membar @gol
580 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
581 -mvliw-branch -mno-vliw-branch @gol
582 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
583 -mno-nested-cond-exec -mtomcat-stats @gol
587 @emph{GNU/Linux Options}
588 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
589 -tno-android-cc -tno-android-ld}
591 @emph{H8/300 Options}
592 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
595 @gccoptlist{-march=@var{architecture-type} @gol
596 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
597 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
598 -mfixed-range=@var{register-range} @gol
599 -mjump-in-delay -mlinker-opt -mlong-calls @gol
600 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
601 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
602 -mno-jump-in-delay -mno-long-load-store @gol
603 -mno-portable-runtime -mno-soft-float @gol
604 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
605 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
606 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
607 -munix=@var{unix-std} -nolibdld -static -threads}
609 @emph{i386 and x86-64 Options}
610 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
611 -mfpmath=@var{unit} @gol
612 -masm=@var{dialect} -mno-fancy-math-387 @gol
613 -mno-fp-ret-in-387 -msoft-float @gol
614 -mno-wide-multiply -mrtd -malign-double @gol
615 -mpreferred-stack-boundary=@var{num} @gol
616 -mincoming-stack-boundary=@var{num} @gol
617 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
618 -mrecip -mrecip=@var{opt} @gol
620 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
621 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
622 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
623 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
624 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
625 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
626 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
627 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
628 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
629 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
630 -mcmodel=@var{code-model} -mabi=@var{name} @gol
631 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
632 -msse2avx -mfentry -m8bit-idiv @gol
633 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
635 @emph{i386 and x86-64 Windows Options}
636 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
637 -mnop-fun-dllimport -mthread @gol
638 -municode -mwin32 -mwindows -fno-set-stack-executable}
641 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
642 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
643 -mconstant-gp -mauto-pic -mfused-madd @gol
644 -minline-float-divide-min-latency @gol
645 -minline-float-divide-max-throughput @gol
646 -mno-inline-float-divide @gol
647 -minline-int-divide-min-latency @gol
648 -minline-int-divide-max-throughput @gol
649 -mno-inline-int-divide @gol
650 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
651 -mno-inline-sqrt @gol
652 -mdwarf2-asm -mearly-stop-bits @gol
653 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
654 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
655 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
656 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
657 -msched-spec-ldc -msched-spec-control-ldc @gol
658 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
659 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
660 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
661 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
663 @emph{IA-64/VMS Options}
664 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
667 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
668 -msign-extend-enabled -muser-enabled}
670 @emph{M32R/D Options}
671 @gccoptlist{-m32r2 -m32rx -m32r @gol
673 -malign-loops -mno-align-loops @gol
674 -missue-rate=@var{number} @gol
675 -mbranch-cost=@var{number} @gol
676 -mmodel=@var{code-size-model-type} @gol
677 -msdata=@var{sdata-type} @gol
678 -mno-flush-func -mflush-func=@var{name} @gol
679 -mno-flush-trap -mflush-trap=@var{number} @gol
683 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
685 @emph{M680x0 Options}
686 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
687 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
688 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
689 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
690 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
691 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
692 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
693 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
697 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
698 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
699 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
700 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
701 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
704 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
705 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
706 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
707 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
710 @emph{MicroBlaze Options}
711 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
712 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
713 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
714 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
715 -mxl-mode-@var{app-model}}
718 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
719 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
720 -mips64 -mips64r2 @gol
721 -mips16 -mno-mips16 -mflip-mips16 @gol
722 -minterlink-mips16 -mno-interlink-mips16 @gol
723 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
724 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
725 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
726 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
727 -mfpu=@var{fpu-type} @gol
728 -msmartmips -mno-smartmips @gol
729 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
730 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
731 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
732 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
733 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
734 -membedded-data -mno-embedded-data @gol
735 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
736 -mcode-readable=@var{setting} @gol
737 -msplit-addresses -mno-split-addresses @gol
738 -mexplicit-relocs -mno-explicit-relocs @gol
739 -mcheck-zero-division -mno-check-zero-division @gol
740 -mdivide-traps -mdivide-breaks @gol
741 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
742 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
743 -mfix-24k -mno-fix-24k @gol
744 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
745 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
746 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
747 -mflush-func=@var{func} -mno-flush-func @gol
748 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
749 -mfp-exceptions -mno-fp-exceptions @gol
750 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
751 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
754 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
755 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
756 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
757 -mno-base-addresses -msingle-exit -mno-single-exit}
759 @emph{MN10300 Options}
760 @gccoptlist{-mmult-bug -mno-mult-bug @gol
761 -mno-am33 -mam33 -mam33-2 -mam34 @gol
762 -mtune=@var{cpu-type} @gol
763 -mreturn-pointer-on-d0 @gol
764 -mno-crt0 -mrelax -mliw -msetlb}
766 @emph{PDP-11 Options}
767 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
768 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
769 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
770 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
771 -mbranch-expensive -mbranch-cheap @gol
772 -munix-asm -mdec-asm}
774 @emph{picoChip Options}
775 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
776 -msymbol-as-address -mno-inefficient-warnings}
778 @emph{PowerPC Options}
779 See RS/6000 and PowerPC Options.
781 @emph{RS/6000 and PowerPC Options}
782 @gccoptlist{-mcpu=@var{cpu-type} @gol
783 -mtune=@var{cpu-type} @gol
784 -mcmodel=@var{code-model} @gol
785 -mpower -mno-power -mpower2 -mno-power2 @gol
786 -mpowerpc -mpowerpc64 -mno-powerpc @gol
787 -maltivec -mno-altivec @gol
788 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
789 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
790 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
791 -mfprnd -mno-fprnd @gol
792 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
793 -mnew-mnemonics -mold-mnemonics @gol
794 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
795 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
796 -malign-power -malign-natural @gol
797 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
798 -msingle-float -mdouble-float -msimple-fpu @gol
799 -mstring -mno-string -mupdate -mno-update @gol
800 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
801 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
802 -mstrict-align -mno-strict-align -mrelocatable @gol
803 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
804 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
805 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
806 -mprioritize-restricted-insns=@var{priority} @gol
807 -msched-costly-dep=@var{dependence_type} @gol
808 -minsert-sched-nops=@var{scheme} @gol
809 -mcall-sysv -mcall-netbsd @gol
810 -maix-struct-return -msvr4-struct-return @gol
811 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
812 -mblock-move-inline-limit=@var{num} @gol
813 -misel -mno-isel @gol
814 -misel=yes -misel=no @gol
816 -mspe=yes -mspe=no @gol
818 -mgen-cell-microcode -mwarn-cell-microcode @gol
819 -mvrsave -mno-vrsave @gol
820 -mmulhw -mno-mulhw @gol
821 -mdlmzb -mno-dlmzb @gol
822 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
823 -mprototype -mno-prototype @gol
824 -msim -mmvme -mads -myellowknife -memb -msdata @gol
825 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
826 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
827 -mno-recip-precision @gol
828 -mveclibabi=@var{type} -mfriz -mno-friz @gol
829 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
830 -msave-toc-indirect -mno-save-toc-indirect}
833 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
835 -mbig-endian-data -mlittle-endian-data @gol
838 -mas100-syntax -mno-as100-syntax@gol
840 -mmax-constant-size=@gol
843 -msave-acc-in-interrupts}
845 @emph{S/390 and zSeries Options}
846 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
847 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
848 -mlong-double-64 -mlong-double-128 @gol
849 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
850 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
851 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
852 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
853 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
856 @gccoptlist{-meb -mel @gol
860 -mscore5 -mscore5u -mscore7 -mscore7d}
863 @gccoptlist{-m1 -m2 -m2e @gol
864 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
866 -m4-nofpu -m4-single-only -m4-single -m4 @gol
867 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
868 -m5-64media -m5-64media-nofpu @gol
869 -m5-32media -m5-32media-nofpu @gol
870 -m5-compact -m5-compact-nofpu @gol
871 -mb -ml -mdalign -mrelax @gol
872 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
873 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
874 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
875 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
876 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
877 -maccumulate-outgoing-args -minvalid-symbols}
879 @emph{Solaris 2 Options}
880 @gccoptlist{-mimpure-text -mno-impure-text @gol
884 @gccoptlist{-mcpu=@var{cpu-type} @gol
885 -mtune=@var{cpu-type} @gol
886 -mcmodel=@var{code-model} @gol
887 -m32 -m64 -mapp-regs -mno-app-regs @gol
888 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
889 -mfpu -mno-fpu -mhard-float -msoft-float @gol
890 -mhard-quad-float -msoft-quad-float @gol
892 -mstack-bias -mno-stack-bias @gol
893 -munaligned-doubles -mno-unaligned-doubles @gol
894 -mv8plus -mno-v8plus -mvis -mno-vis @gol
895 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
896 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
900 @gccoptlist{-mwarn-reloc -merror-reloc @gol
901 -msafe-dma -munsafe-dma @gol
903 -msmall-mem -mlarge-mem -mstdmain @gol
904 -mfixed-range=@var{register-range} @gol
906 -maddress-space-conversion -mno-address-space-conversion @gol
907 -mcache-size=@var{cache-size} @gol
908 -matomic-updates -mno-atomic-updates}
910 @emph{System V Options}
911 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
914 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
915 -mprolog-function -mno-prolog-function -mspace @gol
916 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
917 -mapp-regs -mno-app-regs @gol
918 -mdisable-callt -mno-disable-callt @gol
921 -mv850e1 -mv850es @gol
926 @gccoptlist{-mg -mgnu -munix}
928 @emph{VxWorks Options}
929 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
930 -Xbind-lazy -Xbind-now}
932 @emph{x86-64 Options}
933 See i386 and x86-64 Options.
935 @emph{Xstormy16 Options}
938 @emph{Xtensa Options}
939 @gccoptlist{-mconst16 -mno-const16 @gol
940 -mfused-madd -mno-fused-madd @gol
942 -mserialize-volatile -mno-serialize-volatile @gol
943 -mtext-section-literals -mno-text-section-literals @gol
944 -mtarget-align -mno-target-align @gol
945 -mlongcalls -mno-longcalls}
947 @emph{zSeries Options}
948 See S/390 and zSeries Options.
950 @item Code Generation Options
951 @xref{Code Gen Options,,Options for Code Generation Conventions}.
952 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
953 -ffixed-@var{reg} -fexceptions @gol
954 -fnon-call-exceptions -funwind-tables @gol
955 -fasynchronous-unwind-tables @gol
956 -finhibit-size-directive -finstrument-functions @gol
957 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
958 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
959 -fno-common -fno-ident @gol
960 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
961 -fno-jump-tables @gol
962 -frecord-gcc-switches @gol
963 -freg-struct-return -fshort-enums @gol
964 -fshort-double -fshort-wchar @gol
965 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
966 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
967 -fno-stack-limit -fsplit-stack @gol
968 -fleading-underscore -ftls-model=@var{model} @gol
969 -ftrapv -fwrapv -fbounds-check @gol
970 -fvisibility -fstrict-volatile-bitfields}
974 * Overall Options:: Controlling the kind of output:
975 an executable, object files, assembler files,
976 or preprocessed source.
977 * C Dialect Options:: Controlling the variant of C language compiled.
978 * C++ Dialect Options:: Variations on C++.
979 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
981 * Language Independent Options:: Controlling how diagnostics should be
983 * Warning Options:: How picky should the compiler be?
984 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
985 * Optimize Options:: How much optimization?
986 * Preprocessor Options:: Controlling header files and macro definitions.
987 Also, getting dependency information for Make.
988 * Assembler Options:: Passing options to the assembler.
989 * Link Options:: Specifying libraries and so on.
990 * Directory Options:: Where to find header files and libraries.
991 Where to find the compiler executable files.
992 * Spec Files:: How to pass switches to sub-processes.
993 * Target Options:: Running a cross-compiler, or an old version of GCC.
996 @node Overall Options
997 @section Options Controlling the Kind of Output
999 Compilation can involve up to four stages: preprocessing, compilation
1000 proper, assembly and linking, always in that order. GCC is capable of
1001 preprocessing and compiling several files either into several
1002 assembler input files, or into one assembler input file; then each
1003 assembler input file produces an object file, and linking combines all
1004 the object files (those newly compiled, and those specified as input)
1005 into an executable file.
1007 @cindex file name suffix
1008 For any given input file, the file name suffix determines what kind of
1009 compilation is done:
1013 C source code which must be preprocessed.
1016 C source code which should not be preprocessed.
1019 C++ source code which should not be preprocessed.
1022 Objective-C source code. Note that you must link with the @file{libobjc}
1023 library to make an Objective-C program work.
1026 Objective-C source code which should not be preprocessed.
1030 Objective-C++ source code. Note that you must link with the @file{libobjc}
1031 library to make an Objective-C++ program work. Note that @samp{.M} refers
1032 to a literal capital M@.
1034 @item @var{file}.mii
1035 Objective-C++ source code which should not be preprocessed.
1038 C, C++, Objective-C or Objective-C++ header file to be turned into a
1039 precompiled header (default), or C, C++ header file to be turned into an
1040 Ada spec (via the @option{-fdump-ada-spec} switch).
1043 @itemx @var{file}.cp
1044 @itemx @var{file}.cxx
1045 @itemx @var{file}.cpp
1046 @itemx @var{file}.CPP
1047 @itemx @var{file}.c++
1049 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1050 the last two letters must both be literally @samp{x}. Likewise,
1051 @samp{.C} refers to a literal capital C@.
1055 Objective-C++ source code which must be preprocessed.
1057 @item @var{file}.mii
1058 Objective-C++ source code which should not be preprocessed.
1062 @itemx @var{file}.hp
1063 @itemx @var{file}.hxx
1064 @itemx @var{file}.hpp
1065 @itemx @var{file}.HPP
1066 @itemx @var{file}.h++
1067 @itemx @var{file}.tcc
1068 C++ header file to be turned into a precompiled header or Ada spec.
1071 @itemx @var{file}.for
1072 @itemx @var{file}.ftn
1073 Fixed form Fortran source code which should not be preprocessed.
1076 @itemx @var{file}.FOR
1077 @itemx @var{file}.fpp
1078 @itemx @var{file}.FPP
1079 @itemx @var{file}.FTN
1080 Fixed form Fortran source code which must be preprocessed (with the traditional
1083 @item @var{file}.f90
1084 @itemx @var{file}.f95
1085 @itemx @var{file}.f03
1086 @itemx @var{file}.f08
1087 Free form Fortran source code which should not be preprocessed.
1089 @item @var{file}.F90
1090 @itemx @var{file}.F95
1091 @itemx @var{file}.F03
1092 @itemx @var{file}.F08
1093 Free form Fortran source code which must be preprocessed (with the
1094 traditional preprocessor).
1099 @c FIXME: Descriptions of Java file types.
1105 @item @var{file}.ads
1106 Ada source code file which contains a library unit declaration (a
1107 declaration of a package, subprogram, or generic, or a generic
1108 instantiation), or a library unit renaming declaration (a package,
1109 generic, or subprogram renaming declaration). Such files are also
1112 @item @var{file}.adb
1113 Ada source code file containing a library unit body (a subprogram or
1114 package body). Such files are also called @dfn{bodies}.
1116 @c GCC also knows about some suffixes for languages not yet included:
1127 @itemx @var{file}.sx
1128 Assembler code which must be preprocessed.
1131 An object file to be fed straight into linking.
1132 Any file name with no recognized suffix is treated this way.
1136 You can specify the input language explicitly with the @option{-x} option:
1139 @item -x @var{language}
1140 Specify explicitly the @var{language} for the following input files
1141 (rather than letting the compiler choose a default based on the file
1142 name suffix). This option applies to all following input files until
1143 the next @option{-x} option. Possible values for @var{language} are:
1145 c c-header cpp-output
1146 c++ c++-header c++-cpp-output
1147 objective-c objective-c-header objective-c-cpp-output
1148 objective-c++ objective-c++-header objective-c++-cpp-output
1149 assembler assembler-with-cpp
1151 f77 f77-cpp-input f95 f95-cpp-input
1157 Turn off any specification of a language, so that subsequent files are
1158 handled according to their file name suffixes (as they are if @option{-x}
1159 has not been used at all).
1161 @item -pass-exit-codes
1162 @opindex pass-exit-codes
1163 Normally the @command{gcc} program will exit with the code of 1 if any
1164 phase of the compiler returns a non-success return code. If you specify
1165 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1166 numerically highest error produced by any phase that returned an error
1167 indication. The C, C++, and Fortran frontends return 4, if an internal
1168 compiler error is encountered.
1171 If you only want some of the stages of compilation, you can use
1172 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1173 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1174 @command{gcc} is to stop. Note that some combinations (for example,
1175 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1180 Compile or assemble the source files, but do not link. The linking
1181 stage simply is not done. The ultimate output is in the form of an
1182 object file for each source file.
1184 By default, the object file name for a source file is made by replacing
1185 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1187 Unrecognized input files, not requiring compilation or assembly, are
1192 Stop after the stage of compilation proper; do not assemble. The output
1193 is in the form of an assembler code file for each non-assembler input
1196 By default, the assembler file name for a source file is made by
1197 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1199 Input files that don't require compilation are ignored.
1203 Stop after the preprocessing stage; do not run the compiler proper. The
1204 output is in the form of preprocessed source code, which is sent to the
1207 Input files which don't require preprocessing are ignored.
1209 @cindex output file option
1212 Place output in file @var{file}. This applies regardless to whatever
1213 sort of output is being produced, whether it be an executable file,
1214 an object file, an assembler file or preprocessed C code.
1216 If @option{-o} is not specified, the default is to put an executable
1217 file in @file{a.out}, the object file for
1218 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1219 assembler file in @file{@var{source}.s}, a precompiled header file in
1220 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1225 Print (on standard error output) the commands executed to run the stages
1226 of compilation. Also print the version number of the compiler driver
1227 program and of the preprocessor and the compiler proper.
1231 Like @option{-v} except the commands are not executed and arguments
1232 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1233 This is useful for shell scripts to capture the driver-generated command lines.
1237 Use pipes rather than temporary files for communication between the
1238 various stages of compilation. This fails to work on some systems where
1239 the assembler is unable to read from a pipe; but the GNU assembler has
1244 Print (on the standard output) a description of the command line options
1245 understood by @command{gcc}. If the @option{-v} option is also specified
1246 then @option{--help} will also be passed on to the various processes
1247 invoked by @command{gcc}, so that they can display the command line options
1248 they accept. If the @option{-Wextra} option has also been specified
1249 (prior to the @option{--help} option), then command line options which
1250 have no documentation associated with them will also be displayed.
1253 @opindex target-help
1254 Print (on the standard output) a description of target-specific command
1255 line options for each tool. For some targets extra target-specific
1256 information may also be printed.
1258 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1259 Print (on the standard output) a description of the command line
1260 options understood by the compiler that fit into all specified classes
1261 and qualifiers. These are the supported classes:
1264 @item @samp{optimizers}
1265 This will display all of the optimization options supported by the
1268 @item @samp{warnings}
1269 This will display all of the options controlling warning messages
1270 produced by the compiler.
1273 This will display target-specific options. Unlike the
1274 @option{--target-help} option however, target-specific options of the
1275 linker and assembler will not be displayed. This is because those
1276 tools do not currently support the extended @option{--help=} syntax.
1279 This will display the values recognized by the @option{--param}
1282 @item @var{language}
1283 This will display the options supported for @var{language}, where
1284 @var{language} is the name of one of the languages supported in this
1288 This will display the options that are common to all languages.
1291 These are the supported qualifiers:
1294 @item @samp{undocumented}
1295 Display only those options which are undocumented.
1298 Display options which take an argument that appears after an equal
1299 sign in the same continuous piece of text, such as:
1300 @samp{--help=target}.
1302 @item @samp{separate}
1303 Display options which take an argument that appears as a separate word
1304 following the original option, such as: @samp{-o output-file}.
1307 Thus for example to display all the undocumented target-specific
1308 switches supported by the compiler the following can be used:
1311 --help=target,undocumented
1314 The sense of a qualifier can be inverted by prefixing it with the
1315 @samp{^} character, so for example to display all binary warning
1316 options (i.e., ones that are either on or off and that do not take an
1317 argument), which have a description the following can be used:
1320 --help=warnings,^joined,^undocumented
1323 The argument to @option{--help=} should not consist solely of inverted
1326 Combining several classes is possible, although this usually
1327 restricts the output by so much that there is nothing to display. One
1328 case where it does work however is when one of the classes is
1329 @var{target}. So for example to display all the target-specific
1330 optimization options the following can be used:
1333 --help=target,optimizers
1336 The @option{--help=} option can be repeated on the command line. Each
1337 successive use will display its requested class of options, skipping
1338 those that have already been displayed.
1340 If the @option{-Q} option appears on the command line before the
1341 @option{--help=} option, then the descriptive text displayed by
1342 @option{--help=} is changed. Instead of describing the displayed
1343 options, an indication is given as to whether the option is enabled,
1344 disabled or set to a specific value (assuming that the compiler
1345 knows this at the point where the @option{--help=} option is used).
1347 Here is a truncated example from the ARM port of @command{gcc}:
1350 % gcc -Q -mabi=2 --help=target -c
1351 The following options are target specific:
1353 -mabort-on-noreturn [disabled]
1357 The output is sensitive to the effects of previous command line
1358 options, so for example it is possible to find out which optimizations
1359 are enabled at @option{-O2} by using:
1362 -Q -O2 --help=optimizers
1365 Alternatively you can discover which binary optimizations are enabled
1366 by @option{-O3} by using:
1369 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1370 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1371 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1374 @item -no-canonical-prefixes
1375 @opindex no-canonical-prefixes
1376 Do not expand any symbolic links, resolve references to @samp{/../}
1377 or @samp{/./}, or make the path absolute when generating a relative
1382 Display the version number and copyrights of the invoked GCC@.
1386 Invoke all subcommands under a wrapper program. The name of the
1387 wrapper program and its parameters are passed as a comma separated
1391 gcc -c t.c -wrapper gdb,--args
1394 This will invoke all subprograms of @command{gcc} under
1395 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1396 @samp{gdb --args cc1 @dots{}}.
1398 @item -fplugin=@var{name}.so
1399 Load the plugin code in file @var{name}.so, assumed to be a
1400 shared object to be dlopen'd by the compiler. The base name of
1401 the shared object file is used to identify the plugin for the
1402 purposes of argument parsing (See
1403 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1404 Each plugin should define the callback functions specified in the
1407 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1408 Define an argument called @var{key} with a value of @var{value}
1409 for the plugin called @var{name}.
1411 @item -fdump-ada-spec@r{[}-slim@r{]}
1412 For C and C++ source and include files, generate corresponding Ada
1413 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1414 GNAT User's Guide}, which provides detailed documentation on this feature.
1416 @item -fdump-go-spec=@var{file}
1417 For input files in any language, generate corresponding Go
1418 declarations in @var{file}. This generates Go @code{const},
1419 @code{type}, @code{var}, and @code{func} declarations which may be a
1420 useful way to start writing a Go interface to code written in some
1423 @include @value{srcdir}/../libiberty/at-file.texi
1427 @section Compiling C++ Programs
1429 @cindex suffixes for C++ source
1430 @cindex C++ source file suffixes
1431 C++ source files conventionally use one of the suffixes @samp{.C},
1432 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1433 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1434 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1435 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1436 files with these names and compiles them as C++ programs even if you
1437 call the compiler the same way as for compiling C programs (usually
1438 with the name @command{gcc}).
1442 However, the use of @command{gcc} does not add the C++ library.
1443 @command{g++} is a program that calls GCC and treats @samp{.c},
1444 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1445 files unless @option{-x} is used, and automatically specifies linking
1446 against the C++ library. This program is also useful when
1447 precompiling a C header file with a @samp{.h} extension for use in C++
1448 compilations. On many systems, @command{g++} is also installed with
1449 the name @command{c++}.
1451 @cindex invoking @command{g++}
1452 When you compile C++ programs, you may specify many of the same
1453 command-line options that you use for compiling programs in any
1454 language; or command-line options meaningful for C and related
1455 languages; or options that are meaningful only for C++ programs.
1456 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1457 explanations of options for languages related to C@.
1458 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1459 explanations of options that are meaningful only for C++ programs.
1461 @node C Dialect Options
1462 @section Options Controlling C Dialect
1463 @cindex dialect options
1464 @cindex language dialect options
1465 @cindex options, dialect
1467 The following options control the dialect of C (or languages derived
1468 from C, such as C++, Objective-C and Objective-C++) that the compiler
1472 @cindex ANSI support
1476 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1477 equivalent to @samp{-std=c++98}.
1479 This turns off certain features of GCC that are incompatible with ISO
1480 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1481 such as the @code{asm} and @code{typeof} keywords, and
1482 predefined macros such as @code{unix} and @code{vax} that identify the
1483 type of system you are using. It also enables the undesirable and
1484 rarely used ISO trigraph feature. For the C compiler,
1485 it disables recognition of C++ style @samp{//} comments as well as
1486 the @code{inline} keyword.
1488 The alternate keywords @code{__asm__}, @code{__extension__},
1489 @code{__inline__} and @code{__typeof__} continue to work despite
1490 @option{-ansi}. You would not want to use them in an ISO C program, of
1491 course, but it is useful to put them in header files that might be included
1492 in compilations done with @option{-ansi}. Alternate predefined macros
1493 such as @code{__unix__} and @code{__vax__} are also available, with or
1494 without @option{-ansi}.
1496 The @option{-ansi} option does not cause non-ISO programs to be
1497 rejected gratuitously. For that, @option{-pedantic} is required in
1498 addition to @option{-ansi}. @xref{Warning Options}.
1500 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1501 option is used. Some header files may notice this macro and refrain
1502 from declaring certain functions or defining certain macros that the
1503 ISO standard doesn't call for; this is to avoid interfering with any
1504 programs that might use these names for other things.
1506 Functions that would normally be built in but do not have semantics
1507 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1508 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1509 built-in functions provided by GCC}, for details of the functions
1514 Determine the language standard. @xref{Standards,,Language Standards
1515 Supported by GCC}, for details of these standard versions. This option
1516 is currently only supported when compiling C or C++.
1518 The compiler can accept several base standards, such as @samp{c90} or
1519 @samp{c++98}, and GNU dialects of those standards, such as
1520 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1521 compiler will accept all programs following that standard and those
1522 using GNU extensions that do not contradict it. For example,
1523 @samp{-std=c90} turns off certain features of GCC that are
1524 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1525 keywords, but not other GNU extensions that do not have a meaning in
1526 ISO C90, such as omitting the middle term of a @code{?:}
1527 expression. On the other hand, by specifying a GNU dialect of a
1528 standard, all features the compiler support are enabled, even when
1529 those features change the meaning of the base standard and some
1530 strict-conforming programs may be rejected. The particular standard
1531 is used by @option{-pedantic} to identify which features are GNU
1532 extensions given that version of the standard. For example
1533 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1534 comments, while @samp{-std=gnu99 -pedantic} would not.
1536 A value for this option must be provided; possible values are
1542 Support all ISO C90 programs (certain GNU extensions that conflict
1543 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1545 @item iso9899:199409
1546 ISO C90 as modified in amendment 1.
1552 ISO C99. Note that this standard is not yet fully supported; see
1553 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1554 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1557 ISO C1X, the draft of the next revision of the ISO C standard.
1558 Support is limited and experimental and features enabled by this
1559 option may be changed or removed if changed in or removed from the
1564 GNU dialect of ISO C90 (including some C99 features). This
1565 is the default for C code.
1569 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1570 this will become the default. The name @samp{gnu9x} is deprecated.
1573 GNU dialect of ISO C1X. Support is limited and experimental and
1574 features enabled by this option may be changed or removed if changed
1575 in or removed from the standard draft.
1578 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1582 GNU dialect of @option{-std=c++98}. This is the default for
1586 The 2011 ISO C++ standard plus amendments. Support for C++11 is still
1587 experimental, and may change in incompatible ways in future releases.
1590 GNU dialect of @option{-std=c++11}. Support for C++11 is still
1591 experimental, and may change in incompatible ways in future releases.
1594 @item -fgnu89-inline
1595 @opindex fgnu89-inline
1596 The option @option{-fgnu89-inline} tells GCC to use the traditional
1597 GNU semantics for @code{inline} functions when in C99 mode.
1598 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1599 is accepted and ignored by GCC versions 4.1.3 up to but not including
1600 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1601 C99 mode. Using this option is roughly equivalent to adding the
1602 @code{gnu_inline} function attribute to all inline functions
1603 (@pxref{Function Attributes}).
1605 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1606 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1607 specifies the default behavior). This option was first supported in
1608 GCC 4.3. This option is not supported in @option{-std=c90} or
1609 @option{-std=gnu90} mode.
1611 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1612 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1613 in effect for @code{inline} functions. @xref{Common Predefined
1614 Macros,,,cpp,The C Preprocessor}.
1616 @item -aux-info @var{filename}
1618 Output to the given filename prototyped declarations for all functions
1619 declared and/or defined in a translation unit, including those in header
1620 files. This option is silently ignored in any language other than C@.
1622 Besides declarations, the file indicates, in comments, the origin of
1623 each declaration (source file and line), whether the declaration was
1624 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1625 @samp{O} for old, respectively, in the first character after the line
1626 number and the colon), and whether it came from a declaration or a
1627 definition (@samp{C} or @samp{F}, respectively, in the following
1628 character). In the case of function definitions, a K&R-style list of
1629 arguments followed by their declarations is also provided, inside
1630 comments, after the declaration.
1632 @item -fallow-parameterless-variadic-functions
1633 Accept variadic functions without named parameters.
1635 Although it is possible to define such a function, this is not very
1636 useful as it is not possible to read the arguments. This is only
1637 supported for C as this construct is allowed by C++.
1641 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1642 keyword, so that code can use these words as identifiers. You can use
1643 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1644 instead. @option{-ansi} implies @option{-fno-asm}.
1646 In C++, this switch only affects the @code{typeof} keyword, since
1647 @code{asm} and @code{inline} are standard keywords. You may want to
1648 use the @option{-fno-gnu-keywords} flag instead, which has the same
1649 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1650 switch only affects the @code{asm} and @code{typeof} keywords, since
1651 @code{inline} is a standard keyword in ISO C99.
1654 @itemx -fno-builtin-@var{function}
1655 @opindex fno-builtin
1656 @cindex built-in functions
1657 Don't recognize built-in functions that do not begin with
1658 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1659 functions provided by GCC}, for details of the functions affected,
1660 including those which are not built-in functions when @option{-ansi} or
1661 @option{-std} options for strict ISO C conformance are used because they
1662 do not have an ISO standard meaning.
1664 GCC normally generates special code to handle certain built-in functions
1665 more efficiently; for instance, calls to @code{alloca} may become single
1666 instructions that adjust the stack directly, and calls to @code{memcpy}
1667 may become inline copy loops. The resulting code is often both smaller
1668 and faster, but since the function calls no longer appear as such, you
1669 cannot set a breakpoint on those calls, nor can you change the behavior
1670 of the functions by linking with a different library. In addition,
1671 when a function is recognized as a built-in function, GCC may use
1672 information about that function to warn about problems with calls to
1673 that function, or to generate more efficient code, even if the
1674 resulting code still contains calls to that function. For example,
1675 warnings are given with @option{-Wformat} for bad calls to
1676 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1677 known not to modify global memory.
1679 With the @option{-fno-builtin-@var{function}} option
1680 only the built-in function @var{function} is
1681 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1682 function is named that is not built-in in this version of GCC, this
1683 option is ignored. There is no corresponding
1684 @option{-fbuiltin-@var{function}} option; if you wish to enable
1685 built-in functions selectively when using @option{-fno-builtin} or
1686 @option{-ffreestanding}, you may define macros such as:
1689 #define abs(n) __builtin_abs ((n))
1690 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1695 @cindex hosted environment
1697 Assert that compilation takes place in a hosted environment. This implies
1698 @option{-fbuiltin}. A hosted environment is one in which the
1699 entire standard library is available, and in which @code{main} has a return
1700 type of @code{int}. Examples are nearly everything except a kernel.
1701 This is equivalent to @option{-fno-freestanding}.
1703 @item -ffreestanding
1704 @opindex ffreestanding
1705 @cindex hosted environment
1707 Assert that compilation takes place in a freestanding environment. This
1708 implies @option{-fno-builtin}. A freestanding environment
1709 is one in which the standard library may not exist, and program startup may
1710 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1711 This is equivalent to @option{-fno-hosted}.
1713 @xref{Standards,,Language Standards Supported by GCC}, for details of
1714 freestanding and hosted environments.
1718 @cindex OpenMP parallel
1719 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1720 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1721 compiler generates parallel code according to the OpenMP Application
1722 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1723 implies @option{-pthread}, and thus is only supported on targets that
1724 have support for @option{-pthread}.
1728 When the option @option{-fgnu-tm} is specified, the compiler will
1729 generate code for the Linux variant of Intel's current Transactional
1730 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1731 an experimental feature whose interface may change in future versions
1732 of GCC, as the official specification changes. Please note that not
1733 all architectures are supported for this feature.
1735 For more information on GCC's support for transactional memory,
1736 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1737 Transactional Memory Library}.
1739 @item -fms-extensions
1740 @opindex fms-extensions
1741 Accept some non-standard constructs used in Microsoft header files.
1743 In C++ code, this allows member names in structures to be similar
1744 to previous types declarations.
1753 Some cases of unnamed fields in structures and unions are only
1754 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1755 fields within structs/unions}, for details.
1757 @item -fplan9-extensions
1758 Accept some non-standard constructs used in Plan 9 code.
1760 This enables @option{-fms-extensions}, permits passing pointers to
1761 structures with anonymous fields to functions which expect pointers to
1762 elements of the type of the field, and permits referring to anonymous
1763 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1764 struct/union fields within structs/unions}, for details. This is only
1765 supported for C, not C++.
1769 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1770 options for strict ISO C conformance) implies @option{-trigraphs}.
1772 @item -no-integrated-cpp
1773 @opindex no-integrated-cpp
1774 Performs a compilation in two passes: preprocessing and compiling. This
1775 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1776 @option{-B} option. The user supplied compilation step can then add in
1777 an additional preprocessing step after normal preprocessing but before
1778 compiling. The default is to use the integrated cpp (internal cpp)
1780 The semantics of this option will change if "cc1", "cc1plus", and
1781 "cc1obj" are merged.
1783 @cindex traditional C language
1784 @cindex C language, traditional
1786 @itemx -traditional-cpp
1787 @opindex traditional-cpp
1788 @opindex traditional
1789 Formerly, these options caused GCC to attempt to emulate a pre-standard
1790 C compiler. They are now only supported with the @option{-E} switch.
1791 The preprocessor continues to support a pre-standard mode. See the GNU
1792 CPP manual for details.
1794 @item -fcond-mismatch
1795 @opindex fcond-mismatch
1796 Allow conditional expressions with mismatched types in the second and
1797 third arguments. The value of such an expression is void. This option
1798 is not supported for C++.
1800 @item -flax-vector-conversions
1801 @opindex flax-vector-conversions
1802 Allow implicit conversions between vectors with differing numbers of
1803 elements and/or incompatible element types. This option should not be
1806 @item -funsigned-char
1807 @opindex funsigned-char
1808 Let the type @code{char} be unsigned, like @code{unsigned char}.
1810 Each kind of machine has a default for what @code{char} should
1811 be. It is either like @code{unsigned char} by default or like
1812 @code{signed char} by default.
1814 Ideally, a portable program should always use @code{signed char} or
1815 @code{unsigned char} when it depends on the signedness of an object.
1816 But many programs have been written to use plain @code{char} and
1817 expect it to be signed, or expect it to be unsigned, depending on the
1818 machines they were written for. This option, and its inverse, let you
1819 make such a program work with the opposite default.
1821 The type @code{char} is always a distinct type from each of
1822 @code{signed char} or @code{unsigned char}, even though its behavior
1823 is always just like one of those two.
1826 @opindex fsigned-char
1827 Let the type @code{char} be signed, like @code{signed char}.
1829 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1830 the negative form of @option{-funsigned-char}. Likewise, the option
1831 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1833 @item -fsigned-bitfields
1834 @itemx -funsigned-bitfields
1835 @itemx -fno-signed-bitfields
1836 @itemx -fno-unsigned-bitfields
1837 @opindex fsigned-bitfields
1838 @opindex funsigned-bitfields
1839 @opindex fno-signed-bitfields
1840 @opindex fno-unsigned-bitfields
1841 These options control whether a bit-field is signed or unsigned, when the
1842 declaration does not use either @code{signed} or @code{unsigned}. By
1843 default, such a bit-field is signed, because this is consistent: the
1844 basic integer types such as @code{int} are signed types.
1847 @node C++ Dialect Options
1848 @section Options Controlling C++ Dialect
1850 @cindex compiler options, C++
1851 @cindex C++ options, command line
1852 @cindex options, C++
1853 This section describes the command-line options that are only meaningful
1854 for C++ programs; but you can also use most of the GNU compiler options
1855 regardless of what language your program is in. For example, you
1856 might compile a file @code{firstClass.C} like this:
1859 g++ -g -frepo -O -c firstClass.C
1863 In this example, only @option{-frepo} is an option meant
1864 only for C++ programs; you can use the other options with any
1865 language supported by GCC@.
1867 Here is a list of options that are @emph{only} for compiling C++ programs:
1871 @item -fabi-version=@var{n}
1872 @opindex fabi-version
1873 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1874 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1875 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1876 the version that conforms most closely to the C++ ABI specification.
1877 Therefore, the ABI obtained using version 0 will change as ABI bugs
1880 The default is version 2.
1882 Version 3 corrects an error in mangling a constant address as a
1885 Version 4 implements a standard mangling for vector types.
1887 Version 5 corrects the mangling of attribute const/volatile on
1888 function pointer types, decltype of a plain decl, and use of a
1889 function parameter in the declaration of another parameter.
1891 Version 6 corrects the promotion behavior of C++11 scoped enums.
1893 See also @option{-Wabi}.
1895 @item -fno-access-control
1896 @opindex fno-access-control
1897 Turn off all access checking. This switch is mainly useful for working
1898 around bugs in the access control code.
1902 Check that the pointer returned by @code{operator new} is non-null
1903 before attempting to modify the storage allocated. This check is
1904 normally unnecessary because the C++ standard specifies that
1905 @code{operator new} will only return @code{0} if it is declared
1906 @samp{throw()}, in which case the compiler will always check the
1907 return value even without this option. In all other cases, when
1908 @code{operator new} has a non-empty exception specification, memory
1909 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1910 @samp{new (nothrow)}.
1912 @item -fconserve-space
1913 @opindex fconserve-space
1914 Put uninitialized or runtime-initialized global variables into the
1915 common segment, as C does. This saves space in the executable at the
1916 cost of not diagnosing duplicate definitions. If you compile with this
1917 flag and your program mysteriously crashes after @code{main()} has
1918 completed, you may have an object that is being destroyed twice because
1919 two definitions were merged.
1921 This option is no longer useful on most targets, now that support has
1922 been added for putting variables into BSS without making them common.
1924 @item -fconstexpr-depth=@var{n}
1925 @opindex fconstexpr-depth
1926 Set the maximum nested evaluation depth for C++11 constexpr functions
1927 to @var{n}. A limit is needed to detect endless recursion during
1928 constant expression evaluation. The minimum specified by the standard
1931 @item -fdeduce-init-list
1932 @opindex fdeduce-init-list
1933 Enable deduction of a template type parameter as
1934 std::initializer_list from a brace-enclosed initializer list, i.e.
1937 template <class T> auto forward(T t) -> decltype (realfn (t))
1944 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1948 This deduction was implemented as a possible extension to the
1949 originally proposed semantics for the C++11 standard, but was not part
1950 of the final standard, so it is disabled by default. This option is
1951 deprecated, and may be removed in a future version of G++.
1953 @item -ffriend-injection
1954 @opindex ffriend-injection
1955 Inject friend functions into the enclosing namespace, so that they are
1956 visible outside the scope of the class in which they are declared.
1957 Friend functions were documented to work this way in the old Annotated
1958 C++ Reference Manual, and versions of G++ before 4.1 always worked
1959 that way. However, in ISO C++ a friend function which is not declared
1960 in an enclosing scope can only be found using argument dependent
1961 lookup. This option causes friends to be injected as they were in
1964 This option is for compatibility, and may be removed in a future
1967 @item -fno-elide-constructors
1968 @opindex fno-elide-constructors
1969 The C++ standard allows an implementation to omit creating a temporary
1970 which is only used to initialize another object of the same type.
1971 Specifying this option disables that optimization, and forces G++ to
1972 call the copy constructor in all cases.
1974 @item -fno-enforce-eh-specs
1975 @opindex fno-enforce-eh-specs
1976 Don't generate code to check for violation of exception specifications
1977 at runtime. This option violates the C++ standard, but may be useful
1978 for reducing code size in production builds, much like defining
1979 @samp{NDEBUG}. This does not give user code permission to throw
1980 exceptions in violation of the exception specifications; the compiler
1981 will still optimize based on the specifications, so throwing an
1982 unexpected exception will result in undefined behavior.
1985 @itemx -fno-for-scope
1987 @opindex fno-for-scope
1988 If @option{-ffor-scope} is specified, the scope of variables declared in
1989 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1990 as specified by the C++ standard.
1991 If @option{-fno-for-scope} is specified, the scope of variables declared in
1992 a @i{for-init-statement} extends to the end of the enclosing scope,
1993 as was the case in old versions of G++, and other (traditional)
1994 implementations of C++.
1996 The default if neither flag is given to follow the standard,
1997 but to allow and give a warning for old-style code that would
1998 otherwise be invalid, or have different behavior.
2000 @item -fno-gnu-keywords
2001 @opindex fno-gnu-keywords
2002 Do not recognize @code{typeof} as a keyword, so that code can use this
2003 word as an identifier. You can use the keyword @code{__typeof__} instead.
2004 @option{-ansi} implies @option{-fno-gnu-keywords}.
2006 @item -fno-implicit-templates
2007 @opindex fno-implicit-templates
2008 Never emit code for non-inline templates which are instantiated
2009 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2010 @xref{Template Instantiation}, for more information.
2012 @item -fno-implicit-inline-templates
2013 @opindex fno-implicit-inline-templates
2014 Don't emit code for implicit instantiations of inline templates, either.
2015 The default is to handle inlines differently so that compiles with and
2016 without optimization will need the same set of explicit instantiations.
2018 @item -fno-implement-inlines
2019 @opindex fno-implement-inlines
2020 To save space, do not emit out-of-line copies of inline functions
2021 controlled by @samp{#pragma implementation}. This will cause linker
2022 errors if these functions are not inlined everywhere they are called.
2024 @item -fms-extensions
2025 @opindex fms-extensions
2026 Disable pedantic warnings about constructs used in MFC, such as implicit
2027 int and getting a pointer to member function via non-standard syntax.
2029 @item -fno-nonansi-builtins
2030 @opindex fno-nonansi-builtins
2031 Disable built-in declarations of functions that are not mandated by
2032 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2033 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2036 @opindex fnothrow-opt
2037 Treat a @code{throw()} exception specification as though it were a
2038 @code{noexcept} specification to reduce or eliminate the text size
2039 overhead relative to a function with no exception specification. If
2040 the function has local variables of types with non-trivial
2041 destructors, the exception specification will actually make the
2042 function smaller because the EH cleanups for those variables can be
2043 optimized away. The semantic effect is that an exception thrown out of
2044 a function with such an exception specification will result in a call
2045 to @code{terminate} rather than @code{unexpected}.
2047 @item -fno-operator-names
2048 @opindex fno-operator-names
2049 Do not treat the operator name keywords @code{and}, @code{bitand},
2050 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2051 synonyms as keywords.
2053 @item -fno-optional-diags
2054 @opindex fno-optional-diags
2055 Disable diagnostics that the standard says a compiler does not need to
2056 issue. Currently, the only such diagnostic issued by G++ is the one for
2057 a name having multiple meanings within a class.
2060 @opindex fpermissive
2061 Downgrade some diagnostics about nonconformant code from errors to
2062 warnings. Thus, using @option{-fpermissive} will allow some
2063 nonconforming code to compile.
2065 @item -fno-pretty-templates
2066 @opindex fno-pretty-templates
2067 When an error message refers to a specialization of a function
2068 template, the compiler will normally print the signature of the
2069 template followed by the template arguments and any typedefs or
2070 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2071 rather than @code{void f(int)}) so that it's clear which template is
2072 involved. When an error message refers to a specialization of a class
2073 template, the compiler will omit any template arguments which match
2074 the default template arguments for that template. If either of these
2075 behaviors make it harder to understand the error message rather than
2076 easier, using @option{-fno-pretty-templates} will disable them.
2080 Enable automatic template instantiation at link time. This option also
2081 implies @option{-fno-implicit-templates}. @xref{Template
2082 Instantiation}, for more information.
2086 Disable generation of information about every class with virtual
2087 functions for use by the C++ runtime type identification features
2088 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2089 of the language, you can save some space by using this flag. Note that
2090 exception handling uses the same information, but it will generate it as
2091 needed. The @samp{dynamic_cast} operator can still be used for casts that
2092 do not require runtime type information, i.e.@: casts to @code{void *} or to
2093 unambiguous base classes.
2097 Emit statistics about front-end processing at the end of the compilation.
2098 This information is generally only useful to the G++ development team.
2100 @item -fstrict-enums
2101 @opindex fstrict-enums
2102 Allow the compiler to optimize using the assumption that a value of
2103 enumeration type can only be one of the values of the enumeration (as
2104 defined in the C++ standard; basically, a value which can be
2105 represented in the minimum number of bits needed to represent all the
2106 enumerators). This assumption may not be valid if the program uses a
2107 cast to convert an arbitrary integer value to the enumeration type.
2109 @item -ftemplate-depth=@var{n}
2110 @opindex ftemplate-depth
2111 Set the maximum instantiation depth for template classes to @var{n}.
2112 A limit on the template instantiation depth is needed to detect
2113 endless recursions during template class instantiation. ANSI/ISO C++
2114 conforming programs must not rely on a maximum depth greater than 17
2115 (changed to 1024 in C++11). The default value is 900, as the compiler
2116 can run out of stack space before hitting 1024 in some situations.
2118 @item -fno-threadsafe-statics
2119 @opindex fno-threadsafe-statics
2120 Do not emit the extra code to use the routines specified in the C++
2121 ABI for thread-safe initialization of local statics. You can use this
2122 option to reduce code size slightly in code that doesn't need to be
2125 @item -fuse-cxa-atexit
2126 @opindex fuse-cxa-atexit
2127 Register destructors for objects with static storage duration with the
2128 @code{__cxa_atexit} function rather than the @code{atexit} function.
2129 This option is required for fully standards-compliant handling of static
2130 destructors, but will only work if your C library supports
2131 @code{__cxa_atexit}.
2133 @item -fno-use-cxa-get-exception-ptr
2134 @opindex fno-use-cxa-get-exception-ptr
2135 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2136 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2137 if the runtime routine is not available.
2139 @item -fvisibility-inlines-hidden
2140 @opindex fvisibility-inlines-hidden
2141 This switch declares that the user does not attempt to compare
2142 pointers to inline functions or methods where the addresses of the two functions
2143 were taken in different shared objects.
2145 The effect of this is that GCC may, effectively, mark inline methods with
2146 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2147 appear in the export table of a DSO and do not require a PLT indirection
2148 when used within the DSO@. Enabling this option can have a dramatic effect
2149 on load and link times of a DSO as it massively reduces the size of the
2150 dynamic export table when the library makes heavy use of templates.
2152 The behavior of this switch is not quite the same as marking the
2153 methods as hidden directly, because it does not affect static variables
2154 local to the function or cause the compiler to deduce that
2155 the function is defined in only one shared object.
2157 You may mark a method as having a visibility explicitly to negate the
2158 effect of the switch for that method. For example, if you do want to
2159 compare pointers to a particular inline method, you might mark it as
2160 having default visibility. Marking the enclosing class with explicit
2161 visibility will have no effect.
2163 Explicitly instantiated inline methods are unaffected by this option
2164 as their linkage might otherwise cross a shared library boundary.
2165 @xref{Template Instantiation}.
2167 @item -fvisibility-ms-compat
2168 @opindex fvisibility-ms-compat
2169 This flag attempts to use visibility settings to make GCC's C++
2170 linkage model compatible with that of Microsoft Visual Studio.
2172 The flag makes these changes to GCC's linkage model:
2176 It sets the default visibility to @code{hidden}, like
2177 @option{-fvisibility=hidden}.
2180 Types, but not their members, are not hidden by default.
2183 The One Definition Rule is relaxed for types without explicit
2184 visibility specifications which are defined in more than one different
2185 shared object: those declarations are permitted if they would have
2186 been permitted when this option was not used.
2189 In new code it is better to use @option{-fvisibility=hidden} and
2190 export those classes which are intended to be externally visible.
2191 Unfortunately it is possible for code to rely, perhaps accidentally,
2192 on the Visual Studio behavior.
2194 Among the consequences of these changes are that static data members
2195 of the same type with the same name but defined in different shared
2196 objects will be different, so changing one will not change the other;
2197 and that pointers to function members defined in different shared
2198 objects may not compare equal. When this flag is given, it is a
2199 violation of the ODR to define types with the same name differently.
2203 Do not use weak symbol support, even if it is provided by the linker.
2204 By default, G++ will use weak symbols if they are available. This
2205 option exists only for testing, and should not be used by end-users;
2206 it will result in inferior code and has no benefits. This option may
2207 be removed in a future release of G++.
2211 Do not search for header files in the standard directories specific to
2212 C++, but do still search the other standard directories. (This option
2213 is used when building the C++ library.)
2216 In addition, these optimization, warning, and code generation options
2217 have meanings only for C++ programs:
2220 @item -fno-default-inline
2221 @opindex fno-default-inline
2222 Do not assume @samp{inline} for functions defined inside a class scope.
2223 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2224 functions will have linkage like inline functions; they just won't be
2227 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2230 Warn when G++ generates code that is probably not compatible with the
2231 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2232 all such cases, there are probably some cases that are not warned about,
2233 even though G++ is generating incompatible code. There may also be
2234 cases where warnings are emitted even though the code that is generated
2237 You should rewrite your code to avoid these warnings if you are
2238 concerned about the fact that code generated by G++ may not be binary
2239 compatible with code generated by other compilers.
2241 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2246 A template with a non-type template parameter of reference type is
2247 mangled incorrectly:
2250 template <int &> struct S @{@};
2254 This is fixed in @option{-fabi-version=3}.
2257 SIMD vector types declared using @code{__attribute ((vector_size))} are
2258 mangled in a non-standard way that does not allow for overloading of
2259 functions taking vectors of different sizes.
2261 The mangling is changed in @option{-fabi-version=4}.
2264 The known incompatibilities in @option{-fabi-version=1} include:
2269 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2270 pack data into the same byte as a base class. For example:
2273 struct A @{ virtual void f(); int f1 : 1; @};
2274 struct B : public A @{ int f2 : 1; @};
2278 In this case, G++ will place @code{B::f2} into the same byte
2279 as@code{A::f1}; other compilers will not. You can avoid this problem
2280 by explicitly padding @code{A} so that its size is a multiple of the
2281 byte size on your platform; that will cause G++ and other compilers to
2282 layout @code{B} identically.
2285 Incorrect handling of tail-padding for virtual bases. G++ does not use
2286 tail padding when laying out virtual bases. For example:
2289 struct A @{ virtual void f(); char c1; @};
2290 struct B @{ B(); char c2; @};
2291 struct C : public A, public virtual B @{@};
2295 In this case, G++ will not place @code{B} into the tail-padding for
2296 @code{A}; other compilers will. You can avoid this problem by
2297 explicitly padding @code{A} so that its size is a multiple of its
2298 alignment (ignoring virtual base classes); that will cause G++ and other
2299 compilers to layout @code{C} identically.
2302 Incorrect handling of bit-fields with declared widths greater than that
2303 of their underlying types, when the bit-fields appear in a union. For
2307 union U @{ int i : 4096; @};
2311 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2312 union too small by the number of bits in an @code{int}.
2315 Empty classes can be placed at incorrect offsets. For example:
2325 struct C : public B, public A @{@};
2329 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2330 it should be placed at offset zero. G++ mistakenly believes that the
2331 @code{A} data member of @code{B} is already at offset zero.
2334 Names of template functions whose types involve @code{typename} or
2335 template template parameters can be mangled incorrectly.
2338 template <typename Q>
2339 void f(typename Q::X) @{@}
2341 template <template <typename> class Q>
2342 void f(typename Q<int>::X) @{@}
2346 Instantiations of these templates may be mangled incorrectly.
2350 It also warns psABI related changes. The known psABI changes at this
2356 For SYSV/x86-64, when passing union with long double, it is changed to
2357 pass in memory as specified in psABI. For example:
2367 @code{union U} will always be passed in memory.
2371 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2372 @opindex Wctor-dtor-privacy
2373 @opindex Wno-ctor-dtor-privacy
2374 Warn when a class seems unusable because all the constructors or
2375 destructors in that class are private, and it has neither friends nor
2376 public static member functions.
2378 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2379 @opindex Wdelete-non-virtual-dtor
2380 @opindex Wno-delete-non-virtual-dtor
2381 Warn when @samp{delete} is used to destroy an instance of a class which
2382 has virtual functions and non-virtual destructor. It is unsafe to delete
2383 an instance of a derived class through a pointer to a base class if the
2384 base class does not have a virtual destructor. This warning is enabled
2387 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2389 @opindex Wno-narrowing
2390 Warn when a narrowing conversion prohibited by C++11 occurs within
2394 int i = @{ 2.2 @}; // error: narrowing from double to int
2397 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2399 With -std=c++11, @option{-Wno-narrowing} suppresses the diagnostic
2400 required by the standard. Note that this does not affect the meaning
2401 of well-formed code; narrowing conversions are still considered
2402 ill-formed in SFINAE context.
2404 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2406 @opindex Wno-noexcept
2407 Warn when a noexcept-expression evaluates to false because of a call
2408 to a function that does not have a non-throwing exception
2409 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2410 the compiler to never throw an exception.
2412 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2413 @opindex Wnon-virtual-dtor
2414 @opindex Wno-non-virtual-dtor
2415 Warn when a class has virtual functions and accessible non-virtual
2416 destructor, in which case it would be possible but unsafe to delete
2417 an instance of a derived class through a pointer to the base class.
2418 This warning is also enabled if -Weffc++ is specified.
2420 @item -Wreorder @r{(C++ and Objective-C++ only)}
2422 @opindex Wno-reorder
2423 @cindex reordering, warning
2424 @cindex warning for reordering of member initializers
2425 Warn when the order of member initializers given in the code does not
2426 match the order in which they must be executed. For instance:
2432 A(): j (0), i (1) @{ @}
2436 The compiler will rearrange the member initializers for @samp{i}
2437 and @samp{j} to match the declaration order of the members, emitting
2438 a warning to that effect. This warning is enabled by @option{-Wall}.
2441 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2444 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2447 Warn about violations of the following style guidelines from Scott Meyers'
2448 @cite{Effective C++} book:
2452 Item 11: Define a copy constructor and an assignment operator for classes
2453 with dynamically allocated memory.
2456 Item 12: Prefer initialization to assignment in constructors.
2459 Item 14: Make destructors virtual in base classes.
2462 Item 15: Have @code{operator=} return a reference to @code{*this}.
2465 Item 23: Don't try to return a reference when you must return an object.
2469 Also warn about violations of the following style guidelines from
2470 Scott Meyers' @cite{More Effective C++} book:
2474 Item 6: Distinguish between prefix and postfix forms of increment and
2475 decrement operators.
2478 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2482 When selecting this option, be aware that the standard library
2483 headers do not obey all of these guidelines; use @samp{grep -v}
2484 to filter out those warnings.
2486 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2487 @opindex Wstrict-null-sentinel
2488 @opindex Wno-strict-null-sentinel
2489 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2490 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2491 to @code{__null}. Although it is a null pointer constant not a null pointer,
2492 it is guaranteed to be of the same size as a pointer. But this use is
2493 not portable across different compilers.
2495 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2496 @opindex Wno-non-template-friend
2497 @opindex Wnon-template-friend
2498 Disable warnings when non-templatized friend functions are declared
2499 within a template. Since the advent of explicit template specification
2500 support in G++, if the name of the friend is an unqualified-id (i.e.,
2501 @samp{friend foo(int)}), the C++ language specification demands that the
2502 friend declare or define an ordinary, nontemplate function. (Section
2503 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2504 could be interpreted as a particular specialization of a templatized
2505 function. Because this non-conforming behavior is no longer the default
2506 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2507 check existing code for potential trouble spots and is on by default.
2508 This new compiler behavior can be turned off with
2509 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2510 but disables the helpful warning.
2512 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2513 @opindex Wold-style-cast
2514 @opindex Wno-old-style-cast
2515 Warn if an old-style (C-style) cast to a non-void type is used within
2516 a C++ program. The new-style casts (@samp{dynamic_cast},
2517 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2518 less vulnerable to unintended effects and much easier to search for.
2520 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2521 @opindex Woverloaded-virtual
2522 @opindex Wno-overloaded-virtual
2523 @cindex overloaded virtual function, warning
2524 @cindex warning for overloaded virtual function
2525 Warn when a function declaration hides virtual functions from a
2526 base class. For example, in:
2533 struct B: public A @{
2538 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2546 will fail to compile.
2548 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2549 @opindex Wno-pmf-conversions
2550 @opindex Wpmf-conversions
2551 Disable the diagnostic for converting a bound pointer to member function
2554 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2555 @opindex Wsign-promo
2556 @opindex Wno-sign-promo
2557 Warn when overload resolution chooses a promotion from unsigned or
2558 enumerated type to a signed type, over a conversion to an unsigned type of
2559 the same size. Previous versions of G++ would try to preserve
2560 unsignedness, but the standard mandates the current behavior.
2565 A& operator = (int);
2575 In this example, G++ will synthesize a default @samp{A& operator =
2576 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2579 @node Objective-C and Objective-C++ Dialect Options
2580 @section Options Controlling Objective-C and Objective-C++ Dialects
2582 @cindex compiler options, Objective-C and Objective-C++
2583 @cindex Objective-C and Objective-C++ options, command line
2584 @cindex options, Objective-C and Objective-C++
2585 (NOTE: This manual does not describe the Objective-C and Objective-C++
2586 languages themselves. @xref{Standards,,Language Standards
2587 Supported by GCC}, for references.)
2589 This section describes the command-line options that are only meaningful
2590 for Objective-C and Objective-C++ programs, but you can also use most of
2591 the language-independent GNU compiler options.
2592 For example, you might compile a file @code{some_class.m} like this:
2595 gcc -g -fgnu-runtime -O -c some_class.m
2599 In this example, @option{-fgnu-runtime} is an option meant only for
2600 Objective-C and Objective-C++ programs; you can use the other options with
2601 any language supported by GCC@.
2603 Note that since Objective-C is an extension of the C language, Objective-C
2604 compilations may also use options specific to the C front-end (e.g.,
2605 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2606 C++-specific options (e.g., @option{-Wabi}).
2608 Here is a list of options that are @emph{only} for compiling Objective-C
2609 and Objective-C++ programs:
2612 @item -fconstant-string-class=@var{class-name}
2613 @opindex fconstant-string-class
2614 Use @var{class-name} as the name of the class to instantiate for each
2615 literal string specified with the syntax @code{@@"@dots{}"}. The default
2616 class name is @code{NXConstantString} if the GNU runtime is being used, and
2617 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2618 @option{-fconstant-cfstrings} option, if also present, will override the
2619 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2620 to be laid out as constant CoreFoundation strings.
2623 @opindex fgnu-runtime
2624 Generate object code compatible with the standard GNU Objective-C
2625 runtime. This is the default for most types of systems.
2627 @item -fnext-runtime
2628 @opindex fnext-runtime
2629 Generate output compatible with the NeXT runtime. This is the default
2630 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2631 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2634 @item -fno-nil-receivers
2635 @opindex fno-nil-receivers
2636 Assume that all Objective-C message dispatches (@code{[receiver
2637 message:arg]}) in this translation unit ensure that the receiver is
2638 not @code{nil}. This allows for more efficient entry points in the
2639 runtime to be used. This option is only available in conjunction with
2640 the NeXT runtime and ABI version 0 or 1.
2642 @item -fobjc-abi-version=@var{n}
2643 @opindex fobjc-abi-version
2644 Use version @var{n} of the Objective-C ABI for the selected runtime.
2645 This option is currently supported only for the NeXT runtime. In that
2646 case, Version 0 is the traditional (32-bit) ABI without support for
2647 properties and other Objective-C 2.0 additions. Version 1 is the
2648 traditional (32-bit) ABI with support for properties and other
2649 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2650 nothing is specified, the default is Version 0 on 32-bit target
2651 machines, and Version 2 on 64-bit target machines.
2653 @item -fobjc-call-cxx-cdtors
2654 @opindex fobjc-call-cxx-cdtors
2655 For each Objective-C class, check if any of its instance variables is a
2656 C++ object with a non-trivial default constructor. If so, synthesize a
2657 special @code{- (id) .cxx_construct} instance method that will run
2658 non-trivial default constructors on any such instance variables, in order,
2659 and then return @code{self}. Similarly, check if any instance variable
2660 is a C++ object with a non-trivial destructor, and if so, synthesize a
2661 special @code{- (void) .cxx_destruct} method that will run
2662 all such default destructors, in reverse order.
2664 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2665 methods thusly generated will only operate on instance variables
2666 declared in the current Objective-C class, and not those inherited
2667 from superclasses. It is the responsibility of the Objective-C
2668 runtime to invoke all such methods in an object's inheritance
2669 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2670 by the runtime immediately after a new object instance is allocated;
2671 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2672 before the runtime deallocates an object instance.
2674 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2675 support for invoking the @code{- (id) .cxx_construct} and
2676 @code{- (void) .cxx_destruct} methods.
2678 @item -fobjc-direct-dispatch
2679 @opindex fobjc-direct-dispatch
2680 Allow fast jumps to the message dispatcher. On Darwin this is
2681 accomplished via the comm page.
2683 @item -fobjc-exceptions
2684 @opindex fobjc-exceptions
2685 Enable syntactic support for structured exception handling in
2686 Objective-C, similar to what is offered by C++ and Java. This option
2687 is required to use the Objective-C keywords @code{@@try},
2688 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2689 @code{@@synchronized}. This option is available with both the GNU
2690 runtime and the NeXT runtime (but not available in conjunction with
2691 the NeXT runtime on Mac OS X 10.2 and earlier).
2695 Enable garbage collection (GC) in Objective-C and Objective-C++
2696 programs. This option is only available with the NeXT runtime; the
2697 GNU runtime has a different garbage collection implementation that
2698 does not require special compiler flags.
2700 @item -fobjc-nilcheck
2701 @opindex fobjc-nilcheck
2702 For the NeXT runtime with version 2 of the ABI, check for a nil
2703 receiver in method invocations before doing the actual method call.
2704 This is the default and can be disabled using
2705 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2706 checked for nil in this way no matter what this flag is set to.
2707 Currently this flag does nothing when the GNU runtime, or an older
2708 version of the NeXT runtime ABI, is used.
2710 @item -fobjc-std=objc1
2712 Conform to the language syntax of Objective-C 1.0, the language
2713 recognized by GCC 4.0. This only affects the Objective-C additions to
2714 the C/C++ language; it does not affect conformance to C/C++ standards,
2715 which is controlled by the separate C/C++ dialect option flags. When
2716 this option is used with the Objective-C or Objective-C++ compiler,
2717 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2718 This is useful if you need to make sure that your Objective-C code can
2719 be compiled with older versions of GCC.
2721 @item -freplace-objc-classes
2722 @opindex freplace-objc-classes
2723 Emit a special marker instructing @command{ld(1)} not to statically link in
2724 the resulting object file, and allow @command{dyld(1)} to load it in at
2725 run time instead. This is used in conjunction with the Fix-and-Continue
2726 debugging mode, where the object file in question may be recompiled and
2727 dynamically reloaded in the course of program execution, without the need
2728 to restart the program itself. Currently, Fix-and-Continue functionality
2729 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2734 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2735 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2736 compile time) with static class references that get initialized at load time,
2737 which improves run-time performance. Specifying the @option{-fzero-link} flag
2738 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2739 to be retained. This is useful in Zero-Link debugging mode, since it allows
2740 for individual class implementations to be modified during program execution.
2741 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2742 regardless of command line options.
2746 Dump interface declarations for all classes seen in the source file to a
2747 file named @file{@var{sourcename}.decl}.
2749 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2750 @opindex Wassign-intercept
2751 @opindex Wno-assign-intercept
2752 Warn whenever an Objective-C assignment is being intercepted by the
2755 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2756 @opindex Wno-protocol
2758 If a class is declared to implement a protocol, a warning is issued for
2759 every method in the protocol that is not implemented by the class. The
2760 default behavior is to issue a warning for every method not explicitly
2761 implemented in the class, even if a method implementation is inherited
2762 from the superclass. If you use the @option{-Wno-protocol} option, then
2763 methods inherited from the superclass are considered to be implemented,
2764 and no warning is issued for them.
2766 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2768 @opindex Wno-selector
2769 Warn if multiple methods of different types for the same selector are
2770 found during compilation. The check is performed on the list of methods
2771 in the final stage of compilation. Additionally, a check is performed
2772 for each selector appearing in a @code{@@selector(@dots{})}
2773 expression, and a corresponding method for that selector has been found
2774 during compilation. Because these checks scan the method table only at
2775 the end of compilation, these warnings are not produced if the final
2776 stage of compilation is not reached, for example because an error is
2777 found during compilation, or because the @option{-fsyntax-only} option is
2780 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2781 @opindex Wstrict-selector-match
2782 @opindex Wno-strict-selector-match
2783 Warn if multiple methods with differing argument and/or return types are
2784 found for a given selector when attempting to send a message using this
2785 selector to a receiver of type @code{id} or @code{Class}. When this flag
2786 is off (which is the default behavior), the compiler will omit such warnings
2787 if any differences found are confined to types which share the same size
2790 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2791 @opindex Wundeclared-selector
2792 @opindex Wno-undeclared-selector
2793 Warn if a @code{@@selector(@dots{})} expression referring to an
2794 undeclared selector is found. A selector is considered undeclared if no
2795 method with that name has been declared before the
2796 @code{@@selector(@dots{})} expression, either explicitly in an
2797 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2798 an @code{@@implementation} section. This option always performs its
2799 checks as soon as a @code{@@selector(@dots{})} expression is found,
2800 while @option{-Wselector} only performs its checks in the final stage of
2801 compilation. This also enforces the coding style convention
2802 that methods and selectors must be declared before being used.
2804 @item -print-objc-runtime-info
2805 @opindex print-objc-runtime-info
2806 Generate C header describing the largest structure that is passed by
2811 @node Language Independent Options
2812 @section Options to Control Diagnostic Messages Formatting
2813 @cindex options to control diagnostics formatting
2814 @cindex diagnostic messages
2815 @cindex message formatting
2817 Traditionally, diagnostic messages have been formatted irrespective of
2818 the output device's aspect (e.g.@: its width, @dots{}). The options described
2819 below can be used to control the diagnostic messages formatting
2820 algorithm, e.g.@: how many characters per line, how often source location
2821 information should be reported. Right now, only the C++ front end can
2822 honor these options. However it is expected, in the near future, that
2823 the remaining front ends would be able to digest them correctly.
2826 @item -fmessage-length=@var{n}
2827 @opindex fmessage-length
2828 Try to format error messages so that they fit on lines of about @var{n}
2829 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2830 the front ends supported by GCC@. If @var{n} is zero, then no
2831 line-wrapping will be done; each error message will appear on a single
2834 @opindex fdiagnostics-show-location
2835 @item -fdiagnostics-show-location=once
2836 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2837 reporter to emit @emph{once} source location information; that is, in
2838 case the message is too long to fit on a single physical line and has to
2839 be wrapped, the source location won't be emitted (as prefix) again,
2840 over and over, in subsequent continuation lines. This is the default
2843 @item -fdiagnostics-show-location=every-line
2844 Only meaningful in line-wrapping mode. Instructs the diagnostic
2845 messages reporter to emit the same source location information (as
2846 prefix) for physical lines that result from the process of breaking
2847 a message which is too long to fit on a single line.
2849 @item -fno-diagnostics-show-option
2850 @opindex fno-diagnostics-show-option
2851 @opindex fdiagnostics-show-option
2852 By default, each diagnostic emitted includes text which indicates the
2853 command line option that directly controls the diagnostic (if such an
2854 option is known to the diagnostic machinery). Specifying the
2855 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2857 @item -Wcoverage-mismatch
2858 @opindex Wcoverage-mismatch
2859 Warn if feedback profiles do not match when using the
2860 @option{-fprofile-use} option.
2861 If a source file was changed between @option{-fprofile-gen} and
2862 @option{-fprofile-use}, the files with the profile feedback can fail
2863 to match the source file and GCC can not use the profile feedback
2864 information. By default, this warning is enabled and is treated as an
2865 error. @option{-Wno-coverage-mismatch} can be used to disable the
2866 warning or @option{-Wno-error=coverage-mismatch} can be used to
2867 disable the error. Disable the error for this warning can result in
2868 poorly optimized code, so disabling the error is useful only in the
2869 case of very minor changes such as bug fixes to an existing code-base.
2870 Completely disabling the warning is not recommended.
2874 @node Warning Options
2875 @section Options to Request or Suppress Warnings
2876 @cindex options to control warnings
2877 @cindex warning messages
2878 @cindex messages, warning
2879 @cindex suppressing warnings
2881 Warnings are diagnostic messages that report constructions which
2882 are not inherently erroneous but which are risky or suggest there
2883 may have been an error.
2885 The following language-independent options do not enable specific
2886 warnings but control the kinds of diagnostics produced by GCC.
2889 @cindex syntax checking
2891 @opindex fsyntax-only
2892 Check the code for syntax errors, but don't do anything beyond that.
2894 @item -fmax-errors=@var{n}
2895 @opindex fmax-errors
2896 Limits the maximum number of error messages to @var{n}, at which point
2897 GCC bails out rather than attempting to continue processing the source
2898 code. If @var{n} is 0 (the default), there is no limit on the number
2899 of error messages produced. If @option{-Wfatal-errors} is also
2900 specified, then @option{-Wfatal-errors} takes precedence over this
2905 Inhibit all warning messages.
2910 Make all warnings into errors.
2915 Make the specified warning into an error. The specifier for a warning
2916 is appended, for example @option{-Werror=switch} turns the warnings
2917 controlled by @option{-Wswitch} into errors. This switch takes a
2918 negative form, to be used to negate @option{-Werror} for specific
2919 warnings, for example @option{-Wno-error=switch} makes
2920 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2923 The warning message for each controllable warning includes the
2924 option which controls the warning. That option can then be used with
2925 @option{-Werror=} and @option{-Wno-error=} as described above.
2926 (Printing of the option in the warning message can be disabled using the
2927 @option{-fno-diagnostics-show-option} flag.)
2929 Note that specifying @option{-Werror=}@var{foo} automatically implies
2930 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2933 @item -Wfatal-errors
2934 @opindex Wfatal-errors
2935 @opindex Wno-fatal-errors
2936 This option causes the compiler to abort compilation on the first error
2937 occurred rather than trying to keep going and printing further error
2942 You can request many specific warnings with options beginning
2943 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2944 implicit declarations. Each of these specific warning options also
2945 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2946 example, @option{-Wno-implicit}. This manual lists only one of the
2947 two forms, whichever is not the default. For further,
2948 language-specific options also refer to @ref{C++ Dialect Options} and
2949 @ref{Objective-C and Objective-C++ Dialect Options}.
2951 When an unrecognized warning option is requested (e.g.,
2952 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2953 that the option is not recognized. However, if the @option{-Wno-} form
2954 is used, the behavior is slightly different: No diagnostic will be
2955 produced for @option{-Wno-unknown-warning} unless other diagnostics
2956 are being produced. This allows the use of new @option{-Wno-} options
2957 with old compilers, but if something goes wrong, the compiler will
2958 warn that an unrecognized option was used.
2963 Issue all the warnings demanded by strict ISO C and ISO C++;
2964 reject all programs that use forbidden extensions, and some other
2965 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2966 version of the ISO C standard specified by any @option{-std} option used.
2968 Valid ISO C and ISO C++ programs should compile properly with or without
2969 this option (though a rare few will require @option{-ansi} or a
2970 @option{-std} option specifying the required version of ISO C)@. However,
2971 without this option, certain GNU extensions and traditional C and C++
2972 features are supported as well. With this option, they are rejected.
2974 @option{-pedantic} does not cause warning messages for use of the
2975 alternate keywords whose names begin and end with @samp{__}. Pedantic
2976 warnings are also disabled in the expression that follows
2977 @code{__extension__}. However, only system header files should use
2978 these escape routes; application programs should avoid them.
2979 @xref{Alternate Keywords}.
2981 Some users try to use @option{-pedantic} to check programs for strict ISO
2982 C conformance. They soon find that it does not do quite what they want:
2983 it finds some non-ISO practices, but not all---only those for which
2984 ISO C @emph{requires} a diagnostic, and some others for which
2985 diagnostics have been added.
2987 A feature to report any failure to conform to ISO C might be useful in
2988 some instances, but would require considerable additional work and would
2989 be quite different from @option{-pedantic}. We don't have plans to
2990 support such a feature in the near future.
