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, 2012
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,
16 Free Software Foundation, Inc.
18 Permission is granted to copy, distribute and/or modify this document
19 under the terms of the GNU Free Documentation License, Version 1.3 or
20 any later version published by the Free Software Foundation; with the
21 Invariant Sections being ``GNU General Public License'' and ``Funding
22 Free Software'', the Front-Cover texts being (a) (see below), and with
23 the Back-Cover Texts being (b) (see below). A copy of the license is
24 included in the gfdl(7) man page.
26 (a) The FSF's Front-Cover Text is:
30 (b) The FSF's Back-Cover Text is:
32 You have freedom to copy and modify this GNU Manual, like GNU
33 software. Copies published by the Free Software Foundation raise
34 funds for GNU development.
36 @c Set file name and title for the man page.
38 @settitle GNU project C and C++ compiler
40 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
41 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
42 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
43 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
44 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
45 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
46 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
48 Only the most useful options are listed here; see below for the
49 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
52 gpl(7), gfdl(7), fsf-funding(7),
53 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
54 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
55 @file{ld}, @file{binutils} and @file{gdb}.
58 For instructions on reporting bugs, see
62 See the Info entry for @command{gcc}, or
63 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
64 for contributors to GCC@.
69 @chapter GCC Command Options
70 @cindex GCC command options
71 @cindex command options
72 @cindex options, GCC command
74 @c man begin DESCRIPTION
75 When you invoke GCC, it normally does preprocessing, compilation,
76 assembly and linking. The ``overall options'' allow you to stop this
77 process at an intermediate stage. For example, the @option{-c} option
78 says not to run the linker. Then the output consists of object files
79 output by the assembler.
81 Other options are passed on to one stage of processing. Some options
82 control the preprocessor and others the compiler itself. Yet other
83 options control the assembler and linker; most of these are not
84 documented here, since you rarely need to use any of them.
86 @cindex C compilation options
87 Most of the command-line options that you can use with GCC are useful
88 for C programs; when an option is only useful with another language
89 (usually C++), the explanation says so explicitly. If the description
90 for a particular option does not mention a source language, you can use
91 that option with all supported languages.
93 @cindex C++ compilation options
94 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
95 options for compiling C++ programs.
97 @cindex grouping options
98 @cindex options, grouping
99 The @command{gcc} program accepts options and file names as operands. Many
100 options have multi-letter names; therefore multiple single-letter options
101 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
104 @cindex order of options
105 @cindex options, order
106 You can mix options and other arguments. For the most part, the order
107 you use doesn't matter. Order does matter when you use several
108 options of the same kind; for example, if you specify @option{-L} more
109 than once, the directories are searched in the order specified. Also,
110 the placement of the @option{-l} option is significant.
112 Many options have long names starting with @samp{-f} or with
113 @samp{-W}---for example,
114 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
115 these have both positive and negative forms; the negative form of
116 @option{-ffoo} would be @option{-fno-foo}. This manual documents
117 only one of these two forms, whichever one is not the default.
121 @xref{Option Index}, for an index to GCC's options.
124 * Option Summary:: Brief list of all options, without explanations.
125 * Overall Options:: Controlling the kind of output:
126 an executable, object files, assembler files,
127 or preprocessed source.
128 * Invoking G++:: Compiling C++ programs.
129 * C Dialect Options:: Controlling the variant of C language compiled.
130 * C++ Dialect Options:: Variations on C++.
131 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
133 * Language Independent Options:: Controlling how diagnostics should be
135 * Warning Options:: How picky should the compiler be?
136 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
137 * Optimize Options:: How much optimization?
138 * Preprocessor Options:: Controlling header files and macro definitions.
139 Also, getting dependency information for Make.
140 * Assembler Options:: Passing options to the assembler.
141 * Link Options:: Specifying libraries and so on.
142 * Directory Options:: Where to find header files and libraries.
143 Where to find the compiler executable files.
144 * Spec Files:: How to pass switches to sub-processes.
145 * Target Options:: Running a cross-compiler, or an old version of GCC.
146 * Submodel Options:: Specifying minor hardware or convention variations,
147 such as 68010 vs 68020.
148 * Code Gen Options:: Specifying conventions for function calls, data layout
150 * Environment Variables:: Env vars that affect GCC.
151 * Precompiled Headers:: Compiling a header once, and using it many times.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
166 -pipe -pass-exit-codes @gol
167 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
168 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
169 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
171 @item C Language Options
172 @xref{C Dialect Options,,Options Controlling C Dialect}.
173 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
174 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
175 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
176 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
177 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
178 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
179 -fsigned-bitfields -fsigned-char @gol
180 -funsigned-bitfields -funsigned-char}
182 @item C++ Language Options
183 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
184 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
185 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
186 -fno-elide-constructors @gol
187 -fno-enforce-eh-specs @gol
188 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
189 -fno-implicit-templates @gol
190 -fno-implicit-inline-templates @gol
191 -fno-implement-inlines -fms-extensions @gol
192 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
193 -fno-optional-diags -fpermissive @gol
194 -fno-pretty-templates @gol
195 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
196 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
197 -fno-default-inline -fvisibility-inlines-hidden @gol
198 -fvisibility-ms-compat @gol
199 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
200 -Wdelete-non-virtual-dtor -Wnarrowing -Wnoexcept @gol
201 -Wnon-virtual-dtor -Wreorder @gol
202 -Weffc++ -Wstrict-null-sentinel @gol
203 -Wno-non-template-friend -Wold-style-cast @gol
204 -Woverloaded-virtual -Wno-pmf-conversions @gol
207 @item Objective-C and Objective-C++ Language Options
208 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
209 Objective-C and Objective-C++ Dialects}.
210 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
211 -fgnu-runtime -fnext-runtime @gol
212 -fno-nil-receivers @gol
213 -fobjc-abi-version=@var{n} @gol
214 -fobjc-call-cxx-cdtors @gol
215 -fobjc-direct-dispatch @gol
216 -fobjc-exceptions @gol
219 -fobjc-std=objc1 @gol
220 -freplace-objc-classes @gol
223 -Wassign-intercept @gol
224 -Wno-protocol -Wselector @gol
225 -Wstrict-selector-match @gol
226 -Wundeclared-selector}
228 @item Language Independent Options
229 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
230 @gccoptlist{-fmessage-length=@var{n} @gol
231 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
232 -fno-diagnostics-show-option}
234 @item Warning Options
235 @xref{Warning Options,,Options to Request or Suppress Warnings}.
236 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
237 -pedantic-errors @gol
238 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
239 -Wno-attributes -Wno-builtin-macro-redefined @gol
240 -Wc++-compat -Wc++11-compat -Wcast-align -Wcast-qual @gol
241 -Wchar-subscripts -Wclobbered -Wcomment @gol
242 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
243 -Wno-deprecated-declarations -Wdisabled-optimization @gol
244 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
245 -Wno-endif-labels -Werror -Werror=* @gol
246 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
247 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
248 -Wformat-security -Wformat-y2k @gol
249 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
250 -Wignored-qualifiers @gol
251 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
252 -Winit-self -Winline -Wmaybe-uninitialized @gol
253 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
254 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
255 -Wlogical-op -Wlong-long @gol
256 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
257 -Wmissing-format-attribute -Wmissing-include-dirs @gol
259 -Wno-multichar -Wnonnull -Wno-overflow @gol
260 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
261 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
262 -Wpointer-arith -Wno-pointer-to-int-cast @gol
263 -Wredundant-decls @gol
264 -Wreturn-type -Wsequence-point -Wshadow @gol
265 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
266 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
267 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
268 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
269 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
270 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
271 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
272 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
273 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
274 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
275 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
276 -Wvariadic-macros -Wvector-operation-performance -Wvla
277 -Wvolatile-register-var -Wwrite-strings -Wzero-as-null-pointer-constant}
279 @item C and Objective-C-only Warning Options
280 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
281 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
282 -Wold-style-declaration -Wold-style-definition @gol
283 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
284 -Wdeclaration-after-statement -Wpointer-sign}
286 @item Debugging Options
287 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
288 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
289 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
290 -fdisable-ipa-@var{pass_name} @gol
291 -fdisable-rtl-@var{pass_name} @gol
292 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
293 -fdisable-tree-@var{pass_name} @gol
294 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
295 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
296 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
297 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
298 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
300 -fdump-statistics @gol
302 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
306 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-nrv -fdump-tree-vect @gol
316 -fdump-tree-sink @gol
317 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
318 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
319 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
320 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
321 -ftree-vectorizer-verbose=@var{n} @gol
322 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
323 -fdump-final-insns=@var{file} @gol
324 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
325 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
326 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
327 -fenable-@var{kind}-@var{pass} @gol
328 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
329 -fdebug-types-section @gol
330 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
331 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
332 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
333 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
334 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
335 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
336 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
337 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
338 -gvms -gxcoff -gxcoff+ @gol
339 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
340 -fdebug-prefix-map=@var{old}=@var{new} @gol
341 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
342 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
343 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
344 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
345 -print-prog-name=@var{program} -print-search-dirs -Q @gol
346 -print-sysroot -print-sysroot-headers-suffix @gol
347 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
349 @item Optimization Options
350 @xref{Optimize Options,,Options that Control Optimization}.
351 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
352 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
353 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
354 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
355 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
356 -fcompare-elim -fcprop-registers -fcrossjumping @gol
357 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
358 -fcx-limited-range @gol
359 -fdata-sections -fdce -fdce -fdelayed-branch @gol
360 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
361 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
362 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
363 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
364 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
365 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
366 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
367 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
368 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
369 -fira-algorithm=@var{algorithm} @gol
370 -fira-region=@var{region} @gol
371 -fira-loop-pressure -fno-ira-share-save-slots @gol
372 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
373 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
374 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
375 -floop-parallelize-all -flto -flto-compression-level @gol
376 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
377 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
378 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
379 -fno-default-inline @gol
380 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
381 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
382 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
383 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
384 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
385 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
386 -fprefetch-loop-arrays @gol
387 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
388 -fprofile-generate=@var{path} @gol
389 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
390 -freciprocal-math -free -fregmove -frename-registers -freorder-blocks @gol
391 -freorder-blocks-and-partition -freorder-functions @gol
392 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
393 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
394 -fsched-spec-load -fsched-spec-load-dangerous @gol
395 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
396 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
397 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
398 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
399 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
400 -fselective-scheduling -fselective-scheduling2 @gol
401 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
402 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
403 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
404 -fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol
405 -fthread-jumps -ftracer -ftree-bit-ccp @gol
406 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
407 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
408 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
409 -ftree-loop-if-convert-stores -ftree-loop-im @gol
410 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
411 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
412 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
413 -ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol
414 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
415 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
416 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
417 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
418 -fwhole-program -fwpa -fuse-linker-plugin @gol
419 --param @var{name}=@var{value}
420 -O -O0 -O1 -O2 -O3 -Os -Ofast}
422 @item Preprocessor Options
423 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
424 @gccoptlist{-A@var{question}=@var{answer} @gol
425 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
426 -C -dD -dI -dM -dN @gol
427 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
428 -idirafter @var{dir} @gol
429 -include @var{file} -imacros @var{file} @gol
430 -iprefix @var{file} -iwithprefix @var{dir} @gol
431 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
432 -imultilib @var{dir} -isysroot @var{dir} @gol
433 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
434 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
435 -remap -trigraphs -undef -U@var{macro} @gol
436 -Wp,@var{option} -Xpreprocessor @var{option}}
438 @item Assembler Option
439 @xref{Assembler Options,,Passing Options to the Assembler}.
440 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
443 @xref{Link Options,,Options for Linking}.
444 @gccoptlist{@var{object-file-name} -l@var{library} @gol
445 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
446 -s -static -static-libgcc -static-libstdc++ -shared @gol
447 -shared-libgcc -symbolic @gol
448 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
451 @item Directory Options
452 @xref{Directory Options,,Options for Directory Search}.
453 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
454 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
457 @item Machine Dependent Options
458 @xref{Submodel Options,,Hardware Models and Configurations}.
459 @c This list is ordered alphanumerically by subsection name.
460 @c Try and put the significant identifier (CPU or system) first,
461 @c so users have a clue at guessing where the ones they want will be.
463 @emph{Adapteva Epiphany Options}
464 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
465 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
466 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
467 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
468 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
469 -msplit-vecmove-early -m1reg-@var{reg}}
472 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
473 -mabi=@var{name} @gol
474 -mapcs-stack-check -mno-apcs-stack-check @gol
475 -mapcs-float -mno-apcs-float @gol
476 -mapcs-reentrant -mno-apcs-reentrant @gol
477 -msched-prolog -mno-sched-prolog @gol
478 -mlittle-endian -mbig-endian -mwords-little-endian @gol
479 -mfloat-abi=@var{name} -mfpe @gol
480 -mfp16-format=@var{name}
481 -mthumb-interwork -mno-thumb-interwork @gol
482 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
483 -mstructure-size-boundary=@var{n} @gol
484 -mabort-on-noreturn @gol
485 -mlong-calls -mno-long-calls @gol
486 -msingle-pic-base -mno-single-pic-base @gol
487 -mpic-register=@var{reg} @gol
488 -mnop-fun-dllimport @gol
489 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
490 -mpoke-function-name @gol
492 -mtpcs-frame -mtpcs-leaf-frame @gol
493 -mcaller-super-interworking -mcallee-super-interworking @gol
494 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
495 -mword-relocations @gol
496 -mfix-cortex-m3-ldrd}
499 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
500 -mcall-prologues -mint8 -mno-interrupts -mrelax -mshort-calls @gol
501 -mstrict-X -mtiny-stack}
503 @emph{Blackfin Options}
504 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
505 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
506 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
507 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
508 -mno-id-shared-library -mshared-library-id=@var{n} @gol
509 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
510 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
511 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
515 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
516 -msim -msdata=@var{sdata-type}}
519 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
520 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
521 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
522 -mstack-align -mdata-align -mconst-align @gol
523 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
524 -melf -maout -melinux -mlinux -sim -sim2 @gol
525 -mmul-bug-workaround -mno-mul-bug-workaround}
528 @gccoptlist{-mmac @gol
529 -mcr16cplus -mcr16c @gol
530 -msim -mint32 -mbit-ops
531 -mdata-model=@var{model}}
533 @emph{Darwin Options}
534 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
535 -arch_only -bind_at_load -bundle -bundle_loader @gol
536 -client_name -compatibility_version -current_version @gol
538 -dependency-file -dylib_file -dylinker_install_name @gol
539 -dynamic -dynamiclib -exported_symbols_list @gol
540 -filelist -flat_namespace -force_cpusubtype_ALL @gol
541 -force_flat_namespace -headerpad_max_install_names @gol
543 -image_base -init -install_name -keep_private_externs @gol
544 -multi_module -multiply_defined -multiply_defined_unused @gol
545 -noall_load -no_dead_strip_inits_and_terms @gol
546 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
547 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
548 -private_bundle -read_only_relocs -sectalign @gol
549 -sectobjectsymbols -whyload -seg1addr @gol
550 -sectcreate -sectobjectsymbols -sectorder @gol
551 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
552 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
553 -segprot -segs_read_only_addr -segs_read_write_addr @gol
554 -single_module -static -sub_library -sub_umbrella @gol
555 -twolevel_namespace -umbrella -undefined @gol
556 -unexported_symbols_list -weak_reference_mismatches @gol
557 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
558 -mkernel -mone-byte-bool}
560 @emph{DEC Alpha Options}
561 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
562 -mieee -mieee-with-inexact -mieee-conformant @gol
563 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
564 -mtrap-precision=@var{mode} -mbuild-constants @gol
565 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
566 -mbwx -mmax -mfix -mcix @gol
567 -mfloat-vax -mfloat-ieee @gol
568 -mexplicit-relocs -msmall-data -mlarge-data @gol
569 -msmall-text -mlarge-text @gol
570 -mmemory-latency=@var{time}}
572 @emph{DEC Alpha/VMS Options}
573 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
576 @gccoptlist{-msmall-model -mno-lsim}
579 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
580 -mhard-float -msoft-float @gol
581 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
582 -mdouble -mno-double @gol
583 -mmedia -mno-media -mmuladd -mno-muladd @gol
584 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
585 -mlinked-fp -mlong-calls -malign-labels @gol
586 -mlibrary-pic -macc-4 -macc-8 @gol
587 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
588 -moptimize-membar -mno-optimize-membar @gol
589 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
590 -mvliw-branch -mno-vliw-branch @gol
591 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
592 -mno-nested-cond-exec -mtomcat-stats @gol
596 @emph{GNU/Linux Options}
597 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
598 -tno-android-cc -tno-android-ld}
600 @emph{H8/300 Options}
601 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
604 @gccoptlist{-march=@var{architecture-type} @gol
605 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
606 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
607 -mfixed-range=@var{register-range} @gol
608 -mjump-in-delay -mlinker-opt -mlong-calls @gol
609 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
610 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
611 -mno-jump-in-delay -mno-long-load-store @gol
612 -mno-portable-runtime -mno-soft-float @gol
613 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
614 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
615 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
616 -munix=@var{unix-std} -nolibdld -static -threads}
618 @emph{i386 and x86-64 Options}
619 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
620 -mfpmath=@var{unit} @gol
621 -masm=@var{dialect} -mno-fancy-math-387 @gol
622 -mno-fp-ret-in-387 -msoft-float @gol
623 -mno-wide-multiply -mrtd -malign-double @gol
624 -mpreferred-stack-boundary=@var{num} @gol
625 -mincoming-stack-boundary=@var{num} @gol
626 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
627 -mrecip -mrecip=@var{opt} @gol
629 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
630 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
631 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
632 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
633 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
634 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
635 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
636 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
637 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
638 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
639 -mcmodel=@var{code-model} -mabi=@var{name} @gol
640 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
641 -msse2avx -mfentry -m8bit-idiv @gol
642 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
644 @emph{i386 and x86-64 Windows Options}
645 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
646 -mnop-fun-dllimport -mthread @gol
647 -municode -mwin32 -mwindows -fno-set-stack-executable}
650 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
651 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
652 -mconstant-gp -mauto-pic -mfused-madd @gol
653 -minline-float-divide-min-latency @gol
654 -minline-float-divide-max-throughput @gol
655 -mno-inline-float-divide @gol
656 -minline-int-divide-min-latency @gol
657 -minline-int-divide-max-throughput @gol
658 -mno-inline-int-divide @gol
659 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
660 -mno-inline-sqrt @gol
661 -mdwarf2-asm -mearly-stop-bits @gol
662 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
663 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
664 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
665 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
666 -msched-spec-ldc -msched-spec-control-ldc @gol
667 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
668 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
669 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
670 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
672 @emph{IA-64/VMS Options}
673 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
676 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
677 -msign-extend-enabled -muser-enabled}
679 @emph{M32R/D Options}
680 @gccoptlist{-m32r2 -m32rx -m32r @gol
682 -malign-loops -mno-align-loops @gol
683 -missue-rate=@var{number} @gol
684 -mbranch-cost=@var{number} @gol
685 -mmodel=@var{code-size-model-type} @gol
686 -msdata=@var{sdata-type} @gol
687 -mno-flush-func -mflush-func=@var{name} @gol
688 -mno-flush-trap -mflush-trap=@var{number} @gol
692 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
694 @emph{M680x0 Options}
695 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
696 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
697 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
698 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
699 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
700 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
701 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
702 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
706 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
707 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
708 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
709 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
710 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
713 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
714 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
715 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
716 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
719 @emph{MicroBlaze Options}
720 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
721 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
722 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
723 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
724 -mxl-mode-@var{app-model}}
727 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
728 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
729 -mips64 -mips64r2 @gol
730 -mips16 -mno-mips16 -mflip-mips16 @gol
731 -minterlink-mips16 -mno-interlink-mips16 @gol
732 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
733 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
734 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
735 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
736 -mfpu=@var{fpu-type} @gol
737 -msmartmips -mno-smartmips @gol
738 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
739 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
740 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
741 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
742 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
743 -membedded-data -mno-embedded-data @gol
744 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
745 -mcode-readable=@var{setting} @gol
746 -msplit-addresses -mno-split-addresses @gol
747 -mexplicit-relocs -mno-explicit-relocs @gol
748 -mcheck-zero-division -mno-check-zero-division @gol
749 -mdivide-traps -mdivide-breaks @gol
750 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
751 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
752 -mfix-24k -mno-fix-24k @gol
753 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
754 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
755 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
756 -mflush-func=@var{func} -mno-flush-func @gol
757 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
758 -mfp-exceptions -mno-fp-exceptions @gol
759 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
760 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
763 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
764 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
765 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
766 -mno-base-addresses -msingle-exit -mno-single-exit}
768 @emph{MN10300 Options}
769 @gccoptlist{-mmult-bug -mno-mult-bug @gol
770 -mno-am33 -mam33 -mam33-2 -mam34 @gol
771 -mtune=@var{cpu-type} @gol
772 -mreturn-pointer-on-d0 @gol
773 -mno-crt0 -mrelax -mliw -msetlb}
775 @emph{PDP-11 Options}
776 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
777 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
778 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
779 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
780 -mbranch-expensive -mbranch-cheap @gol
781 -munix-asm -mdec-asm}
783 @emph{picoChip Options}
784 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
785 -msymbol-as-address -mno-inefficient-warnings}
787 @emph{PowerPC Options}
788 See RS/6000 and PowerPC Options.
791 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78}
793 @emph{RS/6000 and PowerPC Options}
794 @gccoptlist{-mcpu=@var{cpu-type} @gol
795 -mtune=@var{cpu-type} @gol
796 -mcmodel=@var{code-model} @gol
797 -mpower -mno-power -mpower2 -mno-power2 @gol
798 -mpowerpc -mpowerpc64 -mno-powerpc @gol
799 -maltivec -mno-altivec @gol
800 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
801 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
802 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
803 -mfprnd -mno-fprnd @gol
804 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
805 -mnew-mnemonics -mold-mnemonics @gol
806 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
807 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
808 -malign-power -malign-natural @gol
809 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
810 -msingle-float -mdouble-float -msimple-fpu @gol
811 -mstring -mno-string -mupdate -mno-update @gol
812 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
813 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
814 -mstrict-align -mno-strict-align -mrelocatable @gol
815 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
816 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
817 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
818 -mprioritize-restricted-insns=@var{priority} @gol
819 -msched-costly-dep=@var{dependence_type} @gol
820 -minsert-sched-nops=@var{scheme} @gol
821 -mcall-sysv -mcall-netbsd @gol
822 -maix-struct-return -msvr4-struct-return @gol
823 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
824 -mblock-move-inline-limit=@var{num} @gol
825 -misel -mno-isel @gol
826 -misel=yes -misel=no @gol
828 -mspe=yes -mspe=no @gol
830 -mgen-cell-microcode -mwarn-cell-microcode @gol
831 -mvrsave -mno-vrsave @gol
832 -mmulhw -mno-mulhw @gol
833 -mdlmzb -mno-dlmzb @gol
834 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
835 -mprototype -mno-prototype @gol
836 -msim -mmvme -mads -myellowknife -memb -msdata @gol
837 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
838 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
839 -mno-recip-precision @gol
840 -mveclibabi=@var{type} -mfriz -mno-friz @gol
841 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
842 -msave-toc-indirect -mno-save-toc-indirect}
845 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
847 -mbig-endian-data -mlittle-endian-data @gol
850 -mas100-syntax -mno-as100-syntax@gol
852 -mmax-constant-size=@gol
855 -msave-acc-in-interrupts}
857 @emph{S/390 and zSeries Options}
858 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
859 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
860 -mlong-double-64 -mlong-double-128 @gol
861 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
862 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
863 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
864 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
865 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
868 @gccoptlist{-meb -mel @gol
872 -mscore5 -mscore5u -mscore7 -mscore7d}
875 @gccoptlist{-m1 -m2 -m2e @gol
876 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
878 -m4-nofpu -m4-single-only -m4-single -m4 @gol
879 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
880 -m5-64media -m5-64media-nofpu @gol
881 -m5-32media -m5-32media-nofpu @gol
882 -m5-compact -m5-compact-nofpu @gol
883 -mb -ml -mdalign -mrelax @gol
884 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
885 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
886 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
887 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
888 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
889 -maccumulate-outgoing-args -minvalid-symbols -msoft-atomic}
891 @emph{Solaris 2 Options}
892 @gccoptlist{-mimpure-text -mno-impure-text @gol
896 @gccoptlist{-mcpu=@var{cpu-type} @gol
897 -mtune=@var{cpu-type} @gol
898 -mcmodel=@var{code-model} @gol
899 -mmemory-model=@var{mem-model} @gol
900 -m32 -m64 -mapp-regs -mno-app-regs @gol
901 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
902 -mfpu -mno-fpu -mhard-float -msoft-float @gol
903 -mhard-quad-float -msoft-quad-float @gol
905 -mstack-bias -mno-stack-bias @gol
906 -munaligned-doubles -mno-unaligned-doubles @gol
907 -mv8plus -mno-v8plus -mvis -mno-vis @gol
908 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
909 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
913 @gccoptlist{-mwarn-reloc -merror-reloc @gol
914 -msafe-dma -munsafe-dma @gol
916 -msmall-mem -mlarge-mem -mstdmain @gol
917 -mfixed-range=@var{register-range} @gol
919 -maddress-space-conversion -mno-address-space-conversion @gol
920 -mcache-size=@var{cache-size} @gol
921 -matomic-updates -mno-atomic-updates}
923 @emph{System V Options}
924 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
927 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
928 -mprolog-function -mno-prolog-function -mspace @gol
929 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
930 -mapp-regs -mno-app-regs @gol
931 -mdisable-callt -mno-disable-callt @gol
934 -mv850e1 -mv850es @gol
939 @gccoptlist{-mg -mgnu -munix}
941 @emph{VxWorks Options}
942 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
943 -Xbind-lazy -Xbind-now}
945 @emph{x86-64 Options}
946 See i386 and x86-64 Options.
948 @emph{Xstormy16 Options}
951 @emph{Xtensa Options}
952 @gccoptlist{-mconst16 -mno-const16 @gol
953 -mfused-madd -mno-fused-madd @gol
955 -mserialize-volatile -mno-serialize-volatile @gol
956 -mtext-section-literals -mno-text-section-literals @gol
957 -mtarget-align -mno-target-align @gol
958 -mlongcalls -mno-longcalls}
960 @emph{zSeries Options}
961 See S/390 and zSeries Options.
963 @item Code Generation Options
964 @xref{Code Gen Options,,Options for Code Generation Conventions}.
965 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
966 -ffixed-@var{reg} -fexceptions @gol
967 -fnon-call-exceptions -funwind-tables @gol
968 -fasynchronous-unwind-tables @gol
969 -finhibit-size-directive -finstrument-functions @gol
970 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
971 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
972 -fno-common -fno-ident @gol
973 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
974 -fno-jump-tables @gol
975 -frecord-gcc-switches @gol
976 -freg-struct-return -fshort-enums @gol
977 -fshort-double -fshort-wchar @gol
978 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
979 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
980 -fno-stack-limit -fsplit-stack @gol
981 -fleading-underscore -ftls-model=@var{model} @gol
982 -ftrapv -fwrapv -fbounds-check @gol
983 -fvisibility -fstrict-volatile-bitfields}
987 * Overall Options:: Controlling the kind of output:
988 an executable, object files, assembler files,
989 or preprocessed source.
990 * C Dialect Options:: Controlling the variant of C language compiled.
991 * C++ Dialect Options:: Variations on C++.
992 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
994 * Language Independent Options:: Controlling how diagnostics should be
996 * Warning Options:: How picky should the compiler be?
997 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
998 * Optimize Options:: How much optimization?
999 * Preprocessor Options:: Controlling header files and macro definitions.
1000 Also, getting dependency information for Make.
1001 * Assembler Options:: Passing options to the assembler.
1002 * Link Options:: Specifying libraries and so on.
1003 * Directory Options:: Where to find header files and libraries.
1004 Where to find the compiler executable files.
1005 * Spec Files:: How to pass switches to sub-processes.
1006 * Target Options:: Running a cross-compiler, or an old version of GCC.
1009 @node Overall Options
1010 @section Options Controlling the Kind of Output
1012 Compilation can involve up to four stages: preprocessing, compilation
1013 proper, assembly and linking, always in that order. GCC is capable of
1014 preprocessing and compiling several files either into several
1015 assembler input files, or into one assembler input file; then each
1016 assembler input file produces an object file, and linking combines all
1017 the object files (those newly compiled, and those specified as input)
1018 into an executable file.
1020 @cindex file name suffix
1021 For any given input file, the file name suffix determines what kind of
1022 compilation is done:
1026 C source code which must be preprocessed.
1029 C source code which should not be preprocessed.
1032 C++ source code which should not be preprocessed.
1035 Objective-C source code. Note that you must link with the @file{libobjc}
1036 library to make an Objective-C program work.
1039 Objective-C source code which should not be preprocessed.
1043 Objective-C++ source code. Note that you must link with the @file{libobjc}
1044 library to make an Objective-C++ program work. Note that @samp{.M} refers
1045 to a literal capital M@.
1047 @item @var{file}.mii
1048 Objective-C++ source code which should not be preprocessed.
1051 C, C++, Objective-C or Objective-C++ header file to be turned into a
1052 precompiled header (default), or C, C++ header file to be turned into an
1053 Ada spec (via the @option{-fdump-ada-spec} switch).
1056 @itemx @var{file}.cp
1057 @itemx @var{file}.cxx
1058 @itemx @var{file}.cpp
1059 @itemx @var{file}.CPP
1060 @itemx @var{file}.c++
1062 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1063 the last two letters must both be literally @samp{x}. Likewise,
1064 @samp{.C} refers to a literal capital C@.
1068 Objective-C++ source code which must be preprocessed.
1070 @item @var{file}.mii
1071 Objective-C++ source code which should not be preprocessed.
1075 @itemx @var{file}.hp
1076 @itemx @var{file}.hxx
1077 @itemx @var{file}.hpp
1078 @itemx @var{file}.HPP
1079 @itemx @var{file}.h++
1080 @itemx @var{file}.tcc
1081 C++ header file to be turned into a precompiled header or Ada spec.
1084 @itemx @var{file}.for
1085 @itemx @var{file}.ftn
1086 Fixed form Fortran source code which should not be preprocessed.
1089 @itemx @var{file}.FOR
1090 @itemx @var{file}.fpp
1091 @itemx @var{file}.FPP
1092 @itemx @var{file}.FTN
1093 Fixed form Fortran source code which must be preprocessed (with the traditional
1096 @item @var{file}.f90
1097 @itemx @var{file}.f95
1098 @itemx @var{file}.f03
1099 @itemx @var{file}.f08
1100 Free form Fortran source code which should not be preprocessed.
1102 @item @var{file}.F90
1103 @itemx @var{file}.F95
1104 @itemx @var{file}.F03
1105 @itemx @var{file}.F08
1106 Free form Fortran source code which must be preprocessed (with the
1107 traditional preprocessor).
1112 @c FIXME: Descriptions of Java file types.
1118 @item @var{file}.ads
1119 Ada source code file which contains a library unit declaration (a
1120 declaration of a package, subprogram, or generic, or a generic
1121 instantiation), or a library unit renaming declaration (a package,
1122 generic, or subprogram renaming declaration). Such files are also
1125 @item @var{file}.adb
1126 Ada source code file containing a library unit body (a subprogram or
1127 package body). Such files are also called @dfn{bodies}.
1129 @c GCC also knows about some suffixes for languages not yet included:
1140 @itemx @var{file}.sx
1141 Assembler code which must be preprocessed.
1144 An object file to be fed straight into linking.
1145 Any file name with no recognized suffix is treated this way.
1149 You can specify the input language explicitly with the @option{-x} option:
1152 @item -x @var{language}
1153 Specify explicitly the @var{language} for the following input files
1154 (rather than letting the compiler choose a default based on the file
1155 name suffix). This option applies to all following input files until
1156 the next @option{-x} option. Possible values for @var{language} are:
1158 c c-header cpp-output
1159 c++ c++-header c++-cpp-output
1160 objective-c objective-c-header objective-c-cpp-output
1161 objective-c++ objective-c++-header objective-c++-cpp-output
1162 assembler assembler-with-cpp
1164 f77 f77-cpp-input f95 f95-cpp-input
1170 Turn off any specification of a language, so that subsequent files are
1171 handled according to their file name suffixes (as they are if @option{-x}
1172 has not been used at all).
1174 @item -pass-exit-codes
1175 @opindex pass-exit-codes
1176 Normally the @command{gcc} program will exit with the code of 1 if any
1177 phase of the compiler returns a non-success return code. If you specify
1178 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1179 numerically highest error produced by any phase that returned an error
1180 indication. The C, C++, and Fortran frontends return 4, if an internal
1181 compiler error is encountered.
1184 If you only want some of the stages of compilation, you can use
1185 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1186 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1187 @command{gcc} is to stop. Note that some combinations (for example,
1188 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1193 Compile or assemble the source files, but do not link. The linking
1194 stage simply is not done. The ultimate output is in the form of an
1195 object file for each source file.
1197 By default, the object file name for a source file is made by replacing
1198 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1200 Unrecognized input files, not requiring compilation or assembly, are
1205 Stop after the stage of compilation proper; do not assemble. The output
1206 is in the form of an assembler code file for each non-assembler input
1209 By default, the assembler file name for a source file is made by
1210 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1212 Input files that don't require compilation are ignored.
1216 Stop after the preprocessing stage; do not run the compiler proper. The
1217 output is in the form of preprocessed source code, which is sent to the
1220 Input files which don't require preprocessing are ignored.
1222 @cindex output file option
1225 Place output in file @var{file}. This applies regardless to whatever
1226 sort of output is being produced, whether it be an executable file,
1227 an object file, an assembler file or preprocessed C code.
1229 If @option{-o} is not specified, the default is to put an executable
1230 file in @file{a.out}, the object file for
1231 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1232 assembler file in @file{@var{source}.s}, a precompiled header file in
1233 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1238 Print (on standard error output) the commands executed to run the stages
1239 of compilation. Also print the version number of the compiler driver
1240 program and of the preprocessor and the compiler proper.
1244 Like @option{-v} except the commands are not executed and arguments
1245 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1246 This is useful for shell scripts to capture the driver-generated command lines.
1250 Use pipes rather than temporary files for communication between the
1251 various stages of compilation. This fails to work on some systems where
1252 the assembler is unable to read from a pipe; but the GNU assembler has
1257 Print (on the standard output) a description of the command-line options
1258 understood by @command{gcc}. If the @option{-v} option is also specified
1259 then @option{--help} will also be passed on to the various processes
1260 invoked by @command{gcc}, so that they can display the command-line options
1261 they accept. If the @option{-Wextra} option has also been specified
1262 (prior to the @option{--help} option), then command-line options which
1263 have no documentation associated with them will also be displayed.
1266 @opindex target-help
1267 Print (on the standard output) a description of target-specific command-line
1268 options for each tool. For some targets extra target-specific
1269 information may also be printed.
1271 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1272 Print (on the standard output) a description of the command-line
1273 options understood by the compiler that fit into all specified classes
1274 and qualifiers. These are the supported classes:
1277 @item @samp{optimizers}
1278 This will display all of the optimization options supported by the
1281 @item @samp{warnings}
1282 This will display all of the options controlling warning messages
1283 produced by the compiler.
1286 This will display target-specific options. Unlike the
1287 @option{--target-help} option however, target-specific options of the
1288 linker and assembler will not be displayed. This is because those
1289 tools do not currently support the extended @option{--help=} syntax.
1292 This will display the values recognized by the @option{--param}
1295 @item @var{language}
1296 This will display the options supported for @var{language}, where
1297 @var{language} is the name of one of the languages supported in this
1301 This will display the options that are common to all languages.
1304 These are the supported qualifiers:
1307 @item @samp{undocumented}
1308 Display only those options which are undocumented.
1311 Display options which take an argument that appears after an equal
1312 sign in the same continuous piece of text, such as:
1313 @samp{--help=target}.
1315 @item @samp{separate}
1316 Display options which take an argument that appears as a separate word
1317 following the original option, such as: @samp{-o output-file}.
1320 Thus for example to display all the undocumented target-specific
1321 switches supported by the compiler the following can be used:
1324 --help=target,undocumented
1327 The sense of a qualifier can be inverted by prefixing it with the
1328 @samp{^} character, so for example to display all binary warning
1329 options (i.e., ones that are either on or off and that do not take an
1330 argument), which have a description the following can be used:
1333 --help=warnings,^joined,^undocumented
1336 The argument to @option{--help=} should not consist solely of inverted
1339 Combining several classes is possible, although this usually
1340 restricts the output by so much that there is nothing to display. One
1341 case where it does work however is when one of the classes is
1342 @var{target}. So for example to display all the target-specific
1343 optimization options the following can be used:
1346 --help=target,optimizers
1349 The @option{--help=} option can be repeated on the command line. Each
1350 successive use will display its requested class of options, skipping
1351 those that have already been displayed.
1353 If the @option{-Q} option appears on the command line before the
1354 @option{--help=} option, then the descriptive text displayed by
1355 @option{--help=} is changed. Instead of describing the displayed
1356 options, an indication is given as to whether the option is enabled,
1357 disabled or set to a specific value (assuming that the compiler
1358 knows this at the point where the @option{--help=} option is used).
1360 Here is a truncated example from the ARM port of @command{gcc}:
1363 % gcc -Q -mabi=2 --help=target -c
1364 The following options are target specific:
1366 -mabort-on-noreturn [disabled]
1370 The output is sensitive to the effects of previous command-line
1371 options, so for example it is possible to find out which optimizations
1372 are enabled at @option{-O2} by using:
1375 -Q -O2 --help=optimizers
1378 Alternatively you can discover which binary optimizations are enabled
1379 by @option{-O3} by using:
1382 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1383 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1384 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1387 @item -no-canonical-prefixes
1388 @opindex no-canonical-prefixes
1389 Do not expand any symbolic links, resolve references to @samp{/../}
1390 or @samp{/./}, or make the path absolute when generating a relative
1395 Display the version number and copyrights of the invoked GCC@.
1399 Invoke all subcommands under a wrapper program. The name of the
1400 wrapper program and its parameters are passed as a comma separated
1404 gcc -c t.c -wrapper gdb,--args
1407 This will invoke all subprograms of @command{gcc} under
1408 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1409 @samp{gdb --args cc1 @dots{}}.
1411 @item -fplugin=@var{name}.so
1412 Load the plugin code in file @var{name}.so, assumed to be a
1413 shared object to be dlopen'd by the compiler. The base name of
1414 the shared object file is used to identify the plugin for the
1415 purposes of argument parsing (See
1416 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1417 Each plugin should define the callback functions specified in the
1420 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1421 Define an argument called @var{key} with a value of @var{value}
1422 for the plugin called @var{name}.
1424 @item -fdump-ada-spec@r{[}-slim@r{]}
1425 For C and C++ source and include files, generate corresponding Ada
1426 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1427 GNAT User's Guide}, which provides detailed documentation on this feature.
1429 @item -fdump-go-spec=@var{file}
1430 For input files in any language, generate corresponding Go
1431 declarations in @var{file}. This generates Go @code{const},
1432 @code{type}, @code{var}, and @code{func} declarations which may be a
1433 useful way to start writing a Go interface to code written in some
1436 @include @value{srcdir}/../libiberty/at-file.texi
1440 @section Compiling C++ Programs
1442 @cindex suffixes for C++ source
1443 @cindex C++ source file suffixes
1444 C++ source files conventionally use one of the suffixes @samp{.C},
1445 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1446 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1447 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1448 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1449 files with these names and compiles them as C++ programs even if you
1450 call the compiler the same way as for compiling C programs (usually
1451 with the name @command{gcc}).
1455 However, the use of @command{gcc} does not add the C++ library.
1456 @command{g++} is a program that calls GCC and treats @samp{.c},
1457 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1458 files unless @option{-x} is used, and automatically specifies linking
1459 against the C++ library. This program is also useful when
1460 precompiling a C header file with a @samp{.h} extension for use in C++
1461 compilations. On many systems, @command{g++} is also installed with
1462 the name @command{c++}.
1464 @cindex invoking @command{g++}
1465 When you compile C++ programs, you may specify many of the same
1466 command-line options that you use for compiling programs in any
1467 language; or command-line options meaningful for C and related
1468 languages; or options that are meaningful only for C++ programs.
1469 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1470 explanations of options for languages related to C@.
1471 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1472 explanations of options that are meaningful only for C++ programs.
1474 @node C Dialect Options
1475 @section Options Controlling C Dialect
1476 @cindex dialect options
1477 @cindex language dialect options
1478 @cindex options, dialect
1480 The following options control the dialect of C (or languages derived
1481 from C, such as C++, Objective-C and Objective-C++) that the compiler
1485 @cindex ANSI support
1489 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1490 equivalent to @samp{-std=c++98}.
1492 This turns off certain features of GCC that are incompatible with ISO
1493 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1494 such as the @code{asm} and @code{typeof} keywords, and
1495 predefined macros such as @code{unix} and @code{vax} that identify the
1496 type of system you are using. It also enables the undesirable and
1497 rarely used ISO trigraph feature. For the C compiler,
1498 it disables recognition of C++ style @samp{//} comments as well as
1499 the @code{inline} keyword.
1501 The alternate keywords @code{__asm__}, @code{__extension__},
1502 @code{__inline__} and @code{__typeof__} continue to work despite
1503 @option{-ansi}. You would not want to use them in an ISO C program, of
1504 course, but it is useful to put them in header files that might be included
1505 in compilations done with @option{-ansi}. Alternate predefined macros
1506 such as @code{__unix__} and @code{__vax__} are also available, with or
1507 without @option{-ansi}.
1509 The @option{-ansi} option does not cause non-ISO programs to be
1510 rejected gratuitously. For that, @option{-pedantic} is required in
1511 addition to @option{-ansi}. @xref{Warning Options}.
1513 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1514 option is used. Some header files may notice this macro and refrain
1515 from declaring certain functions or defining certain macros that the
1516 ISO standard doesn't call for; this is to avoid interfering with any
1517 programs that might use these names for other things.
1519 Functions that would normally be built in but do not have semantics
1520 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1521 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1522 built-in functions provided by GCC}, for details of the functions
1527 Determine the language standard. @xref{Standards,,Language Standards
1528 Supported by GCC}, for details of these standard versions. This option
1529 is currently only supported when compiling C or C++.
1531 The compiler can accept several base standards, such as @samp{c90} or
1532 @samp{c++98}, and GNU dialects of those standards, such as
1533 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1534 compiler will accept all programs following that standard and those
1535 using GNU extensions that do not contradict it. For example,
1536 @samp{-std=c90} turns off certain features of GCC that are
1537 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1538 keywords, but not other GNU extensions that do not have a meaning in
1539 ISO C90, such as omitting the middle term of a @code{?:}
1540 expression. On the other hand, by specifying a GNU dialect of a
1541 standard, all features the compiler support are enabled, even when
1542 those features change the meaning of the base standard and some
1543 strict-conforming programs may be rejected. The particular standard
1544 is used by @option{-pedantic} to identify which features are GNU
1545 extensions given that version of the standard. For example
1546 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1547 comments, while @samp{-std=gnu99 -pedantic} would not.
1549 A value for this option must be provided; possible values are
1555 Support all ISO C90 programs (certain GNU extensions that conflict
1556 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1558 @item iso9899:199409
1559 ISO C90 as modified in amendment 1.
1565 ISO C99. Note that this standard is not yet fully supported; see
1566 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1567 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1572 ISO C11, the 2011 revision of the ISO C standard.
1573 Support is incomplete and experimental. The name @samp{c1x} is
1578 GNU dialect of ISO C90 (including some C99 features). This
1579 is the default for C code.
1583 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1584 this will become the default. The name @samp{gnu9x} is deprecated.
1588 GNU dialect of ISO C11. Support is incomplete and experimental. The
1589 name @samp{gnu1x} is deprecated.
1592 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1596 GNU dialect of @option{-std=c++98}. This is the default for
1600 The 2011 ISO C++ standard plus amendments. Support for C++11 is still
1601 experimental, and may change in incompatible ways in future releases.
1604 GNU dialect of @option{-std=c++11}. Support for C++11 is still
1605 experimental, and may change in incompatible ways in future releases.
1608 @item -fgnu89-inline
1609 @opindex fgnu89-inline
1610 The option @option{-fgnu89-inline} tells GCC to use the traditional
1611 GNU semantics for @code{inline} functions when in C99 mode.
1612 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1613 is accepted and ignored by GCC versions 4.1.3 up to but not including
1614 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1615 C99 mode. Using this option is roughly equivalent to adding the
1616 @code{gnu_inline} function attribute to all inline functions
1617 (@pxref{Function Attributes}).
1619 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1620 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1621 specifies the default behavior). This option was first supported in
1622 GCC 4.3. This option is not supported in @option{-std=c90} or
1623 @option{-std=gnu90} mode.
1625 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1626 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1627 in effect for @code{inline} functions. @xref{Common Predefined
1628 Macros,,,cpp,The C Preprocessor}.
1630 @item -aux-info @var{filename}
1632 Output to the given filename prototyped declarations for all functions
1633 declared and/or defined in a translation unit, including those in header
1634 files. This option is silently ignored in any language other than C@.
1636 Besides declarations, the file indicates, in comments, the origin of
1637 each declaration (source file and line), whether the declaration was
1638 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1639 @samp{O} for old, respectively, in the first character after the line
1640 number and the colon), and whether it came from a declaration or a
1641 definition (@samp{C} or @samp{F}, respectively, in the following
1642 character). In the case of function definitions, a K&R-style list of
1643 arguments followed by their declarations is also provided, inside
1644 comments, after the declaration.
1646 @item -fallow-parameterless-variadic-functions
1647 Accept variadic functions without named parameters.
1649 Although it is possible to define such a function, this is not very
1650 useful as it is not possible to read the arguments. This is only
1651 supported for C as this construct is allowed by C++.
1655 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1656 keyword, so that code can use these words as identifiers. You can use
1657 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1658 instead. @option{-ansi} implies @option{-fno-asm}.
1660 In C++, this switch only affects the @code{typeof} keyword, since
1661 @code{asm} and @code{inline} are standard keywords. You may want to
1662 use the @option{-fno-gnu-keywords} flag instead, which has the same
1663 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1664 switch only affects the @code{asm} and @code{typeof} keywords, since
1665 @code{inline} is a standard keyword in ISO C99.
1668 @itemx -fno-builtin-@var{function}
1669 @opindex fno-builtin
1670 @cindex built-in functions
1671 Don't recognize built-in functions that do not begin with
1672 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1673 functions provided by GCC}, for details of the functions affected,
1674 including those which are not built-in functions when @option{-ansi} or
1675 @option{-std} options for strict ISO C conformance are used because they
1676 do not have an ISO standard meaning.
1678 GCC normally generates special code to handle certain built-in functions
1679 more efficiently; for instance, calls to @code{alloca} may become single
1680 instructions that adjust the stack directly, and calls to @code{memcpy}
1681 may become inline copy loops. The resulting code is often both smaller
1682 and faster, but since the function calls no longer appear as such, you
1683 cannot set a breakpoint on those calls, nor can you change the behavior
1684 of the functions by linking with a different library. In addition,
1685 when a function is recognized as a built-in function, GCC may use
1686 information about that function to warn about problems with calls to
1687 that function, or to generate more efficient code, even if the
1688 resulting code still contains calls to that function. For example,
1689 warnings are given with @option{-Wformat} for bad calls to
1690 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1691 known not to modify global memory.
1693 With the @option{-fno-builtin-@var{function}} option
1694 only the built-in function @var{function} is
1695 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1696 function is named that is not built-in in this version of GCC, this
1697 option is ignored. There is no corresponding
1698 @option{-fbuiltin-@var{function}} option; if you wish to enable
1699 built-in functions selectively when using @option{-fno-builtin} or
1700 @option{-ffreestanding}, you may define macros such as:
1703 #define abs(n) __builtin_abs ((n))
1704 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1709 @cindex hosted environment
1711 Assert that compilation takes place in a hosted environment. This implies
1712 @option{-fbuiltin}. A hosted environment is one in which the
1713 entire standard library is available, and in which @code{main} has a return
1714 type of @code{int}. Examples are nearly everything except a kernel.
1715 This is equivalent to @option{-fno-freestanding}.
1717 @item -ffreestanding
1718 @opindex ffreestanding
1719 @cindex hosted environment
1721 Assert that compilation takes place in a freestanding environment. This
1722 implies @option{-fno-builtin}. A freestanding environment
1723 is one in which the standard library may not exist, and program startup may
1724 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1725 This is equivalent to @option{-fno-hosted}.
1727 @xref{Standards,,Language Standards Supported by GCC}, for details of
1728 freestanding and hosted environments.
1732 @cindex OpenMP parallel
1733 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1734 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1735 compiler generates parallel code according to the OpenMP Application
1736 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1737 implies @option{-pthread}, and thus is only supported on targets that
1738 have support for @option{-pthread}.
1742 When the option @option{-fgnu-tm} is specified, the compiler will
1743 generate code for the Linux variant of Intel's current Transactional
1744 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1745 an experimental feature whose interface may change in future versions
1746 of GCC, as the official specification changes. Please note that not
1747 all architectures are supported for this feature.
1749 For more information on GCC's support for transactional memory,
1750 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1751 Transactional Memory Library}.
1753 Note that the transactional memory feature is not supported with
1754 non-call exceptions (@option{-fnon-call-exceptions}).
1756 @item -fms-extensions
1757 @opindex fms-extensions
1758 Accept some non-standard constructs used in Microsoft header files.
1760 In C++ code, this allows member names in structures to be similar
1761 to previous types declarations.
1770 Some cases of unnamed fields in structures and unions are only
1771 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1772 fields within structs/unions}, for details.
1774 @item -fplan9-extensions
1775 Accept some non-standard constructs used in Plan 9 code.
1777 This enables @option{-fms-extensions}, permits passing pointers to
1778 structures with anonymous fields to functions which expect pointers to
1779 elements of the type of the field, and permits referring to anonymous
1780 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1781 struct/union fields within structs/unions}, for details. This is only
1782 supported for C, not C++.
1786 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1787 options for strict ISO C conformance) implies @option{-trigraphs}.
1789 @item -no-integrated-cpp
1790 @opindex no-integrated-cpp
1791 Performs a compilation in two passes: preprocessing and compiling. This
1792 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1793 @option{-B} option. The user supplied compilation step can then add in
1794 an additional preprocessing step after normal preprocessing but before
1795 compiling. The default is to use the integrated cpp (internal cpp)
1797 The semantics of this option will change if "cc1", "cc1plus", and
1798 "cc1obj" are merged.
1800 @cindex traditional C language
1801 @cindex C language, traditional
1803 @itemx -traditional-cpp
1804 @opindex traditional-cpp
1805 @opindex traditional
1806 Formerly, these options caused GCC to attempt to emulate a pre-standard
1807 C compiler. They are now only supported with the @option{-E} switch.
1808 The preprocessor continues to support a pre-standard mode. See the GNU
1809 CPP manual for details.
1811 @item -fcond-mismatch
1812 @opindex fcond-mismatch
1813 Allow conditional expressions with mismatched types in the second and
1814 third arguments. The value of such an expression is void. This option
1815 is not supported for C++.
1817 @item -flax-vector-conversions
1818 @opindex flax-vector-conversions
1819 Allow implicit conversions between vectors with differing numbers of
1820 elements and/or incompatible element types. This option should not be
1823 @item -funsigned-char
1824 @opindex funsigned-char
1825 Let the type @code{char} be unsigned, like @code{unsigned char}.
1827 Each kind of machine has a default for what @code{char} should
1828 be. It is either like @code{unsigned char} by default or like
1829 @code{signed char} by default.
1831 Ideally, a portable program should always use @code{signed char} or
1832 @code{unsigned char} when it depends on the signedness of an object.
1833 But many programs have been written to use plain @code{char} and
1834 expect it to be signed, or expect it to be unsigned, depending on the
1835 machines they were written for. This option, and its inverse, let you
1836 make such a program work with the opposite default.
1838 The type @code{char} is always a distinct type from each of
1839 @code{signed char} or @code{unsigned char}, even though its behavior
1840 is always just like one of those two.
1843 @opindex fsigned-char
1844 Let the type @code{char} be signed, like @code{signed char}.
1846 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1847 the negative form of @option{-funsigned-char}. Likewise, the option
1848 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1850 @item -fsigned-bitfields
1851 @itemx -funsigned-bitfields
1852 @itemx -fno-signed-bitfields
1853 @itemx -fno-unsigned-bitfields
1854 @opindex fsigned-bitfields
1855 @opindex funsigned-bitfields
1856 @opindex fno-signed-bitfields
1857 @opindex fno-unsigned-bitfields
1858 These options control whether a bit-field is signed or unsigned, when the
1859 declaration does not use either @code{signed} or @code{unsigned}. By
1860 default, such a bit-field is signed, because this is consistent: the
1861 basic integer types such as @code{int} are signed types.
1864 @node C++ Dialect Options
1865 @section Options Controlling C++ Dialect
1867 @cindex compiler options, C++
1868 @cindex C++ options, command-line
1869 @cindex options, C++
1870 This section describes the command-line options that are only meaningful
1871 for C++ programs; but you can also use most of the GNU compiler options
1872 regardless of what language your program is in. For example, you
1873 might compile a file @code{firstClass.C} like this:
1876 g++ -g -frepo -O -c firstClass.C
1880 In this example, only @option{-frepo} is an option meant
1881 only for C++ programs; you can use the other options with any
1882 language supported by GCC@.
1884 Here is a list of options that are @emph{only} for compiling C++ programs:
1888 @item -fabi-version=@var{n}
1889 @opindex fabi-version
1890 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1891 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1892 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1893 the version that conforms most closely to the C++ ABI specification.
1894 Therefore, the ABI obtained using version 0 will change as ABI bugs
1897 The default is version 2.
1899 Version 3 corrects an error in mangling a constant address as a
1902 Version 4, which first appeared in G++ 4.5, implements a standard
1903 mangling for vector types.
1905 Version 5, which first appeared in G++ 4.6, corrects the mangling of
1906 attribute const/volatile on function pointer types, decltype of a
1907 plain decl, and use of a function parameter in the declaration of
1910 Version 6, which first appeared in G++ 4.7, corrects the promotion
1911 behavior of C++11 scoped enums and the mangling of template argument
1912 packs, const/static_cast, prefix ++ and --, and a class scope function
1913 used as a template argument.
1915 See also @option{-Wabi}.
1917 @item -fno-access-control
1918 @opindex fno-access-control
1919 Turn off all access checking. This switch is mainly useful for working
1920 around bugs in the access control code.
1924 Check that the pointer returned by @code{operator new} is non-null
1925 before attempting to modify the storage allocated. This check is
1926 normally unnecessary because the C++ standard specifies that
1927 @code{operator new} will only return @code{0} if it is declared
1928 @samp{throw()}, in which case the compiler will always check the
1929 return value even without this option. In all other cases, when
1930 @code{operator new} has a non-empty exception specification, memory
1931 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1932 @samp{new (nothrow)}.
1934 @item -fconserve-space
1935 @opindex fconserve-space
1936 Put uninitialized or run-time-initialized global variables into the
1937 common segment, as C does. This saves space in the executable at the
1938 cost of not diagnosing duplicate definitions. If you compile with this
1939 flag and your program mysteriously crashes after @code{main()} has
1940 completed, you may have an object that is being destroyed twice because
1941 two definitions were merged.
1943 This option is no longer useful on most targets, now that support has
1944 been added for putting variables into BSS without making them common.
1946 @item -fconstexpr-depth=@var{n}
1947 @opindex fconstexpr-depth
1948 Set the maximum nested evaluation depth for C++11 constexpr functions
1949 to @var{n}. A limit is needed to detect endless recursion during
1950 constant expression evaluation. The minimum specified by the standard
1953 @item -fdeduce-init-list
1954 @opindex fdeduce-init-list
1955 Enable deduction of a template type parameter as
1956 std::initializer_list from a brace-enclosed initializer list, i.e.
1959 template <class T> auto forward(T t) -> decltype (realfn (t))
1966 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1970 This deduction was implemented as a possible extension to the
1971 originally proposed semantics for the C++11 standard, but was not part
1972 of the final standard, so it is disabled by default. This option is
1973 deprecated, and may be removed in a future version of G++.
1975 @item -ffriend-injection
1976 @opindex ffriend-injection
1977 Inject friend functions into the enclosing namespace, so that they are
1978 visible outside the scope of the class in which they are declared.
1979 Friend functions were documented to work this way in the old Annotated
1980 C++ Reference Manual, and versions of G++ before 4.1 always worked
1981 that way. However, in ISO C++ a friend function which is not declared
1982 in an enclosing scope can only be found using argument dependent
1983 lookup. This option causes friends to be injected as they were in
1986 This option is for compatibility, and may be removed in a future
1989 @item -fno-elide-constructors
1990 @opindex fno-elide-constructors
1991 The C++ standard allows an implementation to omit creating a temporary
1992 which is only used to initialize another object of the same type.
1993 Specifying this option disables that optimization, and forces G++ to
1994 call the copy constructor in all cases.
1996 @item -fno-enforce-eh-specs
1997 @opindex fno-enforce-eh-specs
1998 Don't generate code to check for violation of exception specifications
1999 at run time. This option violates the C++ standard, but may be useful
2000 for reducing code size in production builds, much like defining
2001 @samp{NDEBUG}. This does not give user code permission to throw
2002 exceptions in violation of the exception specifications; the compiler
2003 will still optimize based on the specifications, so throwing an
2004 unexpected exception will result in undefined behavior.
2007 @itemx -fno-for-scope
2009 @opindex fno-for-scope
2010 If @option{-ffor-scope} is specified, the scope of variables declared in
2011 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2012 as specified by the C++ standard.
2013 If @option{-fno-for-scope} is specified, the scope of variables declared in
2014 a @i{for-init-statement} extends to the end of the enclosing scope,
2015 as was the case in old versions of G++, and other (traditional)
2016 implementations of C++.
2018 The default if neither flag is given to follow the standard,
2019 but to allow and give a warning for old-style code that would
2020 otherwise be invalid, or have different behavior.
2022 @item -fno-gnu-keywords
2023 @opindex fno-gnu-keywords
2024 Do not recognize @code{typeof} as a keyword, so that code can use this
2025 word as an identifier. You can use the keyword @code{__typeof__} instead.
2026 @option{-ansi} implies @option{-fno-gnu-keywords}.
2028 @item -fno-implicit-templates
2029 @opindex fno-implicit-templates
2030 Never emit code for non-inline templates which are instantiated
2031 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2032 @xref{Template Instantiation}, for more information.
2034 @item -fno-implicit-inline-templates
2035 @opindex fno-implicit-inline-templates
2036 Don't emit code for implicit instantiations of inline templates, either.
2037 The default is to handle inlines differently so that compiles with and
2038 without optimization will need the same set of explicit instantiations.
2040 @item -fno-implement-inlines
2041 @opindex fno-implement-inlines
2042 To save space, do not emit out-of-line copies of inline functions
2043 controlled by @samp{#pragma implementation}. This will cause linker
2044 errors if these functions are not inlined everywhere they are called.
2046 @item -fms-extensions
2047 @opindex fms-extensions
2048 Disable pedantic warnings about constructs used in MFC, such as implicit
2049 int and getting a pointer to member function via non-standard syntax.
2051 @item -fno-nonansi-builtins
2052 @opindex fno-nonansi-builtins
2053 Disable built-in declarations of functions that are not mandated by
2054 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2055 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2058 @opindex fnothrow-opt
2059 Treat a @code{throw()} exception specification as though it were a
2060 @code{noexcept} specification to reduce or eliminate the text size
2061 overhead relative to a function with no exception specification. If
2062 the function has local variables of types with non-trivial
2063 destructors, the exception specification will actually make the
2064 function smaller because the EH cleanups for those variables can be
2065 optimized away. The semantic effect is that an exception thrown out of
2066 a function with such an exception specification will result in a call
2067 to @code{terminate} rather than @code{unexpected}.
2069 @item -fno-operator-names
2070 @opindex fno-operator-names
2071 Do not treat the operator name keywords @code{and}, @code{bitand},
2072 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2073 synonyms as keywords.
2075 @item -fno-optional-diags
2076 @opindex fno-optional-diags
2077 Disable diagnostics that the standard says a compiler does not need to
2078 issue. Currently, the only such diagnostic issued by G++ is the one for
2079 a name having multiple meanings within a class.
2082 @opindex fpermissive
2083 Downgrade some diagnostics about nonconformant code from errors to
2084 warnings. Thus, using @option{-fpermissive} will allow some
2085 nonconforming code to compile.
2087 @item -fno-pretty-templates
2088 @opindex fno-pretty-templates
2089 When an error message refers to a specialization of a function
2090 template, the compiler will normally print the signature of the
2091 template followed by the template arguments and any typedefs or
2092 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2093 rather than @code{void f(int)}) so that it's clear which template is
2094 involved. When an error message refers to a specialization of a class
2095 template, the compiler will omit any template arguments which match
2096 the default template arguments for that template. If either of these
2097 behaviors make it harder to understand the error message rather than
2098 easier, using @option{-fno-pretty-templates} will disable them.
2102 Enable automatic template instantiation at link time. This option also
2103 implies @option{-fno-implicit-templates}. @xref{Template
2104 Instantiation}, for more information.
2108 Disable generation of information about every class with virtual
2109 functions for use by the C++ run-time type identification features
2110 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2111 of the language, you can save some space by using this flag. Note that
2112 exception handling uses the same information, but it will generate it as
2113 needed. The @samp{dynamic_cast} operator can still be used for casts that
2114 do not require run-time type information, i.e.@: casts to @code{void *} or to
2115 unambiguous base classes.
2119 Emit statistics about front-end processing at the end of the compilation.
2120 This information is generally only useful to the G++ development team.
2122 @item -fstrict-enums
2123 @opindex fstrict-enums
2124 Allow the compiler to optimize using the assumption that a value of
2125 enumeration type can only be one of the values of the enumeration (as
2126 defined in the C++ standard; basically, a value which can be
2127 represented in the minimum number of bits needed to represent all the
2128 enumerators). This assumption may not be valid if the program uses a
2129 cast to convert an arbitrary integer value to the enumeration type.
2131 @item -ftemplate-depth=@var{n}
2132 @opindex ftemplate-depth
2133 Set the maximum instantiation depth for template classes to @var{n}.
2134 A limit on the template instantiation depth is needed to detect
2135 endless recursions during template class instantiation. ANSI/ISO C++
2136 conforming programs must not rely on a maximum depth greater than 17
2137 (changed to 1024 in C++11). The default value is 900, as the compiler
2138 can run out of stack space before hitting 1024 in some situations.
2140 @item -fno-threadsafe-statics
2141 @opindex fno-threadsafe-statics
2142 Do not emit the extra code to use the routines specified in the C++
2143 ABI for thread-safe initialization of local statics. You can use this
2144 option to reduce code size slightly in code that doesn't need to be
2147 @item -fuse-cxa-atexit
2148 @opindex fuse-cxa-atexit
2149 Register destructors for objects with static storage duration with the
2150 @code{__cxa_atexit} function rather than the @code{atexit} function.
2151 This option is required for fully standards-compliant handling of static
2152 destructors, but will only work if your C library supports
2153 @code{__cxa_atexit}.
2155 @item -fno-use-cxa-get-exception-ptr
2156 @opindex fno-use-cxa-get-exception-ptr
2157 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2158 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2159 if the runtime routine is not available.
2161 @item -fvisibility-inlines-hidden
2162 @opindex fvisibility-inlines-hidden
2163 This switch declares that the user does not attempt to compare
2164 pointers to inline functions or methods where the addresses of the two functions
2165 were taken in different shared objects.
2167 The effect of this is that GCC may, effectively, mark inline methods with
2168 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2169 appear in the export table of a DSO and do not require a PLT indirection
2170 when used within the DSO@. Enabling this option can have a dramatic effect
2171 on load and link times of a DSO as it massively reduces the size of the
2172 dynamic export table when the library makes heavy use of templates.
2174 The behavior of this switch is not quite the same as marking the
2175 methods as hidden directly, because it does not affect static variables
2176 local to the function or cause the compiler to deduce that
2177 the function is defined in only one shared object.
2179 You may mark a method as having a visibility explicitly to negate the
2180 effect of the switch for that method. For example, if you do want to
2181 compare pointers to a particular inline method, you might mark it as
2182 having default visibility. Marking the enclosing class with explicit
2183 visibility will have no effect.
2185 Explicitly instantiated inline methods are unaffected by this option
2186 as their linkage might otherwise cross a shared library boundary.
2187 @xref{Template Instantiation}.
2189 @item -fvisibility-ms-compat
2190 @opindex fvisibility-ms-compat
2191 This flag attempts to use visibility settings to make GCC's C++
2192 linkage model compatible with that of Microsoft Visual Studio.