2992 Where the standard specified with @option{-std} represents a GNU
2993 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2994 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2995 extended dialect is based. Warnings from @option{-pedantic} are given
2996 where they are required by the base standard. (It would not make sense
2997 for such warnings to be given only for features not in the specified GNU
2998 C dialect, since by definition the GNU dialects of C include all
2999 features the compiler supports with the given option, and there would be
3000 nothing to warn about.)
3002 @item -pedantic-errors
3003 @opindex pedantic-errors
3004 Like @option{-pedantic}, except that errors are produced rather than
3010 This enables all the warnings about constructions that some users
3011 consider questionable, and that are easy to avoid (or modify to
3012 prevent the warning), even in conjunction with macros. This also
3013 enables some language-specific warnings described in @ref{C++ Dialect
3014 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3016 @option{-Wall} turns on the following warning flags:
3018 @gccoptlist{-Waddress @gol
3019 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3021 -Wchar-subscripts @gol
3022 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
3023 -Wimplicit-int @r{(C and Objective-C only)} @gol
3024 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3027 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3028 -Wmaybe-uninitialized @gol
3029 -Wmissing-braces @gol
3035 -Wsequence-point @gol
3036 -Wsign-compare @r{(only in C++)} @gol
3037 -Wstrict-aliasing @gol
3038 -Wstrict-overflow=1 @gol
3041 -Wuninitialized @gol
3042 -Wunknown-pragmas @gol
3043 -Wunused-function @gol
3046 -Wunused-variable @gol
3047 -Wvolatile-register-var @gol
3050 Note that some warning flags are not implied by @option{-Wall}. Some of
3051 them warn about constructions that users generally do not consider
3052 questionable, but which occasionally you might wish to check for;
3053 others warn about constructions that are necessary or hard to avoid in
3054 some cases, and there is no simple way to modify the code to suppress
3055 the warning. Some of them are enabled by @option{-Wextra} but many of
3056 them must be enabled individually.
3062 This enables some extra warning flags that are not enabled by
3063 @option{-Wall}. (This option used to be called @option{-W}. The older
3064 name is still supported, but the newer name is more descriptive.)
3066 @gccoptlist{-Wclobbered @gol
3068 -Wignored-qualifiers @gol
3069 -Wmissing-field-initializers @gol
3070 -Wmissing-parameter-type @r{(C only)} @gol
3071 -Wold-style-declaration @r{(C only)} @gol
3072 -Woverride-init @gol
3075 -Wuninitialized @gol
3076 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3077 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3080 The option @option{-Wextra} also prints warning messages for the
3086 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3087 @samp{>}, or @samp{>=}.
3090 (C++ only) An enumerator and a non-enumerator both appear in a
3091 conditional expression.
3094 (C++ only) Ambiguous virtual bases.
3097 (C++ only) Subscripting an array which has been declared @samp{register}.
3100 (C++ only) Taking the address of a variable which has been declared
3104 (C++ only) A base class is not initialized in a derived class' copy
3109 @item -Wchar-subscripts
3110 @opindex Wchar-subscripts
3111 @opindex Wno-char-subscripts
3112 Warn if an array subscript has type @code{char}. This is a common cause
3113 of error, as programmers often forget that this type is signed on some
3115 This warning is enabled by @option{-Wall}.
3119 @opindex Wno-comment
3120 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3121 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3122 This warning is enabled by @option{-Wall}.
3125 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3127 Suppress warning messages emitted by @code{#warning} directives.
3129 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3130 @opindex Wdouble-promotion
3131 @opindex Wno-double-promotion
3132 Give a warning when a value of type @code{float} is implicitly
3133 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3134 floating-point unit implement @code{float} in hardware, but emulate
3135 @code{double} in software. On such a machine, doing computations
3136 using @code{double} values is much more expensive because of the
3137 overhead required for software emulation.
3139 It is easy to accidentally do computations with @code{double} because
3140 floating-point literals are implicitly of type @code{double}. For
3144 float area(float radius)
3146 return 3.14159 * radius * radius;
3150 the compiler will perform the entire computation with @code{double}
3151 because the floating-point literal is a @code{double}.
3156 @opindex ffreestanding
3157 @opindex fno-builtin
3158 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3159 the arguments supplied have types appropriate to the format string
3160 specified, and that the conversions specified in the format string make
3161 sense. This includes standard functions, and others specified by format
3162 attributes (@pxref{Function Attributes}), in the @code{printf},
3163 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3164 not in the C standard) families (or other target-specific families).
3165 Which functions are checked without format attributes having been
3166 specified depends on the standard version selected, and such checks of
3167 functions without the attribute specified are disabled by
3168 @option{-ffreestanding} or @option{-fno-builtin}.
3170 The formats are checked against the format features supported by GNU
3171 libc version 2.2. These include all ISO C90 and C99 features, as well
3172 as features from the Single Unix Specification and some BSD and GNU
3173 extensions. Other library implementations may not support all these
3174 features; GCC does not support warning about features that go beyond a
3175 particular library's limitations. However, if @option{-pedantic} is used
3176 with @option{-Wformat}, warnings will be given about format features not
3177 in the selected standard version (but not for @code{strfmon} formats,
3178 since those are not in any version of the C standard). @xref{C Dialect
3179 Options,,Options Controlling C Dialect}.
3181 Since @option{-Wformat} also checks for null format arguments for
3182 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3184 @option{-Wformat} is included in @option{-Wall}. For more control over some
3185 aspects of format checking, the options @option{-Wformat-y2k},
3186 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3187 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3188 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3191 @opindex Wformat-y2k
3192 @opindex Wno-format-y2k
3193 If @option{-Wformat} is specified, also warn about @code{strftime}
3194 formats which may yield only a two-digit year.
3196 @item -Wno-format-contains-nul
3197 @opindex Wno-format-contains-nul
3198 @opindex Wformat-contains-nul
3199 If @option{-Wformat} is specified, do not warn about format strings that
3202 @item -Wno-format-extra-args
3203 @opindex Wno-format-extra-args
3204 @opindex Wformat-extra-args
3205 If @option{-Wformat} is specified, do not warn about excess arguments to a
3206 @code{printf} or @code{scanf} format function. The C standard specifies
3207 that such arguments are ignored.
3209 Where the unused arguments lie between used arguments that are
3210 specified with @samp{$} operand number specifications, normally
3211 warnings are still given, since the implementation could not know what
3212 type to pass to @code{va_arg} to skip the unused arguments. However,
3213 in the case of @code{scanf} formats, this option will suppress the
3214 warning if the unused arguments are all pointers, since the Single
3215 Unix Specification says that such unused arguments are allowed.
3217 @item -Wno-format-zero-length
3218 @opindex Wno-format-zero-length
3219 @opindex Wformat-zero-length
3220 If @option{-Wformat} is specified, do not warn about zero-length formats.
3221 The C standard specifies that zero-length formats are allowed.
3223 @item -Wformat-nonliteral
3224 @opindex Wformat-nonliteral
3225 @opindex Wno-format-nonliteral
3226 If @option{-Wformat} is specified, also warn if the format string is not a
3227 string literal and so cannot be checked, unless the format function
3228 takes its format arguments as a @code{va_list}.
3230 @item -Wformat-security
3231 @opindex Wformat-security
3232 @opindex Wno-format-security
3233 If @option{-Wformat} is specified, also warn about uses of format
3234 functions that represent possible security problems. At present, this
3235 warns about calls to @code{printf} and @code{scanf} functions where the
3236 format string is not a string literal and there are no format arguments,
3237 as in @code{printf (foo);}. This may be a security hole if the format
3238 string came from untrusted input and contains @samp{%n}. (This is
3239 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3240 in future warnings may be added to @option{-Wformat-security} that are not
3241 included in @option{-Wformat-nonliteral}.)
3245 @opindex Wno-format=2
3246 Enable @option{-Wformat} plus format checks not included in
3247 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3248 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3252 @opindex Wno-nonnull
3253 Warn about passing a null pointer for arguments marked as
3254 requiring a non-null value by the @code{nonnull} function attribute.
3256 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3257 can be disabled with the @option{-Wno-nonnull} option.
3259 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3261 @opindex Wno-init-self
3262 Warn about uninitialized variables which are initialized with themselves.
3263 Note this option can only be used with the @option{-Wuninitialized} option.
3265 For example, GCC will warn about @code{i} being uninitialized in the
3266 following snippet only when @option{-Winit-self} has been specified:
3277 @item -Wimplicit-int @r{(C and Objective-C only)}
3278 @opindex Wimplicit-int
3279 @opindex Wno-implicit-int
3280 Warn when a declaration does not specify a type.
3281 This warning is enabled by @option{-Wall}.
3283 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3284 @opindex Wimplicit-function-declaration
3285 @opindex Wno-implicit-function-declaration
3286 Give a warning whenever a function is used before being declared. In
3287 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3288 enabled by default and it is made into an error by
3289 @option{-pedantic-errors}. This warning is also enabled by
3292 @item -Wimplicit @r{(C and Objective-C only)}
3294 @opindex Wno-implicit
3295 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3296 This warning is enabled by @option{-Wall}.
3298 @item -Wignored-qualifiers @r{(C and C++ only)}
3299 @opindex Wignored-qualifiers
3300 @opindex Wno-ignored-qualifiers
3301 Warn if the return type of a function has a type qualifier
3302 such as @code{const}. For ISO C such a type qualifier has no effect,
3303 since the value returned by a function is not an lvalue.
3304 For C++, the warning is only emitted for scalar types or @code{void}.
3305 ISO C prohibits qualified @code{void} return types on function
3306 definitions, so such return types always receive a warning
3307 even without this option.
3309 This warning is also enabled by @option{-Wextra}.
3314 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3315 a function with external linkage, returning int, taking either zero
3316 arguments, two, or three arguments of appropriate types. This warning
3317 is enabled by default in C++ and is enabled by either @option{-Wall}
3318 or @option{-pedantic}.
3320 @item -Wmissing-braces
3321 @opindex Wmissing-braces
3322 @opindex Wno-missing-braces
3323 Warn if an aggregate or union initializer is not fully bracketed. In
3324 the following example, the initializer for @samp{a} is not fully
3325 bracketed, but that for @samp{b} is fully bracketed.
3328 int a[2][2] = @{ 0, 1, 2, 3 @};
3329 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3332 This warning is enabled by @option{-Wall}.
3334 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3335 @opindex Wmissing-include-dirs
3336 @opindex Wno-missing-include-dirs
3337 Warn if a user-supplied include directory does not exist.
3340 @opindex Wparentheses
3341 @opindex Wno-parentheses
3342 Warn if parentheses are omitted in certain contexts, such
3343 as when there is an assignment in a context where a truth value
3344 is expected, or when operators are nested whose precedence people
3345 often get confused about.
3347 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3348 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3349 interpretation from that of ordinary mathematical notation.
3351 Also warn about constructions where there may be confusion to which
3352 @code{if} statement an @code{else} branch belongs. Here is an example of
3367 In C/C++, every @code{else} branch belongs to the innermost possible
3368 @code{if} statement, which in this example is @code{if (b)}. This is
3369 often not what the programmer expected, as illustrated in the above
3370 example by indentation the programmer chose. When there is the
3371 potential for this confusion, GCC will issue a warning when this flag
3372 is specified. To eliminate the warning, add explicit braces around
3373 the innermost @code{if} statement so there is no way the @code{else}
3374 could belong to the enclosing @code{if}. The resulting code would
3391 Also warn for dangerous uses of the
3392 ?: with omitted middle operand GNU extension. When the condition
3393 in the ?: operator is a boolean expression the omitted value will
3394 be always 1. Often the user expects it to be a value computed
3395 inside the conditional expression instead.
3397 This warning is enabled by @option{-Wall}.
3399 @item -Wsequence-point
3400 @opindex Wsequence-point
3401 @opindex Wno-sequence-point
3402 Warn about code that may have undefined semantics because of violations
3403 of sequence point rules in the C and C++ standards.
3405 The C and C++ standards defines the order in which expressions in a C/C++
3406 program are evaluated in terms of @dfn{sequence points}, which represent
3407 a partial ordering between the execution of parts of the program: those
3408 executed before the sequence point, and those executed after it. These
3409 occur after the evaluation of a full expression (one which is not part
3410 of a larger expression), after the evaluation of the first operand of a
3411 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3412 function is called (but after the evaluation of its arguments and the
3413 expression denoting the called function), and in certain other places.
3414 Other than as expressed by the sequence point rules, the order of
3415 evaluation of subexpressions of an expression is not specified. All
3416 these rules describe only a partial order rather than a total order,
3417 since, for example, if two functions are called within one expression
3418 with no sequence point between them, the order in which the functions
3419 are called is not specified. However, the standards committee have
3420 ruled that function calls do not overlap.
3422 It is not specified when between sequence points modifications to the
3423 values of objects take effect. Programs whose behavior depends on this
3424 have undefined behavior; the C and C++ standards specify that ``Between
3425 the previous and next sequence point an object shall have its stored
3426 value modified at most once by the evaluation of an expression.
3427 Furthermore, the prior value shall be read only to determine the value
3428 to be stored.''. If a program breaks these rules, the results on any
3429 particular implementation are entirely unpredictable.
3431 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3432 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3433 diagnosed by this option, and it may give an occasional false positive
3434 result, but in general it has been found fairly effective at detecting
3435 this sort of problem in programs.
3437 The standard is worded confusingly, therefore there is some debate
3438 over the precise meaning of the sequence point rules in subtle cases.
3439 Links to discussions of the problem, including proposed formal
3440 definitions, may be found on the GCC readings page, at
3441 @uref{http://gcc.gnu.org/@/readings.html}.
3443 This warning is enabled by @option{-Wall} for C and C++.
3446 @opindex Wreturn-type
3447 @opindex Wno-return-type
3448 Warn whenever a function is defined with a return-type that defaults
3449 to @code{int}. Also warn about any @code{return} statement with no
3450 return-value in a function whose return-type is not @code{void}
3451 (falling off the end of the function body is considered returning
3452 without a value), and about a @code{return} statement with an
3453 expression in a function whose return-type is @code{void}.
3455 For C++, a function without return type always produces a diagnostic
3456 message, even when @option{-Wno-return-type} is specified. The only
3457 exceptions are @samp{main} and functions defined in system headers.
3459 This warning is enabled by @option{-Wall}.
3464 Warn whenever a @code{switch} statement has an index of enumerated type
3465 and lacks a @code{case} for one or more of the named codes of that
3466 enumeration. (The presence of a @code{default} label prevents this
3467 warning.) @code{case} labels outside the enumeration range also
3468 provoke warnings when this option is used (even if there is a
3469 @code{default} label).
3470 This warning is enabled by @option{-Wall}.
3472 @item -Wswitch-default
3473 @opindex Wswitch-default
3474 @opindex Wno-switch-default
3475 Warn whenever a @code{switch} statement does not have a @code{default}
3479 @opindex Wswitch-enum
3480 @opindex Wno-switch-enum
3481 Warn whenever a @code{switch} statement has an index of enumerated type
3482 and lacks a @code{case} for one or more of the named codes of that
3483 enumeration. @code{case} labels outside the enumeration range also
3484 provoke warnings when this option is used. The only difference
3485 between @option{-Wswitch} and this option is that this option gives a
3486 warning about an omitted enumeration code even if there is a
3487 @code{default} label.
3489 @item -Wsync-nand @r{(C and C++ only)}
3491 @opindex Wno-sync-nand
3492 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3493 built-in functions are used. These functions changed semantics in GCC 4.4.
3497 @opindex Wno-trigraphs
3498 Warn if any trigraphs are encountered that might change the meaning of
3499 the program (trigraphs within comments are not warned about).
3500 This warning is enabled by @option{-Wall}.
3502 @item -Wunused-but-set-parameter
3503 @opindex Wunused-but-set-parameter
3504 @opindex Wno-unused-but-set-parameter
3505 Warn whenever a function parameter is assigned to, but otherwise unused
3506 (aside from its declaration).
3508 To suppress this warning use the @samp{unused} attribute
3509 (@pxref{Variable Attributes}).
3511 This warning is also enabled by @option{-Wunused} together with
3514 @item -Wunused-but-set-variable
3515 @opindex Wunused-but-set-variable
3516 @opindex Wno-unused-but-set-variable
3517 Warn whenever a local variable is assigned to, but otherwise unused
3518 (aside from its declaration).
3519 This warning is enabled by @option{-Wall}.
3521 To suppress this warning use the @samp{unused} attribute
3522 (@pxref{Variable Attributes}).
3524 This warning is also enabled by @option{-Wunused}, which is enabled
3527 @item -Wunused-function
3528 @opindex Wunused-function
3529 @opindex Wno-unused-function
3530 Warn whenever a static function is declared but not defined or a
3531 non-inline static function is unused.
3532 This warning is enabled by @option{-Wall}.
3534 @item -Wunused-label
3535 @opindex Wunused-label
3536 @opindex Wno-unused-label
3537 Warn whenever a label is declared but not used.
3538 This warning is enabled by @option{-Wall}.
3540 To suppress this warning use the @samp{unused} attribute
3541 (@pxref{Variable Attributes}).
3543 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3544 @opindex Wunused-local-typedefs
3545 Warn when a typedef locally defined in a function is not used.
3547 @item -Wunused-parameter
3548 @opindex Wunused-parameter
3549 @opindex Wno-unused-parameter
3550 Warn whenever a function parameter is unused aside from its declaration.
3552 To suppress this warning use the @samp{unused} attribute
3553 (@pxref{Variable Attributes}).
3555 @item -Wno-unused-result
3556 @opindex Wunused-result
3557 @opindex Wno-unused-result
3558 Do not warn if a caller of a function marked with attribute
3559 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3560 its return value. The default is @option{-Wunused-result}.
3562 @item -Wunused-variable
3563 @opindex Wunused-variable
3564 @opindex Wno-unused-variable
3565 Warn whenever a local variable or non-constant static variable is unused
3566 aside from its declaration.
3567 This warning is enabled by @option{-Wall}.
3569 To suppress this warning use the @samp{unused} attribute
3570 (@pxref{Variable Attributes}).
3572 @item -Wunused-value
3573 @opindex Wunused-value
3574 @opindex Wno-unused-value
3575 Warn whenever a statement computes a result that is explicitly not
3576 used. To suppress this warning cast the unused expression to
3577 @samp{void}. This includes an expression-statement or the left-hand
3578 side of a comma expression that contains no side effects. For example,
3579 an expression such as @samp{x[i,j]} will cause a warning, while
3580 @samp{x[(void)i,j]} will not.
3582 This warning is enabled by @option{-Wall}.
3587 All the above @option{-Wunused} options combined.
3589 In order to get a warning about an unused function parameter, you must
3590 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3591 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3593 @item -Wuninitialized
3594 @opindex Wuninitialized
3595 @opindex Wno-uninitialized
3596 Warn if an automatic variable is used without first being initialized
3597 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3598 warn if a non-static reference or non-static @samp{const} member
3599 appears in a class without constructors.
3601 If you want to warn about code which uses the uninitialized value of the
3602 variable in its own initializer, use the @option{-Winit-self} option.
3604 These warnings occur for individual uninitialized or clobbered
3605 elements of structure, union or array variables as well as for
3606 variables which are uninitialized or clobbered as a whole. They do
3607 not occur for variables or elements declared @code{volatile}. Because
3608 these warnings depend on optimization, the exact variables or elements
3609 for which there are warnings will depend on the precise optimization
3610 options and version of GCC used.
3612 Note that there may be no warning about a variable that is used only
3613 to compute a value that itself is never used, because such
3614 computations may be deleted by data flow analysis before the warnings
3617 @item -Wmaybe-uninitialized
3618 @opindex Wmaybe-uninitialized
3619 @opindex Wno-maybe-uninitialized
3620 For an automatic variable, if there exists a path from the function
3621 entry to a use of the variable that is initialized, but there exist
3622 some other paths the variable is not initialized, the compiler will
3623 emit a warning if it can not prove the uninitialized paths do not
3624 happen at runtime. These warnings are made optional because GCC is
3625 not smart enough to see all the reasons why the code might be correct
3626 despite appearing to have an error. Here is one example of how
3647 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3648 always initialized, but GCC doesn't know this. To suppress the
3649 warning, the user needs to provide a default case with assert(0) or
3652 @cindex @code{longjmp} warnings
3653 This option also warns when a non-volatile automatic variable might be
3654 changed by a call to @code{longjmp}. These warnings as well are possible
3655 only in optimizing compilation.
3657 The compiler sees only the calls to @code{setjmp}. It cannot know
3658 where @code{longjmp} will be called; in fact, a signal handler could
3659 call it at any point in the code. As a result, you may get a warning
3660 even when there is in fact no problem because @code{longjmp} cannot
3661 in fact be called at the place which would cause a problem.
3663 Some spurious warnings can be avoided if you declare all the functions
3664 you use that never return as @code{noreturn}. @xref{Function
3667 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3669 @item -Wunknown-pragmas
3670 @opindex Wunknown-pragmas
3671 @opindex Wno-unknown-pragmas
3672 @cindex warning for unknown pragmas
3673 @cindex unknown pragmas, warning
3674 @cindex pragmas, warning of unknown
3675 Warn when a #pragma directive is encountered which is not understood by
3676 GCC@. If this command line option is used, warnings will even be issued
3677 for unknown pragmas in system header files. This is not the case if
3678 the warnings were only enabled by the @option{-Wall} command line option.
3681 @opindex Wno-pragmas
3683 Do not warn about misuses of pragmas, such as incorrect parameters,
3684 invalid syntax, or conflicts between pragmas. See also
3685 @samp{-Wunknown-pragmas}.
3687 @item -Wstrict-aliasing
3688 @opindex Wstrict-aliasing
3689 @opindex Wno-strict-aliasing
3690 This option is only active when @option{-fstrict-aliasing} is active.
3691 It warns about code which might break the strict aliasing rules that the
3692 compiler is using for optimization. The warning does not catch all
3693 cases, but does attempt to catch the more common pitfalls. It is
3694 included in @option{-Wall}.
3695 It is equivalent to @option{-Wstrict-aliasing=3}
3697 @item -Wstrict-aliasing=n
3698 @opindex Wstrict-aliasing=n
3699 @opindex Wno-strict-aliasing=n
3700 This option is only active when @option{-fstrict-aliasing} is active.
3701 It warns about code which might break the strict aliasing rules that the
3702 compiler is using for optimization.
3703 Higher levels correspond to higher accuracy (fewer false positives).
3704 Higher levels also correspond to more effort, similar to the way -O works.
3705 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3708 Level 1: Most aggressive, quick, least accurate.
3709 Possibly useful when higher levels
3710 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3711 false negatives. However, it has many false positives.
3712 Warns for all pointer conversions between possibly incompatible types,
3713 even if never dereferenced. Runs in the frontend only.
3715 Level 2: Aggressive, quick, not too precise.
3716 May still have many false positives (not as many as level 1 though),
3717 and few false negatives (but possibly more than level 1).
3718 Unlike level 1, it only warns when an address is taken. Warns about
3719 incomplete types. Runs in the frontend only.
3721 Level 3 (default for @option{-Wstrict-aliasing}):
3722 Should have very few false positives and few false
3723 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3724 Takes care of the common pun+dereference pattern in the frontend:
3725 @code{*(int*)&some_float}.
3726 If optimization is enabled, it also runs in the backend, where it deals
3727 with multiple statement cases using flow-sensitive points-to information.
3728 Only warns when the converted pointer is dereferenced.
3729 Does not warn about incomplete types.
3731 @item -Wstrict-overflow
3732 @itemx -Wstrict-overflow=@var{n}
3733 @opindex Wstrict-overflow
3734 @opindex Wno-strict-overflow
3735 This option is only active when @option{-fstrict-overflow} is active.
3736 It warns about cases where the compiler optimizes based on the
3737 assumption that signed overflow does not occur. Note that it does not
3738 warn about all cases where the code might overflow: it only warns
3739 about cases where the compiler implements some optimization. Thus
3740 this warning depends on the optimization level.
3742 An optimization which assumes that signed overflow does not occur is
3743 perfectly safe if the values of the variables involved are such that
3744 overflow never does, in fact, occur. Therefore this warning can
3745 easily give a false positive: a warning about code which is not
3746 actually a problem. To help focus on important issues, several
3747 warning levels are defined. No warnings are issued for the use of
3748 undefined signed overflow when estimating how many iterations a loop
3749 will require, in particular when determining whether a loop will be
3753 @item -Wstrict-overflow=1
3754 Warn about cases which are both questionable and easy to avoid. For
3755 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3756 compiler will simplify this to @code{1}. This level of
3757 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3758 are not, and must be explicitly requested.
3760 @item -Wstrict-overflow=2
3761 Also warn about other cases where a comparison is simplified to a
3762 constant. For example: @code{abs (x) >= 0}. This can only be
3763 simplified when @option{-fstrict-overflow} is in effect, because
3764 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3765 zero. @option{-Wstrict-overflow} (with no level) is the same as
3766 @option{-Wstrict-overflow=2}.
3768 @item -Wstrict-overflow=3
3769 Also warn about other cases where a comparison is simplified. For
3770 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3772 @item -Wstrict-overflow=4
3773 Also warn about other simplifications not covered by the above cases.
3774 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3776 @item -Wstrict-overflow=5
3777 Also warn about cases where the compiler reduces the magnitude of a
3778 constant involved in a comparison. For example: @code{x + 2 > y} will
3779 be simplified to @code{x + 1 >= y}. This is reported only at the
3780 highest warning level because this simplification applies to many
3781 comparisons, so this warning level will give a very large number of
3785 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3786 @opindex Wsuggest-attribute=
3787 @opindex Wno-suggest-attribute=
3788 Warn for cases where adding an attribute may be beneficial. The
3789 attributes currently supported are listed below.
3792 @item -Wsuggest-attribute=pure
3793 @itemx -Wsuggest-attribute=const
3794 @itemx -Wsuggest-attribute=noreturn
3795 @opindex Wsuggest-attribute=pure
3796 @opindex Wno-suggest-attribute=pure
3797 @opindex Wsuggest-attribute=const
3798 @opindex Wno-suggest-attribute=const
3799 @opindex Wsuggest-attribute=noreturn
3800 @opindex Wno-suggest-attribute=noreturn
3802 Warn about functions which might be candidates for attributes
3803 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3804 functions visible in other compilation units or (in the case of @code{pure} and
3805 @code{const}) if it cannot prove that the function returns normally. A function
3806 returns normally if it doesn't contain an infinite loop nor returns abnormally
3807 by throwing, calling @code{abort()} or trapping. This analysis requires option
3808 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3809 higher. Higher optimization levels improve the accuracy of the analysis.
3812 @item -Warray-bounds
3813 @opindex Wno-array-bounds
3814 @opindex Warray-bounds
3815 This option is only active when @option{-ftree-vrp} is active
3816 (default for @option{-O2} and above). It warns about subscripts to arrays
3817 that are always out of bounds. This warning is enabled by @option{-Wall}.
3819 @item -Wno-div-by-zero
3820 @opindex Wno-div-by-zero
3821 @opindex Wdiv-by-zero
3822 Do not warn about compile-time integer division by zero. Floating point
3823 division by zero is not warned about, as it can be a legitimate way of
3824 obtaining infinities and NaNs.
3826 @item -Wsystem-headers
3827 @opindex Wsystem-headers
3828 @opindex Wno-system-headers
3829 @cindex warnings from system headers
3830 @cindex system headers, warnings from
3831 Print warning messages for constructs found in system header files.
3832 Warnings from system headers are normally suppressed, on the assumption
3833 that they usually do not indicate real problems and would only make the
3834 compiler output harder to read. Using this command line option tells
3835 GCC to emit warnings from system headers as if they occurred in user
3836 code. However, note that using @option{-Wall} in conjunction with this
3837 option will @emph{not} warn about unknown pragmas in system
3838 headers---for that, @option{-Wunknown-pragmas} must also be used.
3841 @opindex Wtrampolines
3842 @opindex Wno-trampolines
3843 Warn about trampolines generated for pointers to nested functions.
3845 A trampoline is a small piece of data or code that is created at run
3846 time on the stack when the address of a nested function is taken, and
3847 is used to call the nested function indirectly. For some targets, it
3848 is made up of data only and thus requires no special treatment. But,
3849 for most targets, it is made up of code and thus requires the stack
3850 to be made executable in order for the program to work properly.
3853 @opindex Wfloat-equal
3854 @opindex Wno-float-equal
3855 Warn if floating point values are used in equality comparisons.
3857 The idea behind this is that sometimes it is convenient (for the
3858 programmer) to consider floating-point values as approximations to
3859 infinitely precise real numbers. If you are doing this, then you need
3860 to compute (by analyzing the code, or in some other way) the maximum or
3861 likely maximum error that the computation introduces, and allow for it
3862 when performing comparisons (and when producing output, but that's a
3863 different problem). In particular, instead of testing for equality, you
3864 would check to see whether the two values have ranges that overlap; and
3865 this is done with the relational operators, so equality comparisons are
3868 @item -Wtraditional @r{(C and Objective-C only)}
3869 @opindex Wtraditional
3870 @opindex Wno-traditional
3871 Warn about certain constructs that behave differently in traditional and
3872 ISO C@. Also warn about ISO C constructs that have no traditional C
3873 equivalent, and/or problematic constructs which should be avoided.
3877 Macro parameters that appear within string literals in the macro body.
3878 In traditional C macro replacement takes place within string literals,
3879 but does not in ISO C@.
3882 In traditional C, some preprocessor directives did not exist.
3883 Traditional preprocessors would only consider a line to be a directive
3884 if the @samp{#} appeared in column 1 on the line. Therefore
3885 @option{-Wtraditional} warns about directives that traditional C
3886 understands but would ignore because the @samp{#} does not appear as the
3887 first character on the line. It also suggests you hide directives like
3888 @samp{#pragma} not understood by traditional C by indenting them. Some
3889 traditional implementations would not recognize @samp{#elif}, so it
3890 suggests avoiding it altogether.
3893 A function-like macro that appears without arguments.
3896 The unary plus operator.
3899 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3900 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3901 constants.) Note, these suffixes appear in macros defined in the system
3902 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3903 Use of these macros in user code might normally lead to spurious
3904 warnings, however GCC's integrated preprocessor has enough context to
3905 avoid warning in these cases.
3908 A function declared external in one block and then used after the end of
3912 A @code{switch} statement has an operand of type @code{long}.
3915 A non-@code{static} function declaration follows a @code{static} one.
3916 This construct is not accepted by some traditional C compilers.
3919 The ISO type of an integer constant has a different width or
3920 signedness from its traditional type. This warning is only issued if
3921 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3922 typically represent bit patterns, are not warned about.
3925 Usage of ISO string concatenation is detected.
3928 Initialization of automatic aggregates.
3931 Identifier conflicts with labels. Traditional C lacks a separate
3932 namespace for labels.
3935 Initialization of unions. If the initializer is zero, the warning is
3936 omitted. This is done under the assumption that the zero initializer in
3937 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3938 initializer warnings and relies on default initialization to zero in the
3942 Conversions by prototypes between fixed/floating point values and vice
3943 versa. The absence of these prototypes when compiling with traditional
3944 C would cause serious problems. This is a subset of the possible
3945 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3948 Use of ISO C style function definitions. This warning intentionally is
3949 @emph{not} issued for prototype declarations or variadic functions
3950 because these ISO C features will appear in your code when using
3951 libiberty's traditional C compatibility macros, @code{PARAMS} and
3952 @code{VPARAMS}. This warning is also bypassed for nested functions
3953 because that feature is already a GCC extension and thus not relevant to
3954 traditional C compatibility.
3957 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3958 @opindex Wtraditional-conversion
3959 @opindex Wno-traditional-conversion
3960 Warn if a prototype causes a type conversion that is different from what
3961 would happen to the same argument in the absence of a prototype. This
3962 includes conversions of fixed point to floating and vice versa, and
3963 conversions changing the width or signedness of a fixed point argument
3964 except when the same as the default promotion.
3966 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3967 @opindex Wdeclaration-after-statement
3968 @opindex Wno-declaration-after-statement
3969 Warn when a declaration is found after a statement in a block. This
3970 construct, known from C++, was introduced with ISO C99 and is by default
3971 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3972 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3977 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3979 @item -Wno-endif-labels
3980 @opindex Wno-endif-labels
3981 @opindex Wendif-labels
3982 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3987 Warn whenever a local variable or type declaration shadows another variable,
3988 parameter, type, or class member (in C++), or whenever a built-in function
3989 is shadowed. Note that in C++, the compiler will not warn if a local variable
3990 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3992 @item -Wlarger-than=@var{len}
3993 @opindex Wlarger-than=@var{len}
3994 @opindex Wlarger-than-@var{len}
3995 Warn whenever an object of larger than @var{len} bytes is defined.
3997 @item -Wframe-larger-than=@var{len}
3998 @opindex Wframe-larger-than
3999 Warn if the size of a function frame is larger than @var{len} bytes.
4000 The computation done to determine the stack frame size is approximate
4001 and not conservative.
4002 The actual requirements may be somewhat greater than @var{len}
4003 even if you do not get a warning. In addition, any space allocated
4004 via @code{alloca}, variable-length arrays, or related constructs
4005 is not included by the compiler when determining
4006 whether or not to issue a warning.
4008 @item -Wno-free-nonheap-object
4009 @opindex Wno-free-nonheap-object
4010 @opindex Wfree-nonheap-object
4011 Do not warn when attempting to free an object which was not allocated
4014 @item -Wstack-usage=@var{len}
4015 @opindex Wstack-usage
4016 Warn if the stack usage of a function might be larger than @var{len} bytes.
4017 The computation done to determine the stack usage is conservative.
4018 Any space allocated via @code{alloca}, variable-length arrays, or related
4019 constructs is included by the compiler when determining whether or not to
4022 The message is in keeping with the output of @option{-fstack-usage}.
4026 If the stack usage is fully static but exceeds the specified amount, it's:
4029 warning: stack usage is 1120 bytes
4032 If the stack usage is (partly) dynamic but bounded, it's:
4035 warning: stack usage might be 1648 bytes
4038 If the stack usage is (partly) dynamic and not bounded, it's:
4041 warning: stack usage might be unbounded
4045 @item -Wunsafe-loop-optimizations
4046 @opindex Wunsafe-loop-optimizations
4047 @opindex Wno-unsafe-loop-optimizations
4048 Warn if the loop cannot be optimized because the compiler could not
4049 assume anything on the bounds of the loop indices. With
4050 @option{-funsafe-loop-optimizations} warn if the compiler made
4053 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4054 @opindex Wno-pedantic-ms-format
4055 @opindex Wpedantic-ms-format
4056 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4057 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4058 depending on the MS runtime, when you are using the options @option{-Wformat}
4059 and @option{-pedantic} without gnu-extensions.
4061 @item -Wpointer-arith
4062 @opindex Wpointer-arith
4063 @opindex Wno-pointer-arith
4064 Warn about anything that depends on the ``size of'' a function type or
4065 of @code{void}. GNU C assigns these types a size of 1, for
4066 convenience in calculations with @code{void *} pointers and pointers
4067 to functions. In C++, warn also when an arithmetic operation involves
4068 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4071 @opindex Wtype-limits
4072 @opindex Wno-type-limits
4073 Warn if a comparison is always true or always false due to the limited
4074 range of the data type, but do not warn for constant expressions. For
4075 example, warn if an unsigned variable is compared against zero with
4076 @samp{<} or @samp{>=}. This warning is also enabled by
4079 @item -Wbad-function-cast @r{(C and Objective-C only)}
4080 @opindex Wbad-function-cast
4081 @opindex Wno-bad-function-cast
4082 Warn whenever a function call is cast to a non-matching type.
4083 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4085 @item -Wc++-compat @r{(C and Objective-C only)}
4086 Warn about ISO C constructs that are outside of the common subset of
4087 ISO C and ISO C++, e.g.@: request for implicit conversion from
4088 @code{void *} to a pointer to non-@code{void} type.
4090 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4091 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4092 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4093 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4094 enabled by @option{-Wall}.
4098 @opindex Wno-cast-qual
4099 Warn whenever a pointer is cast so as to remove a type qualifier from
4100 the target type. For example, warn if a @code{const char *} is cast
4101 to an ordinary @code{char *}.
4103 Also warn when making a cast which introduces a type qualifier in an
4104 unsafe way. For example, casting @code{char **} to @code{const char **}
4105 is unsafe, as in this example:
4108 /* p is char ** value. */
4109 const char **q = (const char **) p;
4110 /* Assignment of readonly string to const char * is OK. */
4112 /* Now char** pointer points to read-only memory. */
4117 @opindex Wcast-align
4118 @opindex Wno-cast-align
4119 Warn whenever a pointer is cast such that the required alignment of the
4120 target is increased. For example, warn if a @code{char *} is cast to
4121 an @code{int *} on machines where integers can only be accessed at
4122 two- or four-byte boundaries.
4124 @item -Wwrite-strings
4125 @opindex Wwrite-strings
4126 @opindex Wno-write-strings
4127 When compiling C, give string constants the type @code{const
4128 char[@var{length}]} so that copying the address of one into a
4129 non-@code{const} @code{char *} pointer will get a warning. These
4130 warnings will help you find at compile time code that can try to write
4131 into a string constant, but only if you have been very careful about
4132 using @code{const} in declarations and prototypes. Otherwise, it will
4133 just be a nuisance. This is why we did not make @option{-Wall} request
4136 When compiling C++, warn about the deprecated conversion from string
4137 literals to @code{char *}. This warning is enabled by default for C++
4142 @opindex Wno-clobbered
4143 Warn for variables that might be changed by @samp{longjmp} or
4144 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4147 @opindex Wconversion
4148 @opindex Wno-conversion
4149 Warn for implicit conversions that may alter a value. This includes
4150 conversions between real and integer, like @code{abs (x)} when
4151 @code{x} is @code{double}; conversions between signed and unsigned,
4152 like @code{unsigned ui = -1}; and conversions to smaller types, like
4153 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4154 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4155 changed by the conversion like in @code{abs (2.0)}. Warnings about
4156 conversions between signed and unsigned integers can be disabled by
4157 using @option{-Wno-sign-conversion}.
4159 For C++, also warn for confusing overload resolution for user-defined
4160 conversions; and conversions that will never use a type conversion
4161 operator: conversions to @code{void}, the same type, a base class or a
4162 reference to them. Warnings about conversions between signed and
4163 unsigned integers are disabled by default in C++ unless
4164 @option{-Wsign-conversion} is explicitly enabled.
4166 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4167 @opindex Wconversion-null
4168 @opindex Wno-conversion-null
4169 Do not warn for conversions between @code{NULL} and non-pointer
4170 types. @option{-Wconversion-null} is enabled by default.
4172 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4173 @opindex Wzero-as-null-pointer-constant
4174 @opindex Wno-zero-as-null-pointer-constant
4175 Warn when a literal '0' is used as null pointer constant. This can
4176 be useful to facilitate the conversion to @code{nullptr} in C++11.
4179 @opindex Wempty-body
4180 @opindex Wno-empty-body
4181 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4182 while} statement. This warning is also enabled by @option{-Wextra}.
4184 @item -Wenum-compare
4185 @opindex Wenum-compare
4186 @opindex Wno-enum-compare
4187 Warn about a comparison between values of different enum types. In C++
4188 this warning is enabled by default. In C this warning is enabled by
4191 @item -Wjump-misses-init @r{(C, Objective-C only)}
4192 @opindex Wjump-misses-init
4193 @opindex Wno-jump-misses-init
4194 Warn if a @code{goto} statement or a @code{switch} statement jumps
4195 forward across the initialization of a variable, or jumps backward to a
4196 label after the variable has been initialized. This only warns about
4197 variables which are initialized when they are declared. This warning is
4198 only supported for C and Objective C; in C++ this sort of branch is an
4201 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4202 can be disabled with the @option{-Wno-jump-misses-init} option.
4204 @item -Wsign-compare
4205 @opindex Wsign-compare
4206 @opindex Wno-sign-compare
4207 @cindex warning for comparison of signed and unsigned values
4208 @cindex comparison of signed and unsigned values, warning
4209 @cindex signed and unsigned values, comparison warning
4210 Warn when a comparison between signed and unsigned values could produce
4211 an incorrect result when the signed value is converted to unsigned.
4212 This warning is also enabled by @option{-Wextra}; to get the other warnings
4213 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4215 @item -Wsign-conversion
4216 @opindex Wsign-conversion
4217 @opindex Wno-sign-conversion
4218 Warn for implicit conversions that may change the sign of an integer
4219 value, like assigning a signed integer expression to an unsigned
4220 integer variable. An explicit cast silences the warning. In C, this
4221 option is enabled also by @option{-Wconversion}.
4225 @opindex Wno-address
4226 Warn about suspicious uses of memory addresses. These include using
4227 the address of a function in a conditional expression, such as
4228 @code{void func(void); if (func)}, and comparisons against the memory
4229 address of a string literal, such as @code{if (x == "abc")}. Such
4230 uses typically indicate a programmer error: the address of a function
4231 always evaluates to true, so their use in a conditional usually
4232 indicate that the programmer forgot the parentheses in a function
4233 call; and comparisons against string literals result in unspecified
4234 behavior and are not portable in C, so they usually indicate that the
4235 programmer intended to use @code{strcmp}. This warning is enabled by
4239 @opindex Wlogical-op
4240 @opindex Wno-logical-op
4241 Warn about suspicious uses of logical operators in expressions.
4242 This includes using logical operators in contexts where a
4243 bit-wise operator is likely to be expected.
4245 @item -Waggregate-return
4246 @opindex Waggregate-return
4247 @opindex Wno-aggregate-return
4248 Warn if any functions that return structures or unions are defined or
4249 called. (In languages where you can return an array, this also elicits
4252 @item -Wno-attributes
4253 @opindex Wno-attributes
4254 @opindex Wattributes
4255 Do not warn if an unexpected @code{__attribute__} is used, such as
4256 unrecognized attributes, function attributes applied to variables,
4257 etc. This will not stop errors for incorrect use of supported
4260 @item -Wno-builtin-macro-redefined
4261 @opindex Wno-builtin-macro-redefined
4262 @opindex Wbuiltin-macro-redefined
4263 Do not warn if certain built-in macros are redefined. This suppresses
4264 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4265 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4267 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4268 @opindex Wstrict-prototypes
4269 @opindex Wno-strict-prototypes
4270 Warn if a function is declared or defined without specifying the
4271 argument types. (An old-style function definition is permitted without
4272 a warning if preceded by a declaration which specifies the argument
4275 @item -Wold-style-declaration @r{(C and Objective-C only)}
4276 @opindex Wold-style-declaration
4277 @opindex Wno-old-style-declaration
4278 Warn for obsolescent usages, according to the C Standard, in a
4279 declaration. For example, warn if storage-class specifiers like
4280 @code{static} are not the first things in a declaration. This warning
4281 is also enabled by @option{-Wextra}.
4283 @item -Wold-style-definition @r{(C and Objective-C only)}
4284 @opindex Wold-style-definition
4285 @opindex Wno-old-style-definition
4286 Warn if an old-style function definition is used. A warning is given
4287 even if there is a previous prototype.
4289 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4290 @opindex Wmissing-parameter-type
4291 @opindex Wno-missing-parameter-type
4292 A function parameter is declared without a type specifier in K&R-style
4299 This warning is also enabled by @option{-Wextra}.
4301 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4302 @opindex Wmissing-prototypes
4303 @opindex Wno-missing-prototypes
4304 Warn if a global function is defined without a previous prototype
4305 declaration. This warning is issued even if the definition itself
4306 provides a prototype. The aim is to detect global functions that fail
4307 to be declared in header files.
4309 @item -Wmissing-declarations
4310 @opindex Wmissing-declarations
4311 @opindex Wno-missing-declarations
4312 Warn if a global function is defined without a previous declaration.
4313 Do so even if the definition itself provides a prototype.
4314 Use this option to detect global functions that are not declared in
4315 header files. In C++, no warnings are issued for function templates,
4316 or for inline functions, or for functions in anonymous namespaces.
4318 @item -Wmissing-field-initializers
4319 @opindex Wmissing-field-initializers
4320 @opindex Wno-missing-field-initializers
4324 Warn if a structure's initializer has some fields missing. For
4325 example, the following code would cause such a warning, because
4326 @code{x.h} is implicitly zero:
4329 struct s @{ int f, g, h; @};
4330 struct s x = @{ 3, 4 @};
4333 This option does not warn about designated initializers, so the following
4334 modification would not trigger a warning:
4337 struct s @{ int f, g, h; @};
4338 struct s x = @{ .f = 3, .g = 4 @};
4341 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4342 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4344 @item -Wmissing-format-attribute
4345 @opindex Wmissing-format-attribute
4346 @opindex Wno-missing-format-attribute
4349 Warn about function pointers which might be candidates for @code{format}
4350 attributes. Note these are only possible candidates, not absolute ones.
4351 GCC will guess that function pointers with @code{format} attributes that
4352 are used in assignment, initialization, parameter passing or return
4353 statements should have a corresponding @code{format} attribute in the
4354 resulting type. I.e.@: the left-hand side of the assignment or
4355 initialization, the type of the parameter variable, or the return type
4356 of the containing function respectively should also have a @code{format}
4357 attribute to avoid the warning.
4359 GCC will also warn about function definitions which might be
4360 candidates for @code{format} attributes. Again, these are only
4361 possible candidates. GCC will guess that @code{format} attributes
4362 might be appropriate for any function that calls a function like
4363 @code{vprintf} or @code{vscanf}, but this might not always be the
4364 case, and some functions for which @code{format} attributes are
4365 appropriate may not be detected.
4367 @item -Wno-multichar
4368 @opindex Wno-multichar
4370 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4371 Usually they indicate a typo in the user's code, as they have
4372 implementation-defined values, and should not be used in portable code.
4374 @item -Wnormalized=<none|id|nfc|nfkc>
4375 @opindex Wnormalized=
4378 @cindex character set, input normalization
4379 In ISO C and ISO C++, two identifiers are different if they are
4380 different sequences of characters. However, sometimes when characters
4381 outside the basic ASCII character set are used, you can have two
4382 different character sequences that look the same. To avoid confusion,
4383 the ISO 10646 standard sets out some @dfn{normalization rules} which
4384 when applied ensure that two sequences that look the same are turned into
4385 the same sequence. GCC can warn you if you are using identifiers which
4386 have not been normalized; this option controls that warning.
4388 There are four levels of warning that GCC supports. The default is
4389 @option{-Wnormalized=nfc}, which warns about any identifier which is
4390 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4391 recommended form for most uses.
4393 Unfortunately, there are some characters which ISO C and ISO C++ allow
4394 in identifiers that when turned into NFC aren't allowable as
4395 identifiers. That is, there's no way to use these symbols in portable
4396 ISO C or C++ and have all your identifiers in NFC@.
4397 @option{-Wnormalized=id} suppresses the warning for these characters.
4398 It is hoped that future versions of the standards involved will correct
4399 this, which is why this option is not the default.
4401 You can switch the warning off for all characters by writing
4402 @option{-Wnormalized=none}. You would only want to do this if you
4403 were using some other normalization scheme (like ``D''), because
4404 otherwise you can easily create bugs that are literally impossible to see.
4406 Some characters in ISO 10646 have distinct meanings but look identical
4407 in some fonts or display methodologies, especially once formatting has
4408 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4409 LETTER N'', will display just like a regular @code{n} which has been
4410 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4411 normalization scheme to convert all these into a standard form as
4412 well, and GCC will warn if your code is not in NFKC if you use
4413 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4414 about every identifier that contains the letter O because it might be
4415 confused with the digit 0, and so is not the default, but may be
4416 useful as a local coding convention if the programming environment is
4417 unable to be fixed to display these characters distinctly.
4419 @item -Wno-deprecated
4420 @opindex Wno-deprecated
4421 @opindex Wdeprecated
4422 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4424 @item -Wno-deprecated-declarations
4425 @opindex Wno-deprecated-declarations
4426 @opindex Wdeprecated-declarations
4427 Do not warn about uses of functions (@pxref{Function Attributes}),
4428 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4429 Attributes}) marked as deprecated by using the @code{deprecated}
4433 @opindex Wno-overflow
4435 Do not warn about compile-time overflow in constant expressions.
4437 @item -Woverride-init @r{(C and Objective-C only)}
4438 @opindex Woverride-init
4439 @opindex Wno-override-init
4443 Warn if an initialized field without side effects is overridden when
4444 using designated initializers (@pxref{Designated Inits, , Designated
4447 This warning is included in @option{-Wextra}. To get other
4448 @option{-Wextra} warnings without this one, use @samp{-Wextra
4449 -Wno-override-init}.
4454 Warn if a structure is given the packed attribute, but the packed
4455 attribute has no effect on the layout or size of the structure.
4456 Such structures may be mis-aligned for little benefit. For
4457 instance, in this code, the variable @code{f.x} in @code{struct bar}
4458 will be misaligned even though @code{struct bar} does not itself
4459 have the packed attribute:
4466 @} __attribute__((packed));
4474 @item -Wpacked-bitfield-compat
4475 @opindex Wpacked-bitfield-compat
4476 @opindex Wno-packed-bitfield-compat
4477 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4478 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4479 the change can lead to differences in the structure layout. GCC
4480 informs you when the offset of such a field has changed in GCC 4.4.
4481 For example there is no longer a 4-bit padding between field @code{a}
4482 and @code{b} in this structure:
4489 @} __attribute__ ((packed));
4492 This warning is enabled by default. Use
4493 @option{-Wno-packed-bitfield-compat} to disable this warning.
4498 Warn if padding is included in a structure, either to align an element
4499 of the structure or to align the whole structure. Sometimes when this
4500 happens it is possible to rearrange the fields of the structure to
4501 reduce the padding and so make the structure smaller.
4503 @item -Wredundant-decls
4504 @opindex Wredundant-decls
4505 @opindex Wno-redundant-decls
4506 Warn if anything is declared more than once in the same scope, even in
4507 cases where multiple declaration is valid and changes nothing.
4509 @item -Wnested-externs @r{(C and Objective-C only)}
4510 @opindex Wnested-externs
4511 @opindex Wno-nested-externs
4512 Warn if an @code{extern} declaration is encountered within a function.
4517 Warn if a function can not be inlined and it was declared as inline.
4518 Even with this option, the compiler will not warn about failures to
4519 inline functions declared in system headers.
4521 The compiler uses a variety of heuristics to determine whether or not
4522 to inline a function. For example, the compiler takes into account
4523 the size of the function being inlined and the amount of inlining
4524 that has already been done in the current function. Therefore,
4525 seemingly insignificant changes in the source program can cause the
4526 warnings produced by @option{-Winline} to appear or disappear.
4528 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4529 @opindex Wno-invalid-offsetof
4530 @opindex Winvalid-offsetof
4531 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4532 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4533 to a non-POD type is undefined. In existing C++ implementations,
4534 however, @samp{offsetof} typically gives meaningful results even when
4535 applied to certain kinds of non-POD types. (Such as a simple
4536 @samp{struct} that fails to be a POD type only by virtue of having a
4537 constructor.) This flag is for users who are aware that they are
4538 writing nonportable code and who have deliberately chosen to ignore the
4541 The restrictions on @samp{offsetof} may be relaxed in a future version
4542 of the C++ standard.
4544 @item -Wno-int-to-pointer-cast
4545 @opindex Wno-int-to-pointer-cast
4546 @opindex Wint-to-pointer-cast
4547 Suppress warnings from casts to pointer type of an integer of a
4548 different size. In C++, casting to a pointer type of smaller size is
4549 an error. @option{Wint-to-pointer-cast} is enabled by default.
4552 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4553 @opindex Wno-pointer-to-int-cast
4554 @opindex Wpointer-to-int-cast
4555 Suppress warnings from casts from a pointer to an integer type of a
4559 @opindex Winvalid-pch
4560 @opindex Wno-invalid-pch
4561 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4562 the search path but can't be used.
4566 @opindex Wno-long-long
4567 Warn if @samp{long long} type is used. This is enabled by either
4568 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4569 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4571 @item -Wvariadic-macros
4572 @opindex Wvariadic-macros
4573 @opindex Wno-variadic-macros
4574 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4575 alternate syntax when in pedantic ISO C99 mode. This is default.
4576 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4578 @item -Wvector-operation-performance
4579 @opindex Wvector-operation-performance
4580 @opindex Wno-vector-operation-performance
4581 Warn if vector operation is not implemented via SIMD capabilities of the
4582 architecture. Mainly useful for the performance tuning.
4583 Vector operation can be implemented @code{piecewise} which means that the
4584 scalar operation is performed on every vector element;
4585 @code{in parallel} which means that the vector operation is implemented
4586 using scalars of wider type, which normally is more performance efficient;
4587 and @code{as a single scalar} which means that vector fits into a
4593 Warn if variable length array is used in the code.
4594 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4595 the variable length array.
4597 @item -Wvolatile-register-var
4598 @opindex Wvolatile-register-var
4599 @opindex Wno-volatile-register-var
4600 Warn if a register variable is declared volatile. The volatile
4601 modifier does not inhibit all optimizations that may eliminate reads
4602 and/or writes to register variables. This warning is enabled by
4605 @item -Wdisabled-optimization
4606 @opindex Wdisabled-optimization
4607 @opindex Wno-disabled-optimization
4608 Warn if a requested optimization pass is disabled. This warning does
4609 not generally indicate that there is anything wrong with your code; it
4610 merely indicates that GCC's optimizers were unable to handle the code
4611 effectively. Often, the problem is that your code is too big or too
4612 complex; GCC will refuse to optimize programs when the optimization
4613 itself is likely to take inordinate amounts of time.
4615 @item -Wpointer-sign @r{(C and Objective-C only)}
4616 @opindex Wpointer-sign
4617 @opindex Wno-pointer-sign
4618 Warn for pointer argument passing or assignment with different signedness.
4619 This option is only supported for C and Objective-C@. It is implied by
4620 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4621 @option{-Wno-pointer-sign}.
4623 @item -Wstack-protector
4624 @opindex Wstack-protector
4625 @opindex Wno-stack-protector
4626 This option is only active when @option{-fstack-protector} is active. It
4627 warns about functions that will not be protected against stack smashing.
4630 @opindex Wno-mudflap
4631 Suppress warnings about constructs that cannot be instrumented by
4634 @item -Woverlength-strings
4635 @opindex Woverlength-strings
4636 @opindex Wno-overlength-strings
4637 Warn about string constants which are longer than the ``minimum
4638 maximum'' length specified in the C standard. Modern compilers
4639 generally allow string constants which are much longer than the
4640 standard's minimum limit, but very portable programs should avoid
4641 using longer strings.
4643 The limit applies @emph{after} string constant concatenation, and does
4644 not count the trailing NUL@. In C90, the limit was 509 characters; in
4645 C99, it was raised to 4095. C++98 does not specify a normative
4646 minimum maximum, so we do not diagnose overlength strings in C++@.
4648 This option is implied by @option{-pedantic}, and can be disabled with
4649 @option{-Wno-overlength-strings}.
4651 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4652 @opindex Wunsuffixed-float-constants
4654 GCC will issue a warning for any floating constant that does not have
4655 a suffix. When used together with @option{-Wsystem-headers} it will
4656 warn about such constants in system header files. This can be useful
4657 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4658 from the decimal floating-point extension to C99.
4661 @node Debugging Options
4662 @section Options for Debugging Your Program or GCC
4663 @cindex options, debugging
4664 @cindex debugging information options
4666 GCC has various special options that are used for debugging
4667 either your program or GCC:
4672 Produce debugging information in the operating system's native format
4673 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4676 On most systems that use stabs format, @option{-g} enables use of extra
4677 debugging information that only GDB can use; this extra information
4678 makes debugging work better in GDB but will probably make other debuggers
4680 refuse to read the program. If you want to control for certain whether
4681 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4682 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4684 GCC allows you to use @option{-g} with
4685 @option{-O}. The shortcuts taken by optimized code may occasionally
4686 produce surprising results: some variables you declared may not exist
4687 at all; flow of control may briefly move where you did not expect it;
4688 some statements may not be executed because they compute constant
4689 results or their values were already at hand; some statements may
4690 execute in different places because they were moved out of loops.
4692 Nevertheless it proves possible to debug optimized output. This makes
4693 it reasonable to use the optimizer for programs that might have bugs.
4695 The following options are useful when GCC is generated with the
4696 capability for more than one debugging format.
4700 Produce debugging information for use by GDB@. This means to use the
4701 most expressive format available (DWARF 2, stabs, or the native format
4702 if neither of those are supported), including GDB extensions if at all
4707 Produce debugging information in stabs format (if that is supported),
4708 without GDB extensions. This is the format used by DBX on most BSD
4709 systems. On MIPS, Alpha and System V Release 4 systems this option
4710 produces stabs debugging output which is not understood by DBX or SDB@.
4711 On System V Release 4 systems this option requires the GNU assembler.
4713 @item -feliminate-unused-debug-symbols
4714 @opindex feliminate-unused-debug-symbols
4715 Produce debugging information in stabs format (if that is supported),
4716 for only symbols that are actually used.
4718 @item -femit-class-debug-always
4719 Instead of emitting debugging information for a C++ class in only one
4720 object file, emit it in all object files using the class. This option
4721 should be used only with debuggers that are unable to handle the way GCC
4722 normally emits debugging information for classes because using this
4723 option will increase the size of debugging information by as much as a
4726 @item -fno-debug-types-section
4727 @opindex fno-debug-types-section
4728 @opindex fdebug-types-section
4729 By default when using DWARF v4 or higher type DIEs will be put into
4730 their own .debug_types section instead of making them part of the
4731 .debug_info section. It is more efficient to put them in a separate
4732 comdat sections since the linker will then be able to remove duplicates.
4733 But not all DWARF consumers support .debug_types sections yet.
4737 Produce debugging information in stabs format (if that is supported),
4738 using GNU extensions understood only by the GNU debugger (GDB)@. The
4739 use of these extensions is likely to make other debuggers crash or
4740 refuse to read the program.
4744 Produce debugging information in COFF format (if that is supported).
4745 This is the format used by SDB on most System V systems prior to
4750 Produce debugging information in XCOFF format (if that is supported).
4751 This is the format used by the DBX debugger on IBM RS/6000 systems.
4755 Produce debugging information in XCOFF format (if that is supported),
4756 using GNU extensions understood only by the GNU debugger (GDB)@. The
4757 use of these extensions is likely to make other debuggers crash or
4758 refuse to read the program, and may cause assemblers other than the GNU
4759 assembler (GAS) to fail with an error.
4761 @item -gdwarf-@var{version}
4762 @opindex gdwarf-@var{version}
4763 Produce debugging information in DWARF format (if that is
4764 supported). This is the format used by DBX on IRIX 6. The value
4765 of @var{version} may be either 2, 3 or 4; the default version is 2.
4767 Note that with DWARF version 2 some ports require, and will always
4768 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4770 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4771 for maximum benefit.
4773 @item -grecord-gcc-switches
4774 @opindex grecord-gcc-switches
4775 This switch causes the command line options, that were used to invoke the
4776 compiler and may affect code generation, to be appended to the
4777 DW_AT_producer attribute in DWARF debugging information. The options
4778 are concatenated with spaces separating them from each other and from
4779 the compiler version. See also @option{-frecord-gcc-switches} for another
4780 way of storing compiler options into the object file.
4782 @item -gno-record-gcc-switches
4783 @opindex gno-record-gcc-switches
4784 Disallow appending command line options to the DW_AT_producer attribute
4785 in DWARF debugging information. This is the default.
4787 @item -gstrict-dwarf
4788 @opindex gstrict-dwarf
4789 Disallow using extensions of later DWARF standard version than selected
4790 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4791 DWARF extensions from later standard versions is allowed.
4793 @item -gno-strict-dwarf
4794 @opindex gno-strict-dwarf
4795 Allow using extensions of later DWARF standard version than selected with
4796 @option{-gdwarf-@var{version}}.
4800 Produce debugging information in VMS debug format (if that is
4801 supported). This is the format used by DEBUG on VMS systems.
4804 @itemx -ggdb@var{level}
4805 @itemx -gstabs@var{level}
4806 @itemx -gcoff@var{level}
4807 @itemx -gxcoff@var{level}
4808 @itemx -gvms@var{level}
4809 Request debugging information and also use @var{level} to specify how
4810 much information. The default level is 2.
4812 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4815 Level 1 produces minimal information, enough for making backtraces in
4816 parts of the program that you don't plan to debug. This includes
4817 descriptions of functions and external variables, but no information
4818 about local variables and no line numbers.
4820 Level 3 includes extra information, such as all the macro definitions
4821 present in the program. Some debuggers support macro expansion when
4822 you use @option{-g3}.
4824 @option{-gdwarf-2} does not accept a concatenated debug level, because
4825 GCC used to support an option @option{-gdwarf} that meant to generate
4826 debug information in version 1 of the DWARF format (which is very
4827 different from version 2), and it would have been too confusing. That
4828 debug format is long obsolete, but the option cannot be changed now.
4829 Instead use an additional @option{-g@var{level}} option to change the
4830 debug level for DWARF.
4834 Turn off generation of debug info, if leaving out this option would have
4835 generated it, or turn it on at level 2 otherwise. The position of this
4836 argument in the command line does not matter, it takes effect after all
4837 other options are processed, and it does so only once, no matter how
4838 many times it is given. This is mainly intended to be used with
4839 @option{-fcompare-debug}.
4841 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4842 @opindex fdump-final-insns
4843 Dump the final internal representation (RTL) to @var{file}. If the
4844 optional argument is omitted (or if @var{file} is @code{.}), the name
4845 of the dump file will be determined by appending @code{.gkd} to the
4846 compilation output file name.
4848 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4849 @opindex fcompare-debug
4850 @opindex fno-compare-debug
4851 If no error occurs during compilation, run the compiler a second time,
4852 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4853 passed to the second compilation. Dump the final internal
4854 representation in both compilations, and print an error if they differ.
4856 If the equal sign is omitted, the default @option{-gtoggle} is used.
4858 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4859 and nonzero, implicitly enables @option{-fcompare-debug}. If
4860 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4861 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4864 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4865 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4866 of the final representation and the second compilation, preventing even
4867 @env{GCC_COMPARE_DEBUG} from taking effect.
4869 To verify full coverage during @option{-fcompare-debug} testing, set
4870 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4871 which GCC will reject as an invalid option in any actual compilation
4872 (rather than preprocessing, assembly or linking). To get just a
4873 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4874 not overridden} will do.
4876 @item -fcompare-debug-second
4877 @opindex fcompare-debug-second
4878 This option is implicitly passed to the compiler for the second
4879 compilation requested by @option{-fcompare-debug}, along with options to
4880 silence warnings, and omitting other options that would cause
4881 side-effect compiler outputs to files or to the standard output. Dump
4882 files and preserved temporary files are renamed so as to contain the
4883 @code{.gk} additional extension during the second compilation, to avoid
4884 overwriting those generated by the first.
4886 When this option is passed to the compiler driver, it causes the
4887 @emph{first} compilation to be skipped, which makes it useful for little
4888 other than debugging the compiler proper.
4890 @item -feliminate-dwarf2-dups
4891 @opindex feliminate-dwarf2-dups
4892 Compress DWARF2 debugging information by eliminating duplicated
4893 information about each symbol. This option only makes sense when
4894 generating DWARF2 debugging information with @option{-gdwarf-2}.
4896 @item -femit-struct-debug-baseonly
4897 Emit debug information for struct-like types
4898 only when the base name of the compilation source file
4899 matches the base name of file in which the struct was defined.
4901 This option substantially reduces the size of debugging information,
4902 but at significant potential loss in type information to the debugger.
4903 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4904 See @option{-femit-struct-debug-detailed} for more detailed control.
4906 This option works only with DWARF 2.
4908 @item -femit-struct-debug-reduced
4909 Emit debug information for struct-like types
4910 only when the base name of the compilation source file
4911 matches the base name of file in which the type was defined,
4912 unless the struct is a template or defined in a system header.
4914 This option significantly reduces the size of debugging information,
4915 with some potential loss in type information to the debugger.
4916 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4917 See @option{-femit-struct-debug-detailed} for more detailed control.
4919 This option works only with DWARF 2.
4921 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4922 Specify the struct-like types
4923 for which the compiler will generate debug information.
4924 The intent is to reduce duplicate struct debug information
4925 between different object files within the same program.
4927 This option is a detailed version of
4928 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4929 which will serve for most needs.
4931 A specification has the syntax@*
4932 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4934 The optional first word limits the specification to
4935 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4936 A struct type is used directly when it is the type of a variable, member.
4937 Indirect uses arise through pointers to structs.
4938 That is, when use of an incomplete struct would be legal, the use is indirect.
4940 @samp{struct one direct; struct two * indirect;}.
4942 The optional second word limits the specification to
4943 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4944 Generic structs are a bit complicated to explain.
4945 For C++, these are non-explicit specializations of template classes,
4946 or non-template classes within the above.
4947 Other programming languages have generics,
4948 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4950 The third word specifies the source files for those
4951 structs for which the compiler will emit debug information.
4952 The values @samp{none} and @samp{any} have the normal meaning.
4953 The value @samp{base} means that
4954 the base of name of the file in which the type declaration appears
4955 must match the base of the name of the main compilation file.
4956 In practice, this means that
4957 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4958 but types declared in other header will not.
4959 The value @samp{sys} means those types satisfying @samp{base}
4960 or declared in system or compiler headers.
4962 You may need to experiment to determine the best settings for your application.
4964 The default is @samp{-femit-struct-debug-detailed=all}.
4966 This option works only with DWARF 2.
4968 @item -fno-merge-debug-strings
4969 @opindex fmerge-debug-strings
4970 @opindex fno-merge-debug-strings
4971 Direct the linker to not merge together strings in the debugging
4972 information which are identical in different object files. Merging is
4973 not supported by all assemblers or linkers. Merging decreases the size
4974 of the debug information in the output file at the cost of increasing
4975 link processing time. Merging is enabled by default.
4977 @item -fdebug-prefix-map=@var{old}=@var{new}
4978 @opindex fdebug-prefix-map
4979 When compiling files in directory @file{@var{old}}, record debugging
4980 information describing them as in @file{@var{new}} instead.
4982 @item -fno-dwarf2-cfi-asm
4983 @opindex fdwarf2-cfi-asm
4984 @opindex fno-dwarf2-cfi-asm
4985 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4986 instead of using GAS @code{.cfi_*} directives.
4988 @cindex @command{prof}
4991 Generate extra code to write profile information suitable for the
4992 analysis program @command{prof}. You must use this option when compiling
4993 the source files you want data about, and you must also use it when
4996 @cindex @command{gprof}
4999 Generate extra code to write profile information suitable for the
5000 analysis program @command{gprof}. You must use this option when compiling
5001 the source files you want data about, and you must also use it when
5006 Makes the compiler print out each function name as it is compiled, and
5007 print some statistics about each pass when it finishes.
5010 @opindex ftime-report
5011 Makes the compiler print some statistics about the time consumed by each
5012 pass when it finishes.
5015 @opindex fmem-report
5016 Makes the compiler print some statistics about permanent memory
5017 allocation when it finishes.
5019 @item -fpre-ipa-mem-report
5020 @opindex fpre-ipa-mem-report
5021 @item -fpost-ipa-mem-report
5022 @opindex fpost-ipa-mem-report
5023 Makes the compiler print some statistics about permanent memory
5024 allocation before or after interprocedural optimization.
5027 @opindex fstack-usage
5028 Makes the compiler output stack usage information for the program, on a
5029 per-function basis. The filename for the dump is made by appending
5030 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5031 the output file, if explicitly specified and it is not an executable,
5032 otherwise it is the basename of the source file. An entry is made up
5037 The name of the function.
5041 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5044 The qualifier @code{static} means that the function manipulates the stack
5045 statically: a fixed number of bytes are allocated for the frame on function
5046 entry and released on function exit; no stack adjustments are otherwise made
5047 in the function. The second field is this fixed number of bytes.
5049 The qualifier @code{dynamic} means that the function manipulates the stack
5050 dynamically: in addition to the static allocation described above, stack
5051 adjustments are made in the body of the function, for example to push/pop
5052 arguments around function calls. If the qualifier @code{bounded} is also
5053 present, the amount of these adjustments is bounded at compile-time and
5054 the second field is an upper bound of the total amount of stack used by
5055 the function. If it is not present, the amount of these adjustments is
5056 not bounded at compile-time and the second field only represents the
5059 @item -fprofile-arcs
5060 @opindex fprofile-arcs
5061 Add code so that program flow @dfn{arcs} are instrumented. During
5062 execution the program records how many times each branch and call is
5063 executed and how many times it is taken or returns. When the compiled
5064 program exits it saves this data to a file called
5065 @file{@var{auxname}.gcda} for each source file. The data may be used for
5066 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5067 test coverage analysis (@option{-ftest-coverage}). Each object file's
5068 @var{auxname} is generated from the name of the output file, if
5069 explicitly specified and it is not the final executable, otherwise it is
5070 the basename of the source file. In both cases any suffix is removed
5071 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5072 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5073 @xref{Cross-profiling}.
5075 @cindex @command{gcov}
5079 This option is used to compile and link code instrumented for coverage
5080 analysis. The option is a synonym for @option{-fprofile-arcs}
5081 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5082 linking). See the documentation for those options for more details.
5087 Compile the source files with @option{-fprofile-arcs} plus optimization
5088 and code generation options. For test coverage analysis, use the
5089 additional @option{-ftest-coverage} option. You do not need to profile
5090 every source file in a program.
5093 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5094 (the latter implies the former).
5097 Run the program on a representative workload to generate the arc profile
5098 information. This may be repeated any number of times. You can run
5099 concurrent instances of your program, and provided that the file system
5100 supports locking, the data files will be correctly updated. Also
5101 @code{fork} calls are detected and correctly handled (double counting
5105 For profile-directed optimizations, compile the source files again with
5106 the same optimization and code generation options plus
5107 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5108 Control Optimization}).
5111 For test coverage analysis, use @command{gcov} to produce human readable
5112 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5113 @command{gcov} documentation for further information.
5117 With @option{-fprofile-arcs}, for each function of your program GCC
5118 creates a program flow graph, then finds a spanning tree for the graph.
5119 Only arcs that are not on the spanning tree have to be instrumented: the
5120 compiler adds code to count the number of times that these arcs are
5121 executed. When an arc is the only exit or only entrance to a block, the
5122 instrumentation code can be added to the block; otherwise, a new basic
5123 block must be created to hold the instrumentation code.
5126 @item -ftest-coverage
5127 @opindex ftest-coverage
5128 Produce a notes file that the @command{gcov} code-coverage utility
5129 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5130 show program coverage. Each source file's note file is called
5131 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5132 above for a description of @var{auxname} and instructions on how to
5133 generate test coverage data. Coverage data will match the source files
5134 more closely, if you do not optimize.
5136 @item -fdbg-cnt-list
5137 @opindex fdbg-cnt-list
5138 Print the name and the counter upper bound for all debug counters.
5141 @item -fdbg-cnt=@var{counter-value-list}
5143 Set the internal debug counter upper bound. @var{counter-value-list}
5144 is a comma-separated list of @var{name}:@var{value} pairs
5145 which sets the upper bound of each debug counter @var{name} to @var{value}.
5146 All debug counters have the initial upper bound of @var{UINT_MAX},
5147 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5148 e.g. With -fdbg-cnt=dce:10,tail_call:0
5149 dbg_cnt(dce) will return true only for first 10 invocations
5151 @itemx -fenable-@var{kind}-@var{pass}
5152 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5156 This is a set of debugging options that are used to explicitly disable/enable
5157 optimization passes. For compiler users, regular options for enabling/disabling
5158 passes should be used instead.
5162 @item -fdisable-ipa-@var{pass}
5163 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5164 statically invoked in the compiler multiple times, the pass name should be
5165 appended with a sequential number starting from 1.
5167 @item -fdisable-rtl-@var{pass}
5168 @item -fdisable-rtl-@var{pass}=@var{range-list}
5169 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5170 statically invoked in the compiler multiple times, the pass name should be
5171 appended with a sequential number starting from 1. @var{range-list} is a comma
5172 seperated list of function ranges or assembler names. Each range is a number
5173 pair seperated by a colon. The range is inclusive in both ends. If the range
5174 is trivial, the number pair can be simplified as a single number. If the
5175 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5176 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5177 function header of a dump file, and the pass names can be dumped by using
5178 option @option{-fdump-passes}.
5180 @item -fdisable-tree-@var{pass}
5181 @item -fdisable-tree-@var{pass}=@var{range-list}
5182 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5185 @item -fenable-ipa-@var{pass}
5186 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5187 statically invoked in the compiler multiple times, the pass name should be
5188 appended with a sequential number starting from 1.
5190 @item -fenable-rtl-@var{pass}
5191 @item -fenable-rtl-@var{pass}=@var{range-list}
5192 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5193 description and examples.
5195 @item -fenable-tree-@var{pass}
5196 @item -fenable-tree-@var{pass}=@var{range-list}
5197 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5198 of option arguments.