2194 The flag makes these changes to GCC's linkage model:
2198 It sets the default visibility to @code{hidden}, like
2199 @option{-fvisibility=hidden}.
2202 Types, but not their members, are not hidden by default.
2205 The One Definition Rule is relaxed for types without explicit
2206 visibility specifications which are defined in more than one different
2207 shared object: those declarations are permitted if they would have
2208 been permitted when this option was not used.
2211 In new code it is better to use @option{-fvisibility=hidden} and
2212 export those classes which are intended to be externally visible.
2213 Unfortunately it is possible for code to rely, perhaps accidentally,
2214 on the Visual Studio behavior.
2216 Among the consequences of these changes are that static data members
2217 of the same type with the same name but defined in different shared
2218 objects will be different, so changing one will not change the other;
2219 and that pointers to function members defined in different shared
2220 objects may not compare equal. When this flag is given, it is a
2221 violation of the ODR to define types with the same name differently.
2225 Do not use weak symbol support, even if it is provided by the linker.
2226 By default, G++ will use weak symbols if they are available. This
2227 option exists only for testing, and should not be used by end-users;
2228 it will result in inferior code and has no benefits. This option may
2229 be removed in a future release of G++.
2233 Do not search for header files in the standard directories specific to
2234 C++, but do still search the other standard directories. (This option
2235 is used when building the C++ library.)
2238 In addition, these optimization, warning, and code generation options
2239 have meanings only for C++ programs:
2242 @item -fno-default-inline
2243 @opindex fno-default-inline
2244 Do not assume @samp{inline} for functions defined inside a class scope.
2245 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2246 functions will have linkage like inline functions; they just won't be
2249 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2252 Warn when G++ generates code that is probably not compatible with the
2253 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2254 all such cases, there are probably some cases that are not warned about,
2255 even though G++ is generating incompatible code. There may also be
2256 cases where warnings are emitted even though the code that is generated
2259 You should rewrite your code to avoid these warnings if you are
2260 concerned about the fact that code generated by G++ may not be binary
2261 compatible with code generated by other compilers.
2263 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2268 A template with a non-type template parameter of reference type is
2269 mangled incorrectly:
2272 template <int &> struct S @{@};
2276 This is fixed in @option{-fabi-version=3}.
2279 SIMD vector types declared using @code{__attribute ((vector_size))} are
2280 mangled in a non-standard way that does not allow for overloading of
2281 functions taking vectors of different sizes.
2283 The mangling is changed in @option{-fabi-version=4}.
2286 The known incompatibilities in @option{-fabi-version=1} include:
2291 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2292 pack data into the same byte as a base class. For example:
2295 struct A @{ virtual void f(); int f1 : 1; @};
2296 struct B : public A @{ int f2 : 1; @};
2300 In this case, G++ will place @code{B::f2} into the same byte
2301 as@code{A::f1}; other compilers will not. You can avoid this problem
2302 by explicitly padding @code{A} so that its size is a multiple of the
2303 byte size on your platform; that will cause G++ and other compilers to
2304 layout @code{B} identically.
2307 Incorrect handling of tail-padding for virtual bases. G++ does not use
2308 tail padding when laying out virtual bases. For example:
2311 struct A @{ virtual void f(); char c1; @};
2312 struct B @{ B(); char c2; @};
2313 struct C : public A, public virtual B @{@};
2317 In this case, G++ will not place @code{B} into the tail-padding for
2318 @code{A}; other compilers will. You can avoid this problem by
2319 explicitly padding @code{A} so that its size is a multiple of its
2320 alignment (ignoring virtual base classes); that will cause G++ and other
2321 compilers to layout @code{C} identically.
2324 Incorrect handling of bit-fields with declared widths greater than that
2325 of their underlying types, when the bit-fields appear in a union. For
2329 union U @{ int i : 4096; @};
2333 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2334 union too small by the number of bits in an @code{int}.
2337 Empty classes can be placed at incorrect offsets. For example:
2347 struct C : public B, public A @{@};
2351 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2352 it should be placed at offset zero. G++ mistakenly believes that the
2353 @code{A} data member of @code{B} is already at offset zero.
2356 Names of template functions whose types involve @code{typename} or
2357 template template parameters can be mangled incorrectly.
2360 template <typename Q>
2361 void f(typename Q::X) @{@}
2363 template <template <typename> class Q>
2364 void f(typename Q<int>::X) @{@}
2368 Instantiations of these templates may be mangled incorrectly.
2372 It also warns psABI related changes. The known psABI changes at this
2378 For SYSV/x86-64, when passing union with long double, it is changed to
2379 pass in memory as specified in psABI. For example:
2389 @code{union U} will always be passed in memory.
2393 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2394 @opindex Wctor-dtor-privacy
2395 @opindex Wno-ctor-dtor-privacy
2396 Warn when a class seems unusable because all the constructors or
2397 destructors in that class are private, and it has neither friends nor
2398 public static member functions.
2400 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2401 @opindex Wdelete-non-virtual-dtor
2402 @opindex Wno-delete-non-virtual-dtor
2403 Warn when @samp{delete} is used to destroy an instance of a class which
2404 has virtual functions and non-virtual destructor. It is unsafe to delete
2405 an instance of a derived class through a pointer to a base class if the
2406 base class does not have a virtual destructor. This warning is enabled
2409 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2411 @opindex Wno-narrowing
2412 Warn when a narrowing conversion prohibited by C++11 occurs within
2416 int i = @{ 2.2 @}; // error: narrowing from double to int
2419 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2421 With -std=c++11, @option{-Wno-narrowing} suppresses the diagnostic
2422 required by the standard. Note that this does not affect the meaning
2423 of well-formed code; narrowing conversions are still considered
2424 ill-formed in SFINAE context.
2426 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2428 @opindex Wno-noexcept
2429 Warn when a noexcept-expression evaluates to false because of a call
2430 to a function that does not have a non-throwing exception
2431 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2432 the compiler to never throw an exception.
2434 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2435 @opindex Wnon-virtual-dtor
2436 @opindex Wno-non-virtual-dtor
2437 Warn when a class has virtual functions and accessible non-virtual
2438 destructor, in which case it would be possible but unsafe to delete
2439 an instance of a derived class through a pointer to the base class.
2440 This warning is also enabled if -Weffc++ is specified.
2442 @item -Wreorder @r{(C++ and Objective-C++ only)}
2444 @opindex Wno-reorder
2445 @cindex reordering, warning
2446 @cindex warning for reordering of member initializers
2447 Warn when the order of member initializers given in the code does not
2448 match the order in which they must be executed. For instance:
2454 A(): j (0), i (1) @{ @}
2458 The compiler will rearrange the member initializers for @samp{i}
2459 and @samp{j} to match the declaration order of the members, emitting
2460 a warning to that effect. This warning is enabled by @option{-Wall}.
2463 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2466 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2469 Warn about violations of the following style guidelines from Scott Meyers'
2470 @cite{Effective C++} book:
2474 Item 11: Define a copy constructor and an assignment operator for classes
2475 with dynamically allocated memory.
2478 Item 12: Prefer initialization to assignment in constructors.
2481 Item 14: Make destructors virtual in base classes.
2484 Item 15: Have @code{operator=} return a reference to @code{*this}.
2487 Item 23: Don't try to return a reference when you must return an object.
2491 Also warn about violations of the following style guidelines from
2492 Scott Meyers' @cite{More Effective C++} book:
2496 Item 6: Distinguish between prefix and postfix forms of increment and
2497 decrement operators.
2500 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2504 When selecting this option, be aware that the standard library
2505 headers do not obey all of these guidelines; use @samp{grep -v}
2506 to filter out those warnings.
2508 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2509 @opindex Wstrict-null-sentinel
2510 @opindex Wno-strict-null-sentinel
2511 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2512 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2513 to @code{__null}. Although it is a null pointer constant not a null pointer,
2514 it is guaranteed to be of the same size as a pointer. But this use is
2515 not portable across different compilers.
2517 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2518 @opindex Wno-non-template-friend
2519 @opindex Wnon-template-friend
2520 Disable warnings when non-templatized friend functions are declared
2521 within a template. Since the advent of explicit template specification
2522 support in G++, if the name of the friend is an unqualified-id (i.e.,
2523 @samp{friend foo(int)}), the C++ language specification demands that the
2524 friend declare or define an ordinary, nontemplate function. (Section
2525 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2526 could be interpreted as a particular specialization of a templatized
2527 function. Because this non-conforming behavior is no longer the default
2528 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2529 check existing code for potential trouble spots and is on by default.
2530 This new compiler behavior can be turned off with
2531 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2532 but disables the helpful warning.
2534 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2535 @opindex Wold-style-cast
2536 @opindex Wno-old-style-cast
2537 Warn if an old-style (C-style) cast to a non-void type is used within
2538 a C++ program. The new-style casts (@samp{dynamic_cast},
2539 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2540 less vulnerable to unintended effects and much easier to search for.
2542 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2543 @opindex Woverloaded-virtual
2544 @opindex Wno-overloaded-virtual
2545 @cindex overloaded virtual function, warning
2546 @cindex warning for overloaded virtual function
2547 Warn when a function declaration hides virtual functions from a
2548 base class. For example, in:
2555 struct B: public A @{
2560 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2568 will fail to compile.
2570 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2571 @opindex Wno-pmf-conversions
2572 @opindex Wpmf-conversions
2573 Disable the diagnostic for converting a bound pointer to member function
2576 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2577 @opindex Wsign-promo
2578 @opindex Wno-sign-promo
2579 Warn when overload resolution chooses a promotion from unsigned or
2580 enumerated type to a signed type, over a conversion to an unsigned type of
2581 the same size. Previous versions of G++ would try to preserve
2582 unsignedness, but the standard mandates the current behavior.
2587 A& operator = (int);
2597 In this example, G++ will synthesize a default @samp{A& operator =
2598 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2601 @node Objective-C and Objective-C++ Dialect Options
2602 @section Options Controlling Objective-C and Objective-C++ Dialects
2604 @cindex compiler options, Objective-C and Objective-C++
2605 @cindex Objective-C and Objective-C++ options, command-line
2606 @cindex options, Objective-C and Objective-C++
2607 (NOTE: This manual does not describe the Objective-C and Objective-C++
2608 languages themselves. @xref{Standards,,Language Standards
2609 Supported by GCC}, for references.)
2611 This section describes the command-line options that are only meaningful
2612 for Objective-C and Objective-C++ programs, but you can also use most of
2613 the language-independent GNU compiler options.
2614 For example, you might compile a file @code{some_class.m} like this:
2617 gcc -g -fgnu-runtime -O -c some_class.m
2621 In this example, @option{-fgnu-runtime} is an option meant only for
2622 Objective-C and Objective-C++ programs; you can use the other options with
2623 any language supported by GCC@.
2625 Note that since Objective-C is an extension of the C language, Objective-C
2626 compilations may also use options specific to the C front-end (e.g.,
2627 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2628 C++-specific options (e.g., @option{-Wabi}).
2630 Here is a list of options that are @emph{only} for compiling Objective-C
2631 and Objective-C++ programs:
2634 @item -fconstant-string-class=@var{class-name}
2635 @opindex fconstant-string-class
2636 Use @var{class-name} as the name of the class to instantiate for each
2637 literal string specified with the syntax @code{@@"@dots{}"}. The default
2638 class name is @code{NXConstantString} if the GNU runtime is being used, and
2639 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2640 @option{-fconstant-cfstrings} option, if also present, will override the
2641 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2642 to be laid out as constant CoreFoundation strings.
2645 @opindex fgnu-runtime
2646 Generate object code compatible with the standard GNU Objective-C
2647 runtime. This is the default for most types of systems.
2649 @item -fnext-runtime
2650 @opindex fnext-runtime
2651 Generate output compatible with the NeXT runtime. This is the default
2652 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2653 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2656 @item -fno-nil-receivers
2657 @opindex fno-nil-receivers
2658 Assume that all Objective-C message dispatches (@code{[receiver
2659 message:arg]}) in this translation unit ensure that the receiver is
2660 not @code{nil}. This allows for more efficient entry points in the
2661 runtime to be used. This option is only available in conjunction with
2662 the NeXT runtime and ABI version 0 or 1.
2664 @item -fobjc-abi-version=@var{n}
2665 @opindex fobjc-abi-version
2666 Use version @var{n} of the Objective-C ABI for the selected runtime.
2667 This option is currently supported only for the NeXT runtime. In that
2668 case, Version 0 is the traditional (32-bit) ABI without support for
2669 properties and other Objective-C 2.0 additions. Version 1 is the
2670 traditional (32-bit) ABI with support for properties and other
2671 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2672 nothing is specified, the default is Version 0 on 32-bit target
2673 machines, and Version 2 on 64-bit target machines.
2675 @item -fobjc-call-cxx-cdtors
2676 @opindex fobjc-call-cxx-cdtors
2677 For each Objective-C class, check if any of its instance variables is a
2678 C++ object with a non-trivial default constructor. If so, synthesize a
2679 special @code{- (id) .cxx_construct} instance method that will run
2680 non-trivial default constructors on any such instance variables, in order,
2681 and then return @code{self}. Similarly, check if any instance variable
2682 is a C++ object with a non-trivial destructor, and if so, synthesize a
2683 special @code{- (void) .cxx_destruct} method that will run
2684 all such default destructors, in reverse order.
2686 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2687 methods thusly generated will only operate on instance variables
2688 declared in the current Objective-C class, and not those inherited
2689 from superclasses. It is the responsibility of the Objective-C
2690 runtime to invoke all such methods in an object's inheritance
2691 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2692 by the runtime immediately after a new object instance is allocated;
2693 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2694 before the runtime deallocates an object instance.
2696 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2697 support for invoking the @code{- (id) .cxx_construct} and
2698 @code{- (void) .cxx_destruct} methods.
2700 @item -fobjc-direct-dispatch
2701 @opindex fobjc-direct-dispatch
2702 Allow fast jumps to the message dispatcher. On Darwin this is
2703 accomplished via the comm page.
2705 @item -fobjc-exceptions
2706 @opindex fobjc-exceptions
2707 Enable syntactic support for structured exception handling in
2708 Objective-C, similar to what is offered by C++ and Java. This option
2709 is required to use the Objective-C keywords @code{@@try},
2710 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2711 @code{@@synchronized}. This option is available with both the GNU
2712 runtime and the NeXT runtime (but not available in conjunction with
2713 the NeXT runtime on Mac OS X 10.2 and earlier).
2717 Enable garbage collection (GC) in Objective-C and Objective-C++
2718 programs. This option is only available with the NeXT runtime; the
2719 GNU runtime has a different garbage collection implementation that
2720 does not require special compiler flags.
2722 @item -fobjc-nilcheck
2723 @opindex fobjc-nilcheck
2724 For the NeXT runtime with version 2 of the ABI, check for a nil
2725 receiver in method invocations before doing the actual method call.
2726 This is the default and can be disabled using
2727 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2728 checked for nil in this way no matter what this flag is set to.
2729 Currently this flag does nothing when the GNU runtime, or an older
2730 version of the NeXT runtime ABI, is used.
2732 @item -fobjc-std=objc1
2734 Conform to the language syntax of Objective-C 1.0, the language
2735 recognized by GCC 4.0. This only affects the Objective-C additions to
2736 the C/C++ language; it does not affect conformance to C/C++ standards,
2737 which is controlled by the separate C/C++ dialect option flags. When
2738 this option is used with the Objective-C or Objective-C++ compiler,
2739 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2740 This is useful if you need to make sure that your Objective-C code can
2741 be compiled with older versions of GCC.
2743 @item -freplace-objc-classes
2744 @opindex freplace-objc-classes
2745 Emit a special marker instructing @command{ld(1)} not to statically link in
2746 the resulting object file, and allow @command{dyld(1)} to load it in at
2747 run time instead. This is used in conjunction with the Fix-and-Continue
2748 debugging mode, where the object file in question may be recompiled and
2749 dynamically reloaded in the course of program execution, without the need
2750 to restart the program itself. Currently, Fix-and-Continue functionality
2751 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2756 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2757 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2758 compile time) with static class references that get initialized at load time,
2759 which improves run-time performance. Specifying the @option{-fzero-link} flag
2760 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2761 to be retained. This is useful in Zero-Link debugging mode, since it allows
2762 for individual class implementations to be modified during program execution.
2763 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2764 regardless of command-line options.
2768 Dump interface declarations for all classes seen in the source file to a
2769 file named @file{@var{sourcename}.decl}.
2771 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2772 @opindex Wassign-intercept
2773 @opindex Wno-assign-intercept
2774 Warn whenever an Objective-C assignment is being intercepted by the
2777 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2778 @opindex Wno-protocol
2780 If a class is declared to implement a protocol, a warning is issued for
2781 every method in the protocol that is not implemented by the class. The
2782 default behavior is to issue a warning for every method not explicitly
2783 implemented in the class, even if a method implementation is inherited
2784 from the superclass. If you use the @option{-Wno-protocol} option, then
2785 methods inherited from the superclass are considered to be implemented,
2786 and no warning is issued for them.
2788 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2790 @opindex Wno-selector
2791 Warn if multiple methods of different types for the same selector are
2792 found during compilation. The check is performed on the list of methods
2793 in the final stage of compilation. Additionally, a check is performed
2794 for each selector appearing in a @code{@@selector(@dots{})}
2795 expression, and a corresponding method for that selector has been found
2796 during compilation. Because these checks scan the method table only at
2797 the end of compilation, these warnings are not produced if the final
2798 stage of compilation is not reached, for example because an error is
2799 found during compilation, or because the @option{-fsyntax-only} option is
2802 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2803 @opindex Wstrict-selector-match
2804 @opindex Wno-strict-selector-match
2805 Warn if multiple methods with differing argument and/or return types are
2806 found for a given selector when attempting to send a message using this
2807 selector to a receiver of type @code{id} or @code{Class}. When this flag
2808 is off (which is the default behavior), the compiler will omit such warnings
2809 if any differences found are confined to types which share the same size
2812 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2813 @opindex Wundeclared-selector
2814 @opindex Wno-undeclared-selector
2815 Warn if a @code{@@selector(@dots{})} expression referring to an
2816 undeclared selector is found. A selector is considered undeclared if no
2817 method with that name has been declared before the
2818 @code{@@selector(@dots{})} expression, either explicitly in an
2819 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2820 an @code{@@implementation} section. This option always performs its
2821 checks as soon as a @code{@@selector(@dots{})} expression is found,
2822 while @option{-Wselector} only performs its checks in the final stage of
2823 compilation. This also enforces the coding style convention
2824 that methods and selectors must be declared before being used.
2826 @item -print-objc-runtime-info
2827 @opindex print-objc-runtime-info
2828 Generate C header describing the largest structure that is passed by
2833 @node Language Independent Options
2834 @section Options to Control Diagnostic Messages Formatting
2835 @cindex options to control diagnostics formatting
2836 @cindex diagnostic messages
2837 @cindex message formatting
2839 Traditionally, diagnostic messages have been formatted irrespective of
2840 the output device's aspect (e.g.@: its width, @dots{}). The options described
2841 below can be used to control the diagnostic messages formatting
2842 algorithm, e.g.@: how many characters per line, how often source location
2843 information should be reported. Right now, only the C++ front end can
2844 honor these options. However it is expected, in the near future, that
2845 the remaining front ends would be able to digest them correctly.
2848 @item -fmessage-length=@var{n}
2849 @opindex fmessage-length
2850 Try to format error messages so that they fit on lines of about @var{n}
2851 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2852 the front ends supported by GCC@. If @var{n} is zero, then no
2853 line-wrapping will be done; each error message will appear on a single
2856 @opindex fdiagnostics-show-location
2857 @item -fdiagnostics-show-location=once
2858 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2859 reporter to emit @emph{once} source location information; that is, in
2860 case the message is too long to fit on a single physical line and has to
2861 be wrapped, the source location won't be emitted (as prefix) again,
2862 over and over, in subsequent continuation lines. This is the default
2865 @item -fdiagnostics-show-location=every-line
2866 Only meaningful in line-wrapping mode. Instructs the diagnostic
2867 messages reporter to emit the same source location information (as
2868 prefix) for physical lines that result from the process of breaking
2869 a message which is too long to fit on a single line.
2871 @item -fno-diagnostics-show-option
2872 @opindex fno-diagnostics-show-option
2873 @opindex fdiagnostics-show-option
2874 By default, each diagnostic emitted includes text which indicates the
2875 command-line option that directly controls the diagnostic (if such an
2876 option is known to the diagnostic machinery). Specifying the
2877 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2881 @node Warning Options
2882 @section Options to Request or Suppress Warnings
2883 @cindex options to control warnings
2884 @cindex warning messages
2885 @cindex messages, warning
2886 @cindex suppressing warnings
2888 Warnings are diagnostic messages that report constructions which
2889 are not inherently erroneous but which are risky or suggest there
2890 may have been an error.
2892 The following language-independent options do not enable specific
2893 warnings but control the kinds of diagnostics produced by GCC.
2896 @cindex syntax checking
2898 @opindex fsyntax-only
2899 Check the code for syntax errors, but don't do anything beyond that.
2901 @item -fmax-errors=@var{n}
2902 @opindex fmax-errors
2903 Limits the maximum number of error messages to @var{n}, at which point
2904 GCC bails out rather than attempting to continue processing the source
2905 code. If @var{n} is 0 (the default), there is no limit on the number
2906 of error messages produced. If @option{-Wfatal-errors} is also
2907 specified, then @option{-Wfatal-errors} takes precedence over this
2912 Inhibit all warning messages.
2917 Make all warnings into errors.
2922 Make the specified warning into an error. The specifier for a warning
2923 is appended, for example @option{-Werror=switch} turns the warnings
2924 controlled by @option{-Wswitch} into errors. This switch takes a
2925 negative form, to be used to negate @option{-Werror} for specific
2926 warnings, for example @option{-Wno-error=switch} makes
2927 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2930 The warning message for each controllable warning includes the
2931 option which controls the warning. That option can then be used with
2932 @option{-Werror=} and @option{-Wno-error=} as described above.
2933 (Printing of the option in the warning message can be disabled using the
2934 @option{-fno-diagnostics-show-option} flag.)
2936 Note that specifying @option{-Werror=}@var{foo} automatically implies
2937 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2940 @item -Wfatal-errors
2941 @opindex Wfatal-errors
2942 @opindex Wno-fatal-errors
2943 This option causes the compiler to abort compilation on the first error
2944 occurred rather than trying to keep going and printing further error
2949 You can request many specific warnings with options beginning
2950 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2951 implicit declarations. Each of these specific warning options also
2952 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2953 example, @option{-Wno-implicit}. This manual lists only one of the
2954 two forms, whichever is not the default. For further,
2955 language-specific options also refer to @ref{C++ Dialect Options} and
2956 @ref{Objective-C and Objective-C++ Dialect Options}.
2958 When an unrecognized warning option is requested (e.g.,
2959 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2960 that the option is not recognized. However, if the @option{-Wno-} form
2961 is used, the behavior is slightly different: No diagnostic will be
2962 produced for @option{-Wno-unknown-warning} unless other diagnostics
2963 are being produced. This allows the use of new @option{-Wno-} options
2964 with old compilers, but if something goes wrong, the compiler will
2965 warn that an unrecognized option was used.
2970 Issue all the warnings demanded by strict ISO C and ISO C++;
2971 reject all programs that use forbidden extensions, and some other
2972 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2973 version of the ISO C standard specified by any @option{-std} option used.
2975 Valid ISO C and ISO C++ programs should compile properly with or without
2976 this option (though a rare few will require @option{-ansi} or a
2977 @option{-std} option specifying the required version of ISO C)@. However,
2978 without this option, certain GNU extensions and traditional C and C++
2979 features are supported as well. With this option, they are rejected.
2981 @option{-pedantic} does not cause warning messages for use of the
2982 alternate keywords whose names begin and end with @samp{__}. Pedantic
2983 warnings are also disabled in the expression that follows
2984 @code{__extension__}. However, only system header files should use
2985 these escape routes; application programs should avoid them.
2986 @xref{Alternate Keywords}.
2988 Some users try to use @option{-pedantic} to check programs for strict ISO
2989 C conformance. They soon find that it does not do quite what they want:
2990 it finds some non-ISO practices, but not all---only those for which
2991 ISO C @emph{requires} a diagnostic, and some others for which
2992 diagnostics have been added.
2994 A feature to report any failure to conform to ISO C might be useful in
2995 some instances, but would require considerable additional work and would
2996 be quite different from @option{-pedantic}. We don't have plans to
2997 support such a feature in the near future.
2999 Where the standard specified with @option{-std} represents a GNU
3000 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3001 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3002 extended dialect is based. Warnings from @option{-pedantic} are given
3003 where they are required by the base standard. (It would not make sense
3004 for such warnings to be given only for features not in the specified GNU
3005 C dialect, since by definition the GNU dialects of C include all
3006 features the compiler supports with the given option, and there would be
3007 nothing to warn about.)
3009 @item -pedantic-errors
3010 @opindex pedantic-errors
3011 Like @option{-pedantic}, except that errors are produced rather than
3017 This enables all the warnings about constructions that some users
3018 consider questionable, and that are easy to avoid (or modify to
3019 prevent the warning), even in conjunction with macros. This also
3020 enables some language-specific warnings described in @ref{C++ Dialect
3021 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3023 @option{-Wall} turns on the following warning flags:
3025 @gccoptlist{-Waddress @gol
3026 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3028 -Wchar-subscripts @gol
3029 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
3030 -Wimplicit-int @r{(C and Objective-C only)} @gol
3031 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3034 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3035 -Wmaybe-uninitialized @gol
3036 -Wmissing-braces @gol
3042 -Wsequence-point @gol
3043 -Wsign-compare @r{(only in C++)} @gol
3044 -Wstrict-aliasing @gol
3045 -Wstrict-overflow=1 @gol
3048 -Wuninitialized @gol
3049 -Wunknown-pragmas @gol
3050 -Wunused-function @gol
3053 -Wunused-variable @gol
3054 -Wvolatile-register-var @gol
3057 Note that some warning flags are not implied by @option{-Wall}. Some of
3058 them warn about constructions that users generally do not consider
3059 questionable, but which occasionally you might wish to check for;
3060 others warn about constructions that are necessary or hard to avoid in
3061 some cases, and there is no simple way to modify the code to suppress
3062 the warning. Some of them are enabled by @option{-Wextra} but many of
3063 them must be enabled individually.
3069 This enables some extra warning flags that are not enabled by
3070 @option{-Wall}. (This option used to be called @option{-W}. The older
3071 name is still supported, but the newer name is more descriptive.)
3073 @gccoptlist{-Wclobbered @gol
3075 -Wignored-qualifiers @gol
3076 -Wmissing-field-initializers @gol
3077 -Wmissing-parameter-type @r{(C only)} @gol
3078 -Wold-style-declaration @r{(C only)} @gol
3079 -Woverride-init @gol
3082 -Wuninitialized @gol
3083 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3084 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3087 The option @option{-Wextra} also prints warning messages for the
3093 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3094 @samp{>}, or @samp{>=}.
3097 (C++ only) An enumerator and a non-enumerator both appear in a
3098 conditional expression.
3101 (C++ only) Ambiguous virtual bases.
3104 (C++ only) Subscripting an array which has been declared @samp{register}.
3107 (C++ only) Taking the address of a variable which has been declared
3111 (C++ only) A base class is not initialized in a derived class' copy
3116 @item -Wchar-subscripts
3117 @opindex Wchar-subscripts
3118 @opindex Wno-char-subscripts
3119 Warn if an array subscript has type @code{char}. This is a common cause
3120 of error, as programmers often forget that this type is signed on some
3122 This warning is enabled by @option{-Wall}.
3126 @opindex Wno-comment
3127 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3128 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3129 This warning is enabled by @option{-Wall}.
3131 @item -Wno-coverage-mismatch
3132 @opindex Wno-coverage-mismatch
3133 Warn if feedback profiles do not match when using the
3134 @option{-fprofile-use} option.
3135 If a source file was changed between @option{-fprofile-gen} and
3136 @option{-fprofile-use}, the files with the profile feedback can fail
3137 to match the source file and GCC cannot use the profile feedback
3138 information. By default, this warning is enabled and is treated as an
3139 error. @option{-Wno-coverage-mismatch} can be used to disable the
3140 warning or @option{-Wno-error=coverage-mismatch} can be used to
3141 disable the error. Disabling the error for this warning can result in
3142 poorly optimized code and is useful only in the
3143 case of very minor changes such as bug fixes to an existing code-base.
3144 Completely disabling the warning is not recommended.
3147 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3149 Suppress warning messages emitted by @code{#warning} directives.
3151 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3152 @opindex Wdouble-promotion
3153 @opindex Wno-double-promotion
3154 Give a warning when a value of type @code{float} is implicitly
3155 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3156 floating-point unit implement @code{float} in hardware, but emulate
3157 @code{double} in software. On such a machine, doing computations
3158 using @code{double} values is much more expensive because of the
3159 overhead required for software emulation.
3161 It is easy to accidentally do computations with @code{double} because
3162 floating-point literals are implicitly of type @code{double}. For
3166 float area(float radius)
3168 return 3.14159 * radius * radius;
3172 the compiler will perform the entire computation with @code{double}
3173 because the floating-point literal is a @code{double}.
3178 @opindex ffreestanding
3179 @opindex fno-builtin
3180 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3181 the arguments supplied have types appropriate to the format string
3182 specified, and that the conversions specified in the format string make
3183 sense. This includes standard functions, and others specified by format
3184 attributes (@pxref{Function Attributes}), in the @code{printf},
3185 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3186 not in the C standard) families (or other target-specific families).
3187 Which functions are checked without format attributes having been
3188 specified depends on the standard version selected, and such checks of
3189 functions without the attribute specified are disabled by
3190 @option{-ffreestanding} or @option{-fno-builtin}.
3192 The formats are checked against the format features supported by GNU
3193 libc version 2.2. These include all ISO C90 and C99 features, as well
3194 as features from the Single Unix Specification and some BSD and GNU
3195 extensions. Other library implementations may not support all these
3196 features; GCC does not support warning about features that go beyond a
3197 particular library's limitations. However, if @option{-pedantic} is used
3198 with @option{-Wformat}, warnings will be given about format features not
3199 in the selected standard version (but not for @code{strfmon} formats,
3200 since those are not in any version of the C standard). @xref{C Dialect
3201 Options,,Options Controlling C Dialect}.
3203 Since @option{-Wformat} also checks for null format arguments for
3204 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3206 @option{-Wformat} is included in @option{-Wall}. For more control over some
3207 aspects of format checking, the options @option{-Wformat-y2k},
3208 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3209 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3210 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3213 @opindex Wformat-y2k
3214 @opindex Wno-format-y2k
3215 If @option{-Wformat} is specified, also warn about @code{strftime}
3216 formats which may yield only a two-digit year.
3218 @item -Wno-format-contains-nul
3219 @opindex Wno-format-contains-nul
3220 @opindex Wformat-contains-nul
3221 If @option{-Wformat} is specified, do not warn about format strings that
3224 @item -Wno-format-extra-args
3225 @opindex Wno-format-extra-args
3226 @opindex Wformat-extra-args
3227 If @option{-Wformat} is specified, do not warn about excess arguments to a
3228 @code{printf} or @code{scanf} format function. The C standard specifies
3229 that such arguments are ignored.
3231 Where the unused arguments lie between used arguments that are
3232 specified with @samp{$} operand number specifications, normally
3233 warnings are still given, since the implementation could not know what
3234 type to pass to @code{va_arg} to skip the unused arguments. However,
3235 in the case of @code{scanf} formats, this option will suppress the
3236 warning if the unused arguments are all pointers, since the Single
3237 Unix Specification says that such unused arguments are allowed.
3239 @item -Wno-format-zero-length
3240 @opindex Wno-format-zero-length
3241 @opindex Wformat-zero-length
3242 If @option{-Wformat} is specified, do not warn about zero-length formats.
3243 The C standard specifies that zero-length formats are allowed.
3245 @item -Wformat-nonliteral
3246 @opindex Wformat-nonliteral
3247 @opindex Wno-format-nonliteral
3248 If @option{-Wformat} is specified, also warn if the format string is not a
3249 string literal and so cannot be checked, unless the format function
3250 takes its format arguments as a @code{va_list}.
3252 @item -Wformat-security
3253 @opindex Wformat-security
3254 @opindex Wno-format-security
3255 If @option{-Wformat} is specified, also warn about uses of format
3256 functions that represent possible security problems. At present, this
3257 warns about calls to @code{printf} and @code{scanf} functions where the
3258 format string is not a string literal and there are no format arguments,
3259 as in @code{printf (foo);}. This may be a security hole if the format
3260 string came from untrusted input and contains @samp{%n}. (This is
3261 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3262 in future warnings may be added to @option{-Wformat-security} that are not
3263 included in @option{-Wformat-nonliteral}.)
3267 @opindex Wno-format=2
3268 Enable @option{-Wformat} plus format checks not included in
3269 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3270 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3274 @opindex Wno-nonnull
3275 Warn about passing a null pointer for arguments marked as
3276 requiring a non-null value by the @code{nonnull} function attribute.
3278 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3279 can be disabled with the @option{-Wno-nonnull} option.
3281 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3283 @opindex Wno-init-self
3284 Warn about uninitialized variables which are initialized with themselves.
3285 Note this option can only be used with the @option{-Wuninitialized} option.
3287 For example, GCC will warn about @code{i} being uninitialized in the
3288 following snippet only when @option{-Winit-self} has been specified:
3299 @item -Wimplicit-int @r{(C and Objective-C only)}
3300 @opindex Wimplicit-int
3301 @opindex Wno-implicit-int
3302 Warn when a declaration does not specify a type.
3303 This warning is enabled by @option{-Wall}.
3305 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3306 @opindex Wimplicit-function-declaration
3307 @opindex Wno-implicit-function-declaration
3308 Give a warning whenever a function is used before being declared. In
3309 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3310 enabled by default and it is made into an error by
3311 @option{-pedantic-errors}. This warning is also enabled by
3314 @item -Wimplicit @r{(C and Objective-C only)}
3316 @opindex Wno-implicit
3317 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3318 This warning is enabled by @option{-Wall}.
3320 @item -Wignored-qualifiers @r{(C and C++ only)}
3321 @opindex Wignored-qualifiers
3322 @opindex Wno-ignored-qualifiers
3323 Warn if the return type of a function has a type qualifier
3324 such as @code{const}. For ISO C such a type qualifier has no effect,
3325 since the value returned by a function is not an lvalue.
3326 For C++, the warning is only emitted for scalar types or @code{void}.
3327 ISO C prohibits qualified @code{void} return types on function
3328 definitions, so such return types always receive a warning
3329 even without this option.
3331 This warning is also enabled by @option{-Wextra}.
3336 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3337 a function with external linkage, returning int, taking either zero
3338 arguments, two, or three arguments of appropriate types. This warning
3339 is enabled by default in C++ and is enabled by either @option{-Wall}
3340 or @option{-pedantic}.
3342 @item -Wmissing-braces
3343 @opindex Wmissing-braces
3344 @opindex Wno-missing-braces
3345 Warn if an aggregate or union initializer is not fully bracketed. In
3346 the following example, the initializer for @samp{a} is not fully
3347 bracketed, but that for @samp{b} is fully bracketed.
3350 int a[2][2] = @{ 0, 1, 2, 3 @};
3351 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3354 This warning is enabled by @option{-Wall}.
3356 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3357 @opindex Wmissing-include-dirs
3358 @opindex Wno-missing-include-dirs
3359 Warn if a user-supplied include directory does not exist.
3362 @opindex Wparentheses
3363 @opindex Wno-parentheses
3364 Warn if parentheses are omitted in certain contexts, such
3365 as when there is an assignment in a context where a truth value
3366 is expected, or when operators are nested whose precedence people
3367 often get confused about.
3369 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3370 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3371 interpretation from that of ordinary mathematical notation.
3373 Also warn about constructions where there may be confusion to which
3374 @code{if} statement an @code{else} branch belongs. Here is an example of
3389 In C/C++, every @code{else} branch belongs to the innermost possible
3390 @code{if} statement, which in this example is @code{if (b)}. This is
3391 often not what the programmer expected, as illustrated in the above
3392 example by indentation the programmer chose. When there is the
3393 potential for this confusion, GCC will issue a warning when this flag
3394 is specified. To eliminate the warning, add explicit braces around
3395 the innermost @code{if} statement so there is no way the @code{else}
3396 could belong to the enclosing @code{if}. The resulting code would
3413 Also warn for dangerous uses of the
3414 ?: with omitted middle operand GNU extension. When the condition
3415 in the ?: operator is a boolean expression the omitted value will
3416 be always 1. Often the user expects it to be a value computed
3417 inside the conditional expression instead.
3419 This warning is enabled by @option{-Wall}.
3421 @item -Wsequence-point
3422 @opindex Wsequence-point
3423 @opindex Wno-sequence-point
3424 Warn about code that may have undefined semantics because of violations
3425 of sequence point rules in the C and C++ standards.
3427 The C and C++ standards defines the order in which expressions in a C/C++
3428 program are evaluated in terms of @dfn{sequence points}, which represent
3429 a partial ordering between the execution of parts of the program: those
3430 executed before the sequence point, and those executed after it. These
3431 occur after the evaluation of a full expression (one which is not part
3432 of a larger expression), after the evaluation of the first operand of a
3433 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3434 function is called (but after the evaluation of its arguments and the
3435 expression denoting the called function), and in certain other places.
3436 Other than as expressed by the sequence point rules, the order of
3437 evaluation of subexpressions of an expression is not specified. All
3438 these rules describe only a partial order rather than a total order,
3439 since, for example, if two functions are called within one expression
3440 with no sequence point between them, the order in which the functions
3441 are called is not specified. However, the standards committee have
3442 ruled that function calls do not overlap.
3444 It is not specified when between sequence points modifications to the
3445 values of objects take effect. Programs whose behavior depends on this
3446 have undefined behavior; the C and C++ standards specify that ``Between
3447 the previous and next sequence point an object shall have its stored
3448 value modified at most once by the evaluation of an expression.
3449 Furthermore, the prior value shall be read only to determine the value
3450 to be stored.''. If a program breaks these rules, the results on any
3451 particular implementation are entirely unpredictable.
3453 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3454 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3455 diagnosed by this option, and it may give an occasional false positive
3456 result, but in general it has been found fairly effective at detecting
3457 this sort of problem in programs.
3459 The standard is worded confusingly, therefore there is some debate
3460 over the precise meaning of the sequence point rules in subtle cases.
3461 Links to discussions of the problem, including proposed formal
3462 definitions, may be found on the GCC readings page, at
3463 @uref{http://gcc.gnu.org/@/readings.html}.
3465 This warning is enabled by @option{-Wall} for C and C++.
3468 @opindex Wreturn-type
3469 @opindex Wno-return-type
3470 Warn whenever a function is defined with a return-type that defaults
3471 to @code{int}. Also warn about any @code{return} statement with no
3472 return-value in a function whose return-type is not @code{void}
3473 (falling off the end of the function body is considered returning
3474 without a value), and about a @code{return} statement with an
3475 expression in a function whose return-type is @code{void}.
3477 For C++, a function without return type always produces a diagnostic
3478 message, even when @option{-Wno-return-type} is specified. The only
3479 exceptions are @samp{main} and functions defined in system headers.
3481 This warning is enabled by @option{-Wall}.
3486 Warn whenever a @code{switch} statement has an index of enumerated type
3487 and lacks a @code{case} for one or more of the named codes of that
3488 enumeration. (The presence of a @code{default} label prevents this
3489 warning.) @code{case} labels outside the enumeration range also
3490 provoke warnings when this option is used (even if there is a
3491 @code{default} label).
3492 This warning is enabled by @option{-Wall}.
3494 @item -Wswitch-default
3495 @opindex Wswitch-default
3496 @opindex Wno-switch-default
3497 Warn whenever a @code{switch} statement does not have a @code{default}
3501 @opindex Wswitch-enum
3502 @opindex Wno-switch-enum
3503 Warn whenever a @code{switch} statement has an index of enumerated type
3504 and lacks a @code{case} for one or more of the named codes of that
3505 enumeration. @code{case} labels outside the enumeration range also
3506 provoke warnings when this option is used. The only difference
3507 between @option{-Wswitch} and this option is that this option gives a
3508 warning about an omitted enumeration code even if there is a
3509 @code{default} label.
3511 @item -Wsync-nand @r{(C and C++ only)}
3513 @opindex Wno-sync-nand
3514 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3515 built-in functions are used. These functions changed semantics in GCC 4.4.
3519 @opindex Wno-trigraphs
3520 Warn if any trigraphs are encountered that might change the meaning of
3521 the program (trigraphs within comments are not warned about).
3522 This warning is enabled by @option{-Wall}.
3524 @item -Wunused-but-set-parameter
3525 @opindex Wunused-but-set-parameter
3526 @opindex Wno-unused-but-set-parameter
3527 Warn whenever a function parameter is assigned to, but otherwise unused
3528 (aside from its declaration).
3530 To suppress this warning use the @samp{unused} attribute
3531 (@pxref{Variable Attributes}).
3533 This warning is also enabled by @option{-Wunused} together with
3536 @item -Wunused-but-set-variable
3537 @opindex Wunused-but-set-variable
3538 @opindex Wno-unused-but-set-variable
3539 Warn whenever a local variable is assigned to, but otherwise unused
3540 (aside from its declaration).
3541 This warning is enabled by @option{-Wall}.
3543 To suppress this warning use the @samp{unused} attribute
3544 (@pxref{Variable Attributes}).
3546 This warning is also enabled by @option{-Wunused}, which is enabled
3549 @item -Wunused-function
3550 @opindex Wunused-function
3551 @opindex Wno-unused-function
3552 Warn whenever a static function is declared but not defined or a
3553 non-inline static function is unused.
3554 This warning is enabled by @option{-Wall}.
3556 @item -Wunused-label
3557 @opindex Wunused-label
3558 @opindex Wno-unused-label
3559 Warn whenever a label is declared but not used.
3560 This warning is enabled by @option{-Wall}.
3562 To suppress this warning use the @samp{unused} attribute
3563 (@pxref{Variable Attributes}).
3565 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3566 @opindex Wunused-local-typedefs
3567 Warn when a typedef locally defined in a function is not used.
3569 @item -Wunused-parameter
3570 @opindex Wunused-parameter
3571 @opindex Wno-unused-parameter
3572 Warn whenever a function parameter is unused aside from its declaration.
3574 To suppress this warning use the @samp{unused} attribute
3575 (@pxref{Variable Attributes}).
3577 @item -Wno-unused-result
3578 @opindex Wunused-result
3579 @opindex Wno-unused-result
3580 Do not warn if a caller of a function marked with attribute
3581 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3582 its return value. The default is @option{-Wunused-result}.
3584 @item -Wunused-variable
3585 @opindex Wunused-variable
3586 @opindex Wno-unused-variable
3587 Warn whenever a local variable or non-constant static variable is unused
3588 aside from its declaration.
3589 This warning is enabled by @option{-Wall}.
3591 To suppress this warning use the @samp{unused} attribute
3592 (@pxref{Variable Attributes}).
3594 @item -Wunused-value
3595 @opindex Wunused-value
3596 @opindex Wno-unused-value
3597 Warn whenever a statement computes a result that is explicitly not
3598 used. To suppress this warning cast the unused expression to
3599 @samp{void}. This includes an expression-statement or the left-hand
3600 side of a comma expression that contains no side effects. For example,
3601 an expression such as @samp{x[i,j]} will cause a warning, while
3602 @samp{x[(void)i,j]} will not.
3604 This warning is enabled by @option{-Wall}.
3609 All the above @option{-Wunused} options combined.
3611 In order to get a warning about an unused function parameter, you must
3612 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3613 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3615 @item -Wuninitialized
3616 @opindex Wuninitialized
3617 @opindex Wno-uninitialized
3618 Warn if an automatic variable is used without first being initialized
3619 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3620 warn if a non-static reference or non-static @samp{const} member
3621 appears in a class without constructors.
3623 If you want to warn about code which uses the uninitialized value of the
3624 variable in its own initializer, use the @option{-Winit-self} option.
3626 These warnings occur for individual uninitialized or clobbered
3627 elements of structure, union or array variables as well as for
3628 variables which are uninitialized or clobbered as a whole. They do
3629 not occur for variables or elements declared @code{volatile}. Because
3630 these warnings depend on optimization, the exact variables or elements
3631 for which there are warnings will depend on the precise optimization
3632 options and version of GCC used.
3634 Note that there may be no warning about a variable that is used only
3635 to compute a value that itself is never used, because such
3636 computations may be deleted by data flow analysis before the warnings
3639 @item -Wmaybe-uninitialized
3640 @opindex Wmaybe-uninitialized
3641 @opindex Wno-maybe-uninitialized
3642 For an automatic variable, if there exists a path from the function
3643 entry to a use of the variable that is initialized, but there exist
3644 some other paths the variable is not initialized, the compiler will
3645 emit a warning if it can not prove the uninitialized paths do not
3646 happen at run time. These warnings are made optional because GCC is
3647 not smart enough to see all the reasons why the code might be correct
3648 despite appearing to have an error. Here is one example of how
3669 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3670 always initialized, but GCC doesn't know this. To suppress the
3671 warning, the user needs to provide a default case with assert(0) or
3674 @cindex @code{longjmp} warnings
3675 This option also warns when a non-volatile automatic variable might be
3676 changed by a call to @code{longjmp}. These warnings as well are possible
3677 only in optimizing compilation.
3679 The compiler sees only the calls to @code{setjmp}. It cannot know
3680 where @code{longjmp} will be called; in fact, a signal handler could
3681 call it at any point in the code. As a result, you may get a warning
3682 even when there is in fact no problem because @code{longjmp} cannot
3683 in fact be called at the place which would cause a problem.
3685 Some spurious warnings can be avoided if you declare all the functions
3686 you use that never return as @code{noreturn}. @xref{Function
3689 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3691 @item -Wunknown-pragmas
3692 @opindex Wunknown-pragmas
3693 @opindex Wno-unknown-pragmas
3694 @cindex warning for unknown pragmas
3695 @cindex unknown pragmas, warning
3696 @cindex pragmas, warning of unknown
3697 Warn when a #pragma directive is encountered which is not understood by
3698 GCC@. If this command-line option is used, warnings will even be issued
3699 for unknown pragmas in system header files. This is not the case if
3700 the warnings were only enabled by the @option{-Wall} command-line option.
3703 @opindex Wno-pragmas
3705 Do not warn about misuses of pragmas, such as incorrect parameters,
3706 invalid syntax, or conflicts between pragmas. See also
3707 @samp{-Wunknown-pragmas}.
3709 @item -Wstrict-aliasing
3710 @opindex Wstrict-aliasing
3711 @opindex Wno-strict-aliasing
3712 This option is only active when @option{-fstrict-aliasing} is active.
3713 It warns about code which might break the strict aliasing rules that the
3714 compiler is using for optimization. The warning does not catch all
3715 cases, but does attempt to catch the more common pitfalls. It is
3716 included in @option{-Wall}.
3717 It is equivalent to @option{-Wstrict-aliasing=3}
3719 @item -Wstrict-aliasing=n
3720 @opindex Wstrict-aliasing=n
3721 @opindex Wno-strict-aliasing=n
3722 This option is only active when @option{-fstrict-aliasing} is active.
3723 It warns about code which might break the strict aliasing rules that the
3724 compiler is using for optimization.
3725 Higher levels correspond to higher accuracy (fewer false positives).
3726 Higher levels also correspond to more effort, similar to the way -O works.
3727 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3730 Level 1: Most aggressive, quick, least accurate.
3731 Possibly useful when higher levels
3732 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3733 false negatives. However, it has many false positives.
3734 Warns for all pointer conversions between possibly incompatible types,
3735 even if never dereferenced. Runs in the frontend only.
3737 Level 2: Aggressive, quick, not too precise.
3738 May still have many false positives (not as many as level 1 though),
3739 and few false negatives (but possibly more than level 1).
3740 Unlike level 1, it only warns when an address is taken. Warns about
3741 incomplete types. Runs in the frontend only.
3743 Level 3 (default for @option{-Wstrict-aliasing}):
3744 Should have very few false positives and few false
3745 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3746 Takes care of the common pun+dereference pattern in the frontend:
3747 @code{*(int*)&some_float}.
3748 If optimization is enabled, it also runs in the backend, where it deals
3749 with multiple statement cases using flow-sensitive points-to information.
3750 Only warns when the converted pointer is dereferenced.
3751 Does not warn about incomplete types.
3753 @item -Wstrict-overflow
3754 @itemx -Wstrict-overflow=@var{n}
3755 @opindex Wstrict-overflow
3756 @opindex Wno-strict-overflow
3757 This option is only active when @option{-fstrict-overflow} is active.
3758 It warns about cases where the compiler optimizes based on the
3759 assumption that signed overflow does not occur. Note that it does not
3760 warn about all cases where the code might overflow: it only warns
3761 about cases where the compiler implements some optimization. Thus
3762 this warning depends on the optimization level.
3764 An optimization which assumes that signed overflow does not occur is
3765 perfectly safe if the values of the variables involved are such that
3766 overflow never does, in fact, occur. Therefore this warning can
3767 easily give a false positive: a warning about code which is not
3768 actually a problem. To help focus on important issues, several
3769 warning levels are defined. No warnings are issued for the use of
3770 undefined signed overflow when estimating how many iterations a loop
3771 will require, in particular when determining whether a loop will be
3775 @item -Wstrict-overflow=1
3776 Warn about cases which are both questionable and easy to avoid. For
3777 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3778 compiler will simplify this to @code{1}. This level of
3779 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3780 are not, and must be explicitly requested.
3782 @item -Wstrict-overflow=2
3783 Also warn about other cases where a comparison is simplified to a
3784 constant. For example: @code{abs (x) >= 0}. This can only be
3785 simplified when @option{-fstrict-overflow} is in effect, because
3786 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3787 zero. @option{-Wstrict-overflow} (with no level) is the same as
3788 @option{-Wstrict-overflow=2}.
3790 @item -Wstrict-overflow=3
3791 Also warn about other cases where a comparison is simplified. For
3792 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3794 @item -Wstrict-overflow=4
3795 Also warn about other simplifications not covered by the above cases.
3796 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3798 @item -Wstrict-overflow=5
3799 Also warn about cases where the compiler reduces the magnitude of a
3800 constant involved in a comparison. For example: @code{x + 2 > y} will
3801 be simplified to @code{x + 1 >= y}. This is reported only at the
3802 highest warning level because this simplification applies to many
3803 comparisons, so this warning level will give a very large number of
3807 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3808 @opindex Wsuggest-attribute=
3809 @opindex Wno-suggest-attribute=
3810 Warn for cases where adding an attribute may be beneficial. The
3811 attributes currently supported are listed below.
3814 @item -Wsuggest-attribute=pure
3815 @itemx -Wsuggest-attribute=const
3816 @itemx -Wsuggest-attribute=noreturn
3817 @opindex Wsuggest-attribute=pure
3818 @opindex Wno-suggest-attribute=pure
3819 @opindex Wsuggest-attribute=const
3820 @opindex Wno-suggest-attribute=const
3821 @opindex Wsuggest-attribute=noreturn
3822 @opindex Wno-suggest-attribute=noreturn
3824 Warn about functions which might be candidates for attributes
3825 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3826 functions visible in other compilation units or (in the case of @code{pure} and
3827 @code{const}) if it cannot prove that the function returns normally. A function
3828 returns normally if it doesn't contain an infinite loop nor returns abnormally
3829 by throwing, calling @code{abort()} or trapping. This analysis requires option
3830 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3831 higher. Higher optimization levels improve the accuracy of the analysis.
3834 @item -Warray-bounds
3835 @opindex Wno-array-bounds
3836 @opindex Warray-bounds
3837 This option is only active when @option{-ftree-vrp} is active
3838 (default for @option{-O2} and above). It warns about subscripts to arrays
3839 that are always out of bounds. This warning is enabled by @option{-Wall}.
3841 @item -Wno-div-by-zero
3842 @opindex Wno-div-by-zero
3843 @opindex Wdiv-by-zero
3844 Do not warn about compile-time integer division by zero. Floating-point
3845 division by zero is not warned about, as it can be a legitimate way of
3846 obtaining infinities and NaNs.
3848 @item -Wsystem-headers
3849 @opindex Wsystem-headers
3850 @opindex Wno-system-headers
3851 @cindex warnings from system headers
3852 @cindex system headers, warnings from
3853 Print warning messages for constructs found in system header files.
3854 Warnings from system headers are normally suppressed, on the assumption
3855 that they usually do not indicate real problems and would only make the
3856 compiler output harder to read. Using this command-line option tells
3857 GCC to emit warnings from system headers as if they occurred in user
3858 code. However, note that using @option{-Wall} in conjunction with this
3859 option will @emph{not} warn about unknown pragmas in system
3860 headers---for that, @option{-Wunknown-pragmas} must also be used.
3863 @opindex Wtrampolines
3864 @opindex Wno-trampolines
3865 Warn about trampolines generated for pointers to nested functions.
3867 A trampoline is a small piece of data or code that is created at run
3868 time on the stack when the address of a nested function is taken, and
3869 is used to call the nested function indirectly. For some targets, it
3870 is made up of data only and thus requires no special treatment. But,
3871 for most targets, it is made up of code and thus requires the stack
3872 to be made executable in order for the program to work properly.
3875 @opindex Wfloat-equal
3876 @opindex Wno-float-equal
3877 Warn if floating-point values are used in equality comparisons.
3879 The idea behind this is that sometimes it is convenient (for the
3880 programmer) to consider floating-point values as approximations to
3881 infinitely precise real numbers. If you are doing this, then you need
3882 to compute (by analyzing the code, or in some other way) the maximum or
3883 likely maximum error that the computation introduces, and allow for it
3884 when performing comparisons (and when producing output, but that's a
3885 different problem). In particular, instead of testing for equality, you
3886 would check to see whether the two values have ranges that overlap; and
3887 this is done with the relational operators, so equality comparisons are
3890 @item -Wtraditional @r{(C and Objective-C only)}
3891 @opindex Wtraditional
3892 @opindex Wno-traditional
3893 Warn about certain constructs that behave differently in traditional and
3894 ISO C@. Also warn about ISO C constructs that have no traditional C
3895 equivalent, and/or problematic constructs which should be avoided.
3899 Macro parameters that appear within string literals in the macro body.
3900 In traditional C macro replacement takes place within string literals,
3901 but does not in ISO C@.
3904 In traditional C, some preprocessor directives did not exist.
3905 Traditional preprocessors would only consider a line to be a directive
3906 if the @samp{#} appeared in column 1 on the line. Therefore
3907 @option{-Wtraditional} warns about directives that traditional C
3908 understands but would ignore because the @samp{#} does not appear as the
3909 first character on the line. It also suggests you hide directives like
3910 @samp{#pragma} not understood by traditional C by indenting them. Some
3911 traditional implementations would not recognize @samp{#elif}, so it
3912 suggests avoiding it altogether.
3915 A function-like macro that appears without arguments.
3918 The unary plus operator.
3921 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
3922 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3923 constants.) Note, these suffixes appear in macros defined in the system
3924 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3925 Use of these macros in user code might normally lead to spurious
3926 warnings, however GCC's integrated preprocessor has enough context to
3927 avoid warning in these cases.
3930 A function declared external in one block and then used after the end of
3934 A @code{switch} statement has an operand of type @code{long}.
3937 A non-@code{static} function declaration follows a @code{static} one.
3938 This construct is not accepted by some traditional C compilers.
3941 The ISO type of an integer constant has a different width or
3942 signedness from its traditional type. This warning is only issued if
3943 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3944 typically represent bit patterns, are not warned about.
3947 Usage of ISO string concatenation is detected.
3950 Initialization of automatic aggregates.
3953 Identifier conflicts with labels. Traditional C lacks a separate
3954 namespace for labels.
3957 Initialization of unions. If the initializer is zero, the warning is
3958 omitted. This is done under the assumption that the zero initializer in
3959 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3960 initializer warnings and relies on default initialization to zero in the
3964 Conversions by prototypes between fixed/floating-point values and vice
3965 versa. The absence of these prototypes when compiling with traditional
3966 C would cause serious problems. This is a subset of the possible
3967 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3970 Use of ISO C style function definitions. This warning intentionally is
3971 @emph{not} issued for prototype declarations or variadic functions
3972 because these ISO C features will appear in your code when using
3973 libiberty's traditional C compatibility macros, @code{PARAMS} and
3974 @code{VPARAMS}. This warning is also bypassed for nested functions
3975 because that feature is already a GCC extension and thus not relevant to
3976 traditional C compatibility.
3979 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3980 @opindex Wtraditional-conversion
3981 @opindex Wno-traditional-conversion
3982 Warn if a prototype causes a type conversion that is different from what
3983 would happen to the same argument in the absence of a prototype. This
3984 includes conversions of fixed point to floating and vice versa, and
3985 conversions changing the width or signedness of a fixed-point argument
3986 except when the same as the default promotion.
3988 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3989 @opindex Wdeclaration-after-statement
3990 @opindex Wno-declaration-after-statement
3991 Warn when a declaration is found after a statement in a block. This
3992 construct, known from C++, was introduced with ISO C99 and is by default
3993 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3994 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3999 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4001 @item -Wno-endif-labels
4002 @opindex Wno-endif-labels
4003 @opindex Wendif-labels
4004 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4009 Warn whenever a local variable or type declaration shadows another variable,
4010 parameter, type, or class member (in C++), or whenever a built-in function
4011 is shadowed. Note that in C++, the compiler will not warn if a local variable
4012 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
4014 @item -Wlarger-than=@var{len}
4015 @opindex Wlarger-than=@var{len}
4016 @opindex Wlarger-than-@var{len}
4017 Warn whenever an object of larger than @var{len} bytes is defined.
4019 @item -Wframe-larger-than=@var{len}
4020 @opindex Wframe-larger-than
4021 Warn if the size of a function frame is larger than @var{len} bytes.
4022 The computation done to determine the stack frame size is approximate
4023 and not conservative.
4024 The actual requirements may be somewhat greater than @var{len}
4025 even if you do not get a warning. In addition, any space allocated
4026 via @code{alloca}, variable-length arrays, or related constructs
4027 is not included by the compiler when determining
4028 whether or not to issue a warning.
4030 @item -Wno-free-nonheap-object
4031 @opindex Wno-free-nonheap-object
4032 @opindex Wfree-nonheap-object
4033 Do not warn when attempting to free an object which was not allocated
4036 @item -Wstack-usage=@var{len}
4037 @opindex Wstack-usage
4038 Warn if the stack usage of a function might be larger than @var{len} bytes.
4039 The computation done to determine the stack usage is conservative.
4040 Any space allocated via @code{alloca}, variable-length arrays, or related
4041 constructs is included by the compiler when determining whether or not to
4044 The message is in keeping with the output of @option{-fstack-usage}.
4048 If the stack usage is fully static but exceeds the specified amount, it's:
4051 warning: stack usage is 1120 bytes
4054 If the stack usage is (partly) dynamic but bounded, it's:
4057 warning: stack usage might be 1648 bytes
4060 If the stack usage is (partly) dynamic and not bounded, it's:
4063 warning: stack usage might be unbounded
4067 @item -Wunsafe-loop-optimizations
4068 @opindex Wunsafe-loop-optimizations
4069 @opindex Wno-unsafe-loop-optimizations
4070 Warn if the loop cannot be optimized because the compiler could not
4071 assume anything on the bounds of the loop indices. With
4072 @option{-funsafe-loop-optimizations} warn if the compiler made
4075 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4076 @opindex Wno-pedantic-ms-format
4077 @opindex Wpedantic-ms-format
4078 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4079 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4080 depending on the MS runtime, when you are using the options @option{-Wformat}
4081 and @option{-pedantic} without gnu-extensions.
4083 @item -Wpointer-arith
4084 @opindex Wpointer-arith
4085 @opindex Wno-pointer-arith
4086 Warn about anything that depends on the ``size of'' a function type or
4087 of @code{void}. GNU C assigns these types a size of 1, for
4088 convenience in calculations with @code{void *} pointers and pointers
4089 to functions. In C++, warn also when an arithmetic operation involves
4090 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4093 @opindex Wtype-limits
4094 @opindex Wno-type-limits
4095 Warn if a comparison is always true or always false due to the limited
4096 range of the data type, but do not warn for constant expressions. For
4097 example, warn if an unsigned variable is compared against zero with
4098 @samp{<} or @samp{>=}. This warning is also enabled by
4101 @item -Wbad-function-cast @r{(C and Objective-C only)}
4102 @opindex Wbad-function-cast
4103 @opindex Wno-bad-function-cast
4104 Warn whenever a function call is cast to a non-matching type.
4105 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4107 @item -Wc++-compat @r{(C and Objective-C only)}
4108 Warn about ISO C constructs that are outside of the common subset of
4109 ISO C and ISO C++, e.g.@: request for implicit conversion from
4110 @code{void *} to a pointer to non-@code{void} type.
4112 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4113 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4114 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4115 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4116 enabled by @option{-Wall}.
4120 @opindex Wno-cast-qual
4121 Warn whenever a pointer is cast so as to remove a type qualifier from
4122 the target type. For example, warn if a @code{const char *} is cast
4123 to an ordinary @code{char *}.
4125 Also warn when making a cast which introduces a type qualifier in an
4126 unsafe way. For example, casting @code{char **} to @code{const char **}
4127 is unsafe, as in this example:
4130 /* p is char ** value. */
4131 const char **q = (const char **) p;
4132 /* Assignment of readonly string to const char * is OK. */
4134 /* Now char** pointer points to read-only memory. */
4139 @opindex Wcast-align
4140 @opindex Wno-cast-align
4141 Warn whenever a pointer is cast such that the required alignment of the
4142 target is increased. For example, warn if a @code{char *} is cast to
4143 an @code{int *} on machines where integers can only be accessed at
4144 two- or four-byte boundaries.
4146 @item -Wwrite-strings
4147 @opindex Wwrite-strings
4148 @opindex Wno-write-strings
4149 When compiling C, give string constants the type @code{const
4150 char[@var{length}]} so that copying the address of one into a
4151 non-@code{const} @code{char *} pointer will get a warning. These
4152 warnings will help you find at compile time code that can try to write
4153 into a string constant, but only if you have been very careful about
4154 using @code{const} in declarations and prototypes. Otherwise, it will
4155 just be a nuisance. This is why we did not make @option{-Wall} request
4158 When compiling C++, warn about the deprecated conversion from string
4159 literals to @code{char *}. This warning is enabled by default for C++
4164 @opindex Wno-clobbered
4165 Warn for variables that might be changed by @samp{longjmp} or
4166 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4169 @opindex Wconversion
4170 @opindex Wno-conversion
4171 Warn for implicit conversions that may alter a value. This includes
4172 conversions between real and integer, like @code{abs (x)} when
4173 @code{x} is @code{double}; conversions between signed and unsigned,
4174 like @code{unsigned ui = -1}; and conversions to smaller types, like
4175 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4176 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4177 changed by the conversion like in @code{abs (2.0)}. Warnings about
4178 conversions between signed and unsigned integers can be disabled by
4179 using @option{-Wno-sign-conversion}.
4181 For C++, also warn for confusing overload resolution for user-defined
4182 conversions; and conversions that will never use a type conversion
4183 operator: conversions to @code{void}, the same type, a base class or a
4184 reference to them. Warnings about conversions between signed and
4185 unsigned integers are disabled by default in C++ unless
4186 @option{-Wsign-conversion} is explicitly enabled.
4188 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4189 @opindex Wconversion-null
4190 @opindex Wno-conversion-null
4191 Do not warn for conversions between @code{NULL} and non-pointer
4192 types. @option{-Wconversion-null} is enabled by default.
4194 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4195 @opindex Wzero-as-null-pointer-constant
4196 @opindex Wno-zero-as-null-pointer-constant
4197 Warn when a literal '0' is used as null pointer constant. This can
4198 be useful to facilitate the conversion to @code{nullptr} in C++11.
4201 @opindex Wempty-body
4202 @opindex Wno-empty-body
4203 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4204 while} statement. This warning is also enabled by @option{-Wextra}.
4206 @item -Wenum-compare
4207 @opindex Wenum-compare
4208 @opindex Wno-enum-compare
4209 Warn about a comparison between values of different enum types. In C++
4210 this warning is enabled by default. In C this warning is enabled by
4213 @item -Wjump-misses-init @r{(C, Objective-C only)}
4214 @opindex Wjump-misses-init
4215 @opindex Wno-jump-misses-init
4216 Warn if a @code{goto} statement or a @code{switch} statement jumps
4217 forward across the initialization of a variable, or jumps backward to a
4218 label after the variable has been initialized. This only warns about
4219 variables which are initialized when they are declared. This warning is
4220 only supported for C and Objective C; in C++ this sort of branch is an
4223 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4224 can be disabled with the @option{-Wno-jump-misses-init} option.