5202 # disable ccp1 for all functions
5204 # disable complete unroll for function whose cgraph node uid is 1
5205 -fenable-tree-cunroll=1
5206 # disable gcse2 for functions at the following ranges [1,1],
5207 # [300,400], and [400,1000]
5208 # disable gcse2 for functions foo and foo2
5209 -fdisable-rtl-gcse2=foo,foo2
5210 # disable early inlining
5211 -fdisable-tree-einline
5212 # disable ipa inlining
5213 -fdisable-ipa-inline
5214 # enable tree full unroll
5215 -fenable-tree-unroll
5221 @item -d@var{letters}
5222 @itemx -fdump-rtl-@var{pass}
5224 Says to make debugging dumps during compilation at times specified by
5225 @var{letters}. This is used for debugging the RTL-based passes of the
5226 compiler. The file names for most of the dumps are made by appending
5227 a pass number and a word to the @var{dumpname}, and the files are
5228 created in the directory of the output file. Note that the pass
5229 number is computed statically as passes get registered into the pass
5230 manager. Thus the numbering is not related to the dynamic order of
5231 execution of passes. In particular, a pass installed by a plugin
5232 could have a number over 200 even if it executed quite early.
5233 @var{dumpname} is generated from the name of the output file, if
5234 explicitly specified and it is not an executable, otherwise it is the
5235 basename of the source file. These switches may have different effects
5236 when @option{-E} is used for preprocessing.
5238 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5239 @option{-d} option @var{letters}. Here are the possible
5240 letters for use in @var{pass} and @var{letters}, and their meanings:
5244 @item -fdump-rtl-alignments
5245 @opindex fdump-rtl-alignments
5246 Dump after branch alignments have been computed.
5248 @item -fdump-rtl-asmcons
5249 @opindex fdump-rtl-asmcons
5250 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5252 @item -fdump-rtl-auto_inc_dec
5253 @opindex fdump-rtl-auto_inc_dec
5254 Dump after auto-inc-dec discovery. This pass is only run on
5255 architectures that have auto inc or auto dec instructions.
5257 @item -fdump-rtl-barriers
5258 @opindex fdump-rtl-barriers
5259 Dump after cleaning up the barrier instructions.
5261 @item -fdump-rtl-bbpart
5262 @opindex fdump-rtl-bbpart
5263 Dump after partitioning hot and cold basic blocks.
5265 @item -fdump-rtl-bbro
5266 @opindex fdump-rtl-bbro
5267 Dump after block reordering.
5269 @item -fdump-rtl-btl1
5270 @itemx -fdump-rtl-btl2
5271 @opindex fdump-rtl-btl2
5272 @opindex fdump-rtl-btl2
5273 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5274 after the two branch
5275 target load optimization passes.
5277 @item -fdump-rtl-bypass
5278 @opindex fdump-rtl-bypass
5279 Dump after jump bypassing and control flow optimizations.
5281 @item -fdump-rtl-combine
5282 @opindex fdump-rtl-combine
5283 Dump after the RTL instruction combination pass.
5285 @item -fdump-rtl-compgotos
5286 @opindex fdump-rtl-compgotos
5287 Dump after duplicating the computed gotos.
5289 @item -fdump-rtl-ce1
5290 @itemx -fdump-rtl-ce2
5291 @itemx -fdump-rtl-ce3
5292 @opindex fdump-rtl-ce1
5293 @opindex fdump-rtl-ce2
5294 @opindex fdump-rtl-ce3
5295 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5296 @option{-fdump-rtl-ce3} enable dumping after the three
5297 if conversion passes.
5299 @itemx -fdump-rtl-cprop_hardreg
5300 @opindex fdump-rtl-cprop_hardreg
5301 Dump after hard register copy propagation.
5303 @itemx -fdump-rtl-csa
5304 @opindex fdump-rtl-csa
5305 Dump after combining stack adjustments.
5307 @item -fdump-rtl-cse1
5308 @itemx -fdump-rtl-cse2
5309 @opindex fdump-rtl-cse1
5310 @opindex fdump-rtl-cse2
5311 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5312 the two common sub-expression elimination passes.
5314 @itemx -fdump-rtl-dce
5315 @opindex fdump-rtl-dce
5316 Dump after the standalone dead code elimination passes.
5318 @itemx -fdump-rtl-dbr
5319 @opindex fdump-rtl-dbr
5320 Dump after delayed branch scheduling.
5322 @item -fdump-rtl-dce1
5323 @itemx -fdump-rtl-dce2
5324 @opindex fdump-rtl-dce1
5325 @opindex fdump-rtl-dce2
5326 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5327 the two dead store elimination passes.
5330 @opindex fdump-rtl-eh
5331 Dump after finalization of EH handling code.
5333 @item -fdump-rtl-eh_ranges
5334 @opindex fdump-rtl-eh_ranges
5335 Dump after conversion of EH handling range regions.
5337 @item -fdump-rtl-expand
5338 @opindex fdump-rtl-expand
5339 Dump after RTL generation.
5341 @item -fdump-rtl-fwprop1
5342 @itemx -fdump-rtl-fwprop2
5343 @opindex fdump-rtl-fwprop1
5344 @opindex fdump-rtl-fwprop2
5345 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5346 dumping after the two forward propagation passes.
5348 @item -fdump-rtl-gcse1
5349 @itemx -fdump-rtl-gcse2
5350 @opindex fdump-rtl-gcse1
5351 @opindex fdump-rtl-gcse2
5352 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5353 after global common subexpression elimination.
5355 @item -fdump-rtl-init-regs
5356 @opindex fdump-rtl-init-regs
5357 Dump after the initialization of the registers.
5359 @item -fdump-rtl-initvals
5360 @opindex fdump-rtl-initvals
5361 Dump after the computation of the initial value sets.
5363 @itemx -fdump-rtl-into_cfglayout
5364 @opindex fdump-rtl-into_cfglayout
5365 Dump after converting to cfglayout mode.
5367 @item -fdump-rtl-ira
5368 @opindex fdump-rtl-ira
5369 Dump after iterated register allocation.
5371 @item -fdump-rtl-jump
5372 @opindex fdump-rtl-jump
5373 Dump after the second jump optimization.
5375 @item -fdump-rtl-loop2
5376 @opindex fdump-rtl-loop2
5377 @option{-fdump-rtl-loop2} enables dumping after the rtl
5378 loop optimization passes.
5380 @item -fdump-rtl-mach
5381 @opindex fdump-rtl-mach
5382 Dump after performing the machine dependent reorganization pass, if that
5385 @item -fdump-rtl-mode_sw
5386 @opindex fdump-rtl-mode_sw
5387 Dump after removing redundant mode switches.
5389 @item -fdump-rtl-rnreg
5390 @opindex fdump-rtl-rnreg
5391 Dump after register renumbering.
5393 @itemx -fdump-rtl-outof_cfglayout
5394 @opindex fdump-rtl-outof_cfglayout
5395 Dump after converting from cfglayout mode.
5397 @item -fdump-rtl-peephole2
5398 @opindex fdump-rtl-peephole2
5399 Dump after the peephole pass.
5401 @item -fdump-rtl-postreload
5402 @opindex fdump-rtl-postreload
5403 Dump after post-reload optimizations.
5405 @itemx -fdump-rtl-pro_and_epilogue
5406 @opindex fdump-rtl-pro_and_epilogue
5407 Dump after generating the function pro and epilogues.
5409 @item -fdump-rtl-regmove
5410 @opindex fdump-rtl-regmove
5411 Dump after the register move pass.
5413 @item -fdump-rtl-sched1
5414 @itemx -fdump-rtl-sched2
5415 @opindex fdump-rtl-sched1
5416 @opindex fdump-rtl-sched2
5417 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5418 after the basic block scheduling passes.
5420 @item -fdump-rtl-see
5421 @opindex fdump-rtl-see
5422 Dump after sign extension elimination.
5424 @item -fdump-rtl-seqabstr
5425 @opindex fdump-rtl-seqabstr
5426 Dump after common sequence discovery.
5428 @item -fdump-rtl-shorten
5429 @opindex fdump-rtl-shorten
5430 Dump after shortening branches.
5432 @item -fdump-rtl-sibling
5433 @opindex fdump-rtl-sibling
5434 Dump after sibling call optimizations.
5436 @item -fdump-rtl-split1
5437 @itemx -fdump-rtl-split2
5438 @itemx -fdump-rtl-split3
5439 @itemx -fdump-rtl-split4
5440 @itemx -fdump-rtl-split5
5441 @opindex fdump-rtl-split1
5442 @opindex fdump-rtl-split2
5443 @opindex fdump-rtl-split3
5444 @opindex fdump-rtl-split4
5445 @opindex fdump-rtl-split5
5446 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5447 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5448 @option{-fdump-rtl-split5} enable dumping after five rounds of
5449 instruction splitting.
5451 @item -fdump-rtl-sms
5452 @opindex fdump-rtl-sms
5453 Dump after modulo scheduling. This pass is only run on some
5456 @item -fdump-rtl-stack
5457 @opindex fdump-rtl-stack
5458 Dump after conversion from GCC's "flat register file" registers to the
5459 x87's stack-like registers. This pass is only run on x86 variants.
5461 @item -fdump-rtl-subreg1
5462 @itemx -fdump-rtl-subreg2
5463 @opindex fdump-rtl-subreg1
5464 @opindex fdump-rtl-subreg2
5465 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5466 the two subreg expansion passes.
5468 @item -fdump-rtl-unshare
5469 @opindex fdump-rtl-unshare
5470 Dump after all rtl has been unshared.
5472 @item -fdump-rtl-vartrack
5473 @opindex fdump-rtl-vartrack
5474 Dump after variable tracking.
5476 @item -fdump-rtl-vregs
5477 @opindex fdump-rtl-vregs
5478 Dump after converting virtual registers to hard registers.
5480 @item -fdump-rtl-web
5481 @opindex fdump-rtl-web
5482 Dump after live range splitting.
5484 @item -fdump-rtl-regclass
5485 @itemx -fdump-rtl-subregs_of_mode_init
5486 @itemx -fdump-rtl-subregs_of_mode_finish
5487 @itemx -fdump-rtl-dfinit
5488 @itemx -fdump-rtl-dfinish
5489 @opindex fdump-rtl-regclass
5490 @opindex fdump-rtl-subregs_of_mode_init
5491 @opindex fdump-rtl-subregs_of_mode_finish
5492 @opindex fdump-rtl-dfinit
5493 @opindex fdump-rtl-dfinish
5494 These dumps are defined but always produce empty files.
5496 @item -fdump-rtl-all
5497 @opindex fdump-rtl-all
5498 Produce all the dumps listed above.
5502 Annotate the assembler output with miscellaneous debugging information.
5506 Dump all macro definitions, at the end of preprocessing, in addition to
5511 Produce a core dump whenever an error occurs.
5515 Print statistics on memory usage, at the end of the run, to
5520 Annotate the assembler output with a comment indicating which
5521 pattern and alternative was used. The length of each instruction is
5526 Dump the RTL in the assembler output as a comment before each instruction.
5527 Also turns on @option{-dp} annotation.
5531 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5532 dump a representation of the control flow graph suitable for viewing with VCG
5533 to @file{@var{file}.@var{pass}.vcg}.
5537 Just generate RTL for a function instead of compiling it. Usually used
5538 with @option{-fdump-rtl-expand}.
5542 @opindex fdump-noaddr
5543 When doing debugging dumps, suppress address output. This makes it more
5544 feasible to use diff on debugging dumps for compiler invocations with
5545 different compiler binaries and/or different
5546 text / bss / data / heap / stack / dso start locations.
5548 @item -fdump-unnumbered
5549 @opindex fdump-unnumbered
5550 When doing debugging dumps, suppress instruction numbers and address output.
5551 This makes it more feasible to use diff on debugging dumps for compiler
5552 invocations with different options, in particular with and without
5555 @item -fdump-unnumbered-links
5556 @opindex fdump-unnumbered-links
5557 When doing debugging dumps (see @option{-d} option above), suppress
5558 instruction numbers for the links to the previous and next instructions
5561 @item -fdump-translation-unit @r{(C++ only)}
5562 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5563 @opindex fdump-translation-unit
5564 Dump a representation of the tree structure for the entire translation
5565 unit to a file. The file name is made by appending @file{.tu} to the
5566 source file name, and the file is created in the same directory as the
5567 output file. If the @samp{-@var{options}} form is used, @var{options}
5568 controls the details of the dump as described for the
5569 @option{-fdump-tree} options.
5571 @item -fdump-class-hierarchy @r{(C++ only)}
5572 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5573 @opindex fdump-class-hierarchy
5574 Dump a representation of each class's hierarchy and virtual function
5575 table layout to a file. The file name is made by appending
5576 @file{.class} to the source file name, and the file is created in the
5577 same directory as the output file. If the @samp{-@var{options}} form
5578 is used, @var{options} controls the details of the dump as described
5579 for the @option{-fdump-tree} options.
5581 @item -fdump-ipa-@var{switch}
5583 Control the dumping at various stages of inter-procedural analysis
5584 language tree to a file. The file name is generated by appending a
5585 switch specific suffix to the source file name, and the file is created
5586 in the same directory as the output file. The following dumps are
5591 Enables all inter-procedural analysis dumps.
5594 Dumps information about call-graph optimization, unused function removal,
5595 and inlining decisions.
5598 Dump after function inlining.
5603 @opindex fdump-passes
5604 Dump the list of optimization passes that are turned on and off by
5605 the current command line options.
5607 @item -fdump-statistics-@var{option}
5608 @opindex fdump-statistics
5609 Enable and control dumping of pass statistics in a separate file. The
5610 file name is generated by appending a suffix ending in
5611 @samp{.statistics} to the source file name, and the file is created in
5612 the same directory as the output file. If the @samp{-@var{option}}
5613 form is used, @samp{-stats} will cause counters to be summed over the
5614 whole compilation unit while @samp{-details} will dump every event as
5615 the passes generate them. The default with no option is to sum
5616 counters for each function compiled.
5618 @item -fdump-tree-@var{switch}
5619 @itemx -fdump-tree-@var{switch}-@var{options}
5621 Control the dumping at various stages of processing the intermediate
5622 language tree to a file. The file name is generated by appending a
5623 switch specific suffix to the source file name, and the file is
5624 created in the same directory as the output file. If the
5625 @samp{-@var{options}} form is used, @var{options} is a list of
5626 @samp{-} separated options that control the details of the dump. Not
5627 all options are applicable to all dumps, those which are not
5628 meaningful will be ignored. The following options are available
5632 Print the address of each node. Usually this is not meaningful as it
5633 changes according to the environment and source file. Its primary use
5634 is for tying up a dump file with a debug environment.
5636 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5637 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5638 use working backward from mangled names in the assembly file.
5640 Inhibit dumping of members of a scope or body of a function merely
5641 because that scope has been reached. Only dump such items when they
5642 are directly reachable by some other path. When dumping pretty-printed
5643 trees, this option inhibits dumping the bodies of control structures.
5645 Print a raw representation of the tree. By default, trees are
5646 pretty-printed into a C-like representation.
5648 Enable more detailed dumps (not honored by every dump option).
5650 Enable dumping various statistics about the pass (not honored by every dump
5653 Enable showing basic block boundaries (disabled in raw dumps).
5655 Enable showing virtual operands for every statement.
5657 Enable showing line numbers for statements.
5659 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5661 Enable showing the tree dump for each statement.
5663 Enable showing the EH region number holding each statement.
5665 Enable showing scalar evolution analysis details.
5667 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5668 and @option{lineno}.
5671 The following tree dumps are possible:
5675 @opindex fdump-tree-original
5676 Dump before any tree based optimization, to @file{@var{file}.original}.
5679 @opindex fdump-tree-optimized
5680 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5683 @opindex fdump-tree-gimple
5684 Dump each function before and after the gimplification pass to a file. The
5685 file name is made by appending @file{.gimple} to the source file name.
5688 @opindex fdump-tree-cfg
5689 Dump the control flow graph of each function to a file. The file name is
5690 made by appending @file{.cfg} to the source file name.
5693 @opindex fdump-tree-vcg
5694 Dump the control flow graph of each function to a file in VCG format. The
5695 file name is made by appending @file{.vcg} to the source file name. Note
5696 that if the file contains more than one function, the generated file cannot
5697 be used directly by VCG@. You will need to cut and paste each function's
5698 graph into its own separate file first.
5701 @opindex fdump-tree-ch
5702 Dump each function after copying loop headers. The file name is made by
5703 appending @file{.ch} to the source file name.
5706 @opindex fdump-tree-ssa
5707 Dump SSA related information to a file. The file name is made by appending
5708 @file{.ssa} to the source file name.
5711 @opindex fdump-tree-alias
5712 Dump aliasing information for each function. The file name is made by
5713 appending @file{.alias} to the source file name.
5716 @opindex fdump-tree-ccp
5717 Dump each function after CCP@. The file name is made by appending
5718 @file{.ccp} to the source file name.
5721 @opindex fdump-tree-storeccp
5722 Dump each function after STORE-CCP@. The file name is made by appending
5723 @file{.storeccp} to the source file name.
5726 @opindex fdump-tree-pre
5727 Dump trees after partial redundancy elimination. The file name is made
5728 by appending @file{.pre} to the source file name.
5731 @opindex fdump-tree-fre
5732 Dump trees after full redundancy elimination. The file name is made
5733 by appending @file{.fre} to the source file name.
5736 @opindex fdump-tree-copyprop
5737 Dump trees after copy propagation. The file name is made
5738 by appending @file{.copyprop} to the source file name.
5740 @item store_copyprop
5741 @opindex fdump-tree-store_copyprop
5742 Dump trees after store copy-propagation. The file name is made
5743 by appending @file{.store_copyprop} to the source file name.
5746 @opindex fdump-tree-dce
5747 Dump each function after dead code elimination. The file name is made by
5748 appending @file{.dce} to the source file name.
5751 @opindex fdump-tree-mudflap
5752 Dump each function after adding mudflap instrumentation. The file name is
5753 made by appending @file{.mudflap} to the source file name.
5756 @opindex fdump-tree-sra
5757 Dump each function after performing scalar replacement of aggregates. The
5758 file name is made by appending @file{.sra} to the source file name.
5761 @opindex fdump-tree-sink
5762 Dump each function after performing code sinking. The file name is made
5763 by appending @file{.sink} to the source file name.
5766 @opindex fdump-tree-dom
5767 Dump each function after applying dominator tree optimizations. The file
5768 name is made by appending @file{.dom} to the source file name.
5771 @opindex fdump-tree-dse
5772 Dump each function after applying dead store elimination. The file
5773 name is made by appending @file{.dse} to the source file name.
5776 @opindex fdump-tree-phiopt
5777 Dump each function after optimizing PHI nodes into straightline code. The file
5778 name is made by appending @file{.phiopt} to the source file name.
5781 @opindex fdump-tree-forwprop
5782 Dump each function after forward propagating single use variables. The file
5783 name is made by appending @file{.forwprop} to the source file name.
5786 @opindex fdump-tree-copyrename
5787 Dump each function after applying the copy rename optimization. The file
5788 name is made by appending @file{.copyrename} to the source file name.
5791 @opindex fdump-tree-nrv
5792 Dump each function after applying the named return value optimization on
5793 generic trees. The file name is made by appending @file{.nrv} to the source
5797 @opindex fdump-tree-vect
5798 Dump each function after applying vectorization of loops. The file name is
5799 made by appending @file{.vect} to the source file name.
5802 @opindex fdump-tree-slp
5803 Dump each function after applying vectorization of basic blocks. The file name
5804 is made by appending @file{.slp} to the source file name.
5807 @opindex fdump-tree-vrp
5808 Dump each function after Value Range Propagation (VRP). The file name
5809 is made by appending @file{.vrp} to the source file name.
5812 @opindex fdump-tree-all
5813 Enable all the available tree dumps with the flags provided in this option.
5816 @item -ftree-vectorizer-verbose=@var{n}
5817 @opindex ftree-vectorizer-verbose
5818 This option controls the amount of debugging output the vectorizer prints.
5819 This information is written to standard error, unless
5820 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5821 in which case it is output to the usual dump listing file, @file{.vect}.
5822 For @var{n}=0 no diagnostic information is reported.
5823 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5824 and the total number of loops that got vectorized.
5825 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5826 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5827 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5828 level that @option{-fdump-tree-vect-stats} uses.
5829 Higher verbosity levels mean either more information dumped for each
5830 reported loop, or same amount of information reported for more loops:
5831 if @var{n}=3, vectorizer cost model information is reported.
5832 If @var{n}=4, alignment related information is added to the reports.
5833 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5834 memory access-patterns) is added to the reports.
5835 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5836 that did not pass the first analysis phase (i.e., may not be countable, or
5837 may have complicated control-flow).
5838 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5839 If @var{n}=8, SLP related information is added to the reports.
5840 For @var{n}=9, all the information the vectorizer generates during its
5841 analysis and transformation is reported. This is the same verbosity level
5842 that @option{-fdump-tree-vect-details} uses.
5844 @item -frandom-seed=@var{string}
5845 @opindex frandom-seed
5846 This option provides a seed that GCC uses when it would otherwise use
5847 random numbers. It is used to generate certain symbol names
5848 that have to be different in every compiled file. It is also used to
5849 place unique stamps in coverage data files and the object files that
5850 produce them. You can use the @option{-frandom-seed} option to produce
5851 reproducibly identical object files.
5853 The @var{string} should be different for every file you compile.
5855 @item -fsched-verbose=@var{n}
5856 @opindex fsched-verbose
5857 On targets that use instruction scheduling, this option controls the
5858 amount of debugging output the scheduler prints. This information is
5859 written to standard error, unless @option{-fdump-rtl-sched1} or
5860 @option{-fdump-rtl-sched2} is specified, in which case it is output
5861 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5862 respectively. However for @var{n} greater than nine, the output is
5863 always printed to standard error.
5865 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5866 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5867 For @var{n} greater than one, it also output basic block probabilities,
5868 detailed ready list information and unit/insn info. For @var{n} greater
5869 than two, it includes RTL at abort point, control-flow and regions info.
5870 And for @var{n} over four, @option{-fsched-verbose} also includes
5874 @itemx -save-temps=cwd
5876 Store the usual ``temporary'' intermediate files permanently; place them
5877 in the current directory and name them based on the source file. Thus,
5878 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5879 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5880 preprocessed @file{foo.i} output file even though the compiler now
5881 normally uses an integrated preprocessor.
5883 When used in combination with the @option{-x} command line option,
5884 @option{-save-temps} is sensible enough to avoid over writing an
5885 input source file with the same extension as an intermediate file.
5886 The corresponding intermediate file may be obtained by renaming the
5887 source file before using @option{-save-temps}.
5889 If you invoke GCC in parallel, compiling several different source
5890 files that share a common base name in different subdirectories or the
5891 same source file compiled for multiple output destinations, it is
5892 likely that the different parallel compilers will interfere with each
5893 other, and overwrite the temporary files. For instance:
5896 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5897 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5900 may result in @file{foo.i} and @file{foo.o} being written to
5901 simultaneously by both compilers.
5903 @item -save-temps=obj
5904 @opindex save-temps=obj
5905 Store the usual ``temporary'' intermediate files permanently. If the
5906 @option{-o} option is used, the temporary files are based on the
5907 object file. If the @option{-o} option is not used, the
5908 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5913 gcc -save-temps=obj -c foo.c
5914 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5915 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5918 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5919 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5920 @file{dir2/yfoobar.o}.
5922 @item -time@r{[}=@var{file}@r{]}
5924 Report the CPU time taken by each subprocess in the compilation
5925 sequence. For C source files, this is the compiler proper and assembler
5926 (plus the linker if linking is done).
5928 Without the specification of an output file, the output looks like this:
5935 The first number on each line is the ``user time'', that is time spent
5936 executing the program itself. The second number is ``system time'',
5937 time spent executing operating system routines on behalf of the program.
5938 Both numbers are in seconds.
5940 With the specification of an output file, the output is appended to the
5941 named file, and it looks like this:
5944 0.12 0.01 cc1 @var{options}
5945 0.00 0.01 as @var{options}
5948 The ``user time'' and the ``system time'' are moved before the program
5949 name, and the options passed to the program are displayed, so that one
5950 can later tell what file was being compiled, and with which options.
5952 @item -fvar-tracking
5953 @opindex fvar-tracking
5954 Run variable tracking pass. It computes where variables are stored at each
5955 position in code. Better debugging information is then generated
5956 (if the debugging information format supports this information).
5958 It is enabled by default when compiling with optimization (@option{-Os},
5959 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5960 the debug info format supports it.
5962 @item -fvar-tracking-assignments
5963 @opindex fvar-tracking-assignments
5964 @opindex fno-var-tracking-assignments
5965 Annotate assignments to user variables early in the compilation and
5966 attempt to carry the annotations over throughout the compilation all the
5967 way to the end, in an attempt to improve debug information while
5968 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5970 It can be enabled even if var-tracking is disabled, in which case
5971 annotations will be created and maintained, but discarded at the end.
5973 @item -fvar-tracking-assignments-toggle
5974 @opindex fvar-tracking-assignments-toggle
5975 @opindex fno-var-tracking-assignments-toggle
5976 Toggle @option{-fvar-tracking-assignments}, in the same way that
5977 @option{-gtoggle} toggles @option{-g}.
5979 @item -print-file-name=@var{library}
5980 @opindex print-file-name
5981 Print the full absolute name of the library file @var{library} that
5982 would be used when linking---and don't do anything else. With this
5983 option, GCC does not compile or link anything; it just prints the
5986 @item -print-multi-directory
5987 @opindex print-multi-directory
5988 Print the directory name corresponding to the multilib selected by any
5989 other switches present in the command line. This directory is supposed
5990 to exist in @env{GCC_EXEC_PREFIX}.
5992 @item -print-multi-lib
5993 @opindex print-multi-lib
5994 Print the mapping from multilib directory names to compiler switches
5995 that enable them. The directory name is separated from the switches by
5996 @samp{;}, and each switch starts with an @samp{@@} instead of the
5997 @samp{-}, without spaces between multiple switches. This is supposed to
5998 ease shell-processing.
6000 @item -print-multi-os-directory
6001 @opindex print-multi-os-directory
6002 Print the path to OS libraries for the selected
6003 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6004 present in the @file{lib} subdirectory and no multilibs are used, this is
6005 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6006 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6007 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6008 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6010 @item -print-prog-name=@var{program}
6011 @opindex print-prog-name
6012 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6014 @item -print-libgcc-file-name
6015 @opindex print-libgcc-file-name
6016 Same as @option{-print-file-name=libgcc.a}.
6018 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6019 but you do want to link with @file{libgcc.a}. You can do
6022 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6025 @item -print-search-dirs
6026 @opindex print-search-dirs
6027 Print the name of the configured installation directory and a list of
6028 program and library directories @command{gcc} will search---and don't do anything else.
6030 This is useful when @command{gcc} prints the error message
6031 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6032 To resolve this you either need to put @file{cpp0} and the other compiler
6033 components where @command{gcc} expects to find them, or you can set the environment
6034 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6035 Don't forget the trailing @samp{/}.
6036 @xref{Environment Variables}.
6038 @item -print-sysroot
6039 @opindex print-sysroot
6040 Print the target sysroot directory that will be used during
6041 compilation. This is the target sysroot specified either at configure
6042 time or using the @option{--sysroot} option, possibly with an extra
6043 suffix that depends on compilation options. If no target sysroot is
6044 specified, the option prints nothing.
6046 @item -print-sysroot-headers-suffix
6047 @opindex print-sysroot-headers-suffix
6048 Print the suffix added to the target sysroot when searching for
6049 headers, or give an error if the compiler is not configured with such
6050 a suffix---and don't do anything else.
6053 @opindex dumpmachine
6054 Print the compiler's target machine (for example,
6055 @samp{i686-pc-linux-gnu})---and don't do anything else.
6058 @opindex dumpversion
6059 Print the compiler version (for example, @samp{3.0})---and don't do
6064 Print the compiler's built-in specs---and don't do anything else. (This
6065 is used when GCC itself is being built.) @xref{Spec Files}.
6067 @item -feliminate-unused-debug-types
6068 @opindex feliminate-unused-debug-types
6069 Normally, when producing DWARF2 output, GCC will emit debugging
6070 information for all types declared in a compilation
6071 unit, regardless of whether or not they are actually used
6072 in that compilation unit. Sometimes this is useful, such as
6073 if, in the debugger, you want to cast a value to a type that is
6074 not actually used in your program (but is declared). More often,
6075 however, this results in a significant amount of wasted space.
6076 With this option, GCC will avoid producing debug symbol output
6077 for types that are nowhere used in the source file being compiled.
6080 @node Optimize Options
6081 @section Options That Control Optimization
6082 @cindex optimize options
6083 @cindex options, optimization
6085 These options control various sorts of optimizations.
6087 Without any optimization option, the compiler's goal is to reduce the
6088 cost of compilation and to make debugging produce the expected
6089 results. Statements are independent: if you stop the program with a
6090 breakpoint between statements, you can then assign a new value to any
6091 variable or change the program counter to any other statement in the
6092 function and get exactly the results you would expect from the source
6095 Turning on optimization flags makes the compiler attempt to improve
6096 the performance and/or code size at the expense of compilation time
6097 and possibly the ability to debug the program.
6099 The compiler performs optimization based on the knowledge it has of the
6100 program. Compiling multiple files at once to a single output file mode allows
6101 the compiler to use information gained from all of the files when compiling
6104 Not all optimizations are controlled directly by a flag. Only
6105 optimizations that have a flag are listed in this section.
6107 Most optimizations are only enabled if an @option{-O} level is set on
6108 the command line. Otherwise they are disabled, even if individual
6109 optimization flags are specified.
6111 Depending on the target and how GCC was configured, a slightly different
6112 set of optimizations may be enabled at each @option{-O} level than
6113 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6114 to find out the exact set of optimizations that are enabled at each level.
6115 @xref{Overall Options}, for examples.
6122 Optimize. Optimizing compilation takes somewhat more time, and a lot
6123 more memory for a large function.
6125 With @option{-O}, the compiler tries to reduce code size and execution
6126 time, without performing any optimizations that take a great deal of
6129 @option{-O} turns on the following optimization flags:
6133 -fcprop-registers @gol
6136 -fdelayed-branch @gol
6138 -fguess-branch-probability @gol
6139 -fif-conversion2 @gol
6140 -fif-conversion @gol
6141 -fipa-pure-const @gol
6143 -fipa-reference @gol
6145 -fsplit-wide-types @gol
6147 -ftree-builtin-call-dce @gol
6150 -ftree-copyrename @gol
6152 -ftree-dominator-opts @gol
6154 -ftree-forwprop @gol
6162 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6163 where doing so does not interfere with debugging.
6167 Optimize even more. GCC performs nearly all supported optimizations
6168 that do not involve a space-speed tradeoff.
6169 As compared to @option{-O}, this option increases both compilation time
6170 and the performance of the generated code.
6172 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6173 also turns on the following optimization flags:
6174 @gccoptlist{-fthread-jumps @gol
6175 -falign-functions -falign-jumps @gol
6176 -falign-loops -falign-labels @gol
6179 -fcse-follow-jumps -fcse-skip-blocks @gol
6180 -fdelete-null-pointer-checks @gol
6182 -fexpensive-optimizations @gol
6183 -fgcse -fgcse-lm @gol
6184 -finline-small-functions @gol
6185 -findirect-inlining @gol
6187 -foptimize-sibling-calls @gol
6188 -fpartial-inlining @gol
6191 -freorder-blocks -freorder-functions @gol
6192 -frerun-cse-after-loop @gol
6193 -fsched-interblock -fsched-spec @gol
6194 -fschedule-insns -fschedule-insns2 @gol
6195 -fstrict-aliasing -fstrict-overflow @gol
6196 -ftree-switch-conversion -ftree-tail-merge @gol
6200 Please note the warning under @option{-fgcse} about
6201 invoking @option{-O2} on programs that use computed gotos.
6205 Optimize yet more. @option{-O3} turns on all optimizations specified
6206 by @option{-O2} and also turns on the @option{-finline-functions},
6207 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6208 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6209 @option{-fipa-cp-clone} options.
6213 Reduce compilation time and make debugging produce the expected
6214 results. This is the default.
6218 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6219 do not typically increase code size. It also performs further
6220 optimizations designed to reduce code size.
6222 @option{-Os} disables the following optimization flags:
6223 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6224 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6225 -fprefetch-loop-arrays -ftree-vect-loop-version}
6229 Disregard strict standards compliance. @option{-Ofast} enables all
6230 @option{-O3} optimizations. It also enables optimizations that are not
6231 valid for all standard compliant programs.
6232 It turns on @option{-ffast-math} and the Fortran-specific
6233 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6235 If you use multiple @option{-O} options, with or without level numbers,
6236 the last such option is the one that is effective.
6239 Options of the form @option{-f@var{flag}} specify machine-independent
6240 flags. Most flags have both positive and negative forms; the negative
6241 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6242 below, only one of the forms is listed---the one you typically will
6243 use. You can figure out the other form by either removing @samp{no-}
6246 The following options control specific optimizations. They are either
6247 activated by @option{-O} options or are related to ones that are. You
6248 can use the following flags in the rare cases when ``fine-tuning'' of
6249 optimizations to be performed is desired.
6252 @item -fno-default-inline
6253 @opindex fno-default-inline
6254 Do not make member functions inline by default merely because they are
6255 defined inside the class scope (C++ only). Otherwise, when you specify
6256 @w{@option{-O}}, member functions defined inside class scope are compiled
6257 inline by default; i.e., you don't need to add @samp{inline} in front of
6258 the member function name.
6260 @item -fno-defer-pop
6261 @opindex fno-defer-pop
6262 Always pop the arguments to each function call as soon as that function
6263 returns. For machines which must pop arguments after a function call,
6264 the compiler normally lets arguments accumulate on the stack for several
6265 function calls and pops them all at once.
6267 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6269 @item -fforward-propagate
6270 @opindex fforward-propagate
6271 Perform a forward propagation pass on RTL@. The pass tries to combine two
6272 instructions and checks if the result can be simplified. If loop unrolling
6273 is active, two passes are performed and the second is scheduled after
6276 This option is enabled by default at optimization levels @option{-O},
6277 @option{-O2}, @option{-O3}, @option{-Os}.
6279 @item -ffp-contract=@var{style}
6280 @opindex ffp-contract
6281 @option{-ffp-contract=off} disables floating-point expression contraction.
6282 @option{-ffp-contract=fast} enables floating-point expression contraction
6283 such as forming of fused multiply-add operations if the target has
6284 native support for them.
6285 @option{-ffp-contract=on} enables floating-point expression contraction
6286 if allowed by the language standard. This is currently not implemented
6287 and treated equal to @option{-ffp-contract=off}.
6289 The default is @option{-ffp-contract=fast}.
6291 @item -fomit-frame-pointer
6292 @opindex fomit-frame-pointer
6293 Don't keep the frame pointer in a register for functions that
6294 don't need one. This avoids the instructions to save, set up and
6295 restore frame pointers; it also makes an extra register available
6296 in many functions. @strong{It also makes debugging impossible on
6299 On some machines, such as the VAX, this flag has no effect, because
6300 the standard calling sequence automatically handles the frame pointer
6301 and nothing is saved by pretending it doesn't exist. The
6302 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6303 whether a target machine supports this flag. @xref{Registers,,Register
6304 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6306 Starting with GCC version 4.6, the default setting (when not optimizing for
6307 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6308 @option{-fomit-frame-pointer}. The default can be reverted to
6309 @option{-fno-omit-frame-pointer} by configuring GCC with the
6310 @option{--enable-frame-pointer} configure option.
6312 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6314 @item -foptimize-sibling-calls
6315 @opindex foptimize-sibling-calls
6316 Optimize sibling and tail recursive calls.
6318 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6322 Don't pay attention to the @code{inline} keyword. Normally this option
6323 is used to keep the compiler from expanding any functions inline.
6324 Note that if you are not optimizing, no functions can be expanded inline.
6326 @item -finline-small-functions
6327 @opindex finline-small-functions
6328 Integrate functions into their callers when their body is smaller than expected
6329 function call code (so overall size of program gets smaller). The compiler
6330 heuristically decides which functions are simple enough to be worth integrating
6333 Enabled at level @option{-O2}.
6335 @item -findirect-inlining
6336 @opindex findirect-inlining
6337 Inline also indirect calls that are discovered to be known at compile
6338 time thanks to previous inlining. This option has any effect only
6339 when inlining itself is turned on by the @option{-finline-functions}
6340 or @option{-finline-small-functions} options.
6342 Enabled at level @option{-O2}.
6344 @item -finline-functions
6345 @opindex finline-functions
6346 Integrate all simple functions into their callers. The compiler
6347 heuristically decides which functions are simple enough to be worth
6348 integrating in this way.
6350 If all calls to a given function are integrated, and the function is
6351 declared @code{static}, then the function is normally not output as
6352 assembler code in its own right.
6354 Enabled at level @option{-O3}.
6356 @item -finline-functions-called-once
6357 @opindex finline-functions-called-once
6358 Consider all @code{static} functions called once for inlining into their
6359 caller even if they are not marked @code{inline}. If a call to a given
6360 function is integrated, then the function is not output as assembler code
6363 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6365 @item -fearly-inlining
6366 @opindex fearly-inlining
6367 Inline functions marked by @code{always_inline} and functions whose body seems
6368 smaller than the function call overhead early before doing
6369 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6370 makes profiling significantly cheaper and usually inlining faster on programs
6371 having large chains of nested wrapper functions.
6377 Perform interprocedural scalar replacement of aggregates, removal of
6378 unused parameters and replacement of parameters passed by reference
6379 by parameters passed by value.
6381 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6383 @item -finline-limit=@var{n}
6384 @opindex finline-limit
6385 By default, GCC limits the size of functions that can be inlined. This flag
6386 allows coarse control of this limit. @var{n} is the size of functions that
6387 can be inlined in number of pseudo instructions.
6389 Inlining is actually controlled by a number of parameters, which may be
6390 specified individually by using @option{--param @var{name}=@var{value}}.
6391 The @option{-finline-limit=@var{n}} option sets some of these parameters
6395 @item max-inline-insns-single
6396 is set to @var{n}/2.
6397 @item max-inline-insns-auto
6398 is set to @var{n}/2.
6401 See below for a documentation of the individual
6402 parameters controlling inlining and for the defaults of these parameters.
6404 @emph{Note:} there may be no value to @option{-finline-limit} that results
6405 in default behavior.
6407 @emph{Note:} pseudo instruction represents, in this particular context, an
6408 abstract measurement of function's size. In no way does it represent a count
6409 of assembly instructions and as such its exact meaning might change from one
6410 release to an another.
6412 @item -fno-keep-inline-dllexport
6413 @opindex -fno-keep-inline-dllexport
6414 This is a more fine-grained version of @option{-fkeep-inline-functions},
6415 which applies only to functions that are declared using the @code{dllexport}
6416 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6419 @item -fkeep-inline-functions
6420 @opindex fkeep-inline-functions
6421 In C, emit @code{static} functions that are declared @code{inline}
6422 into the object file, even if the function has been inlined into all
6423 of its callers. This switch does not affect functions using the
6424 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6425 inline functions into the object file.
6427 @item -fkeep-static-consts
6428 @opindex fkeep-static-consts
6429 Emit variables declared @code{static const} when optimization isn't turned
6430 on, even if the variables aren't referenced.
6432 GCC enables this option by default. If you want to force the compiler to
6433 check if the variable was referenced, regardless of whether or not
6434 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6436 @item -fmerge-constants
6437 @opindex fmerge-constants
6438 Attempt to merge identical constants (string constants and floating point
6439 constants) across compilation units.
6441 This option is the default for optimized compilation if the assembler and
6442 linker support it. Use @option{-fno-merge-constants} to inhibit this
6445 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6447 @item -fmerge-all-constants
6448 @opindex fmerge-all-constants
6449 Attempt to merge identical constants and identical variables.
6451 This option implies @option{-fmerge-constants}. In addition to
6452 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6453 arrays or initialized constant variables with integral or floating point
6454 types. Languages like C or C++ require each variable, including multiple
6455 instances of the same variable in recursive calls, to have distinct locations,
6456 so using this option will result in non-conforming
6459 @item -fmodulo-sched
6460 @opindex fmodulo-sched
6461 Perform swing modulo scheduling immediately before the first scheduling
6462 pass. This pass looks at innermost loops and reorders their
6463 instructions by overlapping different iterations.
6465 @item -fmodulo-sched-allow-regmoves
6466 @opindex fmodulo-sched-allow-regmoves
6467 Perform more aggressive SMS based modulo scheduling with register moves
6468 allowed. By setting this flag certain anti-dependences edges will be
6469 deleted which will trigger the generation of reg-moves based on the
6470 life-range analysis. This option is effective only with
6471 @option{-fmodulo-sched} enabled.
6473 @item -fno-branch-count-reg
6474 @opindex fno-branch-count-reg
6475 Do not use ``decrement and branch'' instructions on a count register,
6476 but instead generate a sequence of instructions that decrement a
6477 register, compare it against zero, then branch based upon the result.
6478 This option is only meaningful on architectures that support such
6479 instructions, which include x86, PowerPC, IA-64 and S/390.
6481 The default is @option{-fbranch-count-reg}.
6483 @item -fno-function-cse
6484 @opindex fno-function-cse
6485 Do not put function addresses in registers; make each instruction that
6486 calls a constant function contain the function's address explicitly.
6488 This option results in less efficient code, but some strange hacks
6489 that alter the assembler output may be confused by the optimizations
6490 performed when this option is not used.
6492 The default is @option{-ffunction-cse}
6494 @item -fno-zero-initialized-in-bss
6495 @opindex fno-zero-initialized-in-bss
6496 If the target supports a BSS section, GCC by default puts variables that
6497 are initialized to zero into BSS@. This can save space in the resulting
6500 This option turns off this behavior because some programs explicitly
6501 rely on variables going to the data section. E.g., so that the
6502 resulting executable can find the beginning of that section and/or make
6503 assumptions based on that.
6505 The default is @option{-fzero-initialized-in-bss}.
6507 @item -fmudflap -fmudflapth -fmudflapir
6511 @cindex bounds checking
6513 For front-ends that support it (C and C++), instrument all risky
6514 pointer/array dereferencing operations, some standard library
6515 string/heap functions, and some other associated constructs with
6516 range/validity tests. Modules so instrumented should be immune to
6517 buffer overflows, invalid heap use, and some other classes of C/C++
6518 programming errors. The instrumentation relies on a separate runtime
6519 library (@file{libmudflap}), which will be linked into a program if
6520 @option{-fmudflap} is given at link time. Run-time behavior of the
6521 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6522 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6525 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6526 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6527 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6528 instrumentation should ignore pointer reads. This produces less
6529 instrumentation (and therefore faster execution) and still provides
6530 some protection against outright memory corrupting writes, but allows
6531 erroneously read data to propagate within a program.
6533 @item -fthread-jumps
6534 @opindex fthread-jumps
6535 Perform optimizations where we check to see if a jump branches to a
6536 location where another comparison subsumed by the first is found. If
6537 so, the first branch is redirected to either the destination of the
6538 second branch or a point immediately following it, depending on whether
6539 the condition is known to be true or false.
6541 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6543 @item -fsplit-wide-types
6544 @opindex fsplit-wide-types
6545 When using a type that occupies multiple registers, such as @code{long
6546 long} on a 32-bit system, split the registers apart and allocate them
6547 independently. This normally generates better code for those types,
6548 but may make debugging more difficult.
6550 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6553 @item -fcse-follow-jumps
6554 @opindex fcse-follow-jumps
6555 In common subexpression elimination (CSE), scan through jump instructions
6556 when the target of the jump is not reached by any other path. For
6557 example, when CSE encounters an @code{if} statement with an
6558 @code{else} clause, CSE will follow the jump when the condition
6561 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6563 @item -fcse-skip-blocks
6564 @opindex fcse-skip-blocks
6565 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6566 follow jumps which conditionally skip over blocks. When CSE
6567 encounters a simple @code{if} statement with no else clause,
6568 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6569 body of the @code{if}.
6571 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6573 @item -frerun-cse-after-loop
6574 @opindex frerun-cse-after-loop
6575 Re-run common subexpression elimination after loop optimizations has been
6578 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6582 Perform a global common subexpression elimination pass.
6583 This pass also performs global constant and copy propagation.
6585 @emph{Note:} When compiling a program using computed gotos, a GCC
6586 extension, you may get better runtime performance if you disable
6587 the global common subexpression elimination pass by adding
6588 @option{-fno-gcse} to the command line.
6590 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6594 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6595 attempt to move loads which are only killed by stores into themselves. This
6596 allows a loop containing a load/store sequence to be changed to a load outside
6597 the loop, and a copy/store within the loop.
6599 Enabled by default when gcse is enabled.
6603 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6604 global common subexpression elimination. This pass will attempt to move
6605 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6606 loops containing a load/store sequence can be changed to a load before
6607 the loop and a store after the loop.
6609 Not enabled at any optimization level.
6613 When @option{-fgcse-las} is enabled, the global common subexpression
6614 elimination pass eliminates redundant loads that come after stores to the
6615 same memory location (both partial and full redundancies).
6617 Not enabled at any optimization level.
6619 @item -fgcse-after-reload
6620 @opindex fgcse-after-reload
6621 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6622 pass is performed after reload. The purpose of this pass is to cleanup
6625 @item -funsafe-loop-optimizations
6626 @opindex funsafe-loop-optimizations
6627 If given, the loop optimizer will assume that loop indices do not
6628 overflow, and that the loops with nontrivial exit condition are not
6629 infinite. This enables a wider range of loop optimizations even if
6630 the loop optimizer itself cannot prove that these assumptions are valid.
6631 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6632 if it finds this kind of loop.
6634 @item -fcrossjumping
6635 @opindex fcrossjumping
6636 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6637 resulting code may or may not perform better than without cross-jumping.
6639 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6641 @item -fauto-inc-dec
6642 @opindex fauto-inc-dec
6643 Combine increments or decrements of addresses with memory accesses.
6644 This pass is always skipped on architectures that do not have
6645 instructions to support this. Enabled by default at @option{-O} and
6646 higher on architectures that support this.
6650 Perform dead code elimination (DCE) on RTL@.
6651 Enabled by default at @option{-O} and higher.
6655 Perform dead store elimination (DSE) on RTL@.
6656 Enabled by default at @option{-O} and higher.
6658 @item -fif-conversion
6659 @opindex fif-conversion
6660 Attempt to transform conditional jumps into branch-less equivalents. This
6661 include use of conditional moves, min, max, set flags and abs instructions, and
6662 some tricks doable by standard arithmetics. The use of conditional execution
6663 on chips where it is available is controlled by @code{if-conversion2}.
6665 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6667 @item -fif-conversion2
6668 @opindex fif-conversion2
6669 Use conditional execution (where available) to transform conditional jumps into
6670 branch-less equivalents.
6672 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6674 @item -fdelete-null-pointer-checks
6675 @opindex fdelete-null-pointer-checks
6676 Assume that programs cannot safely dereference null pointers, and that
6677 no code or data element resides there. This enables simple constant
6678 folding optimizations at all optimization levels. In addition, other
6679 optimization passes in GCC use this flag to control global dataflow
6680 analyses that eliminate useless checks for null pointers; these assume
6681 that if a pointer is checked after it has already been dereferenced,
6684 Note however that in some environments this assumption is not true.
6685 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6686 for programs which depend on that behavior.
6688 Some targets, especially embedded ones, disable this option at all levels.
6689 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6690 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6691 are enabled independently at different optimization levels.
6693 @item -fdevirtualize
6694 @opindex fdevirtualize
6695 Attempt to convert calls to virtual functions to direct calls. This
6696 is done both within a procedure and interprocedurally as part of
6697 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6698 propagation (@option{-fipa-cp}).
6699 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6701 @item -fexpensive-optimizations
6702 @opindex fexpensive-optimizations
6703 Perform a number of minor optimizations that are relatively expensive.
6705 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6707 @item -foptimize-register-move
6709 @opindex foptimize-register-move
6711 Attempt to reassign register numbers in move instructions and as
6712 operands of other simple instructions in order to maximize the amount of
6713 register tying. This is especially helpful on machines with two-operand
6716 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6719 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6721 @item -fira-algorithm=@var{algorithm}
6722 Use specified coloring algorithm for the integrated register
6723 allocator. The @var{algorithm} argument should be @code{priority} or
6724 @code{CB}. The first algorithm specifies Chow's priority coloring,
6725 the second one specifies Chaitin-Briggs coloring. The second
6726 algorithm can be unimplemented for some architectures. If it is
6727 implemented, it is the default because Chaitin-Briggs coloring as a
6728 rule generates a better code.
6730 @item -fira-region=@var{region}
6731 Use specified regions for the integrated register allocator. The
6732 @var{region} argument should be one of @code{all}, @code{mixed}, or
6733 @code{one}. The first value means using all loops as register
6734 allocation regions, the second value which is the default means using
6735 all loops except for loops with small register pressure as the
6736 regions, and third one means using all function as a single region.
6737 The first value can give best result for machines with small size and
6738 irregular register set, the third one results in faster and generates
6739 decent code and the smallest size code, and the default value usually
6740 give the best results in most cases and for most architectures.
6742 @item -fira-loop-pressure
6743 @opindex fira-loop-pressure
6744 Use IRA to evaluate register pressure in loops for decision to move
6745 loop invariants. Usage of this option usually results in generation
6746 of faster and smaller code on machines with big register files (>= 32
6747 registers) but it can slow compiler down.
6749 This option is enabled at level @option{-O3} for some targets.
6751 @item -fno-ira-share-save-slots
6752 @opindex fno-ira-share-save-slots
6753 Switch off sharing stack slots used for saving call used hard
6754 registers living through a call. Each hard register will get a
6755 separate stack slot and as a result function stack frame will be
6758 @item -fno-ira-share-spill-slots
6759 @opindex fno-ira-share-spill-slots
6760 Switch off sharing stack slots allocated for pseudo-registers. Each
6761 pseudo-register which did not get a hard register will get a separate
6762 stack slot and as a result function stack frame will be bigger.
6764 @item -fira-verbose=@var{n}
6765 @opindex fira-verbose
6766 Set up how verbose dump file for the integrated register allocator
6767 will be. Default value is 5. If the value is greater or equal to 10,
6768 the dump file will be stderr as if the value were @var{n} minus 10.
6770 @item -fdelayed-branch
6771 @opindex fdelayed-branch
6772 If supported for the target machine, attempt to reorder instructions
6773 to exploit instruction slots available after delayed branch
6776 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6778 @item -fschedule-insns
6779 @opindex fschedule-insns
6780 If supported for the target machine, attempt to reorder instructions to
6781 eliminate execution stalls due to required data being unavailable. This
6782 helps machines that have slow floating point or memory load instructions
6783 by allowing other instructions to be issued until the result of the load
6784 or floating point instruction is required.
6786 Enabled at levels @option{-O2}, @option{-O3}.
6788 @item -fschedule-insns2
6789 @opindex fschedule-insns2
6790 Similar to @option{-fschedule-insns}, but requests an additional pass of
6791 instruction scheduling after register allocation has been done. This is
6792 especially useful on machines with a relatively small number of
6793 registers and where memory load instructions take more than one cycle.
6795 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6797 @item -fno-sched-interblock
6798 @opindex fno-sched-interblock
6799 Don't schedule instructions across basic blocks. This is normally
6800 enabled by default when scheduling before register allocation, i.e.@:
6801 with @option{-fschedule-insns} or at @option{-O2} or higher.
6803 @item -fno-sched-spec
6804 @opindex fno-sched-spec
6805 Don't allow speculative motion of non-load instructions. This is normally
6806 enabled by default when scheduling before register allocation, i.e.@:
6807 with @option{-fschedule-insns} or at @option{-O2} or higher.
6809 @item -fsched-pressure
6810 @opindex fsched-pressure
6811 Enable register pressure sensitive insn scheduling before the register
6812 allocation. This only makes sense when scheduling before register
6813 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6814 @option{-O2} or higher. Usage of this option can improve the
6815 generated code and decrease its size by preventing register pressure
6816 increase above the number of available hard registers and as a
6817 consequence register spills in the register allocation.
6819 @item -fsched-spec-load
6820 @opindex fsched-spec-load
6821 Allow speculative motion of some load instructions. This only makes
6822 sense when scheduling before register allocation, i.e.@: with
6823 @option{-fschedule-insns} or at @option{-O2} or higher.
6825 @item -fsched-spec-load-dangerous
6826 @opindex fsched-spec-load-dangerous
6827 Allow speculative motion of more load instructions. This only makes
6828 sense when scheduling before register allocation, i.e.@: with
6829 @option{-fschedule-insns} or at @option{-O2} or higher.
6831 @item -fsched-stalled-insns
6832 @itemx -fsched-stalled-insns=@var{n}
6833 @opindex fsched-stalled-insns
6834 Define how many insns (if any) can be moved prematurely from the queue
6835 of stalled insns into the ready list, during the second scheduling pass.
6836 @option{-fno-sched-stalled-insns} means that no insns will be moved
6837 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6838 on how many queued insns can be moved prematurely.
6839 @option{-fsched-stalled-insns} without a value is equivalent to
6840 @option{-fsched-stalled-insns=1}.
6842 @item -fsched-stalled-insns-dep
6843 @itemx -fsched-stalled-insns-dep=@var{n}
6844 @opindex fsched-stalled-insns-dep
6845 Define how many insn groups (cycles) will be examined for a dependency
6846 on a stalled insn that is candidate for premature removal from the queue
6847 of stalled insns. This has an effect only during the second scheduling pass,
6848 and only if @option{-fsched-stalled-insns} is used.
6849 @option{-fno-sched-stalled-insns-dep} is equivalent to
6850 @option{-fsched-stalled-insns-dep=0}.
6851 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6852 @option{-fsched-stalled-insns-dep=1}.
6854 @item -fsched2-use-superblocks
6855 @opindex fsched2-use-superblocks
6856 When scheduling after register allocation, do use superblock scheduling
6857 algorithm. Superblock scheduling allows motion across basic block boundaries
6858 resulting on faster schedules. This option is experimental, as not all machine
6859 descriptions used by GCC model the CPU closely enough to avoid unreliable
6860 results from the algorithm.
6862 This only makes sense when scheduling after register allocation, i.e.@: with
6863 @option{-fschedule-insns2} or at @option{-O2} or higher.
6865 @item -fsched-group-heuristic
6866 @opindex fsched-group-heuristic
6867 Enable the group heuristic in the scheduler. This heuristic favors
6868 the instruction that belongs to a schedule group. This is enabled
6869 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6870 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6872 @item -fsched-critical-path-heuristic
6873 @opindex fsched-critical-path-heuristic
6874 Enable the critical-path heuristic in the scheduler. This heuristic favors
6875 instructions on the critical path. This is enabled by default when
6876 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6877 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6879 @item -fsched-spec-insn-heuristic
6880 @opindex fsched-spec-insn-heuristic
6881 Enable the speculative instruction heuristic in the scheduler. This
6882 heuristic favors speculative instructions with greater dependency weakness.
6883 This is enabled by default when scheduling is enabled, i.e.@:
6884 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6885 or at @option{-O2} or higher.
6887 @item -fsched-rank-heuristic
6888 @opindex fsched-rank-heuristic
6889 Enable the rank heuristic in the scheduler. This heuristic favors
6890 the instruction belonging to a basic block with greater size or frequency.
6891 This is enabled by default when scheduling is enabled, i.e.@:
6892 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6893 at @option{-O2} or higher.
6895 @item -fsched-last-insn-heuristic
6896 @opindex fsched-last-insn-heuristic
6897 Enable the last-instruction heuristic in the scheduler. This heuristic
6898 favors the instruction that is less dependent on the last instruction
6899 scheduled. This is enabled by default when scheduling is enabled,
6900 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6901 at @option{-O2} or higher.
6903 @item -fsched-dep-count-heuristic
6904 @opindex fsched-dep-count-heuristic
6905 Enable the dependent-count heuristic in the scheduler. This heuristic
6906 favors the instruction that has more instructions depending on it.
6907 This is enabled by default when scheduling is enabled, i.e.@:
6908 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6909 at @option{-O2} or higher.
6911 @item -freschedule-modulo-scheduled-loops
6912 @opindex freschedule-modulo-scheduled-loops
6913 The modulo scheduling comes before the traditional scheduling, if a loop
6914 was modulo scheduled we may want to prevent the later scheduling passes
6915 from changing its schedule, we use this option to control that.
6917 @item -fselective-scheduling
6918 @opindex fselective-scheduling
6919 Schedule instructions using selective scheduling algorithm. Selective
6920 scheduling runs instead of the first scheduler pass.
6922 @item -fselective-scheduling2
6923 @opindex fselective-scheduling2
6924 Schedule instructions using selective scheduling algorithm. Selective
6925 scheduling runs instead of the second scheduler pass.
6927 @item -fsel-sched-pipelining
6928 @opindex fsel-sched-pipelining
6929 Enable software pipelining of innermost loops during selective scheduling.
6930 This option has no effect until one of @option{-fselective-scheduling} or
6931 @option{-fselective-scheduling2} is turned on.
6933 @item -fsel-sched-pipelining-outer-loops
6934 @opindex fsel-sched-pipelining-outer-loops
6935 When pipelining loops during selective scheduling, also pipeline outer loops.
6936 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6939 @opindex fshrink-wrap
6940 Emit function prologues only before parts of the function that need it,
6941 rather than at the top of the function. This flag is enabled by default at
6942 @option{-O} and higher.
6944 @item -fcaller-saves
6945 @opindex fcaller-saves
6946 Enable values to be allocated in registers that will be clobbered by
6947 function calls, by emitting extra instructions to save and restore the
6948 registers around such calls. Such allocation is done only when it
6949 seems to result in better code than would otherwise be produced.
6951 This option is always enabled by default on certain machines, usually
6952 those which have no call-preserved registers to use instead.
6954 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6956 @item -fcombine-stack-adjustments
6957 @opindex fcombine-stack-adjustments
6958 Tracks stack adjustments (pushes and pops) and stack memory references
6959 and then tries to find ways to combine them.
6961 Enabled by default at @option{-O1} and higher.
6963 @item -fconserve-stack
6964 @opindex fconserve-stack
6965 Attempt to minimize stack usage. The compiler will attempt to use less
6966 stack space, even if that makes the program slower. This option
6967 implies setting the @option{large-stack-frame} parameter to 100
6968 and the @option{large-stack-frame-growth} parameter to 400.
6970 @item -ftree-reassoc
6971 @opindex ftree-reassoc
6972 Perform reassociation on trees. This flag is enabled by default
6973 at @option{-O} and higher.
6977 Perform partial redundancy elimination (PRE) on trees. This flag is
6978 enabled by default at @option{-O2} and @option{-O3}.
6980 @item -ftree-forwprop
6981 @opindex ftree-forwprop
6982 Perform forward propagation on trees. This flag is enabled by default
6983 at @option{-O} and higher.
6987 Perform full redundancy elimination (FRE) on trees. The difference
6988 between FRE and PRE is that FRE only considers expressions
6989 that are computed on all paths leading to the redundant computation.
6990 This analysis is faster than PRE, though it exposes fewer redundancies.
6991 This flag is enabled by default at @option{-O} and higher.
6993 @item -ftree-phiprop
6994 @opindex ftree-phiprop
6995 Perform hoisting of loads from conditional pointers on trees. This
6996 pass is enabled by default at @option{-O} and higher.
6998 @item -ftree-copy-prop
6999 @opindex ftree-copy-prop
7000 Perform copy propagation on trees. This pass eliminates unnecessary
7001 copy operations. This flag is enabled by default at @option{-O} and
7004 @item -fipa-pure-const
7005 @opindex fipa-pure-const
7006 Discover which functions are pure or constant.
7007 Enabled by default at @option{-O} and higher.
7009 @item -fipa-reference
7010 @opindex fipa-reference
7011 Discover which static variables do not escape cannot escape the
7013 Enabled by default at @option{-O} and higher.
7017 Perform interprocedural pointer analysis and interprocedural modification
7018 and reference analysis. This option can cause excessive memory and
7019 compile-time usage on large compilation units. It is not enabled by
7020 default at any optimization level.
7023 @opindex fipa-profile
7024 Perform interprocedural profile propagation. The functions called only from
7025 cold functions are marked as cold. Also functions executed once (such as
7026 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7027 functions and loop less parts of functions executed once are then optimized for
7029 Enabled by default at @option{-O} and higher.
7033 Perform interprocedural constant propagation.
7034 This optimization analyzes the program to determine when values passed
7035 to functions are constants and then optimizes accordingly.
7036 This optimization can substantially increase performance
7037 if the application has constants passed to functions.
7038 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7040 @item -fipa-cp-clone
7041 @opindex fipa-cp-clone
7042 Perform function cloning to make interprocedural constant propagation stronger.
7043 When enabled, interprocedural constant propagation will perform function cloning
7044 when externally visible function can be called with constant arguments.
7045 Because this optimization can create multiple copies of functions,
7046 it may significantly increase code size
7047 (see @option{--param ipcp-unit-growth=@var{value}}).
7048 This flag is enabled by default at @option{-O3}.
7050 @item -fipa-matrix-reorg
7051 @opindex fipa-matrix-reorg
7052 Perform matrix flattening and transposing.
7053 Matrix flattening tries to replace an @math{m}-dimensional matrix
7054 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7055 This reduces the level of indirection needed for accessing the elements
7056 of the matrix. The second optimization is matrix transposing that
7057 attempts to change the order of the matrix's dimensions in order to
7058 improve cache locality.
7059 Both optimizations need the @option{-fwhole-program} flag.
7060 Transposing is enabled only if profiling information is available.
7064 Perform forward store motion on trees. This flag is
7065 enabled by default at @option{-O} and higher.
7067 @item -ftree-bit-ccp
7068 @opindex ftree-bit-ccp
7069 Perform sparse conditional bit constant propagation on trees and propagate
7070 pointer alignment information.
7071 This pass only operates on local scalar variables and is enabled by default
7072 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7076 Perform sparse conditional constant propagation (CCP) on trees. This
7077 pass only operates on local scalar variables and is enabled by default
7078 at @option{-O} and higher.
7080 @item -ftree-switch-conversion
7081 Perform conversion of simple initializations in a switch to
7082 initializations from a scalar array. This flag is enabled by default
7083 at @option{-O2} and higher.
7085 @item -ftree-tail-merge
7086 Look for identical code sequences. When found, replace one with a jump to the
7087 other. This optimization is known as tail merging or cross jumping. This flag
7088 is enabled by default at @option{-O2} and higher. The run time of this pass can
7089 be limited using @option{max-tail-merge-comparisons} parameter and
7090 @option{max-tail-merge-iterations} parameter.
7094 Perform dead code elimination (DCE) on trees. This flag is enabled by
7095 default at @option{-O} and higher.
7097 @item -ftree-builtin-call-dce
7098 @opindex ftree-builtin-call-dce
7099 Perform conditional dead code elimination (DCE) for calls to builtin functions
7100 that may set @code{errno} but are otherwise side-effect free. This flag is
7101 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7104 @item -ftree-dominator-opts
7105 @opindex ftree-dominator-opts
7106 Perform a variety of simple scalar cleanups (constant/copy
7107 propagation, redundancy elimination, range propagation and expression
7108 simplification) based on a dominator tree traversal. This also
7109 performs jump threading (to reduce jumps to jumps). This flag is
7110 enabled by default at @option{-O} and higher.
7114 Perform dead store elimination (DSE) on trees. A dead store is a store into
7115 a memory location which will later be overwritten by another store without
7116 any intervening loads. In this case the earlier store can be deleted. This
7117 flag is enabled by default at @option{-O} and higher.
7121 Perform loop header copying on trees. This is beneficial since it increases
7122 effectiveness of code motion optimizations. It also saves one jump. This flag
7123 is enabled by default at @option{-O} and higher. It is not enabled
7124 for @option{-Os}, since it usually increases code size.
7126 @item -ftree-loop-optimize
7127 @opindex ftree-loop-optimize
7128 Perform loop optimizations on trees. This flag is enabled by default
7129 at @option{-O} and higher.
7131 @item -ftree-loop-linear
7132 @opindex ftree-loop-linear
7133 Perform loop interchange transformations on tree. Same as
7134 @option{-floop-interchange}. To use this code transformation, GCC has
7135 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7136 enable the Graphite loop transformation infrastructure.
7138 @item -floop-interchange
7139 @opindex floop-interchange
7140 Perform loop interchange transformations on loops. Interchanging two
7141 nested loops switches the inner and outer loops. For example, given a
7146 A(J, I) = A(J, I) * C
7150 loop interchange will transform the loop as if the user had written:
7154 A(J, I) = A(J, I) * C
7158 which can be beneficial when @code{N} is larger than the caches,
7159 because in Fortran, the elements of an array are stored in memory
7160 contiguously by column, and the original loop iterates over rows,
7161 potentially creating at each access a cache miss. This optimization
7162 applies to all the languages supported by GCC and is not limited to
7163 Fortran. To use this code transformation, GCC has to be configured
7164 with @option{--with-ppl} and @option{--with-cloog} to enable the
7165 Graphite loop transformation infrastructure.
7167 @item -floop-strip-mine
7168 @opindex floop-strip-mine
7169 Perform loop strip mining transformations on loops. Strip mining
7170 splits a loop into two nested loops. The outer loop has strides
7171 equal to the strip size and the inner loop has strides of the
7172 original loop within a strip. The strip length can be changed
7173 using the @option{loop-block-tile-size} parameter. For example,
7180 loop strip mining will transform the loop as if the user had written:
7183 DO I = II, min (II + 50, N)
7188 This optimization applies to all the languages supported by GCC and is
7189 not limited to Fortran. To use this code transformation, GCC has to
7190 be configured with @option{--with-ppl} and @option{--with-cloog} to
7191 enable the Graphite loop transformation infrastructure.
7194 @opindex floop-block
7195 Perform loop blocking transformations on loops. Blocking strip mines
7196 each loop in the loop nest such that the memory accesses of the
7197 element loops fit inside caches. The strip length can be changed
7198 using the @option{loop-block-tile-size} parameter. For example, given
7203 A(J, I) = B(I) + C(J)
7207 loop blocking will transform the loop as if the user had written:
7211 DO I = II, min (II + 50, N)
7212 DO J = JJ, min (JJ + 50, M)
7213 A(J, I) = B(I) + C(J)
7219 which can be beneficial when @code{M} is larger than the caches,
7220 because the innermost loop will iterate over a smaller amount of data
7221 that can be kept in the caches. This optimization applies to all the
7222 languages supported by GCC and is not limited to Fortran. To use this
7223 code transformation, GCC has to be configured with @option{--with-ppl}
7224 and @option{--with-cloog} to enable the Graphite loop transformation
7227 @item -fgraphite-identity
7228 @opindex fgraphite-identity
7229 Enable the identity transformation for graphite. For every SCoP we generate
7230 the polyhedral representation and transform it back to gimple. Using
7231 @option{-fgraphite-identity} we can check the costs or benefits of the
7232 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7233 are also performed by the code generator CLooG, like index splitting and
7234 dead code elimination in loops.
7236 @item -floop-flatten
7237 @opindex floop-flatten
7238 Removes the loop nesting structure: transforms the loop nest into a
7239 single loop. This transformation can be useful to vectorize all the
7240 levels of the loop nest.
7242 @item -floop-parallelize-all
7243 @opindex floop-parallelize-all
7244 Use the Graphite data dependence analysis to identify loops that can
7245 be parallelized. Parallelize all the loops that can be analyzed to
7246 not contain loop carried dependences without checking that it is
7247 profitable to parallelize the loops.
7249 @item -fcheck-data-deps
7250 @opindex fcheck-data-deps
7251 Compare the results of several data dependence analyzers. This option
7252 is used for debugging the data dependence analyzers.
7254 @item -ftree-loop-if-convert
7255 Attempt to transform conditional jumps in the innermost loops to
7256 branch-less equivalents. The intent is to remove control-flow from
7257 the innermost loops in order to improve the ability of the
7258 vectorization pass to handle these loops. This is enabled by default
7259 if vectorization is enabled.
7261 @item -ftree-loop-if-convert-stores
7262 Attempt to also if-convert conditional jumps containing memory writes.
7263 This transformation can be unsafe for multi-threaded programs as it
7264 transforms conditional memory writes into unconditional memory writes.
7267 for (i = 0; i < N; i++)
7271 would be transformed to
7273 for (i = 0; i < N; i++)
7274 A[i] = cond ? expr : A[i];
7276 potentially producing data races.
7278 @item -ftree-loop-distribution
7279 Perform loop distribution. This flag can improve cache performance on
7280 big loop bodies and allow further loop optimizations, like
7281 parallelization or vectorization, to take place. For example, the loop
7298 @item -ftree-loop-distribute-patterns
7299 Perform loop distribution of patterns that can be code generated with
7300 calls to a library. This flag is enabled by default at @option{-O3}.
7302 This pass distributes the initialization loops and generates a call to
7303 memset zero. For example, the loop
7319 and the initialization loop is transformed into a call to memset zero.
7321 @item -ftree-loop-im
7322 @opindex ftree-loop-im
7323 Perform loop invariant motion on trees. This pass moves only invariants that
7324 would be hard to handle at RTL level (function calls, operations that expand to
7325 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7326 operands of conditions that are invariant out of the loop, so that we can use
7327 just trivial invariantness analysis in loop unswitching. The pass also includes
7330 @item -ftree-loop-ivcanon
7331 @opindex ftree-loop-ivcanon
7332 Create a canonical counter for number of iterations in the loop for that
7333 determining number of iterations requires complicated analysis. Later
7334 optimizations then may determine the number easily. Useful especially
7335 in connection with unrolling.
7339 Perform induction variable optimizations (strength reduction, induction
7340 variable merging and induction variable elimination) on trees.
7342 @item -ftree-parallelize-loops=n
7343 @opindex ftree-parallelize-loops
7344 Parallelize loops, i.e., split their iteration space to run in n threads.
7345 This is only possible for loops whose iterations are independent
7346 and can be arbitrarily reordered. The optimization is only
7347 profitable on multiprocessor machines, for loops that are CPU-intensive,
7348 rather than constrained e.g.@: by memory bandwidth. This option
7349 implies @option{-pthread}, and thus is only supported on targets
7350 that have support for @option{-pthread}.
7354 Perform function-local points-to analysis on trees. This flag is
7355 enabled by default at @option{-O} and higher.
7359 Perform scalar replacement of aggregates. This pass replaces structure
7360 references with scalars to prevent committing structures to memory too
7361 early. This flag is enabled by default at @option{-O} and higher.
7363 @item -ftree-copyrename
7364 @opindex ftree-copyrename
7365 Perform copy renaming on trees. This pass attempts to rename compiler
7366 temporaries to other variables at copy locations, usually resulting in
7367 variable names which more closely resemble the original variables. This flag
7368 is enabled by default at @option{-O} and higher.
7372 Perform temporary expression replacement during the SSA->normal phase. Single
7373 use/single def temporaries are replaced at their use location with their
7374 defining expression. This results in non-GIMPLE code, but gives the expanders
7375 much more complex trees to work on resulting in better RTL generation. This is
7376 enabled by default at @option{-O} and higher.
7378 @item -ftree-vectorize
7379 @opindex ftree-vectorize
7380 Perform loop vectorization on trees. This flag is enabled by default at
7383 @item -ftree-slp-vectorize
7384 @opindex ftree-slp-vectorize
7385 Perform basic block vectorization on trees. This flag is enabled by default at
7386 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7388 @item -ftree-vect-loop-version
7389 @opindex ftree-vect-loop-version
7390 Perform loop versioning when doing loop vectorization on trees. When a loop
7391 appears to be vectorizable except that data alignment or data dependence cannot
7392 be determined at compile time then vectorized and non-vectorized versions of
7393 the loop are generated along with runtime checks for alignment or dependence
7394 to control which version is executed. This option is enabled by default
7395 except at level @option{-Os} where it is disabled.
7397 @item -fvect-cost-model
7398 @opindex fvect-cost-model
7399 Enable cost model for vectorization.
7403 Perform Value Range Propagation on trees. This is similar to the
7404 constant propagation pass, but instead of values, ranges of values are
7405 propagated. This allows the optimizers to remove unnecessary range
7406 checks like array bound checks and null pointer checks. This is
7407 enabled by default at @option{-O2} and higher. Null pointer check
7408 elimination is only done if @option{-fdelete-null-pointer-checks} is
7413 Perform tail duplication to enlarge superblock size. This transformation
7414 simplifies the control flow of the function allowing other optimizations to do
7417 @item -funroll-loops
7418 @opindex funroll-loops
7419 Unroll loops whose number of iterations can be determined at compile
7420 time or upon entry to the loop. @option{-funroll-loops} implies
7421 @option{-frerun-cse-after-loop}. This option makes code larger,
7422 and may or may not make it run faster.
7424 @item -funroll-all-loops
7425 @opindex funroll-all-loops
7426 Unroll all loops, even if their number of iterations is uncertain when
7427 the loop is entered. This usually makes programs run more slowly.
7428 @option{-funroll-all-loops} implies the same options as
7429 @option{-funroll-loops},
7431 @item -fsplit-ivs-in-unroller
7432 @opindex fsplit-ivs-in-unroller
7433 Enables expressing of values of induction variables in later iterations
7434 of the unrolled loop using the value in the first iteration. This breaks
7435 long dependency chains, thus improving efficiency of the scheduling passes.
7437 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7438 same effect. However in cases the loop body is more complicated than
7439 a single basic block, this is not reliable. It also does not work at all
7440 on some of the architectures due to restrictions in the CSE pass.
7442 This optimization is enabled by default.
7444 @item -fvariable-expansion-in-unroller
7445 @opindex fvariable-expansion-in-unroller
7446 With this option, the compiler will create multiple copies of some
7447 local variables when unrolling a loop which can result in superior code.
7449 @item -fpartial-inlining
7450 @opindex fpartial-inlining
7451 Inline parts of functions. This option has any effect only
7452 when inlining itself is turned on by the @option{-finline-functions}
7453 or @option{-finline-small-functions} options.
7455 Enabled at level @option{-O2}.
7457 @item -fpredictive-commoning
7458 @opindex fpredictive-commoning
7459 Perform predictive commoning optimization, i.e., reusing computations
7460 (especially memory loads and stores) performed in previous
7461 iterations of loops.
7463 This option is enabled at level @option{-O3}.
7465 @item -fprefetch-loop-arrays
7466 @opindex fprefetch-loop-arrays
7467 If supported by the target machine, generate instructions to prefetch
7468 memory to improve the performance of loops that access large arrays.
7470 This option may generate better or worse code; results are highly
7471 dependent on the structure of loops within the source code.
7473 Disabled at level @option{-Os}.
7476 @itemx -fno-peephole2
7477 @opindex fno-peephole
7478 @opindex fno-peephole2
7479 Disable any machine-specific peephole optimizations. The difference
7480 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7481 are implemented in the compiler; some targets use one, some use the
7482 other, a few use both.
7484 @option{-fpeephole} is enabled by default.
7485 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7487 @item -fno-guess-branch-probability
7488 @opindex fno-guess-branch-probability
7489 Do not guess branch probabilities using heuristics.
7491 GCC will use heuristics to guess branch probabilities if they are
7492 not provided by profiling feedback (@option{-fprofile-arcs}). These
7493 heuristics are based on the control flow graph. If some branch probabilities
7494 are specified by @samp{__builtin_expect}, then the heuristics will be
7495 used to guess branch probabilities for the rest of the control flow graph,
7496 taking the @samp{__builtin_expect} info into account. The interactions
7497 between the heuristics and @samp{__builtin_expect} can be complex, and in
7498 some cases, it may be useful to disable the heuristics so that the effects
7499 of @samp{__builtin_expect} are easier to understand.
7501 The default is @option{-fguess-branch-probability} at levels
7502 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7504 @item -freorder-blocks
7505 @opindex freorder-blocks
7506 Reorder basic blocks in the compiled function in order to reduce number of
7507 taken branches and improve code locality.
7509 Enabled at levels @option{-O2}, @option{-O3}.
7511 @item -freorder-blocks-and-partition
7512 @opindex freorder-blocks-and-partition
7513 In addition to reordering basic blocks in the compiled function, in order
7514 to reduce number of taken branches, partitions hot and cold basic blocks
7515 into separate sections of the assembly and .o files, to improve
7516 paging and cache locality performance.
7518 This optimization is automatically turned off in the presence of
7519 exception handling, for linkonce sections, for functions with a user-defined
7520 section attribute and on any architecture that does not support named
7523 @item -freorder-functions
7524 @opindex freorder-functions
7525 Reorder functions in the object file in order to
7526 improve code locality. This is implemented by using special
7527 subsections @code{.text.hot} for most frequently executed functions and
7528 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7529 the linker so object file format must support named sections and linker must
7530 place them in a reasonable way.
7532 Also profile feedback must be available in to make this option effective. See
7533 @option{-fprofile-arcs} for details.
7535 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7537 @item -fstrict-aliasing
7538 @opindex fstrict-aliasing
7539 Allow the compiler to assume the strictest aliasing rules applicable to
7540 the language being compiled. For C (and C++), this activates
7541 optimizations based on the type of expressions. In particular, an
7542 object of one type is assumed never to reside at the same address as an
7543 object of a different type, unless the types are almost the same. For
7544 example, an @code{unsigned int} can alias an @code{int}, but not a
7545 @code{void*} or a @code{double}. A character type may alias any other
7548 @anchor{Type-punning}Pay special attention to code like this:
7561 The practice of reading from a different union member than the one most
7562 recently written to (called ``type-punning'') is common. Even with
7563 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7564 is accessed through the union type. So, the code above will work as
7565 expected. @xref{Structures unions enumerations and bit-fields
7566 implementation}. However, this code might not:
7577 Similarly, access by taking the address, casting the resulting pointer
7578 and dereferencing the result has undefined behavior, even if the cast
7579 uses a union type, e.g.:
7583 return ((union a_union *) &d)->i;
7587 The @option{-fstrict-aliasing} option is enabled at levels
7588 @option{-O2}, @option{-O3}, @option{-Os}.
7590 @item -fstrict-overflow
7591 @opindex fstrict-overflow
7592 Allow the compiler to assume strict signed overflow rules, depending
7593 on the language being compiled. For C (and C++) this means that
7594 overflow when doing arithmetic with signed numbers is undefined, which
7595 means that the compiler may assume that it will not happen. This
7596 permits various optimizations. For example, the compiler will assume
7597 that an expression like @code{i + 10 > i} will always be true for
7598 signed @code{i}. This assumption is only valid if signed overflow is
7599 undefined, as the expression is false if @code{i + 10} overflows when
7600 using twos complement arithmetic. When this option is in effect any
7601 attempt to determine whether an operation on signed numbers will
7602 overflow must be written carefully to not actually involve overflow.
7604 This option also allows the compiler to assume strict pointer
7605 semantics: given a pointer to an object, if adding an offset to that
7606 pointer does not produce a pointer to the same object, the addition is
7607 undefined. This permits the compiler to conclude that @code{p + u >
7608 p} is always true for a pointer @code{p} and unsigned integer
7609 @code{u}. This assumption is only valid because pointer wraparound is
7610 undefined, as the expression is false if @code{p + u} overflows using
7611 twos complement arithmetic.
7613 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7614 that integer signed overflow is fully defined: it wraps. When
7615 @option{-fwrapv} is used, there is no difference between
7616 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7617 integers. With @option{-fwrapv} certain types of overflow are
7618 permitted. For example, if the compiler gets an overflow when doing
7619 arithmetic on constants, the overflowed value can still be used with
7620 @option{-fwrapv}, but not otherwise.
7622 The @option{-fstrict-overflow} option is enabled at levels
7623 @option{-O2}, @option{-O3}, @option{-Os}.
7625 @item -falign-functions
7626 @itemx -falign-functions=@var{n}
7627 @opindex falign-functions
7628 Align the start of functions to the next power-of-two greater than
7629 @var{n}, skipping up to @var{n} bytes. For instance,
7630 @option{-falign-functions=32} aligns functions to the next 32-byte
7631 boundary, but @option{-falign-functions=24} would align to the next
7632 32-byte boundary only if this can be done by skipping 23 bytes or less.
7634 @option{-fno-align-functions} and @option{-falign-functions=1} are
7635 equivalent and mean that functions will not be aligned.
7637 Some assemblers only support this flag when @var{n} is a power of two;
7638 in that case, it is rounded up.
7640 If @var{n} is not specified or is zero, use a machine-dependent default.
7642 Enabled at levels @option{-O2}, @option{-O3}.
7644 @item -falign-labels
7645 @itemx -falign-labels=@var{n}
7646 @opindex falign-labels
7647 Align all branch targets to a power-of-two boundary, skipping up to
7648 @var{n} bytes like @option{-falign-functions}. This option can easily
7649 make code slower, because it must insert dummy operations for when the
7650 branch target is reached in the usual flow of the code.
7652 @option{-fno-align-labels} and @option{-falign-labels=1} are
7653 equivalent and mean that labels will not be aligned.
7655 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7656 are greater than this value, then their values are used instead.
7658 If @var{n} is not specified or is zero, use a machine-dependent default
7659 which is very likely to be @samp{1}, meaning no alignment.
7661 Enabled at levels @option{-O2}, @option{-O3}.
7664 @itemx -falign-loops=@var{n}
7665 @opindex falign-loops
7666 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7667 like @option{-falign-functions}. The hope is that the loop will be
7668 executed many times, which will make up for any execution of the dummy
7671 @option{-fno-align-loops} and @option{-falign-loops=1} are
7672 equivalent and mean that loops will not be aligned.
7674 If @var{n} is not specified or is zero, use a machine-dependent default.
7676 Enabled at levels @option{-O2}, @option{-O3}.
7679 @itemx -falign-jumps=@var{n}
7680 @opindex falign-jumps
7681 Align branch targets to a power-of-two boundary, for branch targets
7682 where the targets can only be reached by jumping, skipping up to @var{n}
7683 bytes like @option{-falign-functions}. In this case, no dummy operations
7686 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7687 equivalent and mean that loops will not be aligned.
7689 If @var{n} is not specified or is zero, use a machine-dependent default.
7691 Enabled at levels @option{-O2}, @option{-O3}.
7693 @item -funit-at-a-time
7694 @opindex funit-at-a-time
7695 This option is left for compatibility reasons. @option{-funit-at-a-time}
7696 has no effect, while @option{-fno-unit-at-a-time} implies
7697 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7701 @item -fno-toplevel-reorder
7702 @opindex fno-toplevel-reorder
7703 Do not reorder top-level functions, variables, and @code{asm}
7704 statements. Output them in the same order that they appear in the
7705 input file. When this option is used, unreferenced static variables
7706 will not be removed. This option is intended to support existing code
7707 which relies on a particular ordering. For new code, it is better to
7710 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7711 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7716 Constructs webs as commonly used for register allocation purposes and assign
7717 each web individual pseudo register. This allows the register allocation pass
7718 to operate on pseudos directly, but also strengthens several other optimization
7719 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7720 however, make debugging impossible, since variables will no longer stay in a
7723 Enabled by default with @option{-funroll-loops}.
7725 @item -fwhole-program
7726 @opindex fwhole-program
7727 Assume that the current compilation unit represents the whole program being
7728 compiled. All public functions and variables with the exception of @code{main}
7729 and those merged by attribute @code{externally_visible} become static functions
7730 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.
7731 While this option is equivalent to proper use of the @code{static} keyword for
7732 programs consisting of a single file, in combination with option
7733 @option{-flto} this flag can be used to
7734 compile many smaller scale programs since the functions and variables become
7735 local for the whole combined compilation unit, not for the single source file
7738 This option implies @option{-fwhole-file} for Fortran programs.
7740 @item -flto[=@var{n}]
7742 This option runs the standard link-time optimizer. When invoked
7743 with source code, it generates GIMPLE (one of GCC's internal
7744 representations) and writes it to special ELF sections in the object
7745 file. When the object files are linked together, all the function
7746 bodies are read from these ELF sections and instantiated as if they
7747 had been part of the same translation unit.
7749 To use the link-timer optimizer, @option{-flto} needs to be specified at
7750 compile time and during the final link. For example,
7753 gcc -c -O2 -flto foo.c
7754 gcc -c -O2 -flto bar.c
7755 gcc -o myprog -flto -O2 foo.o bar.o
7758 The first two invocations to GCC will save a bytecode representation
7759 of GIMPLE into special ELF sections inside @file{foo.o} and
7760 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7761 @file{foo.o} and @file{bar.o}, merge the two files into a single
7762 internal image, and compile the result as usual. Since both
7763 @file{foo.o} and @file{bar.o} are merged into a single image, this
7764 causes all the inter-procedural analyses and optimizations in GCC to
7765 work across the two files as if they were a single one. This means,
7766 for example, that the inliner will be able to inline functions in
7767 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7769 Another (simpler) way to enable link-time optimization is,
7772 gcc -o myprog -flto -O2 foo.c bar.c
7775 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7776 merge them together into a single GIMPLE representation and optimize
7777 them as usual to produce @file{myprog}.
7779 The only important thing to keep in mind is that to enable link-time
7780 optimizations the @option{-flto} flag needs to be passed to both the
7781 compile and the link commands.
7783 To make whole program optimization effective, it is necessary to make
7784 certain whole program assumptions. The compiler needs to know
7785 what functions and variables can be accessed by libraries and runtime
7786 outside of the link time optimized unit. When supported by the linker,
7787 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7788 compiler information about used and externally visible symbols. When
7789 the linker plugin is not available, @option{-fwhole-program} should be
7790 used to allow the compiler to make these assumptions, which will lead
7791 to more aggressive optimization decisions.
7793 Note that when a file is compiled with @option{-flto}, the generated
7794 object file will be larger than a regular object file because it will
7795 contain GIMPLE bytecodes and the usual final code. This means that
7796 object files with LTO information can be linked as a normal object
7797 file. So, in the previous example, if the final link is done with
7800 gcc -o myprog foo.o bar.o
7803 The only difference will be that no inter-procedural optimizations
7804 will be applied to produce @file{myprog}. The two object files
7805 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7808 Additionally, the optimization flags used to compile individual files
7809 are not necessarily related to those used at link-time. For instance,
7812 gcc -c -O0 -flto foo.c
7813 gcc -c -O0 -flto bar.c
7814 gcc -o myprog -flto -O3 foo.o bar.o
7817 This will produce individual object files with unoptimized assembler
7818 code, but the resulting binary @file{myprog} will be optimized at
7819 @option{-O3}. Now, if the final binary is generated without
7820 @option{-flto}, then @file{myprog} will not be optimized.
7822 When producing the final binary with @option{-flto}, GCC will only
7823 apply link-time optimizations to those files that contain bytecode.
7824 Therefore, you can mix and match object files and libraries with
7825 GIMPLE bytecodes and final object code. GCC will automatically select
7826 which files to optimize in LTO mode and which files to link without
7829 There are some code generation flags that GCC will preserve when
7830 generating bytecodes, as they need to be used during the final link
7831 stage. Currently, the following options are saved into the GIMPLE
7832 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7833 @option{-m} target flags.
7835 At link time, these options are read-in and reapplied. Note that the
7836 current implementation makes no attempt at recognizing conflicting
7837 values for these options. If two or more files have a conflicting
7838 value (e.g., one file is compiled with @option{-fPIC} and another
7839 isn't), the compiler will simply use the last value read from the
7840 bytecode files. It is recommended, then, that all the files
7841 participating in the same link be compiled with the same options.
7843 Another feature of LTO is that it is possible to apply interprocedural
7844 optimizations on files written in different languages. This requires
7845 some support in the language front end. Currently, the C, C++ and
7846 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7847 something like this should work
7852 gfortran -c -flto baz.f90
7853 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7856 Notice that the final link is done with @command{g++} to get the C++
7857 runtime libraries and @option{-lgfortran} is added to get the Fortran
7858 runtime libraries. In general, when mixing languages in LTO mode, you
7859 should use the same link command used when mixing languages in a
7860 regular (non-LTO) compilation. This means that if your build process
7861 was mixing languages before, all you need to add is @option{-flto} to
7862 all the compile and link commands.
7864 If LTO encounters objects with C linkage declared with incompatible
7865 types in separate translation units to be linked together (undefined
7866 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7867 issued. The behavior is still undefined at runtime.
7869 If object files containing GIMPLE bytecode are stored in a library archive, say
7870 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7871 are using a linker with linker plugin support. To enable this feature, use
7872 the flag @option{-fuse-linker-plugin} at link-time:
7875 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7878 With the linker plugin enabled, the linker will extract the needed
7879 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7880 to make them part of the aggregated GIMPLE image to be optimized.
7882 If you are not using a linker with linker plugin support and/or do not
7883 enable linker plugin then the objects inside @file{libfoo.a}
7884 will be extracted and linked as usual, but they will not participate
7885 in the LTO optimization process.
7887 Link time optimizations do not require the presence of the whole program to
7888 operate. If the program does not require any symbols to be exported, it is
7889 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7890 the interprocedural optimizers to use more aggressive assumptions which may
7891 lead to improved optimization opportunities.
7892 Use of @option{-fwhole-program} is not needed when linker plugin is
7893 active (see @option{-fuse-linker-plugin}).
7895 Regarding portability: the current implementation of LTO makes no
7896 attempt at generating bytecode that can be ported between different
7897 types of hosts. The bytecode files are versioned and there is a
7898 strict version check, so bytecode files generated in one version of
7899 GCC will not work with an older/newer version of GCC.
7901 Link time optimization does not play well with generating debugging
7902 information. Combining @option{-flto} with
7903 @option{-g} is currently experimental and expected to produce wrong
7906 If you specify the optional @var{n}, the optimization and code
7907 generation done at link time is executed in parallel using @var{n}
7908 parallel jobs by utilizing an installed @command{make} program. The
7909 environment variable @env{MAKE} may be used to override the program
7910 used. The default value for @var{n} is 1.
7912 You can also specify @option{-flto=jobserver} to use GNU make's
7913 job server mode to determine the number of parallel jobs. This
7914 is useful when the Makefile calling GCC is already executing in parallel.
7915 The parent Makefile will need a @samp{+} prepended to the command recipe
7916 for this to work. This will likely only work if @env{MAKE} is
7919 This option is disabled by default.
7921 @item -flto-partition=@var{alg}
7922 @opindex flto-partition
7923 Specify the partitioning algorithm used by the link time optimizer.
7924 The value is either @code{1to1} to specify a partitioning mirroring
7925 the original source files or @code{balanced} to specify partitioning
7926 into equally sized chunks (whenever possible). Specifying @code{none}
7927 as an algorithm disables partitioning and streaming completely. The
7928 default value is @code{balanced}.
7930 @item -flto-compression-level=@var{n}
7931 This option specifies the level of compression used for intermediate
7932 language written to LTO object files, and is only meaningful in
7933 conjunction with LTO mode (@option{-flto}). Valid
7934 values are 0 (no compression) to 9 (maximum compression). Values
7935 outside this range are clamped to either 0 or 9. If the option is not
7936 given, a default balanced compression setting is used.
7939 Prints a report with internal details on the workings of the link-time
7940 optimizer. The contents of this report vary from version to version,
7941 it is meant to be useful to GCC developers when processing object
7942 files in LTO mode (via @option{-flto}).
7944 Disabled by default.
7946 @item -fuse-linker-plugin
7947 Enables the use of a linker plugin during link time optimization. This
7948 option relies on plugin support in the linker, which is available in gold
7949 or in GNU ld 2.21 or newer.
7951 This option enables the extraction of object files with GIMPLE bytecode out
7952 of library archives. This improves the quality of optimization by exposing
7953 more code to the link time optimizer. This information specifies what
7954 symbols can be accessed externally (by non-LTO object or during dynamic
7955 linking). Resulting code quality improvements on binaries (and shared
7956 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7957 See @option{-flto} for a description of the effect of this flag and how to
7960 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7961 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7963 @item -ffat-lto-objects
7964 @opindex ffat-lto-objects
7965 Fat LTO objects are object files that contain both the intermediate language
7966 and the object code. This makes them useable for both LTO linking and normal
7967 linking. This option makes effect only with @option{-flto} and is ignored
7970 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
7971 requires the complete toolchain to be aware of LTO. It requires a linker with
7972 linker plugin support for basic functionality. Additionally, nm, ar and ranlib
7973 need to support linker plugins to allow a full-featured build environment
7974 (capable of building static libraries etc).
7976 The default is @option{-ffat-lto-objects} but this default is intended to
7977 change in future releases when linker plugin enabled environments become more
7979 @item -fcompare-elim
7980 @opindex fcompare-elim
7981 After register allocation and post-register allocation instruction splitting,
7982 identify arithmetic instructions that compute processor flags similar to a
7983 comparison operation based on that arithmetic. If possible, eliminate the
7984 explicit comparison operation.
7986 This pass only applies to certain targets that cannot explicitly represent
7987 the comparison operation before register allocation is complete.
7989 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7991 @item -fcprop-registers
7992 @opindex fcprop-registers
7993 After register allocation and post-register allocation instruction splitting,
7994 we perform a copy-propagation pass to try to reduce scheduling dependencies
7995 and occasionally eliminate the copy.
7997 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7999 @item -fprofile-correction
8000 @opindex fprofile-correction
8001 Profiles collected using an instrumented binary for multi-threaded programs may
8002 be inconsistent due to missed counter updates. When this option is specified,
8003 GCC will use heuristics to correct or smooth out such inconsistencies. By
8004 default, GCC will emit an error message when an inconsistent profile is detected.
8006 @item -fprofile-dir=@var{path}
8007 @opindex fprofile-dir
8009 Set the directory to search for the profile data files in to @var{path}.
8010 This option affects only the profile data generated by
8011 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8012 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8013 and its related options. Both absolute and relative paths can be used.
8014 By default, GCC will use the current directory as @var{path}, thus the
8015 profile data file will appear in the same directory as the object file.
8017 @item -fprofile-generate
8018 @itemx -fprofile-generate=@var{path}
8019 @opindex fprofile-generate
8021 Enable options usually used for instrumenting application to produce
8022 profile useful for later recompilation with profile feedback based
8023 optimization. You must use @option{-fprofile-generate} both when
8024 compiling and when linking your program.
8026 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
8028 If @var{path} is specified, GCC will look at the @var{path} to find
8029 the profile feedback data files. See @option{-fprofile-dir}.
8032 @itemx -fprofile-use=@var{path}
8033 @opindex fprofile-use
8034 Enable profile feedback directed optimizations, and optimizations
8035 generally profitable only with profile feedback available.
8037 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8038 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
8040 By default, GCC emits an error message if the feedback profiles do not
8041 match the source code. This error can be turned into a warning by using
8042 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8045 If @var{path} is specified, GCC will look at the @var{path} to find
8046 the profile feedback data files. See @option{-fprofile-dir}.
8049 The following options control compiler behavior regarding floating
8050 point arithmetic. These options trade off between speed and
8051 correctness. All must be specifically enabled.
8055 @opindex ffloat-store
8056 Do not store floating point variables in registers, and inhibit other
8057 options that might change whether a floating point value is taken from a
8060 @cindex floating point precision
8061 This option prevents undesirable excess precision on machines such as
8062 the 68000 where the floating registers (of the 68881) keep more
8063 precision than a @code{double} is supposed to have. Similarly for the
8064 x86 architecture. For most programs, the excess precision does only
8065 good, but a few programs rely on the precise definition of IEEE floating
8066 point. Use @option{-ffloat-store} for such programs, after modifying
8067 them to store all pertinent intermediate computations into variables.
8069 @item -fexcess-precision=@var{style}
8070 @opindex fexcess-precision
8071 This option allows further control over excess precision on machines
8072 where floating-point registers have more precision than the IEEE
8073 @code{float} and @code{double} types and the processor does not
8074 support operations rounding to those types. By default,
8075 @option{-fexcess-precision=fast} is in effect; this means that
8076 operations are carried out in the precision of the registers and that
8077 it is unpredictable when rounding to the types specified in the source
8078 code takes place. When compiling C, if
8079 @option{-fexcess-precision=standard} is specified then excess
8080 precision will follow the rules specified in ISO C99; in particular,
8081 both casts and assignments cause values to be rounded to their
8082 semantic types (whereas @option{-ffloat-store} only affects
8083 assignments). This option is enabled by default for C if a strict
8084 conformance option such as @option{-std=c99} is used.
8087 @option{-fexcess-precision=standard} is not implemented for languages
8088 other than C, and has no effect if
8089 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8090 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8091 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8092 semantics apply without excess precision, and in the latter, rounding
8097 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8098 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8099 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8101 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8103 This option is not turned on by any @option{-O} option besides
8104 @option{-Ofast} since it can result in incorrect output for programs
8105 which depend on an exact implementation of IEEE or ISO rules/specifications
8106 for math functions. It may, however, yield faster code for programs
8107 that do not require the guarantees of these specifications.
8109 @item -fno-math-errno
8110 @opindex fno-math-errno
8111 Do not set ERRNO after calling math functions that are executed
8112 with a single instruction, e.g., sqrt. A program that relies on
8113 IEEE exceptions for math error handling may want to use this flag
8114 for speed while maintaining IEEE arithmetic compatibility.
8116 This option is not turned on by any @option{-O} option since
8117 it can result in incorrect output for programs which depend on
8118 an exact implementation of IEEE or ISO rules/specifications for
8119 math functions. It may, however, yield faster code for programs
8120 that do not require the guarantees of these specifications.
8122 The default is @option{-fmath-errno}.
8124 On Darwin systems, the math library never sets @code{errno}. There is
8125 therefore no reason for the compiler to consider the possibility that
8126 it might, and @option{-fno-math-errno} is the default.
8128 @item -funsafe-math-optimizations
8129 @opindex funsafe-math-optimizations
8131 Allow optimizations for floating-point arithmetic that (a) assume
8132 that arguments and results are valid and (b) may violate IEEE or
8133 ANSI standards. When used at link-time, it may include libraries
8134 or startup files that change the default FPU control word or other
8135 similar optimizations.
8137 This option is not turned on by any @option{-O} option since
8138 it can result in incorrect output for programs which depend on
8139 an exact implementation of IEEE or ISO rules/specifications for
8140 math functions. It may, however, yield faster code for programs
8141 that do not require the guarantees of these specifications.
8142 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8143 @option{-fassociative-math} and @option{-freciprocal-math}.
8145 The default is @option{-fno-unsafe-math-optimizations}.
8147 @item -fassociative-math
8148 @opindex fassociative-math
8150 Allow re-association of operands in series of floating-point operations.
8151 This violates the ISO C and C++ language standard by possibly changing
8152 computation result. NOTE: re-ordering may change the sign of zero as
8153 well as ignore NaNs and inhibit or create underflow or overflow (and
8154 thus cannot be used on a code which relies on rounding behavior like
8155 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
8156 and thus may not be used when ordered comparisons are required.
8157 This option requires that both @option{-fno-signed-zeros} and
8158 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8159 much sense with @option{-frounding-math}. For Fortran the option
8160 is automatically enabled when both @option{-fno-signed-zeros} and
8161 @option{-fno-trapping-math} are in effect.
8163 The default is @option{-fno-associative-math}.
8165 @item -freciprocal-math
8166 @opindex freciprocal-math
8168 Allow the reciprocal of a value to be used instead of dividing by
8169 the value if this enables optimizations. For example @code{x / y}
8170 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8171 is subject to common subexpression elimination. Note that this loses
8172 precision and increases the number of flops operating on the value.
8174 The default is @option{-fno-reciprocal-math}.
8176 @item -ffinite-math-only
8177 @opindex ffinite-math-only
8178 Allow optimizations for floating-point arithmetic that assume
8179 that arguments and results are not NaNs or +-Infs.
8181 This option is not turned on by any @option{-O} option since
8182 it can result in incorrect output for programs which depend on
8183 an exact implementation of IEEE or ISO rules/specifications for
8184 math functions. It may, however, yield faster code for programs
8185 that do not require the guarantees of these specifications.
8187 The default is @option{-fno-finite-math-only}.
8189 @item -fno-signed-zeros
8190 @opindex fno-signed-zeros
8191 Allow optimizations for floating point arithmetic that ignore the
8192 signedness of zero. IEEE arithmetic specifies the behavior of
8193 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8194 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8195 This option implies that the sign of a zero result isn't significant.
8197 The default is @option{-fsigned-zeros}.
8199 @item -fno-trapping-math
8200 @opindex fno-trapping-math
8201 Compile code assuming that floating-point operations cannot generate
8202 user-visible traps. These traps include division by zero, overflow,
8203 underflow, inexact result and invalid operation. This option requires
8204 that @option{-fno-signaling-nans} be in effect. Setting this option may
8205 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8207 This option should never be turned on by any @option{-O} option since
8208 it can result in incorrect output for programs which depend on
8209 an exact implementation of IEEE or ISO rules/specifications for
8212 The default is @option{-ftrapping-math}.
8214 @item -frounding-math
8215 @opindex frounding-math
8216 Disable transformations and optimizations that assume default floating
8217 point rounding behavior. This is round-to-zero for all floating point
8218 to integer conversions, and round-to-nearest for all other arithmetic
8219 truncations. This option should be specified for programs that change
8220 the FP rounding mode dynamically, or that may be executed with a
8221 non-default rounding mode. This option disables constant folding of
8222 floating point expressions at compile-time (which may be affected by
8223 rounding mode) and arithmetic transformations that are unsafe in the
8224 presence of sign-dependent rounding modes.
8226 The default is @option{-fno-rounding-math}.
8228 This option is experimental and does not currently guarantee to
8229 disable all GCC optimizations that are affected by rounding mode.
8230 Future versions of GCC may provide finer control of this setting
8231 using C99's @code{FENV_ACCESS} pragma. This command line option
8232 will be used to specify the default state for @code{FENV_ACCESS}.
8234 @item -fsignaling-nans
8235 @opindex fsignaling-nans
8236 Compile code assuming that IEEE signaling NaNs may generate user-visible
8237 traps during floating-point operations. Setting this option disables
8238 optimizations that may change the number of exceptions visible with
8239 signaling NaNs. This option implies @option{-ftrapping-math}.
8241 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8244 The default is @option{-fno-signaling-nans}.
8246 This option is experimental and does not currently guarantee to
8247 disable all GCC optimizations that affect signaling NaN behavior.
8249 @item -fsingle-precision-constant
8250 @opindex fsingle-precision-constant
8251 Treat floating point constant as single precision constant instead of
8252 implicitly converting it to double precision constant.
8254 @item -fcx-limited-range
8255 @opindex fcx-limited-range
8256 When enabled, this option states that a range reduction step is not
8257 needed when performing complex division. Also, there is no checking
8258 whether the result of a complex multiplication or division is @code{NaN
8259 + I*NaN}, with an attempt to rescue the situation in that case. The
8260 default is @option{-fno-cx-limited-range}, but is enabled by
8261 @option{-ffast-math}.
8263 This option controls the default setting of the ISO C99
8264 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8267 @item -fcx-fortran-rules
8268 @opindex fcx-fortran-rules
8269 Complex multiplication and division follow Fortran rules. Range
8270 reduction is done as part of complex division, but there is no checking
8271 whether the result of a complex multiplication or division is @code{NaN
8272 + I*NaN}, with an attempt to rescue the situation in that case.
8274 The default is @option{-fno-cx-fortran-rules}.
8278 The following options control optimizations that may improve
8279 performance, but are not enabled by any @option{-O} options. This
8280 section includes experimental options that may produce broken code.
8283 @item -fbranch-probabilities
8284 @opindex fbranch-probabilities
8285 After running a program compiled with @option{-fprofile-arcs}
8286 (@pxref{Debugging Options,, Options for Debugging Your Program or
8287 @command{gcc}}), you can compile it a second time using
8288 @option{-fbranch-probabilities}, to improve optimizations based on
8289 the number of times each branch was taken. When the program
8290 compiled with @option{-fprofile-arcs} exits it saves arc execution
8291 counts to a file called @file{@var{sourcename}.gcda} for each source
8292 file. The information in this data file is very dependent on the
8293 structure of the generated code, so you must use the same source code
8294 and the same optimization options for both compilations.
8296 With @option{-fbranch-probabilities}, GCC puts a
8297 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8298 These can be used to improve optimization. Currently, they are only
8299 used in one place: in @file{reorg.c}, instead of guessing which path a
8300 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8301 exactly determine which path is taken more often.
8303 @item -fprofile-values
8304 @opindex fprofile-values
8305 If combined with @option{-fprofile-arcs}, it adds code so that some
8306 data about values of expressions in the program is gathered.
8308 With @option{-fbranch-probabilities}, it reads back the data gathered
8309 from profiling values of expressions for usage in optimizations.
8311 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8315 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8316 a code to gather information about values of expressions.
8318 With @option{-fbranch-probabilities}, it reads back the data gathered
8319 and actually performs the optimizations based on them.
8320 Currently the optimizations include specialization of division operation
8321 using the knowledge about the value of the denominator.
8323 @item -frename-registers
8324 @opindex frename-registers
8325 Attempt to avoid false dependencies in scheduled code by making use
8326 of registers left over after register allocation. This optimization
8327 will most benefit processors with lots of registers. Depending on the
8328 debug information format adopted by the target, however, it can
8329 make debugging impossible, since variables will no longer stay in
8330 a ``home register''.
8332 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8336 Perform tail duplication to enlarge superblock size. This transformation
8337 simplifies the control flow of the function allowing other optimizations to do
8340 Enabled with @option{-fprofile-use}.
8342 @item -funroll-loops
8343 @opindex funroll-loops
8344 Unroll loops whose number of iterations can be determined at compile time or
8345 upon entry to the loop. @option{-funroll-loops} implies
8346 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8347 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8348 small constant number of iterations). This option makes code larger, and may
8349 or may not make it run faster.
8351 Enabled with @option{-fprofile-use}.
8353 @item -funroll-all-loops
8354 @opindex funroll-all-loops
8355 Unroll all loops, even if their number of iterations is uncertain when
8356 the loop is entered. This usually makes programs run more slowly.
8357 @option{-funroll-all-loops} implies the same options as
8358 @option{-funroll-loops}.
8361 @opindex fpeel-loops
8362 Peels the loops for that there is enough information that they do not
8363 roll much (from profile feedback). It also turns on complete loop peeling
8364 (i.e.@: complete removal of loops with small constant number of iterations).
8366 Enabled with @option{-fprofile-use}.
8368 @item -fmove-loop-invariants
8369 @opindex fmove-loop-invariants
8370 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8371 at level @option{-O1}
8373 @item -funswitch-loops
8374 @opindex funswitch-loops
8375 Move branches with loop invariant conditions out of the loop, with duplicates
8376 of the loop on both branches (modified according to result of the condition).
8378 @item -ffunction-sections
8379 @itemx -fdata-sections
8380 @opindex ffunction-sections
8381 @opindex fdata-sections
8382 Place each function or data item into its own section in the output
8383 file if the target supports arbitrary sections. The name of the
8384 function or the name of the data item determines the section's name
8387 Use these options on systems where the linker can perform optimizations
8388 to improve locality of reference in the instruction space. Most systems
8389 using the ELF object format and SPARC processors running Solaris 2 have
8390 linkers with such optimizations. AIX may have these optimizations in
8393 Only use these options when there are significant benefits from doing
8394 so. When you specify these options, the assembler and linker will
8395 create larger object and executable files and will also be slower.
8396 You will not be able to use @code{gprof} on all systems if you
8397 specify this option and you may have problems with debugging if
8398 you specify both this option and @option{-g}.
8400 @item -fbranch-target-load-optimize
8401 @opindex fbranch-target-load-optimize
8402 Perform branch target register load optimization before prologue / epilogue
8404 The use of target registers can typically be exposed only during reload,
8405 thus hoisting loads out of loops and doing inter-block scheduling needs
8406 a separate optimization pass.
8408 @item -fbranch-target-load-optimize2
8409 @opindex fbranch-target-load-optimize2
8410 Perform branch target register load optimization after prologue / epilogue
8413 @item -fbtr-bb-exclusive
8414 @opindex fbtr-bb-exclusive
8415 When performing branch target register load optimization, don't reuse
8416 branch target registers in within any basic block.
8418 @item -fstack-protector
8419 @opindex fstack-protector
8420 Emit extra code to check for buffer overflows, such as stack smashing
8421 attacks. This is done by adding a guard variable to functions with
8422 vulnerable objects. This includes functions that call alloca, and
8423 functions with buffers larger than 8 bytes. The guards are initialized
8424 when a function is entered and then checked when the function exits.
8425 If a guard check fails, an error message is printed and the program exits.
8427 @item -fstack-protector-all
8428 @opindex fstack-protector-all
8429 Like @option{-fstack-protector} except that all functions are protected.
8431 @item -fsection-anchors
8432 @opindex fsection-anchors
8433 Try to reduce the number of symbolic address calculations by using
8434 shared ``anchor'' symbols to address nearby objects. This transformation
8435 can help to reduce the number of GOT entries and GOT accesses on some
8438 For example, the implementation of the following function @code{foo}:
8442 int foo (void) @{ return a + b + c; @}
8445 would usually calculate the addresses of all three variables, but if you
8446 compile it with @option{-fsection-anchors}, it will access the variables
8447 from a common anchor point instead. The effect is similar to the
8448 following pseudocode (which isn't valid C):
8453 register int *xr = &x;
8454 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8458 Not all targets support this option.
8460 @item --param @var{name}=@var{value}
8462 In some places, GCC uses various constants to control the amount of
8463 optimization that is done. For example, GCC will not inline functions
8464 that contain more that a certain number of instructions. You can
8465 control some of these constants on the command-line using the
8466 @option{--param} option.
8468 The names of specific parameters, and the meaning of the values, are
8469 tied to the internals of the compiler, and are subject to change
8470 without notice in future releases.
8472 In each case, the @var{value} is an integer. The allowable choices for
8473 @var{name} are given in the following table:
8476 @item predictable-branch-outcome
8477 When branch is predicted to be taken with probability lower than this threshold
8478 (in percent), then it is considered well predictable. The default is 10.
8480 @item max-crossjump-edges
8481 The maximum number of incoming edges to consider for crossjumping.
8482 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8483 the number of edges incoming to each block. Increasing values mean
8484 more aggressive optimization, making the compile time increase with
8485 probably small improvement in executable size.
8487 @item min-crossjump-insns
8488 The minimum number of instructions which must be matched at the end
8489 of two blocks before crossjumping will be performed on them. This
8490 value is ignored in the case where all instructions in the block being
8491 crossjumped from are matched. The default value is 5.
8493 @item max-grow-copy-bb-insns
8494 The maximum code size expansion factor when copying basic blocks
8495 instead of jumping. The expansion is relative to a jump instruction.
8496 The default value is 8.
8498 @item max-goto-duplication-insns
8499 The maximum number of instructions to duplicate to a block that jumps
8500 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8501 passes, GCC factors computed gotos early in the compilation process,
8502 and unfactors them as late as possible. Only computed jumps at the
8503 end of a basic blocks with no more than max-goto-duplication-insns are
8504 unfactored. The default value is 8.
8506 @item max-delay-slot-insn-search
8507 The maximum number of instructions to consider when looking for an
8508 instruction to fill a delay slot. If more than this arbitrary number of
8509 instructions is searched, the time savings from filling the delay slot
8510 will be minimal so stop searching. Increasing values mean more
8511 aggressive optimization, making the compile time increase with probably
8512 small improvement in executable run time.
8514 @item max-delay-slot-live-search
8515 When trying to fill delay slots, the maximum number of instructions to
8516 consider when searching for a block with valid live register
8517 information. Increasing this arbitrarily chosen value means more
8518 aggressive optimization, increasing the compile time. This parameter
8519 should be removed when the delay slot code is rewritten to maintain the
8522 @item max-gcse-memory
8523 The approximate maximum amount of memory that will be allocated in
8524 order to perform the global common subexpression elimination
8525 optimization. If more memory than specified is required, the
8526 optimization will not be done.
8528 @item max-gcse-insertion-ratio
8529 If the ratio of expression insertions to deletions is larger than this value
8530 for any expression, then RTL PRE will insert or remove the expression and thus
8531 leave partially redundant computations in the instruction stream. The default value is 20.
8533 @item max-pending-list-length
8534 The maximum number of pending dependencies scheduling will allow
8535 before flushing the current state and starting over. Large functions
8536 with few branches or calls can create excessively large lists which
8537 needlessly consume memory and resources.
8539 @item max-modulo-backtrack-attempts
8540 The maximum number of backtrack attempts the scheduler should make
8541 when modulo scheduling a loop. Larger values can exponentially increase
8544 @item max-inline-insns-single
8545 Several parameters control the tree inliner used in gcc.
8546 This number sets the maximum number of instructions (counted in GCC's
8547 internal representation) in a single function that the tree inliner
8548 will consider for inlining. This only affects functions declared
8549 inline and methods implemented in a class declaration (C++).
8550 The default value is 400.
8552 @item max-inline-insns-auto
8553 When you use @option{-finline-functions} (included in @option{-O3}),
8554 a lot of functions that would otherwise not be considered for inlining
8555 by the compiler will be investigated. To those functions, a different
8556 (more restrictive) limit compared to functions declared inline can
8558 The default value is 40.
8560 @item large-function-insns
8561 The limit specifying really large functions. For functions larger than this
8562 limit after inlining, inlining is constrained by
8563 @option{--param large-function-growth}. This parameter is useful primarily
8564 to avoid extreme compilation time caused by non-linear algorithms used by the
8566 The default value is 2700.
8568 @item large-function-growth
8569 Specifies maximal growth of large function caused by inlining in percents.
8570 The default value is 100 which limits large function growth to 2.0 times
8573 @item large-unit-insns
8574 The limit specifying large translation unit. Growth caused by inlining of
8575 units larger than this limit is limited by @option{--param inline-unit-growth}.
8576 For small units this might be too tight (consider unit consisting of function A
8577 that is inline and B that just calls A three time. If B is small relative to
8578 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8579 large units consisting of small inlineable functions however the overall unit
8580 growth limit is needed to avoid exponential explosion of code size. Thus for
8581 smaller units, the size is increased to @option{--param large-unit-insns}
8582 before applying @option{--param inline-unit-growth}. The default is 10000
8584 @item inline-unit-growth
8585 Specifies maximal overall growth of the compilation unit caused by inlining.
8586 The default value is 30 which limits unit growth to 1.3 times the original
8589 @item ipcp-unit-growth
8590 Specifies maximal overall growth of the compilation unit caused by
8591 interprocedural constant propagation. The default value is 10 which limits
8592 unit growth to 1.1 times the original size.
8594 @item large-stack-frame
8595 The limit specifying large stack frames. While inlining the algorithm is trying
8596 to not grow past this limit too much. Default value is 256 bytes.
8598 @item large-stack-frame-growth
8599 Specifies maximal growth of large stack frames caused by inlining in percents.
8600 The default value is 1000 which limits large stack frame growth to 11 times
8603 @item max-inline-insns-recursive
8604 @itemx max-inline-insns-recursive-auto
8605 Specifies maximum number of instructions out-of-line copy of self recursive inline
8606 function can grow into by performing recursive inlining.
8608 For functions declared inline @option{--param max-inline-insns-recursive} is
8609 taken into account. For function not declared inline, recursive inlining
8610 happens only when @option{-finline-functions} (included in @option{-O3}) is
8611 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8612 default value is 450.
8614 @item max-inline-recursive-depth
8615 @itemx max-inline-recursive-depth-auto
8616 Specifies maximum recursion depth used by the recursive inlining.
8618 For functions declared inline @option{--param max-inline-recursive-depth} is
8619 taken into account. For function not declared inline, recursive inlining
8620 happens only when @option{-finline-functions} (included in @option{-O3}) is
8621 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8624 @item min-inline-recursive-probability
8625 Recursive inlining is profitable only for function having deep recursion
8626 in average and can hurt for function having little recursion depth by
8627 increasing the prologue size or complexity of function body to other
8630 When profile feedback is available (see @option{-fprofile-generate}) the actual
8631 recursion depth can be guessed from probability that function will recurse via
8632 given call expression. This parameter limits inlining only to call expression
8633 whose probability exceeds given threshold (in percents). The default value is
8636 @item early-inlining-insns
8637 Specify growth that early inliner can make. In effect it increases amount of
8638 inlining for code having large abstraction penalty. The default value is 10.
8640 @item max-early-inliner-iterations
8641 @itemx max-early-inliner-iterations
8642 Limit of iterations of early inliner. This basically bounds number of nested
8643 indirect calls early inliner can resolve. Deeper chains are still handled by
8646 @item comdat-sharing-probability
8647 @itemx comdat-sharing-probability
8648 Probability (in percent) that C++ inline function with comdat visibility
8649 will be shared across multiple compilation units. The default value is 20.
8651 @item min-vect-loop-bound
8652 The minimum number of iterations under which a loop will not get vectorized
8653 when @option{-ftree-vectorize} is used. The number of iterations after
8654 vectorization needs to be greater than the value specified by this option
8655 to allow vectorization. The default value is 0.
8657 @item gcse-cost-distance-ratio
8658 Scaling factor in calculation of maximum distance an expression
8659 can be moved by GCSE optimizations. This is currently supported only in the
8660 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8661 will be with simple expressions, i.e., the expressions which have cost
8662 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8663 hoisting of simple expressions. The default value is 10.
8665 @item gcse-unrestricted-cost
8666 Cost, roughly measured as the cost of a single typical machine
8667 instruction, at which GCSE optimizations will not constrain
8668 the distance an expression can travel. This is currently
8669 supported only in the code hoisting pass. The lesser the cost,
8670 the more aggressive code hoisting will be. Specifying 0 will
8671 allow all expressions to travel unrestricted distances.
8672 The default value is 3.
8674 @item max-hoist-depth
8675 The depth of search in the dominator tree for expressions to hoist.
8676 This is used to avoid quadratic behavior in hoisting algorithm.
8677 The value of 0 will avoid limiting the search, but may slow down compilation
8678 of huge functions. The default value is 30.
8680 @item max-tail-merge-comparisons
8681 The maximum amount of similar bbs to compare a bb with. This is used to
8682 avoid quadratic behaviour in tree tail merging. The default value is 10.
8684 @item max-tail-merge-iterations
8685 The maximum amount of iterations of the pass over the function. This is used to
8686 limit run time in tree tail merging. The default value is 2.
8688 @item max-unrolled-insns
8689 The maximum number of instructions that a loop should have if that loop
8690 is unrolled, and if the loop is unrolled, it determines how many times
8691 the loop code is unrolled.
8693 @item max-average-unrolled-insns
8694 The maximum number of instructions biased by probabilities of their execution
8695 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8696 it determines how many times the loop code is unrolled.
8698 @item max-unroll-times
8699 The maximum number of unrollings of a single loop.
8701 @item max-peeled-insns
8702 The maximum number of instructions that a loop should have if that loop
8703 is peeled, and if the loop is peeled, it determines how many times
8704 the loop code is peeled.
8706 @item max-peel-times
8707 The maximum number of peelings of a single loop.
8709 @item max-completely-peeled-insns
8710 The maximum number of insns of a completely peeled loop.
8712 @item max-completely-peel-times
8713 The maximum number of iterations of a loop to be suitable for complete peeling.
8715 @item max-completely-peel-loop-nest-depth
8716 The maximum depth of a loop nest suitable for complete peeling.
8718 @item max-unswitch-insns
8719 The maximum number of insns of an unswitched loop.
8721 @item max-unswitch-level
8722 The maximum number of branches unswitched in a single loop.
8725 The minimum cost of an expensive expression in the loop invariant motion.
8727 @item iv-consider-all-candidates-bound
8728 Bound on number of candidates for induction variables below that
8729 all candidates are considered for each use in induction variable
8730 optimizations. Only the most relevant candidates are considered
8731 if there are more candidates, to avoid quadratic time complexity.
8733 @item iv-max-considered-uses
8734 The induction variable optimizations give up on loops that contain more
8735 induction variable uses.
8737 @item iv-always-prune-cand-set-bound
8738 If number of candidates in the set is smaller than this value,
8739 we always try to remove unnecessary ivs from the set during its
8740 optimization when a new iv is added to the set.
8742 @item scev-max-expr-size
8743 Bound on size of expressions used in the scalar evolutions analyzer.
8744 Large expressions slow the analyzer.
8746 @item scev-max-expr-complexity
8747 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8748 Complex expressions slow the analyzer.
8750 @item omega-max-vars
8751 The maximum number of variables in an Omega constraint system.
8752 The default value is 128.
8754 @item omega-max-geqs
8755 The maximum number of inequalities in an Omega constraint system.
8756 The default value is 256.
8759 The maximum number of equalities in an Omega constraint system.
8760 The default value is 128.
8762 @item omega-max-wild-cards
8763 The maximum number of wildcard variables that the Omega solver will
8764 be able to insert. The default value is 18.
8766 @item omega-hash-table-size
8767 The size of the hash table in the Omega solver. The default value is
8770 @item omega-max-keys
8771 The maximal number of keys used by the Omega solver. The default
8774 @item omega-eliminate-redundant-constraints
8775 When set to 1, use expensive methods to eliminate all redundant
8776 constraints. The default value is 0.
8778 @item vect-max-version-for-alignment-checks
8779 The maximum number of runtime checks that can be performed when
8780 doing loop versioning for alignment in the vectorizer. See option
8781 ftree-vect-loop-version for more information.
8783 @item vect-max-version-for-alias-checks
8784 The maximum number of runtime checks that can be performed when
8785 doing loop versioning for alias in the vectorizer. See option
8786 ftree-vect-loop-version for more information.
8788 @item max-iterations-to-track
8790 The maximum number of iterations of a loop the brute force algorithm
8791 for analysis of # of iterations of the loop tries to evaluate.
8793 @item hot-bb-count-fraction
8794 Select fraction of the maximal count of repetitions of basic block in program
8795 given basic block needs to have to be considered hot.
8797 @item hot-bb-frequency-fraction
8798 Select fraction of the entry block frequency of executions of basic block in
8799 function given basic block needs to have to be considered hot.
8801 @item max-predicted-iterations
8802 The maximum number of loop iterations we predict statically. This is useful
8803 in cases where function contain single loop with known bound and other loop
8804 with unknown. We predict the known number of iterations correctly, while
8805 the unknown number of iterations average to roughly 10. This means that the
8806 loop without bounds would appear artificially cold relative to the other one.
8808 @item align-threshold
8810 Select fraction of the maximal frequency of executions of basic block in
8811 function given basic block will get aligned.
8813 @item align-loop-iterations
8815 A loop expected to iterate at lest the selected number of iterations will get
8818 @item tracer-dynamic-coverage
8819 @itemx tracer-dynamic-coverage-feedback
8821 This value is used to limit superblock formation once the given percentage of
8822 executed instructions is covered. This limits unnecessary code size
8825 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8826 feedback is available. The real profiles (as opposed to statically estimated
8827 ones) are much less balanced allowing the threshold to be larger value.
8829 @item tracer-max-code-growth
8830 Stop tail duplication once code growth has reached given percentage. This is
8831 rather hokey argument, as most of the duplicates will be eliminated later in
8832 cross jumping, so it may be set to much higher values than is the desired code
8835 @item tracer-min-branch-ratio
8837 Stop reverse growth when the reverse probability of best edge is less than this
8838 threshold (in percent).
8840 @item tracer-min-branch-ratio
8841 @itemx tracer-min-branch-ratio-feedback
8843 Stop forward growth if the best edge do have probability lower than this
8846 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8847 compilation for profile feedback and one for compilation without. The value
8848 for compilation with profile feedback needs to be more conservative (higher) in
8849 order to make tracer effective.
8851 @item max-cse-path-length
8853 Maximum number of basic blocks on path that cse considers. The default is 10.
8856 The maximum instructions CSE process before flushing. The default is 1000.
8858 @item ggc-min-expand
8860 GCC uses a garbage collector to manage its own memory allocation. This
8861 parameter specifies the minimum percentage by which the garbage
8862 collector's heap should be allowed to expand between collections.
8863 Tuning this may improve compilation speed; it has no effect on code
8866 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8867 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8868 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8869 GCC is not able to calculate RAM on a particular platform, the lower
8870 bound of 30% is used. Setting this parameter and
8871 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8872 every opportunity. This is extremely slow, but can be useful for
8875 @item ggc-min-heapsize
8877 Minimum size of the garbage collector's heap before it begins bothering
8878 to collect garbage. The first collection occurs after the heap expands
8879 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8880 tuning this may improve compilation speed, and has no effect on code
8883 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8884 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8885 with a lower bound of 4096 (four megabytes) and an upper bound of
8886 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8887 particular platform, the lower bound is used. Setting this parameter
8888 very large effectively disables garbage collection. Setting this
8889 parameter and @option{ggc-min-expand} to zero causes a full collection
8890 to occur at every opportunity.
8892 @item max-reload-search-insns
8893 The maximum number of instruction reload should look backward for equivalent
8894 register. Increasing values mean more aggressive optimization, making the
8895 compile time increase with probably slightly better performance. The default
8898 @item max-cselib-memory-locations
8899 The maximum number of memory locations cselib should take into account.
8900 Increasing values mean more aggressive optimization, making the compile time
8901 increase with probably slightly better performance. The default value is 500.
8903 @item reorder-blocks-duplicate
8904 @itemx reorder-blocks-duplicate-feedback
8906 Used by basic block reordering pass to decide whether to use unconditional
8907 branch or duplicate the code on its destination. Code is duplicated when its
8908 estimated size is smaller than this value multiplied by the estimated size of
8909 unconditional jump in the hot spots of the program.
8911 The @option{reorder-block-duplicate-feedback} is used only when profile
8912 feedback is available and may be set to higher values than
8913 @option{reorder-block-duplicate} since information about the hot spots is more
8916 @item max-sched-ready-insns
8917 The maximum number of instructions ready to be issued the scheduler should
8918 consider at any given time during the first scheduling pass. Increasing
8919 values mean more thorough searches, making the compilation time increase
8920 with probably little benefit. The default value is 100.
8922 @item max-sched-region-blocks
8923 The maximum number of blocks in a region to be considered for
8924 interblock scheduling. The default value is 10.
8926 @item max-pipeline-region-blocks
8927 The maximum number of blocks in a region to be considered for
8928 pipelining in the selective scheduler. The default value is 15.
8930 @item max-sched-region-insns
8931 The maximum number of insns in a region to be considered for
8932 interblock scheduling. The default value is 100.
8934 @item max-pipeline-region-insns
8935 The maximum number of insns in a region to be considered for
8936 pipelining in the selective scheduler. The default value is 200.
8939 The minimum probability (in percents) of reaching a source block
8940 for interblock speculative scheduling. The default value is 40.
8942 @item max-sched-extend-regions-iters
8943 The maximum number of iterations through CFG to extend regions.
8944 0 - disable region extension,
8945 N - do at most N iterations.
8946 The default value is 0.
8948 @item max-sched-insn-conflict-delay
8949 The maximum conflict delay for an insn to be considered for speculative motion.
8950 The default value is 3.
8952 @item sched-spec-prob-cutoff
8953 The minimal probability of speculation success (in percents), so that
8954 speculative insn will be scheduled.
8955 The default value is 40.
8957 @item sched-mem-true-dep-cost
8958 Minimal distance (in CPU cycles) between store and load targeting same
8959 memory locations. The default value is 1.
8961 @item selsched-max-lookahead
8962 The maximum size of the lookahead window of selective scheduling. It is a
8963 depth of search for available instructions.
8964 The default value is 50.
8966 @item selsched-max-sched-times
8967 The maximum number of times that an instruction will be scheduled during
8968 selective scheduling. This is the limit on the number of iterations
8969 through which the instruction may be pipelined. The default value is 2.
8971 @item selsched-max-insns-to-rename
8972 The maximum number of best instructions in the ready list that are considered
8973 for renaming in the selective scheduler. The default value is 2.
8976 The minimum value of stage count that swing modulo scheduler will
8977 generate. The default value is 2.
8979 @item max-last-value-rtl
8980 The maximum size measured as number of RTLs that can be recorded in an expression
8981 in combiner for a pseudo register as last known value of that register. The default
8984 @item integer-share-limit
8985 Small integer constants can use a shared data structure, reducing the
8986 compiler's memory usage and increasing its speed. This sets the maximum
8987 value of a shared integer constant. The default value is 256.
8989 @item min-virtual-mappings
8990 Specifies the minimum number of virtual mappings in the incremental
8991 SSA updater that should be registered to trigger the virtual mappings
8992 heuristic defined by virtual-mappings-ratio. The default value is
8995 @item virtual-mappings-ratio
8996 If the number of virtual mappings is virtual-mappings-ratio bigger
8997 than the number of virtual symbols to be updated, then the incremental
8998 SSA updater switches to a full update for those symbols. The default
9001 @item ssp-buffer-size
9002 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
9003 protection when @option{-fstack-protection} is used.
9005 @item max-jump-thread-duplication-stmts
9006 Maximum number of statements allowed in a block that needs to be
9007 duplicated when threading jumps.
9009 @item max-fields-for-field-sensitive
9010 Maximum number of fields in a structure we will treat in
9011 a field sensitive manner during pointer analysis. The default is zero
9012 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
9014 @item prefetch-latency
9015 Estimate on average number of instructions that are executed before
9016 prefetch finishes. The distance we prefetch ahead is proportional
9017 to this constant. Increasing this number may also lead to less
9018 streams being prefetched (see @option{simultaneous-prefetches}).
9020 @item simultaneous-prefetches
9021 Maximum number of prefetches that can run at the same time.
9023 @item l1-cache-line-size
9024 The size of cache line in L1 cache, in bytes.
9027 The size of L1 cache, in kilobytes.
9030 The size of L2 cache, in kilobytes.
9032 @item min-insn-to-prefetch-ratio
9033 The minimum ratio between the number of instructions and the
9034 number of prefetches to enable prefetching in a loop.
9036 @item prefetch-min-insn-to-mem-ratio
9037 The minimum ratio between the number of instructions and the
9038 number of memory references to enable prefetching in a loop.
9040 @item use-canonical-types
9041 Whether the compiler should use the ``canonical'' type system. By
9042 default, this should always be 1, which uses a more efficient internal
9043 mechanism for comparing types in C++ and Objective-C++. However, if
9044 bugs in the canonical type system are causing compilation failures,
9045 set this value to 0 to disable canonical types.
9047 @item switch-conversion-max-branch-ratio
9048 Switch initialization conversion will refuse to create arrays that are
9049 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9050 branches in the switch.
9052 @item max-partial-antic-length
9053 Maximum length of the partial antic set computed during the tree
9054 partial redundancy elimination optimization (@option{-ftree-pre}) when
9055 optimizing at @option{-O3} and above. For some sorts of source code
9056 the enhanced partial redundancy elimination optimization can run away,
9057 consuming all of the memory available on the host machine. This
9058 parameter sets a limit on the length of the sets that are computed,
9059 which prevents the runaway behavior. Setting a value of 0 for
9060 this parameter will allow an unlimited set length.
9062 @item sccvn-max-scc-size
9063 Maximum size of a strongly connected component (SCC) during SCCVN
9064 processing. If this limit is hit, SCCVN processing for the whole
9065 function will not be done and optimizations depending on it will
9066 be disabled. The default maximum SCC size is 10000.
9068 @item ira-max-loops-num
9069 IRA uses a regional register allocation by default. If a function
9070 contains loops more than number given by the parameter, only at most
9071 given number of the most frequently executed loops will form regions
9072 for the regional register allocation. The default value of the
9075 @item ira-max-conflict-table-size
9076 Although IRA uses a sophisticated algorithm of compression conflict
9077 table, the table can be still big for huge functions. If the conflict
9078 table for a function could be more than size in MB given by the
9079 parameter, the conflict table is not built and faster, simpler, and
9080 lower quality register allocation algorithm will be used. The
9081 algorithm do not use pseudo-register conflicts. The default value of
9082 the parameter is 2000.
9084 @item ira-loop-reserved-regs
9085 IRA can be used to evaluate more accurate register pressure in loops
9086 for decision to move loop invariants (see @option{-O3}). The number
9087 of available registers reserved for some other purposes is described
9088 by this parameter. The default value of the parameter is 2 which is
9089 minimal number of registers needed for execution of typical
9090 instruction. This value is the best found from numerous experiments.
9092 @item loop-invariant-max-bbs-in-loop
9093 Loop invariant motion can be very expensive, both in compile time and
9094 in amount of needed compile time memory, with very large loops. Loops
9095 with more basic blocks than this parameter won't have loop invariant
9096 motion optimization performed on them. The default value of the
9097 parameter is 1000 for -O1 and 10000 for -O2 and above.
9099 @item max-vartrack-size
9100 Sets a maximum number of hash table slots to use during variable
9101 tracking dataflow analysis of any function. If this limit is exceeded
9102 with variable tracking at assignments enabled, analysis for that
9103 function is retried without it, after removing all debug insns from
9104 the function. If the limit is exceeded even without debug insns, var
9105 tracking analysis is completely disabled for the function. Setting
9106 the parameter to zero makes it unlimited.
9108 @item max-vartrack-expr-depth
9109 Sets a maximum number of recursion levels when attempting to map
9110 variable names or debug temporaries to value expressions. This trades
9111 compile time for more complete debug information. If this is set too
9112 low, value expressions that are available and could be represented in
9113 debug information may end up not being used; setting this higher may
9114 enable the compiler to find more complex debug expressions, but compile
9115 time and memory use may grow. The default is 12.
9117 @item min-nondebug-insn-uid
9118 Use uids starting at this parameter for nondebug insns. The range below
9119 the parameter is reserved exclusively for debug insns created by
9120 @option{-fvar-tracking-assignments}, but debug insns may get
9121 (non-overlapping) uids above it if the reserved range is exhausted.
9123 @item ipa-sra-ptr-growth-factor
9124 IPA-SRA will replace a pointer to an aggregate with one or more new
9125 parameters only when their cumulative size is less or equal to
9126 @option{ipa-sra-ptr-growth-factor} times the size of the original
9129 @item tm-max-aggregate-size
9130 When making copies of thread-local variables in a transaction, this
9131 parameter specifies the size in bytes after which variables will be
9132 saved with the logging functions as opposed to save/restore code
9133 sequence pairs. This option only applies when using
9136 @item graphite-max-nb-scop-params
9137 To avoid exponential effects in the Graphite loop transforms, the
9138 number of parameters in a Static Control Part (SCoP) is bounded. The
9139 default value is 10 parameters. A variable whose value is unknown at
9140 compile time and defined outside a SCoP is a parameter of the SCoP.
9142 @item graphite-max-bbs-per-function
9143 To avoid exponential effects in the detection of SCoPs, the size of
9144 the functions analyzed by Graphite is bounded. The default value is
9147 @item loop-block-tile-size
9148 Loop blocking or strip mining transforms, enabled with
9149 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9150 loop in the loop nest by a given number of iterations. The strip
9151 length can be changed using the @option{loop-block-tile-size}
9152 parameter. The default value is 51 iterations.
9154 @item ipa-cp-value-list-size
9155 IPA-CP attempts to track all possible values and types passed to a function's
9156 parameter in order to propagate them and perform devirtualization.
9157 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9158 stores per one formal parameter of a function.
9160 @item lto-partitions
9161 Specify desired number of partitions produced during WHOPR compilation.
9162 The number of partitions should exceed the number of CPUs used for compilation.
9163 The default value is 32.
9165 @item lto-minpartition
9166 Size of minimal partition for WHOPR (in estimated instructions).
9167 This prevents expenses of splitting very small programs into too many
9170 @item cxx-max-namespaces-for-diagnostic-help
9171 The maximum number of namespaces to consult for suggestions when C++
9172 name lookup fails for an identifier. The default is 1000.
9174 @item sink-frequency-threshold
9175 The maximum relative execution frequency (in percents) of the target block
9176 relative to a statement's original block to allow statement sinking of a
9177 statement. Larger numbers result in more aggressive statement sinking.
9178 The default value is 75. A small positive adjustment is applied for
9179 statements with memory operands as those are even more profitable so sink.
9181 @item max-stores-to-sink
9182 The maximum number of conditional stores paires that can be sunk. Set to 0
9183 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9184 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9186 @item allow-load-data-races
9187 Allow optimizers to introduce new data races on loads.
9188 Set to 1 to allow, otherwise to 0. This option is enabled by default
9189 unless implicitly set by the @option{-fmemory-model=} option.
9191 @item allow-store-data-races
9192 Allow optimizers to introduce new data races on stores.
9193 Set to 1 to allow, otherwise to 0. This option is enabled by default
9194 unless implicitly set by the @option{-fmemory-model=} option.
9196 @item allow-packed-load-data-races
9197 Allow optimizers to introduce new data races on packed data loads.
9198 Set to 1 to allow, otherwise to 0. This option is enabled by default
9199 unless implicitly set by the @option{-fmemory-model=} option.
9201 @item allow-packed-store-data-races
9202 Allow optimizers to introduce new data races on packed data stores.
9203 Set to 1 to allow, otherwise to 0. This option is enabled by default
9204 unless implicitly set by the @option{-fmemory-model=} option.
9206 @item case-values-threshold
9207 The smallest number of different values for which it is best to use a
9208 jump-table instead of a tree of conditional branches. If the value is
9209 0, use the default for the machine. The default is 0.
9211 @item tree-reassoc-width
9212 Set the maximum number of instructions executed in parallel in
9213 reassociated tree. This parameter overrides target dependent
9214 heuristics used by default if has non zero value.
9219 @node Preprocessor Options
9220 @section Options Controlling the Preprocessor
9221 @cindex preprocessor options
9222 @cindex options, preprocessor
9224 These options control the C preprocessor, which is run on each C source
9225 file before actual compilation.
9227 If you use the @option{-E} option, nothing is done except preprocessing.
9228 Some of these options make sense only together with @option{-E} because
9229 they cause the preprocessor output to be unsuitable for actual
9233 @item -Wp,@var{option}
9235 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9236 and pass @var{option} directly through to the preprocessor. If
9237 @var{option} contains commas, it is split into multiple options at the
9238 commas. However, many options are modified, translated or interpreted
9239 by the compiler driver before being passed to the preprocessor, and
9240 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9241 interface is undocumented and subject to change, so whenever possible
9242 you should avoid using @option{-Wp} and let the driver handle the
9245 @item -Xpreprocessor @var{option}
9246 @opindex Xpreprocessor
9247 Pass @var{option} as an option to the preprocessor. You can use this to
9248 supply system-specific preprocessor options which GCC does not know how to
9251 If you want to pass an option that takes an argument, you must use
9252 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9255 @include cppopts.texi
9257 @node Assembler Options
9258 @section Passing Options to the Assembler
9260 @c prevent bad page break with this line
9261 You can pass options to the assembler.
9264 @item -Wa,@var{option}
9266 Pass @var{option} as an option to the assembler. If @var{option}
9267 contains commas, it is split into multiple options at the commas.
9269 @item -Xassembler @var{option}
9271 Pass @var{option} as an option to the assembler. You can use this to
9272 supply system-specific assembler options which GCC does not know how to
9275 If you want to pass an option that takes an argument, you must use
9276 @option{-Xassembler} twice, once for the option and once for the argument.
9281 @section Options for Linking
9282 @cindex link options
9283 @cindex options, linking
9285 These options come into play when the compiler links object files into
9286 an executable output file. They are meaningless if the compiler is
9287 not doing a link step.
9291 @item @var{object-file-name}
9292 A file name that does not end in a special recognized suffix is
9293 considered to name an object file or library. (Object files are
9294 distinguished from libraries by the linker according to the file
9295 contents.) If linking is done, these object files are used as input
9304 If any of these options is used, then the linker is not run, and
9305 object file names should not be used as arguments. @xref{Overall
9309 @item -l@var{library}
9310 @itemx -l @var{library}
9312 Search the library named @var{library} when linking. (The second
9313 alternative with the library as a separate argument is only for
9314 POSIX compliance and is not recommended.)
9316 It makes a difference where in the command you write this option; the
9317 linker searches and processes libraries and object files in the order they
9318 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9319 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9320 to functions in @samp{z}, those functions may not be loaded.
9322 The linker searches a standard list of directories for the library,
9323 which is actually a file named @file{lib@var{library}.a}. The linker
9324 then uses this file as if it had been specified precisely by name.
9326 The directories searched include several standard system directories
9327 plus any that you specify with @option{-L}.
9329 Normally the files found this way are library files---archive files
9330 whose members are object files. The linker handles an archive file by
9331 scanning through it for members which define symbols that have so far
9332 been referenced but not defined. But if the file that is found is an
9333 ordinary object file, it is linked in the usual fashion. The only
9334 difference between using an @option{-l} option and specifying a file name
9335 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9336 and searches several directories.
9340 You need this special case of the @option{-l} option in order to
9341 link an Objective-C or Objective-C++ program.
9344 @opindex nostartfiles
9345 Do not use the standard system startup files when linking.
9346 The standard system libraries are used normally, unless @option{-nostdlib}
9347 or @option{-nodefaultlibs} is used.
9349 @item -nodefaultlibs
9350 @opindex nodefaultlibs
9351 Do not use the standard system libraries when linking.
9352 Only the libraries you specify will be passed to the linker, options
9353 specifying linkage of the system libraries, such as @code{-static-libgcc}
9354 or @code{-shared-libgcc}, will be ignored.
9355 The standard startup files are used normally, unless @option{-nostartfiles}
9356 is used. The compiler may generate calls to @code{memcmp},
9357 @code{memset}, @code{memcpy} and @code{memmove}.
9358 These entries are usually resolved by entries in
9359 libc. These entry points should be supplied through some other
9360 mechanism when this option is specified.
9364 Do not use the standard system startup files or libraries when linking.
9365 No startup files and only the libraries you specify will be passed to
9366 the linker, options specifying linkage of the system libraries, such as
9367 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9368 The compiler may generate calls to @code{memcmp}, @code{memset},
9369 @code{memcpy} and @code{memmove}.
9370 These entries are usually resolved by entries in
9371 libc. These entry points should be supplied through some other
9372 mechanism when this option is specified.
9374 @cindex @option{-lgcc}, use with @option{-nostdlib}
9375 @cindex @option{-nostdlib} and unresolved references
9376 @cindex unresolved references and @option{-nostdlib}
9377 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9378 @cindex @option{-nodefaultlibs} and unresolved references
9379 @cindex unresolved references and @option{-nodefaultlibs}
9380 One of the standard libraries bypassed by @option{-nostdlib} and
9381 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9382 that GCC uses to overcome shortcomings of particular machines, or special
9383 needs for some languages.
9384 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9385 Collection (GCC) Internals},
9386 for more discussion of @file{libgcc.a}.)
9387 In most cases, you need @file{libgcc.a} even when you want to avoid
9388 other standard libraries. In other words, when you specify @option{-nostdlib}
9389 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9390 This ensures that you have no unresolved references to internal GCC
9391 library subroutines. (For example, @samp{__main}, used to ensure C++
9392 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9393 GNU Compiler Collection (GCC) Internals}.)
9397 Produce a position independent executable on targets which support it.
9398 For predictable results, you must also specify the same set of options
9399 that were used to generate code (@option{-fpie}, @option{-fPIE},
9400 or model suboptions) when you specify this option.
9404 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9405 that support it. This instructs the linker to add all symbols, not
9406 only used ones, to the dynamic symbol table. This option is needed
9407 for some uses of @code{dlopen} or to allow obtaining backtraces
9408 from within a program.
9412 Remove all symbol table and relocation information from the executable.
9416 On systems that support dynamic linking, this prevents linking with the shared
9417 libraries. On other systems, this option has no effect.
9421 Produce a shared object which can then be linked with other objects to
9422 form an executable. Not all systems support this option. For predictable
9423 results, you must also specify the same set of options that were used to
9424 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9425 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9426 needs to build supplementary stub code for constructors to work. On
9427 multi-libbed systems, @samp{gcc -shared} must select the correct support
9428 libraries to link against. Failing to supply the correct flags may lead
9429 to subtle defects. Supplying them in cases where they are not necessary
9432 @item -shared-libgcc
9433 @itemx -static-libgcc
9434 @opindex shared-libgcc
9435 @opindex static-libgcc
9436 On systems that provide @file{libgcc} as a shared library, these options
9437 force the use of either the shared or static version respectively.
9438 If no shared version of @file{libgcc} was built when the compiler was
9439 configured, these options have no effect.
9441 There are several situations in which an application should use the
9442 shared @file{libgcc} instead of the static version. The most common
9443 of these is when the application wishes to throw and catch exceptions
9444 across different shared libraries. In that case, each of the libraries
9445 as well as the application itself should use the shared @file{libgcc}.
9447 Therefore, the G++ and GCJ drivers automatically add
9448 @option{-shared-libgcc} whenever you build a shared library or a main
9449 executable, because C++ and Java programs typically use exceptions, so
9450 this is the right thing to do.
9452 If, instead, you use the GCC driver to create shared libraries, you may
9453 find that they will not always be linked with the shared @file{libgcc}.
9454 If GCC finds, at its configuration time, that you have a non-GNU linker
9455 or a GNU linker that does not support option @option{--eh-frame-hdr},
9456 it will link the shared version of @file{libgcc} into shared libraries
9457 by default. Otherwise, it will take advantage of the linker and optimize
9458 away the linking with the shared version of @file{libgcc}, linking with
9459 the static version of libgcc by default. This allows exceptions to
9460 propagate through such shared libraries, without incurring relocation
9461 costs at library load time.
9463 However, if a library or main executable is supposed to throw or catch
9464 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9465 for the languages used in the program, or using the option
9466 @option{-shared-libgcc}, such that it is linked with the shared
9469 @item -static-libstdc++
9470 When the @command{g++} program is used to link a C++ program, it will
9471 normally automatically link against @option{libstdc++}. If
9472 @file{libstdc++} is available as a shared library, and the
9473 @option{-static} option is not used, then this will link against the
9474 shared version of @file{libstdc++}. That is normally fine. However, it
9475 is sometimes useful to freeze the version of @file{libstdc++} used by
9476 the program without going all the way to a fully static link. The
9477 @option{-static-libstdc++} option directs the @command{g++} driver to
9478 link @file{libstdc++} statically, without necessarily linking other
9479 libraries statically.
9483 Bind references to global symbols when building a shared object. Warn
9484 about any unresolved references (unless overridden by the link editor
9485 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9488 @item -T @var{script}
9490 @cindex linker script
9491 Use @var{script} as the linker script. This option is supported by most
9492 systems using the GNU linker. On some targets, such as bare-board
9493 targets without an operating system, the @option{-T} option may be required
9494 when linking to avoid references to undefined symbols.
9496 @item -Xlinker @var{option}
9498 Pass @var{option} as an option to the linker. You can use this to
9499 supply system-specific linker options which GCC does not know how to
9502 If you want to pass an option that takes a separate argument, you must use
9503 @option{-Xlinker} twice, once for the option and once for the argument.