4226 @item -Wsign-compare
4227 @opindex Wsign-compare
4228 @opindex Wno-sign-compare
4229 @cindex warning for comparison of signed and unsigned values
4230 @cindex comparison of signed and unsigned values, warning
4231 @cindex signed and unsigned values, comparison warning
4232 Warn when a comparison between signed and unsigned values could produce
4233 an incorrect result when the signed value is converted to unsigned.
4234 This warning is also enabled by @option{-Wextra}; to get the other warnings
4235 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4237 @item -Wsign-conversion
4238 @opindex Wsign-conversion
4239 @opindex Wno-sign-conversion
4240 Warn for implicit conversions that may change the sign of an integer
4241 value, like assigning a signed integer expression to an unsigned
4242 integer variable. An explicit cast silences the warning. In C, this
4243 option is enabled also by @option{-Wconversion}.
4247 @opindex Wno-address
4248 Warn about suspicious uses of memory addresses. These include using
4249 the address of a function in a conditional expression, such as
4250 @code{void func(void); if (func)}, and comparisons against the memory
4251 address of a string literal, such as @code{if (x == "abc")}. Such
4252 uses typically indicate a programmer error: the address of a function
4253 always evaluates to true, so their use in a conditional usually
4254 indicate that the programmer forgot the parentheses in a function
4255 call; and comparisons against string literals result in unspecified
4256 behavior and are not portable in C, so they usually indicate that the
4257 programmer intended to use @code{strcmp}. This warning is enabled by
4261 @opindex Wlogical-op
4262 @opindex Wno-logical-op
4263 Warn about suspicious uses of logical operators in expressions.
4264 This includes using logical operators in contexts where a
4265 bit-wise operator is likely to be expected.
4267 @item -Waggregate-return
4268 @opindex Waggregate-return
4269 @opindex Wno-aggregate-return
4270 Warn if any functions that return structures or unions are defined or
4271 called. (In languages where you can return an array, this also elicits
4274 @item -Wno-attributes
4275 @opindex Wno-attributes
4276 @opindex Wattributes
4277 Do not warn if an unexpected @code{__attribute__} is used, such as
4278 unrecognized attributes, function attributes applied to variables,
4279 etc. This will not stop errors for incorrect use of supported
4282 @item -Wno-builtin-macro-redefined
4283 @opindex Wno-builtin-macro-redefined
4284 @opindex Wbuiltin-macro-redefined
4285 Do not warn if certain built-in macros are redefined. This suppresses
4286 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4287 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4289 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4290 @opindex Wstrict-prototypes
4291 @opindex Wno-strict-prototypes
4292 Warn if a function is declared or defined without specifying the
4293 argument types. (An old-style function definition is permitted without
4294 a warning if preceded by a declaration which specifies the argument
4297 @item -Wold-style-declaration @r{(C and Objective-C only)}
4298 @opindex Wold-style-declaration
4299 @opindex Wno-old-style-declaration
4300 Warn for obsolescent usages, according to the C Standard, in a
4301 declaration. For example, warn if storage-class specifiers like
4302 @code{static} are not the first things in a declaration. This warning
4303 is also enabled by @option{-Wextra}.
4305 @item -Wold-style-definition @r{(C and Objective-C only)}
4306 @opindex Wold-style-definition
4307 @opindex Wno-old-style-definition
4308 Warn if an old-style function definition is used. A warning is given
4309 even if there is a previous prototype.
4311 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4312 @opindex Wmissing-parameter-type
4313 @opindex Wno-missing-parameter-type
4314 A function parameter is declared without a type specifier in K&R-style
4321 This warning is also enabled by @option{-Wextra}.
4323 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4324 @opindex Wmissing-prototypes
4325 @opindex Wno-missing-prototypes
4326 Warn if a global function is defined without a previous prototype
4327 declaration. This warning is issued even if the definition itself
4328 provides a prototype. The aim is to detect global functions that fail
4329 to be declared in header files.
4331 @item -Wmissing-declarations
4332 @opindex Wmissing-declarations
4333 @opindex Wno-missing-declarations
4334 Warn if a global function is defined without a previous declaration.
4335 Do so even if the definition itself provides a prototype.
4336 Use this option to detect global functions that are not declared in
4337 header files. In C++, no warnings are issued for function templates,
4338 or for inline functions, or for functions in anonymous namespaces.
4340 @item -Wmissing-field-initializers
4341 @opindex Wmissing-field-initializers
4342 @opindex Wno-missing-field-initializers
4346 Warn if a structure's initializer has some fields missing. For
4347 example, the following code would cause such a warning, because
4348 @code{x.h} is implicitly zero:
4351 struct s @{ int f, g, h; @};
4352 struct s x = @{ 3, 4 @};
4355 This option does not warn about designated initializers, so the following
4356 modification would not trigger a warning:
4359 struct s @{ int f, g, h; @};
4360 struct s x = @{ .f = 3, .g = 4 @};
4363 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4364 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4366 @item -Wmissing-format-attribute
4367 @opindex Wmissing-format-attribute
4368 @opindex Wno-missing-format-attribute
4371 Warn about function pointers which might be candidates for @code{format}
4372 attributes. Note these are only possible candidates, not absolute ones.
4373 GCC will guess that function pointers with @code{format} attributes that
4374 are used in assignment, initialization, parameter passing or return
4375 statements should have a corresponding @code{format} attribute in the
4376 resulting type. I.e.@: the left-hand side of the assignment or
4377 initialization, the type of the parameter variable, or the return type
4378 of the containing function respectively should also have a @code{format}
4379 attribute to avoid the warning.
4381 GCC will also warn about function definitions which might be
4382 candidates for @code{format} attributes. Again, these are only
4383 possible candidates. GCC will guess that @code{format} attributes
4384 might be appropriate for any function that calls a function like
4385 @code{vprintf} or @code{vscanf}, but this might not always be the
4386 case, and some functions for which @code{format} attributes are
4387 appropriate may not be detected.
4389 @item -Wno-multichar
4390 @opindex Wno-multichar
4392 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4393 Usually they indicate a typo in the user's code, as they have
4394 implementation-defined values, and should not be used in portable code.
4396 @item -Wnormalized=<none|id|nfc|nfkc>
4397 @opindex Wnormalized=
4400 @cindex character set, input normalization
4401 In ISO C and ISO C++, two identifiers are different if they are
4402 different sequences of characters. However, sometimes when characters
4403 outside the basic ASCII character set are used, you can have two
4404 different character sequences that look the same. To avoid confusion,
4405 the ISO 10646 standard sets out some @dfn{normalization rules} which
4406 when applied ensure that two sequences that look the same are turned into
4407 the same sequence. GCC can warn you if you are using identifiers which
4408 have not been normalized; this option controls that warning.
4410 There are four levels of warning that GCC supports. The default is
4411 @option{-Wnormalized=nfc}, which warns about any identifier which is
4412 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4413 recommended form for most uses.
4415 Unfortunately, there are some characters which ISO C and ISO C++ allow
4416 in identifiers that when turned into NFC aren't allowable as
4417 identifiers. That is, there's no way to use these symbols in portable
4418 ISO C or C++ and have all your identifiers in NFC@.
4419 @option{-Wnormalized=id} suppresses the warning for these characters.
4420 It is hoped that future versions of the standards involved will correct
4421 this, which is why this option is not the default.
4423 You can switch the warning off for all characters by writing
4424 @option{-Wnormalized=none}. You would only want to do this if you
4425 were using some other normalization scheme (like ``D''), because
4426 otherwise you can easily create bugs that are literally impossible to see.
4428 Some characters in ISO 10646 have distinct meanings but look identical
4429 in some fonts or display methodologies, especially once formatting has
4430 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4431 LETTER N'', will display just like a regular @code{n} which has been
4432 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4433 normalization scheme to convert all these into a standard form as
4434 well, and GCC will warn if your code is not in NFKC if you use
4435 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4436 about every identifier that contains the letter O because it might be
4437 confused with the digit 0, and so is not the default, but may be
4438 useful as a local coding convention if the programming environment is
4439 unable to be fixed to display these characters distinctly.
4441 @item -Wno-deprecated
4442 @opindex Wno-deprecated
4443 @opindex Wdeprecated
4444 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4446 @item -Wno-deprecated-declarations
4447 @opindex Wno-deprecated-declarations
4448 @opindex Wdeprecated-declarations
4449 Do not warn about uses of functions (@pxref{Function Attributes}),
4450 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4451 Attributes}) marked as deprecated by using the @code{deprecated}
4455 @opindex Wno-overflow
4457 Do not warn about compile-time overflow in constant expressions.
4459 @item -Woverride-init @r{(C and Objective-C only)}
4460 @opindex Woverride-init
4461 @opindex Wno-override-init
4465 Warn if an initialized field without side effects is overridden when
4466 using designated initializers (@pxref{Designated Inits, , Designated
4469 This warning is included in @option{-Wextra}. To get other
4470 @option{-Wextra} warnings without this one, use @samp{-Wextra
4471 -Wno-override-init}.
4476 Warn if a structure is given the packed attribute, but the packed
4477 attribute has no effect on the layout or size of the structure.
4478 Such structures may be mis-aligned for little benefit. For
4479 instance, in this code, the variable @code{f.x} in @code{struct bar}
4480 will be misaligned even though @code{struct bar} does not itself
4481 have the packed attribute:
4488 @} __attribute__((packed));
4496 @item -Wpacked-bitfield-compat
4497 @opindex Wpacked-bitfield-compat
4498 @opindex Wno-packed-bitfield-compat
4499 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4500 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4501 the change can lead to differences in the structure layout. GCC
4502 informs you when the offset of such a field has changed in GCC 4.4.
4503 For example there is no longer a 4-bit padding between field @code{a}
4504 and @code{b} in this structure:
4511 @} __attribute__ ((packed));
4514 This warning is enabled by default. Use
4515 @option{-Wno-packed-bitfield-compat} to disable this warning.
4520 Warn if padding is included in a structure, either to align an element
4521 of the structure or to align the whole structure. Sometimes when this
4522 happens it is possible to rearrange the fields of the structure to
4523 reduce the padding and so make the structure smaller.
4525 @item -Wredundant-decls
4526 @opindex Wredundant-decls
4527 @opindex Wno-redundant-decls
4528 Warn if anything is declared more than once in the same scope, even in
4529 cases where multiple declaration is valid and changes nothing.
4531 @item -Wnested-externs @r{(C and Objective-C only)}
4532 @opindex Wnested-externs
4533 @opindex Wno-nested-externs
4534 Warn if an @code{extern} declaration is encountered within a function.
4539 Warn if a function can not be inlined and it was declared as inline.
4540 Even with this option, the compiler will not warn about failures to
4541 inline functions declared in system headers.
4543 The compiler uses a variety of heuristics to determine whether or not
4544 to inline a function. For example, the compiler takes into account
4545 the size of the function being inlined and the amount of inlining
4546 that has already been done in the current function. Therefore,
4547 seemingly insignificant changes in the source program can cause the
4548 warnings produced by @option{-Winline} to appear or disappear.
4550 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4551 @opindex Wno-invalid-offsetof
4552 @opindex Winvalid-offsetof
4553 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4554 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4555 to a non-POD type is undefined. In existing C++ implementations,
4556 however, @samp{offsetof} typically gives meaningful results even when
4557 applied to certain kinds of non-POD types. (Such as a simple
4558 @samp{struct} that fails to be a POD type only by virtue of having a
4559 constructor.) This flag is for users who are aware that they are
4560 writing nonportable code and who have deliberately chosen to ignore the
4563 The restrictions on @samp{offsetof} may be relaxed in a future version
4564 of the C++ standard.
4566 @item -Wno-int-to-pointer-cast
4567 @opindex Wno-int-to-pointer-cast
4568 @opindex Wint-to-pointer-cast
4569 Suppress warnings from casts to pointer type of an integer of a
4570 different size. In C++, casting to a pointer type of smaller size is
4571 an error. @option{Wint-to-pointer-cast} is enabled by default.
4574 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4575 @opindex Wno-pointer-to-int-cast
4576 @opindex Wpointer-to-int-cast
4577 Suppress warnings from casts from a pointer to an integer type of a
4581 @opindex Winvalid-pch
4582 @opindex Wno-invalid-pch
4583 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4584 the search path but can't be used.
4588 @opindex Wno-long-long
4589 Warn if @samp{long long} type is used. This is enabled by either
4590 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4591 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4593 @item -Wvariadic-macros
4594 @opindex Wvariadic-macros
4595 @opindex Wno-variadic-macros
4596 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4597 alternate syntax when in pedantic ISO C99 mode. This is default.
4598 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4600 @item -Wvector-operation-performance
4601 @opindex Wvector-operation-performance
4602 @opindex Wno-vector-operation-performance
4603 Warn if vector operation is not implemented via SIMD capabilities of the
4604 architecture. Mainly useful for the performance tuning.
4605 Vector operation can be implemented @code{piecewise} which means that the
4606 scalar operation is performed on every vector element;
4607 @code{in parallel} which means that the vector operation is implemented
4608 using scalars of wider type, which normally is more performance efficient;
4609 and @code{as a single scalar} which means that vector fits into a
4615 Warn if variable length array is used in the code.
4616 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4617 the variable length array.
4619 @item -Wvolatile-register-var
4620 @opindex Wvolatile-register-var
4621 @opindex Wno-volatile-register-var
4622 Warn if a register variable is declared volatile. The volatile
4623 modifier does not inhibit all optimizations that may eliminate reads
4624 and/or writes to register variables. This warning is enabled by
4627 @item -Wdisabled-optimization
4628 @opindex Wdisabled-optimization
4629 @opindex Wno-disabled-optimization
4630 Warn if a requested optimization pass is disabled. This warning does
4631 not generally indicate that there is anything wrong with your code; it
4632 merely indicates that GCC's optimizers were unable to handle the code
4633 effectively. Often, the problem is that your code is too big or too
4634 complex; GCC will refuse to optimize programs when the optimization
4635 itself is likely to take inordinate amounts of time.
4637 @item -Wpointer-sign @r{(C and Objective-C only)}
4638 @opindex Wpointer-sign
4639 @opindex Wno-pointer-sign
4640 Warn for pointer argument passing or assignment with different signedness.
4641 This option is only supported for C and Objective-C@. It is implied by
4642 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4643 @option{-Wno-pointer-sign}.
4645 @item -Wstack-protector
4646 @opindex Wstack-protector
4647 @opindex Wno-stack-protector
4648 This option is only active when @option{-fstack-protector} is active. It
4649 warns about functions that will not be protected against stack smashing.
4652 @opindex Wno-mudflap
4653 Suppress warnings about constructs that cannot be instrumented by
4656 @item -Woverlength-strings
4657 @opindex Woverlength-strings
4658 @opindex Wno-overlength-strings
4659 Warn about string constants which are longer than the ``minimum
4660 maximum'' length specified in the C standard. Modern compilers
4661 generally allow string constants which are much longer than the
4662 standard's minimum limit, but very portable programs should avoid
4663 using longer strings.
4665 The limit applies @emph{after} string constant concatenation, and does
4666 not count the trailing NUL@. In C90, the limit was 509 characters; in
4667 C99, it was raised to 4095. C++98 does not specify a normative
4668 minimum maximum, so we do not diagnose overlength strings in C++@.
4670 This option is implied by @option{-pedantic}, and can be disabled with
4671 @option{-Wno-overlength-strings}.
4673 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4674 @opindex Wunsuffixed-float-constants
4676 GCC will issue a warning for any floating constant that does not have
4677 a suffix. When used together with @option{-Wsystem-headers} it will
4678 warn about such constants in system header files. This can be useful
4679 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4680 from the decimal floating-point extension to C99.
4683 @node Debugging Options
4684 @section Options for Debugging Your Program or GCC
4685 @cindex options, debugging
4686 @cindex debugging information options
4688 GCC has various special options that are used for debugging
4689 either your program or GCC:
4694 Produce debugging information in the operating system's native format
4695 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4698 On most systems that use stabs format, @option{-g} enables use of extra
4699 debugging information that only GDB can use; this extra information
4700 makes debugging work better in GDB but will probably make other debuggers
4702 refuse to read the program. If you want to control for certain whether
4703 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4704 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4706 GCC allows you to use @option{-g} with
4707 @option{-O}. The shortcuts taken by optimized code may occasionally
4708 produce surprising results: some variables you declared may not exist
4709 at all; flow of control may briefly move where you did not expect it;
4710 some statements may not be executed because they compute constant
4711 results or their values were already at hand; some statements may
4712 execute in different places because they were moved out of loops.
4714 Nevertheless it proves possible to debug optimized output. This makes
4715 it reasonable to use the optimizer for programs that might have bugs.
4717 The following options are useful when GCC is generated with the
4718 capability for more than one debugging format.
4722 Produce debugging information for use by GDB@. This means to use the
4723 most expressive format available (DWARF 2, stabs, or the native format
4724 if neither of those are supported), including GDB extensions if at all
4729 Produce debugging information in stabs format (if that is supported),
4730 without GDB extensions. This is the format used by DBX on most BSD
4731 systems. On MIPS, Alpha and System V Release 4 systems this option
4732 produces stabs debugging output which is not understood by DBX or SDB@.
4733 On System V Release 4 systems this option requires the GNU assembler.
4735 @item -feliminate-unused-debug-symbols
4736 @opindex feliminate-unused-debug-symbols
4737 Produce debugging information in stabs format (if that is supported),
4738 for only symbols that are actually used.
4740 @item -femit-class-debug-always
4741 Instead of emitting debugging information for a C++ class in only one
4742 object file, emit it in all object files using the class. This option
4743 should be used only with debuggers that are unable to handle the way GCC
4744 normally emits debugging information for classes because using this
4745 option will increase the size of debugging information by as much as a
4748 @item -fno-debug-types-section
4749 @opindex fno-debug-types-section
4750 @opindex fdebug-types-section
4751 By default when using DWARF v4 or higher type DIEs will be put into
4752 their own .debug_types section instead of making them part of the
4753 .debug_info section. It is more efficient to put them in a separate
4754 comdat sections since the linker will then be able to remove duplicates.
4755 But not all DWARF consumers support .debug_types sections yet.
4759 Produce debugging information in stabs format (if that is supported),
4760 using GNU extensions understood only by the GNU debugger (GDB)@. The
4761 use of these extensions is likely to make other debuggers crash or
4762 refuse to read the program.
4766 Produce debugging information in COFF format (if that is supported).
4767 This is the format used by SDB on most System V systems prior to
4772 Produce debugging information in XCOFF format (if that is supported).
4773 This is the format used by the DBX debugger on IBM RS/6000 systems.
4777 Produce debugging information in XCOFF format (if that is supported),
4778 using GNU extensions understood only by the GNU debugger (GDB)@. The
4779 use of these extensions is likely to make other debuggers crash or
4780 refuse to read the program, and may cause assemblers other than the GNU
4781 assembler (GAS) to fail with an error.
4783 @item -gdwarf-@var{version}
4784 @opindex gdwarf-@var{version}
4785 Produce debugging information in DWARF format (if that is
4786 supported). This is the format used by DBX on IRIX 6. The value
4787 of @var{version} may be either 2, 3 or 4; the default version is 2.
4789 Note that with DWARF version 2 some ports require, and will always
4790 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4792 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4793 for maximum benefit.
4795 @item -grecord-gcc-switches
4796 @opindex grecord-gcc-switches
4797 This switch causes the command-line options, that were used to invoke the
4798 compiler and may affect code generation, to be appended to the
4799 DW_AT_producer attribute in DWARF debugging information. The options
4800 are concatenated with spaces separating them from each other and from
4801 the compiler version. See also @option{-frecord-gcc-switches} for another
4802 way of storing compiler options into the object file.
4804 @item -gno-record-gcc-switches
4805 @opindex gno-record-gcc-switches
4806 Disallow appending command-line options to the DW_AT_producer attribute
4807 in DWARF debugging information. This is the default.
4809 @item -gstrict-dwarf
4810 @opindex gstrict-dwarf
4811 Disallow using extensions of later DWARF standard version than selected
4812 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4813 DWARF extensions from later standard versions is allowed.
4815 @item -gno-strict-dwarf
4816 @opindex gno-strict-dwarf
4817 Allow using extensions of later DWARF standard version than selected with
4818 @option{-gdwarf-@var{version}}.
4822 Produce debugging information in VMS debug format (if that is
4823 supported). This is the format used by DEBUG on VMS systems.
4826 @itemx -ggdb@var{level}
4827 @itemx -gstabs@var{level}
4828 @itemx -gcoff@var{level}
4829 @itemx -gxcoff@var{level}
4830 @itemx -gvms@var{level}
4831 Request debugging information and also use @var{level} to specify how
4832 much information. The default level is 2.
4834 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4837 Level 1 produces minimal information, enough for making backtraces in
4838 parts of the program that you don't plan to debug. This includes
4839 descriptions of functions and external variables, but no information
4840 about local variables and no line numbers.
4842 Level 3 includes extra information, such as all the macro definitions
4843 present in the program. Some debuggers support macro expansion when
4844 you use @option{-g3}.
4846 @option{-gdwarf-2} does not accept a concatenated debug level, because
4847 GCC used to support an option @option{-gdwarf} that meant to generate
4848 debug information in version 1 of the DWARF format (which is very
4849 different from version 2), and it would have been too confusing. That
4850 debug format is long obsolete, but the option cannot be changed now.
4851 Instead use an additional @option{-g@var{level}} option to change the
4852 debug level for DWARF.
4856 Turn off generation of debug info, if leaving out this option would have
4857 generated it, or turn it on at level 2 otherwise. The position of this
4858 argument in the command line does not matter, it takes effect after all
4859 other options are processed, and it does so only once, no matter how
4860 many times it is given. This is mainly intended to be used with
4861 @option{-fcompare-debug}.
4863 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4864 @opindex fdump-final-insns
4865 Dump the final internal representation (RTL) to @var{file}. If the
4866 optional argument is omitted (or if @var{file} is @code{.}), the name
4867 of the dump file will be determined by appending @code{.gkd} to the
4868 compilation output file name.
4870 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4871 @opindex fcompare-debug
4872 @opindex fno-compare-debug
4873 If no error occurs during compilation, run the compiler a second time,
4874 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4875 passed to the second compilation. Dump the final internal
4876 representation in both compilations, and print an error if they differ.
4878 If the equal sign is omitted, the default @option{-gtoggle} is used.
4880 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4881 and nonzero, implicitly enables @option{-fcompare-debug}. If
4882 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4883 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4886 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4887 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4888 of the final representation and the second compilation, preventing even
4889 @env{GCC_COMPARE_DEBUG} from taking effect.
4891 To verify full coverage during @option{-fcompare-debug} testing, set
4892 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4893 which GCC will reject as an invalid option in any actual compilation
4894 (rather than preprocessing, assembly or linking). To get just a
4895 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4896 not overridden} will do.
4898 @item -fcompare-debug-second
4899 @opindex fcompare-debug-second
4900 This option is implicitly passed to the compiler for the second
4901 compilation requested by @option{-fcompare-debug}, along with options to
4902 silence warnings, and omitting other options that would cause
4903 side-effect compiler outputs to files or to the standard output. Dump
4904 files and preserved temporary files are renamed so as to contain the
4905 @code{.gk} additional extension during the second compilation, to avoid
4906 overwriting those generated by the first.
4908 When this option is passed to the compiler driver, it causes the
4909 @emph{first} compilation to be skipped, which makes it useful for little
4910 other than debugging the compiler proper.
4912 @item -feliminate-dwarf2-dups
4913 @opindex feliminate-dwarf2-dups
4914 Compress DWARF2 debugging information by eliminating duplicated
4915 information about each symbol. This option only makes sense when
4916 generating DWARF2 debugging information with @option{-gdwarf-2}.
4918 @item -femit-struct-debug-baseonly
4919 Emit debug information for struct-like types
4920 only when the base name of the compilation source file
4921 matches the base name of file in which the struct was defined.
4923 This option substantially reduces the size of debugging information,
4924 but at significant potential loss in type information to the debugger.
4925 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4926 See @option{-femit-struct-debug-detailed} for more detailed control.
4928 This option works only with DWARF 2.
4930 @item -femit-struct-debug-reduced
4931 Emit debug information for struct-like types
4932 only when the base name of the compilation source file
4933 matches the base name of file in which the type was defined,
4934 unless the struct is a template or defined in a system header.
4936 This option significantly reduces the size of debugging information,
4937 with some potential loss in type information to the debugger.
4938 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4939 See @option{-femit-struct-debug-detailed} for more detailed control.
4941 This option works only with DWARF 2.
4943 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4944 Specify the struct-like types
4945 for which the compiler will generate debug information.
4946 The intent is to reduce duplicate struct debug information
4947 between different object files within the same program.
4949 This option is a detailed version of
4950 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4951 which will serve for most needs.
4953 A specification has the syntax@*
4954 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4956 The optional first word limits the specification to
4957 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4958 A struct type is used directly when it is the type of a variable, member.
4959 Indirect uses arise through pointers to structs.
4960 That is, when use of an incomplete struct would be legal, the use is indirect.
4962 @samp{struct one direct; struct two * indirect;}.
4964 The optional second word limits the specification to
4965 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4966 Generic structs are a bit complicated to explain.
4967 For C++, these are non-explicit specializations of template classes,
4968 or non-template classes within the above.
4969 Other programming languages have generics,
4970 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4972 The third word specifies the source files for those
4973 structs for which the compiler will emit debug information.
4974 The values @samp{none} and @samp{any} have the normal meaning.
4975 The value @samp{base} means that
4976 the base of name of the file in which the type declaration appears
4977 must match the base of the name of the main compilation file.
4978 In practice, this means that
4979 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4980 but types declared in other header will not.
4981 The value @samp{sys} means those types satisfying @samp{base}
4982 or declared in system or compiler headers.
4984 You may need to experiment to determine the best settings for your application.
4986 The default is @samp{-femit-struct-debug-detailed=all}.
4988 This option works only with DWARF 2.
4990 @item -fno-merge-debug-strings
4991 @opindex fmerge-debug-strings
4992 @opindex fno-merge-debug-strings
4993 Direct the linker to not merge together strings in the debugging
4994 information which are identical in different object files. Merging is
4995 not supported by all assemblers or linkers. Merging decreases the size
4996 of the debug information in the output file at the cost of increasing
4997 link processing time. Merging is enabled by default.
4999 @item -fdebug-prefix-map=@var{old}=@var{new}
5000 @opindex fdebug-prefix-map
5001 When compiling files in directory @file{@var{old}}, record debugging
5002 information describing them as in @file{@var{new}} instead.
5004 @item -fno-dwarf2-cfi-asm
5005 @opindex fdwarf2-cfi-asm
5006 @opindex fno-dwarf2-cfi-asm
5007 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5008 instead of using GAS @code{.cfi_*} directives.
5010 @cindex @command{prof}
5013 Generate extra code to write profile information suitable for the
5014 analysis program @command{prof}. You must use this option when compiling
5015 the source files you want data about, and you must also use it when
5018 @cindex @command{gprof}
5021 Generate extra code to write profile information suitable for the
5022 analysis program @command{gprof}. You must use this option when compiling
5023 the source files you want data about, and you must also use it when
5028 Makes the compiler print out each function name as it is compiled, and
5029 print some statistics about each pass when it finishes.
5032 @opindex ftime-report
5033 Makes the compiler print some statistics about the time consumed by each
5034 pass when it finishes.
5037 @opindex fmem-report
5038 Makes the compiler print some statistics about permanent memory
5039 allocation when it finishes.
5041 @item -fpre-ipa-mem-report
5042 @opindex fpre-ipa-mem-report
5043 @item -fpost-ipa-mem-report
5044 @opindex fpost-ipa-mem-report
5045 Makes the compiler print some statistics about permanent memory
5046 allocation before or after interprocedural optimization.
5049 @opindex fstack-usage
5050 Makes the compiler output stack usage information for the program, on a
5051 per-function basis. The filename for the dump is made by appending
5052 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5053 the output file, if explicitly specified and it is not an executable,
5054 otherwise it is the basename of the source file. An entry is made up
5059 The name of the function.
5063 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5066 The qualifier @code{static} means that the function manipulates the stack
5067 statically: a fixed number of bytes are allocated for the frame on function
5068 entry and released on function exit; no stack adjustments are otherwise made
5069 in the function. The second field is this fixed number of bytes.
5071 The qualifier @code{dynamic} means that the function manipulates the stack
5072 dynamically: in addition to the static allocation described above, stack
5073 adjustments are made in the body of the function, for example to push/pop
5074 arguments around function calls. If the qualifier @code{bounded} is also
5075 present, the amount of these adjustments is bounded at compile time and
5076 the second field is an upper bound of the total amount of stack used by
5077 the function. If it is not present, the amount of these adjustments is
5078 not bounded at compile time and the second field only represents the
5081 @item -fprofile-arcs
5082 @opindex fprofile-arcs
5083 Add code so that program flow @dfn{arcs} are instrumented. During
5084 execution the program records how many times each branch and call is
5085 executed and how many times it is taken or returns. When the compiled
5086 program exits it saves this data to a file called
5087 @file{@var{auxname}.gcda} for each source file. The data may be used for
5088 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5089 test coverage analysis (@option{-ftest-coverage}). Each object file's
5090 @var{auxname} is generated from the name of the output file, if
5091 explicitly specified and it is not the final executable, otherwise it is
5092 the basename of the source file. In both cases any suffix is removed
5093 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5094 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5095 @xref{Cross-profiling}.
5097 @cindex @command{gcov}
5101 This option is used to compile and link code instrumented for coverage
5102 analysis. The option is a synonym for @option{-fprofile-arcs}
5103 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5104 linking). See the documentation for those options for more details.
5109 Compile the source files with @option{-fprofile-arcs} plus optimization
5110 and code generation options. For test coverage analysis, use the
5111 additional @option{-ftest-coverage} option. You do not need to profile
5112 every source file in a program.
5115 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5116 (the latter implies the former).
5119 Run the program on a representative workload to generate the arc profile
5120 information. This may be repeated any number of times. You can run
5121 concurrent instances of your program, and provided that the file system
5122 supports locking, the data files will be correctly updated. Also
5123 @code{fork} calls are detected and correctly handled (double counting
5127 For profile-directed optimizations, compile the source files again with
5128 the same optimization and code generation options plus
5129 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5130 Control Optimization}).
5133 For test coverage analysis, use @command{gcov} to produce human readable
5134 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5135 @command{gcov} documentation for further information.
5139 With @option{-fprofile-arcs}, for each function of your program GCC
5140 creates a program flow graph, then finds a spanning tree for the graph.
5141 Only arcs that are not on the spanning tree have to be instrumented: the
5142 compiler adds code to count the number of times that these arcs are
5143 executed. When an arc is the only exit or only entrance to a block, the
5144 instrumentation code can be added to the block; otherwise, a new basic
5145 block must be created to hold the instrumentation code.
5148 @item -ftest-coverage
5149 @opindex ftest-coverage
5150 Produce a notes file that the @command{gcov} code-coverage utility
5151 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5152 show program coverage. Each source file's note file is called
5153 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5154 above for a description of @var{auxname} and instructions on how to
5155 generate test coverage data. Coverage data will match the source files
5156 more closely, if you do not optimize.
5158 @item -fdbg-cnt-list
5159 @opindex fdbg-cnt-list
5160 Print the name and the counter upper bound for all debug counters.
5163 @item -fdbg-cnt=@var{counter-value-list}
5165 Set the internal debug counter upper bound. @var{counter-value-list}
5166 is a comma-separated list of @var{name}:@var{value} pairs
5167 which sets the upper bound of each debug counter @var{name} to @var{value}.
5168 All debug counters have the initial upper bound of @var{UINT_MAX},
5169 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5170 e.g. With -fdbg-cnt=dce:10,tail_call:0
5171 dbg_cnt(dce) will return true only for first 10 invocations
5173 @itemx -fenable-@var{kind}-@var{pass}
5174 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5178 This is a set of debugging options that are used to explicitly disable/enable
5179 optimization passes. For compiler users, regular options for enabling/disabling
5180 passes should be used instead.
5184 @item -fdisable-ipa-@var{pass}
5185 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5186 statically invoked in the compiler multiple times, the pass name should be
5187 appended with a sequential number starting from 1.
5189 @item -fdisable-rtl-@var{pass}
5190 @item -fdisable-rtl-@var{pass}=@var{range-list}
5191 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5192 statically invoked in the compiler multiple times, the pass name should be
5193 appended with a sequential number starting from 1. @var{range-list} is a comma
5194 seperated list of function ranges or assembler names. Each range is a number
5195 pair seperated by a colon. The range is inclusive in both ends. If the range
5196 is trivial, the number pair can be simplified as a single number. If the
5197 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5198 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5199 function header of a dump file, and the pass names can be dumped by using
5200 option @option{-fdump-passes}.
5202 @item -fdisable-tree-@var{pass}
5203 @item -fdisable-tree-@var{pass}=@var{range-list}
5204 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5207 @item -fenable-ipa-@var{pass}
5208 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5209 statically invoked in the compiler multiple times, the pass name should be
5210 appended with a sequential number starting from 1.
5212 @item -fenable-rtl-@var{pass}
5213 @item -fenable-rtl-@var{pass}=@var{range-list}
5214 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5215 description and examples.
5217 @item -fenable-tree-@var{pass}
5218 @item -fenable-tree-@var{pass}=@var{range-list}
5219 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5220 of option arguments.
5224 # disable ccp1 for all functions
5226 # disable complete unroll for function whose cgraph node uid is 1
5227 -fenable-tree-cunroll=1
5228 # disable gcse2 for functions at the following ranges [1,1],
5229 # [300,400], and [400,1000]
5230 # disable gcse2 for functions foo and foo2
5231 -fdisable-rtl-gcse2=foo,foo2
5232 # disable early inlining
5233 -fdisable-tree-einline
5234 # disable ipa inlining
5235 -fdisable-ipa-inline
5236 # enable tree full unroll
5237 -fenable-tree-unroll
5243 @item -d@var{letters}
5244 @itemx -fdump-rtl-@var{pass}
5246 Says to make debugging dumps during compilation at times specified by
5247 @var{letters}. This is used for debugging the RTL-based passes of the
5248 compiler. The file names for most of the dumps are made by appending
5249 a pass number and a word to the @var{dumpname}, and the files are
5250 created in the directory of the output file. Note that the pass
5251 number is computed statically as passes get registered into the pass
5252 manager. Thus the numbering is not related to the dynamic order of
5253 execution of passes. In particular, a pass installed by a plugin
5254 could have a number over 200 even if it executed quite early.
5255 @var{dumpname} is generated from the name of the output file, if
5256 explicitly specified and it is not an executable, otherwise it is the
5257 basename of the source file. These switches may have different effects
5258 when @option{-E} is used for preprocessing.
5260 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5261 @option{-d} option @var{letters}. Here are the possible
5262 letters for use in @var{pass} and @var{letters}, and their meanings:
5266 @item -fdump-rtl-alignments
5267 @opindex fdump-rtl-alignments
5268 Dump after branch alignments have been computed.
5270 @item -fdump-rtl-asmcons
5271 @opindex fdump-rtl-asmcons
5272 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5274 @item -fdump-rtl-auto_inc_dec
5275 @opindex fdump-rtl-auto_inc_dec
5276 Dump after auto-inc-dec discovery. This pass is only run on
5277 architectures that have auto inc or auto dec instructions.
5279 @item -fdump-rtl-barriers
5280 @opindex fdump-rtl-barriers
5281 Dump after cleaning up the barrier instructions.
5283 @item -fdump-rtl-bbpart
5284 @opindex fdump-rtl-bbpart
5285 Dump after partitioning hot and cold basic blocks.
5287 @item -fdump-rtl-bbro
5288 @opindex fdump-rtl-bbro
5289 Dump after block reordering.
5291 @item -fdump-rtl-btl1
5292 @itemx -fdump-rtl-btl2
5293 @opindex fdump-rtl-btl2
5294 @opindex fdump-rtl-btl2
5295 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5296 after the two branch
5297 target load optimization passes.
5299 @item -fdump-rtl-bypass
5300 @opindex fdump-rtl-bypass
5301 Dump after jump bypassing and control flow optimizations.
5303 @item -fdump-rtl-combine
5304 @opindex fdump-rtl-combine
5305 Dump after the RTL instruction combination pass.
5307 @item -fdump-rtl-compgotos
5308 @opindex fdump-rtl-compgotos
5309 Dump after duplicating the computed gotos.
5311 @item -fdump-rtl-ce1
5312 @itemx -fdump-rtl-ce2
5313 @itemx -fdump-rtl-ce3
5314 @opindex fdump-rtl-ce1
5315 @opindex fdump-rtl-ce2
5316 @opindex fdump-rtl-ce3
5317 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5318 @option{-fdump-rtl-ce3} enable dumping after the three
5319 if conversion passes.
5321 @itemx -fdump-rtl-cprop_hardreg
5322 @opindex fdump-rtl-cprop_hardreg
5323 Dump after hard register copy propagation.
5325 @itemx -fdump-rtl-csa
5326 @opindex fdump-rtl-csa
5327 Dump after combining stack adjustments.
5329 @item -fdump-rtl-cse1
5330 @itemx -fdump-rtl-cse2
5331 @opindex fdump-rtl-cse1
5332 @opindex fdump-rtl-cse2
5333 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5334 the two common sub-expression elimination passes.
5336 @itemx -fdump-rtl-dce
5337 @opindex fdump-rtl-dce
5338 Dump after the standalone dead code elimination passes.
5340 @itemx -fdump-rtl-dbr
5341 @opindex fdump-rtl-dbr
5342 Dump after delayed branch scheduling.
5344 @item -fdump-rtl-dce1
5345 @itemx -fdump-rtl-dce2
5346 @opindex fdump-rtl-dce1
5347 @opindex fdump-rtl-dce2
5348 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5349 the two dead store elimination passes.
5352 @opindex fdump-rtl-eh
5353 Dump after finalization of EH handling code.
5355 @item -fdump-rtl-eh_ranges
5356 @opindex fdump-rtl-eh_ranges
5357 Dump after conversion of EH handling range regions.
5359 @item -fdump-rtl-expand
5360 @opindex fdump-rtl-expand
5361 Dump after RTL generation.
5363 @item -fdump-rtl-fwprop1
5364 @itemx -fdump-rtl-fwprop2
5365 @opindex fdump-rtl-fwprop1
5366 @opindex fdump-rtl-fwprop2
5367 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5368 dumping after the two forward propagation passes.
5370 @item -fdump-rtl-gcse1
5371 @itemx -fdump-rtl-gcse2
5372 @opindex fdump-rtl-gcse1
5373 @opindex fdump-rtl-gcse2
5374 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5375 after global common subexpression elimination.
5377 @item -fdump-rtl-init-regs
5378 @opindex fdump-rtl-init-regs
5379 Dump after the initialization of the registers.
5381 @item -fdump-rtl-initvals
5382 @opindex fdump-rtl-initvals
5383 Dump after the computation of the initial value sets.
5385 @itemx -fdump-rtl-into_cfglayout
5386 @opindex fdump-rtl-into_cfglayout
5387 Dump after converting to cfglayout mode.
5389 @item -fdump-rtl-ira
5390 @opindex fdump-rtl-ira
5391 Dump after iterated register allocation.
5393 @item -fdump-rtl-jump
5394 @opindex fdump-rtl-jump
5395 Dump after the second jump optimization.
5397 @item -fdump-rtl-loop2
5398 @opindex fdump-rtl-loop2
5399 @option{-fdump-rtl-loop2} enables dumping after the rtl
5400 loop optimization passes.
5402 @item -fdump-rtl-mach
5403 @opindex fdump-rtl-mach
5404 Dump after performing the machine dependent reorganization pass, if that
5407 @item -fdump-rtl-mode_sw
5408 @opindex fdump-rtl-mode_sw
5409 Dump after removing redundant mode switches.
5411 @item -fdump-rtl-rnreg
5412 @opindex fdump-rtl-rnreg
5413 Dump after register renumbering.
5415 @itemx -fdump-rtl-outof_cfglayout
5416 @opindex fdump-rtl-outof_cfglayout
5417 Dump after converting from cfglayout mode.
5419 @item -fdump-rtl-peephole2
5420 @opindex fdump-rtl-peephole2
5421 Dump after the peephole pass.
5423 @item -fdump-rtl-postreload
5424 @opindex fdump-rtl-postreload
5425 Dump after post-reload optimizations.
5427 @itemx -fdump-rtl-pro_and_epilogue
5428 @opindex fdump-rtl-pro_and_epilogue
5429 Dump after generating the function pro and epilogues.
5431 @item -fdump-rtl-regmove
5432 @opindex fdump-rtl-regmove
5433 Dump after the register move pass.
5435 @item -fdump-rtl-sched1
5436 @itemx -fdump-rtl-sched2
5437 @opindex fdump-rtl-sched1
5438 @opindex fdump-rtl-sched2
5439 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5440 after the basic block scheduling passes.
5442 @item -fdump-rtl-see
5443 @opindex fdump-rtl-see
5444 Dump after sign extension elimination.
5446 @item -fdump-rtl-seqabstr
5447 @opindex fdump-rtl-seqabstr
5448 Dump after common sequence discovery.
5450 @item -fdump-rtl-shorten
5451 @opindex fdump-rtl-shorten
5452 Dump after shortening branches.
5454 @item -fdump-rtl-sibling
5455 @opindex fdump-rtl-sibling
5456 Dump after sibling call optimizations.
5458 @item -fdump-rtl-split1
5459 @itemx -fdump-rtl-split2
5460 @itemx -fdump-rtl-split3
5461 @itemx -fdump-rtl-split4
5462 @itemx -fdump-rtl-split5
5463 @opindex fdump-rtl-split1
5464 @opindex fdump-rtl-split2
5465 @opindex fdump-rtl-split3
5466 @opindex fdump-rtl-split4
5467 @opindex fdump-rtl-split5
5468 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5469 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5470 @option{-fdump-rtl-split5} enable dumping after five rounds of
5471 instruction splitting.
5473 @item -fdump-rtl-sms
5474 @opindex fdump-rtl-sms
5475 Dump after modulo scheduling. This pass is only run on some
5478 @item -fdump-rtl-stack
5479 @opindex fdump-rtl-stack
5480 Dump after conversion from GCC's "flat register file" registers to the
5481 x87's stack-like registers. This pass is only run on x86 variants.
5483 @item -fdump-rtl-subreg1
5484 @itemx -fdump-rtl-subreg2
5485 @opindex fdump-rtl-subreg1
5486 @opindex fdump-rtl-subreg2
5487 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5488 the two subreg expansion passes.
5490 @item -fdump-rtl-unshare
5491 @opindex fdump-rtl-unshare
5492 Dump after all rtl has been unshared.
5494 @item -fdump-rtl-vartrack
5495 @opindex fdump-rtl-vartrack
5496 Dump after variable tracking.
5498 @item -fdump-rtl-vregs
5499 @opindex fdump-rtl-vregs
5500 Dump after converting virtual registers to hard registers.
5502 @item -fdump-rtl-web
5503 @opindex fdump-rtl-web
5504 Dump after live range splitting.
5506 @item -fdump-rtl-regclass
5507 @itemx -fdump-rtl-subregs_of_mode_init
5508 @itemx -fdump-rtl-subregs_of_mode_finish
5509 @itemx -fdump-rtl-dfinit
5510 @itemx -fdump-rtl-dfinish
5511 @opindex fdump-rtl-regclass
5512 @opindex fdump-rtl-subregs_of_mode_init
5513 @opindex fdump-rtl-subregs_of_mode_finish
5514 @opindex fdump-rtl-dfinit
5515 @opindex fdump-rtl-dfinish
5516 These dumps are defined but always produce empty files.
5518 @item -fdump-rtl-all
5519 @opindex fdump-rtl-all
5520 Produce all the dumps listed above.
5524 Annotate the assembler output with miscellaneous debugging information.
5528 Dump all macro definitions, at the end of preprocessing, in addition to
5533 Produce a core dump whenever an error occurs.
5537 Print statistics on memory usage, at the end of the run, to
5542 Annotate the assembler output with a comment indicating which
5543 pattern and alternative was used. The length of each instruction is
5548 Dump the RTL in the assembler output as a comment before each instruction.
5549 Also turns on @option{-dp} annotation.
5553 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5554 dump a representation of the control flow graph suitable for viewing with VCG
5555 to @file{@var{file}.@var{pass}.vcg}.
5559 Just generate RTL for a function instead of compiling it. Usually used
5560 with @option{-fdump-rtl-expand}.
5564 @opindex fdump-noaddr
5565 When doing debugging dumps, suppress address output. This makes it more
5566 feasible to use diff on debugging dumps for compiler invocations with
5567 different compiler binaries and/or different
5568 text / bss / data / heap / stack / dso start locations.
5570 @item -fdump-unnumbered
5571 @opindex fdump-unnumbered
5572 When doing debugging dumps, suppress instruction numbers and address output.
5573 This makes it more feasible to use diff on debugging dumps for compiler
5574 invocations with different options, in particular with and without
5577 @item -fdump-unnumbered-links
5578 @opindex fdump-unnumbered-links
5579 When doing debugging dumps (see @option{-d} option above), suppress
5580 instruction numbers for the links to the previous and next instructions
5583 @item -fdump-translation-unit @r{(C++ only)}
5584 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5585 @opindex fdump-translation-unit
5586 Dump a representation of the tree structure for the entire translation
5587 unit to a file. The file name is made by appending @file{.tu} to the
5588 source file name, and the file is created in the same directory as the
5589 output file. If the @samp{-@var{options}} form is used, @var{options}
5590 controls the details of the dump as described for the
5591 @option{-fdump-tree} options.
5593 @item -fdump-class-hierarchy @r{(C++ only)}
5594 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5595 @opindex fdump-class-hierarchy
5596 Dump a representation of each class's hierarchy and virtual function
5597 table layout to a file. The file name is made by appending
5598 @file{.class} to the source file name, and the file is created in the
5599 same directory as the output file. If the @samp{-@var{options}} form
5600 is used, @var{options} controls the details of the dump as described
5601 for the @option{-fdump-tree} options.
5603 @item -fdump-ipa-@var{switch}
5605 Control the dumping at various stages of inter-procedural analysis
5606 language tree to a file. The file name is generated by appending a
5607 switch specific suffix to the source file name, and the file is created
5608 in the same directory as the output file. The following dumps are
5613 Enables all inter-procedural analysis dumps.
5616 Dumps information about call-graph optimization, unused function removal,
5617 and inlining decisions.
5620 Dump after function inlining.
5625 @opindex fdump-passes
5626 Dump the list of optimization passes that are turned on and off by
5627 the current command-line options.
5629 @item -fdump-statistics-@var{option}
5630 @opindex fdump-statistics
5631 Enable and control dumping of pass statistics in a separate file. The
5632 file name is generated by appending a suffix ending in
5633 @samp{.statistics} to the source file name, and the file is created in
5634 the same directory as the output file. If the @samp{-@var{option}}
5635 form is used, @samp{-stats} will cause counters to be summed over the
5636 whole compilation unit while @samp{-details} will dump every event as
5637 the passes generate them. The default with no option is to sum
5638 counters for each function compiled.
5640 @item -fdump-tree-@var{switch}
5641 @itemx -fdump-tree-@var{switch}-@var{options}
5643 Control the dumping at various stages of processing the intermediate
5644 language tree to a file. The file name is generated by appending a
5645 switch specific suffix to the source file name, and the file is
5646 created in the same directory as the output file. If the
5647 @samp{-@var{options}} form is used, @var{options} is a list of
5648 @samp{-} separated options that control the details of the dump. Not
5649 all options are applicable to all dumps, those which are not
5650 meaningful will be ignored. The following options are available
5654 Print the address of each node. Usually this is not meaningful as it
5655 changes according to the environment and source file. Its primary use
5656 is for tying up a dump file with a debug environment.
5658 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5659 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5660 use working backward from mangled names in the assembly file.
5662 Inhibit dumping of members of a scope or body of a function merely
5663 because that scope has been reached. Only dump such items when they
5664 are directly reachable by some other path. When dumping pretty-printed
5665 trees, this option inhibits dumping the bodies of control structures.
5667 Print a raw representation of the tree. By default, trees are
5668 pretty-printed into a C-like representation.
5670 Enable more detailed dumps (not honored by every dump option).
5672 Enable dumping various statistics about the pass (not honored by every dump
5675 Enable showing basic block boundaries (disabled in raw dumps).
5677 Enable showing virtual operands for every statement.
5679 Enable showing line numbers for statements.
5681 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5683 Enable showing the tree dump for each statement.
5685 Enable showing the EH region number holding each statement.
5687 Enable showing scalar evolution analysis details.
5689 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5690 and @option{lineno}.
5693 The following tree dumps are possible:
5697 @opindex fdump-tree-original
5698 Dump before any tree based optimization, to @file{@var{file}.original}.
5701 @opindex fdump-tree-optimized
5702 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5705 @opindex fdump-tree-gimple
5706 Dump each function before and after the gimplification pass to a file. The
5707 file name is made by appending @file{.gimple} to the source file name.
5710 @opindex fdump-tree-cfg
5711 Dump the control flow graph of each function to a file. The file name is
5712 made by appending @file{.cfg} to the source file name.
5715 @opindex fdump-tree-vcg
5716 Dump the control flow graph of each function to a file in VCG format. The
5717 file name is made by appending @file{.vcg} to the source file name. Note
5718 that if the file contains more than one function, the generated file cannot
5719 be used directly by VCG@. You will need to cut and paste each function's
5720 graph into its own separate file first.
5723 @opindex fdump-tree-ch
5724 Dump each function after copying loop headers. The file name is made by
5725 appending @file{.ch} to the source file name.
5728 @opindex fdump-tree-ssa
5729 Dump SSA related information to a file. The file name is made by appending
5730 @file{.ssa} to the source file name.
5733 @opindex fdump-tree-alias
5734 Dump aliasing information for each function. The file name is made by
5735 appending @file{.alias} to the source file name.
5738 @opindex fdump-tree-ccp
5739 Dump each function after CCP@. The file name is made by appending
5740 @file{.ccp} to the source file name.
5743 @opindex fdump-tree-storeccp
5744 Dump each function after STORE-CCP@. The file name is made by appending
5745 @file{.storeccp} to the source file name.
5748 @opindex fdump-tree-pre
5749 Dump trees after partial redundancy elimination. The file name is made
5750 by appending @file{.pre} to the source file name.
5753 @opindex fdump-tree-fre
5754 Dump trees after full redundancy elimination. The file name is made
5755 by appending @file{.fre} to the source file name.
5758 @opindex fdump-tree-copyprop
5759 Dump trees after copy propagation. The file name is made
5760 by appending @file{.copyprop} to the source file name.
5762 @item store_copyprop
5763 @opindex fdump-tree-store_copyprop
5764 Dump trees after store copy-propagation. The file name is made
5765 by appending @file{.store_copyprop} to the source file name.
5768 @opindex fdump-tree-dce
5769 Dump each function after dead code elimination. The file name is made by
5770 appending @file{.dce} to the source file name.
5773 @opindex fdump-tree-mudflap
5774 Dump each function after adding mudflap instrumentation. The file name is
5775 made by appending @file{.mudflap} to the source file name.
5778 @opindex fdump-tree-sra
5779 Dump each function after performing scalar replacement of aggregates. The
5780 file name is made by appending @file{.sra} to the source file name.
5783 @opindex fdump-tree-sink
5784 Dump each function after performing code sinking. The file name is made
5785 by appending @file{.sink} to the source file name.
5788 @opindex fdump-tree-dom
5789 Dump each function after applying dominator tree optimizations. The file
5790 name is made by appending @file{.dom} to the source file name.
5793 @opindex fdump-tree-dse
5794 Dump each function after applying dead store elimination. The file
5795 name is made by appending @file{.dse} to the source file name.
5798 @opindex fdump-tree-phiopt
5799 Dump each function after optimizing PHI nodes into straightline code. The file
5800 name is made by appending @file{.phiopt} to the source file name.
5803 @opindex fdump-tree-forwprop
5804 Dump each function after forward propagating single use variables. The file
5805 name is made by appending @file{.forwprop} to the source file name.
5808 @opindex fdump-tree-copyrename
5809 Dump each function after applying the copy rename optimization. The file
5810 name is made by appending @file{.copyrename} to the source file name.
5813 @opindex fdump-tree-nrv
5814 Dump each function after applying the named return value optimization on
5815 generic trees. The file name is made by appending @file{.nrv} to the source
5819 @opindex fdump-tree-vect
5820 Dump each function after applying vectorization of loops. The file name is
5821 made by appending @file{.vect} to the source file name.
5824 @opindex fdump-tree-slp
5825 Dump each function after applying vectorization of basic blocks. The file name
5826 is made by appending @file{.slp} to the source file name.
5829 @opindex fdump-tree-vrp
5830 Dump each function after Value Range Propagation (VRP). The file name
5831 is made by appending @file{.vrp} to the source file name.
5834 @opindex fdump-tree-all
5835 Enable all the available tree dumps with the flags provided in this option.
5838 @item -ftree-vectorizer-verbose=@var{n}
5839 @opindex ftree-vectorizer-verbose
5840 This option controls the amount of debugging output the vectorizer prints.
5841 This information is written to standard error, unless
5842 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5843 in which case it is output to the usual dump listing file, @file{.vect}.
5844 For @var{n}=0 no diagnostic information is reported.
5845 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5846 and the total number of loops that got vectorized.
5847 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5848 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5849 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5850 level that @option{-fdump-tree-vect-stats} uses.
5851 Higher verbosity levels mean either more information dumped for each
5852 reported loop, or same amount of information reported for more loops:
5853 if @var{n}=3, vectorizer cost model information is reported.
5854 If @var{n}=4, alignment related information is added to the reports.
5855 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5856 memory access-patterns) is added to the reports.
5857 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5858 that did not pass the first analysis phase (i.e., may not be countable, or
5859 may have complicated control-flow).
5860 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5861 If @var{n}=8, SLP related information is added to the reports.
5862 For @var{n}=9, all the information the vectorizer generates during its
5863 analysis and transformation is reported. This is the same verbosity level
5864 that @option{-fdump-tree-vect-details} uses.
5866 @item -frandom-seed=@var{string}
5867 @opindex frandom-seed
5868 This option provides a seed that GCC uses when it would otherwise use
5869 random numbers. It is used to generate certain symbol names
5870 that have to be different in every compiled file. It is also used to
5871 place unique stamps in coverage data files and the object files that
5872 produce them. You can use the @option{-frandom-seed} option to produce
5873 reproducibly identical object files.
5875 The @var{string} should be different for every file you compile.
5877 @item -fsched-verbose=@var{n}
5878 @opindex fsched-verbose
5879 On targets that use instruction scheduling, this option controls the
5880 amount of debugging output the scheduler prints. This information is
5881 written to standard error, unless @option{-fdump-rtl-sched1} or
5882 @option{-fdump-rtl-sched2} is specified, in which case it is output
5883 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5884 respectively. However for @var{n} greater than nine, the output is
5885 always printed to standard error.
5887 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5888 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5889 For @var{n} greater than one, it also output basic block probabilities,
5890 detailed ready list information and unit/insn info. For @var{n} greater
5891 than two, it includes RTL at abort point, control-flow and regions info.
5892 And for @var{n} over four, @option{-fsched-verbose} also includes
5896 @itemx -save-temps=cwd
5898 Store the usual ``temporary'' intermediate files permanently; place them
5899 in the current directory and name them based on the source file. Thus,
5900 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5901 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5902 preprocessed @file{foo.i} output file even though the compiler now
5903 normally uses an integrated preprocessor.
5905 When used in combination with the @option{-x} command-line option,
5906 @option{-save-temps} is sensible enough to avoid over writing an
5907 input source file with the same extension as an intermediate file.
5908 The corresponding intermediate file may be obtained by renaming the
5909 source file before using @option{-save-temps}.
5911 If you invoke GCC in parallel, compiling several different source
5912 files that share a common base name in different subdirectories or the
5913 same source file compiled for multiple output destinations, it is
5914 likely that the different parallel compilers will interfere with each
5915 other, and overwrite the temporary files. For instance:
5918 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5919 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5922 may result in @file{foo.i} and @file{foo.o} being written to
5923 simultaneously by both compilers.
5925 @item -save-temps=obj
5926 @opindex save-temps=obj
5927 Store the usual ``temporary'' intermediate files permanently. If the
5928 @option{-o} option is used, the temporary files are based on the
5929 object file. If the @option{-o} option is not used, the
5930 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5935 gcc -save-temps=obj -c foo.c
5936 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5937 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5940 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5941 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5942 @file{dir2/yfoobar.o}.
5944 @item -time@r{[}=@var{file}@r{]}
5946 Report the CPU time taken by each subprocess in the compilation
5947 sequence. For C source files, this is the compiler proper and assembler
5948 (plus the linker if linking is done).
5950 Without the specification of an output file, the output looks like this:
5957 The first number on each line is the ``user time'', that is time spent
5958 executing the program itself. The second number is ``system time'',
5959 time spent executing operating system routines on behalf of the program.
5960 Both numbers are in seconds.
5962 With the specification of an output file, the output is appended to the
5963 named file, and it looks like this:
5966 0.12 0.01 cc1 @var{options}
5967 0.00 0.01 as @var{options}
5970 The ``user time'' and the ``system time'' are moved before the program
5971 name, and the options passed to the program are displayed, so that one
5972 can later tell what file was being compiled, and with which options.
5974 @item -fvar-tracking
5975 @opindex fvar-tracking
5976 Run variable tracking pass. It computes where variables are stored at each
5977 position in code. Better debugging information is then generated
5978 (if the debugging information format supports this information).
5980 It is enabled by default when compiling with optimization (@option{-Os},
5981 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5982 the debug info format supports it.
5984 @item -fvar-tracking-assignments
5985 @opindex fvar-tracking-assignments
5986 @opindex fno-var-tracking-assignments
5987 Annotate assignments to user variables early in the compilation and
5988 attempt to carry the annotations over throughout the compilation all the
5989 way to the end, in an attempt to improve debug information while
5990 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5992 It can be enabled even if var-tracking is disabled, in which case
5993 annotations will be created and maintained, but discarded at the end.
5995 @item -fvar-tracking-assignments-toggle
5996 @opindex fvar-tracking-assignments-toggle
5997 @opindex fno-var-tracking-assignments-toggle
5998 Toggle @option{-fvar-tracking-assignments}, in the same way that
5999 @option{-gtoggle} toggles @option{-g}.
6001 @item -print-file-name=@var{library}
6002 @opindex print-file-name
6003 Print the full absolute name of the library file @var{library} that
6004 would be used when linking---and don't do anything else. With this
6005 option, GCC does not compile or link anything; it just prints the
6008 @item -print-multi-directory
6009 @opindex print-multi-directory
6010 Print the directory name corresponding to the multilib selected by any
6011 other switches present in the command line. This directory is supposed
6012 to exist in @env{GCC_EXEC_PREFIX}.
6014 @item -print-multi-lib
6015 @opindex print-multi-lib
6016 Print the mapping from multilib directory names to compiler switches
6017 that enable them. The directory name is separated from the switches by
6018 @samp{;}, and each switch starts with an @samp{@@} instead of the
6019 @samp{-}, without spaces between multiple switches. This is supposed to
6020 ease shell-processing.
6022 @item -print-multi-os-directory
6023 @opindex print-multi-os-directory
6024 Print the path to OS libraries for the selected
6025 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6026 present in the @file{lib} subdirectory and no multilibs are used, this is
6027 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6028 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6029 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6030 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6032 @item -print-prog-name=@var{program}
6033 @opindex print-prog-name
6034 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6036 @item -print-libgcc-file-name
6037 @opindex print-libgcc-file-name
6038 Same as @option{-print-file-name=libgcc.a}.
6040 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6041 but you do want to link with @file{libgcc.a}. You can do
6044 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6047 @item -print-search-dirs
6048 @opindex print-search-dirs
6049 Print the name of the configured installation directory and a list of
6050 program and library directories @command{gcc} will search---and don't do anything else.
6052 This is useful when @command{gcc} prints the error message
6053 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6054 To resolve this you either need to put @file{cpp0} and the other compiler
6055 components where @command{gcc} expects to find them, or you can set the environment
6056 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6057 Don't forget the trailing @samp{/}.
6058 @xref{Environment Variables}.
6060 @item -print-sysroot
6061 @opindex print-sysroot
6062 Print the target sysroot directory that will be used during
6063 compilation. This is the target sysroot specified either at configure
6064 time or using the @option{--sysroot} option, possibly with an extra
6065 suffix that depends on compilation options. If no target sysroot is
6066 specified, the option prints nothing.
6068 @item -print-sysroot-headers-suffix
6069 @opindex print-sysroot-headers-suffix
6070 Print the suffix added to the target sysroot when searching for
6071 headers, or give an error if the compiler is not configured with such
6072 a suffix---and don't do anything else.
6075 @opindex dumpmachine
6076 Print the compiler's target machine (for example,
6077 @samp{i686-pc-linux-gnu})---and don't do anything else.
6080 @opindex dumpversion
6081 Print the compiler version (for example, @samp{3.0})---and don't do
6086 Print the compiler's built-in specs---and don't do anything else. (This
6087 is used when GCC itself is being built.) @xref{Spec Files}.
6089 @item -feliminate-unused-debug-types
6090 @opindex feliminate-unused-debug-types
6091 Normally, when producing DWARF2 output, GCC will emit debugging
6092 information for all types declared in a compilation
6093 unit, regardless of whether or not they are actually used
6094 in that compilation unit. Sometimes this is useful, such as
6095 if, in the debugger, you want to cast a value to a type that is
6096 not actually used in your program (but is declared). More often,
6097 however, this results in a significant amount of wasted space.
6098 With this option, GCC will avoid producing debug symbol output
6099 for types that are nowhere used in the source file being compiled.
6102 @node Optimize Options
6103 @section Options That Control Optimization
6104 @cindex optimize options
6105 @cindex options, optimization
6107 These options control various sorts of optimizations.
6109 Without any optimization option, the compiler's goal is to reduce the
6110 cost of compilation and to make debugging produce the expected
6111 results. Statements are independent: if you stop the program with a
6112 breakpoint between statements, you can then assign a new value to any
6113 variable or change the program counter to any other statement in the
6114 function and get exactly the results you would expect from the source
6117 Turning on optimization flags makes the compiler attempt to improve
6118 the performance and/or code size at the expense of compilation time
6119 and possibly the ability to debug the program.
6121 The compiler performs optimization based on the knowledge it has of the
6122 program. Compiling multiple files at once to a single output file mode allows
6123 the compiler to use information gained from all of the files when compiling
6126 Not all optimizations are controlled directly by a flag. Only
6127 optimizations that have a flag are listed in this section.
6129 Most optimizations are only enabled if an @option{-O} level is set on
6130 the command line. Otherwise they are disabled, even if individual
6131 optimization flags are specified.
6133 Depending on the target and how GCC was configured, a slightly different
6134 set of optimizations may be enabled at each @option{-O} level than
6135 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6136 to find out the exact set of optimizations that are enabled at each level.
6137 @xref{Overall Options}, for examples.
6144 Optimize. Optimizing compilation takes somewhat more time, and a lot
6145 more memory for a large function.
6147 With @option{-O}, the compiler tries to reduce code size and execution
6148 time, without performing any optimizations that take a great deal of
6151 @option{-O} turns on the following optimization flags:
6155 -fcprop-registers @gol
6158 -fdelayed-branch @gol
6160 -fguess-branch-probability @gol
6161 -fif-conversion2 @gol
6162 -fif-conversion @gol
6163 -fipa-pure-const @gol
6165 -fipa-reference @gol
6167 -fsplit-wide-types @gol
6169 -ftree-builtin-call-dce @gol
6172 -ftree-copyrename @gol
6174 -ftree-dominator-opts @gol
6176 -ftree-forwprop @gol
6184 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6185 where doing so does not interfere with debugging.
6189 Optimize even more. GCC performs nearly all supported optimizations
6190 that do not involve a space-speed tradeoff.
6191 As compared to @option{-O}, this option increases both compilation time
6192 and the performance of the generated code.
6194 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6195 also turns on the following optimization flags:
6196 @gccoptlist{-fthread-jumps @gol
6197 -falign-functions -falign-jumps @gol
6198 -falign-loops -falign-labels @gol
6201 -fcse-follow-jumps -fcse-skip-blocks @gol
6202 -fdelete-null-pointer-checks @gol
6204 -fexpensive-optimizations @gol
6205 -fgcse -fgcse-lm @gol
6206 -finline-small-functions @gol
6207 -findirect-inlining @gol
6209 -foptimize-sibling-calls @gol
6210 -fpartial-inlining @gol
6213 -freorder-blocks -freorder-functions @gol
6214 -frerun-cse-after-loop @gol
6215 -fsched-interblock -fsched-spec @gol
6216 -fschedule-insns -fschedule-insns2 @gol
6217 -fstrict-aliasing -fstrict-overflow @gol
6218 -ftree-switch-conversion -ftree-tail-merge @gol
6222 Please note the warning under @option{-fgcse} about
6223 invoking @option{-O2} on programs that use computed gotos.
6227 Optimize yet more. @option{-O3} turns on all optimizations specified
6228 by @option{-O2} and also turns on the @option{-finline-functions},
6229 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6230 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6231 @option{-fipa-cp-clone} options.
6235 Reduce compilation time and make debugging produce the expected
6236 results. This is the default.
6240 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6241 do not typically increase code size. It also performs further
6242 optimizations designed to reduce code size.
6244 @option{-Os} disables the following optimization flags:
6245 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6246 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6247 -fprefetch-loop-arrays -ftree-vect-loop-version}
6251 Disregard strict standards compliance. @option{-Ofast} enables all
6252 @option{-O3} optimizations. It also enables optimizations that are not
6253 valid for all standard compliant programs.
6254 It turns on @option{-ffast-math} and the Fortran-specific
6255 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6257 If you use multiple @option{-O} options, with or without level numbers,
6258 the last such option is the one that is effective.
6261 Options of the form @option{-f@var{flag}} specify machine-independent
6262 flags. Most flags have both positive and negative forms; the negative
6263 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6264 below, only one of the forms is listed---the one you typically will
6265 use. You can figure out the other form by either removing @samp{no-}
6268 The following options control specific optimizations. They are either
6269 activated by @option{-O} options or are related to ones that are. You
6270 can use the following flags in the rare cases when ``fine-tuning'' of
6271 optimizations to be performed is desired.
6274 @item -fno-default-inline
6275 @opindex fno-default-inline
6276 Do not make member functions inline by default merely because they are
6277 defined inside the class scope (C++ only). Otherwise, when you specify
6278 @w{@option{-O}}, member functions defined inside class scope are compiled
6279 inline by default; i.e., you don't need to add @samp{inline} in front of
6280 the member function name.
6282 @item -fno-defer-pop
6283 @opindex fno-defer-pop
6284 Always pop the arguments to each function call as soon as that function
6285 returns. For machines which must pop arguments after a function call,
6286 the compiler normally lets arguments accumulate on the stack for several
6287 function calls and pops them all at once.
6289 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6291 @item -fforward-propagate
6292 @opindex fforward-propagate
6293 Perform a forward propagation pass on RTL@. The pass tries to combine two
6294 instructions and checks if the result can be simplified. If loop unrolling
6295 is active, two passes are performed and the second is scheduled after
6298 This option is enabled by default at optimization levels @option{-O},
6299 @option{-O2}, @option{-O3}, @option{-Os}.
6301 @item -ffp-contract=@var{style}
6302 @opindex ffp-contract
6303 @option{-ffp-contract=off} disables floating-point expression contraction.
6304 @option{-ffp-contract=fast} enables floating-point expression contraction
6305 such as forming of fused multiply-add operations if the target has
6306 native support for them.
6307 @option{-ffp-contract=on} enables floating-point expression contraction
6308 if allowed by the language standard. This is currently not implemented
6309 and treated equal to @option{-ffp-contract=off}.
6311 The default is @option{-ffp-contract=fast}.
6313 @item -fomit-frame-pointer
6314 @opindex fomit-frame-pointer
6315 Don't keep the frame pointer in a register for functions that
6316 don't need one. This avoids the instructions to save, set up and
6317 restore frame pointers; it also makes an extra register available
6318 in many functions. @strong{It also makes debugging impossible on
6321 On some machines, such as the VAX, this flag has no effect, because
6322 the standard calling sequence automatically handles the frame pointer
6323 and nothing is saved by pretending it doesn't exist. The
6324 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6325 whether a target machine supports this flag. @xref{Registers,,Register
6326 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6328 Starting with GCC version 4.6, the default setting (when not optimizing for
6329 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6330 @option{-fomit-frame-pointer}. The default can be reverted to
6331 @option{-fno-omit-frame-pointer} by configuring GCC with the
6332 @option{--enable-frame-pointer} configure option.
6334 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6336 @item -foptimize-sibling-calls
6337 @opindex foptimize-sibling-calls
6338 Optimize sibling and tail recursive calls.
6340 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6344 Do not expand any functions inline apart from those marked with
6345 the @code{always_inline} attribute. This is the default when not
6348 Single functions can be exempted from inlining by marking them
6349 with the @code{noinline} attribute.
6351 @item -finline-small-functions
6352 @opindex finline-small-functions
6353 Integrate functions into their callers when their body is smaller than expected
6354 function call code (so overall size of program gets smaller). The compiler
6355 heuristically decides which functions are simple enough to be worth integrating
6356 in this way. This inlining applies to all functions, even those not declared
6359 Enabled at level @option{-O2}.
6361 @item -findirect-inlining
6362 @opindex findirect-inlining
6363 Inline also indirect calls that are discovered to be known at compile
6364 time thanks to previous inlining. This option has any effect only
6365 when inlining itself is turned on by the @option{-finline-functions}
6366 or @option{-finline-small-functions} options.
6368 Enabled at level @option{-O2}.
6370 @item -finline-functions
6371 @opindex finline-functions
6372 Consider all functions for inlining, even if they are not declared inline.
6373 The compiler heuristically decides which functions are worth integrating
6376 If all calls to a given function are integrated, and the function is
6377 declared @code{static}, then the function is normally not output as
6378 assembler code in its own right.
6380 Enabled at level @option{-O3}.
6382 @item -finline-functions-called-once
6383 @opindex finline-functions-called-once
6384 Consider all @code{static} functions called once for inlining into their
6385 caller even if they are not marked @code{inline}. If a call to a given
6386 function is integrated, then the function is not output as assembler code
6389 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6391 @item -fearly-inlining
6392 @opindex fearly-inlining
6393 Inline functions marked by @code{always_inline} and functions whose body seems
6394 smaller than the function call overhead early before doing
6395 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6396 makes profiling significantly cheaper and usually inlining faster on programs
6397 having large chains of nested wrapper functions.
6403 Perform interprocedural scalar replacement of aggregates, removal of
6404 unused parameters and replacement of parameters passed by reference
6405 by parameters passed by value.
6407 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6409 @item -finline-limit=@var{n}
6410 @opindex finline-limit
6411 By default, GCC limits the size of functions that can be inlined. This flag
6412 allows coarse control of this limit. @var{n} is the size of functions that
6413 can be inlined in number of pseudo instructions.
6415 Inlining is actually controlled by a number of parameters, which may be
6416 specified individually by using @option{--param @var{name}=@var{value}}.
6417 The @option{-finline-limit=@var{n}} option sets some of these parameters
6421 @item max-inline-insns-single
6422 is set to @var{n}/2.
6423 @item max-inline-insns-auto
6424 is set to @var{n}/2.
6427 See below for a documentation of the individual
6428 parameters controlling inlining and for the defaults of these parameters.
6430 @emph{Note:} there may be no value to @option{-finline-limit} that results
6431 in default behavior.
6433 @emph{Note:} pseudo instruction represents, in this particular context, an
6434 abstract measurement of function's size. In no way does it represent a count
6435 of assembly instructions and as such its exact meaning might change from one
6436 release to an another.
6438 @item -fno-keep-inline-dllexport
6439 @opindex -fno-keep-inline-dllexport
6440 This is a more fine-grained version of @option{-fkeep-inline-functions},
6441 which applies only to functions that are declared using the @code{dllexport}
6442 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6445 @item -fkeep-inline-functions
6446 @opindex fkeep-inline-functions
6447 In C, emit @code{static} functions that are declared @code{inline}
6448 into the object file, even if the function has been inlined into all
6449 of its callers. This switch does not affect functions using the
6450 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6451 inline functions into the object file.
6453 @item -fkeep-static-consts
6454 @opindex fkeep-static-consts
6455 Emit variables declared @code{static const} when optimization isn't turned
6456 on, even if the variables aren't referenced.
6458 GCC enables this option by default. If you want to force the compiler to
6459 check if the variable was referenced, regardless of whether or not
6460 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6462 @item -fmerge-constants
6463 @opindex fmerge-constants
6464 Attempt to merge identical constants (string constants and floating-point
6465 constants) across compilation units.
6467 This option is the default for optimized compilation if the assembler and
6468 linker support it. Use @option{-fno-merge-constants} to inhibit this
6471 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6473 @item -fmerge-all-constants
6474 @opindex fmerge-all-constants
6475 Attempt to merge identical constants and identical variables.
6477 This option implies @option{-fmerge-constants}. In addition to
6478 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6479 arrays or initialized constant variables with integral or floating-point
6480 types. Languages like C or C++ require each variable, including multiple
6481 instances of the same variable in recursive calls, to have distinct locations,
6482 so using this option will result in non-conforming
6485 @item -fmodulo-sched
6486 @opindex fmodulo-sched
6487 Perform swing modulo scheduling immediately before the first scheduling
6488 pass. This pass looks at innermost loops and reorders their
6489 instructions by overlapping different iterations.
6491 @item -fmodulo-sched-allow-regmoves
6492 @opindex fmodulo-sched-allow-regmoves
6493 Perform more aggressive SMS based modulo scheduling with register moves
6494 allowed. By setting this flag certain anti-dependences edges will be
6495 deleted which will trigger the generation of reg-moves based on the
6496 life-range analysis. This option is effective only with
6497 @option{-fmodulo-sched} enabled.
6499 @item -fno-branch-count-reg
6500 @opindex fno-branch-count-reg
6501 Do not use ``decrement and branch'' instructions on a count register,
6502 but instead generate a sequence of instructions that decrement a
6503 register, compare it against zero, then branch based upon the result.
6504 This option is only meaningful on architectures that support such
6505 instructions, which include x86, PowerPC, IA-64 and S/390.
6507 The default is @option{-fbranch-count-reg}.
6509 @item -fno-function-cse
6510 @opindex fno-function-cse
6511 Do not put function addresses in registers; make each instruction that
6512 calls a constant function contain the function's address explicitly.
6514 This option results in less efficient code, but some strange hacks
6515 that alter the assembler output may be confused by the optimizations
6516 performed when this option is not used.
6518 The default is @option{-ffunction-cse}
6520 @item -fno-zero-initialized-in-bss
6521 @opindex fno-zero-initialized-in-bss
6522 If the target supports a BSS section, GCC by default puts variables that
6523 are initialized to zero into BSS@. This can save space in the resulting
6526 This option turns off this behavior because some programs explicitly
6527 rely on variables going to the data section. E.g., so that the
6528 resulting executable can find the beginning of that section and/or make
6529 assumptions based on that.
6531 The default is @option{-fzero-initialized-in-bss}.
6533 @item -fmudflap -fmudflapth -fmudflapir
6537 @cindex bounds checking
6539 For front-ends that support it (C and C++), instrument all risky
6540 pointer/array dereferencing operations, some standard library
6541 string/heap functions, and some other associated constructs with
6542 range/validity tests. Modules so instrumented should be immune to
6543 buffer overflows, invalid heap use, and some other classes of C/C++
6544 programming errors. The instrumentation relies on a separate runtime
6545 library (@file{libmudflap}), which will be linked into a program if
6546 @option{-fmudflap} is given at link time. Run-time behavior of the
6547 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6548 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6551 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6552 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6553 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6554 instrumentation should ignore pointer reads. This produces less
6555 instrumentation (and therefore faster execution) and still provides
6556 some protection against outright memory corrupting writes, but allows
6557 erroneously read data to propagate within a program.
6559 @item -fthread-jumps
6560 @opindex fthread-jumps
6561 Perform optimizations where we check to see if a jump branches to a
6562 location where another comparison subsumed by the first is found. If
6563 so, the first branch is redirected to either the destination of the
6564 second branch or a point immediately following it, depending on whether
6565 the condition is known to be true or false.
6567 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6569 @item -fsplit-wide-types
6570 @opindex fsplit-wide-types
6571 When using a type that occupies multiple registers, such as @code{long
6572 long} on a 32-bit system, split the registers apart and allocate them
6573 independently. This normally generates better code for those types,
6574 but may make debugging more difficult.
6576 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6579 @item -fcse-follow-jumps
6580 @opindex fcse-follow-jumps
6581 In common subexpression elimination (CSE), scan through jump instructions
6582 when the target of the jump is not reached by any other path. For
6583 example, when CSE encounters an @code{if} statement with an
6584 @code{else} clause, CSE will follow the jump when the condition
6587 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6589 @item -fcse-skip-blocks
6590 @opindex fcse-skip-blocks
6591 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6592 follow jumps which conditionally skip over blocks. When CSE
6593 encounters a simple @code{if} statement with no else clause,
6594 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6595 body of the @code{if}.
6597 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6599 @item -frerun-cse-after-loop
6600 @opindex frerun-cse-after-loop
6601 Re-run common subexpression elimination after loop optimizations has been
6604 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6608 Perform a global common subexpression elimination pass.
6609 This pass also performs global constant and copy propagation.
6611 @emph{Note:} When compiling a program using computed gotos, a GCC
6612 extension, you may get better run-time performance if you disable
6613 the global common subexpression elimination pass by adding
6614 @option{-fno-gcse} to the command line.
6616 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6620 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6621 attempt to move loads which are only killed by stores into themselves. This
6622 allows a loop containing a load/store sequence to be changed to a load outside
6623 the loop, and a copy/store within the loop.
6625 Enabled by default when gcse is enabled.
6629 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6630 global common subexpression elimination. This pass will attempt to move
6631 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6632 loops containing a load/store sequence can be changed to a load before
6633 the loop and a store after the loop.
6635 Not enabled at any optimization level.
6639 When @option{-fgcse-las} is enabled, the global common subexpression
6640 elimination pass eliminates redundant loads that come after stores to the
6641 same memory location (both partial and full redundancies).
6643 Not enabled at any optimization level.
6645 @item -fgcse-after-reload
6646 @opindex fgcse-after-reload
6647 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6648 pass is performed after reload. The purpose of this pass is to cleanup
6651 @item -funsafe-loop-optimizations
6652 @opindex funsafe-loop-optimizations
6653 If given, the loop optimizer will assume that loop indices do not
6654 overflow, and that the loops with nontrivial exit condition are not
6655 infinite. This enables a wider range of loop optimizations even if
6656 the loop optimizer itself cannot prove that these assumptions are valid.
6657 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6658 if it finds this kind of loop.
6660 @item -fcrossjumping
6661 @opindex fcrossjumping
6662 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6663 resulting code may or may not perform better than without cross-jumping.
6665 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6667 @item -fauto-inc-dec
6668 @opindex fauto-inc-dec
6669 Combine increments or decrements of addresses with memory accesses.
6670 This pass is always skipped on architectures that do not have
6671 instructions to support this. Enabled by default at @option{-O} and
6672 higher on architectures that support this.
6676 Perform dead code elimination (DCE) on RTL@.
6677 Enabled by default at @option{-O} and higher.
6681 Perform dead store elimination (DSE) on RTL@.
6682 Enabled by default at @option{-O} and higher.
6684 @item -fif-conversion
6685 @opindex fif-conversion
6686 Attempt to transform conditional jumps into branch-less equivalents. This
6687 include use of conditional moves, min, max, set flags and abs instructions, and
6688 some tricks doable by standard arithmetics. The use of conditional execution
6689 on chips where it is available is controlled by @code{if-conversion2}.
6691 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6693 @item -fif-conversion2
6694 @opindex fif-conversion2
6695 Use conditional execution (where available) to transform conditional jumps into
6696 branch-less equivalents.
6698 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6700 @item -fdelete-null-pointer-checks
6701 @opindex fdelete-null-pointer-checks
6702 Assume that programs cannot safely dereference null pointers, and that
6703 no code or data element resides there. This enables simple constant
6704 folding optimizations at all optimization levels. In addition, other
6705 optimization passes in GCC use this flag to control global dataflow
6706 analyses that eliminate useless checks for null pointers; these assume
6707 that if a pointer is checked after it has already been dereferenced,
6710 Note however that in some environments this assumption is not true.
6711 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6712 for programs which depend on that behavior.
6714 Some targets, especially embedded ones, disable this option at all levels.
6715 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6716 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6717 are enabled independently at different optimization levels.
6719 @item -fdevirtualize
6720 @opindex fdevirtualize
6721 Attempt to convert calls to virtual functions to direct calls. This
6722 is done both within a procedure and interprocedurally as part of
6723 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6724 propagation (@option{-fipa-cp}).
6725 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6727 @item -fexpensive-optimizations
6728 @opindex fexpensive-optimizations
6729 Perform a number of minor optimizations that are relatively expensive.
6731 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6735 Attempt to remove redundant extension instructions. This is especially
6736 helpful for the x86-64 architecture which implicitly zero-extends in 64-bit
6737 registers after writing to their lower 32-bit half.
6739 Enabled for x86 at levels @option{-O2}, @option{-O3}.
6741 @item -foptimize-register-move
6743 @opindex foptimize-register-move
6745 Attempt to reassign register numbers in move instructions and as
6746 operands of other simple instructions in order to maximize the amount of
6747 register tying. This is especially helpful on machines with two-operand
6750 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6753 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6755 @item -fira-algorithm=@var{algorithm}
6756 Use specified coloring algorithm for the integrated register
6757 allocator. The @var{algorithm} argument should be @code{priority} or
6758 @code{CB}. The first algorithm specifies Chow's priority coloring,
6759 the second one specifies Chaitin-Briggs coloring. The second
6760 algorithm can be unimplemented for some architectures. If it is
6761 implemented, it is the default because Chaitin-Briggs coloring as a
6762 rule generates a better code.
6764 @item -fira-region=@var{region}
6765 Use specified regions for the integrated register allocator. The
6766 @var{region} argument should be one of @code{all}, @code{mixed}, or
6767 @code{one}. The first value means using all loops as register
6768 allocation regions, the second value which is enabled by default when
6769 compiling with optimization for speed (@option{-O}, @option{-O2},
6770 @dots{}) means using all loops except for loops with small register
6771 pressure as the regions, and third one which is enabled by default for
6772 @option{-Os} or @option{-O0} means using all function as a single
6773 region. The first value can give best result for machines with small
6774 size and irregular register set, the third one results in faster and
6775 generates decent code and the smallest size code, and the second value
6776 usually give the best results in most cases and for most
6779 @item -fira-loop-pressure
6780 @opindex fira-loop-pressure
6781 Use IRA to evaluate register pressure in loops for decision to move
6782 loop invariants. Usage of this option usually results in generation
6783 of faster and smaller code on machines with big register files (>= 32
6784 registers) but it can slow compiler down.
6786 This option is enabled at level @option{-O3} for some targets.
6788 @item -fno-ira-share-save-slots
6789 @opindex fno-ira-share-save-slots
6790 Switch off sharing stack slots used for saving call used hard
6791 registers living through a call. Each hard register will get a
6792 separate stack slot and as a result function stack frame will be
6795 @item -fno-ira-share-spill-slots
6796 @opindex fno-ira-share-spill-slots
6797 Switch off sharing stack slots allocated for pseudo-registers. Each
6798 pseudo-register which did not get a hard register will get a separate
6799 stack slot and as a result function stack frame will be bigger.
6801 @item -fira-verbose=@var{n}
6802 @opindex fira-verbose
6803 Set up how verbose dump file for the integrated register allocator
6804 will be. Default value is 5. If the value is greater or equal to 10,
6805 the dump file will be stderr as if the value were @var{n} minus 10.
6807 @item -fdelayed-branch
6808 @opindex fdelayed-branch
6809 If supported for the target machine, attempt to reorder instructions
6810 to exploit instruction slots available after delayed branch
6813 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6815 @item -fschedule-insns
6816 @opindex fschedule-insns
6817 If supported for the target machine, attempt to reorder instructions to
6818 eliminate execution stalls due to required data being unavailable. This
6819 helps machines that have slow floating point or memory load instructions
6820 by allowing other instructions to be issued until the result of the load
6821 or floating-point instruction is required.
6823 Enabled at levels @option{-O2}, @option{-O3}.
6825 @item -fschedule-insns2
6826 @opindex fschedule-insns2
6827 Similar to @option{-fschedule-insns}, but requests an additional pass of
6828 instruction scheduling after register allocation has been done. This is
6829 especially useful on machines with a relatively small number of
6830 registers and where memory load instructions take more than one cycle.
6832 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6834 @item -fno-sched-interblock
6835 @opindex fno-sched-interblock
6836 Don't schedule instructions across basic blocks. This is normally
6837 enabled by default when scheduling before register allocation, i.e.@:
6838 with @option{-fschedule-insns} or at @option{-O2} or higher.
6840 @item -fno-sched-spec
6841 @opindex fno-sched-spec
6842 Don't allow speculative motion of non-load instructions. This is normally
6843 enabled by default when scheduling before register allocation, i.e.@:
6844 with @option{-fschedule-insns} or at @option{-O2} or higher.
6846 @item -fsched-pressure
6847 @opindex fsched-pressure
6848 Enable register pressure sensitive insn scheduling before the register
6849 allocation. This only makes sense when scheduling before register
6850 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6851 @option{-O2} or higher. Usage of this option can improve the
6852 generated code and decrease its size by preventing register pressure
6853 increase above the number of available hard registers and as a
6854 consequence register spills in the register allocation.
6856 @item -fsched-spec-load
6857 @opindex fsched-spec-load
6858 Allow speculative motion of some load instructions. This only makes
6859 sense when scheduling before register allocation, i.e.@: with
6860 @option{-fschedule-insns} or at @option{-O2} or higher.
6862 @item -fsched-spec-load-dangerous
6863 @opindex fsched-spec-load-dangerous
6864 Allow speculative motion of more load instructions. This only makes
6865 sense when scheduling before register allocation, i.e.@: with
6866 @option{-fschedule-insns} or at @option{-O2} or higher.
6868 @item -fsched-stalled-insns
6869 @itemx -fsched-stalled-insns=@var{n}
6870 @opindex fsched-stalled-insns
6871 Define how many insns (if any) can be moved prematurely from the queue
6872 of stalled insns into the ready list, during the second scheduling pass.
6873 @option{-fno-sched-stalled-insns} means that no insns will be moved
6874 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6875 on how many queued insns can be moved prematurely.
6876 @option{-fsched-stalled-insns} without a value is equivalent to
6877 @option{-fsched-stalled-insns=1}.
6879 @item -fsched-stalled-insns-dep
6880 @itemx -fsched-stalled-insns-dep=@var{n}
6881 @opindex fsched-stalled-insns-dep
6882 Define how many insn groups (cycles) will be examined for a dependency
6883 on a stalled insn that is candidate for premature removal from the queue
6884 of stalled insns. This has an effect only during the second scheduling pass,
6885 and only if @option{-fsched-stalled-insns} is used.
6886 @option{-fno-sched-stalled-insns-dep} is equivalent to
6887 @option{-fsched-stalled-insns-dep=0}.
6888 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6889 @option{-fsched-stalled-insns-dep=1}.
6891 @item -fsched2-use-superblocks
6892 @opindex fsched2-use-superblocks
6893 When scheduling after register allocation, do use superblock scheduling
6894 algorithm. Superblock scheduling allows motion across basic block boundaries
6895 resulting on faster schedules. This option is experimental, as not all machine
6896 descriptions used by GCC model the CPU closely enough to avoid unreliable
6897 results from the algorithm.
6899 This only makes sense when scheduling after register allocation, i.e.@: with
6900 @option{-fschedule-insns2} or at @option{-O2} or higher.
6902 @item -fsched-group-heuristic
6903 @opindex fsched-group-heuristic
6904 Enable the group heuristic in the scheduler. This heuristic favors
6905 the instruction that belongs to a schedule group. This is enabled
6906 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6907 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6909 @item -fsched-critical-path-heuristic
6910 @opindex fsched-critical-path-heuristic
6911 Enable the critical-path heuristic in the scheduler. This heuristic favors
6912 instructions on the critical path. This is enabled by default when
6913 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6914 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6916 @item -fsched-spec-insn-heuristic
6917 @opindex fsched-spec-insn-heuristic
6918 Enable the speculative instruction heuristic in the scheduler. This
6919 heuristic favors speculative instructions with greater dependency weakness.
6920 This is enabled by default when scheduling is enabled, i.e.@:
6921 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6922 or at @option{-O2} or higher.
6924 @item -fsched-rank-heuristic
6925 @opindex fsched-rank-heuristic
6926 Enable the rank heuristic in the scheduler. This heuristic favors
6927 the instruction belonging to a basic block with greater size or frequency.
6928 This is enabled by default when scheduling is enabled, i.e.@:
6929 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6930 at @option{-O2} or higher.
6932 @item -fsched-last-insn-heuristic
6933 @opindex fsched-last-insn-heuristic
6934 Enable the last-instruction heuristic in the scheduler. This heuristic
6935 favors the instruction that is less dependent on the last instruction
6936 scheduled. This is enabled by default when scheduling is enabled,
6937 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6938 at @option{-O2} or higher.
6940 @item -fsched-dep-count-heuristic
6941 @opindex fsched-dep-count-heuristic
6942 Enable the dependent-count heuristic in the scheduler. This heuristic
6943 favors the instruction that has more instructions depending on it.
6944 This is enabled by default when scheduling is enabled, i.e.@:
6945 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6946 at @option{-O2} or higher.
6948 @item -freschedule-modulo-scheduled-loops
6949 @opindex freschedule-modulo-scheduled-loops
6950 The modulo scheduling comes before the traditional scheduling, if a loop
6951 was modulo scheduled we may want to prevent the later scheduling passes
6952 from changing its schedule, we use this option to control that.
6954 @item -fselective-scheduling
6955 @opindex fselective-scheduling
6956 Schedule instructions using selective scheduling algorithm. Selective
6957 scheduling runs instead of the first scheduler pass.
6959 @item -fselective-scheduling2
6960 @opindex fselective-scheduling2
6961 Schedule instructions using selective scheduling algorithm. Selective
6962 scheduling runs instead of the second scheduler pass.
6964 @item -fsel-sched-pipelining
6965 @opindex fsel-sched-pipelining
6966 Enable software pipelining of innermost loops during selective scheduling.
6967 This option has no effect until one of @option{-fselective-scheduling} or
6968 @option{-fselective-scheduling2} is turned on.
6970 @item -fsel-sched-pipelining-outer-loops
6971 @opindex fsel-sched-pipelining-outer-loops
6972 When pipelining loops during selective scheduling, also pipeline outer loops.
6973 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6976 @opindex fshrink-wrap
6977 Emit function prologues only before parts of the function that need it,
6978 rather than at the top of the function. This flag is enabled by default at
6979 @option{-O} and higher.
6981 @item -fcaller-saves
6982 @opindex fcaller-saves
6983 Enable values to be allocated in registers that will be clobbered by
6984 function calls, by emitting extra instructions to save and restore the
6985 registers around such calls. Such allocation is done only when it
6986 seems to result in better code than would otherwise be produced.
6988 This option is always enabled by default on certain machines, usually
6989 those which have no call-preserved registers to use instead.
6991 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6993 @item -fcombine-stack-adjustments
6994 @opindex fcombine-stack-adjustments
6995 Tracks stack adjustments (pushes and pops) and stack memory references
6996 and then tries to find ways to combine them.
6998 Enabled by default at @option{-O1} and higher.
7000 @item -fconserve-stack
7001 @opindex fconserve-stack
7002 Attempt to minimize stack usage. The compiler will attempt to use less
7003 stack space, even if that makes the program slower. This option
7004 implies setting the @option{large-stack-frame} parameter to 100
7005 and the @option{large-stack-frame-growth} parameter to 400.
7007 @item -ftree-reassoc
7008 @opindex ftree-reassoc
7009 Perform reassociation on trees. This flag is enabled by default
7010 at @option{-O} and higher.
7014 Perform partial redundancy elimination (PRE) on trees. This flag is
7015 enabled by default at @option{-O2} and @option{-O3}.
7017 @item -ftree-forwprop
7018 @opindex ftree-forwprop
7019 Perform forward propagation on trees. This flag is enabled by default
7020 at @option{-O} and higher.
7024 Perform full redundancy elimination (FRE) on trees. The difference
7025 between FRE and PRE is that FRE only considers expressions
7026 that are computed on all paths leading to the redundant computation.
7027 This analysis is faster than PRE, though it exposes fewer redundancies.
7028 This flag is enabled by default at @option{-O} and higher.
7030 @item -ftree-phiprop
7031 @opindex ftree-phiprop
7032 Perform hoisting of loads from conditional pointers on trees. This
7033 pass is enabled by default at @option{-O} and higher.
7035 @item -ftree-copy-prop
7036 @opindex ftree-copy-prop
7037 Perform copy propagation on trees. This pass eliminates unnecessary
7038 copy operations. This flag is enabled by default at @option{-O} and
7041 @item -fipa-pure-const
7042 @opindex fipa-pure-const
7043 Discover which functions are pure or constant.
7044 Enabled by default at @option{-O} and higher.
7046 @item -fipa-reference
7047 @opindex fipa-reference
7048 Discover which static variables do not escape cannot escape the
7050 Enabled by default at @option{-O} and higher.
7054 Perform interprocedural pointer analysis and interprocedural modification
7055 and reference analysis. This option can cause excessive memory and
7056 compile-time usage on large compilation units. It is not enabled by
7057 default at any optimization level.
7060 @opindex fipa-profile
7061 Perform interprocedural profile propagation. The functions called only from
7062 cold functions are marked as cold. Also functions executed once (such as
7063 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7064 functions and loop less parts of functions executed once are then optimized for
7066 Enabled by default at @option{-O} and higher.
7070 Perform interprocedural constant propagation.
7071 This optimization analyzes the program to determine when values passed
7072 to functions are constants and then optimizes accordingly.
7073 This optimization can substantially increase performance
7074 if the application has constants passed to functions.
7075 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7077 @item -fipa-cp-clone
7078 @opindex fipa-cp-clone
7079 Perform function cloning to make interprocedural constant propagation stronger.
7080 When enabled, interprocedural constant propagation will perform function cloning
7081 when externally visible function can be called with constant arguments.
7082 Because this optimization can create multiple copies of functions,
7083 it may significantly increase code size
7084 (see @option{--param ipcp-unit-growth=@var{value}}).
7085 This flag is enabled by default at @option{-O3}.
7087 @item -fipa-matrix-reorg
7088 @opindex fipa-matrix-reorg
7089 Perform matrix flattening and transposing.
7090 Matrix flattening tries to replace an @math{m}-dimensional matrix
7091 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7092 This reduces the level of indirection needed for accessing the elements
7093 of the matrix. The second optimization is matrix transposing that
7094 attempts to change the order of the matrix's dimensions in order to
7095 improve cache locality.
7096 Both optimizations need the @option{-fwhole-program} flag.
7097 Transposing is enabled only if profiling information is available.
7101 Perform forward store motion on trees. This flag is
7102 enabled by default at @option{-O} and higher.
7104 @item -ftree-bit-ccp
7105 @opindex ftree-bit-ccp
7106 Perform sparse conditional bit constant propagation on trees and propagate
7107 pointer alignment information.
7108 This pass only operates on local scalar variables and is enabled by default
7109 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7113 Perform sparse conditional constant propagation (CCP) on trees. This
7114 pass only operates on local scalar variables and is enabled by default
7115 at @option{-O} and higher.
7117 @item -ftree-switch-conversion
7118 Perform conversion of simple initializations in a switch to
7119 initializations from a scalar array. This flag is enabled by default
7120 at @option{-O2} and higher.
7122 @item -ftree-tail-merge
7123 Look for identical code sequences. When found, replace one with a jump to the
7124 other. This optimization is known as tail merging or cross jumping. This flag
7125 is enabled by default at @option{-O2} and higher. The compilation time
7127 be limited using @option{max-tail-merge-comparisons} parameter and
7128 @option{max-tail-merge-iterations} parameter.
7132 Perform dead code elimination (DCE) on trees. This flag is enabled by
7133 default at @option{-O} and higher.
7135 @item -ftree-builtin-call-dce
7136 @opindex ftree-builtin-call-dce
7137 Perform conditional dead code elimination (DCE) for calls to builtin functions
7138 that may set @code{errno} but are otherwise side-effect free. This flag is
7139 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7142 @item -ftree-dominator-opts
7143 @opindex ftree-dominator-opts
7144 Perform a variety of simple scalar cleanups (constant/copy
7145 propagation, redundancy elimination, range propagation and expression
7146 simplification) based on a dominator tree traversal. This also
7147 performs jump threading (to reduce jumps to jumps). This flag is
7148 enabled by default at @option{-O} and higher.
7152 Perform dead store elimination (DSE) on trees. A dead store is a store into
7153 a memory location which will later be overwritten by another store without
7154 any intervening loads. In this case the earlier store can be deleted. This
7155 flag is enabled by default at @option{-O} and higher.
7159 Perform loop header copying on trees. This is beneficial since it increases
7160 effectiveness of code motion optimizations. It also saves one jump. This flag
7161 is enabled by default at @option{-O} and higher. It is not enabled
7162 for @option{-Os}, since it usually increases code size.
7164 @item -ftree-loop-optimize
7165 @opindex ftree-loop-optimize
7166 Perform loop optimizations on trees. This flag is enabled by default
7167 at @option{-O} and higher.
7169 @item -ftree-loop-linear
7170 @opindex ftree-loop-linear
7171 Perform loop interchange transformations on tree. Same as
7172 @option{-floop-interchange}. To use this code transformation, GCC has
7173 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7174 enable the Graphite loop transformation infrastructure.
7176 @item -floop-interchange
7177 @opindex floop-interchange
7178 Perform loop interchange transformations on loops. Interchanging two
7179 nested loops switches the inner and outer loops. For example, given a
7184 A(J, I) = A(J, I) * C
7188 loop interchange will transform the loop as if the user had written:
7192 A(J, I) = A(J, I) * C
7196 which can be beneficial when @code{N} is larger than the caches,
7197 because in Fortran, the elements of an array are stored in memory
7198 contiguously by column, and the original loop iterates over rows,
7199 potentially creating at each access a cache miss. This optimization
7200 applies to all the languages supported by GCC and is not limited to
7201 Fortran. To use this code transformation, GCC has to be configured
7202 with @option{--with-ppl} and @option{--with-cloog} to enable the
7203 Graphite loop transformation infrastructure.
7205 @item -floop-strip-mine
7206 @opindex floop-strip-mine
7207 Perform loop strip mining transformations on loops. Strip mining
7208 splits a loop into two nested loops. The outer loop has strides
7209 equal to the strip size and the inner loop has strides of the
7210 original loop within a strip. The strip length can be changed
7211 using the @option{loop-block-tile-size} parameter. For example,
7218 loop strip mining will transform the loop as if the user had written:
7221 DO I = II, min (II + 50, N)
7226 This optimization applies to all the languages supported by GCC and is
7227 not limited to Fortran. To use this code transformation, GCC has to
7228 be configured with @option{--with-ppl} and @option{--with-cloog} to
7229 enable the Graphite loop transformation infrastructure.
7232 @opindex floop-block
7233 Perform loop blocking transformations on loops. Blocking strip mines
7234 each loop in the loop nest such that the memory accesses of the
7235 element loops fit inside caches. The strip length can be changed
7236 using the @option{loop-block-tile-size} parameter. For example, given
7241 A(J, I) = B(I) + C(J)
7245 loop blocking will transform the loop as if the user had written:
7249 DO I = II, min (II + 50, N)
7250 DO J = JJ, min (JJ + 50, M)
7251 A(J, I) = B(I) + C(J)
7257 which can be beneficial when @code{M} is larger than the caches,
7258 because the innermost loop will iterate over a smaller amount of data
7259 that can be kept in the caches. This optimization applies to all the
7260 languages supported by GCC and is not limited to Fortran. To use this
7261 code transformation, GCC has to be configured with @option{--with-ppl}
7262 and @option{--with-cloog} to enable the Graphite loop transformation
7265 @item -fgraphite-identity
7266 @opindex fgraphite-identity
7267 Enable the identity transformation for graphite. For every SCoP we generate
7268 the polyhedral representation and transform it back to gimple. Using
7269 @option{-fgraphite-identity} we can check the costs or benefits of the
7270 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7271 are also performed by the code generator CLooG, like index splitting and
7272 dead code elimination in loops.
7274 @item -floop-flatten
7275 @opindex floop-flatten
7276 Removes the loop nesting structure: transforms the loop nest into a
7277 single loop. This transformation can be useful to vectorize all the
7278 levels of the loop nest.
7280 @item -floop-parallelize-all
7281 @opindex floop-parallelize-all
7282 Use the Graphite data dependence analysis to identify loops that can
7283 be parallelized. Parallelize all the loops that can be analyzed to
7284 not contain loop carried dependences without checking that it is
7285 profitable to parallelize the loops.
7287 @item -fcheck-data-deps
7288 @opindex fcheck-data-deps
7289 Compare the results of several data dependence analyzers. This option
7290 is used for debugging the data dependence analyzers.
7292 @item -ftree-loop-if-convert
7293 Attempt to transform conditional jumps in the innermost loops to
7294 branch-less equivalents. The intent is to remove control-flow from
7295 the innermost loops in order to improve the ability of the
7296 vectorization pass to handle these loops. This is enabled by default
7297 if vectorization is enabled.
7299 @item -ftree-loop-if-convert-stores
7300 Attempt to also if-convert conditional jumps containing memory writes.
7301 This transformation can be unsafe for multi-threaded programs as it
7302 transforms conditional memory writes into unconditional memory writes.
7305 for (i = 0; i < N; i++)
7309 would be transformed to
7311 for (i = 0; i < N; i++)
7312 A[i] = cond ? expr : A[i];
7314 potentially producing data races.
7316 @item -ftree-loop-distribution
7317 Perform loop distribution. This flag can improve cache performance on
7318 big loop bodies and allow further loop optimizations, like
7319 parallelization or vectorization, to take place. For example, the loop
7336 @item -ftree-loop-distribute-patterns
7337 Perform loop distribution of patterns that can be code generated with
7338 calls to a library. This flag is enabled by default at @option{-O3}.
7340 This pass distributes the initialization loops and generates a call to
7341 memset zero. For example, the loop
7357 and the initialization loop is transformed into a call to memset zero.
7359 @item -ftree-loop-im
7360 @opindex ftree-loop-im
7361 Perform loop invariant motion on trees. This pass moves only invariants that
7362 would be hard to handle at RTL level (function calls, operations that expand to
7363 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7364 operands of conditions that are invariant out of the loop, so that we can use
7365 just trivial invariantness analysis in loop unswitching. The pass also includes
7368 @item -ftree-loop-ivcanon
7369 @opindex ftree-loop-ivcanon
7370 Create a canonical counter for number of iterations in the loop for that
7371 determining number of iterations requires complicated analysis. Later
7372 optimizations then may determine the number easily. Useful especially
7373 in connection with unrolling.
7377 Perform induction variable optimizations (strength reduction, induction
7378 variable merging and induction variable elimination) on trees.
7380 @item -ftree-parallelize-loops=n
7381 @opindex ftree-parallelize-loops
7382 Parallelize loops, i.e., split their iteration space to run in n threads.
7383 This is only possible for loops whose iterations are independent
7384 and can be arbitrarily reordered. The optimization is only
7385 profitable on multiprocessor machines, for loops that are CPU-intensive,
7386 rather than constrained e.g.@: by memory bandwidth. This option
7387 implies @option{-pthread}, and thus is only supported on targets
7388 that have support for @option{-pthread}.
7392 Perform function-local points-to analysis on trees. This flag is
7393 enabled by default at @option{-O} and higher.
7397 Perform scalar replacement of aggregates. This pass replaces structure
7398 references with scalars to prevent committing structures to memory too
7399 early. This flag is enabled by default at @option{-O} and higher.
7401 @item -ftree-copyrename
7402 @opindex ftree-copyrename
7403 Perform copy renaming on trees. This pass attempts to rename compiler
7404 temporaries to other variables at copy locations, usually resulting in
7405 variable names which more closely resemble the original variables. This flag
7406 is enabled by default at @option{-O} and higher.
7410 Perform temporary expression replacement during the SSA->normal phase. Single
7411 use/single def temporaries are replaced at their use location with their
7412 defining expression. This results in non-GIMPLE code, but gives the expanders
7413 much more complex trees to work on resulting in better RTL generation. This is
7414 enabled by default at @option{-O} and higher.
7416 @item -ftree-vectorize
7417 @opindex ftree-vectorize
7418 Perform loop vectorization on trees. This flag is enabled by default at
7421 @item -ftree-slp-vectorize
7422 @opindex ftree-slp-vectorize
7423 Perform basic block vectorization on trees. This flag is enabled by default at
7424 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7426 @item -ftree-vect-loop-version
7427 @opindex ftree-vect-loop-version
7428 Perform loop versioning when doing loop vectorization on trees. When a loop
7429 appears to be vectorizable except that data alignment or data dependence cannot
7430 be determined at compile time, then vectorized and non-vectorized versions of
7431 the loop are generated along with run-time checks for alignment or dependence
7432 to control which version is executed. This option is enabled by default
7433 except at level @option{-Os} where it is disabled.
7435 @item -fvect-cost-model
7436 @opindex fvect-cost-model
7437 Enable cost model for vectorization.
7441 Perform Value Range Propagation on trees. This is similar to the
7442 constant propagation pass, but instead of values, ranges of values are
7443 propagated. This allows the optimizers to remove unnecessary range
7444 checks like array bound checks and null pointer checks. This is
7445 enabled by default at @option{-O2} and higher. Null pointer check
7446 elimination is only done if @option{-fdelete-null-pointer-checks} is
7451 Perform tail duplication to enlarge superblock size. This transformation
7452 simplifies the control flow of the function allowing other optimizations to do
7455 @item -funroll-loops
7456 @opindex funroll-loops
7457 Unroll loops whose number of iterations can be determined at compile
7458 time or upon entry to the loop. @option{-funroll-loops} implies
7459 @option{-frerun-cse-after-loop}. This option makes code larger,
7460 and may or may not make it run faster.
7462 @item -funroll-all-loops
7463 @opindex funroll-all-loops
7464 Unroll all loops, even if their number of iterations is uncertain when
7465 the loop is entered. This usually makes programs run more slowly.
7466 @option{-funroll-all-loops} implies the same options as
7467 @option{-funroll-loops},
7469 @item -fsplit-ivs-in-unroller
7470 @opindex fsplit-ivs-in-unroller
7471 Enables expressing of values of induction variables in later iterations
7472 of the unrolled loop using the value in the first iteration. This breaks
7473 long dependency chains, thus improving efficiency of the scheduling passes.
7475 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7476 same effect. However in cases the loop body is more complicated than
7477 a single basic block, this is not reliable. It also does not work at all
7478 on some of the architectures due to restrictions in the CSE pass.
7480 This optimization is enabled by default.
7482 @item -fvariable-expansion-in-unroller
7483 @opindex fvariable-expansion-in-unroller
7484 With this option, the compiler will create multiple copies of some
7485 local variables when unrolling a loop which can result in superior code.
7487 @item -fpartial-inlining
7488 @opindex fpartial-inlining
7489 Inline parts of functions. This option has any effect only
7490 when inlining itself is turned on by the @option{-finline-functions}
7491 or @option{-finline-small-functions} options.
7493 Enabled at level @option{-O2}.
7495 @item -fpredictive-commoning
7496 @opindex fpredictive-commoning
7497 Perform predictive commoning optimization, i.e., reusing computations
7498 (especially memory loads and stores) performed in previous
7499 iterations of loops.
7501 This option is enabled at level @option{-O3}.
7503 @item -fprefetch-loop-arrays
7504 @opindex fprefetch-loop-arrays
7505 If supported by the target machine, generate instructions to prefetch
7506 memory to improve the performance of loops that access large arrays.
7508 This option may generate better or worse code; results are highly
7509 dependent on the structure of loops within the source code.
7511 Disabled at level @option{-Os}.
7514 @itemx -fno-peephole2
7515 @opindex fno-peephole
7516 @opindex fno-peephole2
7517 Disable any machine-specific peephole optimizations. The difference
7518 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7519 are implemented in the compiler; some targets use one, some use the
7520 other, a few use both.
7522 @option{-fpeephole} is enabled by default.
7523 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7525 @item -fno-guess-branch-probability
7526 @opindex fno-guess-branch-probability
7527 Do not guess branch probabilities using heuristics.
7529 GCC will use heuristics to guess branch probabilities if they are
7530 not provided by profiling feedback (@option{-fprofile-arcs}). These
7531 heuristics are based on the control flow graph. If some branch probabilities
7532 are specified by @samp{__builtin_expect}, then the heuristics will be
7533 used to guess branch probabilities for the rest of the control flow graph,
7534 taking the @samp{__builtin_expect} info into account. The interactions
7535 between the heuristics and @samp{__builtin_expect} can be complex, and in
7536 some cases, it may be useful to disable the heuristics so that the effects
7537 of @samp{__builtin_expect} are easier to understand.
7539 The default is @option{-fguess-branch-probability} at levels
7540 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7542 @item -freorder-blocks
7543 @opindex freorder-blocks
7544 Reorder basic blocks in the compiled function in order to reduce number of
7545 taken branches and improve code locality.
7547 Enabled at levels @option{-O2}, @option{-O3}.
7549 @item -freorder-blocks-and-partition
7550 @opindex freorder-blocks-and-partition
7551 In addition to reordering basic blocks in the compiled function, in order
7552 to reduce number of taken branches, partitions hot and cold basic blocks
7553 into separate sections of the assembly and .o files, to improve
7554 paging and cache locality performance.
7556 This optimization is automatically turned off in the presence of
7557 exception handling, for linkonce sections, for functions with a user-defined
7558 section attribute and on any architecture that does not support named
7561 @item -freorder-functions
7562 @opindex freorder-functions
7563 Reorder functions in the object file in order to
7564 improve code locality. This is implemented by using special
7565 subsections @code{.text.hot} for most frequently executed functions and
7566 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7567 the linker so object file format must support named sections and linker must
7568 place them in a reasonable way.
7570 Also profile feedback must be available in to make this option effective. See
7571 @option{-fprofile-arcs} for details.
7573 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7575 @item -fstrict-aliasing
7576 @opindex fstrict-aliasing
7577 Allow the compiler to assume the strictest aliasing rules applicable to
7578 the language being compiled. For C (and C++), this activates
7579 optimizations based on the type of expressions. In particular, an
7580 object of one type is assumed never to reside at the same address as an
7581 object of a different type, unless the types are almost the same. For
7582 example, an @code{unsigned int} can alias an @code{int}, but not a
7583 @code{void*} or a @code{double}. A character type may alias any other
7586 @anchor{Type-punning}Pay special attention to code like this:
7599 The practice of reading from a different union member than the one most
7600 recently written to (called ``type-punning'') is common. Even with
7601 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7602 is accessed through the union type. So, the code above will work as
7603 expected. @xref{Structures unions enumerations and bit-fields
7604 implementation}. However, this code might not:
7615 Similarly, access by taking the address, casting the resulting pointer
7616 and dereferencing the result has undefined behavior, even if the cast
7617 uses a union type, e.g.:
7621 return ((union a_union *) &d)->i;
7625 The @option{-fstrict-aliasing} option is enabled at levels
7626 @option{-O2}, @option{-O3}, @option{-Os}.
7628 @item -fstrict-overflow
7629 @opindex fstrict-overflow
7630 Allow the compiler to assume strict signed overflow rules, depending
7631 on the language being compiled. For C (and C++) this means that
7632 overflow when doing arithmetic with signed numbers is undefined, which
7633 means that the compiler may assume that it will not happen. This
7634 permits various optimizations. For example, the compiler will assume
7635 that an expression like @code{i + 10 > i} will always be true for
7636 signed @code{i}. This assumption is only valid if signed overflow is
7637 undefined, as the expression is false if @code{i + 10} overflows when
7638 using twos complement arithmetic. When this option is in effect any
7639 attempt to determine whether an operation on signed numbers will
7640 overflow must be written carefully to not actually involve overflow.
7642 This option also allows the compiler to assume strict pointer
7643 semantics: given a pointer to an object, if adding an offset to that
7644 pointer does not produce a pointer to the same object, the addition is
7645 undefined. This permits the compiler to conclude that @code{p + u >
7646 p} is always true for a pointer @code{p} and unsigned integer
7647 @code{u}. This assumption is only valid because pointer wraparound is
7648 undefined, as the expression is false if @code{p + u} overflows using
7649 twos complement arithmetic.
7651 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7652 that integer signed overflow is fully defined: it wraps. When
7653 @option{-fwrapv} is used, there is no difference between
7654 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7655 integers. With @option{-fwrapv} certain types of overflow are
7656 permitted. For example, if the compiler gets an overflow when doing
7657 arithmetic on constants, the overflowed value can still be used with
7658 @option{-fwrapv}, but not otherwise.
7660 The @option{-fstrict-overflow} option is enabled at levels
7661 @option{-O2}, @option{-O3}, @option{-Os}.
7663 @item -falign-functions
7664 @itemx -falign-functions=@var{n}
7665 @opindex falign-functions
7666 Align the start of functions to the next power-of-two greater than
7667 @var{n}, skipping up to @var{n} bytes. For instance,
7668 @option{-falign-functions=32} aligns functions to the next 32-byte
7669 boundary, but @option{-falign-functions=24} would align to the next
7670 32-byte boundary only if this can be done by skipping 23 bytes or less.
7672 @option{-fno-align-functions} and @option{-falign-functions=1} are
7673 equivalent and mean that functions will not be aligned.
7675 Some assemblers only support this flag when @var{n} is a power of two;
7676 in that case, it is rounded up.
7678 If @var{n} is not specified or is zero, use a machine-dependent default.
7680 Enabled at levels @option{-O2}, @option{-O3}.
7682 @item -falign-labels
7683 @itemx -falign-labels=@var{n}
7684 @opindex falign-labels
7685 Align all branch targets to a power-of-two boundary, skipping up to
7686 @var{n} bytes like @option{-falign-functions}. This option can easily
7687 make code slower, because it must insert dummy operations for when the
7688 branch target is reached in the usual flow of the code.
7690 @option{-fno-align-labels} and @option{-falign-labels=1} are
7691 equivalent and mean that labels will not be aligned.
7693 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7694 are greater than this value, then their values are used instead.
7696 If @var{n} is not specified or is zero, use a machine-dependent default
7697 which is very likely to be @samp{1}, meaning no alignment.
7699 Enabled at levels @option{-O2}, @option{-O3}.
7702 @itemx -falign-loops=@var{n}
7703 @opindex falign-loops
7704 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7705 like @option{-falign-functions}. The hope is that the loop will be
7706 executed many times, which will make up for any execution of the dummy
7709 @option{-fno-align-loops} and @option{-falign-loops=1} are
7710 equivalent and mean that loops will not be aligned.
7712 If @var{n} is not specified or is zero, use a machine-dependent default.
7714 Enabled at levels @option{-O2}, @option{-O3}.
7717 @itemx -falign-jumps=@var{n}
7718 @opindex falign-jumps
7719 Align branch targets to a power-of-two boundary, for branch targets
7720 where the targets can only be reached by jumping, skipping up to @var{n}
7721 bytes like @option{-falign-functions}. In this case, no dummy operations
7724 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7725 equivalent and mean that loops will not be aligned.
7727 If @var{n} is not specified or is zero, use a machine-dependent default.
7729 Enabled at levels @option{-O2}, @option{-O3}.
7731 @item -funit-at-a-time
7732 @opindex funit-at-a-time
7733 This option is left for compatibility reasons. @option{-funit-at-a-time}
7734 has no effect, while @option{-fno-unit-at-a-time} implies
7735 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7739 @item -fno-toplevel-reorder
7740 @opindex fno-toplevel-reorder
7741 Do not reorder top-level functions, variables, and @code{asm}
7742 statements. Output them in the same order that they appear in the
7743 input file. When this option is used, unreferenced static variables
7744 will not be removed. This option is intended to support existing code
7745 which relies on a particular ordering. For new code, it is better to
7748 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7749 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7754 Constructs webs as commonly used for register allocation purposes and assign
7755 each web individual pseudo register. This allows the register allocation pass
7756 to operate on pseudos directly, but also strengthens several other optimization
7757 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7758 however, make debugging impossible, since variables will no longer stay in a
7761 Enabled by default with @option{-funroll-loops}.
7763 @item -fwhole-program
7764 @opindex fwhole-program
7765 Assume that the current compilation unit represents the whole program being
7766 compiled. All public functions and variables with the exception of @code{main}
7767 and those merged by attribute @code{externally_visible} become static functions
7768 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.
7769 While this option is equivalent to proper use of the @code{static} keyword for
7770 programs consisting of a single file, in combination with option
7771 @option{-flto} this flag can be used to
7772 compile many smaller scale programs since the functions and variables become
7773 local for the whole combined compilation unit, not for the single source file
7776 This option implies @option{-fwhole-file} for Fortran programs.
7778 @item -flto[=@var{n}]
7780 This option runs the standard link-time optimizer. When invoked
7781 with source code, it generates GIMPLE (one of GCC's internal
7782 representations) and writes it to special ELF sections in the object
7783 file. When the object files are linked together, all the function
7784 bodies are read from these ELF sections and instantiated as if they
7785 had been part of the same translation unit.
7787 To use the link-time optimizer, @option{-flto} needs to be specified at
7788 compile time and during the final link. For example:
7791 gcc -c -O2 -flto foo.c
7792 gcc -c -O2 -flto bar.c
7793 gcc -o myprog -flto -O2 foo.o bar.o
7796 The first two invocations to GCC save a bytecode representation
7797 of GIMPLE into special ELF sections inside @file{foo.o} and
7798 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
7799 @file{foo.o} and @file{bar.o}, merges the two files into a single
7800 internal image, and compiles the result as usual. Since both
7801 @file{foo.o} and @file{bar.o} are merged into a single image, this
7802 causes all the interprocedural analyses and optimizations in GCC to
7803 work across the two files as if they were a single one. This means,
7804 for example, that the inliner is able to inline functions in
7805 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7807 Another (simpler) way to enable link-time optimization is:
7810 gcc -o myprog -flto -O2 foo.c bar.c
7813 The above generates bytecode for @file{foo.c} and @file{bar.c},
7814 merges them together into a single GIMPLE representation and optimizes
7815 them as usual to produce @file{myprog}.
7817 The only important thing to keep in mind is that to enable link-time
7818 optimizations the @option{-flto} flag needs to be passed to both the
7819 compile and the link commands.
7821 To make whole program optimization effective, it is necessary to make
7822 certain whole program assumptions. The compiler needs to know
7823 what functions and variables can be accessed by libraries and runtime
7824 outside of the link-time optimized unit. When supported by the linker,
7825 the linker plugin (see @option{-fuse-linker-plugin}) passes information
7826 to the compiler about used and externally visible symbols. When
7827 the linker plugin is not available, @option{-fwhole-program} should be
7828 used to allow the compiler to make these assumptions, which leads
7829 to more aggressive optimization decisions.
7831 Note that when a file is compiled with @option{-flto}, the generated
7832 object file is larger than a regular object file because it
7833 contains GIMPLE bytecodes and the usual final code. This means that
7834 object files with LTO information can be linked as normal object
7835 files; if @option{-flto} is not passed to the linker, no
7836 interprocedural optimizations are applied.
7838 Additionally, the optimization flags used to compile individual files
7839 are not necessarily related to those used at link time. For instance,
7842 gcc -c -O0 -flto foo.c
7843 gcc -c -O0 -flto bar.c
7844 gcc -o myprog -flto -O3 foo.o bar.o
7847 This produces individual object files with unoptimized assembler
7848 code, but the resulting binary @file{myprog} is optimized at
7849 @option{-O3}. If, instead, the final binary is generated without
7850 @option{-flto}, then @file{myprog} is not optimized.
7852 When producing the final binary with @option{-flto}, GCC only
7853 applies link-time optimizations to those files that contain bytecode.
7854 Therefore, you can mix and match object files and libraries with
7855 GIMPLE bytecodes and final object code. GCC automatically selects
7856 which files to optimize in LTO mode and which files to link without
7859 There are some code generation flags that GCC preserves when
7860 generating bytecodes, as they need to be used during the final link
7861 stage. Currently, the following options are saved into the GIMPLE
7862 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7863 @option{-m} target flags.
7865 At link time, these options are read in and reapplied. Note that the
7866 current implementation makes no attempt to recognize conflicting
7867 values for these options. If different files have conflicting option
7868 values (e.g., one file is compiled with @option{-fPIC} and another
7869 isn't), the compiler simply uses the last value read from the
7870 bytecode files. It is recommended, then, that you compile all the files
7871 participating in the same link with the same options.
7873 If LTO encounters objects with C linkage declared with incompatible
7874 types in separate translation units to be linked together (undefined
7875 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7876 issued. The behavior is still undefined at run time.
7878 Another feature of LTO is that it is possible to apply interprocedural
7879 optimizations on files written in different languages. This requires
7880 support in the language front end. Currently, the C, C++ and
7881 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7882 something like this should work:
7887 gfortran -c -flto baz.f90
7888 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7891 Notice that the final link is done with @command{g++} to get the C++
7892 runtime libraries and @option{-lgfortran} is added to get the Fortran
7893 runtime libraries. In general, when mixing languages in LTO mode, you
7894 should use the same link command options as when mixing languages in a
7895 regular (non-LTO) compilation; all you need to add is @option{-flto} to
7896 all the compile and link commands.
7898 If object files containing GIMPLE bytecode are stored in a library archive, say
7899 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7900 are using a linker with plugin support. To enable this feature, use
7901 the flag @option{-fuse-linker-plugin} at link time:
7904 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7907 With the linker plugin enabled, the linker extracts the needed
7908 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
7909 to make them part of the aggregated GIMPLE image to be optimized.
7911 If you are not using a linker with plugin support and/or do not
7912 enable the linker plugin, then the objects inside @file{libfoo.a}
7913 are extracted and linked as usual, but they do not participate
7914 in the LTO optimization process.
7916 Link-time optimizations do not require the presence of the whole program to
7917 operate. If the program does not require any symbols to be exported, it is
7918 possible to combine @option{-flto} and @option{-fwhole-program} to allow
7919 the interprocedural optimizers to use more aggressive assumptions which may
7920 lead to improved optimization opportunities.
7921 Use of @option{-fwhole-program} is not needed when linker plugin is
7922 active (see @option{-fuse-linker-plugin}).
7924 The current implementation of LTO makes no
7925 attempt to generate bytecode that is portable between different
7926 types of hosts. The bytecode files are versioned and there is a
7927 strict version check, so bytecode files generated in one version of
7928 GCC will not work with an older/newer version of GCC.
7930 Link-time optimization does not work well with generation of debugging
7931 information. Combining @option{-flto} with
7932 @option{-g} is currently experimental and expected to produce wrong
7935 If you specify the optional @var{n}, the optimization and code
7936 generation done at link time is executed in parallel using @var{n}
7937 parallel jobs by utilizing an installed @command{make} program. The
7938 environment variable @env{MAKE} may be used to override the program
7939 used. The default value for @var{n} is 1.
7941 You can also specify @option{-flto=jobserver} to use GNU make's
7942 job server mode to determine the number of parallel jobs. This
7943 is useful when the Makefile calling GCC is already executing in parallel.
7944 You must prepend a @samp{+} to the command recipe in the parent Makefile
7945 for this to work. This option likely only works if @env{MAKE} is
7948 This option is disabled by default
7950 @item -flto-partition=@var{alg}
7951 @opindex flto-partition
7952 Specify the partitioning algorithm used by the link-time optimizer.
7953 The value is either @code{1to1} to specify a partitioning mirroring
7954 the original source files or @code{balanced} to specify partitioning
7955 into equally sized chunks (whenever possible). Specifying @code{none}
7956 as an algorithm disables partitioning and streaming completely. The
7957 default value is @code{balanced}.
7959 @item -flto-compression-level=@var{n}
7960 This option specifies the level of compression used for intermediate
7961 language written to LTO object files, and is only meaningful in
7962 conjunction with LTO mode (@option{-flto}). Valid
7963 values are 0 (no compression) to 9 (maximum compression). Values
7964 outside this range are clamped to either 0 or 9. If the option is not
7965 given, a default balanced compression setting is used.
7968 Prints a report with internal details on the workings of the link-time
7969 optimizer. The contents of this report vary from version to version.
7970 It is meant to be useful to GCC developers when processing object
7971 files in LTO mode (via @option{-flto}).
7973 Disabled by default.
7975 @item -fuse-linker-plugin
7976 Enables the use of a linker plugin during link-time optimization. This
7977 option relies on plugin support in the linker, which is available in gold
7978 or in GNU ld 2.21 or newer.
7980 This option enables the extraction of object files with GIMPLE bytecode out
7981 of library archives. This improves the quality of optimization by exposing
7982 more code to the link-time optimizer. This information specifies what
7983 symbols can be accessed externally (by non-LTO object or during dynamic
7984 linking). Resulting code quality improvements on binaries (and shared
7985 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7986 See @option{-flto} for a description of the effect of this flag and how to
7989 This option is enabled by default when LTO support in GCC is enabled
7990 and GCC was configured for use with
7991 a linker supporting plugins (GNU ld 2.21 or newer or gold).
7993 @item -ffat-lto-objects
7994 @opindex ffat-lto-objects
7995 Fat LTO objects are object files that contain both the intermediate language
7996 and the object code. This makes them usable for both LTO linking and normal
7997 linking. This option is effective only when compiling with @option{-flto}
7998 and is ignored at link time.
8000 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8001 requires the complete toolchain to be aware of LTO. It requires a linker with
8002 linker plugin support for basic functionality. Additionally, nm, ar and ranlib
8003 need to support linker plugins to allow a full-featured build environment
8004 (capable of building static libraries etc).
8006 The default is @option{-ffat-lto-objects} but this default is intended to
8007 change in future releases when linker plugin enabled environments become more
8010 @item -fcompare-elim
8011 @opindex fcompare-elim
8012 After register allocation and post-register allocation instruction splitting,
8013 identify arithmetic instructions that compute processor flags similar to a
8014 comparison operation based on that arithmetic. If possible, eliminate the
8015 explicit comparison operation.
8017 This pass only applies to certain targets that cannot explicitly represent
8018 the comparison operation before register allocation is complete.
8020 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8022 @item -fcprop-registers
8023 @opindex fcprop-registers
8024 After register allocation and post-register allocation instruction splitting,
8025 we perform a copy-propagation pass to try to reduce scheduling dependencies
8026 and occasionally eliminate the copy.
8028 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8030 @item -fprofile-correction
8031 @opindex fprofile-correction
8032 Profiles collected using an instrumented binary for multi-threaded programs may
8033 be inconsistent due to missed counter updates. When this option is specified,
8034 GCC will use heuristics to correct or smooth out such inconsistencies. By
8035 default, GCC will emit an error message when an inconsistent profile is detected.
8037 @item -fprofile-dir=@var{path}
8038 @opindex fprofile-dir
8040 Set the directory to search for the profile data files in to @var{path}.
8041 This option affects only the profile data generated by
8042 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8043 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8044 and its related options. Both absolute and relative paths can be used.
8045 By default, GCC will use the current directory as @var{path}, thus the
8046 profile data file will appear in the same directory as the object file.
8048 @item -fprofile-generate
8049 @itemx -fprofile-generate=@var{path}
8050 @opindex fprofile-generate
8052 Enable options usually used for instrumenting application to produce
8053 profile useful for later recompilation with profile feedback based
8054 optimization. You must use @option{-fprofile-generate} both when
8055 compiling and when linking your program.
8057 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
8059 If @var{path} is specified, GCC will look at the @var{path} to find
8060 the profile feedback data files. See @option{-fprofile-dir}.
8063 @itemx -fprofile-use=@var{path}
8064 @opindex fprofile-use
8065 Enable profile feedback directed optimizations, and optimizations
8066 generally profitable only with profile feedback available.
8068 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8069 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
8071 By default, GCC emits an error message if the feedback profiles do not
8072 match the source code. This error can be turned into a warning by using
8073 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8076 If @var{path} is specified, GCC will look at the @var{path} to find
8077 the profile feedback data files. See @option{-fprofile-dir}.
8080 The following options control compiler behavior regarding floating-point
8081 arithmetic. These options trade off between speed and
8082 correctness. All must be specifically enabled.
8086 @opindex ffloat-store
8087 Do not store floating-point variables in registers, and inhibit other
8088 options that might change whether a floating-point value is taken from a
8091 @cindex floating-point precision
8092 This option prevents undesirable excess precision on machines such as
8093 the 68000 where the floating registers (of the 68881) keep more
8094 precision than a @code{double} is supposed to have. Similarly for the
8095 x86 architecture. For most programs, the excess precision does only
8096 good, but a few programs rely on the precise definition of IEEE floating
8097 point. Use @option{-ffloat-store} for such programs, after modifying
8098 them to store all pertinent intermediate computations into variables.