9504 For example, to pass @option{-assert definitions}, you must write
9505 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9506 @option{-Xlinker "-assert definitions"}, because this passes the entire
9507 string as a single argument, which is not what the linker expects.
9509 When using the GNU linker, it is usually more convenient to pass
9510 arguments to linker options using the @option{@var{option}=@var{value}}
9511 syntax than as separate arguments. For example, you can specify
9512 @samp{-Xlinker -Map=output.map} rather than
9513 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9514 this syntax for command-line options.
9516 @item -Wl,@var{option}
9518 Pass @var{option} as an option to the linker. If @var{option} contains
9519 commas, it is split into multiple options at the commas. You can use this
9520 syntax to pass an argument to the option.
9521 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9522 linker. When using the GNU linker, you can also get the same effect with
9523 @samp{-Wl,-Map=output.map}.
9525 @item -u @var{symbol}
9527 Pretend the symbol @var{symbol} is undefined, to force linking of
9528 library modules to define it. You can use @option{-u} multiple times with
9529 different symbols to force loading of additional library modules.
9532 @node Directory Options
9533 @section Options for Directory Search
9534 @cindex directory options
9535 @cindex options, directory search
9538 These options specify directories to search for header files, for
9539 libraries and for parts of the compiler:
9544 Add the directory @var{dir} to the head of the list of directories to be
9545 searched for header files. This can be used to override a system header
9546 file, substituting your own version, since these directories are
9547 searched before the system header file directories. However, you should
9548 not use this option to add directories that contain vendor-supplied
9549 system header files (use @option{-isystem} for that). If you use more than
9550 one @option{-I} option, the directories are scanned in left-to-right
9551 order; the standard system directories come after.
9553 If a standard system include directory, or a directory specified with
9554 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9555 option will be ignored. The directory will still be searched but as a
9556 system directory at its normal position in the system include chain.
9557 This is to ensure that GCC's procedure to fix buggy system headers and
9558 the ordering for the include_next directive are not inadvertently changed.
9559 If you really need to change the search order for system directories,
9560 use the @option{-nostdinc} and/or @option{-isystem} options.
9562 @item -iplugindir=@var{dir}
9563 Set the directory to search for plugins which are passed
9564 by @option{-fplugin=@var{name}} instead of
9565 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9566 to be used by the user, but only passed by the driver.
9568 @item -iquote@var{dir}
9570 Add the directory @var{dir} to the head of the list of directories to
9571 be searched for header files only for the case of @samp{#include
9572 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9573 otherwise just like @option{-I}.
9577 Add directory @var{dir} to the list of directories to be searched
9580 @item -B@var{prefix}
9582 This option specifies where to find the executables, libraries,
9583 include files, and data files of the compiler itself.
9585 The compiler driver program runs one or more of the subprograms
9586 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9587 @var{prefix} as a prefix for each program it tries to run, both with and
9588 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9590 For each subprogram to be run, the compiler driver first tries the
9591 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9592 was not specified, the driver tries two standard prefixes, which are
9593 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9594 those results in a file name that is found, the unmodified program
9595 name is searched for using the directories specified in your
9596 @env{PATH} environment variable.
9598 The compiler will check to see if the path provided by the @option{-B}
9599 refers to a directory, and if necessary it will add a directory
9600 separator character at the end of the path.
9602 @option{-B} prefixes that effectively specify directory names also apply
9603 to libraries in the linker, because the compiler translates these
9604 options into @option{-L} options for the linker. They also apply to
9605 includes files in the preprocessor, because the compiler translates these
9606 options into @option{-isystem} options for the preprocessor. In this case,
9607 the compiler appends @samp{include} to the prefix.
9609 The run-time support file @file{libgcc.a} can also be searched for using
9610 the @option{-B} prefix, if needed. If it is not found there, the two
9611 standard prefixes above are tried, and that is all. The file is left
9612 out of the link if it is not found by those means.
9614 Another way to specify a prefix much like the @option{-B} prefix is to use
9615 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9618 As a special kludge, if the path provided by @option{-B} is
9619 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9620 9, then it will be replaced by @file{[dir/]include}. This is to help
9621 with boot-strapping the compiler.
9623 @item -specs=@var{file}
9625 Process @var{file} after the compiler reads in the standard @file{specs}
9626 file, in order to override the defaults that the @file{gcc} driver
9627 program uses when determining what switches to pass to @file{cc1},
9628 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9629 @option{-specs=@var{file}} can be specified on the command line, and they
9630 are processed in order, from left to right.
9632 @item --sysroot=@var{dir}
9634 Use @var{dir} as the logical root directory for headers and libraries.
9635 For example, if the compiler would normally search for headers in
9636 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9637 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9639 If you use both this option and the @option{-isysroot} option, then
9640 the @option{--sysroot} option will apply to libraries, but the
9641 @option{-isysroot} option will apply to header files.
9643 The GNU linker (beginning with version 2.16) has the necessary support
9644 for this option. If your linker does not support this option, the
9645 header file aspect of @option{--sysroot} will still work, but the
9646 library aspect will not.
9650 This option has been deprecated. Please use @option{-iquote} instead for
9651 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9652 Any directories you specify with @option{-I} options before the @option{-I-}
9653 option are searched only for the case of @samp{#include "@var{file}"};
9654 they are not searched for @samp{#include <@var{file}>}.
9656 If additional directories are specified with @option{-I} options after
9657 the @option{-I-}, these directories are searched for all @samp{#include}
9658 directives. (Ordinarily @emph{all} @option{-I} directories are used
9661 In addition, the @option{-I-} option inhibits the use of the current
9662 directory (where the current input file came from) as the first search
9663 directory for @samp{#include "@var{file}"}. There is no way to
9664 override this effect of @option{-I-}. With @option{-I.} you can specify
9665 searching the directory which was current when the compiler was
9666 invoked. That is not exactly the same as what the preprocessor does
9667 by default, but it is often satisfactory.
9669 @option{-I-} does not inhibit the use of the standard system directories
9670 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9677 @section Specifying subprocesses and the switches to pass to them
9680 @command{gcc} is a driver program. It performs its job by invoking a
9681 sequence of other programs to do the work of compiling, assembling and
9682 linking. GCC interprets its command-line parameters and uses these to
9683 deduce which programs it should invoke, and which command-line options
9684 it ought to place on their command lines. This behavior is controlled
9685 by @dfn{spec strings}. In most cases there is one spec string for each
9686 program that GCC can invoke, but a few programs have multiple spec
9687 strings to control their behavior. The spec strings built into GCC can
9688 be overridden by using the @option{-specs=} command-line switch to specify
9691 @dfn{Spec files} are plaintext files that are used to construct spec
9692 strings. They consist of a sequence of directives separated by blank
9693 lines. The type of directive is determined by the first non-whitespace
9694 character on the line and it can be one of the following:
9697 @item %@var{command}
9698 Issues a @var{command} to the spec file processor. The commands that can
9702 @item %include <@var{file}>
9703 @cindex @code{%include}
9704 Search for @var{file} and insert its text at the current point in the
9707 @item %include_noerr <@var{file}>
9708 @cindex @code{%include_noerr}
9709 Just like @samp{%include}, but do not generate an error message if the include
9710 file cannot be found.
9712 @item %rename @var{old_name} @var{new_name}
9713 @cindex @code{%rename}
9714 Rename the spec string @var{old_name} to @var{new_name}.
9718 @item *[@var{spec_name}]:
9719 This tells the compiler to create, override or delete the named spec
9720 string. All lines after this directive up to the next directive or
9721 blank line are considered to be the text for the spec string. If this
9722 results in an empty string then the spec will be deleted. (Or, if the
9723 spec did not exist, then nothing will happen.) Otherwise, if the spec
9724 does not currently exist a new spec will be created. If the spec does
9725 exist then its contents will be overridden by the text of this
9726 directive, unless the first character of that text is the @samp{+}
9727 character, in which case the text will be appended to the spec.
9729 @item [@var{suffix}]:
9730 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9731 and up to the next directive or blank line are considered to make up the
9732 spec string for the indicated suffix. When the compiler encounters an
9733 input file with the named suffix, it will processes the spec string in
9734 order to work out how to compile that file. For example:
9741 This says that any input file whose name ends in @samp{.ZZ} should be
9742 passed to the program @samp{z-compile}, which should be invoked with the
9743 command-line switch @option{-input} and with the result of performing the
9744 @samp{%i} substitution. (See below.)
9746 As an alternative to providing a spec string, the text that follows a
9747 suffix directive can be one of the following:
9750 @item @@@var{language}
9751 This says that the suffix is an alias for a known @var{language}. This is
9752 similar to using the @option{-x} command-line switch to GCC to specify a
9753 language explicitly. For example:
9760 Says that .ZZ files are, in fact, C++ source files.
9763 This causes an error messages saying:
9766 @var{name} compiler not installed on this system.
9770 GCC already has an extensive list of suffixes built into it.
9771 This directive will add an entry to the end of the list of suffixes, but
9772 since the list is searched from the end backwards, it is effectively
9773 possible to override earlier entries using this technique.
9777 GCC has the following spec strings built into it. Spec files can
9778 override these strings or create their own. Note that individual
9779 targets can also add their own spec strings to this list.
9782 asm Options to pass to the assembler
9783 asm_final Options to pass to the assembler post-processor
9784 cpp Options to pass to the C preprocessor
9785 cc1 Options to pass to the C compiler
9786 cc1plus Options to pass to the C++ compiler
9787 endfile Object files to include at the end of the link
9788 link Options to pass to the linker
9789 lib Libraries to include on the command line to the linker
9790 libgcc Decides which GCC support library to pass to the linker
9791 linker Sets the name of the linker
9792 predefines Defines to be passed to the C preprocessor
9793 signed_char Defines to pass to CPP to say whether @code{char} is signed
9795 startfile Object files to include at the start of the link
9798 Here is a small example of a spec file:
9804 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9807 This example renames the spec called @samp{lib} to @samp{old_lib} and
9808 then overrides the previous definition of @samp{lib} with a new one.
9809 The new definition adds in some extra command-line options before
9810 including the text of the old definition.
9812 @dfn{Spec strings} are a list of command-line options to be passed to their
9813 corresponding program. In addition, the spec strings can contain
9814 @samp{%}-prefixed sequences to substitute variable text or to
9815 conditionally insert text into the command line. Using these constructs
9816 it is possible to generate quite complex command lines.
9818 Here is a table of all defined @samp{%}-sequences for spec
9819 strings. Note that spaces are not generated automatically around the
9820 results of expanding these sequences. Therefore you can concatenate them
9821 together or combine them with constant text in a single argument.
9825 Substitute one @samp{%} into the program name or argument.
9828 Substitute the name of the input file being processed.
9831 Substitute the basename of the input file being processed.
9832 This is the substring up to (and not including) the last period
9833 and not including the directory.
9836 This is the same as @samp{%b}, but include the file suffix (text after
9840 Marks the argument containing or following the @samp{%d} as a
9841 temporary file name, so that that file will be deleted if GCC exits
9842 successfully. Unlike @samp{%g}, this contributes no text to the
9845 @item %g@var{suffix}
9846 Substitute a file name that has suffix @var{suffix} and is chosen
9847 once per compilation, and mark the argument in the same way as
9848 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9849 name is now chosen in a way that is hard to predict even when previously
9850 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9851 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9852 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9853 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9854 was simply substituted with a file name chosen once per compilation,
9855 without regard to any appended suffix (which was therefore treated
9856 just like ordinary text), making such attacks more likely to succeed.
9858 @item %u@var{suffix}
9859 Like @samp{%g}, but generates a new temporary file name even if
9860 @samp{%u@var{suffix}} was already seen.
9862 @item %U@var{suffix}
9863 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9864 new one if there is no such last file name. In the absence of any
9865 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9866 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9867 would involve the generation of two distinct file names, one
9868 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9869 simply substituted with a file name chosen for the previous @samp{%u},
9870 without regard to any appended suffix.
9872 @item %j@var{suffix}
9873 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9874 writable, and if save-temps is off; otherwise, substitute the name
9875 of a temporary file, just like @samp{%u}. This temporary file is not
9876 meant for communication between processes, but rather as a junk
9879 @item %|@var{suffix}
9880 @itemx %m@var{suffix}
9881 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9882 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9883 all. These are the two most common ways to instruct a program that it
9884 should read from standard input or write to standard output. If you
9885 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9886 construct: see for example @file{f/lang-specs.h}.
9888 @item %.@var{SUFFIX}
9889 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9890 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9891 terminated by the next space or %.
9894 Marks the argument containing or following the @samp{%w} as the
9895 designated output file of this compilation. This puts the argument
9896 into the sequence of arguments that @samp{%o} will substitute later.
9899 Substitutes the names of all the output files, with spaces
9900 automatically placed around them. You should write spaces
9901 around the @samp{%o} as well or the results are undefined.
9902 @samp{%o} is for use in the specs for running the linker.
9903 Input files whose names have no recognized suffix are not compiled
9904 at all, but they are included among the output files, so they will
9908 Substitutes the suffix for object files. Note that this is
9909 handled specially when it immediately follows @samp{%g, %u, or %U},
9910 because of the need for those to form complete file names. The
9911 handling is such that @samp{%O} is treated exactly as if it had already
9912 been substituted, except that @samp{%g, %u, and %U} do not currently
9913 support additional @var{suffix} characters following @samp{%O} as they would
9914 following, for example, @samp{.o}.
9917 Substitutes the standard macro predefinitions for the
9918 current target machine. Use this when running @code{cpp}.
9921 Like @samp{%p}, but puts @samp{__} before and after the name of each
9922 predefined macro, except for macros that start with @samp{__} or with
9923 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9927 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9928 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9929 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9930 and @option{-imultilib} as necessary.
9933 Current argument is the name of a library or startup file of some sort.
9934 Search for that file in a standard list of directories and substitute
9935 the full name found. The current working directory is included in the
9936 list of directories scanned.
9939 Current argument is the name of a linker script. Search for that file
9940 in the current list of directories to scan for libraries. If the file
9941 is located insert a @option{--script} option into the command line
9942 followed by the full path name found. If the file is not found then
9943 generate an error message. Note: the current working directory is not
9947 Print @var{str} as an error message. @var{str} is terminated by a newline.
9948 Use this when inconsistent options are detected.
9951 Substitute the contents of spec string @var{name} at this point.
9953 @item %x@{@var{option}@}
9954 Accumulate an option for @samp{%X}.
9957 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9961 Output the accumulated assembler options specified by @option{-Wa}.
9964 Output the accumulated preprocessor options specified by @option{-Wp}.
9967 Process the @code{asm} spec. This is used to compute the
9968 switches to be passed to the assembler.
9971 Process the @code{asm_final} spec. This is a spec string for
9972 passing switches to an assembler post-processor, if such a program is
9976 Process the @code{link} spec. This is the spec for computing the
9977 command line passed to the linker. Typically it will make use of the
9978 @samp{%L %G %S %D and %E} sequences.
9981 Dump out a @option{-L} option for each directory that GCC believes might
9982 contain startup files. If the target supports multilibs then the
9983 current multilib directory will be prepended to each of these paths.
9986 Process the @code{lib} spec. This is a spec string for deciding which
9987 libraries should be included on the command line to the linker.
9990 Process the @code{libgcc} spec. This is a spec string for deciding
9991 which GCC support library should be included on the command line to the linker.
9994 Process the @code{startfile} spec. This is a spec for deciding which
9995 object files should be the first ones passed to the linker. Typically
9996 this might be a file named @file{crt0.o}.
9999 Process the @code{endfile} spec. This is a spec string that specifies
10000 the last object files that will be passed to the linker.
10003 Process the @code{cpp} spec. This is used to construct the arguments
10004 to be passed to the C preprocessor.
10007 Process the @code{cc1} spec. This is used to construct the options to be
10008 passed to the actual C compiler (@samp{cc1}).
10011 Process the @code{cc1plus} spec. This is used to construct the options to be
10012 passed to the actual C++ compiler (@samp{cc1plus}).
10015 Substitute the variable part of a matched option. See below.
10016 Note that each comma in the substituted string is replaced by
10020 Remove all occurrences of @code{-S} from the command line. Note---this
10021 command is position dependent. @samp{%} commands in the spec string
10022 before this one will see @code{-S}, @samp{%} commands in the spec string
10023 after this one will not.
10025 @item %:@var{function}(@var{args})
10026 Call the named function @var{function}, passing it @var{args}.
10027 @var{args} is first processed as a nested spec string, then split
10028 into an argument vector in the usual fashion. The function returns
10029 a string which is processed as if it had appeared literally as part
10030 of the current spec.
10032 The following built-in spec functions are provided:
10035 @item @code{getenv}
10036 The @code{getenv} spec function takes two arguments: an environment
10037 variable name and a string. If the environment variable is not
10038 defined, a fatal error is issued. Otherwise, the return value is the
10039 value of the environment variable concatenated with the string. For
10040 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
10043 %:getenv(TOPDIR /include)
10046 expands to @file{/path/to/top/include}.
10048 @item @code{if-exists}
10049 The @code{if-exists} spec function takes one argument, an absolute
10050 pathname to a file. If the file exists, @code{if-exists} returns the
10051 pathname. Here is a small example of its usage:
10055 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
10058 @item @code{if-exists-else}
10059 The @code{if-exists-else} spec function is similar to the @code{if-exists}
10060 spec function, except that it takes two arguments. The first argument is
10061 an absolute pathname to a file. If the file exists, @code{if-exists-else}
10062 returns the pathname. If it does not exist, it returns the second argument.
10063 This way, @code{if-exists-else} can be used to select one file or another,
10064 based on the existence of the first. Here is a small example of its usage:
10068 crt0%O%s %:if-exists(crti%O%s) \
10069 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
10072 @item @code{replace-outfile}
10073 The @code{replace-outfile} spec function takes two arguments. It looks for the
10074 first argument in the outfiles array and replaces it with the second argument. Here
10075 is a small example of its usage:
10078 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10081 @item @code{remove-outfile}
10082 The @code{remove-outfile} spec function takes one argument. It looks for the
10083 first argument in the outfiles array and removes it. Here is a small example
10087 %:remove-outfile(-lm)
10090 @item @code{pass-through-libs}
10091 The @code{pass-through-libs} spec function takes any number of arguments. It
10092 finds any @option{-l} options and any non-options ending in ".a" (which it
10093 assumes are the names of linker input library archive files) and returns a
10094 result containing all the found arguments each prepended by
10095 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10096 intended to be passed to the LTO linker plugin.
10099 %:pass-through-libs(%G %L %G)
10102 @item @code{print-asm-header}
10103 The @code{print-asm-header} function takes no arguments and simply
10104 prints a banner like:
10110 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10113 It is used to separate compiler options from assembler options
10114 in the @option{--target-help} output.
10117 @item %@{@code{S}@}
10118 Substitutes the @code{-S} switch, if that switch was given to GCC@.
10119 If that switch was not specified, this substitutes nothing. Note that
10120 the leading dash is omitted when specifying this option, and it is
10121 automatically inserted if the substitution is performed. Thus the spec
10122 string @samp{%@{foo@}} would match the command-line option @option{-foo}
10123 and would output the command line option @option{-foo}.
10125 @item %W@{@code{S}@}
10126 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10127 deleted on failure.
10129 @item %@{@code{S}*@}
10130 Substitutes all the switches specified to GCC whose names start
10131 with @code{-S}, but which also take an argument. This is used for
10132 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10133 GCC considers @option{-o foo} as being
10134 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10135 text, including the space. Thus two arguments would be generated.
10137 @item %@{@code{S}*&@code{T}*@}
10138 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10139 (the order of @code{S} and @code{T} in the spec is not significant).
10140 There can be any number of ampersand-separated variables; for each the
10141 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10143 @item %@{@code{S}:@code{X}@}
10144 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10146 @item %@{!@code{S}:@code{X}@}
10147 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10149 @item %@{@code{S}*:@code{X}@}
10150 Substitutes @code{X} if one or more switches whose names start with
10151 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10152 once, no matter how many such switches appeared. However, if @code{%*}
10153 appears somewhere in @code{X}, then @code{X} will be substituted once
10154 for each matching switch, with the @code{%*} replaced by the part of
10155 that switch that matched the @code{*}.
10157 @item %@{.@code{S}:@code{X}@}
10158 Substitutes @code{X}, if processing a file with suffix @code{S}.
10160 @item %@{!.@code{S}:@code{X}@}
10161 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10163 @item %@{,@code{S}:@code{X}@}
10164 Substitutes @code{X}, if processing a file for language @code{S}.
10166 @item %@{!,@code{S}:@code{X}@}
10167 Substitutes @code{X}, if not processing a file for language @code{S}.
10169 @item %@{@code{S}|@code{P}:@code{X}@}
10170 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10171 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10172 @code{*} sequences as well, although they have a stronger binding than
10173 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10174 alternatives must be starred, and only the first matching alternative
10177 For example, a spec string like this:
10180 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10183 will output the following command-line options from the following input
10184 command-line options:
10189 -d fred.c -foo -baz -boggle
10190 -d jim.d -bar -baz -boggle
10193 @item %@{S:X; T:Y; :D@}
10195 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10196 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10197 be as many clauses as you need. This may be combined with @code{.},
10198 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10203 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10204 construct may contain other nested @samp{%} constructs or spaces, or
10205 even newlines. They are processed as usual, as described above.
10206 Trailing white space in @code{X} is ignored. White space may also
10207 appear anywhere on the left side of the colon in these constructs,
10208 except between @code{.} or @code{*} and the corresponding word.
10210 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10211 handled specifically in these constructs. If another value of
10212 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10213 @option{-W} switch is found later in the command line, the earlier
10214 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10215 just one letter, which passes all matching options.
10217 The character @samp{|} at the beginning of the predicate text is used to
10218 indicate that a command should be piped to the following command, but
10219 only if @option{-pipe} is specified.
10221 It is built into GCC which switches take arguments and which do not.
10222 (You might think it would be useful to generalize this to allow each
10223 compiler's spec to say which switches take arguments. But this cannot
10224 be done in a consistent fashion. GCC cannot even decide which input
10225 files have been specified without knowing which switches take arguments,
10226 and it must know which input files to compile in order to tell which
10229 GCC also knows implicitly that arguments starting in @option{-l} are to be
10230 treated as compiler output files, and passed to the linker in their
10231 proper position among the other output files.
10233 @c man begin OPTIONS
10235 @node Target Options
10236 @section Specifying Target Machine and Compiler Version
10237 @cindex target options
10238 @cindex cross compiling
10239 @cindex specifying machine version
10240 @cindex specifying compiler version and target machine
10241 @cindex compiler version, specifying
10242 @cindex target machine, specifying
10244 The usual way to run GCC is to run the executable called @command{gcc}, or
10245 @command{@var{machine}-gcc} when cross-compiling, or
10246 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10247 one that was installed last.
10249 @node Submodel Options
10250 @section Hardware Models and Configurations
10251 @cindex submodel options
10252 @cindex specifying hardware config
10253 @cindex hardware models and configurations, specifying
10254 @cindex machine dependent options
10256 Each target machine types can have its own
10257 special options, starting with @samp{-m}, to choose among various
10258 hardware models or configurations---for example, 68010 vs 68020,
10259 floating coprocessor or none. A single installed version of the
10260 compiler can compile for any model or configuration, according to the
10263 Some configurations of the compiler also support additional special
10264 options, usually for compatibility with other compilers on the same
10267 @c This list is ordered alphanumerically by subsection name.
10268 @c It should be the same order and spelling as these options are listed
10269 @c in Machine Dependent Options
10272 * Adapteva Epiphany Options::
10275 * Blackfin Options::
10279 * DEC Alpha Options::
10280 * DEC Alpha/VMS Options::
10283 * GNU/Linux Options::
10286 * i386 and x86-64 Options::
10287 * i386 and x86-64 Windows Options::
10289 * IA-64/VMS Options::
10296 * MicroBlaze Options::
10299 * MN10300 Options::
10301 * picoChip Options::
10302 * PowerPC Options::
10303 * RS/6000 and PowerPC Options::
10305 * S/390 and zSeries Options::
10308 * Solaris 2 Options::
10311 * System V Options::
10314 * VxWorks Options::
10316 * Xstormy16 Options::
10318 * zSeries Options::
10321 @node Adapteva Epiphany Options
10322 @subsection Adapteva Epiphany Options
10324 These @samp{-m} options are defined for Adapteva Epiphany:
10327 @item -mhalf-reg-file
10328 @opindex mhalf-reg-file
10329 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
10330 That allows code to run on hardware variants that lack these registers.
10332 @item -mprefer-short-insn-regs
10333 @opindex mprefer-short-insn-regs
10334 Preferrentially allocate registers that allow short instruction generation.
10335 This can result in increasesd instruction count, so if this reduces or
10336 increases code size might vary from case to case.
10338 @item -mbranch-cost=@var{num}
10339 @opindex mbranch-cost
10340 Set the cost of branches to roughly @var{num} ``simple'' instructions.
10341 This cost is only a heuristic and is not guaranteed to produce
10342 consistent results across releases.
10346 Enable the generation of conditional moves.
10348 @item -mnops=@var{num}
10350 Emit @var{num} nops before every other generated instruction.
10352 @item -mno-soft-cmpsf
10353 @opindex mno-soft-cmpsf
10354 For single-precision floating point comparisons, emit an fsub instruction
10355 and test the flags. This is faster than a software comparison, but can
10356 get incorrect results in the presence of NaNs, or when two different small
10357 numbers are compared such that their difference is calculated as zero.
10358 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
10359 software comparisons.
10361 @item -mstack-offset=@var{num}
10362 @opindex mstack-offset
10363 Set the offset between the top of the stack and the stack pointer.
10364 E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7
10365 can be used by leaf functions without stack allocation.
10366 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
10367 Note also that this option changes the ABI, compiling a program with a
10368 different stack offset than the libraries have been compiled with
10369 will generally not work.
10370 This option can be useful if you want to evaluate if a different stack
10371 offset would give you better code, but to actually use a different stack
10372 offset to build working programs, it is recommended to configure the
10373 toolchain with the appropriate @samp{--with-stack-offset=@var{num}} option.
10375 @item -mno-round-nearest
10376 @opindex mno-round-nearest
10377 Make the scheduler assume that the rounding mode has been set to
10378 truncating. The default is @option{-mround-nearest}.
10381 @opindex mlong-calls
10382 If not otherwise specified by an attribute, assume all calls might be beyond
10383 the offset range of the b / bl instructions, and therefore load the
10384 function address into a register before performing a (otherwise direct) call.
10385 This is the default.
10387 @item -mshort-calls
10388 @opindex short-calls
10389 If not otherwise specified by an attribute, assume all direct calls are
10390 in the range of the b / bl instructions, so use these instructions
10391 for direct calls. The default is @option{-mlong-calls}.
10395 Assume addresses can be loaded as 16 bit unsigned values. This does not
10396 apply to function addresses for which @option{-mlong-calls} semantics
10399 @item -mfp-mode=@var{mode}
10401 Set the prevailing mode of the floating point unit.
10402 This determines the floating point mode that is provided and expected
10403 at function call and return time. Making this mode match the mode you
10404 predominantly need at function start can make your programs smaller and
10405 faster by avoiding unnecessary mode switches.
10407 @var{mode} can be set to one the following values:
10411 Any mode at function entry is valid, and retained or restored when
10412 the function returns, and when it calls other functions.
10413 This mode is useful for compiling libraries or other compilation units
10414 you might want to incorporate into different programs with different
10415 prevailing FPU modes, and the convenience of being able to use a single
10416 object file outweighs the size and speed overhead for any extra
10417 mode switching that might be needed, compared with what would be needed
10418 with a more specific choice of prevailing FPU mode.
10421 This is the mode used for floating point calculations with
10422 truncating (i.e.@: round towards zero) rounding mode. That includes
10423 conversion from floating point to integer.
10425 @item round-nearest
10426 This is the mode used for floating point calculations with
10427 round-to-nearest-or-even rounding mode.
10430 This is the mode used to perform integer calculations in the FPU, e.g.@:
10431 integer multiply, or integer multiply-and-accumulate.
10434 The default is @option{-mfp-mode=caller}
10436 @item -mnosplit-lohi
10437 @opindex mnosplit-lohi
10439 @opindex mno-postinc
10440 @item -mno-postmodify
10441 @opindex mno-postmodify
10442 Code generation tweaks that disable, respectively, splitting of 32
10443 bit loads, generation of post-increment addresses, and generation of
10444 post-modify addresses. The defaults are @option{msplit-lohi},
10445 @option{-mpost-inc}, and @option{-mpost-modify}.
10447 @item -mnovect-double
10448 @opindex mno-vect-double
10449 Change the preferred SIMD mode to SImode. The default is
10450 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
10452 @item -max-vect-align=@var{num}
10453 @opindex max-vect-align
10454 The maximum alignment for SIMD vector mode types.
10455 @var{num} may be 4 or 8. The default is 8.
10456 Note that this is an ABI change, even though many library function
10457 interfaces will be unaffected, if they don't use SIMD vector modes
10458 in places where they affect size and/or alignment of relevant types.
10460 @item -msplit-vecmove-early
10461 @opindex msplit-vecmove-early
10462 Split vector moves into single word moves before reload. In theory this
10463 could give better register allocation, but so far the reverse seems to be
10464 generally the case.
10466 @item -m1reg-@var{reg}
10468 Specify a register to hold the constant @minus{}1, which makes loading small negative
10469 constants and certain bitmasks faster.
10470 Allowable values for reg are r43 and r63, which specify to use that register
10471 as a fixed register, and none, which means that no register is used for this
10472 purpose. The default is @option{-m1reg-none}.
10477 @subsection ARM Options
10478 @cindex ARM options
10480 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10484 @item -mabi=@var{name}
10486 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10487 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10490 @opindex mapcs-frame
10491 Generate a stack frame that is compliant with the ARM Procedure Call
10492 Standard for all functions, even if this is not strictly necessary for
10493 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10494 with this option will cause the stack frames not to be generated for
10495 leaf functions. The default is @option{-mno-apcs-frame}.
10499 This is a synonym for @option{-mapcs-frame}.
10502 @c not currently implemented
10503 @item -mapcs-stack-check
10504 @opindex mapcs-stack-check
10505 Generate code to check the amount of stack space available upon entry to
10506 every function (that actually uses some stack space). If there is
10507 insufficient space available then either the function
10508 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10509 called, depending upon the amount of stack space required. The run time
10510 system is required to provide these functions. The default is
10511 @option{-mno-apcs-stack-check}, since this produces smaller code.
10513 @c not currently implemented
10515 @opindex mapcs-float
10516 Pass floating point arguments using the float point registers. This is
10517 one of the variants of the APCS@. This option is recommended if the
10518 target hardware has a floating point unit or if a lot of floating point
10519 arithmetic is going to be performed by the code. The default is
10520 @option{-mno-apcs-float}, since integer only code is slightly increased in
10521 size if @option{-mapcs-float} is used.
10523 @c not currently implemented
10524 @item -mapcs-reentrant
10525 @opindex mapcs-reentrant
10526 Generate reentrant, position independent code. The default is
10527 @option{-mno-apcs-reentrant}.
10530 @item -mthumb-interwork
10531 @opindex mthumb-interwork
10532 Generate code which supports calling between the ARM and Thumb
10533 instruction sets. Without this option, on pre-v5 architectures, the
10534 two instruction sets cannot be reliably used inside one program. The
10535 default is @option{-mno-thumb-interwork}, since slightly larger code
10536 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10537 configurations this option is meaningless.
10539 @item -mno-sched-prolog
10540 @opindex mno-sched-prolog
10541 Prevent the reordering of instructions in the function prolog, or the
10542 merging of those instruction with the instructions in the function's
10543 body. This means that all functions will start with a recognizable set
10544 of instructions (or in fact one of a choice from a small set of
10545 different function prologues), and this information can be used to
10546 locate the start if functions inside an executable piece of code. The
10547 default is @option{-msched-prolog}.
10549 @item -mfloat-abi=@var{name}
10550 @opindex mfloat-abi
10551 Specifies which floating-point ABI to use. Permissible values
10552 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10554 Specifying @samp{soft} causes GCC to generate output containing
10555 library calls for floating-point operations.
10556 @samp{softfp} allows the generation of code using hardware floating-point
10557 instructions, but still uses the soft-float calling conventions.
10558 @samp{hard} allows generation of floating-point instructions
10559 and uses FPU-specific calling conventions.
10561 The default depends on the specific target configuration. Note that
10562 the hard-float and soft-float ABIs are not link-compatible; you must
10563 compile your entire program with the same ABI, and link with a
10564 compatible set of libraries.
10566 @item -mlittle-endian
10567 @opindex mlittle-endian
10568 Generate code for a processor running in little-endian mode. This is
10569 the default for all standard configurations.
10572 @opindex mbig-endian
10573 Generate code for a processor running in big-endian mode; the default is
10574 to compile code for a little-endian processor.
10576 @item -mwords-little-endian
10577 @opindex mwords-little-endian
10578 This option only applies when generating code for big-endian processors.
10579 Generate code for a little-endian word order but a big-endian byte
10580 order. That is, a byte order of the form @samp{32107654}. Note: this
10581 option should only be used if you require compatibility with code for
10582 big-endian ARM processors generated by versions of the compiler prior to
10583 2.8. This option is now deprecated.
10585 @item -mcpu=@var{name}
10587 This specifies the name of the target ARM processor. GCC uses this name
10588 to determine what kind of instructions it can emit when generating
10589 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10590 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10591 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10592 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10593 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10595 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10596 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10597 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10598 @samp{strongarm1110},
10599 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10600 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10601 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10602 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10603 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10604 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10605 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10606 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
10607 @samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10608 @samp{cortex-m4}, @samp{cortex-m3},
10611 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10612 @samp{fa526}, @samp{fa626},
10613 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10616 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10617 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10618 See @option{-mtune} for more information.
10620 @option{-mcpu=native} causes the compiler to auto-detect the CPU
10621 of the build computer. At present, this feature is only supported on
10622 Linux, and not all architectures are recognised. If the auto-detect is
10623 unsuccessful the option has no effect.
10625 @item -mtune=@var{name}
10627 This option is very similar to the @option{-mcpu=} option, except that
10628 instead of specifying the actual target processor type, and hence
10629 restricting which instructions can be used, it specifies that GCC should
10630 tune the performance of the code as if the target were of the type
10631 specified in this option, but still choosing the instructions that it
10632 will generate based on the CPU specified by a @option{-mcpu=} option.
10633 For some ARM implementations better performance can be obtained by using
10636 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10637 performance for a blend of processors within architecture @var{arch}.
10638 The aim is to generate code that run well on the current most popular
10639 processors, balancing between optimizations that benefit some CPUs in the
10640 range, and avoiding performance pitfalls of other CPUs. The effects of
10641 this option may change in future GCC versions as CPU models come and go.
10643 @option{-mtune=native} causes the compiler to auto-detect the CPU
10644 of the build computer. At present, this feature is only supported on
10645 Linux, and not all architectures are recognised. If the auto-detect is
10646 unsuccessful the option has no effect.
10648 @item -march=@var{name}
10650 This specifies the name of the target ARM architecture. GCC uses this
10651 name to determine what kind of instructions it can emit when generating
10652 assembly code. This option can be used in conjunction with or instead
10653 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10654 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10655 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10656 @samp{armv6}, @samp{armv6j},
10657 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10658 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10659 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10661 @option{-march=native} causes the compiler to auto-detect the architecture
10662 of the build computer. At present, this feature is only supported on
10663 Linux, and not all architectures are recognised. If the auto-detect is
10664 unsuccessful the option has no effect.
10666 @item -mfpu=@var{name}
10667 @itemx -mfpe=@var{number}
10668 @itemx -mfp=@var{number}
10672 This specifies what floating point hardware (or hardware emulation) is
10673 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10674 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10675 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10676 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10677 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10678 @option{-mfp} and @option{-mfpe} are synonyms for
10679 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10682 If @option{-msoft-float} is specified this specifies the format of
10683 floating point values.
10685 If the selected floating-point hardware includes the NEON extension
10686 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10687 operations will not be used by GCC's auto-vectorization pass unless
10688 @option{-funsafe-math-optimizations} is also specified. This is
10689 because NEON hardware does not fully implement the IEEE 754 standard for
10690 floating-point arithmetic (in particular denormal values are treated as
10691 zero), so the use of NEON instructions may lead to a loss of precision.
10693 @item -mfp16-format=@var{name}
10694 @opindex mfp16-format
10695 Specify the format of the @code{__fp16} half-precision floating-point type.
10696 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10697 the default is @samp{none}, in which case the @code{__fp16} type is not
10698 defined. @xref{Half-Precision}, for more information.
10700 @item -mstructure-size-boundary=@var{n}
10701 @opindex mstructure-size-boundary
10702 The size of all structures and unions will be rounded up to a multiple
10703 of the number of bits set by this option. Permissible values are 8, 32
10704 and 64. The default value varies for different toolchains. For the COFF
10705 targeted toolchain the default value is 8. A value of 64 is only allowed
10706 if the underlying ABI supports it.
10708 Specifying the larger number can produce faster, more efficient code, but
10709 can also increase the size of the program. Different values are potentially
10710 incompatible. Code compiled with one value cannot necessarily expect to
10711 work with code or libraries compiled with another value, if they exchange
10712 information using structures or unions.
10714 @item -mabort-on-noreturn
10715 @opindex mabort-on-noreturn
10716 Generate a call to the function @code{abort} at the end of a
10717 @code{noreturn} function. It will be executed if the function tries to
10721 @itemx -mno-long-calls
10722 @opindex mlong-calls
10723 @opindex mno-long-calls
10724 Tells the compiler to perform function calls by first loading the
10725 address of the function into a register and then performing a subroutine
10726 call on this register. This switch is needed if the target function
10727 will lie outside of the 64 megabyte addressing range of the offset based
10728 version of subroutine call instruction.
10730 Even if this switch is enabled, not all function calls will be turned
10731 into long calls. The heuristic is that static functions, functions
10732 which have the @samp{short-call} attribute, functions that are inside
10733 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10734 definitions have already been compiled within the current compilation
10735 unit, will not be turned into long calls. The exception to this rule is
10736 that weak function definitions, functions with the @samp{long-call}
10737 attribute or the @samp{section} attribute, and functions that are within
10738 the scope of a @samp{#pragma long_calls} directive, will always be
10739 turned into long calls.
10741 This feature is not enabled by default. Specifying
10742 @option{-mno-long-calls} will restore the default behavior, as will
10743 placing the function calls within the scope of a @samp{#pragma
10744 long_calls_off} directive. Note these switches have no effect on how
10745 the compiler generates code to handle function calls via function
10748 @item -msingle-pic-base
10749 @opindex msingle-pic-base
10750 Treat the register used for PIC addressing as read-only, rather than
10751 loading it in the prologue for each function. The run-time system is
10752 responsible for initializing this register with an appropriate value
10753 before execution begins.
10755 @item -mpic-register=@var{reg}
10756 @opindex mpic-register
10757 Specify the register to be used for PIC addressing. The default is R10
10758 unless stack-checking is enabled, when R9 is used.
10760 @item -mcirrus-fix-invalid-insns
10761 @opindex mcirrus-fix-invalid-insns
10762 @opindex mno-cirrus-fix-invalid-insns
10763 Insert NOPs into the instruction stream to in order to work around
10764 problems with invalid Maverick instruction combinations. This option
10765 is only valid if the @option{-mcpu=ep9312} option has been used to
10766 enable generation of instructions for the Cirrus Maverick floating
10767 point co-processor. This option is not enabled by default, since the
10768 problem is only present in older Maverick implementations. The default
10769 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10772 @item -mpoke-function-name
10773 @opindex mpoke-function-name
10774 Write the name of each function into the text section, directly
10775 preceding the function prologue. The generated code is similar to this:
10779 .ascii "arm_poke_function_name", 0
10782 .word 0xff000000 + (t1 - t0)
10783 arm_poke_function_name
10785 stmfd sp!, @{fp, ip, lr, pc@}
10789 When performing a stack backtrace, code can inspect the value of
10790 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10791 location @code{pc - 12} and the top 8 bits are set, then we know that
10792 there is a function name embedded immediately preceding this location
10793 and has length @code{((pc[-3]) & 0xff000000)}.
10800 Select between generating code that executes in ARM and Thumb
10801 states. The default for most configurations is to generate code
10802 that executes in ARM state, but the default can be changed by
10803 configuring GCC with the @option{--with-mode=}@var{state}
10807 @opindex mtpcs-frame
10808 Generate a stack frame that is compliant with the Thumb Procedure Call
10809 Standard for all non-leaf functions. (A leaf function is one that does
10810 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10812 @item -mtpcs-leaf-frame
10813 @opindex mtpcs-leaf-frame
10814 Generate a stack frame that is compliant with the Thumb Procedure Call
10815 Standard for all leaf functions. (A leaf function is one that does
10816 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10818 @item -mcallee-super-interworking
10819 @opindex mcallee-super-interworking
10820 Gives all externally visible functions in the file being compiled an ARM
10821 instruction set header which switches to Thumb mode before executing the
10822 rest of the function. This allows these functions to be called from
10823 non-interworking code. This option is not valid in AAPCS configurations
10824 because interworking is enabled by default.
10826 @item -mcaller-super-interworking
10827 @opindex mcaller-super-interworking
10828 Allows calls via function pointers (including virtual functions) to
10829 execute correctly regardless of whether the target code has been
10830 compiled for interworking or not. There is a small overhead in the cost
10831 of executing a function pointer if this option is enabled. This option
10832 is not valid in AAPCS configurations because interworking is enabled
10835 @item -mtp=@var{name}
10837 Specify the access model for the thread local storage pointer. The valid
10838 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10839 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10840 (supported in the arm6k architecture), and @option{auto}, which uses the
10841 best available method for the selected processor. The default setting is
10844 @item -mtls-dialect=@var{dialect}
10845 @opindex mtls-dialect
10846 Specify the dialect to use for accessing thread local storage. Two
10847 dialects are supported --- @option{gnu} and @option{gnu2}. The
10848 @option{gnu} dialect selects the original GNU scheme for supporting
10849 local and global dynamic TLS models. The @option{gnu2} dialect
10850 selects the GNU descriptor scheme, which provides better performance
10851 for shared libraries. The GNU descriptor scheme is compatible with
10852 the original scheme, but does require new assembler, linker and
10853 library support. Initial and local exec TLS models are unaffected by
10854 this option and always use the original scheme.
10856 @item -mword-relocations
10857 @opindex mword-relocations
10858 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10859 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10860 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10863 @item -mfix-cortex-m3-ldrd
10864 @opindex mfix-cortex-m3-ldrd
10865 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10866 with overlapping destination and base registers are used. This option avoids
10867 generating these instructions. This option is enabled by default when
10868 @option{-mcpu=cortex-m3} is specified.
10873 @subsection AVR Options
10874 @cindex AVR Options
10876 These options are defined for AVR implementations:
10879 @item -mmcu=@var{mcu}
10881 Specify ATMEL AVR instruction set or MCU type.
10883 Instruction set avr1 is for the minimal AVR core, not supported by the C
10884 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10885 attiny11, attiny12, attiny15, attiny28).
10887 Instruction set avr2 (default) is for the classic AVR core with up to
10888 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10889 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10890 at90c8534, at90s8535).
10892 Instruction set avr3 is for the classic AVR core with up to 128K program
10893 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10895 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10896 memory space (MCU types: atmega8, atmega83, atmega85).
10898 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10899 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10900 atmega64, atmega128, at43usb355, at94k).
10902 @item -mno-interrupts
10903 @opindex mno-interrupts
10904 Generated code is not compatible with hardware interrupts.
10905 Code size will be smaller.
10907 @item -mcall-prologues
10908 @opindex mcall-prologues
10909 Functions prologues/epilogues expanded as call to appropriate
10910 subroutines. Code size will be smaller.
10913 @opindex mtiny-stack
10914 Change only the low 8 bits of the stack pointer.
10918 Assume int to be 8 bit integer. This affects the sizes of all types: A
10919 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10920 and long long will be 4 bytes. Please note that this option does not
10921 comply to the C standards, but it will provide you with smaller code
10926 Use register @code{X} in a way proposed by the hardware. This means
10927 that @code{X} will only be used in indirect, post-increment or
10928 pre-decrement addressing.
10930 Without this option, the @code{X} register may be used in the same way
10931 as @code{Y} or @code{Z} which then is emulated by additional
10933 For example, loading a value with @code{X+const} addressing with a
10934 small @code{const <= 63} to a register @var{Rn} will be printed as
10942 @subsubsection @code{EIND} and Devices with more than 128k Bytes of Flash
10944 Pointers in the implementation are 16 bits wide.
10945 The address of a function or label is represented as word address so
10946 that indirect jumps and calls can address any code address in the
10947 range of 64k words.
10949 In order to faciliate indirect jump on devices with more than 128k
10950 bytes of program memory space, there is a special function register called
10951 @code{EIND} that serves as most significant part of the target address
10952 when @code{EICALL} or @code{EIJMP} instructions are used.
10954 Indirect jumps and calls on these devices are handled as follows and
10955 are subject to some limitations:
10960 The compiler never sets @code{EIND}.
10963 The startup code from libgcc never sets @code{EIND}.
10964 Notice that startup code is a blend of code from libgcc and avr-libc.
10965 For the impact of avr-libc on @code{EIND}, see the
10966 @w{@uref{http://nongnu.org/avr-libc/user-manual,avr-libc user manual}}.
10969 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
10970 instructions or might read @code{EIND} directly.
10973 The compiler assumes that @code{EIND} never changes during the startup
10974 code or run of the application. In particular, @code{EIND} is not
10975 saved/restored in function or interrupt service routine
10979 It is legitimate for user-specific startup code to set up @code{EIND}
10980 early, for example by means of initialization code located in
10981 section @code{.init3}, and thus prior to general startup code that
10982 initializes RAM and calls constructors.
10985 For indirect calls to functions and computed goto, the linker will
10986 generate @emph{stubs}. Stubs are jump pads sometimes also called
10987 @emph{trampolines}. Thus, the indirect call/jump will jump to such a stub.
10988 The stub contains a direct jump to the desired address.
10991 Stubs will be generated automatically by the linker if
10992 the following two conditions are met:
10995 @item The address of a label is taken by means of the @code{gs} modifier
10996 (short for @emph{generate stubs}) like so:
10998 LDI r24, lo8(gs(@var{func}))
10999 LDI r25, hi8(gs(@var{func}))
11001 @item The final location of that label is in a code segment
11002 @emph{outside} the segment where the stubs are located.
11006 The compiler will emit such @code{gs} modifiers for code labels in the
11007 following situations:
11009 @item Taking address of a function or code label.
11010 @item Computed goto.
11011 @item If prologue-save function is used, see @option{-mcall-prologues}
11012 command line option.
11013 @item Switch/case dispatch tables. If you do not want such dispatch
11014 tables you can specify the @option{-fno-jump-tables} command line option.
11015 @item C and C++ constructors/destructors called during startup/shutdown.
11016 @item If the tools hit a @code{gs()} modifier explained above.
11020 The default linker script is arranged for code with @code{EIND = 0}.
11021 If code is supposed to work for a setup with @code{EIND != 0}, a custom
11022 linker script has to be used in order to place the sections whose
11023 name start with @code{.trampolines} into the segment where @code{EIND}
11027 Jumping to non-symbolic addresses like so is @emph{not} supported:
11032 /* Call function at word address 0x2 */
11033 return ((int(*)(void)) 0x2)();
11037 Instead, a stub has to be set up:
11042 extern int func_4 (void);
11044 /* Call function at byte address 0x4 */
11049 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
11050 Alternatively, @code{func_4} can be defined in the linker script.
11053 @node Blackfin Options
11054 @subsection Blackfin Options
11055 @cindex Blackfin Options
11058 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
11060 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
11061 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
11062 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
11063 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
11064 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
11065 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
11066 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
11067 @samp{bf561}, @samp{bf592}.
11068 The optional @var{sirevision} specifies the silicon revision of the target
11069 Blackfin processor. Any workarounds available for the targeted silicon revision
11070 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
11071 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
11072 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
11073 hexadecimal digits representing the major and minor numbers in the silicon
11074 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
11075 is not defined. If @var{sirevision} is @samp{any}, the
11076 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
11077 If this optional @var{sirevision} is not used, GCC assumes the latest known
11078 silicon revision of the targeted Blackfin processor.
11080 Support for @samp{bf561} is incomplete. For @samp{bf561},
11081 Only the processor macro is defined.
11082 Without this option, @samp{bf532} is used as the processor by default.
11083 The corresponding predefined processor macros for @var{cpu} is to
11084 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
11085 provided by libgloss to be linked in if @option{-msim} is not given.
11089 Specifies that the program will be run on the simulator. This causes
11090 the simulator BSP provided by libgloss to be linked in. This option
11091 has effect only for @samp{bfin-elf} toolchain.
11092 Certain other options, such as @option{-mid-shared-library} and
11093 @option{-mfdpic}, imply @option{-msim}.
11095 @item -momit-leaf-frame-pointer
11096 @opindex momit-leaf-frame-pointer
11097 Don't keep the frame pointer in a register for leaf functions. This
11098 avoids the instructions to save, set up and restore frame pointers and
11099 makes an extra register available in leaf functions. The option
11100 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11101 which might make debugging harder.
11103 @item -mspecld-anomaly
11104 @opindex mspecld-anomaly
11105 When enabled, the compiler will ensure that the generated code does not
11106 contain speculative loads after jump instructions. If this option is used,
11107 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
11109 @item -mno-specld-anomaly
11110 @opindex mno-specld-anomaly
11111 Don't generate extra code to prevent speculative loads from occurring.
11113 @item -mcsync-anomaly
11114 @opindex mcsync-anomaly
11115 When enabled, the compiler will ensure that the generated code does not
11116 contain CSYNC or SSYNC instructions too soon after conditional branches.
11117 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
11119 @item -mno-csync-anomaly
11120 @opindex mno-csync-anomaly
11121 Don't generate extra code to prevent CSYNC or SSYNC instructions from
11122 occurring too soon after a conditional branch.
11126 When enabled, the compiler is free to take advantage of the knowledge that
11127 the entire program fits into the low 64k of memory.
11130 @opindex mno-low-64k
11131 Assume that the program is arbitrarily large. This is the default.
11133 @item -mstack-check-l1
11134 @opindex mstack-check-l1
11135 Do stack checking using information placed into L1 scratchpad memory by the
11138 @item -mid-shared-library
11139 @opindex mid-shared-library
11140 Generate code that supports shared libraries via the library ID method.
11141 This allows for execute in place and shared libraries in an environment
11142 without virtual memory management. This option implies @option{-fPIC}.
11143 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11145 @item -mno-id-shared-library
11146 @opindex mno-id-shared-library
11147 Generate code that doesn't assume ID based shared libraries are being used.
11148 This is the default.
11150 @item -mleaf-id-shared-library
11151 @opindex mleaf-id-shared-library
11152 Generate code that supports shared libraries via the library ID method,
11153 but assumes that this library or executable won't link against any other
11154 ID shared libraries. That allows the compiler to use faster code for jumps
11157 @item -mno-leaf-id-shared-library
11158 @opindex mno-leaf-id-shared-library
11159 Do not assume that the code being compiled won't link against any ID shared
11160 libraries. Slower code will be generated for jump and call insns.
11162 @item -mshared-library-id=n
11163 @opindex mshared-library-id
11164 Specified the identification number of the ID based shared library being
11165 compiled. Specifying a value of 0 will generate more compact code, specifying
11166 other values will force the allocation of that number to the current
11167 library but is no more space or time efficient than omitting this option.
11171 Generate code that allows the data segment to be located in a different
11172 area of memory from the text segment. This allows for execute in place in
11173 an environment without virtual memory management by eliminating relocations
11174 against the text section.
11176 @item -mno-sep-data
11177 @opindex mno-sep-data
11178 Generate code that assumes that the data segment follows the text segment.
11179 This is the default.
11182 @itemx -mno-long-calls
11183 @opindex mlong-calls
11184 @opindex mno-long-calls
11185 Tells the compiler to perform function calls by first loading the
11186 address of the function into a register and then performing a subroutine
11187 call on this register. This switch is needed if the target function
11188 will lie outside of the 24 bit addressing range of the offset based
11189 version of subroutine call instruction.
11191 This feature is not enabled by default. Specifying
11192 @option{-mno-long-calls} will restore the default behavior. Note these
11193 switches have no effect on how the compiler generates code to handle
11194 function calls via function pointers.
11198 Link with the fast floating-point library. This library relaxes some of
11199 the IEEE floating-point standard's rules for checking inputs against
11200 Not-a-Number (NAN), in the interest of performance.
11203 @opindex minline-plt
11204 Enable inlining of PLT entries in function calls to functions that are
11205 not known to bind locally. It has no effect without @option{-mfdpic}.
11208 @opindex mmulticore
11209 Build standalone application for multicore Blackfin processor. Proper
11210 start files and link scripts will be used to support multicore.
11211 This option defines @code{__BFIN_MULTICORE}. It can only be used with
11212 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
11213 @option{-mcorea} or @option{-mcoreb}. If it's used without
11214 @option{-mcorea} or @option{-mcoreb}, single application/dual core
11215 programming model is used. In this model, the main function of Core B
11216 should be named as coreb_main. If it's used with @option{-mcorea} or
11217 @option{-mcoreb}, one application per core programming model is used.
11218 If this option is not used, single core application programming
11223 Build standalone application for Core A of BF561 when using
11224 one application per core programming model. Proper start files
11225 and link scripts will be used to support Core A. This option
11226 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
11230 Build standalone application for Core B of BF561 when using
11231 one application per core programming model. Proper start files
11232 and link scripts will be used to support Core B. This option
11233 defines @code{__BFIN_COREB}. When this option is used, coreb_main
11234 should be used instead of main. It must be used with
11235 @option{-mmulticore}.
11239 Build standalone application for SDRAM. Proper start files and
11240 link scripts will be used to put the application into SDRAM.
11241 Loader should initialize SDRAM before loading the application
11242 into SDRAM. This option defines @code{__BFIN_SDRAM}.
11246 Assume that ICPLBs are enabled at runtime. This has an effect on certain
11247 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
11248 are enabled; for standalone applications the default is off.
11252 @subsection C6X Options
11253 @cindex C6X Options
11256 @item -march=@var{name}
11258 This specifies the name of the target architecture. GCC uses this
11259 name to determine what kind of instructions it can emit when generating
11260 assembly code. Permissible names are: @samp{c62x},
11261 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
11264 @opindex mbig-endian
11265 Generate code for a big endian target.
11267 @item -mlittle-endian
11268 @opindex mlittle-endian
11269 Generate code for a little endian target. This is the default.
11273 Choose startup files and linker script suitable for the simulator.
11275 @item -msdata=default
11276 @opindex msdata=default
11277 Put small global and static data in the @samp{.neardata} section,
11278 which is pointed to by register @code{B14}. Put small uninitialized
11279 global and static data in the @samp{.bss} section, which is adjacent
11280 to the @samp{.neardata} section. Put small read-only data into the
11281 @samp{.rodata} section. The corresponding sections used for large
11282 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
11285 @opindex msdata=all
11286 Put all data, not just small objets, into the sections reserved for
11287 small data, and use addressing relative to the @code{B14} register to
11291 @opindex msdata=none
11292 Make no use of the sections reserved for small data, and use absolute
11293 addresses to access all data. Put all initialized global and static
11294 data in the @samp{.fardata} section, and all uninitialized data in the
11295 @samp{.far} section. Put all constant data into the @samp{.const}
11300 @subsection CRIS Options
11301 @cindex CRIS Options
11303 These options are defined specifically for the CRIS ports.
11306 @item -march=@var{architecture-type}
11307 @itemx -mcpu=@var{architecture-type}
11310 Generate code for the specified architecture. The choices for
11311 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
11312 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
11313 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
11316 @item -mtune=@var{architecture-type}
11318 Tune to @var{architecture-type} everything applicable about the generated
11319 code, except for the ABI and the set of available instructions. The
11320 choices for @var{architecture-type} are the same as for
11321 @option{-march=@var{architecture-type}}.
11323 @item -mmax-stack-frame=@var{n}
11324 @opindex mmax-stack-frame
11325 Warn when the stack frame of a function exceeds @var{n} bytes.
11331 The options @option{-metrax4} and @option{-metrax100} are synonyms for
11332 @option{-march=v3} and @option{-march=v8} respectively.
11334 @item -mmul-bug-workaround
11335 @itemx -mno-mul-bug-workaround
11336 @opindex mmul-bug-workaround
11337 @opindex mno-mul-bug-workaround
11338 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
11339 models where it applies. This option is active by default.
11343 Enable CRIS-specific verbose debug-related information in the assembly
11344 code. This option also has the effect to turn off the @samp{#NO_APP}
11345 formatted-code indicator to the assembler at the beginning of the
11350 Do not use condition-code results from previous instruction; always emit
11351 compare and test instructions before use of condition codes.
11353 @item -mno-side-effects
11354 @opindex mno-side-effects
11355 Do not emit instructions with side-effects in addressing modes other than
11358 @item -mstack-align
11359 @itemx -mno-stack-align
11360 @itemx -mdata-align
11361 @itemx -mno-data-align
11362 @itemx -mconst-align
11363 @itemx -mno-const-align
11364 @opindex mstack-align
11365 @opindex mno-stack-align
11366 @opindex mdata-align
11367 @opindex mno-data-align
11368 @opindex mconst-align
11369 @opindex mno-const-align
11370 These options (no-options) arranges (eliminate arrangements) for the
11371 stack-frame, individual data and constants to be aligned for the maximum
11372 single data access size for the chosen CPU model. The default is to
11373 arrange for 32-bit alignment. ABI details such as structure layout are
11374 not affected by these options.
11382 Similar to the stack- data- and const-align options above, these options
11383 arrange for stack-frame, writable data and constants to all be 32-bit,
11384 16-bit or 8-bit aligned. The default is 32-bit alignment.
11386 @item -mno-prologue-epilogue
11387 @itemx -mprologue-epilogue
11388 @opindex mno-prologue-epilogue
11389 @opindex mprologue-epilogue
11390 With @option{-mno-prologue-epilogue}, the normal function prologue and
11391 epilogue that sets up the stack-frame are omitted and no return
11392 instructions or return sequences are generated in the code. Use this
11393 option only together with visual inspection of the compiled code: no
11394 warnings or errors are generated when call-saved registers must be saved,
11395 or storage for local variable needs to be allocated.
11399 @opindex mno-gotplt
11401 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11402 instruction sequences that load addresses for functions from the PLT part
11403 of the GOT rather than (traditional on other architectures) calls to the
11404 PLT@. The default is @option{-mgotplt}.
11408 Legacy no-op option only recognized with the cris-axis-elf and
11409 cris-axis-linux-gnu targets.
11413 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11417 This option, recognized for the cris-axis-elf arranges
11418 to link with input-output functions from a simulator library. Code,
11419 initialized data and zero-initialized data are allocated consecutively.
11423 Like @option{-sim}, but pass linker options to locate initialized data at
11424 0x40000000 and zero-initialized data at 0x80000000.
11427 @node Darwin Options
11428 @subsection Darwin Options
11429 @cindex Darwin options
11431 These options are defined for all architectures running the Darwin operating
11434 FSF GCC on Darwin does not create ``fat'' object files; it will create
11435 an object file for the single architecture that it was built to
11436 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11437 @option{-arch} options are used; it does so by running the compiler or
11438 linker multiple times and joining the results together with
11441 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11442 @samp{i686}) is determined by the flags that specify the ISA
11443 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11444 @option{-force_cpusubtype_ALL} option can be used to override this.
11446 The Darwin tools vary in their behavior when presented with an ISA
11447 mismatch. The assembler, @file{as}, will only permit instructions to
11448 be used that are valid for the subtype of the file it is generating,
11449 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11450 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11451 and print an error if asked to create a shared library with a less
11452 restrictive subtype than its input files (for instance, trying to put
11453 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11454 for executables, @file{ld}, will quietly give the executable the most
11455 restrictive subtype of any of its input files.
11460 Add the framework directory @var{dir} to the head of the list of
11461 directories to be searched for header files. These directories are
11462 interleaved with those specified by @option{-I} options and are
11463 scanned in a left-to-right order.
11465 A framework directory is a directory with frameworks in it. A
11466 framework is a directory with a @samp{"Headers"} and/or
11467 @samp{"PrivateHeaders"} directory contained directly in it that ends
11468 in @samp{".framework"}. The name of a framework is the name of this
11469 directory excluding the @samp{".framework"}. Headers associated with
11470 the framework are found in one of those two directories, with
11471 @samp{"Headers"} being searched first. A subframework is a framework
11472 directory that is in a framework's @samp{"Frameworks"} directory.
11473 Includes of subframework headers can only appear in a header of a
11474 framework that contains the subframework, or in a sibling subframework
11475 header. Two subframeworks are siblings if they occur in the same
11476 framework. A subframework should not have the same name as a
11477 framework, a warning will be issued if this is violated. Currently a
11478 subframework cannot have subframeworks, in the future, the mechanism
11479 may be extended to support this. The standard frameworks can be found
11480 in @samp{"/System/Library/Frameworks"} and
11481 @samp{"/Library/Frameworks"}. An example include looks like
11482 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11483 the name of the framework and header.h is found in the
11484 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11486 @item -iframework@var{dir}
11487 @opindex iframework
11488 Like @option{-F} except the directory is a treated as a system
11489 directory. The main difference between this @option{-iframework} and
11490 @option{-F} is that with @option{-iframework} the compiler does not
11491 warn about constructs contained within header files found via
11492 @var{dir}. This option is valid only for the C family of languages.
11496 Emit debugging information for symbols that are used. For STABS
11497 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11498 This is by default ON@.
11502 Emit debugging information for all symbols and types.
11504 @item -mmacosx-version-min=@var{version}
11505 The earliest version of MacOS X that this executable will run on
11506 is @var{version}. Typical values of @var{version} include @code{10.1},
11507 @code{10.2}, and @code{10.3.9}.
11509 If the compiler was built to use the system's headers by default,
11510 then the default for this option is the system version on which the
11511 compiler is running, otherwise the default is to make choices which
11512 are compatible with as many systems and code bases as possible.
11516 Enable kernel development mode. The @option{-mkernel} option sets
11517 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11518 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11519 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11520 applicable. This mode also sets @option{-mno-altivec},
11521 @option{-msoft-float}, @option{-fno-builtin} and
11522 @option{-mlong-branch} for PowerPC targets.
11524 @item -mone-byte-bool
11525 @opindex mone-byte-bool
11526 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11527 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11528 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11529 option has no effect on x86.
11531 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11532 to generate code that is not binary compatible with code generated
11533 without that switch. Using this switch may require recompiling all
11534 other modules in a program, including system libraries. Use this
11535 switch to conform to a non-default data model.
11537 @item -mfix-and-continue
11538 @itemx -ffix-and-continue
11539 @itemx -findirect-data
11540 @opindex mfix-and-continue
11541 @opindex ffix-and-continue
11542 @opindex findirect-data
11543 Generate code suitable for fast turn around development. Needed to
11544 enable gdb to dynamically load @code{.o} files into already running
11545 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11546 are provided for backwards compatibility.
11550 Loads all members of static archive libraries.
11551 See man ld(1) for more information.
11553 @item -arch_errors_fatal
11554 @opindex arch_errors_fatal
11555 Cause the errors having to do with files that have the wrong architecture
11558 @item -bind_at_load
11559 @opindex bind_at_load
11560 Causes the output file to be marked such that the dynamic linker will
11561 bind all undefined references when the file is loaded or launched.
11565 Produce a Mach-o bundle format file.
11566 See man ld(1) for more information.
11568 @item -bundle_loader @var{executable}
11569 @opindex bundle_loader
11570 This option specifies the @var{executable} that will be loading the build
11571 output file being linked. See man ld(1) for more information.
11574 @opindex dynamiclib
11575 When passed this option, GCC will produce a dynamic library instead of
11576 an executable when linking, using the Darwin @file{libtool} command.
11578 @item -force_cpusubtype_ALL
11579 @opindex force_cpusubtype_ALL
11580 This causes GCC's output file to have the @var{ALL} subtype, instead of
11581 one controlled by the @option{-mcpu} or @option{-march} option.
11583 @item -allowable_client @var{client_name}
11584 @itemx -client_name
11585 @itemx -compatibility_version
11586 @itemx -current_version
11588 @itemx -dependency-file
11590 @itemx -dylinker_install_name
11592 @itemx -exported_symbols_list
11595 @itemx -flat_namespace
11596 @itemx -force_flat_namespace
11597 @itemx -headerpad_max_install_names
11600 @itemx -install_name
11601 @itemx -keep_private_externs
11602 @itemx -multi_module
11603 @itemx -multiply_defined
11604 @itemx -multiply_defined_unused
11607 @itemx -no_dead_strip_inits_and_terms
11608 @itemx -nofixprebinding
11609 @itemx -nomultidefs
11611 @itemx -noseglinkedit
11612 @itemx -pagezero_size
11614 @itemx -prebind_all_twolevel_modules
11615 @itemx -private_bundle
11617 @itemx -read_only_relocs
11619 @itemx -sectobjectsymbols
11623 @itemx -sectobjectsymbols
11626 @itemx -segs_read_only_addr
11628 @itemx -segs_read_write_addr
11629 @itemx -seg_addr_table
11630 @itemx -seg_addr_table_filename
11631 @itemx -seglinkedit
11633 @itemx -segs_read_only_addr
11634 @itemx -segs_read_write_addr
11635 @itemx -single_module
11637 @itemx -sub_library
11639 @itemx -sub_umbrella
11640 @itemx -twolevel_namespace
11643 @itemx -unexported_symbols_list
11644 @itemx -weak_reference_mismatches
11645 @itemx -whatsloaded
11646 @opindex allowable_client
11647 @opindex client_name
11648 @opindex compatibility_version
11649 @opindex current_version
11650 @opindex dead_strip
11651 @opindex dependency-file
11652 @opindex dylib_file
11653 @opindex dylinker_install_name
11655 @opindex exported_symbols_list
11657 @opindex flat_namespace
11658 @opindex force_flat_namespace
11659 @opindex headerpad_max_install_names
11660 @opindex image_base
11662 @opindex install_name
11663 @opindex keep_private_externs
11664 @opindex multi_module
11665 @opindex multiply_defined
11666 @opindex multiply_defined_unused
11667 @opindex noall_load
11668 @opindex no_dead_strip_inits_and_terms
11669 @opindex nofixprebinding
11670 @opindex nomultidefs
11672 @opindex noseglinkedit
11673 @opindex pagezero_size
11675 @opindex prebind_all_twolevel_modules
11676 @opindex private_bundle
11677 @opindex read_only_relocs
11679 @opindex sectobjectsymbols
11682 @opindex sectcreate
11683 @opindex sectobjectsymbols
11686 @opindex segs_read_only_addr
11687 @opindex segs_read_write_addr
11688 @opindex seg_addr_table
11689 @opindex seg_addr_table_filename
11690 @opindex seglinkedit
11692 @opindex segs_read_only_addr
11693 @opindex segs_read_write_addr
11694 @opindex single_module
11696 @opindex sub_library
11697 @opindex sub_umbrella
11698 @opindex twolevel_namespace
11701 @opindex unexported_symbols_list
11702 @opindex weak_reference_mismatches
11703 @opindex whatsloaded
11704 These options are passed to the Darwin linker. The Darwin linker man page
11705 describes them in detail.
11708 @node DEC Alpha Options
11709 @subsection DEC Alpha Options
11711 These @samp{-m} options are defined for the DEC Alpha implementations:
11714 @item -mno-soft-float
11715 @itemx -msoft-float
11716 @opindex mno-soft-float
11717 @opindex msoft-float
11718 Use (do not use) the hardware floating-point instructions for
11719 floating-point operations. When @option{-msoft-float} is specified,
11720 functions in @file{libgcc.a} will be used to perform floating-point
11721 operations. Unless they are replaced by routines that emulate the
11722 floating-point operations, or compiled in such a way as to call such
11723 emulations routines, these routines will issue floating-point
11724 operations. If you are compiling for an Alpha without floating-point
11725 operations, you must ensure that the library is built so as not to call
11728 Note that Alpha implementations without floating-point operations are
11729 required to have floating-point registers.
11732 @itemx -mno-fp-regs
11734 @opindex mno-fp-regs
11735 Generate code that uses (does not use) the floating-point register set.
11736 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11737 register set is not used, floating point operands are passed in integer
11738 registers as if they were integers and floating-point results are passed
11739 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11740 so any function with a floating-point argument or return value called by code
11741 compiled with @option{-mno-fp-regs} must also be compiled with that
11744 A typical use of this option is building a kernel that does not use,
11745 and hence need not save and restore, any floating-point registers.
11749 The Alpha architecture implements floating-point hardware optimized for
11750 maximum performance. It is mostly compliant with the IEEE floating
11751 point standard. However, for full compliance, software assistance is
11752 required. This option generates code fully IEEE compliant code
11753 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11754 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11755 defined during compilation. The resulting code is less efficient but is
11756 able to correctly support denormalized numbers and exceptional IEEE
11757 values such as not-a-number and plus/minus infinity. Other Alpha
11758 compilers call this option @option{-ieee_with_no_inexact}.
11760 @item -mieee-with-inexact
11761 @opindex mieee-with-inexact
11762 This is like @option{-mieee} except the generated code also maintains
11763 the IEEE @var{inexact-flag}. Turning on this option causes the
11764 generated code to implement fully-compliant IEEE math. In addition to
11765 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11766 macro. On some Alpha implementations the resulting code may execute
11767 significantly slower than the code generated by default. Since there is
11768 very little code that depends on the @var{inexact-flag}, you should
11769 normally not specify this option. Other Alpha compilers call this
11770 option @option{-ieee_with_inexact}.
11772 @item -mfp-trap-mode=@var{trap-mode}
11773 @opindex mfp-trap-mode
11774 This option controls what floating-point related traps are enabled.
11775 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11776 The trap mode can be set to one of four values:
11780 This is the default (normal) setting. The only traps that are enabled
11781 are the ones that cannot be disabled in software (e.g., division by zero
11785 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11789 Like @samp{u}, but the instructions are marked to be safe for software
11790 completion (see Alpha architecture manual for details).
11793 Like @samp{su}, but inexact traps are enabled as well.
11796 @item -mfp-rounding-mode=@var{rounding-mode}
11797 @opindex mfp-rounding-mode
11798 Selects the IEEE rounding mode. Other Alpha compilers call this option
11799 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11804 Normal IEEE rounding mode. Floating point numbers are rounded towards
11805 the nearest machine number or towards the even machine number in case
11809 Round towards minus infinity.
11812 Chopped rounding mode. Floating point numbers are rounded towards zero.
11815 Dynamic rounding mode. A field in the floating point control register
11816 (@var{fpcr}, see Alpha architecture reference manual) controls the
11817 rounding mode in effect. The C library initializes this register for
11818 rounding towards plus infinity. Thus, unless your program modifies the
11819 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11822 @item -mtrap-precision=@var{trap-precision}
11823 @opindex mtrap-precision
11824 In the Alpha architecture, floating point traps are imprecise. This
11825 means without software assistance it is impossible to recover from a
11826 floating trap and program execution normally needs to be terminated.
11827 GCC can generate code that can assist operating system trap handlers
11828 in determining the exact location that caused a floating point trap.
11829 Depending on the requirements of an application, different levels of
11830 precisions can be selected:
11834 Program precision. This option is the default and means a trap handler
11835 can only identify which program caused a floating point exception.
11838 Function precision. The trap handler can determine the function that
11839 caused a floating point exception.
11842 Instruction precision. The trap handler can determine the exact
11843 instruction that caused a floating point exception.
11846 Other Alpha compilers provide the equivalent options called
11847 @option{-scope_safe} and @option{-resumption_safe}.
11849 @item -mieee-conformant
11850 @opindex mieee-conformant
11851 This option marks the generated code as IEEE conformant. You must not
11852 use this option unless you also specify @option{-mtrap-precision=i} and either
11853 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11854 is to emit the line @samp{.eflag 48} in the function prologue of the
11855 generated assembly file. Under DEC Unix, this has the effect that
11856 IEEE-conformant math library routines will be linked in.
11858 @item -mbuild-constants
11859 @opindex mbuild-constants
11860 Normally GCC examines a 32- or 64-bit integer constant to
11861 see if it can construct it from smaller constants in two or three
11862 instructions. If it cannot, it will output the constant as a literal and
11863 generate code to load it from the data segment at runtime.
11865 Use this option to require GCC to construct @emph{all} integer constants
11866 using code, even if it takes more instructions (the maximum is six).
11868 You would typically use this option to build a shared library dynamic
11869 loader. Itself a shared library, it must relocate itself in memory
11870 before it can find the variables and constants in its own data segment.
11876 Select whether to generate code to be assembled by the vendor-supplied
11877 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11895 Indicate whether GCC should generate code to use the optional BWX,
11896 CIX, FIX and MAX instruction sets. The default is to use the instruction
11897 sets supported by the CPU type specified via @option{-mcpu=} option or that
11898 of the CPU on which GCC was built if none was specified.
11901 @itemx -mfloat-ieee
11902 @opindex mfloat-vax
11903 @opindex mfloat-ieee
11904 Generate code that uses (does not use) VAX F and G floating point
11905 arithmetic instead of IEEE single and double precision.