8100 @item -fexcess-precision=@var{style}
8101 @opindex fexcess-precision
8102 This option allows further control over excess precision on machines
8103 where floating-point registers have more precision than the IEEE
8104 @code{float} and @code{double} types and the processor does not
8105 support operations rounding to those types. By default,
8106 @option{-fexcess-precision=fast} is in effect; this means that
8107 operations are carried out in the precision of the registers and that
8108 it is unpredictable when rounding to the types specified in the source
8109 code takes place. When compiling C, if
8110 @option{-fexcess-precision=standard} is specified then excess
8111 precision will follow the rules specified in ISO C99; in particular,
8112 both casts and assignments cause values to be rounded to their
8113 semantic types (whereas @option{-ffloat-store} only affects
8114 assignments). This option is enabled by default for C if a strict
8115 conformance option such as @option{-std=c99} is used.
8118 @option{-fexcess-precision=standard} is not implemented for languages
8119 other than C, and has no effect if
8120 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8121 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8122 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8123 semantics apply without excess precision, and in the latter, rounding
8128 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8129 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8130 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8132 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8134 This option is not turned on by any @option{-O} option besides
8135 @option{-Ofast} since it can result in incorrect output for programs
8136 which depend on an exact implementation of IEEE or ISO rules/specifications
8137 for math functions. It may, however, yield faster code for programs
8138 that do not require the guarantees of these specifications.
8140 @item -fno-math-errno
8141 @opindex fno-math-errno
8142 Do not set ERRNO after calling math functions that are executed
8143 with a single instruction, e.g., sqrt. A program that relies on
8144 IEEE exceptions for math error handling may want to use this flag
8145 for speed while maintaining IEEE arithmetic compatibility.
8147 This option is not turned on by any @option{-O} option since
8148 it can result in incorrect output for programs which depend on
8149 an exact implementation of IEEE or ISO rules/specifications for
8150 math functions. It may, however, yield faster code for programs
8151 that do not require the guarantees of these specifications.
8153 The default is @option{-fmath-errno}.
8155 On Darwin systems, the math library never sets @code{errno}. There is
8156 therefore no reason for the compiler to consider the possibility that
8157 it might, and @option{-fno-math-errno} is the default.
8159 @item -funsafe-math-optimizations
8160 @opindex funsafe-math-optimizations
8162 Allow optimizations for floating-point arithmetic that (a) assume
8163 that arguments and results are valid and (b) may violate IEEE or
8164 ANSI standards. When used at link-time, it may include libraries
8165 or startup files that change the default FPU control word or other
8166 similar optimizations.
8168 This option is not turned on by any @option{-O} option since
8169 it can result in incorrect output for programs which depend on
8170 an exact implementation of IEEE or ISO rules/specifications for
8171 math functions. It may, however, yield faster code for programs
8172 that do not require the guarantees of these specifications.
8173 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8174 @option{-fassociative-math} and @option{-freciprocal-math}.
8176 The default is @option{-fno-unsafe-math-optimizations}.
8178 @item -fassociative-math
8179 @opindex fassociative-math
8181 Allow re-association of operands in series of floating-point operations.
8182 This violates the ISO C and C++ language standard by possibly changing
8183 computation result. NOTE: re-ordering may change the sign of zero as
8184 well as ignore NaNs and inhibit or create underflow or overflow (and
8185 thus cannot be used on a code which relies on rounding behavior like
8186 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8187 and thus may not be used when ordered comparisons are required.
8188 This option requires that both @option{-fno-signed-zeros} and
8189 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8190 much sense with @option{-frounding-math}. For Fortran the option
8191 is automatically enabled when both @option{-fno-signed-zeros} and
8192 @option{-fno-trapping-math} are in effect.
8194 The default is @option{-fno-associative-math}.
8196 @item -freciprocal-math
8197 @opindex freciprocal-math
8199 Allow the reciprocal of a value to be used instead of dividing by
8200 the value if this enables optimizations. For example @code{x / y}
8201 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8202 is subject to common subexpression elimination. Note that this loses
8203 precision and increases the number of flops operating on the value.
8205 The default is @option{-fno-reciprocal-math}.
8207 @item -ffinite-math-only
8208 @opindex ffinite-math-only
8209 Allow optimizations for floating-point arithmetic that assume
8210 that arguments and results are not NaNs or +-Infs.
8212 This option is not turned on by any @option{-O} option since
8213 it can result in incorrect output for programs which depend on
8214 an exact implementation of IEEE or ISO rules/specifications for
8215 math functions. It may, however, yield faster code for programs
8216 that do not require the guarantees of these specifications.
8218 The default is @option{-fno-finite-math-only}.
8220 @item -fno-signed-zeros
8221 @opindex fno-signed-zeros
8222 Allow optimizations for floating-point arithmetic that ignore the
8223 signedness of zero. IEEE arithmetic specifies the behavior of
8224 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8225 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8226 This option implies that the sign of a zero result isn't significant.
8228 The default is @option{-fsigned-zeros}.
8230 @item -fno-trapping-math
8231 @opindex fno-trapping-math
8232 Compile code assuming that floating-point operations cannot generate
8233 user-visible traps. These traps include division by zero, overflow,
8234 underflow, inexact result and invalid operation. This option requires
8235 that @option{-fno-signaling-nans} be in effect. Setting this option may
8236 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8238 This option should never be turned on by any @option{-O} option since
8239 it can result in incorrect output for programs which depend on
8240 an exact implementation of IEEE or ISO rules/specifications for
8243 The default is @option{-ftrapping-math}.
8245 @item -frounding-math
8246 @opindex frounding-math
8247 Disable transformations and optimizations that assume default floating-point
8248 rounding behavior. This is round-to-zero for all floating point
8249 to integer conversions, and round-to-nearest for all other arithmetic
8250 truncations. This option should be specified for programs that change
8251 the FP rounding mode dynamically, or that may be executed with a
8252 non-default rounding mode. This option disables constant folding of
8253 floating-point expressions at compile time (which may be affected by
8254 rounding mode) and arithmetic transformations that are unsafe in the
8255 presence of sign-dependent rounding modes.
8257 The default is @option{-fno-rounding-math}.
8259 This option is experimental and does not currently guarantee to
8260 disable all GCC optimizations that are affected by rounding mode.
8261 Future versions of GCC may provide finer control of this setting
8262 using C99's @code{FENV_ACCESS} pragma. This command-line option
8263 will be used to specify the default state for @code{FENV_ACCESS}.
8265 @item -fsignaling-nans
8266 @opindex fsignaling-nans
8267 Compile code assuming that IEEE signaling NaNs may generate user-visible
8268 traps during floating-point operations. Setting this option disables
8269 optimizations that may change the number of exceptions visible with
8270 signaling NaNs. This option implies @option{-ftrapping-math}.
8272 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8275 The default is @option{-fno-signaling-nans}.
8277 This option is experimental and does not currently guarantee to
8278 disable all GCC optimizations that affect signaling NaN behavior.
8280 @item -fsingle-precision-constant
8281 @opindex fsingle-precision-constant
8282 Treat floating-point constants as single precision instead of
8283 implicitly converting them to double-precision constants.
8285 @item -fcx-limited-range
8286 @opindex fcx-limited-range
8287 When enabled, this option states that a range reduction step is not
8288 needed when performing complex division. Also, there is no checking
8289 whether the result of a complex multiplication or division is @code{NaN
8290 + I*NaN}, with an attempt to rescue the situation in that case. The
8291 default is @option{-fno-cx-limited-range}, but is enabled by
8292 @option{-ffast-math}.
8294 This option controls the default setting of the ISO C99
8295 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8298 @item -fcx-fortran-rules
8299 @opindex fcx-fortran-rules
8300 Complex multiplication and division follow Fortran rules. Range
8301 reduction is done as part of complex division, but there is no checking
8302 whether the result of a complex multiplication or division is @code{NaN
8303 + I*NaN}, with an attempt to rescue the situation in that case.
8305 The default is @option{-fno-cx-fortran-rules}.
8309 The following options control optimizations that may improve
8310 performance, but are not enabled by any @option{-O} options. This
8311 section includes experimental options that may produce broken code.
8314 @item -fbranch-probabilities
8315 @opindex fbranch-probabilities
8316 After running a program compiled with @option{-fprofile-arcs}
8317 (@pxref{Debugging Options,, Options for Debugging Your Program or
8318 @command{gcc}}), you can compile it a second time using
8319 @option{-fbranch-probabilities}, to improve optimizations based on
8320 the number of times each branch was taken. When the program
8321 compiled with @option{-fprofile-arcs} exits it saves arc execution
8322 counts to a file called @file{@var{sourcename}.gcda} for each source
8323 file. The information in this data file is very dependent on the
8324 structure of the generated code, so you must use the same source code
8325 and the same optimization options for both compilations.
8327 With @option{-fbranch-probabilities}, GCC puts a
8328 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8329 These can be used to improve optimization. Currently, they are only
8330 used in one place: in @file{reorg.c}, instead of guessing which path a
8331 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8332 exactly determine which path is taken more often.
8334 @item -fprofile-values
8335 @opindex fprofile-values
8336 If combined with @option{-fprofile-arcs}, it adds code so that some
8337 data about values of expressions in the program is gathered.
8339 With @option{-fbranch-probabilities}, it reads back the data gathered
8340 from profiling values of expressions for usage in optimizations.
8342 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8346 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8347 a code to gather information about values of expressions.
8349 With @option{-fbranch-probabilities}, it reads back the data gathered
8350 and actually performs the optimizations based on them.
8351 Currently the optimizations include specialization of division operation
8352 using the knowledge about the value of the denominator.
8354 @item -frename-registers
8355 @opindex frename-registers
8356 Attempt to avoid false dependencies in scheduled code by making use
8357 of registers left over after register allocation. This optimization
8358 will most benefit processors with lots of registers. Depending on the
8359 debug information format adopted by the target, however, it can
8360 make debugging impossible, since variables will no longer stay in
8361 a ``home register''.
8363 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8367 Perform tail duplication to enlarge superblock size. This transformation
8368 simplifies the control flow of the function allowing other optimizations to do
8371 Enabled with @option{-fprofile-use}.
8373 @item -funroll-loops
8374 @opindex funroll-loops
8375 Unroll loops whose number of iterations can be determined at compile time or
8376 upon entry to the loop. @option{-funroll-loops} implies
8377 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8378 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8379 small constant number of iterations). This option makes code larger, and may
8380 or may not make it run faster.
8382 Enabled with @option{-fprofile-use}.
8384 @item -funroll-all-loops
8385 @opindex funroll-all-loops
8386 Unroll all loops, even if their number of iterations is uncertain when
8387 the loop is entered. This usually makes programs run more slowly.
8388 @option{-funroll-all-loops} implies the same options as
8389 @option{-funroll-loops}.
8392 @opindex fpeel-loops
8393 Peels the loops for that there is enough information that they do not
8394 roll much (from profile feedback). It also turns on complete loop peeling
8395 (i.e.@: complete removal of loops with small constant number of iterations).
8397 Enabled with @option{-fprofile-use}.
8399 @item -fmove-loop-invariants
8400 @opindex fmove-loop-invariants
8401 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8402 at level @option{-O1}
8404 @item -funswitch-loops
8405 @opindex funswitch-loops
8406 Move branches with loop invariant conditions out of the loop, with duplicates
8407 of the loop on both branches (modified according to result of the condition).
8409 @item -ffunction-sections
8410 @itemx -fdata-sections
8411 @opindex ffunction-sections
8412 @opindex fdata-sections
8413 Place each function or data item into its own section in the output
8414 file if the target supports arbitrary sections. The name of the
8415 function or the name of the data item determines the section's name
8418 Use these options on systems where the linker can perform optimizations
8419 to improve locality of reference in the instruction space. Most systems
8420 using the ELF object format and SPARC processors running Solaris 2 have
8421 linkers with such optimizations. AIX may have these optimizations in
8424 Only use these options when there are significant benefits from doing
8425 so. When you specify these options, the assembler and linker will
8426 create larger object and executable files and will also be slower.
8427 You will not be able to use @code{gprof} on all systems if you
8428 specify this option and you may have problems with debugging if
8429 you specify both this option and @option{-g}.
8431 @item -fbranch-target-load-optimize
8432 @opindex fbranch-target-load-optimize
8433 Perform branch target register load optimization before prologue / epilogue
8435 The use of target registers can typically be exposed only during reload,
8436 thus hoisting loads out of loops and doing inter-block scheduling needs
8437 a separate optimization pass.
8439 @item -fbranch-target-load-optimize2
8440 @opindex fbranch-target-load-optimize2
8441 Perform branch target register load optimization after prologue / epilogue
8444 @item -fbtr-bb-exclusive
8445 @opindex fbtr-bb-exclusive
8446 When performing branch target register load optimization, don't reuse
8447 branch target registers in within any basic block.
8449 @item -fstack-protector
8450 @opindex fstack-protector
8451 Emit extra code to check for buffer overflows, such as stack smashing
8452 attacks. This is done by adding a guard variable to functions with
8453 vulnerable objects. This includes functions that call alloca, and
8454 functions with buffers larger than 8 bytes. The guards are initialized
8455 when a function is entered and then checked when the function exits.
8456 If a guard check fails, an error message is printed and the program exits.
8458 @item -fstack-protector-all
8459 @opindex fstack-protector-all
8460 Like @option{-fstack-protector} except that all functions are protected.
8462 @item -fsection-anchors
8463 @opindex fsection-anchors
8464 Try to reduce the number of symbolic address calculations by using
8465 shared ``anchor'' symbols to address nearby objects. This transformation
8466 can help to reduce the number of GOT entries and GOT accesses on some
8469 For example, the implementation of the following function @code{foo}:
8473 int foo (void) @{ return a + b + c; @}
8476 would usually calculate the addresses of all three variables, but if you
8477 compile it with @option{-fsection-anchors}, it will access the variables
8478 from a common anchor point instead. The effect is similar to the
8479 following pseudocode (which isn't valid C):
8484 register int *xr = &x;
8485 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8489 Not all targets support this option.
8491 @item --param @var{name}=@var{value}
8493 In some places, GCC uses various constants to control the amount of
8494 optimization that is done. For example, GCC will not inline functions
8495 that contain more that a certain number of instructions. You can
8496 control some of these constants on the command line using the
8497 @option{--param} option.
8499 The names of specific parameters, and the meaning of the values, are
8500 tied to the internals of the compiler, and are subject to change
8501 without notice in future releases.
8503 In each case, the @var{value} is an integer. The allowable choices for
8504 @var{name} are given in the following table:
8507 @item predictable-branch-outcome
8508 When branch is predicted to be taken with probability lower than this threshold
8509 (in percent), then it is considered well predictable. The default is 10.
8511 @item max-crossjump-edges
8512 The maximum number of incoming edges to consider for crossjumping.
8513 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8514 the number of edges incoming to each block. Increasing values mean
8515 more aggressive optimization, making the compilation time increase with
8516 probably small improvement in executable size.
8518 @item min-crossjump-insns
8519 The minimum number of instructions which must be matched at the end
8520 of two blocks before crossjumping will be performed on them. This
8521 value is ignored in the case where all instructions in the block being
8522 crossjumped from are matched. The default value is 5.
8524 @item max-grow-copy-bb-insns
8525 The maximum code size expansion factor when copying basic blocks
8526 instead of jumping. The expansion is relative to a jump instruction.
8527 The default value is 8.
8529 @item max-goto-duplication-insns
8530 The maximum number of instructions to duplicate to a block that jumps
8531 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8532 passes, GCC factors computed gotos early in the compilation process,
8533 and unfactors them as late as possible. Only computed jumps at the
8534 end of a basic blocks with no more than max-goto-duplication-insns are
8535 unfactored. The default value is 8.
8537 @item max-delay-slot-insn-search
8538 The maximum number of instructions to consider when looking for an
8539 instruction to fill a delay slot. If more than this arbitrary number of
8540 instructions is searched, the time savings from filling the delay slot
8541 will be minimal so stop searching. Increasing values mean more
8542 aggressive optimization, making the compilation time increase with probably
8543 small improvement in execution time.
8545 @item max-delay-slot-live-search
8546 When trying to fill delay slots, the maximum number of instructions to
8547 consider when searching for a block with valid live register
8548 information. Increasing this arbitrarily chosen value means more
8549 aggressive optimization, increasing the compilation time. This parameter
8550 should be removed when the delay slot code is rewritten to maintain the
8553 @item max-gcse-memory
8554 The approximate maximum amount of memory that will be allocated in
8555 order to perform the global common subexpression elimination
8556 optimization. If more memory than specified is required, the
8557 optimization will not be done.
8559 @item max-gcse-insertion-ratio
8560 If the ratio of expression insertions to deletions is larger than this value
8561 for any expression, then RTL PRE will insert or remove the expression and thus
8562 leave partially redundant computations in the instruction stream. The default value is 20.
8564 @item max-pending-list-length
8565 The maximum number of pending dependencies scheduling will allow
8566 before flushing the current state and starting over. Large functions
8567 with few branches or calls can create excessively large lists which
8568 needlessly consume memory and resources.
8570 @item max-modulo-backtrack-attempts
8571 The maximum number of backtrack attempts the scheduler should make
8572 when modulo scheduling a loop. Larger values can exponentially increase
8575 @item max-inline-insns-single
8576 Several parameters control the tree inliner used in gcc.
8577 This number sets the maximum number of instructions (counted in GCC's
8578 internal representation) in a single function that the tree inliner
8579 will consider for inlining. This only affects functions declared
8580 inline and methods implemented in a class declaration (C++).
8581 The default value is 400.
8583 @item max-inline-insns-auto
8584 When you use @option{-finline-functions} (included in @option{-O3}),
8585 a lot of functions that would otherwise not be considered for inlining
8586 by the compiler will be investigated. To those functions, a different
8587 (more restrictive) limit compared to functions declared inline can
8589 The default value is 40.
8591 @item large-function-insns
8592 The limit specifying really large functions. For functions larger than this
8593 limit after inlining, inlining is constrained by
8594 @option{--param large-function-growth}. This parameter is useful primarily
8595 to avoid extreme compilation time caused by non-linear algorithms used by the
8597 The default value is 2700.
8599 @item large-function-growth
8600 Specifies maximal growth of large function caused by inlining in percents.
8601 The default value is 100 which limits large function growth to 2.0 times
8604 @item large-unit-insns
8605 The limit specifying large translation unit. Growth caused by inlining of
8606 units larger than this limit is limited by @option{--param inline-unit-growth}.
8607 For small units this might be too tight (consider unit consisting of function A
8608 that is inline and B that just calls A three time. If B is small relative to
8609 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8610 large units consisting of small inlineable functions however the overall unit
8611 growth limit is needed to avoid exponential explosion of code size. Thus for
8612 smaller units, the size is increased to @option{--param large-unit-insns}
8613 before applying @option{--param inline-unit-growth}. The default is 10000
8615 @item inline-unit-growth
8616 Specifies maximal overall growth of the compilation unit caused by inlining.
8617 The default value is 30 which limits unit growth to 1.3 times the original
8620 @item ipcp-unit-growth
8621 Specifies maximal overall growth of the compilation unit caused by
8622 interprocedural constant propagation. The default value is 10 which limits
8623 unit growth to 1.1 times the original size.
8625 @item large-stack-frame
8626 The limit specifying large stack frames. While inlining the algorithm is trying
8627 to not grow past this limit too much. Default value is 256 bytes.
8629 @item large-stack-frame-growth
8630 Specifies maximal growth of large stack frames caused by inlining in percents.
8631 The default value is 1000 which limits large stack frame growth to 11 times
8634 @item max-inline-insns-recursive
8635 @itemx max-inline-insns-recursive-auto
8636 Specifies maximum number of instructions out-of-line copy of self recursive inline
8637 function can grow into by performing recursive inlining.
8639 For functions declared inline @option{--param max-inline-insns-recursive} is
8640 taken into account. For function not declared inline, recursive inlining
8641 happens only when @option{-finline-functions} (included in @option{-O3}) is
8642 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8643 default value is 450.
8645 @item max-inline-recursive-depth
8646 @itemx max-inline-recursive-depth-auto
8647 Specifies maximum recursion depth used by the recursive inlining.
8649 For functions declared inline @option{--param max-inline-recursive-depth} is
8650 taken into account. For function not declared inline, recursive inlining
8651 happens only when @option{-finline-functions} (included in @option{-O3}) is
8652 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8655 @item min-inline-recursive-probability
8656 Recursive inlining is profitable only for function having deep recursion
8657 in average and can hurt for function having little recursion depth by
8658 increasing the prologue size or complexity of function body to other
8661 When profile feedback is available (see @option{-fprofile-generate}) the actual
8662 recursion depth can be guessed from probability that function will recurse via
8663 given call expression. This parameter limits inlining only to call expression
8664 whose probability exceeds given threshold (in percents). The default value is
8667 @item early-inlining-insns
8668 Specify growth that early inliner can make. In effect it increases amount of
8669 inlining for code having large abstraction penalty. The default value is 10.
8671 @item max-early-inliner-iterations
8672 @itemx max-early-inliner-iterations
8673 Limit of iterations of early inliner. This basically bounds number of nested
8674 indirect calls early inliner can resolve. Deeper chains are still handled by
8677 @item comdat-sharing-probability
8678 @itemx comdat-sharing-probability
8679 Probability (in percent) that C++ inline function with comdat visibility
8680 will be shared across multiple compilation units. The default value is 20.
8682 @item min-vect-loop-bound
8683 The minimum number of iterations under which a loop will not get vectorized
8684 when @option{-ftree-vectorize} is used. The number of iterations after
8685 vectorization needs to be greater than the value specified by this option
8686 to allow vectorization. The default value is 0.
8688 @item gcse-cost-distance-ratio
8689 Scaling factor in calculation of maximum distance an expression
8690 can be moved by GCSE optimizations. This is currently supported only in the
8691 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8692 will be with simple expressions, i.e., the expressions which have cost
8693 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8694 hoisting of simple expressions. The default value is 10.
8696 @item gcse-unrestricted-cost
8697 Cost, roughly measured as the cost of a single typical machine
8698 instruction, at which GCSE optimizations will not constrain
8699 the distance an expression can travel. This is currently
8700 supported only in the code hoisting pass. The lesser the cost,
8701 the more aggressive code hoisting will be. Specifying 0 will
8702 allow all expressions to travel unrestricted distances.
8703 The default value is 3.
8705 @item max-hoist-depth
8706 The depth of search in the dominator tree for expressions to hoist.
8707 This is used to avoid quadratic behavior in hoisting algorithm.
8708 The value of 0 will avoid limiting the search, but may slow down compilation
8709 of huge functions. The default value is 30.
8711 @item max-tail-merge-comparisons
8712 The maximum amount of similar bbs to compare a bb with. This is used to
8713 avoid quadratic behaviour in tree tail merging. The default value is 10.
8715 @item max-tail-merge-iterations
8716 The maximum amount of iterations of the pass over the function. This is used to
8717 limit compilation time in tree tail merging. The default value is 2.
8719 @item max-unrolled-insns
8720 The maximum number of instructions that a loop should have if that loop
8721 is unrolled, and if the loop is unrolled, it determines how many times
8722 the loop code is unrolled.
8724 @item max-average-unrolled-insns
8725 The maximum number of instructions biased by probabilities of their execution
8726 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8727 it determines how many times the loop code is unrolled.
8729 @item max-unroll-times
8730 The maximum number of unrollings of a single loop.
8732 @item max-peeled-insns
8733 The maximum number of instructions that a loop should have if that loop
8734 is peeled, and if the loop is peeled, it determines how many times
8735 the loop code is peeled.
8737 @item max-peel-times
8738 The maximum number of peelings of a single loop.
8740 @item max-completely-peeled-insns
8741 The maximum number of insns of a completely peeled loop.
8743 @item max-completely-peel-times
8744 The maximum number of iterations of a loop to be suitable for complete peeling.
8746 @item max-completely-peel-loop-nest-depth
8747 The maximum depth of a loop nest suitable for complete peeling.
8749 @item max-unswitch-insns
8750 The maximum number of insns of an unswitched loop.
8752 @item max-unswitch-level
8753 The maximum number of branches unswitched in a single loop.
8756 The minimum cost of an expensive expression in the loop invariant motion.
8758 @item iv-consider-all-candidates-bound
8759 Bound on number of candidates for induction variables below that
8760 all candidates are considered for each use in induction variable
8761 optimizations. Only the most relevant candidates are considered
8762 if there are more candidates, to avoid quadratic time complexity.
8764 @item iv-max-considered-uses
8765 The induction variable optimizations give up on loops that contain more
8766 induction variable uses.
8768 @item iv-always-prune-cand-set-bound
8769 If number of candidates in the set is smaller than this value,
8770 we always try to remove unnecessary ivs from the set during its
8771 optimization when a new iv is added to the set.
8773 @item scev-max-expr-size
8774 Bound on size of expressions used in the scalar evolutions analyzer.
8775 Large expressions slow the analyzer.
8777 @item scev-max-expr-complexity
8778 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8779 Complex expressions slow the analyzer.
8781 @item omega-max-vars
8782 The maximum number of variables in an Omega constraint system.
8783 The default value is 128.
8785 @item omega-max-geqs
8786 The maximum number of inequalities in an Omega constraint system.
8787 The default value is 256.
8790 The maximum number of equalities in an Omega constraint system.
8791 The default value is 128.
8793 @item omega-max-wild-cards
8794 The maximum number of wildcard variables that the Omega solver will
8795 be able to insert. The default value is 18.
8797 @item omega-hash-table-size
8798 The size of the hash table in the Omega solver. The default value is
8801 @item omega-max-keys
8802 The maximal number of keys used by the Omega solver. The default
8805 @item omega-eliminate-redundant-constraints
8806 When set to 1, use expensive methods to eliminate all redundant
8807 constraints. The default value is 0.
8809 @item vect-max-version-for-alignment-checks
8810 The maximum number of run-time checks that can be performed when
8811 doing loop versioning for alignment in the vectorizer. See option
8812 ftree-vect-loop-version for more information.
8814 @item vect-max-version-for-alias-checks
8815 The maximum number of run-time checks that can be performed when
8816 doing loop versioning for alias in the vectorizer. See option
8817 ftree-vect-loop-version for more information.
8819 @item max-iterations-to-track
8821 The maximum number of iterations of a loop the brute force algorithm
8822 for analysis of # of iterations of the loop tries to evaluate.
8824 @item hot-bb-count-fraction
8825 Select fraction of the maximal count of repetitions of basic block in program
8826 given basic block needs to have to be considered hot.
8828 @item hot-bb-frequency-fraction
8829 Select fraction of the entry block frequency of executions of basic block in
8830 function given basic block needs to have to be considered hot.
8832 @item max-predicted-iterations
8833 The maximum number of loop iterations we predict statically. This is useful
8834 in cases where function contain single loop with known bound and other loop
8835 with unknown. We predict the known number of iterations correctly, while
8836 the unknown number of iterations average to roughly 10. This means that the
8837 loop without bounds would appear artificially cold relative to the other one.
8839 @item align-threshold
8841 Select fraction of the maximal frequency of executions of basic block in
8842 function given basic block will get aligned.
8844 @item align-loop-iterations
8846 A loop expected to iterate at lest the selected number of iterations will get
8849 @item tracer-dynamic-coverage
8850 @itemx tracer-dynamic-coverage-feedback
8852 This value is used to limit superblock formation once the given percentage of
8853 executed instructions is covered. This limits unnecessary code size
8856 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8857 feedback is available. The real profiles (as opposed to statically estimated
8858 ones) are much less balanced allowing the threshold to be larger value.
8860 @item tracer-max-code-growth
8861 Stop tail duplication once code growth has reached given percentage. This is
8862 rather hokey argument, as most of the duplicates will be eliminated later in
8863 cross jumping, so it may be set to much higher values than is the desired code
8866 @item tracer-min-branch-ratio
8868 Stop reverse growth when the reverse probability of best edge is less than this
8869 threshold (in percent).
8871 @item tracer-min-branch-ratio
8872 @itemx tracer-min-branch-ratio-feedback
8874 Stop forward growth if the best edge do have probability lower than this
8877 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8878 compilation for profile feedback and one for compilation without. The value
8879 for compilation with profile feedback needs to be more conservative (higher) in
8880 order to make tracer effective.
8882 @item max-cse-path-length
8884 Maximum number of basic blocks on path that cse considers. The default is 10.
8887 The maximum instructions CSE process before flushing. The default is 1000.
8889 @item ggc-min-expand
8891 GCC uses a garbage collector to manage its own memory allocation. This
8892 parameter specifies the minimum percentage by which the garbage
8893 collector's heap should be allowed to expand between collections.
8894 Tuning this may improve compilation speed; it has no effect on code
8897 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8898 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8899 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8900 GCC is not able to calculate RAM on a particular platform, the lower
8901 bound of 30% is used. Setting this parameter and
8902 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8903 every opportunity. This is extremely slow, but can be useful for
8906 @item ggc-min-heapsize
8908 Minimum size of the garbage collector's heap before it begins bothering
8909 to collect garbage. The first collection occurs after the heap expands
8910 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8911 tuning this may improve compilation speed, and has no effect on code
8914 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8915 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8916 with a lower bound of 4096 (four megabytes) and an upper bound of
8917 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8918 particular platform, the lower bound is used. Setting this parameter
8919 very large effectively disables garbage collection. Setting this
8920 parameter and @option{ggc-min-expand} to zero causes a full collection
8921 to occur at every opportunity.
8923 @item max-reload-search-insns
8924 The maximum number of instruction reload should look backward for equivalent
8925 register. Increasing values mean more aggressive optimization, making the
8926 compilation time increase with probably slightly better performance.
8927 The default value is 100.
8929 @item max-cselib-memory-locations
8930 The maximum number of memory locations cselib should take into account.
8931 Increasing values mean more aggressive optimization, making the compilation time
8932 increase with probably slightly better performance. The default value is 500.
8934 @item reorder-blocks-duplicate
8935 @itemx reorder-blocks-duplicate-feedback
8937 Used by basic block reordering pass to decide whether to use unconditional
8938 branch or duplicate the code on its destination. Code is duplicated when its
8939 estimated size is smaller than this value multiplied by the estimated size of
8940 unconditional jump in the hot spots of the program.
8942 The @option{reorder-block-duplicate-feedback} is used only when profile
8943 feedback is available and may be set to higher values than
8944 @option{reorder-block-duplicate} since information about the hot spots is more
8947 @item max-sched-ready-insns
8948 The maximum number of instructions ready to be issued the scheduler should
8949 consider at any given time during the first scheduling pass. Increasing
8950 values mean more thorough searches, making the compilation time increase
8951 with probably little benefit. The default value is 100.
8953 @item max-sched-region-blocks
8954 The maximum number of blocks in a region to be considered for
8955 interblock scheduling. The default value is 10.
8957 @item max-pipeline-region-blocks
8958 The maximum number of blocks in a region to be considered for
8959 pipelining in the selective scheduler. The default value is 15.
8961 @item max-sched-region-insns
8962 The maximum number of insns in a region to be considered for
8963 interblock scheduling. The default value is 100.
8965 @item max-pipeline-region-insns
8966 The maximum number of insns in a region to be considered for
8967 pipelining in the selective scheduler. The default value is 200.
8970 The minimum probability (in percents) of reaching a source block
8971 for interblock speculative scheduling. The default value is 40.
8973 @item max-sched-extend-regions-iters
8974 The maximum number of iterations through CFG to extend regions.
8975 0 - disable region extension,
8976 N - do at most N iterations.
8977 The default value is 0.
8979 @item max-sched-insn-conflict-delay
8980 The maximum conflict delay for an insn to be considered for speculative motion.
8981 The default value is 3.
8983 @item sched-spec-prob-cutoff
8984 The minimal probability of speculation success (in percents), so that
8985 speculative insn will be scheduled.
8986 The default value is 40.
8988 @item sched-mem-true-dep-cost
8989 Minimal distance (in CPU cycles) between store and load targeting same
8990 memory locations. The default value is 1.
8992 @item selsched-max-lookahead
8993 The maximum size of the lookahead window of selective scheduling. It is a
8994 depth of search for available instructions.
8995 The default value is 50.
8997 @item selsched-max-sched-times
8998 The maximum number of times that an instruction will be scheduled during
8999 selective scheduling. This is the limit on the number of iterations
9000 through which the instruction may be pipelined. The default value is 2.
9002 @item selsched-max-insns-to-rename
9003 The maximum number of best instructions in the ready list that are considered
9004 for renaming in the selective scheduler. The default value is 2.
9007 The minimum value of stage count that swing modulo scheduler will
9008 generate. The default value is 2.
9010 @item max-last-value-rtl
9011 The maximum size measured as number of RTLs that can be recorded in an expression
9012 in combiner for a pseudo register as last known value of that register. The default
9015 @item integer-share-limit
9016 Small integer constants can use a shared data structure, reducing the
9017 compiler's memory usage and increasing its speed. This sets the maximum
9018 value of a shared integer constant. The default value is 256.
9020 @item min-virtual-mappings
9021 Specifies the minimum number of virtual mappings in the incremental
9022 SSA updater that should be registered to trigger the virtual mappings
9023 heuristic defined by virtual-mappings-ratio. The default value is
9026 @item virtual-mappings-ratio
9027 If the number of virtual mappings is virtual-mappings-ratio bigger
9028 than the number of virtual symbols to be updated, then the incremental
9029 SSA updater switches to a full update for those symbols. The default
9032 @item ssp-buffer-size
9033 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
9034 protection when @option{-fstack-protection} is used.
9036 @item max-jump-thread-duplication-stmts
9037 Maximum number of statements allowed in a block that needs to be
9038 duplicated when threading jumps.
9040 @item max-fields-for-field-sensitive
9041 Maximum number of fields in a structure we will treat in
9042 a field sensitive manner during pointer analysis. The default is zero
9043 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
9045 @item prefetch-latency
9046 Estimate on average number of instructions that are executed before
9047 prefetch finishes. The distance we prefetch ahead is proportional
9048 to this constant. Increasing this number may also lead to less
9049 streams being prefetched (see @option{simultaneous-prefetches}).
9051 @item simultaneous-prefetches
9052 Maximum number of prefetches that can run at the same time.
9054 @item l1-cache-line-size
9055 The size of cache line in L1 cache, in bytes.
9058 The size of L1 cache, in kilobytes.
9061 The size of L2 cache, in kilobytes.
9063 @item min-insn-to-prefetch-ratio
9064 The minimum ratio between the number of instructions and the
9065 number of prefetches to enable prefetching in a loop.
9067 @item prefetch-min-insn-to-mem-ratio
9068 The minimum ratio between the number of instructions and the
9069 number of memory references to enable prefetching in a loop.
9071 @item use-canonical-types
9072 Whether the compiler should use the ``canonical'' type system. By
9073 default, this should always be 1, which uses a more efficient internal
9074 mechanism for comparing types in C++ and Objective-C++. However, if
9075 bugs in the canonical type system are causing compilation failures,
9076 set this value to 0 to disable canonical types.
9078 @item switch-conversion-max-branch-ratio
9079 Switch initialization conversion will refuse to create arrays that are
9080 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9081 branches in the switch.
9083 @item max-partial-antic-length
9084 Maximum length of the partial antic set computed during the tree
9085 partial redundancy elimination optimization (@option{-ftree-pre}) when
9086 optimizing at @option{-O3} and above. For some sorts of source code
9087 the enhanced partial redundancy elimination optimization can run away,
9088 consuming all of the memory available on the host machine. This
9089 parameter sets a limit on the length of the sets that are computed,
9090 which prevents the runaway behavior. Setting a value of 0 for
9091 this parameter will allow an unlimited set length.
9093 @item sccvn-max-scc-size
9094 Maximum size of a strongly connected component (SCC) during SCCVN
9095 processing. If this limit is hit, SCCVN processing for the whole
9096 function will not be done and optimizations depending on it will
9097 be disabled. The default maximum SCC size is 10000.
9099 @item ira-max-loops-num
9100 IRA uses a regional register allocation by default. If a function
9101 contains loops more than number given by the parameter, only at most
9102 given number of the most frequently executed loops will form regions
9103 for the regional register allocation. The default value of the
9106 @item ira-max-conflict-table-size
9107 Although IRA uses a sophisticated algorithm of compression conflict
9108 table, the table can be still big for huge functions. If the conflict
9109 table for a function could be more than size in MB given by the
9110 parameter, the conflict table is not built and faster, simpler, and
9111 lower quality register allocation algorithm will be used. The
9112 algorithm do not use pseudo-register conflicts. The default value of
9113 the parameter is 2000.
9115 @item ira-loop-reserved-regs
9116 IRA can be used to evaluate more accurate register pressure in loops
9117 for decision to move loop invariants (see @option{-O3}). The number
9118 of available registers reserved for some other purposes is described
9119 by this parameter. The default value of the parameter is 2 which is
9120 minimal number of registers needed for execution of typical
9121 instruction. This value is the best found from numerous experiments.
9123 @item loop-invariant-max-bbs-in-loop
9124 Loop invariant motion can be very expensive, both in compilation time and
9125 in amount of needed compile-time memory, with very large loops. Loops
9126 with more basic blocks than this parameter won't have loop invariant
9127 motion optimization performed on them. The default value of the
9128 parameter is 1000 for -O1 and 10000 for -O2 and above.
9130 @item loop-max-datarefs-for-datadeps
9131 Building data dapendencies is expensive for very large loops. This
9132 parameter limits the number of data references in loops that are
9133 considered for data dependence analysis. These large loops will not
9134 be handled then by the optimizations using loop data dependencies.
9135 The default value is 1000.
9137 @item max-vartrack-size
9138 Sets a maximum number of hash table slots to use during variable
9139 tracking dataflow analysis of any function. If this limit is exceeded
9140 with variable tracking at assignments enabled, analysis for that
9141 function is retried without it, after removing all debug insns from
9142 the function. If the limit is exceeded even without debug insns, var
9143 tracking analysis is completely disabled for the function. Setting
9144 the parameter to zero makes it unlimited.
9146 @item max-vartrack-expr-depth
9147 Sets a maximum number of recursion levels when attempting to map
9148 variable names or debug temporaries to value expressions. This trades
9149 compilation time for more complete debug information. If this is set too
9150 low, value expressions that are available and could be represented in
9151 debug information may end up not being used; setting this higher may
9152 enable the compiler to find more complex debug expressions, but compile
9153 time and memory use may grow. The default is 12.
9155 @item min-nondebug-insn-uid
9156 Use uids starting at this parameter for nondebug insns. The range below
9157 the parameter is reserved exclusively for debug insns created by
9158 @option{-fvar-tracking-assignments}, but debug insns may get
9159 (non-overlapping) uids above it if the reserved range is exhausted.
9161 @item ipa-sra-ptr-growth-factor
9162 IPA-SRA will replace a pointer to an aggregate with one or more new
9163 parameters only when their cumulative size is less or equal to
9164 @option{ipa-sra-ptr-growth-factor} times the size of the original
9167 @item tm-max-aggregate-size
9168 When making copies of thread-local variables in a transaction, this
9169 parameter specifies the size in bytes after which variables will be
9170 saved with the logging functions as opposed to save/restore code
9171 sequence pairs. This option only applies when using
9174 @item graphite-max-nb-scop-params
9175 To avoid exponential effects in the Graphite loop transforms, the
9176 number of parameters in a Static Control Part (SCoP) is bounded. The
9177 default value is 10 parameters. A variable whose value is unknown at
9178 compilation time and defined outside a SCoP is a parameter of the SCoP.
9180 @item graphite-max-bbs-per-function
9181 To avoid exponential effects in the detection of SCoPs, the size of
9182 the functions analyzed by Graphite is bounded. The default value is
9185 @item loop-block-tile-size
9186 Loop blocking or strip mining transforms, enabled with
9187 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9188 loop in the loop nest by a given number of iterations. The strip
9189 length can be changed using the @option{loop-block-tile-size}
9190 parameter. The default value is 51 iterations.
9192 @item ipa-cp-value-list-size
9193 IPA-CP attempts to track all possible values and types passed to a function's
9194 parameter in order to propagate them and perform devirtualization.
9195 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9196 stores per one formal parameter of a function.
9198 @item lto-partitions
9199 Specify desired number of partitions produced during WHOPR compilation.
9200 The number of partitions should exceed the number of CPUs used for compilation.
9201 The default value is 32.
9203 @item lto-minpartition
9204 Size of minimal partition for WHOPR (in estimated instructions).
9205 This prevents expenses of splitting very small programs into too many
9208 @item cxx-max-namespaces-for-diagnostic-help
9209 The maximum number of namespaces to consult for suggestions when C++
9210 name lookup fails for an identifier. The default is 1000.
9212 @item sink-frequency-threshold
9213 The maximum relative execution frequency (in percents) of the target block
9214 relative to a statement's original block to allow statement sinking of a
9215 statement. Larger numbers result in more aggressive statement sinking.
9216 The default value is 75. A small positive adjustment is applied for
9217 statements with memory operands as those are even more profitable so sink.
9219 @item max-stores-to-sink
9220 The maximum number of conditional stores paires that can be sunk. Set to 0
9221 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9222 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9224 @item allow-load-data-races
9225 Allow optimizers to introduce new data races on loads.
9226 Set to 1 to allow, otherwise to 0. This option is enabled by default
9227 unless implicitly set by the @option{-fmemory-model=} option.
9229 @item allow-store-data-races
9230 Allow optimizers to introduce new data races on stores.
9231 Set to 1 to allow, otherwise to 0. This option is enabled by default
9232 unless implicitly set by the @option{-fmemory-model=} option.
9234 @item allow-packed-load-data-races
9235 Allow optimizers to introduce new data races on packed data loads.
9236 Set to 1 to allow, otherwise to 0. This option is enabled by default
9237 unless implicitly set by the @option{-fmemory-model=} option.
9239 @item allow-packed-store-data-races
9240 Allow optimizers to introduce new data races on packed data stores.
9241 Set to 1 to allow, otherwise to 0. This option is enabled by default
9242 unless implicitly set by the @option{-fmemory-model=} option.
9244 @item case-values-threshold
9245 The smallest number of different values for which it is best to use a
9246 jump-table instead of a tree of conditional branches. If the value is
9247 0, use the default for the machine. The default is 0.
9249 @item tree-reassoc-width
9250 Set the maximum number of instructions executed in parallel in
9251 reassociated tree. This parameter overrides target dependent
9252 heuristics used by default if has non zero value.
9257 @node Preprocessor Options
9258 @section Options Controlling the Preprocessor
9259 @cindex preprocessor options
9260 @cindex options, preprocessor
9262 These options control the C preprocessor, which is run on each C source
9263 file before actual compilation.
9265 If you use the @option{-E} option, nothing is done except preprocessing.
9266 Some of these options make sense only together with @option{-E} because
9267 they cause the preprocessor output to be unsuitable for actual
9271 @item -Wp,@var{option}
9273 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9274 and pass @var{option} directly through to the preprocessor. If
9275 @var{option} contains commas, it is split into multiple options at the
9276 commas. However, many options are modified, translated or interpreted
9277 by the compiler driver before being passed to the preprocessor, and
9278 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9279 interface is undocumented and subject to change, so whenever possible
9280 you should avoid using @option{-Wp} and let the driver handle the
9283 @item -Xpreprocessor @var{option}
9284 @opindex Xpreprocessor
9285 Pass @var{option} as an option to the preprocessor. You can use this to
9286 supply system-specific preprocessor options which GCC does not know how to
9289 If you want to pass an option that takes an argument, you must use
9290 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9293 @include cppopts.texi
9295 @node Assembler Options
9296 @section Passing Options to the Assembler
9298 @c prevent bad page break with this line
9299 You can pass options to the assembler.
9302 @item -Wa,@var{option}
9304 Pass @var{option} as an option to the assembler. If @var{option}
9305 contains commas, it is split into multiple options at the commas.
9307 @item -Xassembler @var{option}
9309 Pass @var{option} as an option to the assembler. You can use this to
9310 supply system-specific assembler options which GCC does not know how to
9313 If you want to pass an option that takes an argument, you must use
9314 @option{-Xassembler} twice, once for the option and once for the argument.
9319 @section Options for Linking
9320 @cindex link options
9321 @cindex options, linking
9323 These options come into play when the compiler links object files into
9324 an executable output file. They are meaningless if the compiler is
9325 not doing a link step.
9329 @item @var{object-file-name}
9330 A file name that does not end in a special recognized suffix is
9331 considered to name an object file or library. (Object files are
9332 distinguished from libraries by the linker according to the file
9333 contents.) If linking is done, these object files are used as input
9342 If any of these options is used, then the linker is not run, and
9343 object file names should not be used as arguments. @xref{Overall
9347 @item -l@var{library}
9348 @itemx -l @var{library}
9350 Search the library named @var{library} when linking. (The second
9351 alternative with the library as a separate argument is only for
9352 POSIX compliance and is not recommended.)
9354 It makes a difference where in the command you write this option; the
9355 linker searches and processes libraries and object files in the order they
9356 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9357 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9358 to functions in @samp{z}, those functions may not be loaded.
9360 The linker searches a standard list of directories for the library,
9361 which is actually a file named @file{lib@var{library}.a}. The linker
9362 then uses this file as if it had been specified precisely by name.
9364 The directories searched include several standard system directories
9365 plus any that you specify with @option{-L}.
9367 Normally the files found this way are library files---archive files
9368 whose members are object files. The linker handles an archive file by
9369 scanning through it for members which define symbols that have so far
9370 been referenced but not defined. But if the file that is found is an
9371 ordinary object file, it is linked in the usual fashion. The only
9372 difference between using an @option{-l} option and specifying a file name
9373 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9374 and searches several directories.
9378 You need this special case of the @option{-l} option in order to
9379 link an Objective-C or Objective-C++ program.
9382 @opindex nostartfiles
9383 Do not use the standard system startup files when linking.
9384 The standard system libraries are used normally, unless @option{-nostdlib}
9385 or @option{-nodefaultlibs} is used.
9387 @item -nodefaultlibs
9388 @opindex nodefaultlibs
9389 Do not use the standard system libraries when linking.
9390 Only the libraries you specify will be passed to the linker, options
9391 specifying linkage of the system libraries, such as @code{-static-libgcc}
9392 or @code{-shared-libgcc}, will be ignored.
9393 The standard startup files are used normally, unless @option{-nostartfiles}
9394 is used. The compiler may generate calls to @code{memcmp},
9395 @code{memset}, @code{memcpy} and @code{memmove}.
9396 These entries are usually resolved by entries in
9397 libc. These entry points should be supplied through some other
9398 mechanism when this option is specified.
9402 Do not use the standard system startup files or libraries when linking.
9403 No startup files and only the libraries you specify will be passed to
9404 the linker, options specifying linkage of the system libraries, such as
9405 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9406 The compiler may generate calls to @code{memcmp}, @code{memset},
9407 @code{memcpy} and @code{memmove}.
9408 These entries are usually resolved by entries in
9409 libc. These entry points should be supplied through some other
9410 mechanism when this option is specified.
9412 @cindex @option{-lgcc}, use with @option{-nostdlib}
9413 @cindex @option{-nostdlib} and unresolved references
9414 @cindex unresolved references and @option{-nostdlib}
9415 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9416 @cindex @option{-nodefaultlibs} and unresolved references
9417 @cindex unresolved references and @option{-nodefaultlibs}
9418 One of the standard libraries bypassed by @option{-nostdlib} and
9419 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9420 that GCC uses to overcome shortcomings of particular machines, or special
9421 needs for some languages.
9422 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9423 Collection (GCC) Internals},
9424 for more discussion of @file{libgcc.a}.)
9425 In most cases, you need @file{libgcc.a} even when you want to avoid
9426 other standard libraries. In other words, when you specify @option{-nostdlib}
9427 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9428 This ensures that you have no unresolved references to internal GCC
9429 library subroutines. (For example, @samp{__main}, used to ensure C++
9430 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9431 GNU Compiler Collection (GCC) Internals}.)
9435 Produce a position independent executable on targets which support it.
9436 For predictable results, you must also specify the same set of options
9437 that were used to generate code (@option{-fpie}, @option{-fPIE},
9438 or model suboptions) when you specify this option.
9442 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9443 that support it. This instructs the linker to add all symbols, not
9444 only used ones, to the dynamic symbol table. This option is needed
9445 for some uses of @code{dlopen} or to allow obtaining backtraces
9446 from within a program.
9450 Remove all symbol table and relocation information from the executable.
9454 On systems that support dynamic linking, this prevents linking with the shared
9455 libraries. On other systems, this option has no effect.
9459 Produce a shared object which can then be linked with other objects to
9460 form an executable. Not all systems support this option. For predictable
9461 results, you must also specify the same set of options that were used to
9462 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9463 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9464 needs to build supplementary stub code for constructors to work. On
9465 multi-libbed systems, @samp{gcc -shared} must select the correct support
9466 libraries to link against. Failing to supply the correct flags may lead
9467 to subtle defects. Supplying them in cases where they are not necessary
9470 @item -shared-libgcc
9471 @itemx -static-libgcc
9472 @opindex shared-libgcc
9473 @opindex static-libgcc
9474 On systems that provide @file{libgcc} as a shared library, these options
9475 force the use of either the shared or static version respectively.
9476 If no shared version of @file{libgcc} was built when the compiler was
9477 configured, these options have no effect.
9479 There are several situations in which an application should use the
9480 shared @file{libgcc} instead of the static version. The most common
9481 of these is when the application wishes to throw and catch exceptions
9482 across different shared libraries. In that case, each of the libraries
9483 as well as the application itself should use the shared @file{libgcc}.
9485 Therefore, the G++ and GCJ drivers automatically add
9486 @option{-shared-libgcc} whenever you build a shared library or a main
9487 executable, because C++ and Java programs typically use exceptions, so
9488 this is the right thing to do.
9490 If, instead, you use the GCC driver to create shared libraries, you may
9491 find that they will not always be linked with the shared @file{libgcc}.
9492 If GCC finds, at its configuration time, that you have a non-GNU linker
9493 or a GNU linker that does not support option @option{--eh-frame-hdr},
9494 it will link the shared version of @file{libgcc} into shared libraries
9495 by default. Otherwise, it will take advantage of the linker and optimize
9496 away the linking with the shared version of @file{libgcc}, linking with
9497 the static version of libgcc by default. This allows exceptions to
9498 propagate through such shared libraries, without incurring relocation
9499 costs at library load time.
9501 However, if a library or main executable is supposed to throw or catch
9502 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9503 for the languages used in the program, or using the option
9504 @option{-shared-libgcc}, such that it is linked with the shared
9507 @item -static-libstdc++
9508 When the @command{g++} program is used to link a C++ program, it will
9509 normally automatically link against @option{libstdc++}. If
9510 @file{libstdc++} is available as a shared library, and the
9511 @option{-static} option is not used, then this will link against the
9512 shared version of @file{libstdc++}. That is normally fine. However, it
9513 is sometimes useful to freeze the version of @file{libstdc++} used by
9514 the program without going all the way to a fully static link. The
9515 @option{-static-libstdc++} option directs the @command{g++} driver to
9516 link @file{libstdc++} statically, without necessarily linking other
9517 libraries statically.
9521 Bind references to global symbols when building a shared object. Warn
9522 about any unresolved references (unless overridden by the link editor
9523 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9526 @item -T @var{script}
9528 @cindex linker script
9529 Use @var{script} as the linker script. This option is supported by most
9530 systems using the GNU linker. On some targets, such as bare-board
9531 targets without an operating system, the @option{-T} option may be required
9532 when linking to avoid references to undefined symbols.
9534 @item -Xlinker @var{option}
9536 Pass @var{option} as an option to the linker. You can use this to
9537 supply system-specific linker options which GCC does not know how to
9540 If you want to pass an option that takes a separate argument, you must use
9541 @option{-Xlinker} twice, once for the option and once for the argument.
9542 For example, to pass @option{-assert definitions}, you must write
9543 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9544 @option{-Xlinker "-assert definitions"}, because this passes the entire
9545 string as a single argument, which is not what the linker expects.
9547 When using the GNU linker, it is usually more convenient to pass
9548 arguments to linker options using the @option{@var{option}=@var{value}}
9549 syntax than as separate arguments. For example, you can specify
9550 @samp{-Xlinker -Map=output.map} rather than
9551 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9552 this syntax for command-line options.
9554 @item -Wl,@var{option}
9556 Pass @var{option} as an option to the linker. If @var{option} contains
9557 commas, it is split into multiple options at the commas. You can use this
9558 syntax to pass an argument to the option.
9559 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9560 linker. When using the GNU linker, you can also get the same effect with
9561 @samp{-Wl,-Map=output.map}.
9563 @item -u @var{symbol}
9565 Pretend the symbol @var{symbol} is undefined, to force linking of
9566 library modules to define it. You can use @option{-u} multiple times with
9567 different symbols to force loading of additional library modules.
9570 @node Directory Options
9571 @section Options for Directory Search
9572 @cindex directory options
9573 @cindex options, directory search
9576 These options specify directories to search for header files, for
9577 libraries and for parts of the compiler:
9582 Add the directory @var{dir} to the head of the list of directories to be
9583 searched for header files. This can be used to override a system header
9584 file, substituting your own version, since these directories are
9585 searched before the system header file directories. However, you should
9586 not use this option to add directories that contain vendor-supplied
9587 system header files (use @option{-isystem} for that). If you use more than
9588 one @option{-I} option, the directories are scanned in left-to-right
9589 order; the standard system directories come after.
9591 If a standard system include directory, or a directory specified with
9592 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9593 option will be ignored. The directory will still be searched but as a
9594 system directory at its normal position in the system include chain.
9595 This is to ensure that GCC's procedure to fix buggy system headers and
9596 the ordering for the include_next directive are not inadvertently changed.
9597 If you really need to change the search order for system directories,
9598 use the @option{-nostdinc} and/or @option{-isystem} options.
9600 @item -iplugindir=@var{dir}
9601 Set the directory to search for plugins which are passed
9602 by @option{-fplugin=@var{name}} instead of
9603 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9604 to be used by the user, but only passed by the driver.
9606 @item -iquote@var{dir}
9608 Add the directory @var{dir} to the head of the list of directories to
9609 be searched for header files only for the case of @samp{#include
9610 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9611 otherwise just like @option{-I}.
9615 Add directory @var{dir} to the list of directories to be searched
9618 @item -B@var{prefix}
9620 This option specifies where to find the executables, libraries,
9621 include files, and data files of the compiler itself.
9623 The compiler driver program runs one or more of the subprograms
9624 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9625 @var{prefix} as a prefix for each program it tries to run, both with and
9626 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9628 For each subprogram to be run, the compiler driver first tries the
9629 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9630 was not specified, the driver tries two standard prefixes, which are
9631 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9632 those results in a file name that is found, the unmodified program
9633 name is searched for using the directories specified in your
9634 @env{PATH} environment variable.
9636 The compiler will check to see if the path provided by the @option{-B}
9637 refers to a directory, and if necessary it will add a directory
9638 separator character at the end of the path.
9640 @option{-B} prefixes that effectively specify directory names also apply
9641 to libraries in the linker, because the compiler translates these
9642 options into @option{-L} options for the linker. They also apply to
9643 includes files in the preprocessor, because the compiler translates these
9644 options into @option{-isystem} options for the preprocessor. In this case,
9645 the compiler appends @samp{include} to the prefix.
9647 The runtime support file @file{libgcc.a} can also be searched for using
9648 the @option{-B} prefix, if needed. If it is not found there, the two
9649 standard prefixes above are tried, and that is all. The file is left
9650 out of the link if it is not found by those means.
9652 Another way to specify a prefix much like the @option{-B} prefix is to use
9653 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9656 As a special kludge, if the path provided by @option{-B} is
9657 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9658 9, then it will be replaced by @file{[dir/]include}. This is to help
9659 with boot-strapping the compiler.
9661 @item -specs=@var{file}
9663 Process @var{file} after the compiler reads in the standard @file{specs}
9664 file, in order to override the defaults that the @file{gcc} driver
9665 program uses when determining what switches to pass to @file{cc1},
9666 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9667 @option{-specs=@var{file}} can be specified on the command line, and they
9668 are processed in order, from left to right.
9670 @item --sysroot=@var{dir}
9672 Use @var{dir} as the logical root directory for headers and libraries.
9673 For example, if the compiler would normally search for headers in
9674 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9675 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9677 If you use both this option and the @option{-isysroot} option, then
9678 the @option{--sysroot} option will apply to libraries, but the
9679 @option{-isysroot} option will apply to header files.
9681 The GNU linker (beginning with version 2.16) has the necessary support
9682 for this option. If your linker does not support this option, the
9683 header file aspect of @option{--sysroot} will still work, but the
9684 library aspect will not.
9688 This option has been deprecated. Please use @option{-iquote} instead for
9689 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9690 Any directories you specify with @option{-I} options before the @option{-I-}
9691 option are searched only for the case of @samp{#include "@var{file}"};
9692 they are not searched for @samp{#include <@var{file}>}.
9694 If additional directories are specified with @option{-I} options after
9695 the @option{-I-}, these directories are searched for all @samp{#include}
9696 directives. (Ordinarily @emph{all} @option{-I} directories are used
9699 In addition, the @option{-I-} option inhibits the use of the current
9700 directory (where the current input file came from) as the first search
9701 directory for @samp{#include "@var{file}"}. There is no way to
9702 override this effect of @option{-I-}. With @option{-I.} you can specify
9703 searching the directory which was current when the compiler was
9704 invoked. That is not exactly the same as what the preprocessor does
9705 by default, but it is often satisfactory.
9707 @option{-I-} does not inhibit the use of the standard system directories
9708 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9715 @section Specifying subprocesses and the switches to pass to them
9718 @command{gcc} is a driver program. It performs its job by invoking a
9719 sequence of other programs to do the work of compiling, assembling and
9720 linking. GCC interprets its command-line parameters and uses these to
9721 deduce which programs it should invoke, and which command-line options
9722 it ought to place on their command lines. This behavior is controlled
9723 by @dfn{spec strings}. In most cases there is one spec string for each
9724 program that GCC can invoke, but a few programs have multiple spec
9725 strings to control their behavior. The spec strings built into GCC can
9726 be overridden by using the @option{-specs=} command-line switch to specify
9729 @dfn{Spec files} are plaintext files that are used to construct spec
9730 strings. They consist of a sequence of directives separated by blank
9731 lines. The type of directive is determined by the first non-whitespace
9732 character on the line and it can be one of the following:
9735 @item %@var{command}
9736 Issues a @var{command} to the spec file processor. The commands that can
9740 @item %include <@var{file}>
9741 @cindex @code{%include}
9742 Search for @var{file} and insert its text at the current point in the
9745 @item %include_noerr <@var{file}>
9746 @cindex @code{%include_noerr}
9747 Just like @samp{%include}, but do not generate an error message if the include
9748 file cannot be found.
9750 @item %rename @var{old_name} @var{new_name}
9751 @cindex @code{%rename}
9752 Rename the spec string @var{old_name} to @var{new_name}.
9756 @item *[@var{spec_name}]:
9757 This tells the compiler to create, override or delete the named spec
9758 string. All lines after this directive up to the next directive or
9759 blank line are considered to be the text for the spec string. If this
9760 results in an empty string then the spec will be deleted. (Or, if the
9761 spec did not exist, then nothing will happen.) Otherwise, if the spec
9762 does not currently exist a new spec will be created. If the spec does
9763 exist then its contents will be overridden by the text of this
9764 directive, unless the first character of that text is the @samp{+}
9765 character, in which case the text will be appended to the spec.
9767 @item [@var{suffix}]:
9768 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9769 and up to the next directive or blank line are considered to make up the
9770 spec string for the indicated suffix. When the compiler encounters an
9771 input file with the named suffix, it will processes the spec string in
9772 order to work out how to compile that file. For example:
9779 This says that any input file whose name ends in @samp{.ZZ} should be
9780 passed to the program @samp{z-compile}, which should be invoked with the
9781 command-line switch @option{-input} and with the result of performing the
9782 @samp{%i} substitution. (See below.)
9784 As an alternative to providing a spec string, the text that follows a
9785 suffix directive can be one of the following:
9788 @item @@@var{language}
9789 This says that the suffix is an alias for a known @var{language}. This is
9790 similar to using the @option{-x} command-line switch to GCC to specify a
9791 language explicitly. For example:
9798 Says that .ZZ files are, in fact, C++ source files.
9801 This causes an error messages saying:
9804 @var{name} compiler not installed on this system.
9808 GCC already has an extensive list of suffixes built into it.
9809 This directive will add an entry to the end of the list of suffixes, but
9810 since the list is searched from the end backwards, it is effectively
9811 possible to override earlier entries using this technique.
9815 GCC has the following spec strings built into it. Spec files can
9816 override these strings or create their own. Note that individual
9817 targets can also add their own spec strings to this list.
9820 asm Options to pass to the assembler
9821 asm_final Options to pass to the assembler post-processor
9822 cpp Options to pass to the C preprocessor
9823 cc1 Options to pass to the C compiler
9824 cc1plus Options to pass to the C++ compiler
9825 endfile Object files to include at the end of the link
9826 link Options to pass to the linker
9827 lib Libraries to include on the command line to the linker
9828 libgcc Decides which GCC support library to pass to the linker
9829 linker Sets the name of the linker
9830 predefines Defines to be passed to the C preprocessor
9831 signed_char Defines to pass to CPP to say whether @code{char} is signed
9833 startfile Object files to include at the start of the link
9836 Here is a small example of a spec file:
9842 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9845 This example renames the spec called @samp{lib} to @samp{old_lib} and
9846 then overrides the previous definition of @samp{lib} with a new one.
9847 The new definition adds in some extra command-line options before
9848 including the text of the old definition.
9850 @dfn{Spec strings} are a list of command-line options to be passed to their
9851 corresponding program. In addition, the spec strings can contain
9852 @samp{%}-prefixed sequences to substitute variable text or to
9853 conditionally insert text into the command line. Using these constructs
9854 it is possible to generate quite complex command lines.
9856 Here is a table of all defined @samp{%}-sequences for spec
9857 strings. Note that spaces are not generated automatically around the
9858 results of expanding these sequences. Therefore you can concatenate them
9859 together or combine them with constant text in a single argument.
9863 Substitute one @samp{%} into the program name or argument.
9866 Substitute the name of the input file being processed.
9869 Substitute the basename of the input file being processed.
9870 This is the substring up to (and not including) the last period
9871 and not including the directory.
9874 This is the same as @samp{%b}, but include the file suffix (text after
9878 Marks the argument containing or following the @samp{%d} as a
9879 temporary file name, so that that file will be deleted if GCC exits
9880 successfully. Unlike @samp{%g}, this contributes no text to the
9883 @item %g@var{suffix}
9884 Substitute a file name that has suffix @var{suffix} and is chosen
9885 once per compilation, and mark the argument in the same way as
9886 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9887 name is now chosen in a way that is hard to predict even when previously
9888 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9889 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9890 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9891 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9892 was simply substituted with a file name chosen once per compilation,
9893 without regard to any appended suffix (which was therefore treated
9894 just like ordinary text), making such attacks more likely to succeed.
9896 @item %u@var{suffix}
9897 Like @samp{%g}, but generates a new temporary file name even if
9898 @samp{%u@var{suffix}} was already seen.
9900 @item %U@var{suffix}
9901 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9902 new one if there is no such last file name. In the absence of any
9903 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9904 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9905 would involve the generation of two distinct file names, one
9906 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9907 simply substituted with a file name chosen for the previous @samp{%u},
9908 without regard to any appended suffix.
9910 @item %j@var{suffix}
9911 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9912 writable, and if save-temps is off; otherwise, substitute the name
9913 of a temporary file, just like @samp{%u}. This temporary file is not
9914 meant for communication between processes, but rather as a junk
9917 @item %|@var{suffix}
9918 @itemx %m@var{suffix}
9919 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9920 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9921 all. These are the two most common ways to instruct a program that it
9922 should read from standard input or write to standard output. If you
9923 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9924 construct: see for example @file{f/lang-specs.h}.
9926 @item %.@var{SUFFIX}
9927 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9928 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9929 terminated by the next space or %.
9932 Marks the argument containing or following the @samp{%w} as the
9933 designated output file of this compilation. This puts the argument
9934 into the sequence of arguments that @samp{%o} will substitute later.
9937 Substitutes the names of all the output files, with spaces
9938 automatically placed around them. You should write spaces
9939 around the @samp{%o} as well or the results are undefined.
9940 @samp{%o} is for use in the specs for running the linker.
9941 Input files whose names have no recognized suffix are not compiled
9942 at all, but they are included among the output files, so they will
9946 Substitutes the suffix for object files. Note that this is
9947 handled specially when it immediately follows @samp{%g, %u, or %U},
9948 because of the need for those to form complete file names. The
9949 handling is such that @samp{%O} is treated exactly as if it had already
9950 been substituted, except that @samp{%g, %u, and %U} do not currently
9951 support additional @var{suffix} characters following @samp{%O} as they would
9952 following, for example, @samp{.o}.
9955 Substitutes the standard macro predefinitions for the
9956 current target machine. Use this when running @code{cpp}.