11907 @item -mexplicit-relocs
11908 @itemx -mno-explicit-relocs
11909 @opindex mexplicit-relocs
11910 @opindex mno-explicit-relocs
11911 Older Alpha assemblers provided no way to generate symbol relocations
11912 except via assembler macros. Use of these macros does not allow
11913 optimal instruction scheduling. GNU binutils as of version 2.12
11914 supports a new syntax that allows the compiler to explicitly mark
11915 which relocations should apply to which instructions. This option
11916 is mostly useful for debugging, as GCC detects the capabilities of
11917 the assembler when it is built and sets the default accordingly.
11920 @itemx -mlarge-data
11921 @opindex msmall-data
11922 @opindex mlarge-data
11923 When @option{-mexplicit-relocs} is in effect, static data is
11924 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11925 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11926 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11927 16-bit relocations off of the @code{$gp} register. This limits the
11928 size of the small data area to 64KB, but allows the variables to be
11929 directly accessed via a single instruction.
11931 The default is @option{-mlarge-data}. With this option the data area
11932 is limited to just below 2GB@. Programs that require more than 2GB of
11933 data must use @code{malloc} or @code{mmap} to allocate the data in the
11934 heap instead of in the program's data segment.
11936 When generating code for shared libraries, @option{-fpic} implies
11937 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11940 @itemx -mlarge-text
11941 @opindex msmall-text
11942 @opindex mlarge-text
11943 When @option{-msmall-text} is used, the compiler assumes that the
11944 code of the entire program (or shared library) fits in 4MB, and is
11945 thus reachable with a branch instruction. When @option{-msmall-data}
11946 is used, the compiler can assume that all local symbols share the
11947 same @code{$gp} value, and thus reduce the number of instructions
11948 required for a function call from 4 to 1.
11950 The default is @option{-mlarge-text}.
11952 @item -mcpu=@var{cpu_type}
11954 Set the instruction set and instruction scheduling parameters for
11955 machine type @var{cpu_type}. You can specify either the @samp{EV}
11956 style name or the corresponding chip number. GCC supports scheduling
11957 parameters for the EV4, EV5 and EV6 family of processors and will
11958 choose the default values for the instruction set from the processor
11959 you specify. If you do not specify a processor type, GCC will default
11960 to the processor on which the compiler was built.
11962 Supported values for @var{cpu_type} are
11968 Schedules as an EV4 and has no instruction set extensions.
11972 Schedules as an EV5 and has no instruction set extensions.
11976 Schedules as an EV5 and supports the BWX extension.
11981 Schedules as an EV5 and supports the BWX and MAX extensions.
11985 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11989 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11992 Native toolchains also support the value @samp{native},
11993 which selects the best architecture option for the host processor.
11994 @option{-mcpu=native} has no effect if GCC does not recognize
11997 @item -mtune=@var{cpu_type}
11999 Set only the instruction scheduling parameters for machine type
12000 @var{cpu_type}. The instruction set is not changed.
12002 Native toolchains also support the value @samp{native},
12003 which selects the best architecture option for the host processor.
12004 @option{-mtune=native} has no effect if GCC does not recognize
12007 @item -mmemory-latency=@var{time}
12008 @opindex mmemory-latency
12009 Sets the latency the scheduler should assume for typical memory
12010 references as seen by the application. This number is highly
12011 dependent on the memory access patterns used by the application
12012 and the size of the external cache on the machine.
12014 Valid options for @var{time} are
12018 A decimal number representing clock cycles.
12024 The compiler contains estimates of the number of clock cycles for
12025 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
12026 (also called Dcache, Scache, and Bcache), as well as to main memory.
12027 Note that L3 is only valid for EV5.
12032 @node DEC Alpha/VMS Options
12033 @subsection DEC Alpha/VMS Options
12035 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
12038 @item -mvms-return-codes
12039 @opindex mvms-return-codes
12040 Return VMS condition codes from main. The default is to return POSIX
12041 style condition (e.g.@: error) codes.
12043 @item -mdebug-main=@var{prefix}
12044 @opindex mdebug-main=@var{prefix}
12045 Flag the first routine whose name starts with @var{prefix} as the main
12046 routine for the debugger.
12050 Default to 64bit memory allocation routines.
12054 @subsection FR30 Options
12055 @cindex FR30 Options
12057 These options are defined specifically for the FR30 port.
12061 @item -msmall-model
12062 @opindex msmall-model
12063 Use the small address space model. This can produce smaller code, but
12064 it does assume that all symbolic values and addresses will fit into a
12069 Assume that run-time support has been provided and so there is no need
12070 to include the simulator library (@file{libsim.a}) on the linker
12076 @subsection FRV Options
12077 @cindex FRV Options
12083 Only use the first 32 general purpose registers.
12088 Use all 64 general purpose registers.
12093 Use only the first 32 floating point registers.
12098 Use all 64 floating point registers
12101 @opindex mhard-float
12103 Use hardware instructions for floating point operations.
12106 @opindex msoft-float
12108 Use library routines for floating point operations.
12113 Dynamically allocate condition code registers.
12118 Do not try to dynamically allocate condition code registers, only
12119 use @code{icc0} and @code{fcc0}.
12124 Change ABI to use double word insns.
12129 Do not use double word instructions.
12134 Use floating point double instructions.
12137 @opindex mno-double
12139 Do not use floating point double instructions.
12144 Use media instructions.
12149 Do not use media instructions.
12154 Use multiply and add/subtract instructions.
12157 @opindex mno-muladd
12159 Do not use multiply and add/subtract instructions.
12164 Select the FDPIC ABI, that uses function descriptors to represent
12165 pointers to functions. Without any PIC/PIE-related options, it
12166 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
12167 assumes GOT entries and small data are within a 12-bit range from the
12168 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
12169 are computed with 32 bits.
12170 With a @samp{bfin-elf} target, this option implies @option{-msim}.
12173 @opindex minline-plt
12175 Enable inlining of PLT entries in function calls to functions that are
12176 not known to bind locally. It has no effect without @option{-mfdpic}.
12177 It's enabled by default if optimizing for speed and compiling for
12178 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
12179 optimization option such as @option{-O3} or above is present in the
12185 Assume a large TLS segment when generating thread-local code.
12190 Do not assume a large TLS segment when generating thread-local code.
12195 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
12196 that is known to be in read-only sections. It's enabled by default,
12197 except for @option{-fpic} or @option{-fpie}: even though it may help
12198 make the global offset table smaller, it trades 1 instruction for 4.
12199 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
12200 one of which may be shared by multiple symbols, and it avoids the need
12201 for a GOT entry for the referenced symbol, so it's more likely to be a
12202 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
12204 @item -multilib-library-pic
12205 @opindex multilib-library-pic
12207 Link with the (library, not FD) pic libraries. It's implied by
12208 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
12209 @option{-fpic} without @option{-mfdpic}. You should never have to use
12213 @opindex mlinked-fp
12215 Follow the EABI requirement of always creating a frame pointer whenever
12216 a stack frame is allocated. This option is enabled by default and can
12217 be disabled with @option{-mno-linked-fp}.
12220 @opindex mlong-calls
12222 Use indirect addressing to call functions outside the current
12223 compilation unit. This allows the functions to be placed anywhere
12224 within the 32-bit address space.
12226 @item -malign-labels
12227 @opindex malign-labels
12229 Try to align labels to an 8-byte boundary by inserting nops into the
12230 previous packet. This option only has an effect when VLIW packing
12231 is enabled. It doesn't create new packets; it merely adds nops to
12234 @item -mlibrary-pic
12235 @opindex mlibrary-pic
12237 Generate position-independent EABI code.
12242 Use only the first four media accumulator registers.
12247 Use all eight media accumulator registers.
12252 Pack VLIW instructions.
12257 Do not pack VLIW instructions.
12260 @opindex mno-eflags
12262 Do not mark ABI switches in e_flags.
12265 @opindex mcond-move
12267 Enable the use of conditional-move instructions (default).
12269 This switch is mainly for debugging the compiler and will likely be removed
12270 in a future version.
12272 @item -mno-cond-move
12273 @opindex mno-cond-move
12275 Disable the use of conditional-move instructions.
12277 This switch is mainly for debugging the compiler and will likely be removed
12278 in a future version.
12283 Enable the use of conditional set instructions (default).
12285 This switch is mainly for debugging the compiler and will likely be removed
12286 in a future version.
12291 Disable the use of conditional set instructions.
12293 This switch is mainly for debugging the compiler and will likely be removed
12294 in a future version.
12297 @opindex mcond-exec
12299 Enable the use of conditional execution (default).
12301 This switch is mainly for debugging the compiler and will likely be removed
12302 in a future version.
12304 @item -mno-cond-exec
12305 @opindex mno-cond-exec
12307 Disable the use of conditional execution.
12309 This switch is mainly for debugging the compiler and will likely be removed
12310 in a future version.
12312 @item -mvliw-branch
12313 @opindex mvliw-branch
12315 Run a pass to pack branches into VLIW instructions (default).
12317 This switch is mainly for debugging the compiler and will likely be removed
12318 in a future version.
12320 @item -mno-vliw-branch
12321 @opindex mno-vliw-branch
12323 Do not run a pass to pack branches into VLIW instructions.
12325 This switch is mainly for debugging the compiler and will likely be removed
12326 in a future version.
12328 @item -mmulti-cond-exec
12329 @opindex mmulti-cond-exec
12331 Enable optimization of @code{&&} and @code{||} in conditional execution
12334 This switch is mainly for debugging the compiler and will likely be removed
12335 in a future version.
12337 @item -mno-multi-cond-exec
12338 @opindex mno-multi-cond-exec
12340 Disable optimization of @code{&&} and @code{||} in conditional execution.
12342 This switch is mainly for debugging the compiler and will likely be removed
12343 in a future version.
12345 @item -mnested-cond-exec
12346 @opindex mnested-cond-exec
12348 Enable nested conditional execution optimizations (default).
12350 This switch is mainly for debugging the compiler and will likely be removed
12351 in a future version.
12353 @item -mno-nested-cond-exec
12354 @opindex mno-nested-cond-exec
12356 Disable nested conditional execution optimizations.
12358 This switch is mainly for debugging the compiler and will likely be removed
12359 in a future version.
12361 @item -moptimize-membar
12362 @opindex moptimize-membar
12364 This switch removes redundant @code{membar} instructions from the
12365 compiler generated code. It is enabled by default.
12367 @item -mno-optimize-membar
12368 @opindex mno-optimize-membar
12370 This switch disables the automatic removal of redundant @code{membar}
12371 instructions from the generated code.
12373 @item -mtomcat-stats
12374 @opindex mtomcat-stats
12376 Cause gas to print out tomcat statistics.
12378 @item -mcpu=@var{cpu}
12381 Select the processor type for which to generate code. Possible values are
12382 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12383 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12387 @node GNU/Linux Options
12388 @subsection GNU/Linux Options
12390 These @samp{-m} options are defined for GNU/Linux targets:
12395 Use the GNU C library. This is the default except
12396 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12400 Use uClibc C library. This is the default on
12401 @samp{*-*-linux-*uclibc*} targets.
12405 Use Bionic C library. This is the default on
12406 @samp{*-*-linux-*android*} targets.
12410 Compile code compatible with Android platform. This is the default on
12411 @samp{*-*-linux-*android*} targets.
12413 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12414 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12415 this option makes the GCC driver pass Android-specific options to the linker.
12416 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12419 @item -tno-android-cc
12420 @opindex tno-android-cc
12421 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12422 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12423 @option{-fno-rtti} by default.
12425 @item -tno-android-ld
12426 @opindex tno-android-ld
12427 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12428 linking options to the linker.
12432 @node H8/300 Options
12433 @subsection H8/300 Options
12435 These @samp{-m} options are defined for the H8/300 implementations:
12440 Shorten some address references at link time, when possible; uses the
12441 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12442 ld, Using ld}, for a fuller description.
12446 Generate code for the H8/300H@.
12450 Generate code for the H8S@.
12454 Generate code for the H8S and H8/300H in the normal mode. This switch
12455 must be used either with @option{-mh} or @option{-ms}.
12459 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12463 Make @code{int} data 32 bits by default.
12466 @opindex malign-300
12467 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12468 The default for the H8/300H and H8S is to align longs and floats on 4
12470 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12471 This option has no effect on the H8/300.
12475 @subsection HPPA Options
12476 @cindex HPPA Options
12478 These @samp{-m} options are defined for the HPPA family of computers:
12481 @item -march=@var{architecture-type}
12483 Generate code for the specified architecture. The choices for
12484 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12485 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12486 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12487 architecture option for your machine. Code compiled for lower numbered
12488 architectures will run on higher numbered architectures, but not the
12491 @item -mpa-risc-1-0
12492 @itemx -mpa-risc-1-1
12493 @itemx -mpa-risc-2-0
12494 @opindex mpa-risc-1-0
12495 @opindex mpa-risc-1-1
12496 @opindex mpa-risc-2-0
12497 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12500 @opindex mbig-switch
12501 Generate code suitable for big switch tables. Use this option only if
12502 the assembler/linker complain about out of range branches within a switch
12505 @item -mjump-in-delay
12506 @opindex mjump-in-delay
12507 Fill delay slots of function calls with unconditional jump instructions
12508 by modifying the return pointer for the function call to be the target
12509 of the conditional jump.
12511 @item -mdisable-fpregs
12512 @opindex mdisable-fpregs
12513 Prevent floating point registers from being used in any manner. This is
12514 necessary for compiling kernels which perform lazy context switching of
12515 floating point registers. If you use this option and attempt to perform
12516 floating point operations, the compiler will abort.
12518 @item -mdisable-indexing
12519 @opindex mdisable-indexing
12520 Prevent the compiler from using indexing address modes. This avoids some
12521 rather obscure problems when compiling MIG generated code under MACH@.
12523 @item -mno-space-regs
12524 @opindex mno-space-regs
12525 Generate code that assumes the target has no space registers. This allows
12526 GCC to generate faster indirect calls and use unscaled index address modes.
12528 Such code is suitable for level 0 PA systems and kernels.
12530 @item -mfast-indirect-calls
12531 @opindex mfast-indirect-calls
12532 Generate code that assumes calls never cross space boundaries. This
12533 allows GCC to emit code which performs faster indirect calls.
12535 This option will not work in the presence of shared libraries or nested
12538 @item -mfixed-range=@var{register-range}
12539 @opindex mfixed-range
12540 Generate code treating the given register range as fixed registers.
12541 A fixed register is one that the register allocator can not use. This is
12542 useful when compiling kernel code. A register range is specified as
12543 two registers separated by a dash. Multiple register ranges can be
12544 specified separated by a comma.
12546 @item -mlong-load-store
12547 @opindex mlong-load-store
12548 Generate 3-instruction load and store sequences as sometimes required by
12549 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12552 @item -mportable-runtime
12553 @opindex mportable-runtime
12554 Use the portable calling conventions proposed by HP for ELF systems.
12558 Enable the use of assembler directives only GAS understands.
12560 @item -mschedule=@var{cpu-type}
12562 Schedule code according to the constraints for the machine type
12563 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12564 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12565 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12566 proper scheduling option for your machine. The default scheduling is
12570 @opindex mlinker-opt
12571 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12572 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12573 linkers in which they give bogus error messages when linking some programs.
12576 @opindex msoft-float
12577 Generate output containing library calls for floating point.
12578 @strong{Warning:} the requisite libraries are not available for all HPPA
12579 targets. Normally the facilities of the machine's usual C compiler are
12580 used, but this cannot be done directly in cross-compilation. You must make
12581 your own arrangements to provide suitable library functions for
12584 @option{-msoft-float} changes the calling convention in the output file;
12585 therefore, it is only useful if you compile @emph{all} of a program with
12586 this option. In particular, you need to compile @file{libgcc.a}, the
12587 library that comes with GCC, with @option{-msoft-float} in order for
12592 Generate the predefine, @code{_SIO}, for server IO@. The default is
12593 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12594 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12595 options are available under HP-UX and HI-UX@.
12599 Use GNU ld specific options. This passes @option{-shared} to ld when
12600 building a shared library. It is the default when GCC is configured,
12601 explicitly or implicitly, with the GNU linker. This option does not
12602 have any affect on which ld is called, it only changes what parameters
12603 are passed to that ld. The ld that is called is determined by the
12604 @option{--with-ld} configure option, GCC's program search path, and
12605 finally by the user's @env{PATH}. The linker used by GCC can be printed
12606 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12607 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12611 Use HP ld specific options. This passes @option{-b} to ld when building
12612 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12613 links. It is the default when GCC is configured, explicitly or
12614 implicitly, with the HP linker. This option does not have any affect on
12615 which ld is called, it only changes what parameters are passed to that
12616 ld. The ld that is called is determined by the @option{--with-ld}
12617 configure option, GCC's program search path, and finally by the user's
12618 @env{PATH}. The linker used by GCC can be printed using @samp{which
12619 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12620 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12623 @opindex mno-long-calls
12624 Generate code that uses long call sequences. This ensures that a call
12625 is always able to reach linker generated stubs. The default is to generate
12626 long calls only when the distance from the call site to the beginning
12627 of the function or translation unit, as the case may be, exceeds a
12628 predefined limit set by the branch type being used. The limits for
12629 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12630 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12633 Distances are measured from the beginning of functions when using the
12634 @option{-ffunction-sections} option, or when using the @option{-mgas}
12635 and @option{-mno-portable-runtime} options together under HP-UX with
12638 It is normally not desirable to use this option as it will degrade
12639 performance. However, it may be useful in large applications,
12640 particularly when partial linking is used to build the application.
12642 The types of long calls used depends on the capabilities of the
12643 assembler and linker, and the type of code being generated. The
12644 impact on systems that support long absolute calls, and long pic
12645 symbol-difference or pc-relative calls should be relatively small.
12646 However, an indirect call is used on 32-bit ELF systems in pic code
12647 and it is quite long.
12649 @item -munix=@var{unix-std}
12651 Generate compiler predefines and select a startfile for the specified
12652 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12653 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12654 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12655 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12656 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12659 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12660 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12661 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12662 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12663 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12664 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12666 It is @emph{important} to note that this option changes the interfaces
12667 for various library routines. It also affects the operational behavior
12668 of the C library. Thus, @emph{extreme} care is needed in using this
12671 Library code that is intended to operate with more than one UNIX
12672 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12673 as appropriate. Most GNU software doesn't provide this capability.
12677 Suppress the generation of link options to search libdld.sl when the
12678 @option{-static} option is specified on HP-UX 10 and later.
12682 The HP-UX implementation of setlocale in libc has a dependency on
12683 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12684 when the @option{-static} option is specified, special link options
12685 are needed to resolve this dependency.
12687 On HP-UX 10 and later, the GCC driver adds the necessary options to
12688 link with libdld.sl when the @option{-static} option is specified.
12689 This causes the resulting binary to be dynamic. On the 64-bit port,
12690 the linkers generate dynamic binaries by default in any case. The
12691 @option{-nolibdld} option can be used to prevent the GCC driver from
12692 adding these link options.
12696 Add support for multithreading with the @dfn{dce thread} library
12697 under HP-UX@. This option sets flags for both the preprocessor and
12701 @node i386 and x86-64 Options
12702 @subsection Intel 386 and AMD x86-64 Options
12703 @cindex i386 Options
12704 @cindex x86-64 Options
12705 @cindex Intel 386 Options
12706 @cindex AMD x86-64 Options
12708 These @samp{-m} options are defined for the i386 and x86-64 family of
12712 @item -mtune=@var{cpu-type}
12714 Tune to @var{cpu-type} everything applicable about the generated code, except
12715 for the ABI and the set of available instructions. The choices for
12716 @var{cpu-type} are:
12719 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12720 If you know the CPU on which your code will run, then you should use
12721 the corresponding @option{-mtune} option instead of
12722 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12723 of your application will have, then you should use this option.
12725 As new processors are deployed in the marketplace, the behavior of this
12726 option will change. Therefore, if you upgrade to a newer version of
12727 GCC, the code generated option will change to reflect the processors
12728 that were most common when that version of GCC was released.
12730 There is no @option{-march=generic} option because @option{-march}
12731 indicates the instruction set the compiler can use, and there is no
12732 generic instruction set applicable to all processors. In contrast,
12733 @option{-mtune} indicates the processor (or, in this case, collection of
12734 processors) for which the code is optimized.
12736 This selects the CPU to tune for at compilation time by determining
12737 the processor type of the compiling machine. Using @option{-mtune=native}
12738 will produce code optimized for the local machine under the constraints
12739 of the selected instruction set. Using @option{-march=native} will
12740 enable all instruction subsets supported by the local machine (hence
12741 the result might not run on different machines).
12743 Original Intel's i386 CPU@.
12745 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12746 @item i586, pentium
12747 Intel Pentium CPU with no MMX support.
12749 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12751 Intel PentiumPro CPU@.
12753 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12754 instruction set will be used, so the code will run on all i686 family chips.
12756 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12757 @item pentium3, pentium3m
12758 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12761 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12762 support. Used by Centrino notebooks.
12763 @item pentium4, pentium4m
12764 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12766 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12769 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12770 SSE2 and SSE3 instruction set support.
12772 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12773 instruction set support.
12775 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12776 and SSE4.2 instruction set support.
12778 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12779 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12781 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12782 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
12785 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12786 instruction set support.
12788 AMD K6 CPU with MMX instruction set support.
12790 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12791 @item athlon, athlon-tbird
12792 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12794 @item athlon-4, athlon-xp, athlon-mp
12795 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12796 instruction set support.
12797 @item k8, opteron, athlon64, athlon-fx
12798 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12799 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12800 @item k8-sse3, opteron-sse3, athlon64-sse3
12801 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12802 @item amdfam10, barcelona
12803 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12804 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12805 instruction set extensions.)
12807 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12810 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12811 instruction set support.
12813 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12814 implemented for this chip.)
12816 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12817 implemented for this chip.)
12819 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12822 While picking a specific @var{cpu-type} will schedule things appropriately
12823 for that particular chip, the compiler will not generate any code that
12824 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12827 @item -march=@var{cpu-type}
12829 Generate instructions for the machine type @var{cpu-type}. The choices
12830 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12831 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12833 @item -mcpu=@var{cpu-type}
12835 A deprecated synonym for @option{-mtune}.
12837 @item -mfpmath=@var{unit}
12839 Generate floating point arithmetics for selected unit @var{unit}. The choices
12840 for @var{unit} are:
12844 Use the standard 387 floating point coprocessor present majority of chips and
12845 emulated otherwise. Code compiled with this option will run almost everywhere.
12846 The temporary results are computed in 80bit precision instead of precision
12847 specified by the type resulting in slightly different results compared to most
12848 of other chips. See @option{-ffloat-store} for more detailed description.
12850 This is the default choice for i386 compiler.
12853 Use scalar floating point instructions present in the SSE instruction set.
12854 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12855 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12856 instruction set supports only single precision arithmetics, thus the double and
12857 extended precision arithmetics is still done using 387. Later version, present
12858 only in Pentium4 and the future AMD x86-64 chips supports double precision
12861 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12862 or @option{-msse2} switches to enable SSE extensions and make this option
12863 effective. For the x86-64 compiler, these extensions are enabled by default.
12865 The resulting code should be considerably faster in the majority of cases and avoid
12866 the numerical instability problems of 387 code, but may break some existing
12867 code that expects temporaries to be 80bit.
12869 This is the default choice for the x86-64 compiler.
12874 Attempt to utilize both instruction sets at once. This effectively double the
12875 amount of available registers and on chips with separate execution units for
12876 387 and SSE the execution resources too. Use this option with care, as it is
12877 still experimental, because the GCC register allocator does not model separate
12878 functional units well resulting in instable performance.
12881 @item -masm=@var{dialect}
12882 @opindex masm=@var{dialect}
12883 Output asm instructions using selected @var{dialect}. Supported
12884 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12885 not support @samp{intel}.
12888 @itemx -mno-ieee-fp
12890 @opindex mno-ieee-fp
12891 Control whether or not the compiler uses IEEE floating point
12892 comparisons. These handle correctly the case where the result of a
12893 comparison is unordered.
12896 @opindex msoft-float
12897 Generate output containing library calls for floating point.
12898 @strong{Warning:} the requisite libraries are not part of GCC@.
12899 Normally the facilities of the machine's usual C compiler are used, but
12900 this can't be done directly in cross-compilation. You must make your
12901 own arrangements to provide suitable library functions for
12904 On machines where a function returns floating point results in the 80387
12905 register stack, some floating point opcodes may be emitted even if
12906 @option{-msoft-float} is used.
12908 @item -mno-fp-ret-in-387
12909 @opindex mno-fp-ret-in-387
12910 Do not use the FPU registers for return values of functions.
12912 The usual calling convention has functions return values of types
12913 @code{float} and @code{double} in an FPU register, even if there
12914 is no FPU@. The idea is that the operating system should emulate
12917 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12918 in ordinary CPU registers instead.
12920 @item -mno-fancy-math-387
12921 @opindex mno-fancy-math-387
12922 Some 387 emulators do not support the @code{sin}, @code{cos} and
12923 @code{sqrt} instructions for the 387. Specify this option to avoid
12924 generating those instructions. This option is the default on FreeBSD,
12925 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12926 indicates that the target CPU will always have an FPU and so the
12927 instruction will not need emulation. As of revision 2.6.1, these
12928 instructions are not generated unless you also use the
12929 @option{-funsafe-math-optimizations} switch.
12931 @item -malign-double
12932 @itemx -mno-align-double
12933 @opindex malign-double
12934 @opindex mno-align-double
12935 Control whether GCC aligns @code{double}, @code{long double}, and
12936 @code{long long} variables on a two word boundary or a one word
12937 boundary. Aligning @code{double} variables on a two word boundary will
12938 produce code that runs somewhat faster on a @samp{Pentium} at the
12939 expense of more memory.
12941 On x86-64, @option{-malign-double} is enabled by default.
12943 @strong{Warning:} if you use the @option{-malign-double} switch,
12944 structures containing the above types will be aligned differently than
12945 the published application binary interface specifications for the 386
12946 and will not be binary compatible with structures in code compiled
12947 without that switch.
12949 @item -m96bit-long-double
12950 @itemx -m128bit-long-double
12951 @opindex m96bit-long-double
12952 @opindex m128bit-long-double
12953 These switches control the size of @code{long double} type. The i386
12954 application binary interface specifies the size to be 96 bits,
12955 so @option{-m96bit-long-double} is the default in 32 bit mode.
12957 Modern architectures (Pentium and newer) would prefer @code{long double}
12958 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12959 conforming to the ABI, this would not be possible. So specifying a
12960 @option{-m128bit-long-double} will align @code{long double}
12961 to a 16 byte boundary by padding the @code{long double} with an additional
12964 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12965 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12967 Notice that neither of these options enable any extra precision over the x87
12968 standard of 80 bits for a @code{long double}.
12970 @strong{Warning:} if you override the default value for your target ABI, the
12971 structures and arrays containing @code{long double} variables will change
12972 their size as well as function calling convention for function taking
12973 @code{long double} will be modified. Hence they will not be binary
12974 compatible with arrays or structures in code compiled without that switch.
12976 @item -mlarge-data-threshold=@var{number}
12977 @opindex mlarge-data-threshold=@var{number}
12978 When @option{-mcmodel=medium} is specified, the data greater than
12979 @var{threshold} are placed in large data section. This value must be the
12980 same across all object linked into the binary and defaults to 65535.
12984 Use a different function-calling convention, in which functions that
12985 take a fixed number of arguments return with the @code{ret} @var{num}
12986 instruction, which pops their arguments while returning. This saves one
12987 instruction in the caller since there is no need to pop the arguments
12990 You can specify that an individual function is called with this calling
12991 sequence with the function attribute @samp{stdcall}. You can also
12992 override the @option{-mrtd} option by using the function attribute
12993 @samp{cdecl}. @xref{Function Attributes}.
12995 @strong{Warning:} this calling convention is incompatible with the one
12996 normally used on Unix, so you cannot use it if you need to call
12997 libraries compiled with the Unix compiler.
12999 Also, you must provide function prototypes for all functions that
13000 take variable numbers of arguments (including @code{printf});
13001 otherwise incorrect code will be generated for calls to those
13004 In addition, seriously incorrect code will result if you call a
13005 function with too many arguments. (Normally, extra arguments are
13006 harmlessly ignored.)
13008 @item -mregparm=@var{num}
13010 Control how many registers are used to pass integer arguments. By
13011 default, no registers are used to pass arguments, and at most 3
13012 registers can be used. You can control this behavior for a specific
13013 function by using the function attribute @samp{regparm}.
13014 @xref{Function Attributes}.
13016 @strong{Warning:} if you use this switch, and
13017 @var{num} is nonzero, then you must build all modules with the same
13018 value, including any libraries. This includes the system libraries and
13022 @opindex msseregparm
13023 Use SSE register passing conventions for float and double arguments
13024 and return values. You can control this behavior for a specific
13025 function by using the function attribute @samp{sseregparm}.
13026 @xref{Function Attributes}.
13028 @strong{Warning:} if you use this switch then you must build all
13029 modules with the same value, including any libraries. This includes
13030 the system libraries and startup modules.
13032 @item -mvect8-ret-in-mem
13033 @opindex mvect8-ret-in-mem
13034 Return 8-byte vectors in memory instead of MMX registers. This is the
13035 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
13036 Studio compilers until version 12. Later compiler versions (starting
13037 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
13038 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
13039 you need to remain compatible with existing code produced by those
13040 previous compiler versions or older versions of GCC.
13049 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
13050 is specified, the significands of results of floating-point operations are
13051 rounded to 24 bits (single precision); @option{-mpc64} rounds the
13052 significands of results of floating-point operations to 53 bits (double
13053 precision) and @option{-mpc80} rounds the significands of results of
13054 floating-point operations to 64 bits (extended double precision), which is
13055 the default. When this option is used, floating-point operations in higher
13056 precisions are not available to the programmer without setting the FPU
13057 control word explicitly.
13059 Setting the rounding of floating-point operations to less than the default
13060 80 bits can speed some programs by 2% or more. Note that some mathematical
13061 libraries assume that extended precision (80 bit) floating-point operations
13062 are enabled by default; routines in such libraries could suffer significant
13063 loss of accuracy, typically through so-called "catastrophic cancellation",
13064 when this option is used to set the precision to less than extended precision.
13066 @item -mstackrealign
13067 @opindex mstackrealign
13068 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
13069 option will generate an alternate prologue and epilogue that realigns the
13070 runtime stack if necessary. This supports mixing legacy codes that keep
13071 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
13072 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
13073 applicable to individual functions.
13075 @item -mpreferred-stack-boundary=@var{num}
13076 @opindex mpreferred-stack-boundary
13077 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
13078 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
13079 the default is 4 (16 bytes or 128 bits).
13081 @item -mincoming-stack-boundary=@var{num}
13082 @opindex mincoming-stack-boundary
13083 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
13084 boundary. If @option{-mincoming-stack-boundary} is not specified,
13085 the one specified by @option{-mpreferred-stack-boundary} will be used.
13087 On Pentium and PentiumPro, @code{double} and @code{long double} values
13088 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
13089 suffer significant run time performance penalties. On Pentium III, the
13090 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
13091 properly if it is not 16 byte aligned.
13093 To ensure proper alignment of this values on the stack, the stack boundary
13094 must be as aligned as that required by any value stored on the stack.
13095 Further, every function must be generated such that it keeps the stack
13096 aligned. Thus calling a function compiled with a higher preferred
13097 stack boundary from a function compiled with a lower preferred stack
13098 boundary will most likely misalign the stack. It is recommended that
13099 libraries that use callbacks always use the default setting.
13101 This extra alignment does consume extra stack space, and generally
13102 increases code size. Code that is sensitive to stack space usage, such
13103 as embedded systems and operating system kernels, may want to reduce the
13104 preferred alignment to @option{-mpreferred-stack-boundary=2}.
13133 @itemx -mno-fsgsbase
13169 These switches enable or disable the use of instructions in the MMX, SSE,
13170 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
13171 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
13172 @: extended instruction sets.
13173 These extensions are also available as built-in functions: see
13174 @ref{X86 Built-in Functions}, for details of the functions enabled and
13175 disabled by these switches.
13177 To have SSE/SSE2 instructions generated automatically from floating-point
13178 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
13180 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
13181 generates new AVX instructions or AVX equivalence for all SSEx instructions
13184 These options will enable GCC to use these extended instructions in
13185 generated code, even without @option{-mfpmath=sse}. Applications which
13186 perform runtime CPU detection must compile separate files for each
13187 supported architecture, using the appropriate flags. In particular,
13188 the file containing the CPU detection code should be compiled without
13193 This option instructs GCC to emit a @code{cld} instruction in the prologue
13194 of functions that use string instructions. String instructions depend on
13195 the DF flag to select between autoincrement or autodecrement mode. While the
13196 ABI specifies the DF flag to be cleared on function entry, some operating
13197 systems violate this specification by not clearing the DF flag in their
13198 exception dispatchers. The exception handler can be invoked with the DF flag
13199 set which leads to wrong direction mode, when string instructions are used.
13200 This option can be enabled by default on 32-bit x86 targets by configuring
13201 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
13202 instructions can be suppressed with the @option{-mno-cld} compiler option
13206 @opindex mvzeroupper
13207 This option instructs GCC to emit a @code{vzeroupper} instruction
13208 before a transfer of control flow out of the function to minimize
13209 AVX to SSE transition penalty as well as remove unnecessary zeroupper
13214 This option will enable GCC to use CMPXCHG16B instruction in generated code.
13215 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
13216 data types. This is useful for high resolution counters that could be updated
13217 by multiple processors (or cores). This instruction is generated as part of
13218 atomic built-in functions: see @ref{__sync Builtins} or
13219 @ref{__atomic Builtins} for details.
13223 This option will enable GCC to use SAHF instruction in generated 64-bit code.
13224 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
13225 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
13226 SAHF are load and store instructions, respectively, for certain status flags.
13227 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
13228 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
13232 This option will enable GCC to use movbe instruction to implement
13233 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
13237 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
13238 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
13239 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
13243 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
13244 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
13245 to increase precision instead of DIVSS and SQRTSS (and their vectorized
13246 variants) for single precision floating point arguments. These instructions
13247 are generated only when @option{-funsafe-math-optimizations} is enabled
13248 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
13249 Note that while the throughput of the sequence is higher than the throughput
13250 of the non-reciprocal instruction, the precision of the sequence can be
13251 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
13253 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS
13254 (or RSQRTPS) already with @option{-ffast-math} (or the above option
13255 combination), and doesn't need @option{-mrecip}.
13257 Also note that GCC emits the above sequence with additional Newton-Raphson step
13258 for vectorized single float division and vectorized @code{sqrtf(@var{x})}
13259 already with @option{-ffast-math} (or the above option combination), and
13260 doesn't need @option{-mrecip}.
13262 @item -mrecip=@var{opt}
13263 @opindex mrecip=opt
13264 This option allows to control which reciprocal estimate instructions
13265 may be used. @var{opt} is a comma separated list of options, that may
13266 be preceded by a @code{!} to invert the option:
13267 @code{all}: enable all estimate instructions,
13268 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
13269 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip},
13270 @code{div}: enable the approximation for scalar division,
13271 @code{vec-div}: enable the approximation for vectorized division,
13272 @code{sqrt}: enable the approximation for scalar square root,
13273 @code{vec-sqrt}: enable the approximation for vectorized square root.
13275 So for example, @option{-mrecip=all,!sqrt} would enable
13276 all of the reciprocal approximations, except for square root.
13278 @item -mveclibabi=@var{type}
13279 @opindex mveclibabi
13280 Specifies the ABI type to use for vectorizing intrinsics using an
13281 external library. Supported types are @code{svml} for the Intel short
13282 vector math library and @code{acml} for the AMD math core library style
13283 of interfacing. GCC will currently emit calls to @code{vmldExp2},
13284 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
13285 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
13286 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
13287 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
13288 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
13289 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
13290 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
13291 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
13292 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
13293 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
13294 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
13295 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
13296 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
13297 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
13298 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
13299 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
13300 compatible library will have to be specified at link time.
13302 @item -mabi=@var{name}
13304 Generate code for the specified calling convention. Permissible values
13305 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
13306 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
13307 ABI when targeting Windows. On all other systems, the default is the
13308 SYSV ABI. You can control this behavior for a specific function by
13309 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
13310 @xref{Function Attributes}.
13312 @item -mtls-dialect=@var{type}
13313 @opindex mtls-dialect
13314 Generate code to access thread-local storage using the @samp{gnu} or
13315 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
13316 @samp{gnu2} is more efficient, but it may add compile- and run-time
13317 requirements that cannot be satisfied on all systems.
13320 @itemx -mno-push-args
13321 @opindex mpush-args
13322 @opindex mno-push-args
13323 Use PUSH operations to store outgoing parameters. This method is shorter
13324 and usually equally fast as method using SUB/MOV operations and is enabled
13325 by default. In some cases disabling it may improve performance because of
13326 improved scheduling and reduced dependencies.
13328 @item -maccumulate-outgoing-args
13329 @opindex maccumulate-outgoing-args
13330 If enabled, the maximum amount of space required for outgoing arguments will be
13331 computed in the function prologue. This is faster on most modern CPUs
13332 because of reduced dependencies, improved scheduling and reduced stack usage
13333 when preferred stack boundary is not equal to 2. The drawback is a notable
13334 increase in code size. This switch implies @option{-mno-push-args}.
13338 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
13339 on thread-safe exception handling must compile and link all code with the
13340 @option{-mthreads} option. When compiling, @option{-mthreads} defines
13341 @option{-D_MT}; when linking, it links in a special thread helper library
13342 @option{-lmingwthrd} which cleans up per thread exception handling data.
13344 @item -mno-align-stringops
13345 @opindex mno-align-stringops
13346 Do not align destination of inlined string operations. This switch reduces
13347 code size and improves performance in case the destination is already aligned,
13348 but GCC doesn't know about it.
13350 @item -minline-all-stringops
13351 @opindex minline-all-stringops
13352 By default GCC inlines string operations only when destination is known to be
13353 aligned at least to 4 byte boundary. This enables more inlining, increase code
13354 size, but may improve performance of code that depends on fast memcpy, strlen
13355 and memset for short lengths.
13357 @item -minline-stringops-dynamically
13358 @opindex minline-stringops-dynamically
13359 For string operation of unknown size, inline runtime checks so for small
13360 blocks inline code is used, while for large blocks library call is used.
13362 @item -mstringop-strategy=@var{alg}
13363 @opindex mstringop-strategy=@var{alg}
13364 Overwrite internal decision heuristic about particular algorithm to inline
13365 string operation with. The allowed values are @code{rep_byte},
13366 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
13367 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
13368 expanding inline loop, @code{libcall} for always expanding library call.
13370 @item -momit-leaf-frame-pointer
13371 @opindex momit-leaf-frame-pointer
13372 Don't keep the frame pointer in a register for leaf functions. This
13373 avoids the instructions to save, set up and restore frame pointers and
13374 makes an extra register available in leaf functions. The option
13375 @option{-fomit-frame-pointer} removes the frame pointer for all functions
13376 which might make debugging harder.
13378 @item -mtls-direct-seg-refs
13379 @itemx -mno-tls-direct-seg-refs
13380 @opindex mtls-direct-seg-refs
13381 Controls whether TLS variables may be accessed with offsets from the
13382 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
13383 or whether the thread base pointer must be added. Whether or not this
13384 is legal depends on the operating system, and whether it maps the
13385 segment to cover the entire TLS area.
13387 For systems that use GNU libc, the default is on.
13390 @itemx -mno-sse2avx
13392 Specify that the assembler should encode SSE instructions with VEX
13393 prefix. The option @option{-mavx} turns this on by default.
13398 If profiling is active @option{-pg} put the profiling
13399 counter call before prologue.
13400 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13401 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13404 @itemx -mno-8bit-idiv
13406 On some processors, like Intel Atom, 8bit unsigned integer divide is
13407 much faster than 32bit/64bit integer divide. This option will generate a
13408 runt-time check. If both dividend and divisor are within range of 0
13409 to 255, 8bit unsigned integer divide will be used instead of
13410 32bit/64bit integer divide.
13412 @item -mavx256-split-unaligned-load
13413 @item -mavx256-split-unaligned-store
13414 @opindex avx256-split-unaligned-load
13415 @opindex avx256-split-unaligned-store
13416 Split 32-byte AVX unaligned load and store.
13420 These @samp{-m} switches are supported in addition to the above
13421 on AMD x86-64 processors in 64-bit environments.
13430 Generate code for a 32-bit or 64-bit environment.
13431 The @option{-m32} option sets int, long and pointer to 32 bits and
13432 generates code that runs on any i386 system.
13433 The @option{-m64} option sets int to 32 bits and long and pointer
13434 to 64 bits and generates code for AMD's x86-64 architecture.
13435 The @option{-mx32} option sets int, long and pointer to 32 bits and
13436 generates code for AMD's x86-64 architecture.
13437 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
13438 and @option{-mdynamic-no-pic} options.
13440 @item -mno-red-zone
13441 @opindex mno-red-zone
13442 Do not use a so called red zone for x86-64 code. The red zone is mandated
13443 by the x86-64 ABI, it is a 128-byte area beyond the location of the
13444 stack pointer that will not be modified by signal or interrupt handlers
13445 and therefore can be used for temporary data without adjusting the stack
13446 pointer. The flag @option{-mno-red-zone} disables this red zone.
13448 @item -mcmodel=small
13449 @opindex mcmodel=small
13450 Generate code for the small code model: the program and its symbols must
13451 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13452 Programs can be statically or dynamically linked. This is the default
13455 @item -mcmodel=kernel
13456 @opindex mcmodel=kernel
13457 Generate code for the kernel code model. The kernel runs in the
13458 negative 2 GB of the address space.
13459 This model has to be used for Linux kernel code.
13461 @item -mcmodel=medium
13462 @opindex mcmodel=medium
13463 Generate code for the medium model: The program is linked in the lower 2
13464 GB of the address space. Small symbols are also placed there. Symbols
13465 with sizes larger than @option{-mlarge-data-threshold} are put into
13466 large data or bss sections and can be located above 2GB. Programs can
13467 be statically or dynamically linked.
13469 @item -mcmodel=large
13470 @opindex mcmodel=large
13471 Generate code for the large model: This model makes no assumptions
13472 about addresses and sizes of sections.
13475 @node i386 and x86-64 Windows Options
13476 @subsection i386 and x86-64 Windows Options
13477 @cindex i386 and x86-64 Windows Options
13479 These additional options are available for Windows targets:
13484 This option is available for Cygwin and MinGW targets. It
13485 specifies that a console application is to be generated, by
13486 instructing the linker to set the PE header subsystem type
13487 required for console applications.
13488 This is the default behavior for Cygwin and MinGW targets.
13492 This option is available for Cygwin and MinGW targets. It
13493 specifies that a DLL - a dynamic link library - is to be
13494 generated, enabling the selection of the required runtime
13495 startup object and entry point.
13497 @item -mnop-fun-dllimport
13498 @opindex mnop-fun-dllimport
13499 This option is available for Cygwin and MinGW targets. It
13500 specifies that the dllimport attribute should be ignored.
13504 This option is available for MinGW targets. It specifies
13505 that MinGW-specific thread support is to be used.
13509 This option is available for mingw-w64 targets. It specifies
13510 that the UNICODE macro is getting pre-defined and that the
13511 unicode capable runtime startup code is chosen.
13515 This option is available for Cygwin and MinGW targets. It
13516 specifies that the typical Windows pre-defined macros are to
13517 be set in the pre-processor, but does not influence the choice
13518 of runtime library/startup code.
13522 This option is available for Cygwin and MinGW targets. It
13523 specifies that a GUI application is to be generated by
13524 instructing the linker to set the PE header subsystem type
13527 @item -fno-set-stack-executable
13528 @opindex fno-set-stack-executable
13529 This option is available for MinGW targets. It specifies that
13530 the executable flag for stack used by nested functions isn't
13531 set. This is necessary for binaries running in kernel mode of
13532 Windows, as there the user32 API, which is used to set executable
13533 privileges, isn't available.
13535 @item -mpe-aligned-commons
13536 @opindex mpe-aligned-commons
13537 This option is available for Cygwin and MinGW targets. It
13538 specifies that the GNU extension to the PE file format that
13539 permits the correct alignment of COMMON variables should be
13540 used when generating code. It will be enabled by default if
13541 GCC detects that the target assembler found during configuration
13542 supports the feature.
13545 See also under @ref{i386 and x86-64 Options} for standard options.
13547 @node IA-64 Options
13548 @subsection IA-64 Options
13549 @cindex IA-64 Options
13551 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13555 @opindex mbig-endian
13556 Generate code for a big endian target. This is the default for HP-UX@.
13558 @item -mlittle-endian
13559 @opindex mlittle-endian
13560 Generate code for a little endian target. This is the default for AIX5
13566 @opindex mno-gnu-as
13567 Generate (or don't) code for the GNU assembler. This is the default.
13568 @c Also, this is the default if the configure option @option{--with-gnu-as}
13574 @opindex mno-gnu-ld
13575 Generate (or don't) code for the GNU linker. This is the default.
13576 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13581 Generate code that does not use a global pointer register. The result
13582 is not position independent code, and violates the IA-64 ABI@.
13584 @item -mvolatile-asm-stop
13585 @itemx -mno-volatile-asm-stop
13586 @opindex mvolatile-asm-stop
13587 @opindex mno-volatile-asm-stop
13588 Generate (or don't) a stop bit immediately before and after volatile asm
13591 @item -mregister-names
13592 @itemx -mno-register-names
13593 @opindex mregister-names
13594 @opindex mno-register-names
13595 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13596 the stacked registers. This may make assembler output more readable.
13602 Disable (or enable) optimizations that use the small data section. This may
13603 be useful for working around optimizer bugs.
13605 @item -mconstant-gp
13606 @opindex mconstant-gp
13607 Generate code that uses a single constant global pointer value. This is
13608 useful when compiling kernel code.
13612 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13613 This is useful when compiling firmware code.
13615 @item -minline-float-divide-min-latency
13616 @opindex minline-float-divide-min-latency
13617 Generate code for inline divides of floating point values
13618 using the minimum latency algorithm.
13620 @item -minline-float-divide-max-throughput
13621 @opindex minline-float-divide-max-throughput
13622 Generate code for inline divides of floating point values
13623 using the maximum throughput algorithm.
13625 @item -mno-inline-float-divide
13626 @opindex mno-inline-float-divide
13627 Do not generate inline code for divides of floating point values.
13629 @item -minline-int-divide-min-latency
13630 @opindex minline-int-divide-min-latency
13631 Generate code for inline divides of integer values
13632 using the minimum latency algorithm.
13634 @item -minline-int-divide-max-throughput
13635 @opindex minline-int-divide-max-throughput
13636 Generate code for inline divides of integer values
13637 using the maximum throughput algorithm.
13639 @item -mno-inline-int-divide
13640 @opindex mno-inline-int-divide
13641 Do not generate inline code for divides of integer values.
13643 @item -minline-sqrt-min-latency
13644 @opindex minline-sqrt-min-latency
13645 Generate code for inline square roots
13646 using the minimum latency algorithm.
13648 @item -minline-sqrt-max-throughput
13649 @opindex minline-sqrt-max-throughput
13650 Generate code for inline square roots
13651 using the maximum throughput algorithm.
13653 @item -mno-inline-sqrt
13654 @opindex mno-inline-sqrt
13655 Do not generate inline code for sqrt.
13658 @itemx -mno-fused-madd
13659 @opindex mfused-madd
13660 @opindex mno-fused-madd
13661 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13662 instructions. The default is to use these instructions.
13664 @item -mno-dwarf2-asm
13665 @itemx -mdwarf2-asm
13666 @opindex mno-dwarf2-asm
13667 @opindex mdwarf2-asm
13668 Don't (or do) generate assembler code for the DWARF2 line number debugging
13669 info. This may be useful when not using the GNU assembler.
13671 @item -mearly-stop-bits
13672 @itemx -mno-early-stop-bits
13673 @opindex mearly-stop-bits
13674 @opindex mno-early-stop-bits
13675 Allow stop bits to be placed earlier than immediately preceding the
13676 instruction that triggered the stop bit. This can improve instruction
13677 scheduling, but does not always do so.
13679 @item -mfixed-range=@var{register-range}
13680 @opindex mfixed-range
13681 Generate code treating the given register range as fixed registers.
13682 A fixed register is one that the register allocator can not use. This is
13683 useful when compiling kernel code. A register range is specified as
13684 two registers separated by a dash. Multiple register ranges can be
13685 specified separated by a comma.
13687 @item -mtls-size=@var{tls-size}
13689 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13692 @item -mtune=@var{cpu-type}
13694 Tune the instruction scheduling for a particular CPU, Valid values are
13695 itanium, itanium1, merced, itanium2, and mckinley.
13701 Generate code for a 32-bit or 64-bit environment.
13702 The 32-bit environment sets int, long and pointer to 32 bits.
13703 The 64-bit environment sets int to 32 bits and long and pointer
13704 to 64 bits. These are HP-UX specific flags.
13706 @item -mno-sched-br-data-spec
13707 @itemx -msched-br-data-spec
13708 @opindex mno-sched-br-data-spec
13709 @opindex msched-br-data-spec
13710 (Dis/En)able data speculative scheduling before reload.
13711 This will result in generation of the ld.a instructions and
13712 the corresponding check instructions (ld.c / chk.a).
13713 The default is 'disable'.
13715 @item -msched-ar-data-spec
13716 @itemx -mno-sched-ar-data-spec
13717 @opindex msched-ar-data-spec
13718 @opindex mno-sched-ar-data-spec
13719 (En/Dis)able data speculative scheduling after reload.
13720 This will result in generation of the ld.a instructions and
13721 the corresponding check instructions (ld.c / chk.a).
13722 The default is 'enable'.
13724 @item -mno-sched-control-spec
13725 @itemx -msched-control-spec
13726 @opindex mno-sched-control-spec
13727 @opindex msched-control-spec
13728 (Dis/En)able control speculative scheduling. This feature is
13729 available only during region scheduling (i.e.@: before reload).
13730 This will result in generation of the ld.s instructions and
13731 the corresponding check instructions chk.s .
13732 The default is 'disable'.
13734 @item -msched-br-in-data-spec
13735 @itemx -mno-sched-br-in-data-spec
13736 @opindex msched-br-in-data-spec
13737 @opindex mno-sched-br-in-data-spec
13738 (En/Dis)able speculative scheduling of the instructions that
13739 are dependent on the data speculative loads before reload.
13740 This is effective only with @option{-msched-br-data-spec} enabled.
13741 The default is 'enable'.
13743 @item -msched-ar-in-data-spec
13744 @itemx -mno-sched-ar-in-data-spec
13745 @opindex msched-ar-in-data-spec
13746 @opindex mno-sched-ar-in-data-spec
13747 (En/Dis)able speculative scheduling of the instructions that
13748 are dependent on the data speculative loads after reload.
13749 This is effective only with @option{-msched-ar-data-spec} enabled.
13750 The default is 'enable'.
13752 @item -msched-in-control-spec
13753 @itemx -mno-sched-in-control-spec
13754 @opindex msched-in-control-spec
13755 @opindex mno-sched-in-control-spec
13756 (En/Dis)able speculative scheduling of the instructions that
13757 are dependent on the control speculative loads.
13758 This is effective only with @option{-msched-control-spec} enabled.
13759 The default is 'enable'.
13761 @item -mno-sched-prefer-non-data-spec-insns
13762 @itemx -msched-prefer-non-data-spec-insns
13763 @opindex mno-sched-prefer-non-data-spec-insns
13764 @opindex msched-prefer-non-data-spec-insns
13765 If enabled, data speculative instructions will be chosen for schedule
13766 only if there are no other choices at the moment. This will make
13767 the use of the data speculation much more conservative.
13768 The default is 'disable'.
13770 @item -mno-sched-prefer-non-control-spec-insns
13771 @itemx -msched-prefer-non-control-spec-insns
13772 @opindex mno-sched-prefer-non-control-spec-insns
13773 @opindex msched-prefer-non-control-spec-insns
13774 If enabled, control speculative instructions will be chosen for schedule
13775 only if there are no other choices at the moment. This will make
13776 the use of the control speculation much more conservative.
13777 The default is 'disable'.
13779 @item -mno-sched-count-spec-in-critical-path
13780 @itemx -msched-count-spec-in-critical-path
13781 @opindex mno-sched-count-spec-in-critical-path
13782 @opindex msched-count-spec-in-critical-path
13783 If enabled, speculative dependencies will be considered during
13784 computation of the instructions priorities. This will make the use of the
13785 speculation a bit more conservative.
13786 The default is 'disable'.
13788 @item -msched-spec-ldc
13789 @opindex msched-spec-ldc
13790 Use a simple data speculation check. This option is on by default.
13792 @item -msched-control-spec-ldc
13793 @opindex msched-spec-ldc
13794 Use a simple check for control speculation. This option is on by default.
13796 @item -msched-stop-bits-after-every-cycle
13797 @opindex msched-stop-bits-after-every-cycle
13798 Place a stop bit after every cycle when scheduling. This option is on
13801 @item -msched-fp-mem-deps-zero-cost
13802 @opindex msched-fp-mem-deps-zero-cost
13803 Assume that floating-point stores and loads are not likely to cause a conflict
13804 when placed into the same instruction group. This option is disabled by
13807 @item -msel-sched-dont-check-control-spec
13808 @opindex msel-sched-dont-check-control-spec
13809 Generate checks for control speculation in selective scheduling.
13810 This flag is disabled by default.
13812 @item -msched-max-memory-insns=@var{max-insns}
13813 @opindex msched-max-memory-insns
13814 Limit on the number of memory insns per instruction group, giving lower
13815 priority to subsequent memory insns attempting to schedule in the same
13816 instruction group. Frequently useful to prevent cache bank conflicts.
13817 The default value is 1.
13819 @item -msched-max-memory-insns-hard-limit
13820 @opindex msched-max-memory-insns-hard-limit
13821 Disallow more than `msched-max-memory-insns' in instruction group.
13822 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13823 when limit is reached but may still schedule memory operations.
13827 @node IA-64/VMS Options
13828 @subsection IA-64/VMS Options
13830 These @samp{-m} options are defined for the IA-64/VMS implementations:
13833 @item -mvms-return-codes
13834 @opindex mvms-return-codes
13835 Return VMS condition codes from main. The default is to return POSIX
13836 style condition (e.g.@ error) codes.
13838 @item -mdebug-main=@var{prefix}
13839 @opindex mdebug-main=@var{prefix}
13840 Flag the first routine whose name starts with @var{prefix} as the main
13841 routine for the debugger.
13845 Default to 64bit memory allocation routines.
13849 @subsection LM32 Options
13850 @cindex LM32 options
13852 These @option{-m} options are defined for the Lattice Mico32 architecture:
13855 @item -mbarrel-shift-enabled
13856 @opindex mbarrel-shift-enabled
13857 Enable barrel-shift instructions.
13859 @item -mdivide-enabled
13860 @opindex mdivide-enabled
13861 Enable divide and modulus instructions.
13863 @item -mmultiply-enabled
13864 @opindex multiply-enabled
13865 Enable multiply instructions.
13867 @item -msign-extend-enabled
13868 @opindex msign-extend-enabled
13869 Enable sign extend instructions.
13871 @item -muser-enabled
13872 @opindex muser-enabled
13873 Enable user-defined instructions.
13878 @subsection M32C Options
13879 @cindex M32C options
13882 @item -mcpu=@var{name}
13884 Select the CPU for which code is generated. @var{name} may be one of
13885 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13886 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13887 the M32C/80 series.
13891 Specifies that the program will be run on the simulator. This causes
13892 an alternate runtime library to be linked in which supports, for
13893 example, file I/O@. You must not use this option when generating
13894 programs that will run on real hardware; you must provide your own
13895 runtime library for whatever I/O functions are needed.
13897 @item -memregs=@var{number}
13899 Specifies the number of memory-based pseudo-registers GCC will use
13900 during code generation. These pseudo-registers will be used like real
13901 registers, so there is a tradeoff between GCC's ability to fit the
13902 code into available registers, and the performance penalty of using
13903 memory instead of registers. Note that all modules in a program must
13904 be compiled with the same value for this option. Because of that, you
13905 must not use this option with the default runtime libraries gcc
13910 @node M32R/D Options
13911 @subsection M32R/D Options
13912 @cindex M32R/D options
13914 These @option{-m} options are defined for Renesas M32R/D architectures:
13919 Generate code for the M32R/2@.
13923 Generate code for the M32R/X@.
13927 Generate code for the M32R@. This is the default.
13929 @item -mmodel=small
13930 @opindex mmodel=small
13931 Assume all objects live in the lower 16MB of memory (so that their addresses
13932 can be loaded with the @code{ld24} instruction), and assume all subroutines
13933 are reachable with the @code{bl} instruction.
13934 This is the default.
13936 The addressability of a particular object can be set with the
13937 @code{model} attribute.
13939 @item -mmodel=medium
13940 @opindex mmodel=medium
13941 Assume objects may be anywhere in the 32-bit address space (the compiler
13942 will generate @code{seth/add3} instructions to load their addresses), and
13943 assume all subroutines are reachable with the @code{bl} instruction.
13945 @item -mmodel=large
13946 @opindex mmodel=large
13947 Assume objects may be anywhere in the 32-bit address space (the compiler
13948 will generate @code{seth/add3} instructions to load their addresses), and
13949 assume subroutines may not be reachable with the @code{bl} instruction
13950 (the compiler will generate the much slower @code{seth/add3/jl}
13951 instruction sequence).
13954 @opindex msdata=none
13955 Disable use of the small data area. Variables will be put into
13956 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13957 @code{section} attribute has been specified).
13958 This is the default.
13960 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13961 Objects may be explicitly put in the small data area with the
13962 @code{section} attribute using one of these sections.
13964 @item -msdata=sdata
13965 @opindex msdata=sdata
13966 Put small global and static data in the small data area, but do not
13967 generate special code to reference them.
13970 @opindex msdata=use
13971 Put small global and static data in the small data area, and generate
13972 special instructions to reference them.
13976 @cindex smaller data references
13977 Put global and static objects less than or equal to @var{num} bytes
13978 into the small data or bss sections instead of the normal data or bss
13979 sections. The default value of @var{num} is 8.
13980 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13981 for this option to have any effect.
13983 All modules should be compiled with the same @option{-G @var{num}} value.
13984 Compiling with different values of @var{num} may or may not work; if it
13985 doesn't the linker will give an error message---incorrect code will not be
13990 Makes the M32R specific code in the compiler display some statistics
13991 that might help in debugging programs.
13993 @item -malign-loops
13994 @opindex malign-loops
13995 Align all loops to a 32-byte boundary.
13997 @item -mno-align-loops
13998 @opindex mno-align-loops
13999 Do not enforce a 32-byte alignment for loops. This is the default.
14001 @item -missue-rate=@var{number}
14002 @opindex missue-rate=@var{number}
14003 Issue @var{number} instructions per cycle. @var{number} can only be 1
14006 @item -mbranch-cost=@var{number}
14007 @opindex mbranch-cost=@var{number}
14008 @var{number} can only be 1 or 2. If it is 1 then branches will be
14009 preferred over conditional code, if it is 2, then the opposite will
14012 @item -mflush-trap=@var{number}
14013 @opindex mflush-trap=@var{number}
14014 Specifies the trap number to use to flush the cache. The default is
14015 12. Valid numbers are between 0 and 15 inclusive.
14017 @item -mno-flush-trap
14018 @opindex mno-flush-trap
14019 Specifies that the cache cannot be flushed by using a trap.
14021 @item -mflush-func=@var{name}
14022 @opindex mflush-func=@var{name}
14023 Specifies the name of the operating system function to call to flush
14024 the cache. The default is @emph{_flush_cache}, but a function call
14025 will only be used if a trap is not available.
14027 @item -mno-flush-func
14028 @opindex mno-flush-func
14029 Indicates that there is no OS function for flushing the cache.
14033 @node M680x0 Options
14034 @subsection M680x0 Options
14035 @cindex M680x0 options
14037 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
14038 The default settings depend on which architecture was selected when
14039 the compiler was configured; the defaults for the most common choices
14043 @item -march=@var{arch}
14045 Generate code for a specific M680x0 or ColdFire instruction set
14046 architecture. Permissible values of @var{arch} for M680x0
14047 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
14048 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
14049 architectures are selected according to Freescale's ISA classification
14050 and the permissible values are: @samp{isaa}, @samp{isaaplus},
14051 @samp{isab} and @samp{isac}.
14053 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
14054 code for a ColdFire target. The @var{arch} in this macro is one of the
14055 @option{-march} arguments given above.
14057 When used together, @option{-march} and @option{-mtune} select code
14058 that runs on a family of similar processors but that is optimized
14059 for a particular microarchitecture.
14061 @item -mcpu=@var{cpu}
14063 Generate code for a specific M680x0 or ColdFire processor.
14064 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
14065 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
14066 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
14067 below, which also classifies the CPUs into families:
14069 @multitable @columnfractions 0.20 0.80
14070 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
14071 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
14072 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
14073 @item @samp{5206e} @tab @samp{5206e}
14074 @item @samp{5208} @tab @samp{5207} @samp{5208}
14075 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
14076 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
14077 @item @samp{5216} @tab @samp{5214} @samp{5216}
14078 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
14079 @item @samp{5225} @tab @samp{5224} @samp{5225}
14080 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
14081 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
14082 @item @samp{5249} @tab @samp{5249}
14083 @item @samp{5250} @tab @samp{5250}
14084 @item @samp{5271} @tab @samp{5270} @samp{5271}
14085 @item @samp{5272} @tab @samp{5272}
14086 @item @samp{5275} @tab @samp{5274} @samp{5275}
14087 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
14088 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
14089 @item @samp{5307} @tab @samp{5307}
14090 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
14091 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
14092 @item @samp{5407} @tab @samp{5407}
14093 @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}
14096 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
14097 @var{arch} is compatible with @var{cpu}. Other combinations of
14098 @option{-mcpu} and @option{-march} are rejected.
14100 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
14101 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
14102 where the value of @var{family} is given by the table above.
14104 @item -mtune=@var{tune}
14106 Tune the code for a particular microarchitecture, within the
14107 constraints set by @option{-march} and @option{-mcpu}.
14108 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
14109 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
14110 and @samp{cpu32}. The ColdFire microarchitectures
14111 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
14113 You can also use @option{-mtune=68020-40} for code that needs
14114 to run relatively well on 68020, 68030 and 68040 targets.
14115 @option{-mtune=68020-60} is similar but includes 68060 targets
14116 as well. These two options select the same tuning decisions as
14117 @option{-m68020-40} and @option{-m68020-60} respectively.
14119 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
14120 when tuning for 680x0 architecture @var{arch}. It also defines
14121 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
14122 option is used. If gcc is tuning for a range of architectures,
14123 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
14124 it defines the macros for every architecture in the range.
14126 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
14127 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
14128 of the arguments given above.
14134 Generate output for a 68000. This is the default
14135 when the compiler is configured for 68000-based systems.
14136 It is equivalent to @option{-march=68000}.
14138 Use this option for microcontrollers with a 68000 or EC000 core,
14139 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
14143 Generate output for a 68010. This is the default
14144 when the compiler is configured for 68010-based systems.
14145 It is equivalent to @option{-march=68010}.
14151 Generate output for a 68020. This is the default
14152 when the compiler is configured for 68020-based systems.
14153 It is equivalent to @option{-march=68020}.
14157 Generate output for a 68030. This is the default when the compiler is
14158 configured for 68030-based systems. It is equivalent to
14159 @option{-march=68030}.
14163 Generate output for a 68040. This is the default when the compiler is
14164 configured for 68040-based systems. It is equivalent to
14165 @option{-march=68040}.
14167 This option inhibits the use of 68881/68882 instructions that have to be
14168 emulated by software on the 68040. Use this option if your 68040 does not
14169 have code to emulate those instructions.
14173 Generate output for a 68060. This is the default when the compiler is
14174 configured for 68060-based systems. It is equivalent to
14175 @option{-march=68060}.
14177 This option inhibits the use of 68020 and 68881/68882 instructions that
14178 have to be emulated by software on the 68060. Use this option if your 68060
14179 does not have code to emulate those instructions.
14183 Generate output for a CPU32. This is the default
14184 when the compiler is configured for CPU32-based systems.
14185 It is equivalent to @option{-march=cpu32}.
14187 Use this option for microcontrollers with a
14188 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
14189 68336, 68340, 68341, 68349 and 68360.
14193 Generate output for a 520X ColdFire CPU@. This is the default
14194 when the compiler is configured for 520X-based systems.
14195 It is equivalent to @option{-mcpu=5206}, and is now deprecated
14196 in favor of that option.
14198 Use this option for microcontroller with a 5200 core, including
14199 the MCF5202, MCF5203, MCF5204 and MCF5206.
14203 Generate output for a 5206e ColdFire CPU@. The option is now
14204 deprecated in favor of the equivalent @option{-mcpu=5206e}.
14208 Generate output for a member of the ColdFire 528X family.
14209 The option is now deprecated in favor of the equivalent
14210 @option{-mcpu=528x}.
14214 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
14215 in favor of the equivalent @option{-mcpu=5307}.
14219 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
14220 in favor of the equivalent @option{-mcpu=5407}.
14224 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
14225 This includes use of hardware floating point instructions.
14226 The option is equivalent to @option{-mcpu=547x}, and is now
14227 deprecated in favor of that option.
14231 Generate output for a 68040, without using any of the new instructions.
14232 This results in code which can run relatively efficiently on either a
14233 68020/68881 or a 68030 or a 68040. The generated code does use the
14234 68881 instructions that are emulated on the 68040.
14236 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
14240 Generate output for a 68060, without using any of the new instructions.
14241 This results in code which can run relatively efficiently on either a
14242 68020/68881 or a 68030 or a 68040. The generated code does use the
14243 68881 instructions that are emulated on the 68060.
14245 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
14249 @opindex mhard-float
14251 Generate floating-point instructions. This is the default for 68020
14252 and above, and for ColdFire devices that have an FPU@. It defines the
14253 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
14254 on ColdFire targets.
14257 @opindex msoft-float
14258 Do not generate floating-point instructions; use library calls instead.
14259 This is the default for 68000, 68010, and 68832 targets. It is also
14260 the default for ColdFire devices that have no FPU.
14266 Generate (do not generate) ColdFire hardware divide and remainder
14267 instructions. If @option{-march} is used without @option{-mcpu},
14268 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
14269 architectures. Otherwise, the default is taken from the target CPU
14270 (either the default CPU, or the one specified by @option{-mcpu}). For
14271 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
14272 @option{-mcpu=5206e}.
14274 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
14278 Consider type @code{int} to be 16 bits wide, like @code{short int}.
14279 Additionally, parameters passed on the stack are also aligned to a
14280 16-bit boundary even on targets whose API mandates promotion to 32-bit.
14284 Do not consider type @code{int} to be 16 bits wide. This is the default.
14287 @itemx -mno-bitfield
14288 @opindex mnobitfield
14289 @opindex mno-bitfield
14290 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
14291 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
14295 Do use the bit-field instructions. The @option{-m68020} option implies
14296 @option{-mbitfield}. This is the default if you use a configuration
14297 designed for a 68020.
14301 Use a different function-calling convention, in which functions
14302 that take a fixed number of arguments return with the @code{rtd}
14303 instruction, which pops their arguments while returning. This
14304 saves one instruction in the caller since there is no need to pop
14305 the arguments there.
14307 This calling convention is incompatible with the one normally
14308 used on Unix, so you cannot use it if you need to call libraries
14309 compiled with the Unix compiler.
14311 Also, you must provide function prototypes for all functions that
14312 take variable numbers of arguments (including @code{printf});
14313 otherwise incorrect code will be generated for calls to those
14316 In addition, seriously incorrect code will result if you call a
14317 function with too many arguments. (Normally, extra arguments are
14318 harmlessly ignored.)
14320 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
14321 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
14325 Do not use the calling conventions selected by @option{-mrtd}.
14326 This is the default.
14329 @itemx -mno-align-int
14330 @opindex malign-int
14331 @opindex mno-align-int
14332 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
14333 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
14334 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
14335 Aligning variables on 32-bit boundaries produces code that runs somewhat
14336 faster on processors with 32-bit busses at the expense of more memory.
14338 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
14339 align structures containing the above types differently than
14340 most published application binary interface specifications for the m68k.
14344 Use the pc-relative addressing mode of the 68000 directly, instead of
14345 using a global offset table. At present, this option implies @option{-fpic},
14346 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
14347 not presently supported with @option{-mpcrel}, though this could be supported for
14348 68020 and higher processors.
14350 @item -mno-strict-align
14351 @itemx -mstrict-align
14352 @opindex mno-strict-align
14353 @opindex mstrict-align
14354 Do not (do) assume that unaligned memory references will be handled by
14358 Generate code that allows the data segment to be located in a different
14359 area of memory from the text segment. This allows for execute in place in
14360 an environment without virtual memory management. This option implies
14363 @item -mno-sep-data
14364 Generate code that assumes that the data segment follows the text segment.
14365 This is the default.
14367 @item -mid-shared-library
14368 Generate code that supports shared libraries via the library ID method.
14369 This allows for execute in place and shared libraries in an environment
14370 without virtual memory management. This option implies @option{-fPIC}.
14372 @item -mno-id-shared-library
14373 Generate code that doesn't assume ID based shared libraries are being used.
14374 This is the default.
14376 @item -mshared-library-id=n
14377 Specified the identification number of the ID based shared library being
14378 compiled. Specifying a value of 0 will generate more compact code, specifying
14379 other values will force the allocation of that number to the current
14380 library but is no more space or time efficient than omitting this option.
14386 When generating position-independent code for ColdFire, generate code
14387 that works if the GOT has more than 8192 entries. This code is
14388 larger and slower than code generated without this option. On M680x0
14389 processors, this option is not needed; @option{-fPIC} suffices.
14391 GCC normally uses a single instruction to load values from the GOT@.
14392 While this is relatively efficient, it only works if the GOT
14393 is smaller than about 64k. Anything larger causes the linker
14394 to report an error such as:
14396 @cindex relocation truncated to fit (ColdFire)
14398 relocation truncated to fit: R_68K_GOT16O foobar
14401 If this happens, you should recompile your code with @option{-mxgot}.
14402 It should then work with very large GOTs. However, code generated with
14403 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14404 the value of a global symbol.
14406 Note that some linkers, including newer versions of the GNU linker,
14407 can create multiple GOTs and sort GOT entries. If you have such a linker,
14408 you should only need to use @option{-mxgot} when compiling a single
14409 object file that accesses more than 8192 GOT entries. Very few do.
14411 These options have no effect unless GCC is generating
14412 position-independent code.
14416 @node MCore Options
14417 @subsection MCore Options
14418 @cindex MCore options
14420 These are the @samp{-m} options defined for the Motorola M*Core
14426 @itemx -mno-hardlit
14428 @opindex mno-hardlit
14429 Inline constants into the code stream if it can be done in two
14430 instructions or less.
14436 Use the divide instruction. (Enabled by default).
14438 @item -mrelax-immediate
14439 @itemx -mno-relax-immediate
14440 @opindex mrelax-immediate
14441 @opindex mno-relax-immediate
14442 Allow arbitrary sized immediates in bit operations.
14444 @item -mwide-bitfields
14445 @itemx -mno-wide-bitfields
14446 @opindex mwide-bitfields
14447 @opindex mno-wide-bitfields
14448 Always treat bit-fields as int-sized.
14450 @item -m4byte-functions
14451 @itemx -mno-4byte-functions
14452 @opindex m4byte-functions
14453 @opindex mno-4byte-functions
14454 Force all functions to be aligned to a four byte boundary.
14456 @item -mcallgraph-data
14457 @itemx -mno-callgraph-data
14458 @opindex mcallgraph-data
14459 @opindex mno-callgraph-data
14460 Emit callgraph information.
14463 @itemx -mno-slow-bytes
14464 @opindex mslow-bytes
14465 @opindex mno-slow-bytes
14466 Prefer word access when reading byte quantities.
14468 @item -mlittle-endian
14469 @itemx -mbig-endian
14470 @opindex mlittle-endian
14471 @opindex mbig-endian
14472 Generate code for a little endian target.
14478 Generate code for the 210 processor.
14482 Assume that run-time support has been provided and so omit the
14483 simulator library (@file{libsim.a)} from the linker command line.
14485 @item -mstack-increment=@var{size}
14486 @opindex mstack-increment
14487 Set the maximum amount for a single stack increment operation. Large
14488 values can increase the speed of programs which contain functions
14489 that need a large amount of stack space, but they can also trigger a
14490 segmentation fault if the stack is extended too much. The default
14496 @subsection MeP Options
14497 @cindex MeP options
14503 Enables the @code{abs} instruction, which is the absolute difference
14504 between two registers.
14508 Enables all the optional instructions - average, multiply, divide, bit
14509 operations, leading zero, absolute difference, min/max, clip, and
14515 Enables the @code{ave} instruction, which computes the average of two
14518 @item -mbased=@var{n}
14520 Variables of size @var{n} bytes or smaller will be placed in the
14521 @code{.based} section by default. Based variables use the @code{$tp}
14522 register as a base register, and there is a 128 byte limit to the
14523 @code{.based} section.
14527 Enables the bit operation instructions - bit test (@code{btstm}), set
14528 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14529 test-and-set (@code{tas}).
14531 @item -mc=@var{name}
14533 Selects which section constant data will be placed in. @var{name} may
14534 be @code{tiny}, @code{near}, or @code{far}.
14538 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14539 useful unless you also provide @code{-mminmax}.
14541 @item -mconfig=@var{name}
14543 Selects one of the build-in core configurations. Each MeP chip has
14544 one or more modules in it; each module has a core CPU and a variety of
14545 coprocessors, optional instructions, and peripherals. The
14546 @code{MeP-Integrator} tool, not part of GCC, provides these
14547 configurations through this option; using this option is the same as
14548 using all the corresponding command line options. The default
14549 configuration is @code{default}.
14553 Enables the coprocessor instructions. By default, this is a 32-bit
14554 coprocessor. Note that the coprocessor is normally enabled via the
14555 @code{-mconfig=} option.
14559 Enables the 32-bit coprocessor's instructions.
14563 Enables the 64-bit coprocessor's instructions.
14567 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14571 Causes constant variables to be placed in the @code{.near} section.
14575 Enables the @code{div} and @code{divu} instructions.
14579 Generate big-endian code.
14583 Generate little-endian code.
14585 @item -mio-volatile
14586 @opindex mio-volatile
14587 Tells the compiler that any variable marked with the @code{io}
14588 attribute is to be considered volatile.
14592 Causes variables to be assigned to the @code{.far} section by default.
14596 Enables the @code{leadz} (leading zero) instruction.
14600 Causes variables to be assigned to the @code{.near} section by default.
14604 Enables the @code{min} and @code{max} instructions.
14608 Enables the multiplication and multiply-accumulate instructions.
14612 Disables all the optional instructions enabled by @code{-mall-opts}.
14616 Enables the @code{repeat} and @code{erepeat} instructions, used for
14617 low-overhead looping.
14621 Causes all variables to default to the @code{.tiny} section. Note
14622 that there is a 65536 byte limit to this section. Accesses to these
14623 variables use the @code{%gp} base register.
14627 Enables the saturation instructions. Note that the compiler does not
14628 currently generate these itself, but this option is included for
14629 compatibility with other tools, like @code{as}.
14633 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14637 Link the simulator runtime libraries.
14641 Link the simulator runtime libraries, excluding built-in support
14642 for reset and exception vectors and tables.
14646 Causes all functions to default to the @code{.far} section. Without
14647 this option, functions default to the @code{.near} section.
14649 @item -mtiny=@var{n}
14651 Variables that are @var{n} bytes or smaller will be allocated to the
14652 @code{.tiny} section. These variables use the @code{$gp} base
14653 register. The default for this option is 4, but note that there's a
14654 65536 byte limit to the @code{.tiny} section.
14658 @node MicroBlaze Options
14659 @subsection MicroBlaze Options
14660 @cindex MicroBlaze Options
14665 @opindex msoft-float
14666 Use software emulation for floating point (default).
14669 @opindex mhard-float
14670 Use hardware floating point instructions.
14674 Do not optimize block moves, use @code{memcpy}.
14676 @item -mno-clearbss
14677 @opindex mno-clearbss
14678 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14680 @item -mcpu=@var{cpu-type}
14682 Use features of and schedule code for given CPU.
14683 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14684 where @var{X} is a major version, @var{YY} is the minor version, and
14685 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14686 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14688 @item -mxl-soft-mul
14689 @opindex mxl-soft-mul
14690 Use software multiply emulation (default).
14692 @item -mxl-soft-div
14693 @opindex mxl-soft-div
14694 Use software emulation for divides (default).
14696 @item -mxl-barrel-shift
14697 @opindex mxl-barrel-shift
14698 Use the hardware barrel shifter.
14700 @item -mxl-pattern-compare
14701 @opindex mxl-pattern-compare
14702 Use pattern compare instructions.
14704 @item -msmall-divides
14705 @opindex msmall-divides
14706 Use table lookup optimization for small signed integer divisions.
14708 @item -mxl-stack-check
14709 @opindex mxl-stack-check
14710 This option is deprecated. Use -fstack-check instead.
14713 @opindex mxl-gp-opt
14714 Use GP relative sdata/sbss sections.
14716 @item -mxl-multiply-high
14717 @opindex mxl-multiply-high
14718 Use multiply high instructions for high part of 32x32 multiply.
14720 @item -mxl-float-convert
14721 @opindex mxl-float-convert
14722 Use hardware floating point conversion instructions.
14724 @item -mxl-float-sqrt
14725 @opindex mxl-float-sqrt
14726 Use hardware floating point square root instruction.
14728 @item -mxl-mode-@var{app-model}
14729 Select application model @var{app-model}. Valid models are
14732 normal executable (default), uses startup code @file{crt0.o}.
14735 for use with Xilinx Microprocessor Debugger (XMD) based
14736 software intrusive debug agent called xmdstub. This uses startup file
14737 @file{crt1.o} and sets the start address of the program to be 0x800.
14740 for applications that are loaded using a bootloader.
14741 This model uses startup file @file{crt2.o} which does not contain a processor
14742 reset vector handler. This is suitable for transferring control on a
14743 processor reset to the bootloader rather than the application.
14746 for applications that do not require any of the
14747 MicroBlaze vectors. This option may be useful for applications running
14748 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14751 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14752 @option{-mxl-mode-@var{app-model}}.
14757 @subsection MIPS Options
14758 @cindex MIPS options
14764 Generate big-endian code.
14768 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14771 @item -march=@var{arch}
14773 Generate code that will run on @var{arch}, which can be the name of a
14774 generic MIPS ISA, or the name of a particular processor.
14776 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14777 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14778 The processor names are:
14779 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14780 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14781 @samp{5kc}, @samp{5kf},
14783 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14784 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14785 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14786 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14787 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14788 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14792 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14793 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14794 @samp{rm7000}, @samp{rm9000},
14795 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14798 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14799 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14801 The special value @samp{from-abi} selects the
14802 most compatible architecture for the selected ABI (that is,
14803 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14805 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
14806 which selects the best architecture option for the host processor.
14807 @option{-march=native} has no effect if GCC does not recognize
14810 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14811 (for example, @samp{-march=r2k}). Prefixes are optional, and
14812 @samp{vr} may be written @samp{r}.
14814 Names of the form @samp{@var{n}f2_1} refer to processors with
14815 FPUs clocked at half the rate of the core, names of the form
14816 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14817 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14818 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14819 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14820 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14821 accepted as synonyms for @samp{@var{n}f1_1}.
14823 GCC defines two macros based on the value of this option. The first
14824 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14825 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14826 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14827 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14828 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14830 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14831 above. In other words, it will have the full prefix and will not
14832 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14833 the macro names the resolved architecture (either @samp{"mips1"} or
14834 @samp{"mips3"}). It names the default architecture when no
14835 @option{-march} option is given.
14837 @item -mtune=@var{arch}
14839 Optimize for @var{arch}. Among other things, this option controls
14840 the way instructions are scheduled, and the perceived cost of arithmetic
14841 operations. The list of @var{arch} values is the same as for
14844 When this option is not used, GCC will optimize for the processor
14845 specified by @option{-march}. By using @option{-march} and
14846 @option{-mtune} together, it is possible to generate code that will
14847 run on a family of processors, but optimize the code for one
14848 particular member of that family.
14850 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14851 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14852 @samp{-march} ones described above.
14856 Equivalent to @samp{-march=mips1}.
14860 Equivalent to @samp{-march=mips2}.
14864 Equivalent to @samp{-march=mips3}.
14868 Equivalent to @samp{-march=mips4}.
14872 Equivalent to @samp{-march=mips32}.
14876 Equivalent to @samp{-march=mips32r2}.
14880 Equivalent to @samp{-march=mips64}.
14884 Equivalent to @samp{-march=mips64r2}.
14889 @opindex mno-mips16
14890 Generate (do not generate) MIPS16 code. If GCC is targetting a
14891 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14893 MIPS16 code generation can also be controlled on a per-function basis
14894 by means of @code{mips16} and @code{nomips16} attributes.
14895 @xref{Function Attributes}, for more information.
14897 @item -mflip-mips16
14898 @opindex mflip-mips16
14899 Generate MIPS16 code on alternating functions. This option is provided
14900 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14901 not intended for ordinary use in compiling user code.
14903 @item -minterlink-mips16
14904 @itemx -mno-interlink-mips16
14905 @opindex minterlink-mips16
14906 @opindex mno-interlink-mips16
14907 Require (do not require) that non-MIPS16 code be link-compatible with
14910 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14911 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14912 therefore disables direct jumps unless GCC knows that the target of the
14913 jump is not MIPS16.
14925 Generate code for the given ABI@.
14927 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14928 generates 64-bit code when you select a 64-bit architecture, but you
14929 can use @option{-mgp32} to get 32-bit code instead.
14931 For information about the O64 ABI, see
14932 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14934 GCC supports a variant of the o32 ABI in which floating-point registers
14935 are 64 rather than 32 bits wide. You can select this combination with
14936 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14937 and @samp{mfhc1} instructions and is therefore only supported for
14938 MIPS32R2 processors.
14940 The register assignments for arguments and return values remain the
14941 same, but each scalar value is passed in a single 64-bit register
14942 rather than a pair of 32-bit registers. For example, scalar
14943 floating-point values are returned in @samp{$f0} only, not a
14944 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14945 remains the same, but all 64 bits are saved.
14948 @itemx -mno-abicalls
14950 @opindex mno-abicalls
14951 Generate (do not generate) code that is suitable for SVR4-style
14952 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14957 Generate (do not generate) code that is fully position-independent,
14958 and that can therefore be linked into shared libraries. This option
14959 only affects @option{-mabicalls}.
14961 All @option{-mabicalls} code has traditionally been position-independent,
14962 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14963 as an extension, the GNU toolchain allows executables to use absolute
14964 accesses for locally-binding symbols. It can also use shorter GP
14965 initialization sequences and generate direct calls to locally-defined
14966 functions. This mode is selected by @option{-mno-shared}.
14968 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14969 objects that can only be linked by the GNU linker. However, the option
14970 does not affect the ABI of the final executable; it only affects the ABI
14971 of relocatable objects. Using @option{-mno-shared} will generally make
14972 executables both smaller and quicker.
14974 @option{-mshared} is the default.
14980 Assume (do not assume) that the static and dynamic linkers
14981 support PLTs and copy relocations. This option only affects
14982 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14983 has no effect without @samp{-msym32}.
14985 You can make @option{-mplt} the default by configuring
14986 GCC with @option{--with-mips-plt}. The default is
14987 @option{-mno-plt} otherwise.
14993 Lift (do not lift) the usual restrictions on the size of the global
14996 GCC normally uses a single instruction to load values from the GOT@.
14997 While this is relatively efficient, it will only work if the GOT
14998 is smaller than about 64k. Anything larger will cause the linker
14999 to report an error such as:
15001 @cindex relocation truncated to fit (MIPS)
15003 relocation truncated to fit: R_MIPS_GOT16 foobar
15006 If this happens, you should recompile your code with @option{-mxgot}.
15007 It should then work with very large GOTs, although it will also be
15008 less efficient, since it will take three instructions to fetch the
15009 value of a global symbol.
15011 Note that some linkers can create multiple GOTs. If you have such a
15012 linker, you should only need to use @option{-mxgot} when a single object
15013 file accesses more than 64k's worth of GOT entries. Very few do.
15015 These options have no effect unless GCC is generating position
15020 Assume that general-purpose registers are 32 bits wide.
15024 Assume that general-purpose registers are 64 bits wide.
15028 Assume that floating-point registers are 32 bits wide.
15032 Assume that floating-point registers are 64 bits wide.
15035 @opindex mhard-float
15036 Use floating-point coprocessor instructions.
15039 @opindex msoft-float
15040 Do not use floating-point coprocessor instructions. Implement
15041 floating-point calculations using library calls instead.
15043 @item -msingle-float
15044 @opindex msingle-float
15045 Assume that the floating-point coprocessor only supports single-precision
15048 @item -mdouble-float
15049 @opindex mdouble-float
15050 Assume that the floating-point coprocessor supports double-precision
15051 operations. This is the default.
15057 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
15058 implement atomic memory built-in functions. When neither option is
15059 specified, GCC will use the instructions if the target architecture
15062 @option{-mllsc} is useful if the runtime environment can emulate the
15063 instructions and @option{-mno-llsc} can be useful when compiling for
15064 nonstandard ISAs. You can make either option the default by
15065 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
15066 respectively. @option{--with-llsc} is the default for some
15067 configurations; see the installation documentation for details.
15073 Use (do not use) revision 1 of the MIPS DSP ASE@.
15074 @xref{MIPS DSP Built-in Functions}. This option defines the
15075 preprocessor macro @samp{__mips_dsp}. It also defines
15076 @samp{__mips_dsp_rev} to 1.
15082 Use (do not use) revision 2 of the MIPS DSP ASE@.
15083 @xref{MIPS DSP Built-in Functions}. This option defines the
15084 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
15085 It also defines @samp{__mips_dsp_rev} to 2.
15088 @itemx -mno-smartmips
15089 @opindex msmartmips
15090 @opindex mno-smartmips
15091 Use (do not use) the MIPS SmartMIPS ASE.
15093 @item -mpaired-single
15094 @itemx -mno-paired-single
15095 @opindex mpaired-single
15096 @opindex mno-paired-single
15097 Use (do not use) paired-single floating-point instructions.
15098 @xref{MIPS Paired-Single Support}. This option requires
15099 hardware floating-point support to be enabled.
15105 Use (do not use) MIPS Digital Media Extension instructions.
15106 This option can only be used when generating 64-bit code and requires
15107 hardware floating-point support to be enabled.
15112 @opindex mno-mips3d
15113 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
15114 The option @option{-mips3d} implies @option{-mpaired-single}.
15120 Use (do not use) MT Multithreading instructions.
15124 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
15125 an explanation of the default and the way that the pointer size is
15130 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
15132 The default size of @code{int}s, @code{long}s and pointers depends on
15133 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
15134 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
15135 32-bit @code{long}s. Pointers are the same size as @code{long}s,
15136 or the same size as integer registers, whichever is smaller.
15142 Assume (do not assume) that all symbols have 32-bit values, regardless
15143 of the selected ABI@. This option is useful in combination with
15144 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
15145 to generate shorter and faster references to symbolic addresses.
15149 Put definitions of externally-visible data in a small data section
15150 if that data is no bigger than @var{num} bytes. GCC can then access
15151 the data more efficiently; see @option{-mgpopt} for details.
15153 The default @option{-G} option depends on the configuration.
15155 @item -mlocal-sdata
15156 @itemx -mno-local-sdata
15157 @opindex mlocal-sdata
15158 @opindex mno-local-sdata
15159 Extend (do not extend) the @option{-G} behavior to local data too,
15160 such as to static variables in C@. @option{-mlocal-sdata} is the
15161 default for all configurations.
15163 If the linker complains that an application is using too much small data,
15164 you might want to try rebuilding the less performance-critical parts with
15165 @option{-mno-local-sdata}. You might also want to build large
15166 libraries with @option{-mno-local-sdata}, so that the libraries leave
15167 more room for the main program.
15169 @item -mextern-sdata
15170 @itemx -mno-extern-sdata
15171 @opindex mextern-sdata
15172 @opindex mno-extern-sdata
15173 Assume (do not assume) that externally-defined data will be in
15174 a small data section if that data is within the @option{-G} limit.
15175 @option{-mextern-sdata} is the default for all configurations.
15177 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
15178 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
15179 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
15180 is placed in a small data section. If @var{Var} is defined by another
15181 module, you must either compile that module with a high-enough
15182 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
15183 definition. If @var{Var} is common, you must link the application
15184 with a high-enough @option{-G} setting.
15186 The easiest way of satisfying these restrictions is to compile
15187 and link every module with the same @option{-G} option. However,
15188 you may wish to build a library that supports several different
15189 small data limits. You can do this by compiling the library with
15190 the highest supported @option{-G} setting and additionally using
15191 @option{-mno-extern-sdata} to stop the library from making assumptions
15192 about externally-defined data.
15198 Use (do not use) GP-relative accesses for symbols that are known to be
15199 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
15200 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
15203 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
15204 might not hold the value of @code{_gp}. For example, if the code is
15205 part of a library that might be used in a boot monitor, programs that
15206 call boot monitor routines will pass an unknown value in @code{$gp}.
15207 (In such situations, the boot monitor itself would usually be compiled
15208 with @option{-G0}.)
15210 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
15211 @option{-mno-extern-sdata}.
15213 @item -membedded-data
15214 @itemx -mno-embedded-data
15215 @opindex membedded-data
15216 @opindex mno-embedded-data
15217 Allocate variables to the read-only data section first if possible, then
15218 next in the small data section if possible, otherwise in data. This gives
15219 slightly slower code than the default, but reduces the amount of RAM required
15220 when executing, and thus may be preferred for some embedded systems.
15222 @item -muninit-const-in-rodata
15223 @itemx -mno-uninit-const-in-rodata
15224 @opindex muninit-const-in-rodata
15225 @opindex mno-uninit-const-in-rodata
15226 Put uninitialized @code{const} variables in the read-only data section.
15227 This option is only meaningful in conjunction with @option{-membedded-data}.
15229 @item -mcode-readable=@var{setting}
15230 @opindex mcode-readable
15231 Specify whether GCC may generate code that reads from executable sections.
15232 There are three possible settings:
15235 @item -mcode-readable=yes
15236 Instructions may freely access executable sections. This is the
15239 @item -mcode-readable=pcrel
15240 MIPS16 PC-relative load instructions can access executable sections,
15241 but other instructions must not do so. This option is useful on 4KSc
15242 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
15243 It is also useful on processors that can be configured to have a dual
15244 instruction/data SRAM interface and that, like the M4K, automatically
15245 redirect PC-relative loads to the instruction RAM.
15247 @item -mcode-readable=no
15248 Instructions must not access executable sections. This option can be
15249 useful on targets that are configured to have a dual instruction/data
15250 SRAM interface but that (unlike the M4K) do not automatically redirect
15251 PC-relative loads to the instruction RAM.
15254 @item -msplit-addresses
15255 @itemx -mno-split-addresses
15256 @opindex msplit-addresses
15257 @opindex mno-split-addresses
15258 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
15259 relocation operators. This option has been superseded by
15260 @option{-mexplicit-relocs} but is retained for backwards compatibility.
15262 @item -mexplicit-relocs
15263 @itemx -mno-explicit-relocs
15264 @opindex mexplicit-relocs
15265 @opindex mno-explicit-relocs
15266 Use (do not use) assembler relocation operators when dealing with symbolic
15267 addresses. The alternative, selected by @option{-mno-explicit-relocs},
15268 is to use assembler macros instead.
15270 @option{-mexplicit-relocs} is the default if GCC was configured
15271 to use an assembler that supports relocation operators.
15273 @item -mcheck-zero-division
15274 @itemx -mno-check-zero-division
15275 @opindex mcheck-zero-division
15276 @opindex mno-check-zero-division
15277 Trap (do not trap) on integer division by zero.
15279 The default is @option{-mcheck-zero-division}.
15281 @item -mdivide-traps
15282 @itemx -mdivide-breaks
15283 @opindex mdivide-traps
15284 @opindex mdivide-breaks
15285 MIPS systems check for division by zero by generating either a
15286 conditional trap or a break instruction. Using traps results in
15287 smaller code, but is only supported on MIPS II and later. Also, some
15288 versions of the Linux kernel have a bug that prevents trap from
15289 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
15290 allow conditional traps on architectures that support them and
15291 @option{-mdivide-breaks} to force the use of breaks.
15293 The default is usually @option{-mdivide-traps}, but this can be
15294 overridden at configure time using @option{--with-divide=breaks}.
15295 Divide-by-zero checks can be completely disabled using
15296 @option{-mno-check-zero-division}.
15301 @opindex mno-memcpy
15302 Force (do not force) the use of @code{memcpy()} for non-trivial block
15303 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
15304 most constant-sized copies.
15307 @itemx -mno-long-calls
15308 @opindex mlong-calls
15309 @opindex mno-long-calls
15310 Disable (do not disable) use of the @code{jal} instruction. Calling
15311 functions using @code{jal} is more efficient but requires the caller
15312 and callee to be in the same 256 megabyte segment.
15314 This option has no effect on abicalls code. The default is
15315 @option{-mno-long-calls}.
15321 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
15322 instructions, as provided by the R4650 ISA@.
15325 @itemx -mno-fused-madd
15326 @opindex mfused-madd
15327 @opindex mno-fused-madd
15328 Enable (disable) use of the floating point multiply-accumulate
15329 instructions, when they are available. The default is
15330 @option{-mfused-madd}.
15332 When multiply-accumulate instructions are used, the intermediate
15333 product is calculated to infinite precision and is not subject to
15334 the FCSR Flush to Zero bit. This may be undesirable in some
15339 Tell the MIPS assembler to not run its preprocessor over user
15340 assembler files (with a @samp{.s} suffix) when assembling them.
15345 @opindex mno-fix-24k
15346 Work around the 24K E48 (lost data on stores during refill) errata.
15347 The workarounds are implemented by the assembler rather than by GCC.
15350 @itemx -mno-fix-r4000
15351 @opindex mfix-r4000
15352 @opindex mno-fix-r4000
15353 Work around certain R4000 CPU errata:
15356 A double-word or a variable shift may give an incorrect result if executed
15357 immediately after starting an integer division.
15359 A double-word or a variable shift may give an incorrect result if executed
15360 while an integer multiplication is in progress.
15362 An integer division may give an incorrect result if started in a delay slot
15363 of a taken branch or a jump.
15367 @itemx -mno-fix-r4400
15368 @opindex mfix-r4400
15369 @opindex mno-fix-r4400
15370 Work around certain R4400 CPU errata:
15373 A double-word or a variable shift may give an incorrect result if executed
15374 immediately after starting an integer division.
15378 @itemx -mno-fix-r10000
15379 @opindex mfix-r10000
15380 @opindex mno-fix-r10000
15381 Work around certain R10000 errata:
15384 @code{ll}/@code{sc} sequences may not behave atomically on revisions
15385 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15388 This option can only be used if the target architecture supports
15389 branch-likely instructions. @option{-mfix-r10000} is the default when
15390 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15394 @itemx -mno-fix-vr4120
15395 @opindex mfix-vr4120
15396 Work around certain VR4120 errata:
15399 @code{dmultu} does not always produce the correct result.
15401 @code{div} and @code{ddiv} do not always produce the correct result if one
15402 of the operands is negative.
15404 The workarounds for the division errata rely on special functions in
15405 @file{libgcc.a}. At present, these functions are only provided by
15406 the @code{mips64vr*-elf} configurations.
15408 Other VR4120 errata require a nop to be inserted between certain pairs of
15409 instructions. These errata are handled by the assembler, not by GCC itself.
15412 @opindex mfix-vr4130
15413 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15414 workarounds are implemented by the assembler rather than by GCC,
15415 although GCC will avoid using @code{mflo} and @code{mfhi} if the
15416 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15417 instructions are available instead.
15420 @itemx -mno-fix-sb1
15422 Work around certain SB-1 CPU core errata.
15423 (This flag currently works around the SB-1 revision 2
15424 ``F1'' and ``F2'' floating point errata.)
15426 @item -mr10k-cache-barrier=@var{setting}
15427 @opindex mr10k-cache-barrier
15428 Specify whether GCC should insert cache barriers to avoid the
15429 side-effects of speculation on R10K processors.
15431 In common with many processors, the R10K tries to predict the outcome
15432 of a conditional branch and speculatively executes instructions from
15433 the ``taken'' branch. It later aborts these instructions if the
15434 predicted outcome was wrong. However, on the R10K, even aborted
15435 instructions can have side effects.
15437 This problem only affects kernel stores and, depending on the system,
15438 kernel loads. As an example, a speculatively-executed store may load
15439 the target memory into cache and mark the cache line as dirty, even if
15440 the store itself is later aborted. If a DMA operation writes to the
15441 same area of memory before the ``dirty'' line is flushed, the cached
15442 data will overwrite the DMA-ed data. See the R10K processor manual
15443 for a full description, including other potential problems.
15445 One workaround is to insert cache barrier instructions before every memory
15446 access that might be speculatively executed and that might have side
15447 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15448 controls GCC's implementation of this workaround. It assumes that
15449 aborted accesses to any byte in the following regions will not have
15454 the memory occupied by the current function's stack frame;
15457 the memory occupied by an incoming stack argument;
15460 the memory occupied by an object with a link-time-constant address.
15463 It is the kernel's responsibility to ensure that speculative
15464 accesses to these regions are indeed safe.
15466 If the input program contains a function declaration such as:
15472 then the implementation of @code{foo} must allow @code{j foo} and
15473 @code{jal foo} to be executed speculatively. GCC honors this
15474 restriction for functions it compiles itself. It expects non-GCC
15475 functions (such as hand-written assembly code) to do the same.
15477 The option has three forms:
15480 @item -mr10k-cache-barrier=load-store
15481 Insert a cache barrier before a load or store that might be
15482 speculatively executed and that might have side effects even
15485 @item -mr10k-cache-barrier=store
15486 Insert a cache barrier before a store that might be speculatively
15487 executed and that might have side effects even if aborted.
15489 @item -mr10k-cache-barrier=none
15490 Disable the insertion of cache barriers. This is the default setting.
15493 @item -mflush-func=@var{func}
15494 @itemx -mno-flush-func
15495 @opindex mflush-func
15496 Specifies the function to call to flush the I and D caches, or to not
15497 call any such function. If called, the function must take the same
15498 arguments as the common @code{_flush_func()}, that is, the address of the
15499 memory range for which the cache is being flushed, the size of the
15500 memory range, and the number 3 (to flush both caches). The default
15501 depends on the target GCC was configured for, but commonly is either
15502 @samp{_flush_func} or @samp{__cpu_flush}.
15504 @item mbranch-cost=@var{num}
15505 @opindex mbranch-cost
15506 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15507 This cost is only a heuristic and is not guaranteed to produce
15508 consistent results across releases. A zero cost redundantly selects
15509 the default, which is based on the @option{-mtune} setting.
15511 @item -mbranch-likely
15512 @itemx -mno-branch-likely
15513 @opindex mbranch-likely
15514 @opindex mno-branch-likely
15515 Enable or disable use of Branch Likely instructions, regardless of the
15516 default for the selected architecture. By default, Branch Likely
15517 instructions may be generated if they are supported by the selected
15518 architecture. An exception is for the MIPS32 and MIPS64 architectures
15519 and processors which implement those architectures; for those, Branch
15520 Likely instructions will not be generated by default because the MIPS32
15521 and MIPS64 architectures specifically deprecate their use.
15523 @item -mfp-exceptions
15524 @itemx -mno-fp-exceptions
15525 @opindex mfp-exceptions
15526 Specifies whether FP exceptions are enabled. This affects how we schedule
15527 FP instructions for some processors. The default is that FP exceptions are
15530 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15531 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15534 @item -mvr4130-align
15535 @itemx -mno-vr4130-align
15536 @opindex mvr4130-align
15537 The VR4130 pipeline is two-way superscalar, but can only issue two
15538 instructions together if the first one is 8-byte aligned. When this
15539 option is enabled, GCC will align pairs of instructions that it
15540 thinks should execute in parallel.
15542 This option only has an effect when optimizing for the VR4130.
15543 It normally makes code faster, but at the expense of making it bigger.
15544 It is enabled by default at optimization level @option{-O3}.
15549 Enable (disable) generation of @code{synci} instructions on
15550 architectures that support it. The @code{synci} instructions (if
15551 enabled) will be generated when @code{__builtin___clear_cache()} is
15554 This option defaults to @code{-mno-synci}, but the default can be
15555 overridden by configuring with @code{--with-synci}.
15557 When compiling code for single processor systems, it is generally safe
15558 to use @code{synci}. However, on many multi-core (SMP) systems, it
15559 will not invalidate the instruction caches on all cores and may lead
15560 to undefined behavior.
15562 @item -mrelax-pic-calls
15563 @itemx -mno-relax-pic-calls
15564 @opindex mrelax-pic-calls
15565 Try to turn PIC calls that are normally dispatched via register
15566 @code{$25} into direct calls. This is only possible if the linker can
15567 resolve the destination at link-time and if the destination is within
15568 range for a direct call.
15570 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15571 an assembler and a linker that supports the @code{.reloc} assembly
15572 directive and @code{-mexplicit-relocs} is in effect. With
15573 @code{-mno-explicit-relocs}, this optimization can be performed by the
15574 assembler and the linker alone without help from the compiler.
15576 @item -mmcount-ra-address
15577 @itemx -mno-mcount-ra-address
15578 @opindex mmcount-ra-address
15579 @opindex mno-mcount-ra-address
15580 Emit (do not emit) code that allows @code{_mcount} to modify the
15581 calling function's return address. When enabled, this option extends
15582 the usual @code{_mcount} interface with a new @var{ra-address}
15583 parameter, which has type @code{intptr_t *} and is passed in register
15584 @code{$12}. @code{_mcount} can then modify the return address by
15585 doing both of the following:
15588 Returning the new address in register @code{$31}.
15590 Storing the new address in @code{*@var{ra-address}},
15591 if @var{ra-address} is nonnull.
15594 The default is @option{-mno-mcount-ra-address}.
15599 @subsection MMIX Options
15600 @cindex MMIX Options
15602 These options are defined for the MMIX:
15606 @itemx -mno-libfuncs
15608 @opindex mno-libfuncs
15609 Specify that intrinsic library functions are being compiled, passing all
15610 values in registers, no matter the size.
15613 @itemx -mno-epsilon
15615 @opindex mno-epsilon
15616 Generate floating-point comparison instructions that compare with respect
15617 to the @code{rE} epsilon register.
15619 @item -mabi=mmixware
15621 @opindex mabi=mmixware
15623 Generate code that passes function parameters and return values that (in
15624 the called function) are seen as registers @code{$0} and up, as opposed to
15625 the GNU ABI which uses global registers @code{$231} and up.
15627 @item -mzero-extend
15628 @itemx -mno-zero-extend
15629 @opindex mzero-extend
15630 @opindex mno-zero-extend
15631 When reading data from memory in sizes shorter than 64 bits, use (do not
15632 use) zero-extending load instructions by default, rather than
15633 sign-extending ones.
15636 @itemx -mno-knuthdiv
15638 @opindex mno-knuthdiv
15639 Make the result of a division yielding a remainder have the same sign as
15640 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15641 remainder follows the sign of the dividend. Both methods are
15642 arithmetically valid, the latter being almost exclusively used.
15644 @item -mtoplevel-symbols
15645 @itemx -mno-toplevel-symbols
15646 @opindex mtoplevel-symbols
15647 @opindex mno-toplevel-symbols
15648 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15649 code can be used with the @code{PREFIX} assembly directive.
15653 Generate an executable in the ELF format, rather than the default
15654 @samp{mmo} format used by the @command{mmix} simulator.
15656 @item -mbranch-predict
15657 @itemx -mno-branch-predict
15658 @opindex mbranch-predict
15659 @opindex mno-branch-predict
15660 Use (do not use) the probable-branch instructions, when static branch
15661 prediction indicates a probable branch.
15663 @item -mbase-addresses
15664 @itemx -mno-base-addresses
15665 @opindex mbase-addresses
15666 @opindex mno-base-addresses
15667 Generate (do not generate) code that uses @emph{base addresses}. Using a
15668 base address automatically generates a request (handled by the assembler
15669 and the linker) for a constant to be set up in a global register. The
15670 register is used for one or more base address requests within the range 0
15671 to 255 from the value held in the register. The generally leads to short
15672 and fast code, but the number of different data items that can be
15673 addressed is limited. This means that a program that uses lots of static
15674 data may require @option{-mno-base-addresses}.
15676 @item -msingle-exit
15677 @itemx -mno-single-exit
15678 @opindex msingle-exit
15679 @opindex mno-single-exit
15680 Force (do not force) generated code to have a single exit point in each
15684 @node MN10300 Options
15685 @subsection MN10300 Options
15686 @cindex MN10300 options
15688 These @option{-m} options are defined for Matsushita MN10300 architectures:
15693 Generate code to avoid bugs in the multiply instructions for the MN10300
15694 processors. This is the default.
15696 @item -mno-mult-bug
15697 @opindex mno-mult-bug
15698 Do not generate code to avoid bugs in the multiply instructions for the
15699 MN10300 processors.
15703 Generate code which uses features specific to the AM33 processor.
15707 Do not generate code which uses features specific to the AM33 processor. This
15712 Generate code which uses features specific to the AM33/2.0 processor.
15716 Generate code which uses features specific to the AM34 processor.
15718 @item -mtune=@var{cpu-type}
15720 Use the timing characteristics of the indicated CPU type when
15721 scheduling instructions. This does not change the targeted processor
15722 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15723 @samp{am33-2} or @samp{am34}.
15725 @item -mreturn-pointer-on-d0
15726 @opindex mreturn-pointer-on-d0
15727 When generating a function which returns a pointer, return the pointer
15728 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15729 only in a0, and attempts to call such functions without a prototype
15730 would result in errors. Note that this option is on by default; use
15731 @option{-mno-return-pointer-on-d0} to disable it.
15735 Do not link in the C run-time initialization object file.
15739 Indicate to the linker that it should perform a relaxation optimization pass
15740 to shorten branches, calls and absolute memory addresses. This option only
15741 has an effect when used on the command line for the final link step.
15743 This option makes symbolic debugging impossible.
15747 Allow the compiler to generate @emph{Long Instruction Word}
15748 instructions if the target is the @samp{AM33} or later. This is the
15749 default. This option defines the preprocessor macro @samp{__LIW__}.
15753 Do not allow the compiler to generate @emph{Long Instruction Word}
15754 instructions. This option defines the preprocessor macro
15759 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15760 instructions if the target is the @samp{AM33} or later. This is the
15761 default. This option defines the preprocessor macro @samp{__SETLB__}.
15765 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15766 instructions. This option defines the preprocessor macro
15767 @samp{__NO_SETLB__}.
15771 @node PDP-11 Options
15772 @subsection PDP-11 Options
15773 @cindex PDP-11 Options
15775 These options are defined for the PDP-11:
15780 Use hardware FPP floating point. This is the default. (FIS floating
15781 point on the PDP-11/40 is not supported.)
15784 @opindex msoft-float
15785 Do not use hardware floating point.
15789 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15793 Return floating-point results in memory. This is the default.
15797 Generate code for a PDP-11/40.
15801 Generate code for a PDP-11/45. This is the default.
15805 Generate code for a PDP-11/10.
15807 @item -mbcopy-builtin
15808 @opindex mbcopy-builtin
15809 Use inline @code{movmemhi} patterns for copying memory. This is the
15814 Do not use inline @code{movmemhi} patterns for copying memory.
15820 Use 16-bit @code{int}. This is the default.
15826 Use 32-bit @code{int}.
15829 @itemx -mno-float32
15831 @opindex mno-float32
15832 Use 64-bit @code{float}. This is the default.
15835 @itemx -mno-float64
15837 @opindex mno-float64
15838 Use 32-bit @code{float}.
15842 Use @code{abshi2} pattern. This is the default.
15846 Do not use @code{abshi2} pattern.
15848 @item -mbranch-expensive
15849 @opindex mbranch-expensive
15850 Pretend that branches are expensive. This is for experimenting with
15851 code generation only.
15853 @item -mbranch-cheap
15854 @opindex mbranch-cheap
15855 Do not pretend that branches are expensive. This is the default.
15859 Use Unix assembler syntax. This is the default when configured for
15860 @samp{pdp11-*-bsd}.
15864 Use DEC assembler syntax. This is the default when configured for any
15865 PDP-11 target other than @samp{pdp11-*-bsd}.
15868 @node picoChip Options
15869 @subsection picoChip Options
15870 @cindex picoChip options
15872 These @samp{-m} options are defined for picoChip implementations:
15876 @item -mae=@var{ae_type}
15878 Set the instruction set, register set, and instruction scheduling
15879 parameters for array element type @var{ae_type}. Supported values
15880 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15882 @option{-mae=ANY} selects a completely generic AE type. Code
15883 generated with this option will run on any of the other AE types. The
15884 code will not be as efficient as it would be if compiled for a specific
15885 AE type, and some types of operation (e.g., multiplication) will not
15886 work properly on all types of AE.
15888 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15889 for compiled code, and is the default.
15891 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15892 option may suffer from poor performance of byte (char) manipulation,
15893 since the DSP AE does not provide hardware support for byte load/stores.
15895 @item -msymbol-as-address
15896 Enable the compiler to directly use a symbol name as an address in a
15897 load/store instruction, without first loading it into a
15898 register. Typically, the use of this option will generate larger
15899 programs, which run faster than when the option isn't used. However, the
15900 results vary from program to program, so it is left as a user option,
15901 rather than being permanently enabled.
15903 @item -mno-inefficient-warnings
15904 Disables warnings about the generation of inefficient code. These
15905 warnings can be generated, for example, when compiling code which
15906 performs byte-level memory operations on the MAC AE type. The MAC AE has
15907 no hardware support for byte-level memory operations, so all byte
15908 load/stores must be synthesized from word load/store operations. This is
15909 inefficient and a warning will be generated indicating to the programmer
15910 that they should rewrite the code to avoid byte operations, or to target
15911 an AE type which has the necessary hardware support. This option enables
15912 the warning to be turned off.
15916 @node PowerPC Options
15917 @subsection PowerPC Options
15918 @cindex PowerPC options
15920 These are listed under @xref{RS/6000 and PowerPC Options}.
15922 @node RS/6000 and PowerPC Options
15923 @subsection IBM RS/6000 and PowerPC Options
15924 @cindex RS/6000 and PowerPC Options
15925 @cindex IBM RS/6000 and PowerPC Options
15927 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15934 @itemx -mno-powerpc
15935 @itemx -mpowerpc-gpopt
15936 @itemx -mno-powerpc-gpopt
15937 @itemx -mpowerpc-gfxopt
15938 @itemx -mno-powerpc-gfxopt
15941 @itemx -mno-powerpc64
15945 @itemx -mno-popcntb
15947 @itemx -mno-popcntd
15956 @itemx -mno-hard-dfp
15960 @opindex mno-power2
15962 @opindex mno-powerpc
15963 @opindex mpowerpc-gpopt
15964 @opindex mno-powerpc-gpopt
15965 @opindex mpowerpc-gfxopt
15966 @opindex mno-powerpc-gfxopt
15967 @opindex mpowerpc64
15968 @opindex mno-powerpc64
15972 @opindex mno-popcntb
15974 @opindex mno-popcntd
15980 @opindex mno-mfpgpr
15982 @opindex mno-hard-dfp
15983 GCC supports two related instruction set architectures for the
15984 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15985 instructions supported by the @samp{rios} chip set used in the original
15986 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15987 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15988 the IBM 4xx, 6xx, and follow-on microprocessors.
15990 Neither architecture is a subset of the other. However there is a
15991 large common subset of instructions supported by both. An MQ
15992 register is included in processors supporting the POWER architecture.
15994 You use these options to specify which instructions are available on the
15995 processor you are using. The default value of these options is
15996 determined when configuring GCC@. Specifying the
15997 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15998 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15999 rather than the options listed above.
16001 The @option{-mpower} option allows GCC to generate instructions that
16002 are found only in the POWER architecture and to use the MQ register.
16003 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
16004 to generate instructions that are present in the POWER2 architecture but
16005 not the original POWER architecture.
16007 The @option{-mpowerpc} option allows GCC to generate instructions that
16008 are found only in the 32-bit subset of the PowerPC architecture.
16009 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
16010 GCC to use the optional PowerPC architecture instructions in the
16011 General Purpose group, including floating-point square root. Specifying
16012 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
16013 use the optional PowerPC architecture instructions in the Graphics
16014 group, including floating-point select.
16016 The @option{-mmfcrf} option allows GCC to generate the move from
16017 condition register field instruction implemented on the POWER4
16018 processor and other processors that support the PowerPC V2.01
16020 The @option{-mpopcntb} option allows GCC to generate the popcount and
16021 double precision FP reciprocal estimate instruction implemented on the
16022 POWER5 processor and other processors that support the PowerPC V2.02
16024 The @option{-mpopcntd} option allows GCC to generate the popcount
16025 instruction implemented on the POWER7 processor and other processors
16026 that support the PowerPC V2.06 architecture.
16027 The @option{-mfprnd} option allows GCC to generate the FP round to
16028 integer instructions implemented on the POWER5+ processor and other
16029 processors that support the PowerPC V2.03 architecture.
16030 The @option{-mcmpb} option allows GCC to generate the compare bytes
16031 instruction implemented on the POWER6 processor and other processors
16032 that support the PowerPC V2.05 architecture.
16033 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
16034 general purpose register instructions implemented on the POWER6X
16035 processor and other processors that support the extended PowerPC V2.05
16037 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
16038 point instructions implemented on some POWER processors.
16040 The @option{-mpowerpc64} option allows GCC to generate the additional
16041 64-bit instructions that are found in the full PowerPC64 architecture
16042 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
16043 @option{-mno-powerpc64}.
16045 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
16046 will use only the instructions in the common subset of both
16047 architectures plus some special AIX common-mode calls, and will not use
16048 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
16049 permits GCC to use any instruction from either architecture and to
16050 allow use of the MQ register; specify this for the Motorola MPC601.
16052 @item -mnew-mnemonics
16053 @itemx -mold-mnemonics
16054 @opindex mnew-mnemonics
16055 @opindex mold-mnemonics
16056 Select which mnemonics to use in the generated assembler code. With
16057 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
16058 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
16059 assembler mnemonics defined for the POWER architecture. Instructions
16060 defined in only one architecture have only one mnemonic; GCC uses that
16061 mnemonic irrespective of which of these options is specified.
16063 GCC defaults to the mnemonics appropriate for the architecture in
16064 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
16065 value of these option. Unless you are building a cross-compiler, you
16066 should normally not specify either @option{-mnew-mnemonics} or
16067 @option{-mold-mnemonics}, but should instead accept the default.
16069 @item -mcpu=@var{cpu_type}
16071 Set architecture type, register usage, choice of mnemonics, and
16072 instruction scheduling parameters for machine type @var{cpu_type}.
16073 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
16074 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
16075 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
16076 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
16077 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
16078 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
16079 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
16080 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
16081 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
16082 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
16083 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
16085 @option{-mcpu=common} selects a completely generic processor. Code
16086 generated under this option will run on any POWER or PowerPC processor.
16087 GCC will use only the instructions in the common subset of both
16088 architectures, and will not use the MQ register. GCC assumes a generic
16089 processor model for scheduling purposes.
16091 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
16092 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
16093 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
16094 types, with an appropriate, generic processor model assumed for
16095 scheduling purposes.
16097 The other options specify a specific processor. Code generated under
16098 those options will run best on that processor, and may not run at all on
16101 The @option{-mcpu} options automatically enable or disable the
16104 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
16105 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
16106 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
16107 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
16109 The particular options set for any particular CPU will vary between
16110 compiler versions, depending on what setting seems to produce optimal
16111 code for that CPU; it doesn't necessarily reflect the actual hardware's
16112 capabilities. If you wish to set an individual option to a particular
16113 value, you may specify it after the @option{-mcpu} option, like
16114 @samp{-mcpu=970 -mno-altivec}.
16116 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
16117 not enabled or disabled by the @option{-mcpu} option at present because
16118 AIX does not have full support for these options. You may still
16119 enable or disable them individually if you're sure it'll work in your
16122 @item -mtune=@var{cpu_type}
16124 Set the instruction scheduling parameters for machine type
16125 @var{cpu_type}, but do not set the architecture type, register usage, or
16126 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
16127 values for @var{cpu_type} are used for @option{-mtune} as for
16128 @option{-mcpu}. If both are specified, the code generated will use the
16129 architecture, registers, and mnemonics set by @option{-mcpu}, but the
16130 scheduling parameters set by @option{-mtune}.
16132 @item -mcmodel=small
16133 @opindex mcmodel=small
16134 Generate PowerPC64 code for the small model: The TOC is limited to
16137 @item -mcmodel=medium
16138 @opindex mcmodel=medium
16139 Generate PowerPC64 code for the medium model: The TOC and other static
16140 data may be up to a total of 4G in size.
16142 @item -mcmodel=large
16143 @opindex mcmodel=large
16144 Generate PowerPC64 code for the large model: The TOC may be up to 4G
16145 in size. Other data and code is only limited by the 64-bit address
16149 @itemx -mno-altivec
16151 @opindex mno-altivec
16152 Generate code that uses (does not use) AltiVec instructions, and also
16153 enable the use of built-in functions that allow more direct access to
16154 the AltiVec instruction set. You may also need to set
16155 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
16161 @opindex mno-vrsave
16162 Generate VRSAVE instructions when generating AltiVec code.
16164 @item -mgen-cell-microcode
16165 @opindex mgen-cell-microcode
16166 Generate Cell microcode instructions
16168 @item -mwarn-cell-microcode
16169 @opindex mwarn-cell-microcode
16170 Warning when a Cell microcode instruction is going to emitted. An example
16171 of a Cell microcode instruction is a variable shift.
16174 @opindex msecure-plt
16175 Generate code that allows ld and ld.so to build executables and shared
16176 libraries with non-exec .plt and .got sections. This is a PowerPC
16177 32-bit SYSV ABI option.
16181 Generate code that uses a BSS .plt section that ld.so fills in, and
16182 requires .plt and .got sections that are both writable and executable.
16183 This is a PowerPC 32-bit SYSV ABI option.
16189 This switch enables or disables the generation of ISEL instructions.
16191 @item -misel=@var{yes/no}
16192 This switch has been deprecated. Use @option{-misel} and
16193 @option{-mno-isel} instead.
16199 This switch enables or disables the generation of SPE simd
16205 @opindex mno-paired
16206 This switch enables or disables the generation of PAIRED simd
16209 @item -mspe=@var{yes/no}
16210 This option has been deprecated. Use @option{-mspe} and
16211 @option{-mno-spe} instead.
16217 Generate code that uses (does not use) vector/scalar (VSX)
16218 instructions, and also enable the use of built-in functions that allow
16219 more direct access to the VSX instruction set.
16221 @item -mfloat-gprs=@var{yes/single/double/no}
16222 @itemx -mfloat-gprs
16223 @opindex mfloat-gprs
16224 This switch enables or disables the generation of floating point
16225 operations on the general purpose registers for architectures that
16228 The argument @var{yes} or @var{single} enables the use of
16229 single-precision floating point operations.
16231 The argument @var{double} enables the use of single and
16232 double-precision floating point operations.
16234 The argument @var{no} disables floating point operations on the
16235 general purpose registers.
16237 This option is currently only available on the MPC854x.
16243 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
16244 targets (including GNU/Linux). The 32-bit environment sets int, long
16245 and pointer to 32 bits and generates code that runs on any PowerPC
16246 variant. The 64-bit environment sets int to 32 bits and long and
16247 pointer to 64 bits, and generates code for PowerPC64, as for
16248 @option{-mpowerpc64}.
16251 @itemx -mno-fp-in-toc
16252 @itemx -mno-sum-in-toc
16253 @itemx -mminimal-toc
16255 @opindex mno-fp-in-toc
16256 @opindex mno-sum-in-toc
16257 @opindex mminimal-toc
16258 Modify generation of the TOC (Table Of Contents), which is created for
16259 every executable file. The @option{-mfull-toc} option is selected by
16260 default. In that case, GCC will allocate at least one TOC entry for
16261 each unique non-automatic variable reference in your program. GCC
16262 will also place floating-point constants in the TOC@. However, only
16263 16,384 entries are available in the TOC@.
16265 If you receive a linker error message that saying you have overflowed
16266 the available TOC space, you can reduce the amount of TOC space used
16267 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
16268 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
16269 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
16270 generate code to calculate the sum of an address and a constant at
16271 run-time instead of putting that sum into the TOC@. You may specify one
16272 or both of these options. Each causes GCC to produce very slightly
16273 slower and larger code at the expense of conserving TOC space.
16275 If you still run out of space in the TOC even when you specify both of
16276 these options, specify @option{-mminimal-toc} instead. This option causes
16277 GCC to make only one TOC entry for every file. When you specify this
16278 option, GCC will produce code that is slower and larger but which
16279 uses extremely little TOC space. You may wish to use this option
16280 only on files that contain less frequently executed code.
16286 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
16287 @code{long} type, and the infrastructure needed to support them.
16288 Specifying @option{-maix64} implies @option{-mpowerpc64} and
16289 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
16290 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
16293 @itemx -mno-xl-compat
16294 @opindex mxl-compat
16295 @opindex mno-xl-compat
16296 Produce code that conforms more closely to IBM XL compiler semantics
16297 when using AIX-compatible ABI@. Pass floating-point arguments to
16298 prototyped functions beyond the register save area (RSA) on the stack
16299 in addition to argument FPRs. Do not assume that most significant
16300 double in 128-bit long double value is properly rounded when comparing
16301 values and converting to double. Use XL symbol names for long double
16304 The AIX calling convention was extended but not initially documented to
16305 handle an obscure K&R C case of calling a function that takes the
16306 address of its arguments with fewer arguments than declared. IBM XL
16307 compilers access floating point arguments which do not fit in the
16308 RSA from the stack when a subroutine is compiled without
16309 optimization. Because always storing floating-point arguments on the
16310 stack is inefficient and rarely needed, this option is not enabled by
16311 default and only is necessary when calling subroutines compiled by IBM
16312 XL compilers without optimization.
16316 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
16317 application written to use message passing with special startup code to
16318 enable the application to run. The system must have PE installed in the
16319 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
16320 must be overridden with the @option{-specs=} option to specify the
16321 appropriate directory location. The Parallel Environment does not
16322 support threads, so the @option{-mpe} option and the @option{-pthread}
16323 option are incompatible.
16325 @item -malign-natural
16326 @itemx -malign-power
16327 @opindex malign-natural
16328 @opindex malign-power
16329 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
16330 @option{-malign-natural} overrides the ABI-defined alignment of larger
16331 types, such as floating-point doubles, on their natural size-based boundary.
16332 The option @option{-malign-power} instructs GCC to follow the ABI-specified
16333 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
16335 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
16339 @itemx -mhard-float
16340 @opindex msoft-float
16341 @opindex mhard-float
16342 Generate code that does not use (uses) the floating-point register set.
16343 Software floating point emulation is provided if you use the
16344 @option{-msoft-float} option, and pass the option to GCC when linking.
16346 @item -msingle-float
16347 @itemx -mdouble-float
16348 @opindex msingle-float
16349 @opindex mdouble-float
16350 Generate code for single or double-precision floating point operations.
16351 @option{-mdouble-float} implies @option{-msingle-float}.
16354 @opindex msimple-fpu
16355 Do not generate sqrt and div instructions for hardware floating point unit.
16359 Specify type of floating point unit. Valid values are @var{sp_lite}
16360 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
16361 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
16362 and @var{dp_full} (equivalent to -mdouble-float).
16365 @opindex mxilinx-fpu
16366 Perform optimizations for floating point unit on Xilinx PPC 405/440.
16369 @itemx -mno-multiple
16371 @opindex mno-multiple
16372 Generate code that uses (does not use) the load multiple word
16373 instructions and the store multiple word instructions. These
16374 instructions are generated by default on POWER systems, and not
16375 generated on PowerPC systems. Do not use @option{-mmultiple} on little
16376 endian PowerPC systems, since those instructions do not work when the
16377 processor is in little endian mode. The exceptions are PPC740 and
16378 PPC750 which permit the instructions usage in little endian mode.
16383 @opindex mno-string
16384 Generate code that uses (does not use) the load string instructions
16385 and the store string word instructions to save multiple registers and
16386 do small block moves. These instructions are generated by default on
16387 POWER systems, and not generated on PowerPC systems. Do not use
16388 @option{-mstring} on little endian PowerPC systems, since those
16389 instructions do not work when the processor is in little endian mode.
16390 The exceptions are PPC740 and PPC750 which permit the instructions
16391 usage in little endian mode.
16396 @opindex mno-update
16397 Generate code that uses (does not use) the load or store instructions
16398 that update the base register to the address of the calculated memory
16399 location. These instructions are generated by default. If you use
16400 @option{-mno-update}, there is a small window between the time that the
16401 stack pointer is updated and the address of the previous frame is
16402 stored, which means code that walks the stack frame across interrupts or
16403 signals may get corrupted data.
16405 @item -mavoid-indexed-addresses
16406 @itemx -mno-avoid-indexed-addresses
16407 @opindex mavoid-indexed-addresses
16408 @opindex mno-avoid-indexed-addresses
16409 Generate code that tries to avoid (not avoid) the use of indexed load
16410 or store instructions. These instructions can incur a performance
16411 penalty on Power6 processors in certain situations, such as when
16412 stepping through large arrays that cross a 16M boundary. This option
16413 is enabled by default when targetting Power6 and disabled otherwise.
16416 @itemx -mno-fused-madd
16417 @opindex mfused-madd
16418 @opindex mno-fused-madd
16419 Generate code that uses (does not use) the floating point multiply and
16420 accumulate instructions. These instructions are generated by default
16421 if hardware floating point is used. The machine dependent
16422 @option{-mfused-madd} option is now mapped to the machine independent
16423 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16424 mapped to @option{-ffp-contract=off}.
16430 Generate code that uses (does not use) the half-word multiply and
16431 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
16432 These instructions are generated by default when targetting those
16439 Generate code that uses (does not use) the string-search @samp{dlmzb}
16440 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
16441 generated by default when targetting those processors.
16443 @item -mno-bit-align
16445 @opindex mno-bit-align
16446 @opindex mbit-align
16447 On System V.4 and embedded PowerPC systems do not (do) force structures
16448 and unions that contain bit-fields to be aligned to the base type of the
16451 For example, by default a structure containing nothing but 8
16452 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
16453 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
16454 the structure would be aligned to a 1 byte boundary and be one byte in
16457 @item -mno-strict-align
16458 @itemx -mstrict-align
16459 @opindex mno-strict-align
16460 @opindex mstrict-align
16461 On System V.4 and embedded PowerPC systems do not (do) assume that
16462 unaligned memory references will be handled by the system.
16464 @item -mrelocatable
16465 @itemx -mno-relocatable
16466 @opindex mrelocatable
16467 @opindex mno-relocatable
16468 Generate code that allows (does not allow) a static executable to be
16469 relocated to a different address at runtime. A simple embedded
16470 PowerPC system loader should relocate the entire contents of
16471 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16472 a table of 32-bit addresses generated by this option. For this to
16473 work, all objects linked together must be compiled with
16474 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16475 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
16477 @item -mrelocatable-lib
16478 @itemx -mno-relocatable-lib
16479 @opindex mrelocatable-lib
16480 @opindex mno-relocatable-lib
16481 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16482 @code{.fixup} section to allow static executables to be relocated at
16483 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
16484 alignment of @option{-mrelocatable}. Objects compiled with
16485 @option{-mrelocatable-lib} may be linked with objects compiled with
16486 any combination of the @option{-mrelocatable} options.
16492 On System V.4 and embedded PowerPC systems do not (do) assume that
16493 register 2 contains a pointer to a global area pointing to the addresses
16494 used in the program.
16497 @itemx -mlittle-endian
16499 @opindex mlittle-endian
16500 On System V.4 and embedded PowerPC systems compile code for the
16501 processor in little endian mode. The @option{-mlittle-endian} option is
16502 the same as @option{-mlittle}.
16505 @itemx -mbig-endian
16507 @opindex mbig-endian
16508 On System V.4 and embedded PowerPC systems compile code for the
16509 processor in big endian mode. The @option{-mbig-endian} option is
16510 the same as @option{-mbig}.
16512 @item -mdynamic-no-pic
16513 @opindex mdynamic-no-pic
16514 On Darwin and Mac OS X systems, compile code so that it is not
16515 relocatable, but that its external references are relocatable. The
16516 resulting code is suitable for applications, but not shared
16519 @item -msingle-pic-base
16520 @opindex msingle-pic-base
16521 Treat the register used for PIC addressing as read-only, rather than
16522 loading it in the prologue for each function. The run-time system is
16523 responsible for initializing this register with an appropriate value
16524 before execution begins.
16526 @item -mprioritize-restricted-insns=@var{priority}
16527 @opindex mprioritize-restricted-insns
16528 This option controls the priority that is assigned to
16529 dispatch-slot restricted instructions during the second scheduling
16530 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16531 @var{no/highest/second-highest} priority to dispatch slot restricted
16534 @item -msched-costly-dep=@var{dependence_type}
16535 @opindex msched-costly-dep
16536 This option controls which dependences are considered costly
16537 by the target during instruction scheduling. The argument
16538 @var{dependence_type} takes one of the following values:
16539 @var{no}: no dependence is costly,
16540 @var{all}: all dependences are costly,
16541 @var{true_store_to_load}: a true dependence from store to load is costly,
16542 @var{store_to_load}: any dependence from store to load is costly,
16543 @var{number}: any dependence which latency >= @var{number} is costly.
16545 @item -minsert-sched-nops=@var{scheme}
16546 @opindex minsert-sched-nops
16547 This option controls which nop insertion scheme will be used during
16548 the second scheduling pass. The argument @var{scheme} takes one of the
16550 @var{no}: Don't insert nops.
16551 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16552 according to the scheduler's grouping.
16553 @var{regroup_exact}: Insert nops to force costly dependent insns into
16554 separate groups. Insert exactly as many nops as needed to force an insn
16555 to a new group, according to the estimated processor grouping.
16556 @var{number}: Insert nops to force costly dependent insns into
16557 separate groups. Insert @var{number} nops to force an insn to a new group.
16560 @opindex mcall-sysv
16561 On System V.4 and embedded PowerPC systems compile code using calling
16562 conventions that adheres to the March 1995 draft of the System V
16563 Application Binary Interface, PowerPC processor supplement. This is the
16564 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16566 @item -mcall-sysv-eabi
16568 @opindex mcall-sysv-eabi
16569 @opindex mcall-eabi
16570 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16572 @item -mcall-sysv-noeabi
16573 @opindex mcall-sysv-noeabi
16574 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16576 @item -mcall-aixdesc
16578 On System V.4 and embedded PowerPC systems compile code for the AIX
16582 @opindex mcall-linux
16583 On System V.4 and embedded PowerPC systems compile code for the
16584 Linux-based GNU system.
16586 @item -mcall-freebsd
16587 @opindex mcall-freebsd
16588 On System V.4 and embedded PowerPC systems compile code for the
16589 FreeBSD operating system.
16591 @item -mcall-netbsd
16592 @opindex mcall-netbsd
16593 On System V.4 and embedded PowerPC systems compile code for the
16594 NetBSD operating system.
16596 @item -mcall-openbsd
16597 @opindex mcall-netbsd
16598 On System V.4 and embedded PowerPC systems compile code for the
16599 OpenBSD operating system.
16601 @item -maix-struct-return
16602 @opindex maix-struct-return
16603 Return all structures in memory (as specified by the AIX ABI)@.
16605 @item -msvr4-struct-return
16606 @opindex msvr4-struct-return
16607 Return structures smaller than 8 bytes in registers (as specified by the
16610 @item -mabi=@var{abi-type}
16612 Extend the current ABI with a particular extension, or remove such extension.
16613 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16614 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16618 Extend the current ABI with SPE ABI extensions. This does not change
16619 the default ABI, instead it adds the SPE ABI extensions to the current
16623 @opindex mabi=no-spe
16624 Disable Booke SPE ABI extensions for the current ABI@.
16626 @item -mabi=ibmlongdouble
16627 @opindex mabi=ibmlongdouble
16628 Change the current ABI to use IBM extended precision long double.
16629 This is a PowerPC 32-bit SYSV ABI option.
16631 @item -mabi=ieeelongdouble
16632 @opindex mabi=ieeelongdouble
16633 Change the current ABI to use IEEE extended precision long double.
16634 This is a PowerPC 32-bit Linux ABI option.
16637 @itemx -mno-prototype
16638 @opindex mprototype
16639 @opindex mno-prototype
16640 On System V.4 and embedded PowerPC systems assume that all calls to
16641 variable argument functions are properly prototyped. Otherwise, the
16642 compiler must insert an instruction before every non prototyped call to
16643 set or clear bit 6 of the condition code register (@var{CR}) to
16644 indicate whether floating point values were passed in the floating point
16645 registers in case the function takes a variable arguments. With
16646 @option{-mprototype}, only calls to prototyped variable argument functions
16647 will set or clear the bit.
16651 On embedded PowerPC systems, assume that the startup module is called
16652 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16653 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16658 On embedded PowerPC systems, assume that the startup module is called
16659 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16664 On embedded PowerPC systems, assume that the startup module is called
16665 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16668 @item -myellowknife
16669 @opindex myellowknife
16670 On embedded PowerPC systems, assume that the startup module is called
16671 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16676 On System V.4 and embedded PowerPC systems, specify that you are
16677 compiling for a VxWorks system.
16681 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16682 header to indicate that @samp{eabi} extended relocations are used.
16688 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16689 Embedded Applications Binary Interface (eabi) which is a set of
16690 modifications to the System V.4 specifications. Selecting @option{-meabi}
16691 means that the stack is aligned to an 8 byte boundary, a function
16692 @code{__eabi} is called to from @code{main} to set up the eabi
16693 environment, and the @option{-msdata} option can use both @code{r2} and
16694 @code{r13} to point to two separate small data areas. Selecting
16695 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16696 do not call an initialization function from @code{main}, and the
16697 @option{-msdata} option will only use @code{r13} to point to a single
16698 small data area. The @option{-meabi} option is on by default if you
16699 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16702 @opindex msdata=eabi
16703 On System V.4 and embedded PowerPC systems, put small initialized
16704 @code{const} global and static data in the @samp{.sdata2} section, which
16705 is pointed to by register @code{r2}. Put small initialized
16706 non-@code{const} global and static data in the @samp{.sdata} section,
16707 which is pointed to by register @code{r13}. Put small uninitialized
16708 global and static data in the @samp{.sbss} section, which is adjacent to
16709 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16710 incompatible with the @option{-mrelocatable} option. The
16711 @option{-msdata=eabi} option also sets the @option{-memb} option.
16714 @opindex msdata=sysv
16715 On System V.4 and embedded PowerPC systems, put small global and static
16716 data in the @samp{.sdata} section, which is pointed to by register
16717 @code{r13}. Put small uninitialized global and static data in the
16718 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16719 The @option{-msdata=sysv} option is incompatible with the
16720 @option{-mrelocatable} option.
16722 @item -msdata=default
16724 @opindex msdata=default
16726 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16727 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16728 same as @option{-msdata=sysv}.
16731 @opindex msdata=data
16732 On System V.4 and embedded PowerPC systems, put small global
16733 data in the @samp{.sdata} section. Put small uninitialized global
16734 data in the @samp{.sbss} section. Do not use register @code{r13}
16735 to address small data however. This is the default behavior unless
16736 other @option{-msdata} options are used.
16740 @opindex msdata=none
16742 On embedded PowerPC systems, put all initialized global and static data
16743 in the @samp{.data} section, and all uninitialized data in the
16744 @samp{.bss} section.
16746 @item -mblock-move-inline-limit=@var{num}
16747 @opindex mblock-move-inline-limit
16748 Inline all block moves (such as calls to @code{memcpy} or structure
16749 copies) less than or equal to @var{num} bytes. The minimum value for
16750 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16751 targets. The default value is target-specific.
16755 @cindex smaller data references (PowerPC)
16756 @cindex .sdata/.sdata2 references (PowerPC)
16757 On embedded PowerPC systems, put global and static items less than or
16758 equal to @var{num} bytes into the small data or bss sections instead of
16759 the normal data or bss section. By default, @var{num} is 8. The
16760 @option{-G @var{num}} switch is also passed to the linker.
16761 All modules should be compiled with the same @option{-G @var{num}} value.
16764 @itemx -mno-regnames
16766 @opindex mno-regnames
16767 On System V.4 and embedded PowerPC systems do (do not) emit register
16768 names in the assembly language output using symbolic forms.
16771 @itemx -mno-longcall
16773 @opindex mno-longcall
16774 By default assume that all calls are far away so that a longer more
16775 expensive calling sequence is required. This is required for calls
16776 further than 32 megabytes (33,554,432 bytes) from the current location.
16777 A short call will be generated if the compiler knows
16778 the call cannot be that far away. This setting can be overridden by
16779 the @code{shortcall} function attribute, or by @code{#pragma
16782 Some linkers are capable of detecting out-of-range calls and generating
16783 glue code on the fly. On these systems, long calls are unnecessary and
16784 generate slower code. As of this writing, the AIX linker can do this,
16785 as can the GNU linker for PowerPC/64. It is planned to add this feature
16786 to the GNU linker for 32-bit PowerPC systems as well.
16788 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16789 callee, L42'', plus a ``branch island'' (glue code). The two target
16790 addresses represent the callee and the ``branch island''. The
16791 Darwin/PPC linker will prefer the first address and generate a ``bl
16792 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16793 otherwise, the linker will generate ``bl L42'' to call the ``branch
16794 island''. The ``branch island'' is appended to the body of the
16795 calling function; it computes the full 32-bit address of the callee
16798 On Mach-O (Darwin) systems, this option directs the compiler emit to
16799 the glue for every direct call, and the Darwin linker decides whether
16800 to use or discard it.
16802 In the future, we may cause GCC to ignore all longcall specifications
16803 when the linker is known to generate glue.
16805 @item -mtls-markers
16806 @itemx -mno-tls-markers
16807 @opindex mtls-markers
16808 @opindex mno-tls-markers
16809 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16810 specifying the function argument. The relocation allows ld to
16811 reliably associate function call with argument setup instructions for
16812 TLS optimization, which in turn allows gcc to better schedule the
16817 Adds support for multithreading with the @dfn{pthreads} library.
16818 This option sets flags for both the preprocessor and linker.
16823 This option will enable GCC to use the reciprocal estimate and
16824 reciprocal square root estimate instructions with additional
16825 Newton-Raphson steps to increase precision instead of doing a divide or
16826 square root and divide for floating point arguments. You should use
16827 the @option{-ffast-math} option when using @option{-mrecip} (or at
16828 least @option{-funsafe-math-optimizations},
16829 @option{-finite-math-only}, @option{-freciprocal-math} and
16830 @option{-fno-trapping-math}). Note that while the throughput of the
16831 sequence is generally higher than the throughput of the non-reciprocal
16832 instruction, the precision of the sequence can be decreased by up to 2
16833 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16836 @item -mrecip=@var{opt}
16837 @opindex mrecip=opt
16838 This option allows to control which reciprocal estimate instructions
16839 may be used. @var{opt} is a comma separated list of options, that may
16840 be preceded by a @code{!} to invert the option:
16841 @code{all}: enable all estimate instructions,
16842 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16843 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16844 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16845 @code{divf}: enable the single precision reciprocal approximation instructions;
16846 @code{divd}: enable the double precision reciprocal approximation instructions;
16847 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16848 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16849 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16851 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16852 all of the reciprocal estimate instructions, except for the
16853 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16854 which handle the double precision reciprocal square root calculations.
16856 @item -mrecip-precision
16857 @itemx -mno-recip-precision
16858 @opindex mrecip-precision
16859 Assume (do not assume) that the reciprocal estimate instructions
16860 provide higher precision estimates than is mandated by the powerpc
16861 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16862 automatically selects @option{-mrecip-precision}. The double
16863 precision square root estimate instructions are not generated by
16864 default on low precision machines, since they do not provide an
16865 estimate that converges after three steps.
16867 @item -mveclibabi=@var{type}
16868 @opindex mveclibabi
16869 Specifies the ABI type to use for vectorizing intrinsics using an
16870 external library. The only type supported at present is @code{mass},
16871 which specifies to use IBM's Mathematical Acceleration Subsystem
16872 (MASS) libraries for vectorizing intrinsics using external libraries.
16873 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16874 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16875 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16876 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16877 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16878 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16879 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16880 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16881 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16882 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16883 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16884 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16885 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16886 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16887 for power7. Both @option{-ftree-vectorize} and
16888 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16889 libraries will have to be specified at link time.
16894 Generate (do not generate) the @code{friz} instruction when the
16895 @option{-funsafe-math-optimizations} option is used to optimize
16896 rounding a floating point value to 64-bit integer and back to floating
16897 point. The @code{friz} instruction does not return the same value if
16898 the floating point number is too large to fit in an integer.
16900 @item -mpointers-to-nested-functions
16901 @itemx -mno-pointers-to-nested-functions
16902 @opindex mpointers-to-nested-functions
16903 Generate (do not generate) code to load up the static chain register
16904 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16905 systems where a function pointer points to a 3 word descriptor giving
16906 the function address, TOC value to be loaded in register @var{r2}, and
16907 static chain value to be loaded in register @var{r11}. The
16908 @option{-mpointers-to-nested-functions} is on by default. You will
16909 not be able to call through pointers to nested functions or pointers
16910 to functions compiled in other languages that use the static chain if
16911 you use the @option{-mno-pointers-to-nested-functions}.
16913 @item -msave-toc-indirect
16914 @itemx -mno-save-toc-indirect
16915 @opindex msave-toc-indirect
16916 Generate (do not generate) code to save the TOC value in the reserved
16917 stack location in the function prologue if the function calls through
16918 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
16919 saved in the prologue, it is saved just before the call through the
16920 pointer. The @option{-mno-save-toc-indirect} option is the default.
16924 @subsection RX Options
16927 These command line options are defined for RX targets:
16930 @item -m64bit-doubles
16931 @itemx -m32bit-doubles
16932 @opindex m64bit-doubles
16933 @opindex m32bit-doubles
16934 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16935 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16936 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16937 works on 32-bit values, which is why the default is
16938 @option{-m32bit-doubles}.
16944 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16945 floating point hardware. The default is enabled for the @var{RX600}
16946 series and disabled for the @var{RX200} series.
16948 Floating point instructions will only be generated for 32-bit floating
16949 point values however, so if the @option{-m64bit-doubles} option is in
16950 use then the FPU hardware will not be used for doubles.
16952 @emph{Note} If the @option{-fpu} option is enabled then
16953 @option{-funsafe-math-optimizations} is also enabled automatically.
16954 This is because the RX FPU instructions are themselves unsafe.
16956 @item -mcpu=@var{name}
16958 Selects the type of RX CPU to be targeted. Currently three types are
16959 supported, the generic @var{RX600} and @var{RX200} series hardware and
16960 the specific @var{RX610} CPU. The default is @var{RX600}.
16962 The only difference between @var{RX600} and @var{RX610} is that the
16963 @var{RX610} does not support the @code{MVTIPL} instruction.
16965 The @var{RX200} series does not have a hardware floating point unit
16966 and so @option{-nofpu} is enabled by default when this type is
16969 @item -mbig-endian-data
16970 @itemx -mlittle-endian-data
16971 @opindex mbig-endian-data
16972 @opindex mlittle-endian-data
16973 Store data (but not code) in the big-endian format. The default is
16974 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16977 @item -msmall-data-limit=@var{N}
16978 @opindex msmall-data-limit
16979 Specifies the maximum size in bytes of global and static variables
16980 which can be placed into the small data area. Using the small data
16981 area can lead to smaller and faster code, but the size of area is
16982 limited and it is up to the programmer to ensure that the area does
16983 not overflow. Also when the small data area is used one of the RX's
16984 registers (usually @code{r13}) is reserved for use pointing to this
16985 area, so it is no longer available for use by the compiler. This
16986 could result in slower and/or larger code if variables which once
16987 could have been held in the reserved register are now pushed onto the
16990 Note, common variables (variables which have not been initialised) and
16991 constants are not placed into the small data area as they are assigned
16992 to other sections in the output executable.
16994 The default value is zero, which disables this feature. Note, this
16995 feature is not enabled by default with higher optimization levels
16996 (@option{-O2} etc) because of the potentially detrimental effects of
16997 reserving a register. It is up to the programmer to experiment and
16998 discover whether this feature is of benefit to their program. See the
16999 description of the @option{-mpid} option for a description of how the
17000 actual register to hold the small data area pointer is chosen.
17006 Use the simulator runtime. The default is to use the libgloss board
17009 @item -mas100-syntax
17010 @itemx -mno-as100-syntax
17011 @opindex mas100-syntax
17012 @opindex mno-as100-syntax
17013 When generating assembler output use a syntax that is compatible with
17014 Renesas's AS100 assembler. This syntax can also be handled by the GAS
17015 assembler but it has some restrictions so generating it is not the
17018 @item -mmax-constant-size=@var{N}
17019 @opindex mmax-constant-size
17020 Specifies the maximum size, in bytes, of a constant that can be used as
17021 an operand in a RX instruction. Although the RX instruction set does
17022 allow constants of up to 4 bytes in length to be used in instructions,
17023 a longer value equates to a longer instruction. Thus in some
17024 circumstances it can be beneficial to restrict the size of constants
17025 that are used in instructions. Constants that are too big are instead
17026 placed into a constant pool and referenced via register indirection.
17028 The value @var{N} can be between 0 and 4. A value of 0 (the default)
17029 or 4 means that constants of any size are allowed.
17033 Enable linker relaxation. Linker relaxation is a process whereby the
17034 linker will attempt to reduce the size of a program by finding shorter
17035 versions of various instructions. Disabled by default.