9959 Like @samp{%p}, but puts @samp{__} before and after the name of each
9960 predefined macro, except for macros that start with @samp{__} or with
9961 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9965 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9966 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9967 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9968 and @option{-imultilib} as necessary.
9971 Current argument is the name of a library or startup file of some sort.
9972 Search for that file in a standard list of directories and substitute
9973 the full name found. The current working directory is included in the
9974 list of directories scanned.
9977 Current argument is the name of a linker script. Search for that file
9978 in the current list of directories to scan for libraries. If the file
9979 is located insert a @option{--script} option into the command line
9980 followed by the full path name found. If the file is not found then
9981 generate an error message. Note: the current working directory is not
9985 Print @var{str} as an error message. @var{str} is terminated by a newline.
9986 Use this when inconsistent options are detected.
9989 Substitute the contents of spec string @var{name} at this point.
9991 @item %x@{@var{option}@}
9992 Accumulate an option for @samp{%X}.
9995 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9999 Output the accumulated assembler options specified by @option{-Wa}.
10002 Output the accumulated preprocessor options specified by @option{-Wp}.
10005 Process the @code{asm} spec. This is used to compute the
10006 switches to be passed to the assembler.
10009 Process the @code{asm_final} spec. This is a spec string for
10010 passing switches to an assembler post-processor, if such a program is
10014 Process the @code{link} spec. This is the spec for computing the
10015 command line passed to the linker. Typically it will make use of the
10016 @samp{%L %G %S %D and %E} sequences.
10019 Dump out a @option{-L} option for each directory that GCC believes might
10020 contain startup files. If the target supports multilibs then the
10021 current multilib directory will be prepended to each of these paths.
10024 Process the @code{lib} spec. This is a spec string for deciding which
10025 libraries should be included on the command line to the linker.
10028 Process the @code{libgcc} spec. This is a spec string for deciding
10029 which GCC support library should be included on the command line to the linker.
10032 Process the @code{startfile} spec. This is a spec for deciding which
10033 object files should be the first ones passed to the linker. Typically
10034 this might be a file named @file{crt0.o}.
10037 Process the @code{endfile} spec. This is a spec string that specifies
10038 the last object files that will be passed to the linker.
10041 Process the @code{cpp} spec. This is used to construct the arguments
10042 to be passed to the C preprocessor.
10045 Process the @code{cc1} spec. This is used to construct the options to be
10046 passed to the actual C compiler (@samp{cc1}).
10049 Process the @code{cc1plus} spec. This is used to construct the options to be
10050 passed to the actual C++ compiler (@samp{cc1plus}).
10053 Substitute the variable part of a matched option. See below.
10054 Note that each comma in the substituted string is replaced by
10058 Remove all occurrences of @code{-S} from the command line. Note---this
10059 command is position dependent. @samp{%} commands in the spec string
10060 before this one will see @code{-S}, @samp{%} commands in the spec string
10061 after this one will not.
10063 @item %:@var{function}(@var{args})
10064 Call the named function @var{function}, passing it @var{args}.
10065 @var{args} is first processed as a nested spec string, then split
10066 into an argument vector in the usual fashion. The function returns
10067 a string which is processed as if it had appeared literally as part
10068 of the current spec.
10070 The following built-in spec functions are provided:
10073 @item @code{getenv}
10074 The @code{getenv} spec function takes two arguments: an environment
10075 variable name and a string. If the environment variable is not
10076 defined, a fatal error is issued. Otherwise, the return value is the
10077 value of the environment variable concatenated with the string. For
10078 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
10081 %:getenv(TOPDIR /include)
10084 expands to @file{/path/to/top/include}.
10086 @item @code{if-exists}
10087 The @code{if-exists} spec function takes one argument, an absolute
10088 pathname to a file. If the file exists, @code{if-exists} returns the
10089 pathname. Here is a small example of its usage:
10093 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
10096 @item @code{if-exists-else}
10097 The @code{if-exists-else} spec function is similar to the @code{if-exists}
10098 spec function, except that it takes two arguments. The first argument is
10099 an absolute pathname to a file. If the file exists, @code{if-exists-else}
10100 returns the pathname. If it does not exist, it returns the second argument.
10101 This way, @code{if-exists-else} can be used to select one file or another,
10102 based on the existence of the first. Here is a small example of its usage:
10106 crt0%O%s %:if-exists(crti%O%s) \
10107 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
10110 @item @code{replace-outfile}
10111 The @code{replace-outfile} spec function takes two arguments. It looks for the
10112 first argument in the outfiles array and replaces it with the second argument. Here
10113 is a small example of its usage:
10116 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10119 @item @code{remove-outfile}
10120 The @code{remove-outfile} spec function takes one argument. It looks for the
10121 first argument in the outfiles array and removes it. Here is a small example
10125 %:remove-outfile(-lm)
10128 @item @code{pass-through-libs}
10129 The @code{pass-through-libs} spec function takes any number of arguments. It
10130 finds any @option{-l} options and any non-options ending in ".a" (which it
10131 assumes are the names of linker input library archive files) and returns a
10132 result containing all the found arguments each prepended by
10133 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10134 intended to be passed to the LTO linker plugin.
10137 %:pass-through-libs(%G %L %G)
10140 @item @code{print-asm-header}
10141 The @code{print-asm-header} function takes no arguments and simply
10142 prints a banner like:
10148 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10151 It is used to separate compiler options from assembler options
10152 in the @option{--target-help} output.
10155 @item %@{@code{S}@}
10156 Substitutes the @code{-S} switch, if that switch was given to GCC@.
10157 If that switch was not specified, this substitutes nothing. Note that
10158 the leading dash is omitted when specifying this option, and it is
10159 automatically inserted if the substitution is performed. Thus the spec
10160 string @samp{%@{foo@}} would match the command-line option @option{-foo}
10161 and would output the command-line option @option{-foo}.
10163 @item %W@{@code{S}@}
10164 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10165 deleted on failure.
10167 @item %@{@code{S}*@}
10168 Substitutes all the switches specified to GCC whose names start
10169 with @code{-S}, but which also take an argument. This is used for
10170 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10171 GCC considers @option{-o foo} as being
10172 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10173 text, including the space. Thus two arguments would be generated.
10175 @item %@{@code{S}*&@code{T}*@}
10176 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10177 (the order of @code{S} and @code{T} in the spec is not significant).
10178 There can be any number of ampersand-separated variables; for each the
10179 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10181 @item %@{@code{S}:@code{X}@}
10182 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10184 @item %@{!@code{S}:@code{X}@}
10185 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10187 @item %@{@code{S}*:@code{X}@}
10188 Substitutes @code{X} if one or more switches whose names start with
10189 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10190 once, no matter how many such switches appeared. However, if @code{%*}
10191 appears somewhere in @code{X}, then @code{X} will be substituted once
10192 for each matching switch, with the @code{%*} replaced by the part of
10193 that switch that matched the @code{*}.
10195 @item %@{.@code{S}:@code{X}@}
10196 Substitutes @code{X}, if processing a file with suffix @code{S}.
10198 @item %@{!.@code{S}:@code{X}@}
10199 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10201 @item %@{,@code{S}:@code{X}@}
10202 Substitutes @code{X}, if processing a file for language @code{S}.
10204 @item %@{!,@code{S}:@code{X}@}
10205 Substitutes @code{X}, if not processing a file for language @code{S}.
10207 @item %@{@code{S}|@code{P}:@code{X}@}
10208 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10209 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10210 @code{*} sequences as well, although they have a stronger binding than
10211 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10212 alternatives must be starred, and only the first matching alternative
10215 For example, a spec string like this:
10218 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10221 will output the following command-line options from the following input
10222 command-line options:
10227 -d fred.c -foo -baz -boggle
10228 -d jim.d -bar -baz -boggle
10231 @item %@{S:X; T:Y; :D@}
10233 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10234 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10235 be as many clauses as you need. This may be combined with @code{.},
10236 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10241 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10242 construct may contain other nested @samp{%} constructs or spaces, or
10243 even newlines. They are processed as usual, as described above.
10244 Trailing white space in @code{X} is ignored. White space may also
10245 appear anywhere on the left side of the colon in these constructs,
10246 except between @code{.} or @code{*} and the corresponding word.
10248 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10249 handled specifically in these constructs. If another value of
10250 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10251 @option{-W} switch is found later in the command line, the earlier
10252 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10253 just one letter, which passes all matching options.
10255 The character @samp{|} at the beginning of the predicate text is used to
10256 indicate that a command should be piped to the following command, but
10257 only if @option{-pipe} is specified.
10259 It is built into GCC which switches take arguments and which do not.
10260 (You might think it would be useful to generalize this to allow each
10261 compiler's spec to say which switches take arguments. But this cannot
10262 be done in a consistent fashion. GCC cannot even decide which input
10263 files have been specified without knowing which switches take arguments,
10264 and it must know which input files to compile in order to tell which
10267 GCC also knows implicitly that arguments starting in @option{-l} are to be
10268 treated as compiler output files, and passed to the linker in their
10269 proper position among the other output files.
10271 @c man begin OPTIONS
10273 @node Target Options
10274 @section Specifying Target Machine and Compiler Version
10275 @cindex target options
10276 @cindex cross compiling
10277 @cindex specifying machine version
10278 @cindex specifying compiler version and target machine
10279 @cindex compiler version, specifying
10280 @cindex target machine, specifying
10282 The usual way to run GCC is to run the executable called @command{gcc}, or
10283 @command{@var{machine}-gcc} when cross-compiling, or
10284 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10285 one that was installed last.
10287 @node Submodel Options
10288 @section Hardware Models and Configurations
10289 @cindex submodel options
10290 @cindex specifying hardware config
10291 @cindex hardware models and configurations, specifying
10292 @cindex machine dependent options
10294 Each target machine types can have its own
10295 special options, starting with @samp{-m}, to choose among various
10296 hardware models or configurations---for example, 68010 vs 68020,
10297 floating coprocessor or none. A single installed version of the
10298 compiler can compile for any model or configuration, according to the
10301 Some configurations of the compiler also support additional special
10302 options, usually for compatibility with other compilers on the same
10305 @c This list is ordered alphanumerically by subsection name.
10306 @c It should be the same order and spelling as these options are listed
10307 @c in Machine Dependent Options
10310 * Adapteva Epiphany Options::
10313 * Blackfin Options::
10318 * DEC Alpha Options::
10319 * DEC Alpha/VMS Options::
10322 * GNU/Linux Options::
10325 * i386 and x86-64 Options::
10326 * i386 and x86-64 Windows Options::
10328 * IA-64/VMS Options::
10335 * MicroBlaze Options::
10338 * MN10300 Options::
10340 * picoChip Options::
10341 * PowerPC Options::
10343 * RS/6000 and PowerPC Options::
10345 * S/390 and zSeries Options::
10348 * Solaris 2 Options::
10351 * System V Options::
10354 * VxWorks Options::
10356 * Xstormy16 Options::
10358 * zSeries Options::
10361 @node Adapteva Epiphany Options
10362 @subsection Adapteva Epiphany Options
10364 These @samp{-m} options are defined for Adapteva Epiphany:
10367 @item -mhalf-reg-file
10368 @opindex mhalf-reg-file
10369 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
10370 That allows code to run on hardware variants that lack these registers.
10372 @item -mprefer-short-insn-regs
10373 @opindex mprefer-short-insn-regs
10374 Preferrentially allocate registers that allow short instruction generation.
10375 This can result in increasesd instruction count, so if this reduces or
10376 increases code size might vary from case to case.
10378 @item -mbranch-cost=@var{num}
10379 @opindex mbranch-cost
10380 Set the cost of branches to roughly @var{num} ``simple'' instructions.
10381 This cost is only a heuristic and is not guaranteed to produce
10382 consistent results across releases.
10386 Enable the generation of conditional moves.
10388 @item -mnops=@var{num}
10390 Emit @var{num} nops before every other generated instruction.
10392 @item -mno-soft-cmpsf
10393 @opindex mno-soft-cmpsf
10394 For single-precision floating-point comparisons, emit an fsub instruction
10395 and test the flags. This is faster than a software comparison, but can
10396 get incorrect results in the presence of NaNs, or when two different small
10397 numbers are compared such that their difference is calculated as zero.
10398 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
10399 software comparisons.
10401 @item -mstack-offset=@var{num}
10402 @opindex mstack-offset
10403 Set the offset between the top of the stack and the stack pointer.
10404 E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7
10405 can be used by leaf functions without stack allocation.
10406 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
10407 Note also that this option changes the ABI, compiling a program with a
10408 different stack offset than the libraries have been compiled with
10409 will generally not work.
10410 This option can be useful if you want to evaluate if a different stack
10411 offset would give you better code, but to actually use a different stack
10412 offset to build working programs, it is recommended to configure the
10413 toolchain with the appropriate @samp{--with-stack-offset=@var{num}} option.
10415 @item -mno-round-nearest
10416 @opindex mno-round-nearest
10417 Make the scheduler assume that the rounding mode has been set to
10418 truncating. The default is @option{-mround-nearest}.
10421 @opindex mlong-calls
10422 If not otherwise specified by an attribute, assume all calls might be beyond
10423 the offset range of the b / bl instructions, and therefore load the
10424 function address into a register before performing a (otherwise direct) call.
10425 This is the default.
10427 @item -mshort-calls
10428 @opindex short-calls
10429 If not otherwise specified by an attribute, assume all direct calls are
10430 in the range of the b / bl instructions, so use these instructions
10431 for direct calls. The default is @option{-mlong-calls}.
10435 Assume addresses can be loaded as 16 bit unsigned values. This does not
10436 apply to function addresses for which @option{-mlong-calls} semantics
10439 @item -mfp-mode=@var{mode}
10441 Set the prevailing mode of the floating-point unit.
10442 This determines the floating-point mode that is provided and expected
10443 at function call and return time. Making this mode match the mode you
10444 predominantly need at function start can make your programs smaller and
10445 faster by avoiding unnecessary mode switches.
10447 @var{mode} can be set to one the following values:
10451 Any mode at function entry is valid, and retained or restored when
10452 the function returns, and when it calls other functions.
10453 This mode is useful for compiling libraries or other compilation units
10454 you might want to incorporate into different programs with different
10455 prevailing FPU modes, and the convenience of being able to use a single
10456 object file outweighs the size and speed overhead for any extra
10457 mode switching that might be needed, compared with what would be needed
10458 with a more specific choice of prevailing FPU mode.
10461 This is the mode used for floating-point calculations with
10462 truncating (i.e.@: round towards zero) rounding mode. That includes
10463 conversion from floating point to integer.
10465 @item round-nearest
10466 This is the mode used for floating-point calculations with
10467 round-to-nearest-or-even rounding mode.
10470 This is the mode used to perform integer calculations in the FPU, e.g.@:
10471 integer multiply, or integer multiply-and-accumulate.
10474 The default is @option{-mfp-mode=caller}
10476 @item -mnosplit-lohi
10477 @opindex mnosplit-lohi
10479 @opindex mno-postinc
10480 @item -mno-postmodify
10481 @opindex mno-postmodify
10482 Code generation tweaks that disable, respectively, splitting of 32
10483 bit loads, generation of post-increment addresses, and generation of
10484 post-modify addresses. The defaults are @option{msplit-lohi},
10485 @option{-mpost-inc}, and @option{-mpost-modify}.
10487 @item -mnovect-double
10488 @opindex mno-vect-double
10489 Change the preferred SIMD mode to SImode. The default is
10490 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
10492 @item -max-vect-align=@var{num}
10493 @opindex max-vect-align
10494 The maximum alignment for SIMD vector mode types.
10495 @var{num} may be 4 or 8. The default is 8.
10496 Note that this is an ABI change, even though many library function
10497 interfaces will be unaffected, if they don't use SIMD vector modes
10498 in places where they affect size and/or alignment of relevant types.
10500 @item -msplit-vecmove-early
10501 @opindex msplit-vecmove-early
10502 Split vector moves into single word moves before reload. In theory this
10503 could give better register allocation, but so far the reverse seems to be
10504 generally the case.
10506 @item -m1reg-@var{reg}
10508 Specify a register to hold the constant @minus{}1, which makes loading small negative
10509 constants and certain bitmasks faster.
10510 Allowable values for reg are r43 and r63, which specify to use that register
10511 as a fixed register, and none, which means that no register is used for this
10512 purpose. The default is @option{-m1reg-none}.
10517 @subsection ARM Options
10518 @cindex ARM options
10520 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10524 @item -mabi=@var{name}
10526 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10527 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10530 @opindex mapcs-frame
10531 Generate a stack frame that is compliant with the ARM Procedure Call
10532 Standard for all functions, even if this is not strictly necessary for
10533 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10534 with this option will cause the stack frames not to be generated for
10535 leaf functions. The default is @option{-mno-apcs-frame}.
10539 This is a synonym for @option{-mapcs-frame}.
10542 @c not currently implemented
10543 @item -mapcs-stack-check
10544 @opindex mapcs-stack-check
10545 Generate code to check the amount of stack space available upon entry to
10546 every function (that actually uses some stack space). If there is
10547 insufficient space available then either the function
10548 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10549 called, depending upon the amount of stack space required. The runtime
10550 system is required to provide these functions. The default is
10551 @option{-mno-apcs-stack-check}, since this produces smaller code.
10553 @c not currently implemented
10555 @opindex mapcs-float
10556 Pass floating-point arguments using the floating-point registers. This is
10557 one of the variants of the APCS@. This option is recommended if the
10558 target hardware has a floating-point unit or if a lot of floating-point
10559 arithmetic is going to be performed by the code. The default is
10560 @option{-mno-apcs-float}, since integer only code is slightly increased in
10561 size if @option{-mapcs-float} is used.
10563 @c not currently implemented
10564 @item -mapcs-reentrant
10565 @opindex mapcs-reentrant
10566 Generate reentrant, position independent code. The default is
10567 @option{-mno-apcs-reentrant}.
10570 @item -mthumb-interwork
10571 @opindex mthumb-interwork
10572 Generate code which supports calling between the ARM and Thumb
10573 instruction sets. Without this option, on pre-v5 architectures, the
10574 two instruction sets cannot be reliably used inside one program. The
10575 default is @option{-mno-thumb-interwork}, since slightly larger code
10576 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10577 configurations this option is meaningless.
10579 @item -mno-sched-prolog
10580 @opindex mno-sched-prolog
10581 Prevent the reordering of instructions in the function prolog, or the
10582 merging of those instruction with the instructions in the function's
10583 body. This means that all functions will start with a recognizable set
10584 of instructions (or in fact one of a choice from a small set of
10585 different function prologues), and this information can be used to
10586 locate the start if functions inside an executable piece of code. The
10587 default is @option{-msched-prolog}.
10589 @item -mfloat-abi=@var{name}
10590 @opindex mfloat-abi
10591 Specifies which floating-point ABI to use. Permissible values
10592 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10594 Specifying @samp{soft} causes GCC to generate output containing
10595 library calls for floating-point operations.
10596 @samp{softfp} allows the generation of code using hardware floating-point
10597 instructions, but still uses the soft-float calling conventions.
10598 @samp{hard} allows generation of floating-point instructions
10599 and uses FPU-specific calling conventions.
10601 The default depends on the specific target configuration. Note that
10602 the hard-float and soft-float ABIs are not link-compatible; you must
10603 compile your entire program with the same ABI, and link with a
10604 compatible set of libraries.
10606 @item -mlittle-endian
10607 @opindex mlittle-endian
10608 Generate code for a processor running in little-endian mode. This is
10609 the default for all standard configurations.
10612 @opindex mbig-endian
10613 Generate code for a processor running in big-endian mode; the default is
10614 to compile code for a little-endian processor.
10616 @item -mwords-little-endian
10617 @opindex mwords-little-endian
10618 This option only applies when generating code for big-endian processors.
10619 Generate code for a little-endian word order but a big-endian byte
10620 order. That is, a byte order of the form @samp{32107654}. Note: this
10621 option should only be used if you require compatibility with code for
10622 big-endian ARM processors generated by versions of the compiler prior to
10623 2.8. This option is now deprecated.
10625 @item -mcpu=@var{name}
10627 This specifies the name of the target ARM processor. GCC uses this name
10628 to determine what kind of instructions it can emit when generating
10629 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10630 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10631 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10632 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10633 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10635 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10636 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10637 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10638 @samp{strongarm1110},
10639 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10640 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10641 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10642 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10643 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10644 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10645 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10646 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
10647 @samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10648 @samp{cortex-m4}, @samp{cortex-m3},
10651 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10652 @samp{fa526}, @samp{fa626},
10653 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10656 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10657 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10658 See @option{-mtune} for more information.
10660 @option{-mcpu=native} causes the compiler to auto-detect the CPU
10661 of the build computer. At present, this feature is only supported on
10662 Linux, and not all architectures are recognised. If the auto-detect is
10663 unsuccessful the option has no effect.
10665 @item -mtune=@var{name}
10667 This option is very similar to the @option{-mcpu=} option, except that
10668 instead of specifying the actual target processor type, and hence
10669 restricting which instructions can be used, it specifies that GCC should
10670 tune the performance of the code as if the target were of the type
10671 specified in this option, but still choosing the instructions that it
10672 will generate based on the CPU specified by a @option{-mcpu=} option.
10673 For some ARM implementations better performance can be obtained by using
10676 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10677 performance for a blend of processors within architecture @var{arch}.
10678 The aim is to generate code that run well on the current most popular
10679 processors, balancing between optimizations that benefit some CPUs in the
10680 range, and avoiding performance pitfalls of other CPUs. The effects of
10681 this option may change in future GCC versions as CPU models come and go.
10683 @option{-mtune=native} causes the compiler to auto-detect the CPU
10684 of the build computer. At present, this feature is only supported on
10685 Linux, and not all architectures are recognised. If the auto-detect is
10686 unsuccessful the option has no effect.
10688 @item -march=@var{name}
10690 This specifies the name of the target ARM architecture. GCC uses this
10691 name to determine what kind of instructions it can emit when generating
10692 assembly code. This option can be used in conjunction with or instead
10693 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10694 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10695 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10696 @samp{armv6}, @samp{armv6j},
10697 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10698 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10699 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10701 @option{-march=native} causes the compiler to auto-detect the architecture
10702 of the build computer. At present, this feature is only supported on
10703 Linux, and not all architectures are recognised. If the auto-detect is
10704 unsuccessful the option has no effect.
10706 @item -mfpu=@var{name}
10707 @itemx -mfpe=@var{number}
10708 @itemx -mfp=@var{number}
10712 This specifies what floating-point hardware (or hardware emulation) is
10713 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10714 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10715 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10716 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10717 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10718 @option{-mfp} and @option{-mfpe} are synonyms for
10719 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10722 If @option{-msoft-float} is specified this specifies the format of
10723 floating-point values.
10725 If the selected floating-point hardware includes the NEON extension
10726 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10727 operations will not be used by GCC's auto-vectorization pass unless
10728 @option{-funsafe-math-optimizations} is also specified. This is
10729 because NEON hardware does not fully implement the IEEE 754 standard for
10730 floating-point arithmetic (in particular denormal values are treated as
10731 zero), so the use of NEON instructions may lead to a loss of precision.
10733 @item -mfp16-format=@var{name}
10734 @opindex mfp16-format
10735 Specify the format of the @code{__fp16} half-precision floating-point type.
10736 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10737 the default is @samp{none}, in which case the @code{__fp16} type is not
10738 defined. @xref{Half-Precision}, for more information.
10740 @item -mstructure-size-boundary=@var{n}
10741 @opindex mstructure-size-boundary
10742 The size of all structures and unions will be rounded up to a multiple
10743 of the number of bits set by this option. Permissible values are 8, 32
10744 and 64. The default value varies for different toolchains. For the COFF
10745 targeted toolchain the default value is 8. A value of 64 is only allowed
10746 if the underlying ABI supports it.
10748 Specifying the larger number can produce faster, more efficient code, but
10749 can also increase the size of the program. Different values are potentially
10750 incompatible. Code compiled with one value cannot necessarily expect to
10751 work with code or libraries compiled with another value, if they exchange
10752 information using structures or unions.
10754 @item -mabort-on-noreturn
10755 @opindex mabort-on-noreturn
10756 Generate a call to the function @code{abort} at the end of a
10757 @code{noreturn} function. It will be executed if the function tries to
10761 @itemx -mno-long-calls
10762 @opindex mlong-calls
10763 @opindex mno-long-calls
10764 Tells the compiler to perform function calls by first loading the
10765 address of the function into a register and then performing a subroutine
10766 call on this register. This switch is needed if the target function
10767 will lie outside of the 64 megabyte addressing range of the offset based
10768 version of subroutine call instruction.
10770 Even if this switch is enabled, not all function calls will be turned
10771 into long calls. The heuristic is that static functions, functions
10772 which have the @samp{short-call} attribute, functions that are inside
10773 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10774 definitions have already been compiled within the current compilation
10775 unit, will not be turned into long calls. The exception to this rule is
10776 that weak function definitions, functions with the @samp{long-call}
10777 attribute or the @samp{section} attribute, and functions that are within
10778 the scope of a @samp{#pragma long_calls} directive, will always be
10779 turned into long calls.
10781 This feature is not enabled by default. Specifying
10782 @option{-mno-long-calls} will restore the default behavior, as will
10783 placing the function calls within the scope of a @samp{#pragma
10784 long_calls_off} directive. Note these switches have no effect on how
10785 the compiler generates code to handle function calls via function
10788 @item -msingle-pic-base
10789 @opindex msingle-pic-base
10790 Treat the register used for PIC addressing as read-only, rather than
10791 loading it in the prologue for each function. The runtime system is
10792 responsible for initializing this register with an appropriate value
10793 before execution begins.
10795 @item -mpic-register=@var{reg}
10796 @opindex mpic-register
10797 Specify the register to be used for PIC addressing. The default is R10
10798 unless stack-checking is enabled, when R9 is used.
10800 @item -mcirrus-fix-invalid-insns
10801 @opindex mcirrus-fix-invalid-insns
10802 @opindex mno-cirrus-fix-invalid-insns
10803 Insert NOPs into the instruction stream to in order to work around
10804 problems with invalid Maverick instruction combinations. This option
10805 is only valid if the @option{-mcpu=ep9312} option has been used to
10806 enable generation of instructions for the Cirrus Maverick floating-point
10807 co-processor. This option is not enabled by default, since the
10808 problem is only present in older Maverick implementations. The default
10809 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10812 @item -mpoke-function-name
10813 @opindex mpoke-function-name
10814 Write the name of each function into the text section, directly
10815 preceding the function prologue. The generated code is similar to this:
10819 .ascii "arm_poke_function_name", 0
10822 .word 0xff000000 + (t1 - t0)
10823 arm_poke_function_name
10825 stmfd sp!, @{fp, ip, lr, pc@}
10829 When performing a stack backtrace, code can inspect the value of
10830 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10831 location @code{pc - 12} and the top 8 bits are set, then we know that
10832 there is a function name embedded immediately preceding this location
10833 and has length @code{((pc[-3]) & 0xff000000)}.
10840 Select between generating code that executes in ARM and Thumb
10841 states. The default for most configurations is to generate code
10842 that executes in ARM state, but the default can be changed by
10843 configuring GCC with the @option{--with-mode=}@var{state}
10847 @opindex mtpcs-frame
10848 Generate a stack frame that is compliant with the Thumb Procedure Call
10849 Standard for all non-leaf functions. (A leaf function is one that does
10850 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10852 @item -mtpcs-leaf-frame
10853 @opindex mtpcs-leaf-frame
10854 Generate a stack frame that is compliant with the Thumb Procedure Call
10855 Standard for all leaf functions. (A leaf function is one that does
10856 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10858 @item -mcallee-super-interworking
10859 @opindex mcallee-super-interworking
10860 Gives all externally visible functions in the file being compiled an ARM
10861 instruction set header which switches to Thumb mode before executing the
10862 rest of the function. This allows these functions to be called from
10863 non-interworking code. This option is not valid in AAPCS configurations
10864 because interworking is enabled by default.
10866 @item -mcaller-super-interworking
10867 @opindex mcaller-super-interworking
10868 Allows calls via function pointers (including virtual functions) to
10869 execute correctly regardless of whether the target code has been
10870 compiled for interworking or not. There is a small overhead in the cost
10871 of executing a function pointer if this option is enabled. This option
10872 is not valid in AAPCS configurations because interworking is enabled
10875 @item -mtp=@var{name}
10877 Specify the access model for the thread local storage pointer. The valid
10878 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10879 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10880 (supported in the arm6k architecture), and @option{auto}, which uses the
10881 best available method for the selected processor. The default setting is
10884 @item -mtls-dialect=@var{dialect}
10885 @opindex mtls-dialect
10886 Specify the dialect to use for accessing thread local storage. Two
10887 dialects are supported --- @option{gnu} and @option{gnu2}. The
10888 @option{gnu} dialect selects the original GNU scheme for supporting
10889 local and global dynamic TLS models. The @option{gnu2} dialect
10890 selects the GNU descriptor scheme, which provides better performance
10891 for shared libraries. The GNU descriptor scheme is compatible with
10892 the original scheme, but does require new assembler, linker and
10893 library support. Initial and local exec TLS models are unaffected by
10894 this option and always use the original scheme.
10896 @item -mword-relocations
10897 @opindex mword-relocations
10898 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10899 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10900 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10903 @item -mfix-cortex-m3-ldrd
10904 @opindex mfix-cortex-m3-ldrd
10905 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10906 with overlapping destination and base registers are used. This option avoids
10907 generating these instructions. This option is enabled by default when
10908 @option{-mcpu=cortex-m3} is specified.
10913 @subsection AVR Options
10914 @cindex AVR Options
10916 These options are defined for AVR implementations:
10919 @item -mmcu=@var{mcu}
10921 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
10923 For a complete list of @var{mcu} values that are supported by avr-gcc,
10924 see the compiler output when called with the @code{--help=target}
10925 command line option.
10926 The default for this option is@tie{}@code{avr2}.
10928 avr-gcc supports the following AVR devices and ISAs:
10933 This ISA is implemented by the minimal AVR core and supported
10934 for assembler only.
10935 @*@var{mcu}@tie{}= @code{at90s1200},
10936 @code{attiny10}, @code{attiny11}, @code{attiny12}, @code{attiny15},
10940 ``Classic'' devices with up to 8@tie{}KiB of program memory.
10941 @*@var{mcu}@tie{}= @code{at90s2313}, @code{attiny26}, @code{at90c8534},
10945 ``Classic'' devices with up to 8@tie{}KiB of program memory and with
10946 the @code{MOVW} instruction.
10947 @*@var{mcu}@tie{}= @code{attiny2313}, @code{attiny261}, @code{attiny24},
10951 ``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory.
10952 @*@var{mcu}@tie{}= @code{at43usb355}, @code{at76c711}.
10955 ``Classic'' devices with 128@tie{}KiB of program memory.
10956 @*@var{mcu}@tie{}= @code{atmega103}, @code{at43usb320}.
10959 ``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program
10960 memory and with the @code{MOVW} instruction.
10961 @*@var{mcu}@tie{}= @code{at90usb162}, @code{atmega8u2},
10962 @code{attiny167}, @dots{}
10965 ``Enhanced'' devices with up to 8@tie{}KiB of program memory.
10966 @*@var{mcu}@tie{}= @code{atmega8}, @code{atmega88}, @code{at90pwm81},
10970 ``Enhanced'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory.
10971 @*@var{mcu}@tie{}= @code{atmega16}, @code{atmega6490}, @code{at90can64},
10975 ``Enhanced'' devices with 128@tie{}KiB of program memory.
10976 @*@var{mcu}@tie{}= @code{atmega128}, @code{at90can128}, @code{at90usb1287},
10980 ``Enhanced'' devices with 3-byte PC, i.e.@: with at least 256@tie{}KiB
10982 @*@var{mcu}@tie{}= @code{atmega2560}, @code{atmega2561}.
10987 @item -maccumulate-args
10988 @opindex maccumulate-args
10989 Accumulate outgoing function arguments and acquire/release the needed
10990 stack space for outgoing function arguments once in function
10991 prologue/epilogue. Without this option, outgoing arguments are pushed
10992 before calling a function and popped afterwards.
10994 Popping the arguments after the function call can be expensive on
10995 AVR so that accumulating the stack space might lead to smaller
10996 executables because arguments need not to be removed from the
10997 stack after such a function call.
10999 This option can lead to reduced code size for functions that perform
11000 several calls to functions which get their arguments on the stack like
11001 calls to printf-like functions.
11003 @item -mbranch-cost=@var{cost}
11004 @opindex mbranch-cost
11005 Set the branch costs for conditional branch instructions to
11006 @var{cost}. Reasonable values for @var{cost} are small, non-negative
11007 integers. The default branch cost is 0.
11009 @item -mcall-prologues
11010 @opindex mcall-prologues
11011 Functions prologues/epilogues expanded as call to appropriate
11012 subroutines. Code size will be smaller.
11016 Assume int to be 8 bit integer. This affects the sizes of all types: A
11017 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
11018 and long long will be 4 bytes. Please note that this option does not
11019 comply to the C standards, but it will provide you with smaller code
11022 @item -mno-interrupts
11023 @opindex mno-interrupts
11024 Generated code is not compatible with hardware interrupts.
11025 Code size will be smaller.
11029 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
11030 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
11031 Setting @code{-mrelax} just adds the @code{--relax} option to the
11032 linker command line when the linker is called.
11034 Jump relaxing is performed by the linker because jump offsets are not
11035 known before code is located. Therefore, the assembler code generated by the
11036 compiler will be the same, but the instructions in the executable may
11037 differ from instructions in the assembler code.
11039 @item -mshort-calls
11040 @opindex mshort-calls
11041 Use @code{RCALL}/@code{RJMP} instructions even on devices with
11042 16@tie{}KiB or more of program memory, i.e.@: on devices that
11043 have the @code{CALL} and @code{JMP} instructions.
11044 See also the @code{-mrelax} command line option.
11048 Use address register @code{X} in a way proposed by the hardware. This means
11049 that @code{X} will only be used in indirect, post-increment or
11050 pre-decrement addressing.
11052 Without this option, the @code{X} register may be used in the same way
11053 as @code{Y} or @code{Z} which then is emulated by additional
11055 For example, loading a value with @code{X+const} addressing with a
11056 small non-negative @code{const < 64} to a register @var{Rn} will be
11060 adiw r26, const ; X += const
11061 ld @var{Rn}, X ; @var{Rn} = *X
11062 sbiw r26, const ; X -= const
11066 @opindex mtiny-stack
11067 Only use the lower 8@tie{}bits of the stack pointer and assume that the high
11068 byte of SP is always zero.
11071 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
11073 Pointers in the implementation are 16@tie{}bits wide.
11074 The address of a function or label is represented as word address so
11075 that indirect jumps and calls can target any code address in the
11076 range of 64@tie{}Ki words.
11078 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
11079 bytes of program memory space, there is a special function register called
11080 @code{EIND} that serves as most significant part of the target address
11081 when @code{EICALL} or @code{EIJMP} instructions are used.
11083 Indirect jumps and calls on these devices are handled as follows by
11084 the compiler and are subject to some limitations:
11089 The compiler never sets @code{EIND}.
11092 The startup code from libgcc never sets @code{EIND}.
11093 Notice that startup code is a blend of code from libgcc and avr-libc.
11094 For the impact of avr-libc on @code{EIND}, see the
11095 @w{@uref{http://nongnu.org/avr-libc/user-manual,avr-libc user manual}}.
11098 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
11099 instructions or might read @code{EIND} directly in order to emulate an
11100 indirect call/jump by means of a @code{RET} instruction.
11103 The compiler assumes that @code{EIND} never changes during the startup
11104 code or during the application. In particular, @code{EIND} is not
11105 saved/restored in function or interrupt service routine
11109 It is legitimate for user-specific startup code to set up @code{EIND}
11110 early, for example by means of initialization code located in
11111 section @code{.init3}. Such code runs prior to general startup code
11112 that initializes RAM and calls constructors.
11115 For indirect calls to functions and computed goto, the linker will
11116 generate @emph{stubs}. Stubs are jump pads sometimes also called
11117 @emph{trampolines}. Thus, the indirect call/jump will jump to such a stub.
11118 The stub contains a direct jump to the desired address.
11121 Stubs will be generated automatically by the linker if
11122 the following two conditions are met:
11125 @item The address of a label is taken by means of the @code{gs} modifier
11126 (short for @emph{generate stubs}) like so:
11128 LDI r24, lo8(gs(@var{func}))
11129 LDI r25, hi8(gs(@var{func}))
11131 @item The final location of that label is in a code segment
11132 @emph{outside} the segment where the stubs are located.
11136 The compiler will emit such @code{gs} modifiers for code labels in the
11137 following situations:
11139 @item Taking address of a function or code label.
11140 @item Computed goto.
11141 @item If prologue-save function is used, see @option{-mcall-prologues}
11142 command-line option.
11143 @item Switch/case dispatch tables. If you do not want such dispatch
11144 tables you can specify the @option{-fno-jump-tables} command-line option.
11145 @item C and C++ constructors/destructors called during startup/shutdown.
11146 @item If the tools hit a @code{gs()} modifier explained above.
11150 The default linker script is arranged for code with @code{EIND = 0}.
11151 If code is supposed to work for a setup with @code{EIND != 0}, a custom
11152 linker script has to be used in order to place the sections whose
11153 name start with @code{.trampolines} into the segment where @code{EIND}
11157 Jumping to non-symbolic addresses like so is @emph{not} supported:
11162 /* Call function at word address 0x2 */
11163 return ((int(*)(void)) 0x2)();
11167 Instead, a stub has to be set up, i.e.@: the function has to be called
11168 through a symbol (@code{func_4} in the example):
11173 extern int func_4 (void);
11175 /* Call function at byte address 0x4 */
11180 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
11181 Alternatively, @code{func_4} can be defined in the linker script.
11184 @subsubsection AVR Built-in Macros
11186 avr-gcc defines several built-in macros so that the user code can test
11187 for presence of absence of features. Almost any of the following
11188 built-in macros are deduced from device capabilities and thus
11189 triggered by the @code{-mmcu=} command-line option.
11191 For even more AVR-specific built-in macros see
11192 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
11196 @item __AVR_@var{Device}__
11197 Setting @code{-mmcu=@var{device}} defines this built-in macro that reflects
11198 the device's name. For example, @code{-mmcu=atmega8} will define the
11199 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
11200 @code{__AVR_ATtiny261A__}, etc.
11202 The built-in macros' names follow
11203 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
11204 the device name as from the AVR user manual. The difference between
11205 @var{Device} in the built-in macro and @var{device} in
11206 @code{-mmcu=@var{device}} is that the latter is always lower case.
11208 @item __AVR_HAVE_RAMPZ__
11209 @item __AVR_HAVE_ELPM__
11210 The device has the @code{RAMPZ} special function register and thus the
11211 @code{ELPM} instruction.
11213 @item __AVR_HAVE_ELPMX__
11214 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
11215 R@var{n},Z+} instructions.
11217 @item __AVR_HAVE_MOVW__
11218 The device has the @code{MOVW} instruction to perform 16-bit
11219 register-register moves.
11221 @item __AVR_HAVE_LPMX__
11222 The device has the @code{LPM R@var{n},Z} and @code{LPM
11223 R@var{n},Z+} instructions.
11225 @item __AVR_HAVE_MUL__
11226 The device has a hardware multiplier.
11228 @item __AVR_HAVE_JMP_CALL__
11229 The device has the @code{JMP} and @code{CALL} instructions.
11230 This is the case for devices with at least 16@tie{}KiB of program
11231 memory and if @code{-mshort-calls} is not set.
11233 @item __AVR_HAVE_EIJMP_EICALL__
11234 @item __AVR_3_BYTE_PC__
11235 The device has the @code{EIJMP} and @code{EICALL} instructions.
11236 This is the case for devices with at least 256@tie{}KiB of program memory.
11237 This also means that the program counter
11238 (PC) is 3@tie{}bytes wide.
11240 @item __AVR_2_BYTE_PC__
11241 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
11242 with up to 128@tie{}KiB of program memory.
11244 @item __AVR_HAVE_8BIT_SP__
11245 @item __AVR_HAVE_16BIT_SP__
11246 The stack pointer (SP) is 8@tie{}bits resp. 16@tie{}bits wide.
11247 The definition of these macros is affected by @code{-mtiny-stack}.
11249 @item __NO_INTERRUPTS__
11250 This macro reflects the @code{-mno-interrupts} command line option.
11252 @item __AVR_ERRATA_SKIP__
11253 @item __AVR_ERRATA_SKIP_JMP_CALL__
11254 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
11255 instructions because of a hardware erratum. Skip instructions are
11256 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
11257 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
11260 @item __AVR_SFR_OFFSET__=@var{offset}
11261 Instructions that can address I/O special function registers directly
11262 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
11263 address as if addressed by an instruction to access RAM like @code{LD}
11264 or @code{STS}. This offset depends on the device architecture and has
11265 to be subtracted from the RAM address in order to get the
11266 respective I/O@tie{}address.
11270 @node Blackfin Options
11271 @subsection Blackfin Options
11272 @cindex Blackfin Options
11275 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
11277 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
11278 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
11279 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
11280 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
11281 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
11282 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
11283 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
11284 @samp{bf561}, @samp{bf592}.
11285 The optional @var{sirevision} specifies the silicon revision of the target
11286 Blackfin processor. Any workarounds available for the targeted silicon revision
11287 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
11288 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
11289 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
11290 hexadecimal digits representing the major and minor numbers in the silicon
11291 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
11292 is not defined. If @var{sirevision} is @samp{any}, the
11293 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
11294 If this optional @var{sirevision} is not used, GCC assumes the latest known
11295 silicon revision of the targeted Blackfin processor.
11297 Support for @samp{bf561} is incomplete. For @samp{bf561},
11298 Only the processor macro is defined.
11299 Without this option, @samp{bf532} is used as the processor by default.
11300 The corresponding predefined processor macros for @var{cpu} is to
11301 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
11302 provided by libgloss to be linked in if @option{-msim} is not given.
11306 Specifies that the program will be run on the simulator. This causes
11307 the simulator BSP provided by libgloss to be linked in. This option
11308 has effect only for @samp{bfin-elf} toolchain.
11309 Certain other options, such as @option{-mid-shared-library} and
11310 @option{-mfdpic}, imply @option{-msim}.
11312 @item -momit-leaf-frame-pointer
11313 @opindex momit-leaf-frame-pointer
11314 Don't keep the frame pointer in a register for leaf functions. This
11315 avoids the instructions to save, set up and restore frame pointers and
11316 makes an extra register available in leaf functions. The option
11317 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11318 which might make debugging harder.
11320 @item -mspecld-anomaly
11321 @opindex mspecld-anomaly
11322 When enabled, the compiler will ensure that the generated code does not
11323 contain speculative loads after jump instructions. If this option is used,
11324 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
11326 @item -mno-specld-anomaly
11327 @opindex mno-specld-anomaly
11328 Don't generate extra code to prevent speculative loads from occurring.
11330 @item -mcsync-anomaly
11331 @opindex mcsync-anomaly
11332 When enabled, the compiler will ensure that the generated code does not
11333 contain CSYNC or SSYNC instructions too soon after conditional branches.
11334 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
11336 @item -mno-csync-anomaly
11337 @opindex mno-csync-anomaly
11338 Don't generate extra code to prevent CSYNC or SSYNC instructions from
11339 occurring too soon after a conditional branch.
11343 When enabled, the compiler is free to take advantage of the knowledge that
11344 the entire program fits into the low 64k of memory.
11347 @opindex mno-low-64k
11348 Assume that the program is arbitrarily large. This is the default.
11350 @item -mstack-check-l1
11351 @opindex mstack-check-l1
11352 Do stack checking using information placed into L1 scratchpad memory by the
11355 @item -mid-shared-library
11356 @opindex mid-shared-library
11357 Generate code that supports shared libraries via the library ID method.
11358 This allows for execute in place and shared libraries in an environment
11359 without virtual memory management. This option implies @option{-fPIC}.
11360 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11362 @item -mno-id-shared-library
11363 @opindex mno-id-shared-library
11364 Generate code that doesn't assume ID based shared libraries are being used.
11365 This is the default.
11367 @item -mleaf-id-shared-library
11368 @opindex mleaf-id-shared-library
11369 Generate code that supports shared libraries via the library ID method,
11370 but assumes that this library or executable won't link against any other
11371 ID shared libraries. That allows the compiler to use faster code for jumps
11374 @item -mno-leaf-id-shared-library
11375 @opindex mno-leaf-id-shared-library
11376 Do not assume that the code being compiled won't link against any ID shared
11377 libraries. Slower code will be generated for jump and call insns.
11379 @item -mshared-library-id=n
11380 @opindex mshared-library-id
11381 Specified the identification number of the ID based shared library being
11382 compiled. Specifying a value of 0 will generate more compact code, specifying
11383 other values will force the allocation of that number to the current
11384 library but is no more space or time efficient than omitting this option.
11388 Generate code that allows the data segment to be located in a different
11389 area of memory from the text segment. This allows for execute in place in
11390 an environment without virtual memory management by eliminating relocations
11391 against the text section.
11393 @item -mno-sep-data
11394 @opindex mno-sep-data
11395 Generate code that assumes that the data segment follows the text segment.
11396 This is the default.
11399 @itemx -mno-long-calls
11400 @opindex mlong-calls
11401 @opindex mno-long-calls
11402 Tells the compiler to perform function calls by first loading the
11403 address of the function into a register and then performing a subroutine
11404 call on this register. This switch is needed if the target function
11405 will lie outside of the 24 bit addressing range of the offset based
11406 version of subroutine call instruction.
11408 This feature is not enabled by default. Specifying
11409 @option{-mno-long-calls} will restore the default behavior. Note these
11410 switches have no effect on how the compiler generates code to handle
11411 function calls via function pointers.
11415 Link with the fast floating-point library. This library relaxes some of
11416 the IEEE floating-point standard's rules for checking inputs against
11417 Not-a-Number (NAN), in the interest of performance.
11420 @opindex minline-plt
11421 Enable inlining of PLT entries in function calls to functions that are
11422 not known to bind locally. It has no effect without @option{-mfdpic}.
11425 @opindex mmulticore
11426 Build standalone application for multicore Blackfin processor. Proper
11427 start files and link scripts will be used to support multicore.
11428 This option defines @code{__BFIN_MULTICORE}. It can only be used with
11429 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
11430 @option{-mcorea} or @option{-mcoreb}. If it's used without
11431 @option{-mcorea} or @option{-mcoreb}, single application/dual core
11432 programming model is used. In this model, the main function of Core B
11433 should be named as coreb_main. If it's used with @option{-mcorea} or
11434 @option{-mcoreb}, one application per core programming model is used.
11435 If this option is not used, single core application programming
11440 Build standalone application for Core A of BF561 when using
11441 one application per core programming model. Proper start files
11442 and link scripts will be used to support Core A. This option
11443 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
11447 Build standalone application for Core B of BF561 when using
11448 one application per core programming model. Proper start files
11449 and link scripts will be used to support Core B. This option
11450 defines @code{__BFIN_COREB}. When this option is used, coreb_main
11451 should be used instead of main. It must be used with
11452 @option{-mmulticore}.
11456 Build standalone application for SDRAM. Proper start files and
11457 link scripts will be used to put the application into SDRAM.
11458 Loader should initialize SDRAM before loading the application
11459 into SDRAM. This option defines @code{__BFIN_SDRAM}.
11463 Assume that ICPLBs are enabled at run time. This has an effect on certain
11464 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
11465 are enabled; for standalone applications the default is off.
11469 @subsection C6X Options
11470 @cindex C6X Options
11473 @item -march=@var{name}
11475 This specifies the name of the target architecture. GCC uses this
11476 name to determine what kind of instructions it can emit when generating
11477 assembly code. Permissible names are: @samp{c62x},
11478 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
11481 @opindex mbig-endian
11482 Generate code for a big endian target.
11484 @item -mlittle-endian
11485 @opindex mlittle-endian
11486 Generate code for a little endian target. This is the default.
11490 Choose startup files and linker script suitable for the simulator.
11492 @item -msdata=default
11493 @opindex msdata=default
11494 Put small global and static data in the @samp{.neardata} section,
11495 which is pointed to by register @code{B14}. Put small uninitialized
11496 global and static data in the @samp{.bss} section, which is adjacent
11497 to the @samp{.neardata} section. Put small read-only data into the
11498 @samp{.rodata} section. The corresponding sections used for large
11499 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
11502 @opindex msdata=all
11503 Put all data, not just small objets, into the sections reserved for
11504 small data, and use addressing relative to the @code{B14} register to
11508 @opindex msdata=none
11509 Make no use of the sections reserved for small data, and use absolute
11510 addresses to access all data. Put all initialized global and static
11511 data in the @samp{.fardata} section, and all uninitialized data in the
11512 @samp{.far} section. Put all constant data into the @samp{.const}
11517 @subsection CRIS Options
11518 @cindex CRIS Options
11520 These options are defined specifically for the CRIS ports.
11523 @item -march=@var{architecture-type}
11524 @itemx -mcpu=@var{architecture-type}
11527 Generate code for the specified architecture. The choices for
11528 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
11529 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
11530 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
11533 @item -mtune=@var{architecture-type}
11535 Tune to @var{architecture-type} everything applicable about the generated
11536 code, except for the ABI and the set of available instructions. The
11537 choices for @var{architecture-type} are the same as for
11538 @option{-march=@var{architecture-type}}.
11540 @item -mmax-stack-frame=@var{n}
11541 @opindex mmax-stack-frame
11542 Warn when the stack frame of a function exceeds @var{n} bytes.
11548 The options @option{-metrax4} and @option{-metrax100} are synonyms for
11549 @option{-march=v3} and @option{-march=v8} respectively.
11551 @item -mmul-bug-workaround
11552 @itemx -mno-mul-bug-workaround
11553 @opindex mmul-bug-workaround
11554 @opindex mno-mul-bug-workaround
11555 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
11556 models where it applies. This option is active by default.
11560 Enable CRIS-specific verbose debug-related information in the assembly
11561 code. This option also has the effect to turn off the @samp{#NO_APP}
11562 formatted-code indicator to the assembler at the beginning of the
11567 Do not use condition-code results from previous instruction; always emit
11568 compare and test instructions before use of condition codes.
11570 @item -mno-side-effects
11571 @opindex mno-side-effects
11572 Do not emit instructions with side-effects in addressing modes other than
11575 @item -mstack-align
11576 @itemx -mno-stack-align
11577 @itemx -mdata-align
11578 @itemx -mno-data-align
11579 @itemx -mconst-align
11580 @itemx -mno-const-align
11581 @opindex mstack-align
11582 @opindex mno-stack-align
11583 @opindex mdata-align
11584 @opindex mno-data-align
11585 @opindex mconst-align
11586 @opindex mno-const-align
11587 These options (no-options) arranges (eliminate arrangements) for the
11588 stack-frame, individual data and constants to be aligned for the maximum
11589 single data access size for the chosen CPU model. The default is to
11590 arrange for 32-bit alignment. ABI details such as structure layout are
11591 not affected by these options.
11599 Similar to the stack- data- and const-align options above, these options
11600 arrange for stack-frame, writable data and constants to all be 32-bit,
11601 16-bit or 8-bit aligned. The default is 32-bit alignment.
11603 @item -mno-prologue-epilogue
11604 @itemx -mprologue-epilogue
11605 @opindex mno-prologue-epilogue
11606 @opindex mprologue-epilogue
11607 With @option{-mno-prologue-epilogue}, the normal function prologue and
11608 epilogue that sets up the stack-frame are omitted and no return
11609 instructions or return sequences are generated in the code. Use this
11610 option only together with visual inspection of the compiled code: no
11611 warnings or errors are generated when call-saved registers must be saved,
11612 or storage for local variable needs to be allocated.
11616 @opindex mno-gotplt
11618 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11619 instruction sequences that load addresses for functions from the PLT part
11620 of the GOT rather than (traditional on other architectures) calls to the
11621 PLT@. The default is @option{-mgotplt}.
11625 Legacy no-op option only recognized with the cris-axis-elf and
11626 cris-axis-linux-gnu targets.
11630 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11634 This option, recognized for the cris-axis-elf arranges
11635 to link with input-output functions from a simulator library. Code,
11636 initialized data and zero-initialized data are allocated consecutively.
11640 Like @option{-sim}, but pass linker options to locate initialized data at
11641 0x40000000 and zero-initialized data at 0x80000000.
11645 @subsection CR16 Options
11646 @cindex CR16 Options
11648 These options are defined specifically for the CR16 ports.
11654 Enable the use of multiply-accumulate instructions. Disabled by default.
11658 @opindex mcr16cplus
11660 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
11665 Links the library libsim.a which is in compatible with simulator. Applicable
11666 to elf compiler only.
11670 Choose integer type as 32-bit wide.
11674 Generates sbit/cbit instructions for bit manipulations.
11676 @item -mdata-model=@var{model}
11677 @opindex mdata-model
11678 Choose a data model. The choices for @var{model} are @samp{near},
11679 @samp{far} or @samp{medium}. @samp{medium} is default.
11680 However, @samp{far} is not valid when -mcr16c option is chosen as
11681 CR16C architecture does not support far data model.
11684 @node Darwin Options
11685 @subsection Darwin Options
11686 @cindex Darwin options
11688 These options are defined for all architectures running the Darwin operating
11691 FSF GCC on Darwin does not create ``fat'' object files; it will create
11692 an object file for the single architecture that it was built to
11693 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11694 @option{-arch} options are used; it does so by running the compiler or
11695 linker multiple times and joining the results together with
11698 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11699 @samp{i686}) is determined by the flags that specify the ISA
11700 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11701 @option{-force_cpusubtype_ALL} option can be used to override this.
11703 The Darwin tools vary in their behavior when presented with an ISA
11704 mismatch. The assembler, @file{as}, will only permit instructions to
11705 be used that are valid for the subtype of the file it is generating,
11706 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11707 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11708 and print an error if asked to create a shared library with a less
11709 restrictive subtype than its input files (for instance, trying to put
11710 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11711 for executables, @file{ld}, will quietly give the executable the most
11712 restrictive subtype of any of its input files.
11717 Add the framework directory @var{dir} to the head of the list of
11718 directories to be searched for header files. These directories are
11719 interleaved with those specified by @option{-I} options and are
11720 scanned in a left-to-right order.
11722 A framework directory is a directory with frameworks in it. A
11723 framework is a directory with a @samp{"Headers"} and/or
11724 @samp{"PrivateHeaders"} directory contained directly in it that ends
11725 in @samp{".framework"}. The name of a framework is the name of this
11726 directory excluding the @samp{".framework"}. Headers associated with
11727 the framework are found in one of those two directories, with
11728 @samp{"Headers"} being searched first. A subframework is a framework
11729 directory that is in a framework's @samp{"Frameworks"} directory.
11730 Includes of subframework headers can only appear in a header of a
11731 framework that contains the subframework, or in a sibling subframework
11732 header. Two subframeworks are siblings if they occur in the same
11733 framework. A subframework should not have the same name as a
11734 framework, a warning will be issued if this is violated. Currently a
11735 subframework cannot have subframeworks, in the future, the mechanism
11736 may be extended to support this. The standard frameworks can be found
11737 in @samp{"/System/Library/Frameworks"} and
11738 @samp{"/Library/Frameworks"}. An example include looks like
11739 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11740 the name of the framework and header.h is found in the
11741 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11743 @item -iframework@var{dir}
11744 @opindex iframework
11745 Like @option{-F} except the directory is a treated as a system
11746 directory. The main difference between this @option{-iframework} and
11747 @option{-F} is that with @option{-iframework} the compiler does not
11748 warn about constructs contained within header files found via
11749 @var{dir}. This option is valid only for the C family of languages.
11753 Emit debugging information for symbols that are used. For STABS
11754 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11755 This is by default ON@.
11759 Emit debugging information for all symbols and types.
11761 @item -mmacosx-version-min=@var{version}
11762 The earliest version of MacOS X that this executable will run on
11763 is @var{version}. Typical values of @var{version} include @code{10.1},
11764 @code{10.2}, and @code{10.3.9}.
11766 If the compiler was built to use the system's headers by default,
11767 then the default for this option is the system version on which the
11768 compiler is running, otherwise the default is to make choices which
11769 are compatible with as many systems and code bases as possible.
11773 Enable kernel development mode. The @option{-mkernel} option sets
11774 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11775 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11776 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11777 applicable. This mode also sets @option{-mno-altivec},
11778 @option{-msoft-float}, @option{-fno-builtin} and
11779 @option{-mlong-branch} for PowerPC targets.
11781 @item -mone-byte-bool
11782 @opindex mone-byte-bool
11783 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11784 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11785 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11786 option has no effect on x86.
11788 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11789 to generate code that is not binary compatible with code generated
11790 without that switch. Using this switch may require recompiling all
11791 other modules in a program, including system libraries. Use this
11792 switch to conform to a non-default data model.
11794 @item -mfix-and-continue
11795 @itemx -ffix-and-continue
11796 @itemx -findirect-data
11797 @opindex mfix-and-continue
11798 @opindex ffix-and-continue
11799 @opindex findirect-data
11800 Generate code suitable for fast turn around development. Needed to
11801 enable gdb to dynamically load @code{.o} files into already running
11802 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11803 are provided for backwards compatibility.
11807 Loads all members of static archive libraries.
11808 See man ld(1) for more information.
11810 @item -arch_errors_fatal
11811 @opindex arch_errors_fatal
11812 Cause the errors having to do with files that have the wrong architecture
11815 @item -bind_at_load
11816 @opindex bind_at_load
11817 Causes the output file to be marked such that the dynamic linker will
11818 bind all undefined references when the file is loaded or launched.
11822 Produce a Mach-o bundle format file.
11823 See man ld(1) for more information.
11825 @item -bundle_loader @var{executable}
11826 @opindex bundle_loader
11827 This option specifies the @var{executable} that will be loading the build
11828 output file being linked. See man ld(1) for more information.
11831 @opindex dynamiclib
11832 When passed this option, GCC will produce a dynamic library instead of
11833 an executable when linking, using the Darwin @file{libtool} command.
11835 @item -force_cpusubtype_ALL
11836 @opindex force_cpusubtype_ALL
11837 This causes GCC's output file to have the @var{ALL} subtype, instead of
11838 one controlled by the @option{-mcpu} or @option{-march} option.
11840 @item -allowable_client @var{client_name}
11841 @itemx -client_name
11842 @itemx -compatibility_version
11843 @itemx -current_version
11845 @itemx -dependency-file
11847 @itemx -dylinker_install_name
11849 @itemx -exported_symbols_list
11852 @itemx -flat_namespace
11853 @itemx -force_flat_namespace
11854 @itemx -headerpad_max_install_names
11857 @itemx -install_name
11858 @itemx -keep_private_externs
11859 @itemx -multi_module
11860 @itemx -multiply_defined
11861 @itemx -multiply_defined_unused
11864 @itemx -no_dead_strip_inits_and_terms
11865 @itemx -nofixprebinding
11866 @itemx -nomultidefs
11868 @itemx -noseglinkedit
11869 @itemx -pagezero_size
11871 @itemx -prebind_all_twolevel_modules
11872 @itemx -private_bundle
11874 @itemx -read_only_relocs
11876 @itemx -sectobjectsymbols
11880 @itemx -sectobjectsymbols
11883 @itemx -segs_read_only_addr
11885 @itemx -segs_read_write_addr
11886 @itemx -seg_addr_table
11887 @itemx -seg_addr_table_filename
11888 @itemx -seglinkedit
11890 @itemx -segs_read_only_addr
11891 @itemx -segs_read_write_addr
11892 @itemx -single_module
11894 @itemx -sub_library
11896 @itemx -sub_umbrella
11897 @itemx -twolevel_namespace
11900 @itemx -unexported_symbols_list
11901 @itemx -weak_reference_mismatches
11902 @itemx -whatsloaded
11903 @opindex allowable_client
11904 @opindex client_name
11905 @opindex compatibility_version
11906 @opindex current_version
11907 @opindex dead_strip
11908 @opindex dependency-file
11909 @opindex dylib_file
11910 @opindex dylinker_install_name
11912 @opindex exported_symbols_list
11914 @opindex flat_namespace
11915 @opindex force_flat_namespace
11916 @opindex headerpad_max_install_names
11917 @opindex image_base
11919 @opindex install_name
11920 @opindex keep_private_externs
11921 @opindex multi_module
11922 @opindex multiply_defined
11923 @opindex multiply_defined_unused
11924 @opindex noall_load
11925 @opindex no_dead_strip_inits_and_terms
11926 @opindex nofixprebinding
11927 @opindex nomultidefs
11929 @opindex noseglinkedit
11930 @opindex pagezero_size
11932 @opindex prebind_all_twolevel_modules
11933 @opindex private_bundle
11934 @opindex read_only_relocs
11936 @opindex sectobjectsymbols
11939 @opindex sectcreate
11940 @opindex sectobjectsymbols
11943 @opindex segs_read_only_addr
11944 @opindex segs_read_write_addr
11945 @opindex seg_addr_table
11946 @opindex seg_addr_table_filename
11947 @opindex seglinkedit
11949 @opindex segs_read_only_addr
11950 @opindex segs_read_write_addr
11951 @opindex single_module
11953 @opindex sub_library
11954 @opindex sub_umbrella
11955 @opindex twolevel_namespace
11958 @opindex unexported_symbols_list
11959 @opindex weak_reference_mismatches
11960 @opindex whatsloaded
11961 These options are passed to the Darwin linker. The Darwin linker man page
11962 describes them in detail.
11965 @node DEC Alpha Options
11966 @subsection DEC Alpha Options
11968 These @samp{-m} options are defined for the DEC Alpha implementations:
11971 @item -mno-soft-float
11972 @itemx -msoft-float
11973 @opindex mno-soft-float
11974 @opindex msoft-float
11975 Use (do not use) the hardware floating-point instructions for
11976 floating-point operations. When @option{-msoft-float} is specified,
11977 functions in @file{libgcc.a} will be used to perform floating-point
11978 operations. Unless they are replaced by routines that emulate the
11979 floating-point operations, or compiled in such a way as to call such
11980 emulations routines, these routines will issue floating-point
11981 operations. If you are compiling for an Alpha without floating-point
11982 operations, you must ensure that the library is built so as not to call
11985 Note that Alpha implementations without floating-point operations are
11986 required to have floating-point registers.
11989 @itemx -mno-fp-regs
11991 @opindex mno-fp-regs
11992 Generate code that uses (does not use) the floating-point register set.
11993 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11994 register set is not used, floating-point operands are passed in integer
11995 registers as if they were integers and floating-point results are passed
11996 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11997 so any function with a floating-point argument or return value called by code
11998 compiled with @option{-mno-fp-regs} must also be compiled with that
12001 A typical use of this option is building a kernel that does not use,
12002 and hence need not save and restore, any floating-point registers.
12006 The Alpha architecture implements floating-point hardware optimized for
12007 maximum performance. It is mostly compliant with the IEEE floating-point
12008 standard. However, for full compliance, software assistance is
12009 required. This option generates code fully IEEE-compliant code
12010 @emph{except} that the @var{inexact-flag} is not maintained (see below).
12011 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
12012 defined during compilation. The resulting code is less efficient but is
12013 able to correctly support denormalized numbers and exceptional IEEE
12014 values such as not-a-number and plus/minus infinity. Other Alpha
12015 compilers call this option @option{-ieee_with_no_inexact}.
12017 @item -mieee-with-inexact
12018 @opindex mieee-with-inexact
12019 This is like @option{-mieee} except the generated code also maintains
12020 the IEEE @var{inexact-flag}. Turning on this option causes the
12021 generated code to implement fully-compliant IEEE math. In addition to
12022 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
12023 macro. On some Alpha implementations the resulting code may execute
12024 significantly slower than the code generated by default. Since there is
12025 very little code that depends on the @var{inexact-flag}, you should
12026 normally not specify this option. Other Alpha compilers call this
12027 option @option{-ieee_with_inexact}.
12029 @item -mfp-trap-mode=@var{trap-mode}
12030 @opindex mfp-trap-mode
12031 This option controls what floating-point related traps are enabled.
12032 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
12033 The trap mode can be set to one of four values:
12037 This is the default (normal) setting. The only traps that are enabled
12038 are the ones that cannot be disabled in software (e.g., division by zero
12042 In addition to the traps enabled by @samp{n}, underflow traps are enabled
12046 Like @samp{u}, but the instructions are marked to be safe for software
12047 completion (see Alpha architecture manual for details).
12050 Like @samp{su}, but inexact traps are enabled as well.
12053 @item -mfp-rounding-mode=@var{rounding-mode}
12054 @opindex mfp-rounding-mode
12055 Selects the IEEE rounding mode. Other Alpha compilers call this option
12056 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
12061 Normal IEEE rounding mode. Floating-point numbers are rounded towards
12062 the nearest machine number or towards the even machine number in case
12066 Round towards minus infinity.
12069 Chopped rounding mode. Floating-point numbers are rounded towards zero.
12072 Dynamic rounding mode. A field in the floating-point control register
12073 (@var{fpcr}, see Alpha architecture reference manual) controls the
12074 rounding mode in effect. The C library initializes this register for
12075 rounding towards plus infinity. Thus, unless your program modifies the
12076 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
12079 @item -mtrap-precision=@var{trap-precision}
12080 @opindex mtrap-precision
12081 In the Alpha architecture, floating-point traps are imprecise. This
12082 means without software assistance it is impossible to recover from a
12083 floating trap and program execution normally needs to be terminated.
12084 GCC can generate code that can assist operating system trap handlers
12085 in determining the exact location that caused a floating-point trap.
12086 Depending on the requirements of an application, different levels of
12087 precisions can be selected:
12091 Program precision. This option is the default and means a trap handler
12092 can only identify which program caused a floating-point exception.
12095 Function precision. The trap handler can determine the function that
12096 caused a floating-point exception.
12099 Instruction precision. The trap handler can determine the exact
12100 instruction that caused a floating-point exception.
12103 Other Alpha compilers provide the equivalent options called
12104 @option{-scope_safe} and @option{-resumption_safe}.
12106 @item -mieee-conformant
12107 @opindex mieee-conformant
12108 This option marks the generated code as IEEE conformant. You must not
12109 use this option unless you also specify @option{-mtrap-precision=i} and either
12110 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
12111 is to emit the line @samp{.eflag 48} in the function prologue of the
12112 generated assembly file. Under DEC Unix, this has the effect that
12113 IEEE-conformant math library routines will be linked in.
12115 @item -mbuild-constants
12116 @opindex mbuild-constants
12117 Normally GCC examines a 32- or 64-bit integer constant to
12118 see if it can construct it from smaller constants in two or three
12119 instructions. If it cannot, it will output the constant as a literal and
12120 generate code to load it from the data segment at run time.
12122 Use this option to require GCC to construct @emph{all} integer constants
12123 using code, even if it takes more instructions (the maximum is six).
12125 You would typically use this option to build a shared library dynamic
12126 loader. Itself a shared library, it must relocate itself in memory
12127 before it can find the variables and constants in its own data segment.
12133 Select whether to generate code to be assembled by the vendor-supplied
12134 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
12152 Indicate whether GCC should generate code to use the optional BWX,
12153 CIX, FIX and MAX instruction sets. The default is to use the instruction
12154 sets supported by the CPU type specified via @option{-mcpu=} option or that
12155 of the CPU on which GCC was built if none was specified.
12158 @itemx -mfloat-ieee
12159 @opindex mfloat-vax
12160 @opindex mfloat-ieee
12161 Generate code that uses (does not use) VAX F and G floating-point
12162 arithmetic instead of IEEE single and double precision.
12164 @item -mexplicit-relocs
12165 @itemx -mno-explicit-relocs
12166 @opindex mexplicit-relocs
12167 @opindex mno-explicit-relocs
12168 Older Alpha assemblers provided no way to generate symbol relocations
12169 except via assembler macros. Use of these macros does not allow
12170 optimal instruction scheduling. GNU binutils as of version 2.12
12171 supports a new syntax that allows the compiler to explicitly mark
12172 which relocations should apply to which instructions. This option
12173 is mostly useful for debugging, as GCC detects the capabilities of
12174 the assembler when it is built and sets the default accordingly.
12177 @itemx -mlarge-data
12178 @opindex msmall-data
12179 @opindex mlarge-data
12180 When @option{-mexplicit-relocs} is in effect, static data is
12181 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
12182 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
12183 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
12184 16-bit relocations off of the @code{$gp} register. This limits the
12185 size of the small data area to 64KB, but allows the variables to be
12186 directly accessed via a single instruction.
12188 The default is @option{-mlarge-data}. With this option the data area
12189 is limited to just below 2GB@. Programs that require more than 2GB of
12190 data must use @code{malloc} or @code{mmap} to allocate the data in the
12191 heap instead of in the program's data segment.
12193 When generating code for shared libraries, @option{-fpic} implies
12194 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
12197 @itemx -mlarge-text
12198 @opindex msmall-text
12199 @opindex mlarge-text
12200 When @option{-msmall-text} is used, the compiler assumes that the
12201 code of the entire program (or shared library) fits in 4MB, and is
12202 thus reachable with a branch instruction. When @option{-msmall-data}
12203 is used, the compiler can assume that all local symbols share the
12204 same @code{$gp} value, and thus reduce the number of instructions
12205 required for a function call from 4 to 1.
12207 The default is @option{-mlarge-text}.
12209 @item -mcpu=@var{cpu_type}
12211 Set the instruction set and instruction scheduling parameters for
12212 machine type @var{cpu_type}. You can specify either the @samp{EV}
12213 style name or the corresponding chip number. GCC supports scheduling
12214 parameters for the EV4, EV5 and EV6 family of processors and will
12215 choose the default values for the instruction set from the processor
12216 you specify. If you do not specify a processor type, GCC will default
12217 to the processor on which the compiler was built.
12219 Supported values for @var{cpu_type} are
12225 Schedules as an EV4 and has no instruction set extensions.
12229 Schedules as an EV5 and has no instruction set extensions.
12233 Schedules as an EV5 and supports the BWX extension.
12238 Schedules as an EV5 and supports the BWX and MAX extensions.
12242 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
12246 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
12249 Native toolchains also support the value @samp{native},
12250 which selects the best architecture option for the host processor.
12251 @option{-mcpu=native} has no effect if GCC does not recognize
12254 @item -mtune=@var{cpu_type}
12256 Set only the instruction scheduling parameters for machine type
12257 @var{cpu_type}. The instruction set is not changed.
12259 Native toolchains also support the value @samp{native},
12260 which selects the best architecture option for the host processor.
12261 @option{-mtune=native} has no effect if GCC does not recognize
12264 @item -mmemory-latency=@var{time}
12265 @opindex mmemory-latency
12266 Sets the latency the scheduler should assume for typical memory
12267 references as seen by the application. This number is highly
12268 dependent on the memory access patterns used by the application
12269 and the size of the external cache on the machine.
12271 Valid options for @var{time} are
12275 A decimal number representing clock cycles.
12281 The compiler contains estimates of the number of clock cycles for
12282 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
12283 (also called Dcache, Scache, and Bcache), as well as to main memory.
12284 Note that L3 is only valid for EV5.
12289 @node DEC Alpha/VMS Options
12290 @subsection DEC Alpha/VMS Options
12292 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
12295 @item -mvms-return-codes
12296 @opindex mvms-return-codes
12297 Return VMS condition codes from main. The default is to return POSIX
12298 style condition (e.g.@: error) codes.
12300 @item -mdebug-main=@var{prefix}
12301 @opindex mdebug-main=@var{prefix}
12302 Flag the first routine whose name starts with @var{prefix} as the main
12303 routine for the debugger.
12307 Default to 64bit memory allocation routines.
12311 @subsection FR30 Options
12312 @cindex FR30 Options
12314 These options are defined specifically for the FR30 port.
12318 @item -msmall-model
12319 @opindex msmall-model
12320 Use the small address space model. This can produce smaller code, but
12321 it does assume that all symbolic values and addresses will fit into a
12326 Assume that runtime support has been provided and so there is no need
12327 to include the simulator library (@file{libsim.a}) on the linker
12333 @subsection FRV Options
12334 @cindex FRV Options
12340 Only use the first 32 general-purpose registers.
12345 Use all 64 general-purpose registers.
12350 Use only the first 32 floating-point registers.
12355 Use all 64 floating-point registers.
12358 @opindex mhard-float
12360 Use hardware instructions for floating-point operations.
12363 @opindex msoft-float
12365 Use library routines for floating-point operations.
12370 Dynamically allocate condition code registers.
12375 Do not try to dynamically allocate condition code registers, only
12376 use @code{icc0} and @code{fcc0}.
12381 Change ABI to use double word insns.
12386 Do not use double word instructions.
12391 Use floating-point double instructions.
12394 @opindex mno-double
12396 Do not use floating-point double instructions.
12401 Use media instructions.
12406 Do not use media instructions.
12411 Use multiply and add/subtract instructions.
12414 @opindex mno-muladd
12416 Do not use multiply and add/subtract instructions.
12421 Select the FDPIC ABI, that uses function descriptors to represent
12422 pointers to functions. Without any PIC/PIE-related options, it
12423 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
12424 assumes GOT entries and small data are within a 12-bit range from the
12425 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
12426 are computed with 32 bits.
12427 With a @samp{bfin-elf} target, this option implies @option{-msim}.
12430 @opindex minline-plt
12432 Enable inlining of PLT entries in function calls to functions that are
12433 not known to bind locally. It has no effect without @option{-mfdpic}.
12434 It's enabled by default if optimizing for speed and compiling for
12435 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
12436 optimization option such as @option{-O3} or above is present in the
12442 Assume a large TLS segment when generating thread-local code.
12447 Do not assume a large TLS segment when generating thread-local code.
12452 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
12453 that is known to be in read-only sections. It's enabled by default,
12454 except for @option{-fpic} or @option{-fpie}: even though it may help
12455 make the global offset table smaller, it trades 1 instruction for 4.
12456 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
12457 one of which may be shared by multiple symbols, and it avoids the need
12458 for a GOT entry for the referenced symbol, so it's more likely to be a
12459 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
12461 @item -multilib-library-pic
12462 @opindex multilib-library-pic
12464 Link with the (library, not FD) pic libraries. It's implied by
12465 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
12466 @option{-fpic} without @option{-mfdpic}. You should never have to use
12470 @opindex mlinked-fp
12472 Follow the EABI requirement of always creating a frame pointer whenever
12473 a stack frame is allocated. This option is enabled by default and can
12474 be disabled with @option{-mno-linked-fp}.
12477 @opindex mlong-calls
12479 Use indirect addressing to call functions outside the current
12480 compilation unit. This allows the functions to be placed anywhere
12481 within the 32-bit address space.
12483 @item -malign-labels
12484 @opindex malign-labels
12486 Try to align labels to an 8-byte boundary by inserting nops into the
12487 previous packet. This option only has an effect when VLIW packing
12488 is enabled. It doesn't create new packets; it merely adds nops to
12491 @item -mlibrary-pic
12492 @opindex mlibrary-pic
12494 Generate position-independent EABI code.
12499 Use only the first four media accumulator registers.
12504 Use all eight media accumulator registers.
12509 Pack VLIW instructions.
12514 Do not pack VLIW instructions.
12517 @opindex mno-eflags
12519 Do not mark ABI switches in e_flags.
12522 @opindex mcond-move
12524 Enable the use of conditional-move instructions (default).
12526 This switch is mainly for debugging the compiler and will likely be removed
12527 in a future version.
12529 @item -mno-cond-move
12530 @opindex mno-cond-move
12532 Disable the use of conditional-move instructions.
12534 This switch is mainly for debugging the compiler and will likely be removed
12535 in a future version.
12540 Enable the use of conditional set instructions (default).
12542 This switch is mainly for debugging the compiler and will likely be removed
12543 in a future version.
12548 Disable the use of conditional set instructions.
12550 This switch is mainly for debugging the compiler and will likely be removed
12551 in a future version.
12554 @opindex mcond-exec
12556 Enable the use of conditional execution (default).
12558 This switch is mainly for debugging the compiler and will likely be removed
12559 in a future version.
12561 @item -mno-cond-exec
12562 @opindex mno-cond-exec
12564 Disable the use of conditional execution.
12566 This switch is mainly for debugging the compiler and will likely be removed
12567 in a future version.
12569 @item -mvliw-branch
12570 @opindex mvliw-branch
12572 Run a pass to pack branches into VLIW instructions (default).
12574 This switch is mainly for debugging the compiler and will likely be removed
12575 in a future version.
12577 @item -mno-vliw-branch
12578 @opindex mno-vliw-branch
12580 Do not run a pass to pack branches into VLIW instructions.
12582 This switch is mainly for debugging the compiler and will likely be removed
12583 in a future version.
12585 @item -mmulti-cond-exec
12586 @opindex mmulti-cond-exec
12588 Enable optimization of @code{&&} and @code{||} in conditional execution
12591 This switch is mainly for debugging the compiler and will likely be removed
12592 in a future version.
12594 @item -mno-multi-cond-exec
12595 @opindex mno-multi-cond-exec
12597 Disable optimization of @code{&&} and @code{||} in conditional execution.
12599 This switch is mainly for debugging the compiler and will likely be removed
12600 in a future version.
12602 @item -mnested-cond-exec
12603 @opindex mnested-cond-exec
12605 Enable nested conditional execution optimizations (default).
12607 This switch is mainly for debugging the compiler and will likely be removed
12608 in a future version.
12610 @item -mno-nested-cond-exec
12611 @opindex mno-nested-cond-exec
12613 Disable nested conditional execution optimizations.
12615 This switch is mainly for debugging the compiler and will likely be removed
12616 in a future version.
12618 @item -moptimize-membar
12619 @opindex moptimize-membar
12621 This switch removes redundant @code{membar} instructions from the
12622 compiler generated code. It is enabled by default.
12624 @item -mno-optimize-membar
12625 @opindex mno-optimize-membar
12627 This switch disables the automatic removal of redundant @code{membar}
12628 instructions from the generated code.
12630 @item -mtomcat-stats
12631 @opindex mtomcat-stats
12633 Cause gas to print out tomcat statistics.
12635 @item -mcpu=@var{cpu}
12638 Select the processor type for which to generate code. Possible values are
12639 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12640 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12644 @node GNU/Linux Options
12645 @subsection GNU/Linux Options
12647 These @samp{-m} options are defined for GNU/Linux targets:
12652 Use the GNU C library. This is the default except
12653 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12657 Use uClibc C library. This is the default on
12658 @samp{*-*-linux-*uclibc*} targets.
12662 Use Bionic C library. This is the default on
12663 @samp{*-*-linux-*android*} targets.
12667 Compile code compatible with Android platform. This is the default on
12668 @samp{*-*-linux-*android*} targets.
12670 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12671 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12672 this option makes the GCC driver pass Android-specific options to the linker.
12673 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12676 @item -tno-android-cc
12677 @opindex tno-android-cc
12678 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12679 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12680 @option{-fno-rtti} by default.
12682 @item -tno-android-ld
12683 @opindex tno-android-ld
12684 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12685 linking options to the linker.
12689 @node H8/300 Options
12690 @subsection H8/300 Options
12692 These @samp{-m} options are defined for the H8/300 implementations:
12697 Shorten some address references at link time, when possible; uses the
12698 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12699 ld, Using ld}, for a fuller description.
12703 Generate code for the H8/300H@.
12707 Generate code for the H8S@.
12711 Generate code for the H8S and H8/300H in the normal mode. This switch
12712 must be used either with @option{-mh} or @option{-ms}.
12716 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12720 Make @code{int} data 32 bits by default.
12723 @opindex malign-300
12724 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12725 The default for the H8/300H and H8S is to align longs and floats on 4
12727 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12728 This option has no effect on the H8/300.
12732 @subsection HPPA Options
12733 @cindex HPPA Options
12735 These @samp{-m} options are defined for the HPPA family of computers:
12738 @item -march=@var{architecture-type}
12740 Generate code for the specified architecture. The choices for
12741 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12742 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12743 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12744 architecture option for your machine. Code compiled for lower numbered
12745 architectures will run on higher numbered architectures, but not the
12748 @item -mpa-risc-1-0
12749 @itemx -mpa-risc-1-1
12750 @itemx -mpa-risc-2-0
12751 @opindex mpa-risc-1-0
12752 @opindex mpa-risc-1-1
12753 @opindex mpa-risc-2-0
12754 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12757 @opindex mbig-switch
12758 Generate code suitable for big switch tables. Use this option only if
12759 the assembler/linker complain about out of range branches within a switch
12762 @item -mjump-in-delay
12763 @opindex mjump-in-delay
12764 Fill delay slots of function calls with unconditional jump instructions
12765 by modifying the return pointer for the function call to be the target
12766 of the conditional jump.
12768 @item -mdisable-fpregs
12769 @opindex mdisable-fpregs
12770 Prevent floating-point registers from being used in any manner. This is
12771 necessary for compiling kernels that perform lazy context switching of
12772 floating-point registers. If you use this option and attempt to perform
12773 floating-point operations, the compiler aborts.
12775 @item -mdisable-indexing
12776 @opindex mdisable-indexing
12777 Prevent the compiler from using indexing address modes. This avoids some
12778 rather obscure problems when compiling MIG generated code under MACH@.
12780 @item -mno-space-regs
12781 @opindex mno-space-regs
12782 Generate code that assumes the target has no space registers. This allows
12783 GCC to generate faster indirect calls and use unscaled index address modes.
12785 Such code is suitable for level 0 PA systems and kernels.
12787 @item -mfast-indirect-calls
12788 @opindex mfast-indirect-calls
12789 Generate code that assumes calls never cross space boundaries. This
12790 allows GCC to emit code which performs faster indirect calls.
12792 This option will not work in the presence of shared libraries or nested
12795 @item -mfixed-range=@var{register-range}
12796 @opindex mfixed-range
12797 Generate code treating the given register range as fixed registers.
12798 A fixed register is one that the register allocator can not use. This is
12799 useful when compiling kernel code. A register range is specified as
12800 two registers separated by a dash. Multiple register ranges can be
12801 specified separated by a comma.
12803 @item -mlong-load-store
12804 @opindex mlong-load-store
12805 Generate 3-instruction load and store sequences as sometimes required by
12806 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12809 @item -mportable-runtime
12810 @opindex mportable-runtime
12811 Use the portable calling conventions proposed by HP for ELF systems.
12815 Enable the use of assembler directives only GAS understands.
12817 @item -mschedule=@var{cpu-type}
12819 Schedule code according to the constraints for the machine type
12820 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12821 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12822 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12823 proper scheduling option for your machine. The default scheduling is
12827 @opindex mlinker-opt
12828 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12829 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12830 linkers in which they give bogus error messages when linking some programs.
12833 @opindex msoft-float
12834 Generate output containing library calls for floating point.
12835 @strong{Warning:} the requisite libraries are not available for all HPPA
12836 targets. Normally the facilities of the machine's usual C compiler are
12837 used, but this cannot be done directly in cross-compilation. You must make
12838 your own arrangements to provide suitable library functions for
12841 @option{-msoft-float} changes the calling convention in the output file;
12842 therefore, it is only useful if you compile @emph{all} of a program with
12843 this option. In particular, you need to compile @file{libgcc.a}, the
12844 library that comes with GCC, with @option{-msoft-float} in order for
12849 Generate the predefine, @code{_SIO}, for server IO@. The default is
12850 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12851 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12852 options are available under HP-UX and HI-UX@.
12856 Use GNU ld specific options. This passes @option{-shared} to ld when
12857 building a shared library. It is the default when GCC is configured,
12858 explicitly or implicitly, with the GNU linker. This option does not
12859 have any affect on which ld is called, it only changes what parameters
12860 are passed to that ld. The ld that is called is determined by the
12861 @option{--with-ld} configure option, GCC's program search path, and
12862 finally by the user's @env{PATH}. The linker used by GCC can be printed
12863 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12864 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12868 Use HP ld specific options. This passes @option{-b} to ld when building
12869 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12870 links. It is the default when GCC is configured, explicitly or
12871 implicitly, with the HP linker. This option does not have any affect on
12872 which ld is called, it only changes what parameters are passed to that
12873 ld. The ld that is called is determined by the @option{--with-ld}
12874 configure option, GCC's program search path, and finally by the user's
12875 @env{PATH}. The linker used by GCC can be printed using @samp{which
12876 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12877 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12880 @opindex mno-long-calls
12881 Generate code that uses long call sequences. This ensures that a call
12882 is always able to reach linker generated stubs. The default is to generate
12883 long calls only when the distance from the call site to the beginning
12884 of the function or translation unit, as the case may be, exceeds a
12885 predefined limit set by the branch type being used. The limits for
12886 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12887 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12890 Distances are measured from the beginning of functions when using the
12891 @option{-ffunction-sections} option, or when using the @option{-mgas}
12892 and @option{-mno-portable-runtime} options together under HP-UX with
12895 It is normally not desirable to use this option as it will degrade
12896 performance. However, it may be useful in large applications,
12897 particularly when partial linking is used to build the application.
12899 The types of long calls used depends on the capabilities of the
12900 assembler and linker, and the type of code being generated. The
12901 impact on systems that support long absolute calls, and long pic
12902 symbol-difference or pc-relative calls should be relatively small.
12903 However, an indirect call is used on 32-bit ELF systems in pic code
12904 and it is quite long.
12906 @item -munix=@var{unix-std}
12908 Generate compiler predefines and select a startfile for the specified
12909 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12910 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12911 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12912 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12913 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12916 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12917 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12918 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12919 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12920 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12921 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12923 It is @emph{important} to note that this option changes the interfaces
12924 for various library routines. It also affects the operational behavior
12925 of the C library. Thus, @emph{extreme} care is needed in using this
12928 Library code that is intended to operate with more than one UNIX
12929 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12930 as appropriate. Most GNU software doesn't provide this capability.
12934 Suppress the generation of link options to search libdld.sl when the
12935 @option{-static} option is specified on HP-UX 10 and later.
12939 The HP-UX implementation of setlocale in libc has a dependency on
12940 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12941 when the @option{-static} option is specified, special link options
12942 are needed to resolve this dependency.
12944 On HP-UX 10 and later, the GCC driver adds the necessary options to
12945 link with libdld.sl when the @option{-static} option is specified.
12946 This causes the resulting binary to be dynamic. On the 64-bit port,
12947 the linkers generate dynamic binaries by default in any case. The
12948 @option{-nolibdld} option can be used to prevent the GCC driver from
12949 adding these link options.
12953 Add support for multithreading with the @dfn{dce thread} library
12954 under HP-UX@. This option sets flags for both the preprocessor and
12958 @node i386 and x86-64 Options
12959 @subsection Intel 386 and AMD x86-64 Options
12960 @cindex i386 Options
12961 @cindex x86-64 Options
12962 @cindex Intel 386 Options
12963 @cindex AMD x86-64 Options
12965 These @samp{-m} options are defined for the i386 and x86-64 family of
12969 @item -mtune=@var{cpu-type}
12971 Tune to @var{cpu-type} everything applicable about the generated code, except
12972 for the ABI and the set of available instructions. The choices for
12973 @var{cpu-type} are:
12976 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12977 If you know the CPU on which your code will run, then you should use
12978 the corresponding @option{-mtune} option instead of
12979 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12980 of your application will have, then you should use this option.
12982 As new processors are deployed in the marketplace, the behavior of this
12983 option will change. Therefore, if you upgrade to a newer version of
12984 GCC, the code generated option will change to reflect the processors
12985 that were most common when that version of GCC was released.
12987 There is no @option{-march=generic} option because @option{-march}
12988 indicates the instruction set the compiler can use, and there is no
12989 generic instruction set applicable to all processors. In contrast,
12990 @option{-mtune} indicates the processor (or, in this case, collection of
12991 processors) for which the code is optimized.
12993 This selects the CPU to tune for at compilation time by determining
12994 the processor type of the compiling machine. Using @option{-mtune=native}
12995 will produce code optimized for the local machine under the constraints
12996 of the selected instruction set. Using @option{-march=native} will
12997 enable all instruction subsets supported by the local machine (hence
12998 the result might not run on different machines).
13000 Original Intel's i386 CPU@.
13002 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
13003 @item i586, pentium
13004 Intel Pentium CPU with no MMX support.
13006 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
13008 Intel PentiumPro CPU@.
13010 Same as @code{generic}, but when used as @code{march} option, PentiumPro
13011 instruction set will be used, so the code will run on all i686 family chips.
13013 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
13014 @item pentium3, pentium3m
13015 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
13018 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
13019 support. Used by Centrino notebooks.
13020 @item pentium4, pentium4m
13021 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
13023 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
13026 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
13027 SSE2 and SSE3 instruction set support.
13029 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13030 instruction set support.
13032 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
13033 and SSE4.2 instruction set support.
13035 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13036 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
13038 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13039 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
13042 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13043 instruction set support.
13045 AMD K6 CPU with MMX instruction set support.
13047 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
13048 @item athlon, athlon-tbird
13049 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
13051 @item athlon-4, athlon-xp, athlon-mp
13052 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
13053 instruction set support.
13054 @item k8, opteron, athlon64, athlon-fx
13055 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
13056 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
13057 @item k8-sse3, opteron-sse3, athlon64-sse3
13058 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
13059 @item amdfam10, barcelona
13060 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
13061 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
13062 instruction set extensions.)
13064 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
13065 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
13066 SSSE3, SSE4.1, SSE4.2, 3DNow!, enhanced 3DNow!, ABM and 64-bit
13067 instruction set extensions.)
13069 AMD Family 14h core based CPUs with x86-64 instruction set support. (This
13070 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
13071 instruction set extensions.)
13073 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
13076 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
13077 instruction set support.
13079 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
13080 implemented for this chip.)
13082 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
13083 implemented for this chip.)
13085 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
13088 While picking a specific @var{cpu-type} will schedule things appropriately
13089 for that particular chip, the compiler will not generate any code that
13090 does not run on the default machine type without the @option{-march=@var{cpu-type}}
13091 option being used. For example, if GCC is configured for i686-pc-linux-gnu
13092 then @option{-mtune=pentium4} will generate code that is tuned for Pentium4
13093 but will still run on i686 machines.
13095 @item -march=@var{cpu-type}
13097 Generate instructions for the machine type @var{cpu-type}. The choices
13098 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
13099 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
13101 @item -mcpu=@var{cpu-type}
13103 A deprecated synonym for @option{-mtune}.
13105 @item -mfpmath=@var{unit}
13107 Generate floating-point arithmetic for selected unit @var{unit}. The choices
13108 for @var{unit} are:
13112 Use the standard 387 floating-point coprocessor present on the majority of chips and
13113 emulated otherwise. Code compiled with this option runs almost everywhere.
13114 The temporary results are computed in 80-bit precision instead of the precision
13115 specified by the type, resulting in slightly different results compared to most
13116 of other chips. See @option{-ffloat-store} for more detailed description.
13118 This is the default choice for i386 compiler.
13121 Use scalar floating-point instructions present in the SSE instruction set.
13122 This instruction set is supported by Pentium3 and newer chips, in the AMD line
13123 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
13124 instruction set supports only single-precision arithmetic, thus the double and
13125 extended-precision arithmetic are still done using 387. A later version, present
13126 only in Pentium4 and the future AMD x86-64 chips, supports double-precision
13129 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
13130 or @option{-msse2} switches to enable SSE extensions and make this option
13131 effective. For the x86-64 compiler, these extensions are enabled by default.
13133 The resulting code should be considerably faster in the majority of cases and avoid
13134 the numerical instability problems of 387 code, but may break some existing
13135 code that expects temporaries to be 80bit.
13137 This is the default choice for the x86-64 compiler.
13142 Attempt to utilize both instruction sets at once. This effectively double the
13143 amount of available registers and on chips with separate execution units for
13144 387 and SSE the execution resources too. Use this option with care, as it is
13145 still experimental, because the GCC register allocator does not model separate
13146 functional units well resulting in instable performance.
13149 @item -masm=@var{dialect}
13150 @opindex masm=@var{dialect}
13151 Output asm instructions using selected @var{dialect}. Supported
13152 choices are @samp{intel} or @samp{att} (the default one). Darwin does
13153 not support @samp{intel}.
13156 @itemx -mno-ieee-fp
13158 @opindex mno-ieee-fp
13159 Control whether or not the compiler uses IEEE floating-point
13160 comparisons. These handle correctly the case where the result of a
13161 comparison is unordered.
13164 @opindex msoft-float
13165 Generate output containing library calls for floating point.
13166 @strong{Warning:} the requisite libraries are not part of GCC@.
13167 Normally the facilities of the machine's usual C compiler are used, but
13168 this can't be done directly in cross-compilation. You must make your
13169 own arrangements to provide suitable library functions for
13172 On machines where a function returns floating-point results in the 80387
13173 register stack, some floating-point opcodes may be emitted even if
13174 @option{-msoft-float} is used.
13176 @item -mno-fp-ret-in-387
13177 @opindex mno-fp-ret-in-387
13178 Do not use the FPU registers for return values of functions.
13180 The usual calling convention has functions return values of types
13181 @code{float} and @code{double} in an FPU register, even if there
13182 is no FPU@. The idea is that the operating system should emulate
13185 The option @option{-mno-fp-ret-in-387} causes such values to be returned
13186 in ordinary CPU registers instead.
13188 @item -mno-fancy-math-387
13189 @opindex mno-fancy-math-387
13190 Some 387 emulators do not support the @code{sin}, @code{cos} and
13191 @code{sqrt} instructions for the 387. Specify this option to avoid
13192 generating those instructions. This option is the default on FreeBSD,
13193 OpenBSD and NetBSD@. This option is overridden when @option{-march}
13194 indicates that the target CPU will always have an FPU and so the
13195 instruction will not need emulation. As of revision 2.6.1, these
13196 instructions are not generated unless you also use the
13197 @option{-funsafe-math-optimizations} switch.
13199 @item -malign-double
13200 @itemx -mno-align-double
13201 @opindex malign-double
13202 @opindex mno-align-double
13203 Control whether GCC aligns @code{double}, @code{long double}, and
13204 @code{long long} variables on a two word boundary or a one word
13205 boundary. Aligning @code{double} variables on a two word boundary will
13206 produce code that runs somewhat faster on a @samp{Pentium} at the
13207 expense of more memory.
13209 On x86-64, @option{-malign-double} is enabled by default.
13211 @strong{Warning:} if you use the @option{-malign-double} switch,
13212 structures containing the above types will be aligned differently than
13213 the published application binary interface specifications for the 386
13214 and will not be binary compatible with structures in code compiled
13215 without that switch.
13217 @item -m96bit-long-double
13218 @itemx -m128bit-long-double
13219 @opindex m96bit-long-double
13220 @opindex m128bit-long-double
13221 These switches control the size of @code{long double} type. The i386
13222 application binary interface specifies the size to be 96 bits,
13223 so @option{-m96bit-long-double} is the default in 32 bit mode.
13225 Modern architectures (Pentium and newer) would prefer @code{long double}
13226 to be aligned to an 8 or 16 byte boundary. In arrays or structures
13227 conforming to the ABI, this would not be possible. So specifying a
13228 @option{-m128bit-long-double} will align @code{long double}
13229 to a 16 byte boundary by padding the @code{long double} with an additional
13232 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
13233 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
13235 Notice that neither of these options enable any extra precision over the x87
13236 standard of 80 bits for a @code{long double}.
13238 @strong{Warning:} if you override the default value for your target ABI, the
13239 structures and arrays containing @code{long double} variables will change
13240 their size as well as function calling convention for function taking
13241 @code{long double} will be modified. Hence they will not be binary
13242 compatible with arrays or structures in code compiled without that switch.
13244 @item -mlarge-data-threshold=@var{number}
13245 @opindex mlarge-data-threshold=@var{number}
13246 When @option{-mcmodel=medium} is specified, the data greater than
13247 @var{threshold} are placed in large data section. This value must be the
13248 same across all object linked into the binary and defaults to 65535.
13252 Use a different function-calling convention, in which functions that
13253 take a fixed number of arguments return with the @code{ret} @var{num}
13254 instruction, which pops their arguments while returning. This saves one
13255 instruction in the caller since there is no need to pop the arguments
13258 You can specify that an individual function is called with this calling
13259 sequence with the function attribute @samp{stdcall}. You can also
13260 override the @option{-mrtd} option by using the function attribute
13261 @samp{cdecl}. @xref{Function Attributes}.
13263 @strong{Warning:} this calling convention is incompatible with the one
13264 normally used on Unix, so you cannot use it if you need to call
13265 libraries compiled with the Unix compiler.
13267 Also, you must provide function prototypes for all functions that
13268 take variable numbers of arguments (including @code{printf});
13269 otherwise incorrect code will be generated for calls to those
13272 In addition, seriously incorrect code will result if you call a
13273 function with too many arguments. (Normally, extra arguments are
13274 harmlessly ignored.)
13276 @item -mregparm=@var{num}
13278 Control how many registers are used to pass integer arguments. By
13279 default, no registers are used to pass arguments, and at most 3
13280 registers can be used. You can control this behavior for a specific
13281 function by using the function attribute @samp{regparm}.
13282 @xref{Function Attributes}.
13284 @strong{Warning:} if you use this switch, and
13285 @var{num} is nonzero, then you must build all modules with the same
13286 value, including any libraries. This includes the system libraries and
13290 @opindex msseregparm
13291 Use SSE register passing conventions for float and double arguments
13292 and return values. You can control this behavior for a specific
13293 function by using the function attribute @samp{sseregparm}.
13294 @xref{Function Attributes}.
13296 @strong{Warning:} if you use this switch then you must build all
13297 modules with the same value, including any libraries. This includes
13298 the system libraries and startup modules.
13300 @item -mvect8-ret-in-mem
13301 @opindex mvect8-ret-in-mem
13302 Return 8-byte vectors in memory instead of MMX registers. This is the
13303 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
13304 Studio compilers until version 12. Later compiler versions (starting
13305 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
13306 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
13307 you need to remain compatible with existing code produced by those
13308 previous compiler versions or older versions of GCC.
13317 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
13318 is specified, the significands of results of floating-point operations are
13319 rounded to 24 bits (single precision); @option{-mpc64} rounds the
13320 significands of results of floating-point operations to 53 bits (double
13321 precision) and @option{-mpc80} rounds the significands of results of
13322 floating-point operations to 64 bits (extended double precision), which is
13323 the default. When this option is used, floating-point operations in higher
13324 precisions are not available to the programmer without setting the FPU
13325 control word explicitly.
13327 Setting the rounding of floating-point operations to less than the default
13328 80 bits can speed some programs by 2% or more. Note that some mathematical
13329 libraries assume that extended-precision (80-bit) floating-point operations
13330 are enabled by default; routines in such libraries could suffer significant
13331 loss of accuracy, typically through so-called "catastrophic cancellation",
13332 when this option is used to set the precision to less than extended precision.
13334 @item -mstackrealign
13335 @opindex mstackrealign
13336 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
13337 option will generate an alternate prologue and epilogue that realigns the
13338 run-time stack if necessary. This supports mixing legacy codes that keep
13339 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
13340 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
13341 applicable to individual functions.
13343 @item -mpreferred-stack-boundary=@var{num}
13344 @opindex mpreferred-stack-boundary
13345 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
13346 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
13347 the default is 4 (16 bytes or 128 bits).
13349 @item -mincoming-stack-boundary=@var{num}
13350 @opindex mincoming-stack-boundary
13351 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
13352 boundary. If @option{-mincoming-stack-boundary} is not specified,
13353 the one specified by @option{-mpreferred-stack-boundary} will be used.
13355 On Pentium and PentiumPro, @code{double} and @code{long double} values
13356 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
13357 suffer significant run time performance penalties. On Pentium III, the
13358 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
13359 properly if it is not 16 byte aligned.
13361 To ensure proper alignment of this values on the stack, the stack boundary
13362 must be as aligned as that required by any value stored on the stack.
13363 Further, every function must be generated such that it keeps the stack
13364 aligned. Thus calling a function compiled with a higher preferred
13365 stack boundary from a function compiled with a lower preferred stack
13366 boundary will most likely misalign the stack. It is recommended that
13367 libraries that use callbacks always use the default setting.
13369 This extra alignment does consume extra stack space, and generally
13370 increases code size. Code that is sensitive to stack space usage, such
13371 as embedded systems and operating system kernels, may want to reduce the
13372 preferred alignment to @option{-mpreferred-stack-boundary=2}.
13401 @itemx -mno-fsgsbase
13437 These switches enable or disable the use of instructions in the MMX, SSE,
13438 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
13439 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
13440 @: extended instruction sets.
13441 These extensions are also available as built-in functions: see
13442 @ref{X86 Built-in Functions}, for details of the functions enabled and
13443 disabled by these switches.
13445 To have SSE/SSE2 instructions generated automatically from floating-point
13446 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
13448 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
13449 generates new AVX instructions or AVX equivalence for all SSEx instructions
13452 These options will enable GCC to use these extended instructions in
13453 generated code, even without @option{-mfpmath=sse}. Applications which
13454 perform run-time CPU detection must compile separate files for each
13455 supported architecture, using the appropriate flags. In particular,
13456 the file containing the CPU detection code should be compiled without
13461 This option instructs GCC to emit a @code{cld} instruction in the prologue
13462 of functions that use string instructions. String instructions depend on
13463 the DF flag to select between autoincrement or autodecrement mode. While the
13464 ABI specifies the DF flag to be cleared on function entry, some operating
13465 systems violate this specification by not clearing the DF flag in their
13466 exception dispatchers. The exception handler can be invoked with the DF flag
13467 set which leads to wrong direction mode, when string instructions are used.
13468 This option can be enabled by default on 32-bit x86 targets by configuring
13469 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
13470 instructions can be suppressed with the @option{-mno-cld} compiler option
13474 @opindex mvzeroupper
13475 This option instructs GCC to emit a @code{vzeroupper} instruction
13476 before a transfer of control flow out of the function to minimize
13477 AVX to SSE transition penalty as well as remove unnecessary zeroupper
13482 This option will enable GCC to use CMPXCHG16B instruction in generated code.
13483 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
13484 data types. This is useful for high resolution counters that could be updated
13485 by multiple processors (or cores). This instruction is generated as part of
13486 atomic built-in functions: see @ref{__sync Builtins} or
13487 @ref{__atomic Builtins} for details.
13491 This option will enable GCC to use SAHF instruction in generated 64-bit code.
13492 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
13493 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
13494 SAHF are load and store instructions, respectively, for certain status flags.
13495 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
13496 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
13500 This option will enable GCC to use movbe instruction to implement
13501 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
13505 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
13506 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
13507 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
13511 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
13512 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
13513 to increase precision instead of DIVSS and SQRTSS (and their vectorized
13514 variants) for single-precision floating-point arguments. These instructions
13515 are generated only when @option{-funsafe-math-optimizations} is enabled
13516 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
13517 Note that while the throughput of the sequence is higher than the throughput
13518 of the non-reciprocal instruction, the precision of the sequence can be
13519 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
13521 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS
13522 (or RSQRTPS) already with @option{-ffast-math} (or the above option
13523 combination), and doesn't need @option{-mrecip}.
13525 Also note that GCC emits the above sequence with additional Newton-Raphson step
13526 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
13527 already with @option{-ffast-math} (or the above option combination), and
13528 doesn't need @option{-mrecip}.
13530 @item -mrecip=@var{opt}
13531 @opindex mrecip=opt
13532 This option allows to control which reciprocal estimate instructions
13533 may be used. @var{opt} is a comma separated list of options, that may
13534 be preceded by a @code{!} to invert the option:
13535 @code{all}: enable all estimate instructions,
13536 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
13537 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip},
13538 @code{div}: enable the approximation for scalar division,
13539 @code{vec-div}: enable the approximation for vectorized division,
13540 @code{sqrt}: enable the approximation for scalar square root,
13541 @code{vec-sqrt}: enable the approximation for vectorized square root.
13543 So for example, @option{-mrecip=all,!sqrt} would enable
13544 all of the reciprocal approximations, except for square root.
13546 @item -mveclibabi=@var{type}
13547 @opindex mveclibabi
13548 Specifies the ABI type to use for vectorizing intrinsics using an
13549 external library. Supported types are @code{svml} for the Intel short
13550 vector math library and @code{acml} for the AMD math core library style
13551 of interfacing. GCC will currently emit calls to @code{vmldExp2},
13552 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
13553 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
13554 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
13555 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
13556 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
13557 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
13558 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
13559 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
13560 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
13561 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
13562 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
13563 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
13564 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
13565 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
13566 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
13567 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
13568 compatible library will have to be specified at link time.
13570 @item -mabi=@var{name}
13572 Generate code for the specified calling convention. Permissible values
13573 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
13574 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
13575 ABI when targeting Windows. On all other systems, the default is the
13576 SYSV ABI. You can control this behavior for a specific function by
13577 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
13578 @xref{Function Attributes}.
13580 @item -mtls-dialect=@var{type}
13581 @opindex mtls-dialect
13582 Generate code to access thread-local storage using the @samp{gnu} or
13583 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
13584 @samp{gnu2} is more efficient, but it may add compile- and run-time
13585 requirements that cannot be satisfied on all systems.
13588 @itemx -mno-push-args
13589 @opindex mpush-args
13590 @opindex mno-push-args
13591 Use PUSH operations to store outgoing parameters. This method is shorter
13592 and usually equally fast as method using SUB/MOV operations and is enabled
13593 by default. In some cases disabling it may improve performance because of
13594 improved scheduling and reduced dependencies.
13596 @item -maccumulate-outgoing-args
13597 @opindex maccumulate-outgoing-args
13598 If enabled, the maximum amount of space required for outgoing arguments will be
13599 computed in the function prologue. This is faster on most modern CPUs
13600 because of reduced dependencies, improved scheduling and reduced stack usage
13601 when preferred stack boundary is not equal to 2. The drawback is a notable
13602 increase in code size. This switch implies @option{-mno-push-args}.
13606 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
13607 on thread-safe exception handling must compile and link all code with the
13608 @option{-mthreads} option. When compiling, @option{-mthreads} defines
13609 @option{-D_MT}; when linking, it links in a special thread helper library
13610 @option{-lmingwthrd} which cleans up per thread exception handling data.
13612 @item -mno-align-stringops
13613 @opindex mno-align-stringops
13614 Do not align destination of inlined string operations. This switch reduces
13615 code size and improves performance in case the destination is already aligned,
13616 but GCC doesn't know about it.
13618 @item -minline-all-stringops
13619 @opindex minline-all-stringops
13620 By default GCC inlines string operations only when destination is known to be
13621 aligned at least to 4 byte boundary. This enables more inlining, increase code
13622 size, but may improve performance of code that depends on fast memcpy, strlen
13623 and memset for short lengths.
13625 @item -minline-stringops-dynamically
13626 @opindex minline-stringops-dynamically
13627 For string operations of unknown size, use run-time checks with
13628 inline code for small blocks and a library call for large blocks.
13630 @item -mstringop-strategy=@var{alg}
13631 @opindex mstringop-strategy=@var{alg}
13632 Overwrite internal decision heuristic about particular algorithm to inline
13633 string operation with. The allowed values are @code{rep_byte},
13634 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
13635 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
13636 expanding inline loop, @code{libcall} for always expanding library call.
13638 @item -momit-leaf-frame-pointer
13639 @opindex momit-leaf-frame-pointer
13640 Don't keep the frame pointer in a register for leaf functions. This
13641 avoids the instructions to save, set up and restore frame pointers and
13642 makes an extra register available in leaf functions. The option
13643 @option{-fomit-frame-pointer} removes the frame pointer for all functions
13644 which might make debugging harder.
13646 @item -mtls-direct-seg-refs
13647 @itemx -mno-tls-direct-seg-refs
13648 @opindex mtls-direct-seg-refs
13649 Controls whether TLS variables may be accessed with offsets from the
13650 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
13651 or whether the thread base pointer must be added. Whether or not this
13652 is legal depends on the operating system, and whether it maps the
13653 segment to cover the entire TLS area.
13655 For systems that use GNU libc, the default is on.
13658 @itemx -mno-sse2avx
13660 Specify that the assembler should encode SSE instructions with VEX
13661 prefix. The option @option{-mavx} turns this on by default.
13666 If profiling is active @option{-pg} put the profiling
13667 counter call before prologue.
13668 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13669 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13672 @itemx -mno-8bit-idiv
13674 On some processors, like Intel Atom, 8-bit unsigned integer divide is
13675 much faster than 32-bit/64-bit integer divide. This option generates a
13676 run-time check. If both dividend and divisor are within range of 0
13677 to 255, 8-bit unsigned integer divide is used instead of
13678 32-bit/64-bit integer divide.
13680 @item -mavx256-split-unaligned-load
13681 @item -mavx256-split-unaligned-store
13682 @opindex avx256-split-unaligned-load
13683 @opindex avx256-split-unaligned-store
13684 Split 32-byte AVX unaligned load and store.
13688 These @samp{-m} switches are supported in addition to the above
13689 on AMD x86-64 processors in 64-bit environments.
13698 Generate code for a 32-bit or 64-bit environment.
13699 The @option{-m32} option sets int, long and pointer to 32 bits and
13700 generates code that runs on any i386 system.
13701 The @option{-m64} option sets int to 32 bits and long and pointer
13702 to 64 bits and generates code for AMD's x86-64 architecture.
13703 The @option{-mx32} option sets int, long and pointer to 32 bits and
13704 generates code for AMD's x86-64 architecture.
13705 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
13706 and @option{-mdynamic-no-pic} options.
13708 @item -mno-red-zone
13709 @opindex mno-red-zone
13710 Do not use a so called red zone for x86-64 code. The red zone is mandated
13711 by the x86-64 ABI, it is a 128-byte area beyond the location of the
13712 stack pointer that will not be modified by signal or interrupt handlers
13713 and therefore can be used for temporary data without adjusting the stack
13714 pointer. The flag @option{-mno-red-zone} disables this red zone.
13716 @item -mcmodel=small
13717 @opindex mcmodel=small
13718 Generate code for the small code model: the program and its symbols must
13719 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13720 Programs can be statically or dynamically linked. This is the default
13723 @item -mcmodel=kernel
13724 @opindex mcmodel=kernel
13725 Generate code for the kernel code model. The kernel runs in the
13726 negative 2 GB of the address space.
13727 This model has to be used for Linux kernel code.
13729 @item -mcmodel=medium
13730 @opindex mcmodel=medium
13731 Generate code for the medium model: The program is linked in the lower 2
13732 GB of the address space. Small symbols are also placed there. Symbols
13733 with sizes larger than @option{-mlarge-data-threshold} are put into
13734 large data or bss sections and can be located above 2GB. Programs can
13735 be statically or dynamically linked.
13737 @item -mcmodel=large
13738 @opindex mcmodel=large
13739 Generate code for the large model: This model makes no assumptions
13740 about addresses and sizes of sections.
13743 @node i386 and x86-64 Windows Options
13744 @subsection i386 and x86-64 Windows Options
13745 @cindex i386 and x86-64 Windows Options
13747 These additional options are available for Windows targets:
13752 This option is available for Cygwin and MinGW targets. It
13753 specifies that a console application is to be generated, by
13754 instructing the linker to set the PE header subsystem type
13755 required for console applications.
13756 This is the default behavior for Cygwin and MinGW targets.
13760 This option is available for Cygwin and MinGW targets. It
13761 specifies that a DLL - a dynamic link library - is to be
13762 generated, enabling the selection of the required runtime
13763 startup object and entry point.
13765 @item -mnop-fun-dllimport
13766 @opindex mnop-fun-dllimport
13767 This option is available for Cygwin and MinGW targets. It
13768 specifies that the dllimport attribute should be ignored.
13772 This option is available for MinGW targets. It specifies
13773 that MinGW-specific thread support is to be used.
13777 This option is available for mingw-w64 targets. It specifies
13778 that the UNICODE macro is getting pre-defined and that the
13779 unicode capable runtime startup code is chosen.
13783 This option is available for Cygwin and MinGW targets. It
13784 specifies that the typical Windows pre-defined macros are to
13785 be set in the pre-processor, but does not influence the choice
13786 of runtime library/startup code.
13790 This option is available for Cygwin and MinGW targets. It
13791 specifies that a GUI application is to be generated by
13792 instructing the linker to set the PE header subsystem type
13795 @item -fno-set-stack-executable
13796 @opindex fno-set-stack-executable
13797 This option is available for MinGW targets. It specifies that
13798 the executable flag for stack used by nested functions isn't
13799 set. This is necessary for binaries running in kernel mode of
13800 Windows, as there the user32 API, which is used to set executable
13801 privileges, isn't available.
13803 @item -mpe-aligned-commons
13804 @opindex mpe-aligned-commons
13805 This option is available for Cygwin and MinGW targets. It
13806 specifies that the GNU extension to the PE file format that
13807 permits the correct alignment of COMMON variables should be
13808 used when generating code. It will be enabled by default if
13809 GCC detects that the target assembler found during configuration
13810 supports the feature.
13813 See also under @ref{i386 and x86-64 Options} for standard options.
13815 @node IA-64 Options
13816 @subsection IA-64 Options
13817 @cindex IA-64 Options
13819 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13823 @opindex mbig-endian
13824 Generate code for a big endian target. This is the default for HP-UX@.
13826 @item -mlittle-endian
13827 @opindex mlittle-endian
13828 Generate code for a little endian target. This is the default for AIX5
13834 @opindex mno-gnu-as
13835 Generate (or don't) code for the GNU assembler. This is the default.
13836 @c Also, this is the default if the configure option @option{--with-gnu-as}
13842 @opindex mno-gnu-ld
13843 Generate (or don't) code for the GNU linker. This is the default.
13844 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13849 Generate code that does not use a global pointer register. The result
13850 is not position independent code, and violates the IA-64 ABI@.
13852 @item -mvolatile-asm-stop
13853 @itemx -mno-volatile-asm-stop
13854 @opindex mvolatile-asm-stop
13855 @opindex mno-volatile-asm-stop
13856 Generate (or don't) a stop bit immediately before and after volatile asm
13859 @item -mregister-names
13860 @itemx -mno-register-names
13861 @opindex mregister-names
13862 @opindex mno-register-names
13863 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13864 the stacked registers. This may make assembler output more readable.
13870 Disable (or enable) optimizations that use the small data section. This may
13871 be useful for working around optimizer bugs.
13873 @item -mconstant-gp
13874 @opindex mconstant-gp
13875 Generate code that uses a single constant global pointer value. This is
13876 useful when compiling kernel code.
13880 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13881 This is useful when compiling firmware code.
13883 @item -minline-float-divide-min-latency
13884 @opindex minline-float-divide-min-latency
13885 Generate code for inline divides of floating-point values
13886 using the minimum latency algorithm.
13888 @item -minline-float-divide-max-throughput
13889 @opindex minline-float-divide-max-throughput
13890 Generate code for inline divides of floating-point values
13891 using the maximum throughput algorithm.
13893 @item -mno-inline-float-divide
13894 @opindex mno-inline-float-divide
13895 Do not generate inline code for divides of floating-point values.
13897 @item -minline-int-divide-min-latency
13898 @opindex minline-int-divide-min-latency
13899 Generate code for inline divides of integer values
13900 using the minimum latency algorithm.
13902 @item -minline-int-divide-max-throughput
13903 @opindex minline-int-divide-max-throughput
13904 Generate code for inline divides of integer values
13905 using the maximum throughput algorithm.
13907 @item -mno-inline-int-divide
13908 @opindex mno-inline-int-divide
13909 Do not generate inline code for divides of integer values.
13911 @item -minline-sqrt-min-latency
13912 @opindex minline-sqrt-min-latency
13913 Generate code for inline square roots
13914 using the minimum latency algorithm.
13916 @item -minline-sqrt-max-throughput
13917 @opindex minline-sqrt-max-throughput
13918 Generate code for inline square roots
13919 using the maximum throughput algorithm.
13921 @item -mno-inline-sqrt
13922 @opindex mno-inline-sqrt
13923 Do not generate inline code for sqrt.
13926 @itemx -mno-fused-madd
13927 @opindex mfused-madd
13928 @opindex mno-fused-madd
13929 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13930 instructions. The default is to use these instructions.
13932 @item -mno-dwarf2-asm
13933 @itemx -mdwarf2-asm
13934 @opindex mno-dwarf2-asm
13935 @opindex mdwarf2-asm
13936 Don't (or do) generate assembler code for the DWARF2 line number debugging
13937 info. This may be useful when not using the GNU assembler.
13939 @item -mearly-stop-bits
13940 @itemx -mno-early-stop-bits
13941 @opindex mearly-stop-bits
13942 @opindex mno-early-stop-bits
13943 Allow stop bits to be placed earlier than immediately preceding the
13944 instruction that triggered the stop bit. This can improve instruction
13945 scheduling, but does not always do so.
13947 @item -mfixed-range=@var{register-range}
13948 @opindex mfixed-range
13949 Generate code treating the given register range as fixed registers.
13950 A fixed register is one that the register allocator can not use. This is
13951 useful when compiling kernel code. A register range is specified as
13952 two registers separated by a dash. Multiple register ranges can be
13953 specified separated by a comma.
13955 @item -mtls-size=@var{tls-size}
13957 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13960 @item -mtune=@var{cpu-type}
13962 Tune the instruction scheduling for a particular CPU, Valid values are
13963 itanium, itanium1, merced, itanium2, and mckinley.
13969 Generate code for a 32-bit or 64-bit environment.
13970 The 32-bit environment sets int, long and pointer to 32 bits.
13971 The 64-bit environment sets int to 32 bits and long and pointer
13972 to 64 bits. These are HP-UX specific flags.
13974 @item -mno-sched-br-data-spec
13975 @itemx -msched-br-data-spec
13976 @opindex mno-sched-br-data-spec
13977 @opindex msched-br-data-spec
13978 (Dis/En)able data speculative scheduling before reload.
13979 This will result in generation of the ld.a instructions and
13980 the corresponding check instructions (ld.c / chk.a).
13981 The default is 'disable'.
13983 @item -msched-ar-data-spec
13984 @itemx -mno-sched-ar-data-spec
13985 @opindex msched-ar-data-spec
13986 @opindex mno-sched-ar-data-spec
13987 (En/Dis)able data speculative scheduling after reload.
13988 This will result in generation of the ld.a instructions and
13989 the corresponding check instructions (ld.c / chk.a).
13990 The default is 'enable'.
13992 @item -mno-sched-control-spec
13993 @itemx -msched-control-spec
13994 @opindex mno-sched-control-spec
13995 @opindex msched-control-spec
13996 (Dis/En)able control speculative scheduling. This feature is
13997 available only during region scheduling (i.e.@: before reload).
13998 This will result in generation of the ld.s instructions and
13999 the corresponding check instructions chk.s .
14000 The default is 'disable'.
14002 @item -msched-br-in-data-spec
14003 @itemx -mno-sched-br-in-data-spec
14004 @opindex msched-br-in-data-spec
14005 @opindex mno-sched-br-in-data-spec
14006 (En/Dis)able speculative scheduling of the instructions that
14007 are dependent on the data speculative loads before reload.
14008 This is effective only with @option{-msched-br-data-spec} enabled.
14009 The default is 'enable'.
14011 @item -msched-ar-in-data-spec
14012 @itemx -mno-sched-ar-in-data-spec
14013 @opindex msched-ar-in-data-spec
14014 @opindex mno-sched-ar-in-data-spec
14015 (En/Dis)able speculative scheduling of the instructions that
14016 are dependent on the data speculative loads after reload.
14017 This is effective only with @option{-msched-ar-data-spec} enabled.
14018 The default is 'enable'.
14020 @item -msched-in-control-spec
14021 @itemx -mno-sched-in-control-spec
14022 @opindex msched-in-control-spec
14023 @opindex mno-sched-in-control-spec
14024 (En/Dis)able speculative scheduling of the instructions that
14025 are dependent on the control speculative loads.
14026 This is effective only with @option{-msched-control-spec} enabled.
14027 The default is 'enable'.
14029 @item -mno-sched-prefer-non-data-spec-insns
14030 @itemx -msched-prefer-non-data-spec-insns
14031 @opindex mno-sched-prefer-non-data-spec-insns
14032 @opindex msched-prefer-non-data-spec-insns
14033 If enabled, data speculative instructions will be chosen for schedule
14034 only if there are no other choices at the moment. This will make
14035 the use of the data speculation much more conservative.
14036 The default is 'disable'.
14038 @item -mno-sched-prefer-non-control-spec-insns
14039 @itemx -msched-prefer-non-control-spec-insns
14040 @opindex mno-sched-prefer-non-control-spec-insns
14041 @opindex msched-prefer-non-control-spec-insns
14042 If enabled, control speculative instructions will be chosen for schedule
14043 only if there are no other choices at the moment. This will make
14044 the use of the control speculation much more conservative.
14045 The default is 'disable'.
14047 @item -mno-sched-count-spec-in-critical-path
14048 @itemx -msched-count-spec-in-critical-path
14049 @opindex mno-sched-count-spec-in-critical-path
14050 @opindex msched-count-spec-in-critical-path
14051 If enabled, speculative dependencies will be considered during
14052 computation of the instructions priorities. This will make the use of the
14053 speculation a bit more conservative.
14054 The default is 'disable'.
14056 @item -msched-spec-ldc
14057 @opindex msched-spec-ldc
14058 Use a simple data speculation check. This option is on by default.
14060 @item -msched-control-spec-ldc
14061 @opindex msched-spec-ldc
14062 Use a simple check for control speculation. This option is on by default.
14064 @item -msched-stop-bits-after-every-cycle
14065 @opindex msched-stop-bits-after-every-cycle
14066 Place a stop bit after every cycle when scheduling. This option is on
14069 @item -msched-fp-mem-deps-zero-cost
14070 @opindex msched-fp-mem-deps-zero-cost
14071 Assume that floating-point stores and loads are not likely to cause a conflict
14072 when placed into the same instruction group. This option is disabled by
14075 @item -msel-sched-dont-check-control-spec
14076 @opindex msel-sched-dont-check-control-spec
14077 Generate checks for control speculation in selective scheduling.
14078 This flag is disabled by default.
14080 @item -msched-max-memory-insns=@var{max-insns}
14081 @opindex msched-max-memory-insns
14082 Limit on the number of memory insns per instruction group, giving lower
14083 priority to subsequent memory insns attempting to schedule in the same
14084 instruction group. Frequently useful to prevent cache bank conflicts.
14085 The default value is 1.
14087 @item -msched-max-memory-insns-hard-limit
14088 @opindex msched-max-memory-insns-hard-limit
14089 Disallow more than `msched-max-memory-insns' in instruction group.
14090 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
14091 when limit is reached but may still schedule memory operations.
14095 @node IA-64/VMS Options
14096 @subsection IA-64/VMS Options
14098 These @samp{-m} options are defined for the IA-64/VMS implementations:
14101 @item -mvms-return-codes
14102 @opindex mvms-return-codes
14103 Return VMS condition codes from main. The default is to return POSIX
14104 style condition (e.g.@ error) codes.
14106 @item -mdebug-main=@var{prefix}
14107 @opindex mdebug-main=@var{prefix}
14108 Flag the first routine whose name starts with @var{prefix} as the main
14109 routine for the debugger.
14113 Default to 64bit memory allocation routines.
14117 @subsection LM32 Options
14118 @cindex LM32 options
14120 These @option{-m} options are defined for the Lattice Mico32 architecture:
14123 @item -mbarrel-shift-enabled
14124 @opindex mbarrel-shift-enabled
14125 Enable barrel-shift instructions.
14127 @item -mdivide-enabled
14128 @opindex mdivide-enabled
14129 Enable divide and modulus instructions.
14131 @item -mmultiply-enabled
14132 @opindex multiply-enabled
14133 Enable multiply instructions.
14135 @item -msign-extend-enabled
14136 @opindex msign-extend-enabled
14137 Enable sign extend instructions.
14139 @item -muser-enabled
14140 @opindex muser-enabled
14141 Enable user-defined instructions.
14146 @subsection M32C Options
14147 @cindex M32C options
14150 @item -mcpu=@var{name}
14152 Select the CPU for which code is generated. @var{name} may be one of
14153 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
14154 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
14155 the M32C/80 series.
14159 Specifies that the program will be run on the simulator. This causes
14160 an alternate runtime library to be linked in which supports, for
14161 example, file I/O@. You must not use this option when generating
14162 programs that will run on real hardware; you must provide your own
14163 runtime library for whatever I/O functions are needed.
14165 @item -memregs=@var{number}
14167 Specifies the number of memory-based pseudo-registers GCC will use
14168 during code generation. These pseudo-registers will be used like real
14169 registers, so there is a tradeoff between GCC's ability to fit the
14170 code into available registers, and the performance penalty of using
14171 memory instead of registers. Note that all modules in a program must
14172 be compiled with the same value for this option. Because of that, you
14173 must not use this option with the default runtime libraries gcc
14178 @node M32R/D Options
14179 @subsection M32R/D Options
14180 @cindex M32R/D options
14182 These @option{-m} options are defined for Renesas M32R/D architectures:
14187 Generate code for the M32R/2@.
14191 Generate code for the M32R/X@.
14195 Generate code for the M32R@. This is the default.
14197 @item -mmodel=small
14198 @opindex mmodel=small
14199 Assume all objects live in the lower 16MB of memory (so that their addresses
14200 can be loaded with the @code{ld24} instruction), and assume all subroutines
14201 are reachable with the @code{bl} instruction.
14202 This is the default.
14204 The addressability of a particular object can be set with the
14205 @code{model} attribute.
14207 @item -mmodel=medium
14208 @opindex mmodel=medium
14209 Assume objects may be anywhere in the 32-bit address space (the compiler
14210 will generate @code{seth/add3} instructions to load their addresses), and
14211 assume all subroutines are reachable with the @code{bl} instruction.
14213 @item -mmodel=large
14214 @opindex mmodel=large
14215 Assume objects may be anywhere in the 32-bit address space (the compiler
14216 will generate @code{seth/add3} instructions to load their addresses), and
14217 assume subroutines may not be reachable with the @code{bl} instruction
14218 (the compiler will generate the much slower @code{seth/add3/jl}
14219 instruction sequence).
14222 @opindex msdata=none
14223 Disable use of the small data area. Variables will be put into
14224 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
14225 @code{section} attribute has been specified).
14226 This is the default.
14228 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
14229 Objects may be explicitly put in the small data area with the
14230 @code{section} attribute using one of these sections.
14232 @item -msdata=sdata
14233 @opindex msdata=sdata
14234 Put small global and static data in the small data area, but do not
14235 generate special code to reference them.
14238 @opindex msdata=use
14239 Put small global and static data in the small data area, and generate
14240 special instructions to reference them.
14244 @cindex smaller data references
14245 Put global and static objects less than or equal to @var{num} bytes
14246 into the small data or bss sections instead of the normal data or bss
14247 sections. The default value of @var{num} is 8.
14248 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
14249 for this option to have any effect.
14251 All modules should be compiled with the same @option{-G @var{num}} value.
14252 Compiling with different values of @var{num} may or may not work; if it
14253 doesn't the linker will give an error message---incorrect code will not be
14258 Makes the M32R specific code in the compiler display some statistics
14259 that might help in debugging programs.
14261 @item -malign-loops
14262 @opindex malign-loops
14263 Align all loops to a 32-byte boundary.
14265 @item -mno-align-loops
14266 @opindex mno-align-loops
14267 Do not enforce a 32-byte alignment for loops. This is the default.
14269 @item -missue-rate=@var{number}
14270 @opindex missue-rate=@var{number}
14271 Issue @var{number} instructions per cycle. @var{number} can only be 1
14274 @item -mbranch-cost=@var{number}
14275 @opindex mbranch-cost=@var{number}
14276 @var{number} can only be 1 or 2. If it is 1 then branches will be
14277 preferred over conditional code, if it is 2, then the opposite will
14280 @item -mflush-trap=@var{number}
14281 @opindex mflush-trap=@var{number}
14282 Specifies the trap number to use to flush the cache. The default is
14283 12. Valid numbers are between 0 and 15 inclusive.
14285 @item -mno-flush-trap
14286 @opindex mno-flush-trap
14287 Specifies that the cache cannot be flushed by using a trap.
14289 @item -mflush-func=@var{name}
14290 @opindex mflush-func=@var{name}
14291 Specifies the name of the operating system function to call to flush
14292 the cache. The default is @emph{_flush_cache}, but a function call
14293 will only be used if a trap is not available.
14295 @item -mno-flush-func
14296 @opindex mno-flush-func
14297 Indicates that there is no OS function for flushing the cache.
14301 @node M680x0 Options
14302 @subsection M680x0 Options
14303 @cindex M680x0 options
14305 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
14306 The default settings depend on which architecture was selected when
14307 the compiler was configured; the defaults for the most common choices
14311 @item -march=@var{arch}
14313 Generate code for a specific M680x0 or ColdFire instruction set
14314 architecture. Permissible values of @var{arch} for M680x0
14315 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
14316 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
14317 architectures are selected according to Freescale's ISA classification
14318 and the permissible values are: @samp{isaa}, @samp{isaaplus},
14319 @samp{isab} and @samp{isac}.
14321 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
14322 code for a ColdFire target. The @var{arch} in this macro is one of the
14323 @option{-march} arguments given above.
14325 When used together, @option{-march} and @option{-mtune} select code
14326 that runs on a family of similar processors but that is optimized
14327 for a particular microarchitecture.