17037 @item -mint-register=@var{N}
17038 @opindex mint-register
17039 Specify the number of registers to reserve for fast interrupt handler
17040 functions. The value @var{N} can be between 0 and 4. A value of 1
17041 means that register @code{r13} will be reserved for the exclusive use
17042 of fast interrupt handlers. A value of 2 reserves @code{r13} and
17043 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
17044 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
17045 A value of 0, the default, does not reserve any registers.
17047 @item -msave-acc-in-interrupts
17048 @opindex msave-acc-in-interrupts
17049 Specifies that interrupt handler functions should preserve the
17050 accumulator register. This is only necessary if normal code might use
17051 the accumulator register, for example because it performs 64-bit
17052 multiplications. The default is to ignore the accumulator as this
17053 makes the interrupt handlers faster.
17059 Enables the generation of position independent data. When enabled any
17060 access to constant data will done via an offset from a base address
17061 held in a register. This allows the location of constant data to be
17062 determined at run-time without requiring the executable to be
17063 relocated, which is a benefit to embedded applications with tight
17064 memory constraints. Data that can be modified is not affected by this
17067 Note, using this feature reserves a register, usually @code{r13}, for
17068 the constant data base address. This can result in slower and/or
17069 larger code, especially in complicated functions.
17071 The actual register chosen to hold the constant data base address
17072 depends upon whether the @option{-msmall-data-limit} and/or the
17073 @option{-mint-register} command line options are enabled. Starting
17074 with register @code{r13} and proceeding downwards, registers are
17075 allocated first to satisfy the requirements of @option{-mint-register},
17076 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
17077 is possible for the small data area register to be @code{r8} if both
17078 @option{-mint-register=4} and @option{-mpid} are specified on the
17081 By default this feature is not enabled. The default can be restored
17082 via the @option{-mno-pid} command line option.
17086 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
17087 has special significance to the RX port when used with the
17088 @code{interrupt} function attribute. This attribute indicates a
17089 function intended to process fast interrupts. GCC will will ensure
17090 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
17091 and/or @code{r13} and only provided that the normal use of the
17092 corresponding registers have been restricted via the
17093 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
17096 @node S/390 and zSeries Options
17097 @subsection S/390 and zSeries Options
17098 @cindex S/390 and zSeries Options
17100 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
17104 @itemx -msoft-float
17105 @opindex mhard-float
17106 @opindex msoft-float
17107 Use (do not use) the hardware floating-point instructions and registers
17108 for floating-point operations. When @option{-msoft-float} is specified,
17109 functions in @file{libgcc.a} will be used to perform floating-point
17110 operations. When @option{-mhard-float} is specified, the compiler
17111 generates IEEE floating-point instructions. This is the default.
17114 @itemx -mno-hard-dfp
17116 @opindex mno-hard-dfp
17117 Use (do not use) the hardware decimal-floating-point instructions for
17118 decimal-floating-point operations. When @option{-mno-hard-dfp} is
17119 specified, functions in @file{libgcc.a} will be used to perform
17120 decimal-floating-point operations. When @option{-mhard-dfp} is
17121 specified, the compiler generates decimal-floating-point hardware
17122 instructions. This is the default for @option{-march=z9-ec} or higher.
17124 @item -mlong-double-64
17125 @itemx -mlong-double-128
17126 @opindex mlong-double-64
17127 @opindex mlong-double-128
17128 These switches control the size of @code{long double} type. A size
17129 of 64bit makes the @code{long double} type equivalent to the @code{double}
17130 type. This is the default.
17133 @itemx -mno-backchain
17134 @opindex mbackchain
17135 @opindex mno-backchain
17136 Store (do not store) the address of the caller's frame as backchain pointer
17137 into the callee's stack frame.
17138 A backchain may be needed to allow debugging using tools that do not understand
17139 DWARF-2 call frame information.
17140 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
17141 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
17142 the backchain is placed into the topmost word of the 96/160 byte register
17145 In general, code compiled with @option{-mbackchain} is call-compatible with
17146 code compiled with @option{-mmo-backchain}; however, use of the backchain
17147 for debugging purposes usually requires that the whole binary is built with
17148 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
17149 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17150 to build a linux kernel use @option{-msoft-float}.
17152 The default is to not maintain the backchain.
17154 @item -mpacked-stack
17155 @itemx -mno-packed-stack
17156 @opindex mpacked-stack
17157 @opindex mno-packed-stack
17158 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
17159 specified, the compiler uses the all fields of the 96/160 byte register save
17160 area only for their default purpose; unused fields still take up stack space.
17161 When @option{-mpacked-stack} is specified, register save slots are densely
17162 packed at the top of the register save area; unused space is reused for other
17163 purposes, allowing for more efficient use of the available stack space.
17164 However, when @option{-mbackchain} is also in effect, the topmost word of
17165 the save area is always used to store the backchain, and the return address
17166 register is always saved two words below the backchain.
17168 As long as the stack frame backchain is not used, code generated with
17169 @option{-mpacked-stack} is call-compatible with code generated with
17170 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
17171 S/390 or zSeries generated code that uses the stack frame backchain at run
17172 time, not just for debugging purposes. Such code is not call-compatible
17173 with code compiled with @option{-mpacked-stack}. Also, note that the
17174 combination of @option{-mbackchain},
17175 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17176 to build a linux kernel use @option{-msoft-float}.
17178 The default is to not use the packed stack layout.
17181 @itemx -mno-small-exec
17182 @opindex msmall-exec
17183 @opindex mno-small-exec
17184 Generate (or do not generate) code using the @code{bras} instruction
17185 to do subroutine calls.
17186 This only works reliably if the total executable size does not
17187 exceed 64k. The default is to use the @code{basr} instruction instead,
17188 which does not have this limitation.
17194 When @option{-m31} is specified, generate code compliant to the
17195 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
17196 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
17197 particular to generate 64-bit instructions. For the @samp{s390}
17198 targets, the default is @option{-m31}, while the @samp{s390x}
17199 targets default to @option{-m64}.
17205 When @option{-mzarch} is specified, generate code using the
17206 instructions available on z/Architecture.
17207 When @option{-mesa} is specified, generate code using the
17208 instructions available on ESA/390. Note that @option{-mesa} is
17209 not possible with @option{-m64}.
17210 When generating code compliant to the GNU/Linux for S/390 ABI,
17211 the default is @option{-mesa}. When generating code compliant
17212 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
17218 Generate (or do not generate) code using the @code{mvcle} instruction
17219 to perform block moves. When @option{-mno-mvcle} is specified,
17220 use a @code{mvc} loop instead. This is the default unless optimizing for
17227 Print (or do not print) additional debug information when compiling.
17228 The default is to not print debug information.
17230 @item -march=@var{cpu-type}
17232 Generate code that will run on @var{cpu-type}, which is the name of a system
17233 representing a certain processor type. Possible values for
17234 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
17235 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
17236 When generating code using the instructions available on z/Architecture,
17237 the default is @option{-march=z900}. Otherwise, the default is
17238 @option{-march=g5}.
17240 @item -mtune=@var{cpu-type}
17242 Tune to @var{cpu-type} everything applicable about the generated code,
17243 except for the ABI and the set of available instructions.
17244 The list of @var{cpu-type} values is the same as for @option{-march}.
17245 The default is the value used for @option{-march}.
17248 @itemx -mno-tpf-trace
17249 @opindex mtpf-trace
17250 @opindex mno-tpf-trace
17251 Generate code that adds (does not add) in TPF OS specific branches to trace
17252 routines in the operating system. This option is off by default, even
17253 when compiling for the TPF OS@.
17256 @itemx -mno-fused-madd
17257 @opindex mfused-madd
17258 @opindex mno-fused-madd
17259 Generate code that uses (does not use) the floating point multiply and
17260 accumulate instructions. These instructions are generated by default if
17261 hardware floating point is used.
17263 @item -mwarn-framesize=@var{framesize}
17264 @opindex mwarn-framesize
17265 Emit a warning if the current function exceeds the given frame size. Because
17266 this is a compile time check it doesn't need to be a real problem when the program
17267 runs. It is intended to identify functions which most probably cause
17268 a stack overflow. It is useful to be used in an environment with limited stack
17269 size e.g.@: the linux kernel.
17271 @item -mwarn-dynamicstack
17272 @opindex mwarn-dynamicstack
17273 Emit a warning if the function calls alloca or uses dynamically
17274 sized arrays. This is generally a bad idea with a limited stack size.
17276 @item -mstack-guard=@var{stack-guard}
17277 @itemx -mstack-size=@var{stack-size}
17278 @opindex mstack-guard
17279 @opindex mstack-size
17280 If these options are provided the s390 back end emits additional instructions in
17281 the function prologue which trigger a trap if the stack size is @var{stack-guard}
17282 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
17283 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
17284 the frame size of the compiled function is chosen.
17285 These options are intended to be used to help debugging stack overflow problems.
17286 The additionally emitted code causes only little overhead and hence can also be
17287 used in production like systems without greater performance degradation. The given
17288 values have to be exact powers of 2 and @var{stack-size} has to be greater than
17289 @var{stack-guard} without exceeding 64k.
17290 In order to be efficient the extra code makes the assumption that the stack starts
17291 at an address aligned to the value given by @var{stack-size}.
17292 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
17295 @node Score Options
17296 @subsection Score Options
17297 @cindex Score Options
17299 These options are defined for Score implementations:
17304 Compile code for big endian mode. This is the default.
17308 Compile code for little endian mode.
17312 Disable generate bcnz instruction.
17316 Enable generate unaligned load and store instruction.
17320 Enable the use of multiply-accumulate instructions. Disabled by default.
17324 Specify the SCORE5 as the target architecture.
17328 Specify the SCORE5U of the target architecture.
17332 Specify the SCORE7 as the target architecture. This is the default.
17336 Specify the SCORE7D as the target architecture.
17340 @subsection SH Options
17342 These @samp{-m} options are defined for the SH implementations:
17347 Generate code for the SH1.
17351 Generate code for the SH2.
17354 Generate code for the SH2e.
17358 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
17359 that the floating-point unit is not used.
17361 @item -m2a-single-only
17362 @opindex m2a-single-only
17363 Generate code for the SH2a-FPU, in such a way that no double-precision
17364 floating point operations are used.
17367 @opindex m2a-single
17368 Generate code for the SH2a-FPU assuming the floating-point unit is in
17369 single-precision mode by default.
17373 Generate code for the SH2a-FPU assuming the floating-point unit is in
17374 double-precision mode by default.
17378 Generate code for the SH3.
17382 Generate code for the SH3e.
17386 Generate code for the SH4 without a floating-point unit.
17388 @item -m4-single-only
17389 @opindex m4-single-only
17390 Generate code for the SH4 with a floating-point unit that only
17391 supports single-precision arithmetic.
17395 Generate code for the SH4 assuming the floating-point unit is in
17396 single-precision mode by default.
17400 Generate code for the SH4.
17404 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
17405 floating-point unit is not used.
17407 @item -m4a-single-only
17408 @opindex m4a-single-only
17409 Generate code for the SH4a, in such a way that no double-precision
17410 floating point operations are used.
17413 @opindex m4a-single
17414 Generate code for the SH4a assuming the floating-point unit is in
17415 single-precision mode by default.
17419 Generate code for the SH4a.
17423 Same as @option{-m4a-nofpu}, except that it implicitly passes
17424 @option{-dsp} to the assembler. GCC doesn't generate any DSP
17425 instructions at the moment.
17429 Compile code for the processor in big endian mode.
17433 Compile code for the processor in little endian mode.
17437 Align doubles at 64-bit boundaries. Note that this changes the calling
17438 conventions, and thus some functions from the standard C library will
17439 not work unless you recompile it first with @option{-mdalign}.
17443 Shorten some address references at link time, when possible; uses the
17444 linker option @option{-relax}.
17448 Use 32-bit offsets in @code{switch} tables. The default is to use
17453 Enable the use of bit manipulation instructions on SH2A.
17457 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
17458 alignment constraints.
17462 Comply with the calling conventions defined by Renesas.
17466 Comply with the calling conventions defined by Renesas.
17470 Comply with the calling conventions defined for GCC before the Renesas
17471 conventions were available. This option is the default for all
17472 targets of the SH toolchain.
17475 @opindex mnomacsave
17476 Mark the @code{MAC} register as call-clobbered, even if
17477 @option{-mhitachi} is given.
17481 Increase IEEE-compliance of floating-point code.
17482 At the moment, this is equivalent to @option{-fno-finite-math-only}.
17483 When generating 16 bit SH opcodes, getting IEEE-conforming results for
17484 comparisons of NANs / infinities incurs extra overhead in every
17485 floating point comparison, therefore the default is set to
17486 @option{-ffinite-math-only}.
17488 @item -minline-ic_invalidate
17489 @opindex minline-ic_invalidate
17490 Inline code to invalidate instruction cache entries after setting up
17491 nested function trampolines.
17492 This option has no effect if -musermode is in effect and the selected
17493 code generation option (e.g. -m4) does not allow the use of the icbi
17495 If the selected code generation option does not allow the use of the icbi
17496 instruction, and -musermode is not in effect, the inlined code will
17497 manipulate the instruction cache address array directly with an associative
17498 write. This not only requires privileged mode, but it will also
17499 fail if the cache line had been mapped via the TLB and has become unmapped.
17503 Dump instruction size and location in the assembly code.
17506 @opindex mpadstruct
17507 This option is deprecated. It pads structures to multiple of 4 bytes,
17508 which is incompatible with the SH ABI@.
17512 Optimize for space instead of speed. Implied by @option{-Os}.
17515 @opindex mprefergot
17516 When generating position-independent code, emit function calls using
17517 the Global Offset Table instead of the Procedure Linkage Table.
17521 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17522 if the inlined code would not work in user mode.
17523 This is the default when the target is @code{sh-*-linux*}.
17525 @item -multcost=@var{number}
17526 @opindex multcost=@var{number}
17527 Set the cost to assume for a multiply insn.
17529 @item -mdiv=@var{strategy}
17530 @opindex mdiv=@var{strategy}
17531 Set the division strategy to use for SHmedia code. @var{strategy} must be
17532 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17533 inv:call2, inv:fp .
17534 "fp" performs the operation in floating point. This has a very high latency,
17535 but needs only a few instructions, so it might be a good choice if
17536 your code has enough easily exploitable ILP to allow the compiler to
17537 schedule the floating point instructions together with other instructions.
17538 Division by zero causes a floating point exception.
17539 "inv" uses integer operations to calculate the inverse of the divisor,
17540 and then multiplies the dividend with the inverse. This strategy allows
17541 cse and hoisting of the inverse calculation. Division by zero calculates
17542 an unspecified result, but does not trap.
17543 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17544 have been found, or if the entire operation has been hoisted to the same
17545 place, the last stages of the inverse calculation are intertwined with the
17546 final multiply to reduce the overall latency, at the expense of using a few
17547 more instructions, and thus offering fewer scheduling opportunities with
17549 "call" calls a library function that usually implements the inv:minlat
17551 This gives high code density for m5-*media-nofpu compilations.
17552 "call2" uses a different entry point of the same library function, where it
17553 assumes that a pointer to a lookup table has already been set up, which
17554 exposes the pointer load to cse / code hoisting optimizations.
17555 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17556 code generation, but if the code stays unoptimized, revert to the "call",
17557 "call2", or "fp" strategies, respectively. Note that the
17558 potentially-trapping side effect of division by zero is carried by a
17559 separate instruction, so it is possible that all the integer instructions
17560 are hoisted out, but the marker for the side effect stays where it is.
17561 A recombination to fp operations or a call is not possible in that case.
17562 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17563 that the inverse calculation was nor separated from the multiply, they speed
17564 up division where the dividend fits into 20 bits (plus sign where applicable),
17565 by inserting a test to skip a number of operations in this case; this test
17566 slows down the case of larger dividends. inv20u assumes the case of a such
17567 a small dividend to be unlikely, and inv20l assumes it to be likely.
17569 @item -maccumulate-outgoing-args
17570 @opindex maccumulate-outgoing-args
17571 Reserve space once for outgoing arguments in the function prologue rather
17572 than around each call. Generally beneficial for performance and size. Also
17573 needed for unwinding to avoid changing the stack frame around conditional code.
17575 @item -mdivsi3_libfunc=@var{name}
17576 @opindex mdivsi3_libfunc=@var{name}
17577 Set the name of the library function used for 32 bit signed division to
17578 @var{name}. This only affect the name used in the call and inv:call
17579 division strategies, and the compiler will still expect the same
17580 sets of input/output/clobbered registers as if this option was not present.
17582 @item -mfixed-range=@var{register-range}
17583 @opindex mfixed-range
17584 Generate code treating the given register range as fixed registers.
17585 A fixed register is one that the register allocator can not use. This is
17586 useful when compiling kernel code. A register range is specified as
17587 two registers separated by a dash. Multiple register ranges can be
17588 specified separated by a comma.
17590 @item -madjust-unroll
17591 @opindex madjust-unroll
17592 Throttle unrolling to avoid thrashing target registers.
17593 This option only has an effect if the gcc code base supports the
17594 TARGET_ADJUST_UNROLL_MAX target hook.
17596 @item -mindexed-addressing
17597 @opindex mindexed-addressing
17598 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17599 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17600 semantics for the indexed addressing mode. The architecture allows the
17601 implementation of processors with 64 bit MMU, which the OS could use to
17602 get 32 bit addressing, but since no current hardware implementation supports
17603 this or any other way to make the indexed addressing mode safe to use in
17604 the 32 bit ABI, the default is -mno-indexed-addressing.
17606 @item -mgettrcost=@var{number}
17607 @opindex mgettrcost=@var{number}
17608 Set the cost assumed for the gettr instruction to @var{number}.
17609 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17613 Assume pt* instructions won't trap. This will generally generate better
17614 scheduled code, but is unsafe on current hardware. The current architecture
17615 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17616 This has the unintentional effect of making it unsafe to schedule ptabs /
17617 ptrel before a branch, or hoist it out of a loop. For example,
17618 __do_global_ctors, a part of libgcc that runs constructors at program
17619 startup, calls functions in a list which is delimited by @minus{}1. With the
17620 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17621 That means that all the constructors will be run a bit quicker, but when
17622 the loop comes to the end of the list, the program crashes because ptabs
17623 loads @minus{}1 into a target register. Since this option is unsafe for any
17624 hardware implementing the current architecture specification, the default
17625 is -mno-pt-fixed. Unless the user specifies a specific cost with
17626 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17627 this deters register allocation using target registers for storing
17630 @item -minvalid-symbols
17631 @opindex minvalid-symbols
17632 Assume symbols might be invalid. Ordinary function symbols generated by
17633 the compiler will always be valid to load with movi/shori/ptabs or
17634 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17635 to generate symbols that will cause ptabs / ptrel to trap.
17636 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17637 It will then prevent cross-basic-block cse, hoisting and most scheduling
17638 of symbol loads. The default is @option{-mno-invalid-symbols}.
17641 @node Solaris 2 Options
17642 @subsection Solaris 2 Options
17643 @cindex Solaris 2 options
17645 These @samp{-m} options are supported on Solaris 2:
17648 @item -mimpure-text
17649 @opindex mimpure-text
17650 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17651 the compiler to not pass @option{-z text} to the linker when linking a
17652 shared object. Using this option, you can link position-dependent
17653 code into a shared object.
17655 @option{-mimpure-text} suppresses the ``relocations remain against
17656 allocatable but non-writable sections'' linker error message.
17657 However, the necessary relocations will trigger copy-on-write, and the
17658 shared object is not actually shared across processes. Instead of
17659 using @option{-mimpure-text}, you should compile all source code with
17660 @option{-fpic} or @option{-fPIC}.
17664 These switches are supported in addition to the above on Solaris 2:
17669 Add support for multithreading using the POSIX threads library. This
17670 option sets flags for both the preprocessor and linker. This option does
17671 not affect the thread safety of object code produced by the compiler or
17672 that of libraries supplied with it.
17676 This is a synonym for @option{-pthreads}.
17679 @node SPARC Options
17680 @subsection SPARC Options
17681 @cindex SPARC options
17683 These @samp{-m} options are supported on the SPARC:
17686 @item -mno-app-regs
17688 @opindex mno-app-regs
17690 Specify @option{-mapp-regs} to generate output using the global registers
17691 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17694 To be fully SVR4 ABI compliant at the cost of some performance loss,
17695 specify @option{-mno-app-regs}. You should compile libraries and system
17696 software with this option.
17702 With @option{-mflat}, the compiler does not generate save/restore instructions
17703 and uses a ``flat'' or single register window model. This model is compatible
17704 with the regular register window model. The local registers and the input
17705 registers (0--5) are still treated as ``call-saved'' registers and will be
17706 saved on the stack as needed.
17708 With @option{-mno-flat} (the default), the compiler generates save/restore
17709 instructions (except for leaf functions). This is the normal operating mode.
17712 @itemx -mhard-float
17714 @opindex mhard-float
17715 Generate output containing floating point instructions. This is the
17719 @itemx -msoft-float
17721 @opindex msoft-float
17722 Generate output containing library calls for floating point.
17723 @strong{Warning:} the requisite libraries are not available for all SPARC
17724 targets. Normally the facilities of the machine's usual C compiler are
17725 used, but this cannot be done directly in cross-compilation. You must make
17726 your own arrangements to provide suitable library functions for
17727 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17728 @samp{sparclite-*-*} do provide software floating point support.
17730 @option{-msoft-float} changes the calling convention in the output file;
17731 therefore, it is only useful if you compile @emph{all} of a program with
17732 this option. In particular, you need to compile @file{libgcc.a}, the
17733 library that comes with GCC, with @option{-msoft-float} in order for
17736 @item -mhard-quad-float
17737 @opindex mhard-quad-float
17738 Generate output containing quad-word (long double) floating point
17741 @item -msoft-quad-float
17742 @opindex msoft-quad-float
17743 Generate output containing library calls for quad-word (long double)
17744 floating point instructions. The functions called are those specified
17745 in the SPARC ABI@. This is the default.
17747 As of this writing, there are no SPARC implementations that have hardware
17748 support for the quad-word floating point instructions. They all invoke
17749 a trap handler for one of these instructions, and then the trap handler
17750 emulates the effect of the instruction. Because of the trap handler overhead,
17751 this is much slower than calling the ABI library routines. Thus the
17752 @option{-msoft-quad-float} option is the default.
17754 @item -mno-unaligned-doubles
17755 @itemx -munaligned-doubles
17756 @opindex mno-unaligned-doubles
17757 @opindex munaligned-doubles
17758 Assume that doubles have 8 byte alignment. This is the default.
17760 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17761 alignment only if they are contained in another type, or if they have an
17762 absolute address. Otherwise, it assumes they have 4 byte alignment.
17763 Specifying this option avoids some rare compatibility problems with code
17764 generated by other compilers. It is not the default because it results
17765 in a performance loss, especially for floating point code.
17767 @item -mno-faster-structs
17768 @itemx -mfaster-structs
17769 @opindex mno-faster-structs
17770 @opindex mfaster-structs
17771 With @option{-mfaster-structs}, the compiler assumes that structures
17772 should have 8 byte alignment. This enables the use of pairs of
17773 @code{ldd} and @code{std} instructions for copies in structure
17774 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17775 However, the use of this changed alignment directly violates the SPARC
17776 ABI@. Thus, it's intended only for use on targets where the developer
17777 acknowledges that their resulting code will not be directly in line with
17778 the rules of the ABI@.
17780 @item -mcpu=@var{cpu_type}
17782 Set the instruction set, register set, and instruction scheduling parameters
17783 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17784 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17785 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17786 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17787 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
17788 and @samp{niagara4}.
17790 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
17791 which selects the best architecture option for the host processor.
17792 @option{-mcpu=native} has no effect if GCC does not recognize
17795 Default instruction scheduling parameters are used for values that select
17796 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17797 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17799 Here is a list of each supported architecture and their supported
17804 v8: supersparc, hypersparc, leon
17805 sparclite: f930, f934, sparclite86x
17807 v9: ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
17810 By default (unless configured otherwise), GCC generates code for the V7
17811 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17812 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17813 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17814 SPARCStation 1, 2, IPX etc.
17816 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17817 architecture. The only difference from V7 code is that the compiler emits
17818 the integer multiply and integer divide instructions which exist in SPARC-V8
17819 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17820 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17823 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17824 the SPARC architecture. This adds the integer multiply, integer divide step
17825 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17826 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17827 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17828 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17829 MB86934 chip, which is the more recent SPARClite with FPU@.
17831 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17832 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17833 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17834 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17835 optimizes it for the TEMIC SPARClet chip.
17837 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17838 architecture. This adds 64-bit integer and floating-point move instructions,
17839 3 additional floating-point condition code registers and conditional move
17840 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17841 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17842 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17843 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17844 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17845 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17846 additionally optimizes it for Sun UltraSPARC T2 chips. With
17847 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
17848 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
17849 additionally optimizes it for Sun UltraSPARC T4 chips.
17851 @item -mtune=@var{cpu_type}
17853 Set the instruction scheduling parameters for machine type
17854 @var{cpu_type}, but do not set the instruction set or register set that the
17855 option @option{-mcpu=@var{cpu_type}} would.
17857 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17858 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17859 that select a particular CPU implementation. Those are @samp{cypress},
17860 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17861 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17862 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
17863 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
17868 @opindex mno-v8plus
17869 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17870 difference from the V8 ABI is that the global and out registers are
17871 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17872 mode for all SPARC-V9 processors.
17878 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17879 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17885 With @option{-mvis2}, GCC generates code that takes advantage of
17886 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
17887 default is @option{-mvis2} when targetting a cpu that supports such
17888 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
17889 also sets @option{-mvis}.
17895 With @option{-mvis3}, GCC generates code that takes advantage of
17896 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
17897 default is @option{-mvis3} when targetting a cpu that supports such
17898 instructions, such as niagara-3 and later. Setting @option{-mvis3}
17899 also sets @option{-mvis2} and @option{-mvis}.
17905 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
17906 population count instruction. The default is @option{-mpopc}
17907 when targetting a cpu that supports such instructions, such as Niagara-2 and
17914 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
17915 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
17916 when targetting a cpu that supports such instructions, such as Niagara-3 and
17920 @opindex mfix-at697f
17921 Enable the documented workaround for the single erratum of the Atmel AT697F
17922 processor (which corresponds to erratum #13 of the AT697E processor).
17925 These @samp{-m} options are supported in addition to the above
17926 on SPARC-V9 processors in 64-bit environments:
17929 @item -mlittle-endian
17930 @opindex mlittle-endian
17931 Generate code for a processor running in little-endian mode. It is only
17932 available for a few configurations and most notably not on Solaris and Linux.
17938 Generate code for a 32-bit or 64-bit environment.
17939 The 32-bit environment sets int, long and pointer to 32 bits.
17940 The 64-bit environment sets int to 32 bits and long and pointer
17943 @item -mcmodel=medlow
17944 @opindex mcmodel=medlow
17945 Generate code for the Medium/Low code model: 64-bit addresses, programs
17946 must be linked in the low 32 bits of memory. Programs can be statically
17947 or dynamically linked.
17949 @item -mcmodel=medmid
17950 @opindex mcmodel=medmid
17951 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17952 must be linked in the low 44 bits of memory, the text and data segments must
17953 be less than 2GB in size and the data segment must be located within 2GB of
17956 @item -mcmodel=medany
17957 @opindex mcmodel=medany
17958 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17959 may be linked anywhere in memory, the text and data segments must be less
17960 than 2GB in size and the data segment must be located within 2GB of the
17963 @item -mcmodel=embmedany
17964 @opindex mcmodel=embmedany
17965 Generate code for the Medium/Anywhere code model for embedded systems:
17966 64-bit addresses, the text and data segments must be less than 2GB in
17967 size, both starting anywhere in memory (determined at link time). The
17968 global register %g4 points to the base of the data segment. Programs
17969 are statically linked and PIC is not supported.
17972 @itemx -mno-stack-bias
17973 @opindex mstack-bias
17974 @opindex mno-stack-bias
17975 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17976 frame pointer if present, are offset by @minus{}2047 which must be added back
17977 when making stack frame references. This is the default in 64-bit mode.
17978 Otherwise, assume no such offset is present.
17982 @subsection SPU Options
17983 @cindex SPU options
17985 These @samp{-m} options are supported on the SPU:
17989 @itemx -merror-reloc
17990 @opindex mwarn-reloc
17991 @opindex merror-reloc
17993 The loader for SPU does not handle dynamic relocations. By default, GCC
17994 will give an error when it generates code that requires a dynamic
17995 relocation. @option{-mno-error-reloc} disables the error,
17996 @option{-mwarn-reloc} will generate a warning instead.
17999 @itemx -munsafe-dma
18001 @opindex munsafe-dma
18003 Instructions which initiate or test completion of DMA must not be
18004 reordered with respect to loads and stores of the memory which is being
18005 accessed. Users typically address this problem using the volatile
18006 keyword, but that can lead to inefficient code in places where the
18007 memory is known to not change. Rather than mark the memory as volatile
18008 we treat the DMA instructions as potentially effecting all memory. With
18009 @option{-munsafe-dma} users must use the volatile keyword to protect
18012 @item -mbranch-hints
18013 @opindex mbranch-hints
18015 By default, GCC will generate a branch hint instruction to avoid
18016 pipeline stalls for always taken or probably taken branches. A hint
18017 will not be generated closer than 8 instructions away from its branch.
18018 There is little reason to disable them, except for debugging purposes,
18019 or to make an object a little bit smaller.
18023 @opindex msmall-mem
18024 @opindex mlarge-mem
18026 By default, GCC generates code assuming that addresses are never larger
18027 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
18028 a full 32 bit address.
18033 By default, GCC links against startup code that assumes the SPU-style
18034 main function interface (which has an unconventional parameter list).
18035 With @option{-mstdmain}, GCC will link your program against startup
18036 code that assumes a C99-style interface to @code{main}, including a
18037 local copy of @code{argv} strings.
18039 @item -mfixed-range=@var{register-range}
18040 @opindex mfixed-range
18041 Generate code treating the given register range as fixed registers.
18042 A fixed register is one that the register allocator can not use. This is
18043 useful when compiling kernel code. A register range is specified as
18044 two registers separated by a dash. Multiple register ranges can be
18045 specified separated by a comma.
18051 Compile code assuming that pointers to the PPU address space accessed
18052 via the @code{__ea} named address space qualifier are either 32 or 64
18053 bits wide. The default is 32 bits. As this is an ABI changing option,
18054 all object code in an executable must be compiled with the same setting.
18056 @item -maddress-space-conversion
18057 @itemx -mno-address-space-conversion
18058 @opindex maddress-space-conversion
18059 @opindex mno-address-space-conversion
18060 Allow/disallow treating the @code{__ea} address space as superset
18061 of the generic address space. This enables explicit type casts
18062 between @code{__ea} and generic pointer as well as implicit
18063 conversions of generic pointers to @code{__ea} pointers. The
18064 default is to allow address space pointer conversions.
18066 @item -mcache-size=@var{cache-size}
18067 @opindex mcache-size
18068 This option controls the version of libgcc that the compiler links to an
18069 executable and selects a software-managed cache for accessing variables
18070 in the @code{__ea} address space with a particular cache size. Possible
18071 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
18072 and @samp{128}. The default cache size is 64KB.
18074 @item -matomic-updates
18075 @itemx -mno-atomic-updates
18076 @opindex matomic-updates
18077 @opindex mno-atomic-updates
18078 This option controls the version of libgcc that the compiler links to an
18079 executable and selects whether atomic updates to the software-managed
18080 cache of PPU-side variables are used. If you use atomic updates, changes
18081 to a PPU variable from SPU code using the @code{__ea} named address space
18082 qualifier will not interfere with changes to other PPU variables residing
18083 in the same cache line from PPU code. If you do not use atomic updates,
18084 such interference may occur; however, writing back cache lines will be
18085 more efficient. The default behavior is to use atomic updates.
18088 @itemx -mdual-nops=@var{n}
18089 @opindex mdual-nops
18090 By default, GCC will insert nops to increase dual issue when it expects
18091 it to increase performance. @var{n} can be a value from 0 to 10. A
18092 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
18093 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
18095 @item -mhint-max-nops=@var{n}
18096 @opindex mhint-max-nops
18097 Maximum number of nops to insert for a branch hint. A branch hint must
18098 be at least 8 instructions away from the branch it is effecting. GCC
18099 will insert up to @var{n} nops to enforce this, otherwise it will not
18100 generate the branch hint.
18102 @item -mhint-max-distance=@var{n}
18103 @opindex mhint-max-distance
18104 The encoding of the branch hint instruction limits the hint to be within
18105 256 instructions of the branch it is effecting. By default, GCC makes
18106 sure it is within 125.
18109 @opindex msafe-hints
18110 Work around a hardware bug which causes the SPU to stall indefinitely.
18111 By default, GCC will insert the @code{hbrp} instruction to make sure
18112 this stall won't happen.
18116 @node System V Options
18117 @subsection Options for System V
18119 These additional options are available on System V Release 4 for
18120 compatibility with other compilers on those systems:
18125 Create a shared object.
18126 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
18130 Identify the versions of each tool used by the compiler, in a
18131 @code{.ident} assembler directive in the output.
18135 Refrain from adding @code{.ident} directives to the output file (this is
18138 @item -YP,@var{dirs}
18140 Search the directories @var{dirs}, and no others, for libraries
18141 specified with @option{-l}.
18143 @item -Ym,@var{dir}
18145 Look in the directory @var{dir} to find the M4 preprocessor.
18146 The assembler uses this option.
18147 @c This is supposed to go with a -Yd for predefined M4 macro files, but
18148 @c the generic assembler that comes with Solaris takes just -Ym.
18152 @subsection V850 Options
18153 @cindex V850 Options
18155 These @samp{-m} options are defined for V850 implementations:
18159 @itemx -mno-long-calls
18160 @opindex mlong-calls
18161 @opindex mno-long-calls
18162 Treat all calls as being far away (near). If calls are assumed to be
18163 far away, the compiler will always load the functions address up into a
18164 register, and call indirect through the pointer.
18170 Do not optimize (do optimize) basic blocks that use the same index
18171 pointer 4 or more times to copy pointer into the @code{ep} register, and
18172 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
18173 option is on by default if you optimize.
18175 @item -mno-prolog-function
18176 @itemx -mprolog-function
18177 @opindex mno-prolog-function
18178 @opindex mprolog-function
18179 Do not use (do use) external functions to save and restore registers
18180 at the prologue and epilogue of a function. The external functions
18181 are slower, but use less code space if more than one function saves
18182 the same number of registers. The @option{-mprolog-function} option
18183 is on by default if you optimize.
18187 Try to make the code as small as possible. At present, this just turns
18188 on the @option{-mep} and @option{-mprolog-function} options.
18190 @item -mtda=@var{n}
18192 Put static or global variables whose size is @var{n} bytes or less into
18193 the tiny data area that register @code{ep} points to. The tiny data
18194 area can hold up to 256 bytes in total (128 bytes for byte references).
18196 @item -msda=@var{n}
18198 Put static or global variables whose size is @var{n} bytes or less into
18199 the small data area that register @code{gp} points to. The small data
18200 area can hold up to 64 kilobytes.
18202 @item -mzda=@var{n}
18204 Put static or global variables whose size is @var{n} bytes or less into
18205 the first 32 kilobytes of memory.
18209 Specify that the target processor is the V850.
18212 @opindex mbig-switch
18213 Generate code suitable for big switch tables. Use this option only if
18214 the assembler/linker complain about out of range branches within a switch
18219 This option will cause r2 and r5 to be used in the code generated by
18220 the compiler. This setting is the default.
18222 @item -mno-app-regs
18223 @opindex mno-app-regs
18224 This option will cause r2 and r5 to be treated as fixed registers.
18228 Specify that the target processor is the V850E2V3. The preprocessor
18229 constants @samp{__v850e2v3__} will be defined if
18230 this option is used.
18234 Specify that the target processor is the V850E2. The preprocessor
18235 constants @samp{__v850e2__} will be defined if this option is used.
18239 Specify that the target processor is the V850E1. The preprocessor
18240 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
18241 this option is used.
18245 Specify that the target processor is the V850ES. This is an alias for
18246 the @option{-mv850e1} option.
18250 Specify that the target processor is the V850E@. The preprocessor
18251 constant @samp{__v850e__} will be defined if this option is used.
18253 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
18254 nor @option{-mv850e2} nor @option{-mv850e2v3}
18255 are defined then a default target processor will be chosen and the
18256 relevant @samp{__v850*__} preprocessor constant will be defined.
18258 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
18259 defined, regardless of which processor variant is the target.
18261 @item -mdisable-callt
18262 @opindex mdisable-callt
18263 This option will suppress generation of the CALLT instruction for the
18264 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
18265 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
18270 @subsection VAX Options
18271 @cindex VAX options
18273 These @samp{-m} options are defined for the VAX:
18278 Do not output certain jump instructions (@code{aobleq} and so on)
18279 that the Unix assembler for the VAX cannot handle across long
18284 Do output those jump instructions, on the assumption that you
18285 will assemble with the GNU assembler.
18289 Output code for g-format floating point numbers instead of d-format.
18292 @node VxWorks Options
18293 @subsection VxWorks Options
18294 @cindex VxWorks Options
18296 The options in this section are defined for all VxWorks targets.
18297 Options specific to the target hardware are listed with the other
18298 options for that target.
18303 GCC can generate code for both VxWorks kernels and real time processes
18304 (RTPs). This option switches from the former to the latter. It also
18305 defines the preprocessor macro @code{__RTP__}.
18308 @opindex non-static
18309 Link an RTP executable against shared libraries rather than static
18310 libraries. The options @option{-static} and @option{-shared} can
18311 also be used for RTPs (@pxref{Link Options}); @option{-static}
18318 These options are passed down to the linker. They are defined for
18319 compatibility with Diab.
18322 @opindex Xbind-lazy
18323 Enable lazy binding of function calls. This option is equivalent to
18324 @option{-Wl,-z,now} and is defined for compatibility with Diab.
18328 Disable lazy binding of function calls. This option is the default and
18329 is defined for compatibility with Diab.
18332 @node x86-64 Options
18333 @subsection x86-64 Options
18334 @cindex x86-64 options
18336 These are listed under @xref{i386 and x86-64 Options}.
18338 @node Xstormy16 Options
18339 @subsection Xstormy16 Options
18340 @cindex Xstormy16 Options
18342 These options are defined for Xstormy16:
18347 Choose startup files and linker script suitable for the simulator.
18350 @node Xtensa Options
18351 @subsection Xtensa Options
18352 @cindex Xtensa Options
18354 These options are supported for Xtensa targets:
18358 @itemx -mno-const16
18360 @opindex mno-const16
18361 Enable or disable use of @code{CONST16} instructions for loading
18362 constant values. The @code{CONST16} instruction is currently not a
18363 standard option from Tensilica. When enabled, @code{CONST16}
18364 instructions are always used in place of the standard @code{L32R}
18365 instructions. The use of @code{CONST16} is enabled by default only if
18366 the @code{L32R} instruction is not available.
18369 @itemx -mno-fused-madd
18370 @opindex mfused-madd
18371 @opindex mno-fused-madd
18372 Enable or disable use of fused multiply/add and multiply/subtract
18373 instructions in the floating-point option. This has no effect if the
18374 floating-point option is not also enabled. Disabling fused multiply/add
18375 and multiply/subtract instructions forces the compiler to use separate
18376 instructions for the multiply and add/subtract operations. This may be
18377 desirable in some cases where strict IEEE 754-compliant results are
18378 required: the fused multiply add/subtract instructions do not round the
18379 intermediate result, thereby producing results with @emph{more} bits of
18380 precision than specified by the IEEE standard. Disabling fused multiply
18381 add/subtract instructions also ensures that the program output is not
18382 sensitive to the compiler's ability to combine multiply and add/subtract
18385 @item -mserialize-volatile
18386 @itemx -mno-serialize-volatile
18387 @opindex mserialize-volatile
18388 @opindex mno-serialize-volatile
18389 When this option is enabled, GCC inserts @code{MEMW} instructions before
18390 @code{volatile} memory references to guarantee sequential consistency.
18391 The default is @option{-mserialize-volatile}. Use
18392 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
18394 @item -mforce-no-pic
18395 @opindex mforce-no-pic
18396 For targets, like GNU/Linux, where all user-mode Xtensa code must be
18397 position-independent code (PIC), this option disables PIC for compiling
18400 @item -mtext-section-literals
18401 @itemx -mno-text-section-literals
18402 @opindex mtext-section-literals
18403 @opindex mno-text-section-literals
18404 Control the treatment of literal pools. The default is
18405 @option{-mno-text-section-literals}, which places literals in a separate
18406 section in the output file. This allows the literal pool to be placed
18407 in a data RAM/ROM, and it also allows the linker to combine literal
18408 pools from separate object files to remove redundant literals and
18409 improve code size. With @option{-mtext-section-literals}, the literals
18410 are interspersed in the text section in order to keep them as close as
18411 possible to their references. This may be necessary for large assembly
18414 @item -mtarget-align
18415 @itemx -mno-target-align
18416 @opindex mtarget-align
18417 @opindex mno-target-align
18418 When this option is enabled, GCC instructs the assembler to
18419 automatically align instructions to reduce branch penalties at the
18420 expense of some code density. The assembler attempts to widen density
18421 instructions to align branch targets and the instructions following call
18422 instructions. If there are not enough preceding safe density
18423 instructions to align a target, no widening will be performed. The
18424 default is @option{-mtarget-align}. These options do not affect the
18425 treatment of auto-aligned instructions like @code{LOOP}, which the
18426 assembler will always align, either by widening density instructions or
18427 by inserting no-op instructions.
18430 @itemx -mno-longcalls
18431 @opindex mlongcalls
18432 @opindex mno-longcalls
18433 When this option is enabled, GCC instructs the assembler to translate
18434 direct calls to indirect calls unless it can determine that the target
18435 of a direct call is in the range allowed by the call instruction. This
18436 translation typically occurs for calls to functions in other source
18437 files. Specifically, the assembler translates a direct @code{CALL}
18438 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
18439 The default is @option{-mno-longcalls}. This option should be used in
18440 programs where the call target can potentially be out of range. This
18441 option is implemented in the assembler, not the compiler, so the
18442 assembly code generated by GCC will still show direct call
18443 instructions---look at the disassembled object code to see the actual
18444 instructions. Note that the assembler will use an indirect call for
18445 every cross-file call, not just those that really will be out of range.
18448 @node zSeries Options
18449 @subsection zSeries Options
18450 @cindex zSeries options
18452 These are listed under @xref{S/390 and zSeries Options}.
18454 @node Code Gen Options
18455 @section Options for Code Generation Conventions
18456 @cindex code generation conventions
18457 @cindex options, code generation
18458 @cindex run-time options
18460 These machine-independent options control the interface conventions
18461 used in code generation.
18463 Most of them have both positive and negative forms; the negative form
18464 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
18465 one of the forms is listed---the one which is not the default. You
18466 can figure out the other form by either removing @samp{no-} or adding
18470 @item -fbounds-check
18471 @opindex fbounds-check
18472 For front-ends that support it, generate additional code to check that
18473 indices used to access arrays are within the declared range. This is
18474 currently only supported by the Java and Fortran front-ends, where
18475 this option defaults to true and false respectively.
18479 This option generates traps for signed overflow on addition, subtraction,
18480 multiplication operations.
18484 This option instructs the compiler to assume that signed arithmetic
18485 overflow of addition, subtraction and multiplication wraps around
18486 using twos-complement representation. This flag enables some optimizations
18487 and disables others. This option is enabled by default for the Java
18488 front-end, as required by the Java language specification.
18491 @opindex fexceptions
18492 Enable exception handling. Generates extra code needed to propagate
18493 exceptions. For some targets, this implies GCC will generate frame
18494 unwind information for all functions, which can produce significant data
18495 size overhead, although it does not affect execution. If you do not
18496 specify this option, GCC will enable it by default for languages like
18497 C++ which normally require exception handling, and disable it for
18498 languages like C that do not normally require it. However, you may need
18499 to enable this option when compiling C code that needs to interoperate
18500 properly with exception handlers written in C++. You may also wish to
18501 disable this option if you are compiling older C++ programs that don't
18502 use exception handling.
18504 @item -fnon-call-exceptions
18505 @opindex fnon-call-exceptions
18506 Generate code that allows trapping instructions to throw exceptions.
18507 Note that this requires platform-specific runtime support that does
18508 not exist everywhere. Moreover, it only allows @emph{trapping}
18509 instructions to throw exceptions, i.e.@: memory references or floating
18510 point instructions. It does not allow exceptions to be thrown from
18511 arbitrary signal handlers such as @code{SIGALRM}.
18513 @item -funwind-tables
18514 @opindex funwind-tables
18515 Similar to @option{-fexceptions}, except that it will just generate any needed
18516 static data, but will not affect the generated code in any other way.
18517 You will normally not enable this option; instead, a language processor
18518 that needs this handling would enable it on your behalf.
18520 @item -fasynchronous-unwind-tables
18521 @opindex fasynchronous-unwind-tables
18522 Generate unwind table in dwarf2 format, if supported by target machine. The
18523 table is exact at each instruction boundary, so it can be used for stack
18524 unwinding from asynchronous events (such as debugger or garbage collector).
18526 @item -fpcc-struct-return
18527 @opindex fpcc-struct-return
18528 Return ``short'' @code{struct} and @code{union} values in memory like
18529 longer ones, rather than in registers. This convention is less
18530 efficient, but it has the advantage of allowing intercallability between
18531 GCC-compiled files and files compiled with other compilers, particularly
18532 the Portable C Compiler (pcc).
18534 The precise convention for returning structures in memory depends
18535 on the target configuration macros.
18537 Short structures and unions are those whose size and alignment match
18538 that of some integer type.
18540 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
18541 switch is not binary compatible with code compiled with the
18542 @option{-freg-struct-return} switch.
18543 Use it to conform to a non-default application binary interface.
18545 @item -freg-struct-return
18546 @opindex freg-struct-return
18547 Return @code{struct} and @code{union} values in registers when possible.
18548 This is more efficient for small structures than
18549 @option{-fpcc-struct-return}.
18551 If you specify neither @option{-fpcc-struct-return} nor
18552 @option{-freg-struct-return}, GCC defaults to whichever convention is
18553 standard for the target. If there is no standard convention, GCC
18554 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18555 the principal compiler. In those cases, we can choose the standard, and
18556 we chose the more efficient register return alternative.
18558 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18559 switch is not binary compatible with code compiled with the
18560 @option{-fpcc-struct-return} switch.
18561 Use it to conform to a non-default application binary interface.
18563 @item -fshort-enums
18564 @opindex fshort-enums
18565 Allocate to an @code{enum} type only as many bytes as it needs for the
18566 declared range of possible values. Specifically, the @code{enum} type
18567 will be equivalent to the smallest integer type which has enough room.
18569 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18570 code that is not binary compatible with code generated without that switch.
18571 Use it to conform to a non-default application binary interface.
18573 @item -fshort-double
18574 @opindex fshort-double
18575 Use the same size for @code{double} as for @code{float}.
18577 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18578 code that is not binary compatible with code generated without that switch.
18579 Use it to conform to a non-default application binary interface.
18581 @item -fshort-wchar
18582 @opindex fshort-wchar
18583 Override the underlying type for @samp{wchar_t} to be @samp{short
18584 unsigned int} instead of the default for the target. This option is
18585 useful for building programs to run under WINE@.
18587 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18588 code that is not binary compatible with code generated without that switch.
18589 Use it to conform to a non-default application binary interface.
18592 @opindex fno-common
18593 In C code, controls the placement of uninitialized global variables.
18594 Unix C compilers have traditionally permitted multiple definitions of
18595 such variables in different compilation units by placing the variables
18597 This is the behavior specified by @option{-fcommon}, and is the default
18598 for GCC on most targets.
18599 On the other hand, this behavior is not required by ISO C, and on some
18600 targets may carry a speed or code size penalty on variable references.
18601 The @option{-fno-common} option specifies that the compiler should place
18602 uninitialized global variables in the data section of the object file,
18603 rather than generating them as common blocks.
18604 This has the effect that if the same variable is declared
18605 (without @code{extern}) in two different compilations,
18606 you will get a multiple-definition error when you link them.
18607 In this case, you must compile with @option{-fcommon} instead.
18608 Compiling with @option{-fno-common} is useful on targets for which
18609 it provides better performance, or if you wish to verify that the
18610 program will work on other systems which always treat uninitialized
18611 variable declarations this way.
18615 Ignore the @samp{#ident} directive.
18617 @item -finhibit-size-directive
18618 @opindex finhibit-size-directive
18619 Don't output a @code{.size} assembler directive, or anything else that
18620 would cause trouble if the function is split in the middle, and the
18621 two halves are placed at locations far apart in memory. This option is
18622 used when compiling @file{crtstuff.c}; you should not need to use it
18625 @item -fverbose-asm
18626 @opindex fverbose-asm
18627 Put extra commentary information in the generated assembly code to
18628 make it more readable. This option is generally only of use to those
18629 who actually need to read the generated assembly code (perhaps while
18630 debugging the compiler itself).
18632 @option{-fno-verbose-asm}, the default, causes the
18633 extra information to be omitted and is useful when comparing two assembler
18636 @item -frecord-gcc-switches
18637 @opindex frecord-gcc-switches
18638 This switch causes the command line that was used to invoke the
18639 compiler to be recorded into the object file that is being created.
18640 This switch is only implemented on some targets and the exact format
18641 of the recording is target and binary file format dependent, but it
18642 usually takes the form of a section containing ASCII text. This
18643 switch is related to the @option{-fverbose-asm} switch, but that
18644 switch only records information in the assembler output file as
18645 comments, so it never reaches the object file.
18646 See also @option{-grecord-gcc-switches} for another
18647 way of storing compiler options into the object file.
18651 @cindex global offset table
18653 Generate position-independent code (PIC) suitable for use in a shared
18654 library, if supported for the target machine. Such code accesses all
18655 constant addresses through a global offset table (GOT)@. The dynamic
18656 loader resolves the GOT entries when the program starts (the dynamic
18657 loader is not part of GCC; it is part of the operating system). If
18658 the GOT size for the linked executable exceeds a machine-specific
18659 maximum size, you get an error message from the linker indicating that
18660 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18661 instead. (These maximums are 8k on the SPARC and 32k
18662 on the m68k and RS/6000. The 386 has no such limit.)
18664 Position-independent code requires special support, and therefore works
18665 only on certain machines. For the 386, GCC supports PIC for System V
18666 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18667 position-independent.
18669 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18674 If supported for the target machine, emit position-independent code,
18675 suitable for dynamic linking and avoiding any limit on the size of the
18676 global offset table. This option makes a difference on the m68k,
18677 PowerPC and SPARC@.
18679 Position-independent code requires special support, and therefore works
18680 only on certain machines.
18682 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18689 These options are similar to @option{-fpic} and @option{-fPIC}, but
18690 generated position independent code can be only linked into executables.
18691 Usually these options are used when @option{-pie} GCC option will be
18692 used during linking.
18694 @option{-fpie} and @option{-fPIE} both define the macros
18695 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18696 for @option{-fpie} and 2 for @option{-fPIE}.
18698 @item -fno-jump-tables
18699 @opindex fno-jump-tables
18700 Do not use jump tables for switch statements even where it would be
18701 more efficient than other code generation strategies. This option is
18702 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18703 building code which forms part of a dynamic linker and cannot
18704 reference the address of a jump table. On some targets, jump tables
18705 do not require a GOT and this option is not needed.
18707 @item -ffixed-@var{reg}
18709 Treat the register named @var{reg} as a fixed register; generated code
18710 should never refer to it (except perhaps as a stack pointer, frame
18711 pointer or in some other fixed role).
18713 @var{reg} must be the name of a register. The register names accepted
18714 are machine-specific and are defined in the @code{REGISTER_NAMES}
18715 macro in the machine description macro file.
18717 This flag does not have a negative form, because it specifies a
18720 @item -fcall-used-@var{reg}
18721 @opindex fcall-used
18722 Treat the register named @var{reg} as an allocable register that is
18723 clobbered by function calls. It may be allocated for temporaries or
18724 variables that do not live across a call. Functions compiled this way
18725 will not save and restore the register @var{reg}.
18727 It is an error to used this flag with the frame pointer or stack pointer.
18728 Use of this flag for other registers that have fixed pervasive roles in
18729 the machine's execution model will produce disastrous results.
18731 This flag does not have a negative form, because it specifies a
18734 @item -fcall-saved-@var{reg}
18735 @opindex fcall-saved
18736 Treat the register named @var{reg} as an allocable register saved by
18737 functions. It may be allocated even for temporaries or variables that
18738 live across a call. Functions compiled this way will save and restore
18739 the register @var{reg} if they use it.
18741 It is an error to used this flag with the frame pointer or stack pointer.
18742 Use of this flag for other registers that have fixed pervasive roles in
18743 the machine's execution model will produce disastrous results.
18745 A different sort of disaster will result from the use of this flag for
18746 a register in which function values may be returned.
18748 This flag does not have a negative form, because it specifies a
18751 @item -fpack-struct[=@var{n}]
18752 @opindex fpack-struct
18753 Without a value specified, pack all structure members together without
18754 holes. When a value is specified (which must be a small power of two), pack
18755 structure members according to this value, representing the maximum
18756 alignment (that is, objects with default alignment requirements larger than
18757 this will be output potentially unaligned at the next fitting location.
18759 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18760 code that is not binary compatible with code generated without that switch.
18761 Additionally, it makes the code suboptimal.
18762 Use it to conform to a non-default application binary interface.
18764 @item -finstrument-functions
18765 @opindex finstrument-functions
18766 Generate instrumentation calls for entry and exit to functions. Just
18767 after function entry and just before function exit, the following
18768 profiling functions will be called with the address of the current
18769 function and its call site. (On some platforms,
18770 @code{__builtin_return_address} does not work beyond the current
18771 function, so the call site information may not be available to the
18772 profiling functions otherwise.)
18775 void __cyg_profile_func_enter (void *this_fn,
18777 void __cyg_profile_func_exit (void *this_fn,
18781 The first argument is the address of the start of the current function,
18782 which may be looked up exactly in the symbol table.
18784 This instrumentation is also done for functions expanded inline in other
18785 functions. The profiling calls will indicate where, conceptually, the
18786 inline function is entered and exited. This means that addressable
18787 versions of such functions must be available. If all your uses of a
18788 function are expanded inline, this may mean an additional expansion of
18789 code size. If you use @samp{extern inline} in your C code, an
18790 addressable version of such functions must be provided. (This is
18791 normally the case anyways, but if you get lucky and the optimizer always
18792 expands the functions inline, you might have gotten away without
18793 providing static copies.)
18795 A function may be given the attribute @code{no_instrument_function}, in
18796 which case this instrumentation will not be done. This can be used, for
18797 example, for the profiling functions listed above, high-priority
18798 interrupt routines, and any functions from which the profiling functions
18799 cannot safely be called (perhaps signal handlers, if the profiling
18800 routines generate output or allocate memory).
18802 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18803 @opindex finstrument-functions-exclude-file-list
18805 Set the list of functions that are excluded from instrumentation (see
18806 the description of @code{-finstrument-functions}). If the file that
18807 contains a function definition matches with one of @var{file}, then
18808 that function is not instrumented. The match is done on substrings:
18809 if the @var{file} parameter is a substring of the file name, it is
18810 considered to be a match.
18815 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18819 will exclude any inline function defined in files whose pathnames
18820 contain @code{/bits/stl} or @code{include/sys}.
18822 If, for some reason, you want to include letter @code{','} in one of
18823 @var{sym}, write @code{'\,'}. For example,
18824 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18825 (note the single quote surrounding the option).
18827 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18828 @opindex finstrument-functions-exclude-function-list
18830 This is similar to @code{-finstrument-functions-exclude-file-list},
18831 but this option sets the list of function names to be excluded from
18832 instrumentation. The function name to be matched is its user-visible
18833 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18834 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18835 match is done on substrings: if the @var{sym} parameter is a substring
18836 of the function name, it is considered to be a match. For C99 and C++
18837 extended identifiers, the function name must be given in UTF-8, not
18838 using universal character names.
18840 @item -fstack-check
18841 @opindex fstack-check
18842 Generate code to verify that you do not go beyond the boundary of the
18843 stack. You should specify this flag if you are running in an
18844 environment with multiple threads, but only rarely need to specify it in
18845 a single-threaded environment since stack overflow is automatically
18846 detected on nearly all systems if there is only one stack.
18848 Note that this switch does not actually cause checking to be done; the
18849 operating system or the language runtime must do that. The switch causes
18850 generation of code to ensure that they see the stack being extended.
18852 You can additionally specify a string parameter: @code{no} means no
18853 checking, @code{generic} means force the use of old-style checking,
18854 @code{specific} means use the best checking method and is equivalent
18855 to bare @option{-fstack-check}.
18857 Old-style checking is a generic mechanism that requires no specific
18858 target support in the compiler but comes with the following drawbacks:
18862 Modified allocation strategy for large objects: they will always be
18863 allocated dynamically if their size exceeds a fixed threshold.
18866 Fixed limit on the size of the static frame of functions: when it is
18867 topped by a particular function, stack checking is not reliable and
18868 a warning is issued by the compiler.
18871 Inefficiency: because of both the modified allocation strategy and the
18872 generic implementation, the performances of the code are hampered.
18875 Note that old-style stack checking is also the fallback method for
18876 @code{specific} if no target support has been added in the compiler.
18878 @item -fstack-limit-register=@var{reg}
18879 @itemx -fstack-limit-symbol=@var{sym}
18880 @itemx -fno-stack-limit
18881 @opindex fstack-limit-register
18882 @opindex fstack-limit-symbol
18883 @opindex fno-stack-limit
18884 Generate code to ensure that the stack does not grow beyond a certain value,
18885 either the value of a register or the address of a symbol. If the stack
18886 would grow beyond the value, a signal is raised. For most targets,
18887 the signal is raised before the stack overruns the boundary, so
18888 it is possible to catch the signal without taking special precautions.
18890 For instance, if the stack starts at absolute address @samp{0x80000000}
18891 and grows downwards, you can use the flags
18892 @option{-fstack-limit-symbol=__stack_limit} and
18893 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18894 of 128KB@. Note that this may only work with the GNU linker.
18896 @item -fsplit-stack
18897 @opindex fsplit-stack
18898 Generate code to automatically split the stack before it overflows.
18899 The resulting program has a discontiguous stack which can only
18900 overflow if the program is unable to allocate any more memory. This
18901 is most useful when running threaded programs, as it is no longer
18902 necessary to calculate a good stack size to use for each thread. This
18903 is currently only implemented for the i386 and x86_64 backends running
18906 When code compiled with @option{-fsplit-stack} calls code compiled
18907 without @option{-fsplit-stack}, there may not be much stack space
18908 available for the latter code to run. If compiling all code,
18909 including library code, with @option{-fsplit-stack} is not an option,
18910 then the linker can fix up these calls so that the code compiled
18911 without @option{-fsplit-stack} always has a large stack. Support for
18912 this is implemented in the gold linker in GNU binutils release 2.21
18915 @item -fleading-underscore
18916 @opindex fleading-underscore
18917 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18918 change the way C symbols are represented in the object file. One use
18919 is to help link with legacy assembly code.
18921 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18922 generate code that is not binary compatible with code generated without that
18923 switch. Use it to conform to a non-default application binary interface.
18924 Not all targets provide complete support for this switch.
18926 @item -ftls-model=@var{model}
18927 @opindex ftls-model
18928 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18929 The @var{model} argument should be one of @code{global-dynamic},
18930 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18932 The default without @option{-fpic} is @code{initial-exec}; with
18933 @option{-fpic} the default is @code{global-dynamic}.
18935 @item -fvisibility=@var{default|internal|hidden|protected}
18936 @opindex fvisibility
18937 Set the default ELF image symbol visibility to the specified option---all
18938 symbols will be marked with this unless overridden within the code.
18939 Using this feature can very substantially improve linking and
18940 load times of shared object libraries, produce more optimized
18941 code, provide near-perfect API export and prevent symbol clashes.
18942 It is @strong{strongly} recommended that you use this in any shared objects
18945 Despite the nomenclature, @code{default} always means public; i.e.,
18946 available to be linked against from outside the shared object.
18947 @code{protected} and @code{internal} are pretty useless in real-world
18948 usage so the only other commonly used option will be @code{hidden}.
18949 The default if @option{-fvisibility} isn't specified is
18950 @code{default}, i.e., make every
18951 symbol public---this causes the same behavior as previous versions of
18954 A good explanation of the benefits offered by ensuring ELF
18955 symbols have the correct visibility is given by ``How To Write
18956 Shared Libraries'' by Ulrich Drepper (which can be found at
18957 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18958 solution made possible by this option to marking things hidden when
18959 the default is public is to make the default hidden and mark things
18960 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18961 and @code{__attribute__ ((visibility("default")))} instead of
18962 @code{__declspec(dllexport)} you get almost identical semantics with
18963 identical syntax. This is a great boon to those working with
18964 cross-platform projects.
18966 For those adding visibility support to existing code, you may find
18967 @samp{#pragma GCC visibility} of use. This works by you enclosing
18968 the declarations you wish to set visibility for with (for example)
18969 @samp{#pragma GCC visibility push(hidden)} and
18970 @samp{#pragma GCC visibility pop}.
18971 Bear in mind that symbol visibility should be viewed @strong{as
18972 part of the API interface contract} and thus all new code should
18973 always specify visibility when it is not the default; i.e., declarations
18974 only for use within the local DSO should @strong{always} be marked explicitly
18975 as hidden as so to avoid PLT indirection overheads---making this
18976 abundantly clear also aids readability and self-documentation of the code.
18977 Note that due to ISO C++ specification requirements, operator new and
18978 operator delete must always be of default visibility.
18980 Be aware that headers from outside your project, in particular system
18981 headers and headers from any other library you use, may not be
18982 expecting to be compiled with visibility other than the default. You
18983 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18984 before including any such headers.
18986 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18987 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18988 no modifications. However, this means that calls to @samp{extern}
18989 functions with no explicit visibility will use the PLT, so it is more
18990 effective to use @samp{__attribute ((visibility))} and/or
18991 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18992 declarations should be treated as hidden.
18994 Note that @samp{-fvisibility} does affect C++ vague linkage
18995 entities. This means that, for instance, an exception class that will
18996 be thrown between DSOs must be explicitly marked with default
18997 visibility so that the @samp{type_info} nodes will be unified between
19000 An overview of these techniques, their benefits and how to use them
19001 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
19003 @item -fstrict-volatile-bitfields
19004 @opindex fstrict-volatile-bitfields
19005 This option should be used if accesses to volatile bitfields (or other
19006 structure fields, although the compiler usually honors those types
19007 anyway) should use a single access of the width of the
19008 field's type, aligned to a natural alignment if possible. For
19009 example, targets with memory-mapped peripheral registers might require
19010 all such accesses to be 16 bits wide; with this flag the user could
19011 declare all peripheral bitfields as ``unsigned short'' (assuming short
19012 is 16 bits on these targets) to force GCC to use 16 bit accesses
19013 instead of, perhaps, a more efficient 32 bit access.
19015 If this option is disabled, the compiler will use the most efficient
19016 instruction. In the previous example, that might be a 32-bit load
19017 instruction, even though that will access bytes that do not contain
19018 any portion of the bitfield, or memory-mapped registers unrelated to
19019 the one being updated.
19021 If the target requires strict alignment, and honoring the field
19022 type would require violating this alignment, a warning is issued.
19023 If the field has @code{packed} attribute, the access is done without
19024 honoring the field type. If the field doesn't have @code{packed}
19025 attribute, the access is done honoring the field type. In both cases,
19026 GCC assumes that the user knows something about the target hardware
19027 that it is unaware of.
19029 The default value of this option is determined by the application binary
19030 interface for the target processor.
19036 @node Environment Variables
19037 @section Environment Variables Affecting GCC
19038 @cindex environment variables
19040 @c man begin ENVIRONMENT
19041 This section describes several environment variables that affect how GCC
19042 operates. Some of them work by specifying directories or prefixes to use
19043 when searching for various kinds of files. Some are used to specify other
19044 aspects of the compilation environment.
19046 Note that you can also specify places to search using options such as
19047 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
19048 take precedence over places specified using environment variables, which
19049 in turn take precedence over those specified by the configuration of GCC@.
19050 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
19051 GNU Compiler Collection (GCC) Internals}.
19056 @c @itemx LC_COLLATE
19058 @c @itemx LC_MONETARY
19059 @c @itemx LC_NUMERIC
19064 @c @findex LC_COLLATE
19065 @findex LC_MESSAGES
19066 @c @findex LC_MONETARY
19067 @c @findex LC_NUMERIC
19071 These environment variables control the way that GCC uses
19072 localization information that allow GCC to work with different
19073 national conventions. GCC inspects the locale categories
19074 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
19075 so. These locale categories can be set to any value supported by your
19076 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
19077 Kingdom encoded in UTF-8.
19079 The @env{LC_CTYPE} environment variable specifies character
19080 classification. GCC uses it to determine the character boundaries in
19081 a string; this is needed for some multibyte encodings that contain quote
19082 and escape characters that would otherwise be interpreted as a string
19085 The @env{LC_MESSAGES} environment variable specifies the language to
19086 use in diagnostic messages.
19088 If the @env{LC_ALL} environment variable is set, it overrides the value
19089 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
19090 and @env{LC_MESSAGES} default to the value of the @env{LANG}
19091 environment variable. If none of these variables are set, GCC
19092 defaults to traditional C English behavior.
19096 If @env{TMPDIR} is set, it specifies the directory to use for temporary
19097 files. GCC uses temporary files to hold the output of one stage of
19098 compilation which is to be used as input to the next stage: for example,
19099 the output of the preprocessor, which is the input to the compiler
19102 @item GCC_COMPARE_DEBUG
19103 @findex GCC_COMPARE_DEBUG
19104 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
19105 @option{-fcompare-debug} to the compiler driver. See the documentation
19106 of this option for more details.
19108 @item GCC_EXEC_PREFIX
19109 @findex GCC_EXEC_PREFIX
19110 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
19111 names of the subprograms executed by the compiler. No slash is added
19112 when this prefix is combined with the name of a subprogram, but you can
19113 specify a prefix that ends with a slash if you wish.
19115 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
19116 an appropriate prefix to use based on the pathname it was invoked with.
19118 If GCC cannot find the subprogram using the specified prefix, it
19119 tries looking in the usual places for the subprogram.
19121 The default value of @env{GCC_EXEC_PREFIX} is
19122 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
19123 the installed compiler. In many cases @var{prefix} is the value
19124 of @code{prefix} when you ran the @file{configure} script.
19126 Other prefixes specified with @option{-B} take precedence over this prefix.
19128 This prefix is also used for finding files such as @file{crt0.o} that are
19131 In addition, the prefix is used in an unusual way in finding the
19132 directories to search for header files. For each of the standard
19133 directories whose name normally begins with @samp{/usr/local/lib/gcc}
19134 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
19135 replacing that beginning with the specified prefix to produce an
19136 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
19137 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
19138 These alternate directories are searched first; the standard directories
19139 come next. If a standard directory begins with the configured
19140 @var{prefix} then the value of @var{prefix} is replaced by
19141 @env{GCC_EXEC_PREFIX} when looking for header files.
19143 @item COMPILER_PATH
19144 @findex COMPILER_PATH
19145 The value of @env{COMPILER_PATH} is a colon-separated list of
19146 directories, much like @env{PATH}. GCC tries the directories thus
19147 specified when searching for subprograms, if it can't find the
19148 subprograms using @env{GCC_EXEC_PREFIX}.
19151 @findex LIBRARY_PATH
19152 The value of @env{LIBRARY_PATH} is a colon-separated list of
19153 directories, much like @env{PATH}. When configured as a native compiler,
19154 GCC tries the directories thus specified when searching for special
19155 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
19156 using GCC also uses these directories when searching for ordinary
19157 libraries for the @option{-l} option (but directories specified with
19158 @option{-L} come first).
19162 @cindex locale definition
19163 This variable is used to pass locale information to the compiler. One way in
19164 which this information is used is to determine the character set to be used
19165 when character literals, string literals and comments are parsed in C and C++.
19166 When the compiler is configured to allow multibyte characters,
19167 the following values for @env{LANG} are recognized:
19171 Recognize JIS characters.
19173 Recognize SJIS characters.
19175 Recognize EUCJP characters.
19178 If @env{LANG} is not defined, or if it has some other value, then the
19179 compiler will use mblen and mbtowc as defined by the default locale to
19180 recognize and translate multibyte characters.
19184 Some additional environments variables affect the behavior of the
19187 @include cppenv.texi
19191 @node Precompiled Headers
19192 @section Using Precompiled Headers
19193 @cindex precompiled headers
19194 @cindex speed of compilation
19196 Often large projects have many header files that are included in every
19197 source file. The time the compiler takes to process these header files
19198 over and over again can account for nearly all of the time required to
19199 build the project. To make builds faster, GCC allows users to
19200 `precompile' a header file; then, if builds can use the precompiled
19201 header file they will be much faster.
19203 To create a precompiled header file, simply compile it as you would any
19204 other file, if necessary using the @option{-x} option to make the driver
19205 treat it as a C or C++ header file. You will probably want to use a
19206 tool like @command{make} to keep the precompiled header up-to-date when
19207 the headers it contains change.
19209 A precompiled header file will be searched for when @code{#include} is
19210 seen in the compilation. As it searches for the included file
19211 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
19212 compiler looks for a precompiled header in each directory just before it
19213 looks for the include file in that directory. The name searched for is
19214 the name specified in the @code{#include} with @samp{.gch} appended. If
19215 the precompiled header file can't be used, it is ignored.
19217 For instance, if you have @code{#include "all.h"}, and you have
19218 @file{all.h.gch} in the same directory as @file{all.h}, then the
19219 precompiled header file will be used if possible, and the original
19220 header will be used otherwise.
19222 Alternatively, you might decide to put the precompiled header file in a
19223 directory and use @option{-I} to ensure that directory is searched
19224 before (or instead of) the directory containing the original header.
19225 Then, if you want to check that the precompiled header file is always
19226 used, you can put a file of the same name as the original header in this
19227 directory containing an @code{#error} command.
19229 This also works with @option{-include}. So yet another way to use
19230 precompiled headers, good for projects not designed with precompiled
19231 header files in mind, is to simply take most of the header files used by
19232 a project, include them from another header file, precompile that header
19233 file, and @option{-include} the precompiled header. If the header files
19234 have guards against multiple inclusion, they will be skipped because
19235 they've already been included (in the precompiled header).
19237 If you need to precompile the same header file for different
19238 languages, targets, or compiler options, you can instead make a
19239 @emph{directory} named like @file{all.h.gch}, and put each precompiled
19240 header in the directory, perhaps using @option{-o}. It doesn't matter
19241 what you call the files in the directory, every precompiled header in
19242 the directory will be considered. The first precompiled header
19243 encountered in the directory that is valid for this compilation will
19244 be used; they're searched in no particular order.
19246 There are many other possibilities, limited only by your imagination,
19247 good sense, and the constraints of your build system.
19249 A precompiled header file can be used only when these conditions apply:
19253 Only one precompiled header can be used in a particular compilation.
19256 A precompiled header can't be used once the first C token is seen. You
19257 can have preprocessor directives before a precompiled header; you can
19258 even include a precompiled header from inside another header, so long as
19259 there are no C tokens before the @code{#include}.
19262 The precompiled header file must be produced for the same language as
19263 the current compilation. You can't use a C precompiled header for a C++
19267 The precompiled header file must have been produced by the same compiler
19268 binary as the current compilation is using.
19271 Any macros defined before the precompiled header is included must
19272 either be defined in the same way as when the precompiled header was
19273 generated, or must not affect the precompiled header, which usually
19274 means that they don't appear in the precompiled header at all.
19276 The @option{-D} option is one way to define a macro before a
19277 precompiled header is included; using a @code{#define} can also do it.
19278 There are also some options that define macros implicitly, like
19279 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
19282 @item If debugging information is output when using the precompiled
19283 header, using @option{-g} or similar, the same kind of debugging information
19284 must have been output when building the precompiled header. However,
19285 a precompiled header built using @option{-g} can be used in a compilation
19286 when no debugging information is being output.
19288 @item The same @option{-m} options must generally be used when building
19289 and using the precompiled header. @xref{Submodel Options},
19290 for any cases where this rule is relaxed.
19292 @item Each of the following options must be the same when building and using
19293 the precompiled header:
19295 @gccoptlist{-fexceptions}
19298 Some other command-line options starting with @option{-f},
19299 @option{-p}, or @option{-O} must be defined in the same way as when
19300 the precompiled header was generated. At present, it's not clear
19301 which options are safe to change and which are not; the safest choice
19302 is to use exactly the same options when generating and using the
19303 precompiled header. The following are known to be safe:
19305 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
19306 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
19307 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
19312 For all of these except the last, the compiler will automatically
19313 ignore the precompiled header if the conditions aren't met. If you
19314 find an option combination that doesn't work and doesn't cause the
19315 precompiled header to be ignored, please consider filing a bug report,
19318 If you do use differing options when generating and using the
19319 precompiled header, the actual behavior will be a mixture of the
19320 behavior for the options. For instance, if you use @option{-g} to
19321 generate the precompiled header but not when using it, you may or may
19322 not get debugging information for routines in the precompiled header.