14329 @item -mcpu=@var{cpu}
14331 Generate code for a specific M680x0 or ColdFire processor.
14332 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
14333 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
14334 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
14335 below, which also classifies the CPUs into families:
14337 @multitable @columnfractions 0.20 0.80
14338 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
14339 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
14340 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
14341 @item @samp{5206e} @tab @samp{5206e}
14342 @item @samp{5208} @tab @samp{5207} @samp{5208}
14343 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
14344 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
14345 @item @samp{5216} @tab @samp{5214} @samp{5216}
14346 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
14347 @item @samp{5225} @tab @samp{5224} @samp{5225}
14348 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
14349 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
14350 @item @samp{5249} @tab @samp{5249}
14351 @item @samp{5250} @tab @samp{5250}
14352 @item @samp{5271} @tab @samp{5270} @samp{5271}
14353 @item @samp{5272} @tab @samp{5272}
14354 @item @samp{5275} @tab @samp{5274} @samp{5275}
14355 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
14356 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
14357 @item @samp{5307} @tab @samp{5307}
14358 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
14359 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
14360 @item @samp{5407} @tab @samp{5407}
14361 @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}
14364 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
14365 @var{arch} is compatible with @var{cpu}. Other combinations of
14366 @option{-mcpu} and @option{-march} are rejected.
14368 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
14369 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
14370 where the value of @var{family} is given by the table above.
14372 @item -mtune=@var{tune}
14374 Tune the code for a particular microarchitecture, within the
14375 constraints set by @option{-march} and @option{-mcpu}.
14376 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
14377 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
14378 and @samp{cpu32}. The ColdFire microarchitectures
14379 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
14381 You can also use @option{-mtune=68020-40} for code that needs
14382 to run relatively well on 68020, 68030 and 68040 targets.
14383 @option{-mtune=68020-60} is similar but includes 68060 targets
14384 as well. These two options select the same tuning decisions as
14385 @option{-m68020-40} and @option{-m68020-60} respectively.
14387 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
14388 when tuning for 680x0 architecture @var{arch}. It also defines
14389 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
14390 option is used. If gcc is tuning for a range of architectures,
14391 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
14392 it defines the macros for every architecture in the range.
14394 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
14395 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
14396 of the arguments given above.
14402 Generate output for a 68000. This is the default
14403 when the compiler is configured for 68000-based systems.
14404 It is equivalent to @option{-march=68000}.
14406 Use this option for microcontrollers with a 68000 or EC000 core,
14407 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
14411 Generate output for a 68010. This is the default
14412 when the compiler is configured for 68010-based systems.
14413 It is equivalent to @option{-march=68010}.
14419 Generate output for a 68020. This is the default
14420 when the compiler is configured for 68020-based systems.
14421 It is equivalent to @option{-march=68020}.
14425 Generate output for a 68030. This is the default when the compiler is
14426 configured for 68030-based systems. It is equivalent to
14427 @option{-march=68030}.
14431 Generate output for a 68040. This is the default when the compiler is
14432 configured for 68040-based systems. It is equivalent to
14433 @option{-march=68040}.
14435 This option inhibits the use of 68881/68882 instructions that have to be
14436 emulated by software on the 68040. Use this option if your 68040 does not
14437 have code to emulate those instructions.
14441 Generate output for a 68060. This is the default when the compiler is
14442 configured for 68060-based systems. It is equivalent to
14443 @option{-march=68060}.
14445 This option inhibits the use of 68020 and 68881/68882 instructions that
14446 have to be emulated by software on the 68060. Use this option if your 68060
14447 does not have code to emulate those instructions.
14451 Generate output for a CPU32. This is the default
14452 when the compiler is configured for CPU32-based systems.
14453 It is equivalent to @option{-march=cpu32}.
14455 Use this option for microcontrollers with a
14456 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
14457 68336, 68340, 68341, 68349 and 68360.
14461 Generate output for a 520X ColdFire CPU@. This is the default
14462 when the compiler is configured for 520X-based systems.
14463 It is equivalent to @option{-mcpu=5206}, and is now deprecated
14464 in favor of that option.
14466 Use this option for microcontroller with a 5200 core, including
14467 the MCF5202, MCF5203, MCF5204 and MCF5206.
14471 Generate output for a 5206e ColdFire CPU@. The option is now
14472 deprecated in favor of the equivalent @option{-mcpu=5206e}.
14476 Generate output for a member of the ColdFire 528X family.
14477 The option is now deprecated in favor of the equivalent
14478 @option{-mcpu=528x}.
14482 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
14483 in favor of the equivalent @option{-mcpu=5307}.
14487 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
14488 in favor of the equivalent @option{-mcpu=5407}.
14492 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
14493 This includes use of hardware floating-point instructions.
14494 The option is equivalent to @option{-mcpu=547x}, and is now
14495 deprecated in favor of that option.
14499 Generate output for a 68040, without using any of the new instructions.
14500 This results in code which can run relatively efficiently on either a
14501 68020/68881 or a 68030 or a 68040. The generated code does use the
14502 68881 instructions that are emulated on the 68040.
14504 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
14508 Generate output for a 68060, without using any of the new instructions.
14509 This results in code which can run relatively efficiently on either a
14510 68020/68881 or a 68030 or a 68040. The generated code does use the
14511 68881 instructions that are emulated on the 68060.
14513 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
14517 @opindex mhard-float
14519 Generate floating-point instructions. This is the default for 68020
14520 and above, and for ColdFire devices that have an FPU@. It defines the
14521 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
14522 on ColdFire targets.
14525 @opindex msoft-float
14526 Do not generate floating-point instructions; use library calls instead.
14527 This is the default for 68000, 68010, and 68832 targets. It is also
14528 the default for ColdFire devices that have no FPU.
14534 Generate (do not generate) ColdFire hardware divide and remainder
14535 instructions. If @option{-march} is used without @option{-mcpu},
14536 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
14537 architectures. Otherwise, the default is taken from the target CPU
14538 (either the default CPU, or the one specified by @option{-mcpu}). For
14539 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
14540 @option{-mcpu=5206e}.
14542 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
14546 Consider type @code{int} to be 16 bits wide, like @code{short int}.
14547 Additionally, parameters passed on the stack are also aligned to a
14548 16-bit boundary even on targets whose API mandates promotion to 32-bit.
14552 Do not consider type @code{int} to be 16 bits wide. This is the default.
14555 @itemx -mno-bitfield
14556 @opindex mnobitfield
14557 @opindex mno-bitfield
14558 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
14559 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
14563 Do use the bit-field instructions. The @option{-m68020} option implies
14564 @option{-mbitfield}. This is the default if you use a configuration
14565 designed for a 68020.
14569 Use a different function-calling convention, in which functions
14570 that take a fixed number of arguments return with the @code{rtd}
14571 instruction, which pops their arguments while returning. This
14572 saves one instruction in the caller since there is no need to pop
14573 the arguments there.
14575 This calling convention is incompatible with the one normally
14576 used on Unix, so you cannot use it if you need to call libraries
14577 compiled with the Unix compiler.
14579 Also, you must provide function prototypes for all functions that
14580 take variable numbers of arguments (including @code{printf});
14581 otherwise incorrect code will be generated for calls to those
14584 In addition, seriously incorrect code will result if you call a
14585 function with too many arguments. (Normally, extra arguments are
14586 harmlessly ignored.)
14588 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
14589 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
14593 Do not use the calling conventions selected by @option{-mrtd}.
14594 This is the default.
14597 @itemx -mno-align-int
14598 @opindex malign-int
14599 @opindex mno-align-int
14600 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
14601 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
14602 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
14603 Aligning variables on 32-bit boundaries produces code that runs somewhat
14604 faster on processors with 32-bit busses at the expense of more memory.
14606 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
14607 align structures containing the above types differently than
14608 most published application binary interface specifications for the m68k.
14612 Use the pc-relative addressing mode of the 68000 directly, instead of
14613 using a global offset table. At present, this option implies @option{-fpic},
14614 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
14615 not presently supported with @option{-mpcrel}, though this could be supported for
14616 68020 and higher processors.
14618 @item -mno-strict-align
14619 @itemx -mstrict-align
14620 @opindex mno-strict-align
14621 @opindex mstrict-align
14622 Do not (do) assume that unaligned memory references will be handled by
14626 Generate code that allows the data segment to be located in a different
14627 area of memory from the text segment. This allows for execute in place in
14628 an environment without virtual memory management. This option implies
14631 @item -mno-sep-data
14632 Generate code that assumes that the data segment follows the text segment.
14633 This is the default.
14635 @item -mid-shared-library
14636 Generate code that supports shared libraries via the library ID method.
14637 This allows for execute in place and shared libraries in an environment
14638 without virtual memory management. This option implies @option{-fPIC}.
14640 @item -mno-id-shared-library
14641 Generate code that doesn't assume ID based shared libraries are being used.
14642 This is the default.
14644 @item -mshared-library-id=n
14645 Specified the identification number of the ID based shared library being
14646 compiled. Specifying a value of 0 will generate more compact code, specifying
14647 other values will force the allocation of that number to the current
14648 library but is no more space or time efficient than omitting this option.
14654 When generating position-independent code for ColdFire, generate code
14655 that works if the GOT has more than 8192 entries. This code is
14656 larger and slower than code generated without this option. On M680x0
14657 processors, this option is not needed; @option{-fPIC} suffices.
14659 GCC normally uses a single instruction to load values from the GOT@.
14660 While this is relatively efficient, it only works if the GOT
14661 is smaller than about 64k. Anything larger causes the linker
14662 to report an error such as:
14664 @cindex relocation truncated to fit (ColdFire)
14666 relocation truncated to fit: R_68K_GOT16O foobar
14669 If this happens, you should recompile your code with @option{-mxgot}.
14670 It should then work with very large GOTs. However, code generated with
14671 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14672 the value of a global symbol.
14674 Note that some linkers, including newer versions of the GNU linker,
14675 can create multiple GOTs and sort GOT entries. If you have such a linker,
14676 you should only need to use @option{-mxgot} when compiling a single
14677 object file that accesses more than 8192 GOT entries. Very few do.
14679 These options have no effect unless GCC is generating
14680 position-independent code.
14684 @node MCore Options
14685 @subsection MCore Options
14686 @cindex MCore options
14688 These are the @samp{-m} options defined for the Motorola M*Core
14694 @itemx -mno-hardlit
14696 @opindex mno-hardlit
14697 Inline constants into the code stream if it can be done in two
14698 instructions or less.
14704 Use the divide instruction. (Enabled by default).
14706 @item -mrelax-immediate
14707 @itemx -mno-relax-immediate
14708 @opindex mrelax-immediate
14709 @opindex mno-relax-immediate
14710 Allow arbitrary sized immediates in bit operations.
14712 @item -mwide-bitfields
14713 @itemx -mno-wide-bitfields
14714 @opindex mwide-bitfields
14715 @opindex mno-wide-bitfields
14716 Always treat bit-fields as int-sized.
14718 @item -m4byte-functions
14719 @itemx -mno-4byte-functions
14720 @opindex m4byte-functions
14721 @opindex mno-4byte-functions
14722 Force all functions to be aligned to a four byte boundary.
14724 @item -mcallgraph-data
14725 @itemx -mno-callgraph-data
14726 @opindex mcallgraph-data
14727 @opindex mno-callgraph-data
14728 Emit callgraph information.
14731 @itemx -mno-slow-bytes
14732 @opindex mslow-bytes
14733 @opindex mno-slow-bytes
14734 Prefer word access when reading byte quantities.
14736 @item -mlittle-endian
14737 @itemx -mbig-endian
14738 @opindex mlittle-endian
14739 @opindex mbig-endian
14740 Generate code for a little endian target.
14746 Generate code for the 210 processor.
14750 Assume that runtime support has been provided and so omit the
14751 simulator library (@file{libsim.a)} from the linker command line.
14753 @item -mstack-increment=@var{size}
14754 @opindex mstack-increment
14755 Set the maximum amount for a single stack increment operation. Large
14756 values can increase the speed of programs which contain functions
14757 that need a large amount of stack space, but they can also trigger a
14758 segmentation fault if the stack is extended too much. The default
14764 @subsection MeP Options
14765 @cindex MeP options
14771 Enables the @code{abs} instruction, which is the absolute difference
14772 between two registers.
14776 Enables all the optional instructions - average, multiply, divide, bit
14777 operations, leading zero, absolute difference, min/max, clip, and
14783 Enables the @code{ave} instruction, which computes the average of two
14786 @item -mbased=@var{n}
14788 Variables of size @var{n} bytes or smaller will be placed in the
14789 @code{.based} section by default. Based variables use the @code{$tp}
14790 register as a base register, and there is a 128 byte limit to the
14791 @code{.based} section.
14795 Enables the bit operation instructions - bit test (@code{btstm}), set
14796 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14797 test-and-set (@code{tas}).
14799 @item -mc=@var{name}
14801 Selects which section constant data will be placed in. @var{name} may
14802 be @code{tiny}, @code{near}, or @code{far}.
14806 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14807 useful unless you also provide @code{-mminmax}.
14809 @item -mconfig=@var{name}
14811 Selects one of the build-in core configurations. Each MeP chip has
14812 one or more modules in it; each module has a core CPU and a variety of
14813 coprocessors, optional instructions, and peripherals. The
14814 @code{MeP-Integrator} tool, not part of GCC, provides these
14815 configurations through this option; using this option is the same as
14816 using all the corresponding command-line options. The default
14817 configuration is @code{default}.
14821 Enables the coprocessor instructions. By default, this is a 32-bit
14822 coprocessor. Note that the coprocessor is normally enabled via the
14823 @code{-mconfig=} option.
14827 Enables the 32-bit coprocessor's instructions.
14831 Enables the 64-bit coprocessor's instructions.
14835 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14839 Causes constant variables to be placed in the @code{.near} section.
14843 Enables the @code{div} and @code{divu} instructions.
14847 Generate big-endian code.
14851 Generate little-endian code.
14853 @item -mio-volatile
14854 @opindex mio-volatile
14855 Tells the compiler that any variable marked with the @code{io}
14856 attribute is to be considered volatile.
14860 Causes variables to be assigned to the @code{.far} section by default.
14864 Enables the @code{leadz} (leading zero) instruction.
14868 Causes variables to be assigned to the @code{.near} section by default.
14872 Enables the @code{min} and @code{max} instructions.
14876 Enables the multiplication and multiply-accumulate instructions.
14880 Disables all the optional instructions enabled by @code{-mall-opts}.
14884 Enables the @code{repeat} and @code{erepeat} instructions, used for
14885 low-overhead looping.
14889 Causes all variables to default to the @code{.tiny} section. Note
14890 that there is a 65536 byte limit to this section. Accesses to these
14891 variables use the @code{%gp} base register.
14895 Enables the saturation instructions. Note that the compiler does not
14896 currently generate these itself, but this option is included for
14897 compatibility with other tools, like @code{as}.
14901 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14905 Link the simulator runtime libraries.
14909 Link the simulator runtime libraries, excluding built-in support
14910 for reset and exception vectors and tables.
14914 Causes all functions to default to the @code{.far} section. Without
14915 this option, functions default to the @code{.near} section.
14917 @item -mtiny=@var{n}
14919 Variables that are @var{n} bytes or smaller will be allocated to the
14920 @code{.tiny} section. These variables use the @code{$gp} base
14921 register. The default for this option is 4, but note that there's a
14922 65536 byte limit to the @code{.tiny} section.
14926 @node MicroBlaze Options
14927 @subsection MicroBlaze Options
14928 @cindex MicroBlaze Options
14933 @opindex msoft-float
14934 Use software emulation for floating point (default).
14937 @opindex mhard-float
14938 Use hardware floating-point instructions.
14942 Do not optimize block moves, use @code{memcpy}.
14944 @item -mno-clearbss
14945 @opindex mno-clearbss
14946 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14948 @item -mcpu=@var{cpu-type}
14950 Use features of and schedule code for given CPU.
14951 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14952 where @var{X} is a major version, @var{YY} is the minor version, and
14953 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14954 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14956 @item -mxl-soft-mul
14957 @opindex mxl-soft-mul
14958 Use software multiply emulation (default).
14960 @item -mxl-soft-div
14961 @opindex mxl-soft-div
14962 Use software emulation for divides (default).
14964 @item -mxl-barrel-shift
14965 @opindex mxl-barrel-shift
14966 Use the hardware barrel shifter.
14968 @item -mxl-pattern-compare
14969 @opindex mxl-pattern-compare
14970 Use pattern compare instructions.
14972 @item -msmall-divides
14973 @opindex msmall-divides
14974 Use table lookup optimization for small signed integer divisions.
14976 @item -mxl-stack-check
14977 @opindex mxl-stack-check
14978 This option is deprecated. Use -fstack-check instead.
14981 @opindex mxl-gp-opt
14982 Use GP relative sdata/sbss sections.
14984 @item -mxl-multiply-high
14985 @opindex mxl-multiply-high
14986 Use multiply high instructions for high part of 32x32 multiply.
14988 @item -mxl-float-convert
14989 @opindex mxl-float-convert
14990 Use hardware floating-point conversion instructions.
14992 @item -mxl-float-sqrt
14993 @opindex mxl-float-sqrt
14994 Use hardware floating-point square root instruction.
14996 @item -mxl-mode-@var{app-model}
14997 Select application model @var{app-model}. Valid models are
15000 normal executable (default), uses startup code @file{crt0.o}.
15003 for use with Xilinx Microprocessor Debugger (XMD) based
15004 software intrusive debug agent called xmdstub. This uses startup file
15005 @file{crt1.o} and sets the start address of the program to be 0x800.
15008 for applications that are loaded using a bootloader.
15009 This model uses startup file @file{crt2.o} which does not contain a processor
15010 reset vector handler. This is suitable for transferring control on a
15011 processor reset to the bootloader rather than the application.
15014 for applications that do not require any of the
15015 MicroBlaze vectors. This option may be useful for applications running
15016 within a monitoring application. This model uses @file{crt3.o} as a startup file.
15019 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
15020 @option{-mxl-mode-@var{app-model}}.
15025 @subsection MIPS Options
15026 @cindex MIPS options
15032 Generate big-endian code.
15036 Generate little-endian code. This is the default for @samp{mips*el-*-*}
15039 @item -march=@var{arch}
15041 Generate code that will run on @var{arch}, which can be the name of a
15042 generic MIPS ISA, or the name of a particular processor.
15044 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
15045 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
15046 The processor names are:
15047 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
15048 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
15049 @samp{5kc}, @samp{5kf},
15051 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
15052 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
15053 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
15054 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
15055 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
15056 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
15058 @samp{octeon}, @samp{octeon+}, @samp{octeon2},
15060 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
15061 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
15062 @samp{rm7000}, @samp{rm9000},
15063 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
15066 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
15067 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
15069 The special value @samp{from-abi} selects the
15070 most compatible architecture for the selected ABI (that is,
15071 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
15073 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
15074 which selects the best architecture option for the host processor.
15075 @option{-march=native} has no effect if GCC does not recognize
15078 In processor names, a final @samp{000} can be abbreviated as @samp{k}
15079 (for example, @samp{-march=r2k}). Prefixes are optional, and
15080 @samp{vr} may be written @samp{r}.
15082 Names of the form @samp{@var{n}f2_1} refer to processors with
15083 FPUs clocked at half the rate of the core, names of the form
15084 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
15085 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
15086 processors with FPUs clocked a ratio of 3:2 with respect to the core.
15087 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
15088 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
15089 accepted as synonyms for @samp{@var{n}f1_1}.
15091 GCC defines two macros based on the value of this option. The first
15092 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
15093 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
15094 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
15095 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
15096 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
15098 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
15099 above. In other words, it will have the full prefix and will not
15100 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
15101 the macro names the resolved architecture (either @samp{"mips1"} or
15102 @samp{"mips3"}). It names the default architecture when no
15103 @option{-march} option is given.
15105 @item -mtune=@var{arch}
15107 Optimize for @var{arch}. Among other things, this option controls
15108 the way instructions are scheduled, and the perceived cost of arithmetic
15109 operations. The list of @var{arch} values is the same as for
15112 When this option is not used, GCC will optimize for the processor
15113 specified by @option{-march}. By using @option{-march} and
15114 @option{-mtune} together, it is possible to generate code that will
15115 run on a family of processors, but optimize the code for one
15116 particular member of that family.
15118 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
15119 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
15120 @samp{-march} ones described above.
15124 Equivalent to @samp{-march=mips1}.
15128 Equivalent to @samp{-march=mips2}.
15132 Equivalent to @samp{-march=mips3}.
15136 Equivalent to @samp{-march=mips4}.
15140 Equivalent to @samp{-march=mips32}.
15144 Equivalent to @samp{-march=mips32r2}.
15148 Equivalent to @samp{-march=mips64}.
15152 Equivalent to @samp{-march=mips64r2}.
15157 @opindex mno-mips16
15158 Generate (do not generate) MIPS16 code. If GCC is targetting a
15159 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
15161 MIPS16 code generation can also be controlled on a per-function basis
15162 by means of @code{mips16} and @code{nomips16} attributes.
15163 @xref{Function Attributes}, for more information.
15165 @item -mflip-mips16
15166 @opindex mflip-mips16
15167 Generate MIPS16 code on alternating functions. This option is provided
15168 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
15169 not intended for ordinary use in compiling user code.
15171 @item -minterlink-mips16
15172 @itemx -mno-interlink-mips16
15173 @opindex minterlink-mips16
15174 @opindex mno-interlink-mips16
15175 Require (do not require) that non-MIPS16 code be link-compatible with
15178 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
15179 it must either use a call or an indirect jump. @option{-minterlink-mips16}
15180 therefore disables direct jumps unless GCC knows that the target of the
15181 jump is not MIPS16.
15193 Generate code for the given ABI@.
15195 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
15196 generates 64-bit code when you select a 64-bit architecture, but you
15197 can use @option{-mgp32} to get 32-bit code instead.
15199 For information about the O64 ABI, see
15200 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
15202 GCC supports a variant of the o32 ABI in which floating-point registers
15203 are 64 rather than 32 bits wide. You can select this combination with
15204 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
15205 and @samp{mfhc1} instructions and is therefore only supported for
15206 MIPS32R2 processors.
15208 The register assignments for arguments and return values remain the
15209 same, but each scalar value is passed in a single 64-bit register
15210 rather than a pair of 32-bit registers. For example, scalar
15211 floating-point values are returned in @samp{$f0} only, not a
15212 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
15213 remains the same, but all 64 bits are saved.
15216 @itemx -mno-abicalls
15218 @opindex mno-abicalls
15219 Generate (do not generate) code that is suitable for SVR4-style
15220 dynamic objects. @option{-mabicalls} is the default for SVR4-based
15225 Generate (do not generate) code that is fully position-independent,
15226 and that can therefore be linked into shared libraries. This option
15227 only affects @option{-mabicalls}.
15229 All @option{-mabicalls} code has traditionally been position-independent,
15230 regardless of options like @option{-fPIC} and @option{-fpic}. However,
15231 as an extension, the GNU toolchain allows executables to use absolute
15232 accesses for locally-binding symbols. It can also use shorter GP
15233 initialization sequences and generate direct calls to locally-defined
15234 functions. This mode is selected by @option{-mno-shared}.
15236 @option{-mno-shared} depends on binutils 2.16 or higher and generates
15237 objects that can only be linked by the GNU linker. However, the option
15238 does not affect the ABI of the final executable; it only affects the ABI
15239 of relocatable objects. Using @option{-mno-shared} will generally make
15240 executables both smaller and quicker.
15242 @option{-mshared} is the default.
15248 Assume (do not assume) that the static and dynamic linkers
15249 support PLTs and copy relocations. This option only affects
15250 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
15251 has no effect without @samp{-msym32}.
15253 You can make @option{-mplt} the default by configuring
15254 GCC with @option{--with-mips-plt}. The default is
15255 @option{-mno-plt} otherwise.
15261 Lift (do not lift) the usual restrictions on the size of the global
15264 GCC normally uses a single instruction to load values from the GOT@.
15265 While this is relatively efficient, it will only work if the GOT
15266 is smaller than about 64k. Anything larger will cause the linker
15267 to report an error such as:
15269 @cindex relocation truncated to fit (MIPS)
15271 relocation truncated to fit: R_MIPS_GOT16 foobar
15274 If this happens, you should recompile your code with @option{-mxgot}.
15275 It should then work with very large GOTs, although it will also be
15276 less efficient, since it will take three instructions to fetch the
15277 value of a global symbol.
15279 Note that some linkers can create multiple GOTs. If you have such a
15280 linker, you should only need to use @option{-mxgot} when a single object
15281 file accesses more than 64k's worth of GOT entries. Very few do.
15283 These options have no effect unless GCC is generating position
15288 Assume that general-purpose registers are 32 bits wide.
15292 Assume that general-purpose registers are 64 bits wide.
15296 Assume that floating-point registers are 32 bits wide.
15300 Assume that floating-point registers are 64 bits wide.
15303 @opindex mhard-float
15304 Use floating-point coprocessor instructions.
15307 @opindex msoft-float
15308 Do not use floating-point coprocessor instructions. Implement
15309 floating-point calculations using library calls instead.
15311 @item -msingle-float
15312 @opindex msingle-float
15313 Assume that the floating-point coprocessor only supports single-precision
15316 @item -mdouble-float
15317 @opindex mdouble-float
15318 Assume that the floating-point coprocessor supports double-precision
15319 operations. This is the default.
15325 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
15326 implement atomic memory built-in functions. When neither option is
15327 specified, GCC will use the instructions if the target architecture
15330 @option{-mllsc} is useful if the runtime environment can emulate the
15331 instructions and @option{-mno-llsc} can be useful when compiling for
15332 nonstandard ISAs. You can make either option the default by
15333 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
15334 respectively. @option{--with-llsc} is the default for some
15335 configurations; see the installation documentation for details.
15341 Use (do not use) revision 1 of the MIPS DSP ASE@.
15342 @xref{MIPS DSP Built-in Functions}. This option defines the
15343 preprocessor macro @samp{__mips_dsp}. It also defines
15344 @samp{__mips_dsp_rev} to 1.
15350 Use (do not use) revision 2 of the MIPS DSP ASE@.
15351 @xref{MIPS DSP Built-in Functions}. This option defines the
15352 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
15353 It also defines @samp{__mips_dsp_rev} to 2.
15356 @itemx -mno-smartmips
15357 @opindex msmartmips
15358 @opindex mno-smartmips
15359 Use (do not use) the MIPS SmartMIPS ASE.
15361 @item -mpaired-single
15362 @itemx -mno-paired-single
15363 @opindex mpaired-single
15364 @opindex mno-paired-single
15365 Use (do not use) paired-single floating-point instructions.
15366 @xref{MIPS Paired-Single Support}. This option requires
15367 hardware floating-point support to be enabled.
15373 Use (do not use) MIPS Digital Media Extension instructions.
15374 This option can only be used when generating 64-bit code and requires
15375 hardware floating-point support to be enabled.
15380 @opindex mno-mips3d
15381 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
15382 The option @option{-mips3d} implies @option{-mpaired-single}.
15388 Use (do not use) MT Multithreading instructions.
15392 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
15393 an explanation of the default and the way that the pointer size is
15398 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
15400 The default size of @code{int}s, @code{long}s and pointers depends on
15401 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
15402 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
15403 32-bit @code{long}s. Pointers are the same size as @code{long}s,
15404 or the same size as integer registers, whichever is smaller.
15410 Assume (do not assume) that all symbols have 32-bit values, regardless
15411 of the selected ABI@. This option is useful in combination with
15412 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
15413 to generate shorter and faster references to symbolic addresses.
15417 Put definitions of externally-visible data in a small data section
15418 if that data is no bigger than @var{num} bytes. GCC can then access
15419 the data more efficiently; see @option{-mgpopt} for details.
15421 The default @option{-G} option depends on the configuration.
15423 @item -mlocal-sdata
15424 @itemx -mno-local-sdata
15425 @opindex mlocal-sdata
15426 @opindex mno-local-sdata
15427 Extend (do not extend) the @option{-G} behavior to local data too,
15428 such as to static variables in C@. @option{-mlocal-sdata} is the
15429 default for all configurations.
15431 If the linker complains that an application is using too much small data,
15432 you might want to try rebuilding the less performance-critical parts with
15433 @option{-mno-local-sdata}. You might also want to build large
15434 libraries with @option{-mno-local-sdata}, so that the libraries leave
15435 more room for the main program.
15437 @item -mextern-sdata
15438 @itemx -mno-extern-sdata
15439 @opindex mextern-sdata
15440 @opindex mno-extern-sdata
15441 Assume (do not assume) that externally-defined data will be in
15442 a small data section if that data is within the @option{-G} limit.
15443 @option{-mextern-sdata} is the default for all configurations.
15445 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
15446 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
15447 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
15448 is placed in a small data section. If @var{Var} is defined by another
15449 module, you must either compile that module with a high-enough
15450 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
15451 definition. If @var{Var} is common, you must link the application
15452 with a high-enough @option{-G} setting.
15454 The easiest way of satisfying these restrictions is to compile
15455 and link every module with the same @option{-G} option. However,
15456 you may wish to build a library that supports several different
15457 small data limits. You can do this by compiling the library with
15458 the highest supported @option{-G} setting and additionally using
15459 @option{-mno-extern-sdata} to stop the library from making assumptions
15460 about externally-defined data.
15466 Use (do not use) GP-relative accesses for symbols that are known to be
15467 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
15468 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
15471 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
15472 might not hold the value of @code{_gp}. For example, if the code is
15473 part of a library that might be used in a boot monitor, programs that
15474 call boot monitor routines will pass an unknown value in @code{$gp}.
15475 (In such situations, the boot monitor itself would usually be compiled
15476 with @option{-G0}.)
15478 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
15479 @option{-mno-extern-sdata}.
15481 @item -membedded-data
15482 @itemx -mno-embedded-data
15483 @opindex membedded-data
15484 @opindex mno-embedded-data
15485 Allocate variables to the read-only data section first if possible, then
15486 next in the small data section if possible, otherwise in data. This gives
15487 slightly slower code than the default, but reduces the amount of RAM required
15488 when executing, and thus may be preferred for some embedded systems.
15490 @item -muninit-const-in-rodata
15491 @itemx -mno-uninit-const-in-rodata
15492 @opindex muninit-const-in-rodata
15493 @opindex mno-uninit-const-in-rodata
15494 Put uninitialized @code{const} variables in the read-only data section.
15495 This option is only meaningful in conjunction with @option{-membedded-data}.
15497 @item -mcode-readable=@var{setting}
15498 @opindex mcode-readable
15499 Specify whether GCC may generate code that reads from executable sections.
15500 There are three possible settings:
15503 @item -mcode-readable=yes
15504 Instructions may freely access executable sections. This is the
15507 @item -mcode-readable=pcrel
15508 MIPS16 PC-relative load instructions can access executable sections,
15509 but other instructions must not do so. This option is useful on 4KSc
15510 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
15511 It is also useful on processors that can be configured to have a dual
15512 instruction/data SRAM interface and that, like the M4K, automatically
15513 redirect PC-relative loads to the instruction RAM.
15515 @item -mcode-readable=no
15516 Instructions must not access executable sections. This option can be
15517 useful on targets that are configured to have a dual instruction/data
15518 SRAM interface but that (unlike the M4K) do not automatically redirect
15519 PC-relative loads to the instruction RAM.
15522 @item -msplit-addresses
15523 @itemx -mno-split-addresses
15524 @opindex msplit-addresses
15525 @opindex mno-split-addresses
15526 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
15527 relocation operators. This option has been superseded by
15528 @option{-mexplicit-relocs} but is retained for backwards compatibility.
15530 @item -mexplicit-relocs
15531 @itemx -mno-explicit-relocs
15532 @opindex mexplicit-relocs
15533 @opindex mno-explicit-relocs
15534 Use (do not use) assembler relocation operators when dealing with symbolic
15535 addresses. The alternative, selected by @option{-mno-explicit-relocs},
15536 is to use assembler macros instead.
15538 @option{-mexplicit-relocs} is the default if GCC was configured
15539 to use an assembler that supports relocation operators.
15541 @item -mcheck-zero-division
15542 @itemx -mno-check-zero-division
15543 @opindex mcheck-zero-division
15544 @opindex mno-check-zero-division
15545 Trap (do not trap) on integer division by zero.
15547 The default is @option{-mcheck-zero-division}.
15549 @item -mdivide-traps
15550 @itemx -mdivide-breaks
15551 @opindex mdivide-traps
15552 @opindex mdivide-breaks
15553 MIPS systems check for division by zero by generating either a
15554 conditional trap or a break instruction. Using traps results in
15555 smaller code, but is only supported on MIPS II and later. Also, some
15556 versions of the Linux kernel have a bug that prevents trap from
15557 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
15558 allow conditional traps on architectures that support them and
15559 @option{-mdivide-breaks} to force the use of breaks.
15561 The default is usually @option{-mdivide-traps}, but this can be
15562 overridden at configure time using @option{--with-divide=breaks}.
15563 Divide-by-zero checks can be completely disabled using
15564 @option{-mno-check-zero-division}.
15569 @opindex mno-memcpy
15570 Force (do not force) the use of @code{memcpy()} for non-trivial block
15571 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
15572 most constant-sized copies.
15575 @itemx -mno-long-calls
15576 @opindex mlong-calls
15577 @opindex mno-long-calls
15578 Disable (do not disable) use of the @code{jal} instruction. Calling
15579 functions using @code{jal} is more efficient but requires the caller
15580 and callee to be in the same 256 megabyte segment.
15582 This option has no effect on abicalls code. The default is
15583 @option{-mno-long-calls}.
15589 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
15590 instructions, as provided by the R4650 ISA@.
15593 @itemx -mno-fused-madd
15594 @opindex mfused-madd
15595 @opindex mno-fused-madd
15596 Enable (disable) use of the floating-point multiply-accumulate
15597 instructions, when they are available. The default is
15598 @option{-mfused-madd}.
15600 When multiply-accumulate instructions are used, the intermediate
15601 product is calculated to infinite precision and is not subject to
15602 the FCSR Flush to Zero bit. This may be undesirable in some
15607 Tell the MIPS assembler to not run its preprocessor over user
15608 assembler files (with a @samp{.s} suffix) when assembling them.
15613 @opindex mno-fix-24k
15614 Work around the 24K E48 (lost data on stores during refill) errata.
15615 The workarounds are implemented by the assembler rather than by GCC.
15618 @itemx -mno-fix-r4000
15619 @opindex mfix-r4000
15620 @opindex mno-fix-r4000
15621 Work around certain R4000 CPU errata:
15624 A double-word or a variable shift may give an incorrect result if executed
15625 immediately after starting an integer division.
15627 A double-word or a variable shift may give an incorrect result if executed
15628 while an integer multiplication is in progress.
15630 An integer division may give an incorrect result if started in a delay slot
15631 of a taken branch or a jump.
15635 @itemx -mno-fix-r4400
15636 @opindex mfix-r4400
15637 @opindex mno-fix-r4400
15638 Work around certain R4400 CPU errata:
15641 A double-word or a variable shift may give an incorrect result if executed
15642 immediately after starting an integer division.
15646 @itemx -mno-fix-r10000
15647 @opindex mfix-r10000
15648 @opindex mno-fix-r10000
15649 Work around certain R10000 errata:
15652 @code{ll}/@code{sc} sequences may not behave atomically on revisions
15653 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15656 This option can only be used if the target architecture supports
15657 branch-likely instructions. @option{-mfix-r10000} is the default when
15658 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15662 @itemx -mno-fix-vr4120
15663 @opindex mfix-vr4120
15664 Work around certain VR4120 errata:
15667 @code{dmultu} does not always produce the correct result.
15669 @code{div} and @code{ddiv} do not always produce the correct result if one
15670 of the operands is negative.
15672 The workarounds for the division errata rely on special functions in
15673 @file{libgcc.a}. At present, these functions are only provided by
15674 the @code{mips64vr*-elf} configurations.
15676 Other VR4120 errata require a nop to be inserted between certain pairs of
15677 instructions. These errata are handled by the assembler, not by GCC itself.
15680 @opindex mfix-vr4130
15681 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15682 workarounds are implemented by the assembler rather than by GCC,
15683 although GCC will avoid using @code{mflo} and @code{mfhi} if the
15684 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15685 instructions are available instead.
15688 @itemx -mno-fix-sb1
15690 Work around certain SB-1 CPU core errata.
15691 (This flag currently works around the SB-1 revision 2
15692 ``F1'' and ``F2'' floating-point errata.)
15694 @item -mr10k-cache-barrier=@var{setting}
15695 @opindex mr10k-cache-barrier
15696 Specify whether GCC should insert cache barriers to avoid the
15697 side-effects of speculation on R10K processors.
15699 In common with many processors, the R10K tries to predict the outcome
15700 of a conditional branch and speculatively executes instructions from
15701 the ``taken'' branch. It later aborts these instructions if the
15702 predicted outcome was wrong. However, on the R10K, even aborted
15703 instructions can have side effects.
15705 This problem only affects kernel stores and, depending on the system,
15706 kernel loads. As an example, a speculatively-executed store may load
15707 the target memory into cache and mark the cache line as dirty, even if
15708 the store itself is later aborted. If a DMA operation writes to the
15709 same area of memory before the ``dirty'' line is flushed, the cached
15710 data will overwrite the DMA-ed data. See the R10K processor manual
15711 for a full description, including other potential problems.
15713 One workaround is to insert cache barrier instructions before every memory
15714 access that might be speculatively executed and that might have side
15715 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15716 controls GCC's implementation of this workaround. It assumes that
15717 aborted accesses to any byte in the following regions will not have
15722 the memory occupied by the current function's stack frame;
15725 the memory occupied by an incoming stack argument;
15728 the memory occupied by an object with a link-time-constant address.
15731 It is the kernel's responsibility to ensure that speculative
15732 accesses to these regions are indeed safe.
15734 If the input program contains a function declaration such as:
15740 then the implementation of @code{foo} must allow @code{j foo} and
15741 @code{jal foo} to be executed speculatively. GCC honors this
15742 restriction for functions it compiles itself. It expects non-GCC
15743 functions (such as hand-written assembly code) to do the same.
15745 The option has three forms:
15748 @item -mr10k-cache-barrier=load-store
15749 Insert a cache barrier before a load or store that might be
15750 speculatively executed and that might have side effects even
15753 @item -mr10k-cache-barrier=store
15754 Insert a cache barrier before a store that might be speculatively
15755 executed and that might have side effects even if aborted.
15757 @item -mr10k-cache-barrier=none
15758 Disable the insertion of cache barriers. This is the default setting.
15761 @item -mflush-func=@var{func}
15762 @itemx -mno-flush-func
15763 @opindex mflush-func
15764 Specifies the function to call to flush the I and D caches, or to not
15765 call any such function. If called, the function must take the same
15766 arguments as the common @code{_flush_func()}, that is, the address of the
15767 memory range for which the cache is being flushed, the size of the
15768 memory range, and the number 3 (to flush both caches). The default
15769 depends on the target GCC was configured for, but commonly is either
15770 @samp{_flush_func} or @samp{__cpu_flush}.
15772 @item mbranch-cost=@var{num}
15773 @opindex mbranch-cost
15774 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15775 This cost is only a heuristic and is not guaranteed to produce
15776 consistent results across releases. A zero cost redundantly selects
15777 the default, which is based on the @option{-mtune} setting.
15779 @item -mbranch-likely
15780 @itemx -mno-branch-likely
15781 @opindex mbranch-likely
15782 @opindex mno-branch-likely
15783 Enable or disable use of Branch Likely instructions, regardless of the
15784 default for the selected architecture. By default, Branch Likely
15785 instructions may be generated if they are supported by the selected
15786 architecture. An exception is for the MIPS32 and MIPS64 architectures
15787 and processors which implement those architectures; for those, Branch
15788 Likely instructions will not be generated by default because the MIPS32
15789 and MIPS64 architectures specifically deprecate their use.
15791 @item -mfp-exceptions
15792 @itemx -mno-fp-exceptions
15793 @opindex mfp-exceptions
15794 Specifies whether FP exceptions are enabled. This affects how we schedule
15795 FP instructions for some processors. The default is that FP exceptions are
15798 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15799 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15802 @item -mvr4130-align
15803 @itemx -mno-vr4130-align
15804 @opindex mvr4130-align
15805 The VR4130 pipeline is two-way superscalar, but can only issue two
15806 instructions together if the first one is 8-byte aligned. When this
15807 option is enabled, GCC will align pairs of instructions that it
15808 thinks should execute in parallel.
15810 This option only has an effect when optimizing for the VR4130.
15811 It normally makes code faster, but at the expense of making it bigger.
15812 It is enabled by default at optimization level @option{-O3}.
15817 Enable (disable) generation of @code{synci} instructions on
15818 architectures that support it. The @code{synci} instructions (if
15819 enabled) will be generated when @code{__builtin___clear_cache()} is
15822 This option defaults to @code{-mno-synci}, but the default can be
15823 overridden by configuring with @code{--with-synci}.
15825 When compiling code for single processor systems, it is generally safe
15826 to use @code{synci}. However, on many multi-core (SMP) systems, it
15827 will not invalidate the instruction caches on all cores and may lead
15828 to undefined behavior.
15830 @item -mrelax-pic-calls
15831 @itemx -mno-relax-pic-calls
15832 @opindex mrelax-pic-calls
15833 Try to turn PIC calls that are normally dispatched via register
15834 @code{$25} into direct calls. This is only possible if the linker can
15835 resolve the destination at link-time and if the destination is within
15836 range for a direct call.
15838 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15839 an assembler and a linker that supports the @code{.reloc} assembly
15840 directive and @code{-mexplicit-relocs} is in effect. With
15841 @code{-mno-explicit-relocs}, this optimization can be performed by the
15842 assembler and the linker alone without help from the compiler.
15844 @item -mmcount-ra-address
15845 @itemx -mno-mcount-ra-address
15846 @opindex mmcount-ra-address
15847 @opindex mno-mcount-ra-address
15848 Emit (do not emit) code that allows @code{_mcount} to modify the
15849 calling function's return address. When enabled, this option extends
15850 the usual @code{_mcount} interface with a new @var{ra-address}
15851 parameter, which has type @code{intptr_t *} and is passed in register
15852 @code{$12}. @code{_mcount} can then modify the return address by
15853 doing both of the following:
15856 Returning the new address in register @code{$31}.
15858 Storing the new address in @code{*@var{ra-address}},
15859 if @var{ra-address} is nonnull.
15862 The default is @option{-mno-mcount-ra-address}.
15867 @subsection MMIX Options
15868 @cindex MMIX Options
15870 These options are defined for the MMIX:
15874 @itemx -mno-libfuncs
15876 @opindex mno-libfuncs
15877 Specify that intrinsic library functions are being compiled, passing all
15878 values in registers, no matter the size.
15881 @itemx -mno-epsilon
15883 @opindex mno-epsilon
15884 Generate floating-point comparison instructions that compare with respect
15885 to the @code{rE} epsilon register.
15887 @item -mabi=mmixware
15889 @opindex mabi=mmixware
15891 Generate code that passes function parameters and return values that (in
15892 the called function) are seen as registers @code{$0} and up, as opposed to
15893 the GNU ABI which uses global registers @code{$231} and up.
15895 @item -mzero-extend
15896 @itemx -mno-zero-extend
15897 @opindex mzero-extend
15898 @opindex mno-zero-extend
15899 When reading data from memory in sizes shorter than 64 bits, use (do not
15900 use) zero-extending load instructions by default, rather than
15901 sign-extending ones.
15904 @itemx -mno-knuthdiv
15906 @opindex mno-knuthdiv
15907 Make the result of a division yielding a remainder have the same sign as
15908 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15909 remainder follows the sign of the dividend. Both methods are
15910 arithmetically valid, the latter being almost exclusively used.
15912 @item -mtoplevel-symbols
15913 @itemx -mno-toplevel-symbols
15914 @opindex mtoplevel-symbols
15915 @opindex mno-toplevel-symbols
15916 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15917 code can be used with the @code{PREFIX} assembly directive.
15921 Generate an executable in the ELF format, rather than the default
15922 @samp{mmo} format used by the @command{mmix} simulator.
15924 @item -mbranch-predict
15925 @itemx -mno-branch-predict
15926 @opindex mbranch-predict
15927 @opindex mno-branch-predict
15928 Use (do not use) the probable-branch instructions, when static branch
15929 prediction indicates a probable branch.
15931 @item -mbase-addresses
15932 @itemx -mno-base-addresses
15933 @opindex mbase-addresses
15934 @opindex mno-base-addresses
15935 Generate (do not generate) code that uses @emph{base addresses}. Using a
15936 base address automatically generates a request (handled by the assembler
15937 and the linker) for a constant to be set up in a global register. The
15938 register is used for one or more base address requests within the range 0
15939 to 255 from the value held in the register. The generally leads to short
15940 and fast code, but the number of different data items that can be
15941 addressed is limited. This means that a program that uses lots of static
15942 data may require @option{-mno-base-addresses}.
15944 @item -msingle-exit
15945 @itemx -mno-single-exit
15946 @opindex msingle-exit
15947 @opindex mno-single-exit
15948 Force (do not force) generated code to have a single exit point in each
15952 @node MN10300 Options
15953 @subsection MN10300 Options
15954 @cindex MN10300 options
15956 These @option{-m} options are defined for Matsushita MN10300 architectures:
15961 Generate code to avoid bugs in the multiply instructions for the MN10300
15962 processors. This is the default.
15964 @item -mno-mult-bug
15965 @opindex mno-mult-bug
15966 Do not generate code to avoid bugs in the multiply instructions for the
15967 MN10300 processors.
15971 Generate code which uses features specific to the AM33 processor.
15975 Do not generate code which uses features specific to the AM33 processor. This
15980 Generate code which uses features specific to the AM33/2.0 processor.
15984 Generate code which uses features specific to the AM34 processor.
15986 @item -mtune=@var{cpu-type}
15988 Use the timing characteristics of the indicated CPU type when
15989 scheduling instructions. This does not change the targeted processor
15990 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15991 @samp{am33-2} or @samp{am34}.
15993 @item -mreturn-pointer-on-d0
15994 @opindex mreturn-pointer-on-d0
15995 When generating a function which returns a pointer, return the pointer
15996 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15997 only in a0, and attempts to call such functions without a prototype
15998 would result in errors. Note that this option is on by default; use
15999 @option{-mno-return-pointer-on-d0} to disable it.
16003 Do not link in the C run-time initialization object file.
16007 Indicate to the linker that it should perform a relaxation optimization pass
16008 to shorten branches, calls and absolute memory addresses. This option only
16009 has an effect when used on the command line for the final link step.
16011 This option makes symbolic debugging impossible.
16015 Allow the compiler to generate @emph{Long Instruction Word}
16016 instructions if the target is the @samp{AM33} or later. This is the
16017 default. This option defines the preprocessor macro @samp{__LIW__}.
16021 Do not allow the compiler to generate @emph{Long Instruction Word}
16022 instructions. This option defines the preprocessor macro
16027 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
16028 instructions if the target is the @samp{AM33} or later. This is the
16029 default. This option defines the preprocessor macro @samp{__SETLB__}.
16033 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
16034 instructions. This option defines the preprocessor macro
16035 @samp{__NO_SETLB__}.
16039 @node PDP-11 Options
16040 @subsection PDP-11 Options
16041 @cindex PDP-11 Options
16043 These options are defined for the PDP-11:
16048 Use hardware FPP floating point. This is the default. (FIS floating
16049 point on the PDP-11/40 is not supported.)
16052 @opindex msoft-float
16053 Do not use hardware floating point.
16057 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
16061 Return floating-point results in memory. This is the default.
16065 Generate code for a PDP-11/40.
16069 Generate code for a PDP-11/45. This is the default.
16073 Generate code for a PDP-11/10.
16075 @item -mbcopy-builtin
16076 @opindex mbcopy-builtin
16077 Use inline @code{movmemhi} patterns for copying memory. This is the
16082 Do not use inline @code{movmemhi} patterns for copying memory.
16088 Use 16-bit @code{int}. This is the default.
16094 Use 32-bit @code{int}.
16097 @itemx -mno-float32
16099 @opindex mno-float32
16100 Use 64-bit @code{float}. This is the default.
16103 @itemx -mno-float64
16105 @opindex mno-float64
16106 Use 32-bit @code{float}.
16110 Use @code{abshi2} pattern. This is the default.
16114 Do not use @code{abshi2} pattern.
16116 @item -mbranch-expensive
16117 @opindex mbranch-expensive
16118 Pretend that branches are expensive. This is for experimenting with
16119 code generation only.
16121 @item -mbranch-cheap
16122 @opindex mbranch-cheap
16123 Do not pretend that branches are expensive. This is the default.
16127 Use Unix assembler syntax. This is the default when configured for
16128 @samp{pdp11-*-bsd}.
16132 Use DEC assembler syntax. This is the default when configured for any
16133 PDP-11 target other than @samp{pdp11-*-bsd}.
16136 @node picoChip Options
16137 @subsection picoChip Options
16138 @cindex picoChip options
16140 These @samp{-m} options are defined for picoChip implementations:
16144 @item -mae=@var{ae_type}
16146 Set the instruction set, register set, and instruction scheduling
16147 parameters for array element type @var{ae_type}. Supported values
16148 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
16150 @option{-mae=ANY} selects a completely generic AE type. Code
16151 generated with this option will run on any of the other AE types. The
16152 code will not be as efficient as it would be if compiled for a specific
16153 AE type, and some types of operation (e.g., multiplication) will not
16154 work properly on all types of AE.
16156 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
16157 for compiled code, and is the default.
16159 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
16160 option may suffer from poor performance of byte (char) manipulation,
16161 since the DSP AE does not provide hardware support for byte load/stores.
16163 @item -msymbol-as-address
16164 Enable the compiler to directly use a symbol name as an address in a
16165 load/store instruction, without first loading it into a
16166 register. Typically, the use of this option will generate larger
16167 programs, which run faster than when the option isn't used. However, the
16168 results vary from program to program, so it is left as a user option,
16169 rather than being permanently enabled.
16171 @item -mno-inefficient-warnings
16172 Disables warnings about the generation of inefficient code. These
16173 warnings can be generated, for example, when compiling code which
16174 performs byte-level memory operations on the MAC AE type. The MAC AE has
16175 no hardware support for byte-level memory operations, so all byte
16176 load/stores must be synthesized from word load/store operations. This is
16177 inefficient and a warning will be generated indicating to the programmer
16178 that they should rewrite the code to avoid byte operations, or to target
16179 an AE type which has the necessary hardware support. This option enables
16180 the warning to be turned off.
16184 @node PowerPC Options
16185 @subsection PowerPC Options
16186 @cindex PowerPC options
16188 These are listed under @xref{RS/6000 and PowerPC Options}.
16191 @subsection RL78 Options
16192 @cindex RL78 Options
16198 Links in additional target libraries to support operation within a
16205 Specifies the type of hardware multiplication support to be used. The
16206 default is @code{none}, which uses software multiplication functions.
16207 The @code{g13} option is for the hardware multiply/divide peripheral
16208 only on the RL78/G13 targets. The @code{rl78} option is for the
16209 standard hardware multiplication defined in the RL78 software manual.
16213 @node RS/6000 and PowerPC Options
16214 @subsection IBM RS/6000 and PowerPC Options
16215 @cindex RS/6000 and PowerPC Options
16216 @cindex IBM RS/6000 and PowerPC Options
16218 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
16225 @itemx -mno-powerpc
16226 @itemx -mpowerpc-gpopt
16227 @itemx -mno-powerpc-gpopt
16228 @itemx -mpowerpc-gfxopt
16229 @itemx -mno-powerpc-gfxopt
16232 @itemx -mno-powerpc64
16236 @itemx -mno-popcntb
16238 @itemx -mno-popcntd
16247 @itemx -mno-hard-dfp
16251 @opindex mno-power2
16253 @opindex mno-powerpc
16254 @opindex mpowerpc-gpopt
16255 @opindex mno-powerpc-gpopt
16256 @opindex mpowerpc-gfxopt
16257 @opindex mno-powerpc-gfxopt
16258 @opindex mpowerpc64
16259 @opindex mno-powerpc64
16263 @opindex mno-popcntb
16265 @opindex mno-popcntd
16271 @opindex mno-mfpgpr
16273 @opindex mno-hard-dfp
16274 GCC supports two related instruction set architectures for the
16275 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
16276 instructions supported by the @samp{rios} chip set used in the original
16277 RS/6000 systems and the @dfn{PowerPC} instruction set is the
16278 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
16279 the IBM 4xx, 6xx, and follow-on microprocessors.
16281 Neither architecture is a subset of the other. However there is a
16282 large common subset of instructions supported by both. An MQ
16283 register is included in processors supporting the POWER architecture.
16285 You use these options to specify which instructions are available on the
16286 processor you are using. The default value of these options is
16287 determined when configuring GCC@. Specifying the
16288 @option{-mcpu=@var{cpu_type}} overrides the specification of these
16289 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
16290 rather than the options listed above.
16292 The @option{-mpower} option allows GCC to generate instructions that
16293 are found only in the POWER architecture and to use the MQ register.
16294 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
16295 to generate instructions that are present in the POWER2 architecture but
16296 not the original POWER architecture.
16298 The @option{-mpowerpc} option allows GCC to generate instructions that
16299 are found only in the 32-bit subset of the PowerPC architecture.
16300 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
16301 GCC to use the optional PowerPC architecture instructions in the
16302 General Purpose group, including floating-point square root. Specifying
16303 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
16304 use the optional PowerPC architecture instructions in the Graphics
16305 group, including floating-point select.
16307 The @option{-mmfcrf} option allows GCC to generate the move from
16308 condition register field instruction implemented on the POWER4
16309 processor and other processors that support the PowerPC V2.01
16311 The @option{-mpopcntb} option allows GCC to generate the popcount and
16312 double-precision FP reciprocal estimate instruction implemented on the
16313 POWER5 processor and other processors that support the PowerPC V2.02
16315 The @option{-mpopcntd} option allows GCC to generate the popcount
16316 instruction implemented on the POWER7 processor and other processors
16317 that support the PowerPC V2.06 architecture.
16318 The @option{-mfprnd} option allows GCC to generate the FP round to
16319 integer instructions implemented on the POWER5+ processor and other
16320 processors that support the PowerPC V2.03 architecture.
16321 The @option{-mcmpb} option allows GCC to generate the compare bytes
16322 instruction implemented on the POWER6 processor and other processors
16323 that support the PowerPC V2.05 architecture.
16324 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
16325 general-purpose register instructions implemented on the POWER6X
16326 processor and other processors that support the extended PowerPC V2.05
16328 The @option{-mhard-dfp} option allows GCC to generate the decimal
16329 floating-point instructions implemented on some POWER processors.
16331 The @option{-mpowerpc64} option allows GCC to generate the additional
16332 64-bit instructions that are found in the full PowerPC64 architecture
16333 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
16334 @option{-mno-powerpc64}.
16336 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
16337 will use only the instructions in the common subset of both
16338 architectures plus some special AIX common-mode calls, and will not use
16339 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
16340 permits GCC to use any instruction from either architecture and to
16341 allow use of the MQ register; specify this for the Motorola MPC601.
16343 @item -mnew-mnemonics
16344 @itemx -mold-mnemonics
16345 @opindex mnew-mnemonics
16346 @opindex mold-mnemonics
16347 Select which mnemonics to use in the generated assembler code. With
16348 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
16349 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
16350 assembler mnemonics defined for the POWER architecture. Instructions
16351 defined in only one architecture have only one mnemonic; GCC uses that
16352 mnemonic irrespective of which of these options is specified.
16354 GCC defaults to the mnemonics appropriate for the architecture in
16355 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
16356 value of these option. Unless you are building a cross-compiler, you
16357 should normally not specify either @option{-mnew-mnemonics} or
16358 @option{-mold-mnemonics}, but should instead accept the default.
16360 @item -mcpu=@var{cpu_type}
16362 Set architecture type, register usage, choice of mnemonics, and
16363 instruction scheduling parameters for machine type @var{cpu_type}.
16364 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
16365 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
16366 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
16367 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
16368 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
16369 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
16370 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
16371 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
16372 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
16373 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
16374 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
16376 @option{-mcpu=common} selects a completely generic processor. Code
16377 generated under this option will run on any POWER or PowerPC processor.
16378 GCC will use only the instructions in the common subset of both
16379 architectures, and will not use the MQ register. GCC assumes a generic
16380 processor model for scheduling purposes.
16382 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
16383 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
16384 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
16385 types, with an appropriate, generic processor model assumed for
16386 scheduling purposes.
16388 The other options specify a specific processor. Code generated under
16389 those options will run best on that processor, and may not run at all on
16392 The @option{-mcpu} options automatically enable or disable the
16395 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
16396 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
16397 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
16398 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
16400 The particular options set for any particular CPU will vary between
16401 compiler versions, depending on what setting seems to produce optimal
16402 code for that CPU; it doesn't necessarily reflect the actual hardware's
16403 capabilities. If you wish to set an individual option to a particular
16404 value, you may specify it after the @option{-mcpu} option, like
16405 @samp{-mcpu=970 -mno-altivec}.
16407 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
16408 not enabled or disabled by the @option{-mcpu} option at present because
16409 AIX does not have full support for these options. You may still
16410 enable or disable them individually if you're sure it'll work in your
16413 @item -mtune=@var{cpu_type}
16415 Set the instruction scheduling parameters for machine type
16416 @var{cpu_type}, but do not set the architecture type, register usage, or
16417 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
16418 values for @var{cpu_type} are used for @option{-mtune} as for
16419 @option{-mcpu}. If both are specified, the code generated will use the
16420 architecture, registers, and mnemonics set by @option{-mcpu}, but the
16421 scheduling parameters set by @option{-mtune}.
16423 @item -mcmodel=small
16424 @opindex mcmodel=small
16425 Generate PowerPC64 code for the small model: The TOC is limited to
16428 @item -mcmodel=medium
16429 @opindex mcmodel=medium
16430 Generate PowerPC64 code for the medium model: The TOC and other static
16431 data may be up to a total of 4G in size.
16433 @item -mcmodel=large
16434 @opindex mcmodel=large
16435 Generate PowerPC64 code for the large model: The TOC may be up to 4G
16436 in size. Other data and code is only limited by the 64-bit address
16440 @itemx -mno-altivec
16442 @opindex mno-altivec
16443 Generate code that uses (does not use) AltiVec instructions, and also
16444 enable the use of built-in functions that allow more direct access to
16445 the AltiVec instruction set. You may also need to set
16446 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
16452 @opindex mno-vrsave
16453 Generate VRSAVE instructions when generating AltiVec code.
16455 @item -mgen-cell-microcode
16456 @opindex mgen-cell-microcode
16457 Generate Cell microcode instructions
16459 @item -mwarn-cell-microcode
16460 @opindex mwarn-cell-microcode
16461 Warning when a Cell microcode instruction is going to emitted. An example
16462 of a Cell microcode instruction is a variable shift.
16465 @opindex msecure-plt
16466 Generate code that allows ld and ld.so to build executables and shared
16467 libraries with non-exec .plt and .got sections. This is a PowerPC
16468 32-bit SYSV ABI option.
16472 Generate code that uses a BSS .plt section that ld.so fills in, and
16473 requires .plt and .got sections that are both writable and executable.
16474 This is a PowerPC 32-bit SYSV ABI option.
16480 This switch enables or disables the generation of ISEL instructions.
16482 @item -misel=@var{yes/no}
16483 This switch has been deprecated. Use @option{-misel} and
16484 @option{-mno-isel} instead.
16490 This switch enables or disables the generation of SPE simd
16496 @opindex mno-paired
16497 This switch enables or disables the generation of PAIRED simd
16500 @item -mspe=@var{yes/no}
16501 This option has been deprecated. Use @option{-mspe} and
16502 @option{-mno-spe} instead.
16508 Generate code that uses (does not use) vector/scalar (VSX)
16509 instructions, and also enable the use of built-in functions that allow
16510 more direct access to the VSX instruction set.
16512 @item -mfloat-gprs=@var{yes/single/double/no}
16513 @itemx -mfloat-gprs
16514 @opindex mfloat-gprs
16515 This switch enables or disables the generation of floating-point
16516 operations on the general-purpose registers for architectures that
16519 The argument @var{yes} or @var{single} enables the use of
16520 single-precision floating-point operations.
16522 The argument @var{double} enables the use of single and
16523 double-precision floating-point operations.
16525 The argument @var{no} disables floating-point operations on the
16526 general-purpose registers.
16528 This option is currently only available on the MPC854x.
16534 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
16535 targets (including GNU/Linux). The 32-bit environment sets int, long
16536 and pointer to 32 bits and generates code that runs on any PowerPC
16537 variant. The 64-bit environment sets int to 32 bits and long and
16538 pointer to 64 bits, and generates code for PowerPC64, as for
16539 @option{-mpowerpc64}.
16542 @itemx -mno-fp-in-toc
16543 @itemx -mno-sum-in-toc
16544 @itemx -mminimal-toc
16546 @opindex mno-fp-in-toc
16547 @opindex mno-sum-in-toc
16548 @opindex mminimal-toc
16549 Modify generation of the TOC (Table Of Contents), which is created for
16550 every executable file. The @option{-mfull-toc} option is selected by
16551 default. In that case, GCC will allocate at least one TOC entry for
16552 each unique non-automatic variable reference in your program. GCC
16553 will also place floating-point constants in the TOC@. However, only
16554 16,384 entries are available in the TOC@.
16556 If you receive a linker error message that saying you have overflowed
16557 the available TOC space, you can reduce the amount of TOC space used
16558 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
16559 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
16560 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
16561 generate code to calculate the sum of an address and a constant at
16562 run time instead of putting that sum into the TOC@. You may specify one
16563 or both of these options. Each causes GCC to produce very slightly
16564 slower and larger code at the expense of conserving TOC space.
16566 If you still run out of space in the TOC even when you specify both of
16567 these options, specify @option{-mminimal-toc} instead. This option causes
16568 GCC to make only one TOC entry for every file. When you specify this
16569 option, GCC will produce code that is slower and larger but which
16570 uses extremely little TOC space. You may wish to use this option
16571 only on files that contain less frequently executed code.
16577 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
16578 @code{long} type, and the infrastructure needed to support them.
16579 Specifying @option{-maix64} implies @option{-mpowerpc64} and
16580 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
16581 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
16584 @itemx -mno-xl-compat
16585 @opindex mxl-compat
16586 @opindex mno-xl-compat
16587 Produce code that conforms more closely to IBM XL compiler semantics
16588 when using AIX-compatible ABI@. Pass floating-point arguments to
16589 prototyped functions beyond the register save area (RSA) on the stack
16590 in addition to argument FPRs. Do not assume that most significant
16591 double in 128-bit long double value is properly rounded when comparing
16592 values and converting to double. Use XL symbol names for long double
16595 The AIX calling convention was extended but not initially documented to
16596 handle an obscure K&R C case of calling a function that takes the
16597 address of its arguments with fewer arguments than declared. IBM XL
16598 compilers access floating-point arguments that do not fit in the
16599 RSA from the stack when a subroutine is compiled without
16600 optimization. Because always storing floating-point arguments on the
16601 stack is inefficient and rarely needed, this option is not enabled by
16602 default and only is necessary when calling subroutines compiled by IBM
16603 XL compilers without optimization.
16607 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
16608 application written to use message passing with special startup code to
16609 enable the application to run. The system must have PE installed in the
16610 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
16611 must be overridden with the @option{-specs=} option to specify the
16612 appropriate directory location. The Parallel Environment does not
16613 support threads, so the @option{-mpe} option and the @option{-pthread}
16614 option are incompatible.
16616 @item -malign-natural
16617 @itemx -malign-power
16618 @opindex malign-natural
16619 @opindex malign-power
16620 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
16621 @option{-malign-natural} overrides the ABI-defined alignment of larger
16622 types, such as floating-point doubles, on their natural size-based boundary.
16623 The option @option{-malign-power} instructs GCC to follow the ABI-specified
16624 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
16626 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
16630 @itemx -mhard-float
16631 @opindex msoft-float
16632 @opindex mhard-float
16633 Generate code that does not use (uses) the floating-point register set.
16634 Software floating-point emulation is provided if you use the
16635 @option{-msoft-float} option, and pass the option to GCC when linking.
16637 @item -msingle-float
16638 @itemx -mdouble-float
16639 @opindex msingle-float
16640 @opindex mdouble-float
16641 Generate code for single- or double-precision floating-point operations.
16642 @option{-mdouble-float} implies @option{-msingle-float}.
16645 @opindex msimple-fpu
16646 Do not generate sqrt and div instructions for hardware floating-point unit.
16650 Specify type of floating-point unit. Valid values are @var{sp_lite}
16651 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
16652 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
16653 and @var{dp_full} (equivalent to -mdouble-float).
16656 @opindex mxilinx-fpu
16657 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
16660 @itemx -mno-multiple
16662 @opindex mno-multiple
16663 Generate code that uses (does not use) the load multiple word
16664 instructions and the store multiple word instructions. These
16665 instructions are generated by default on POWER systems, and not
16666 generated on PowerPC systems. Do not use @option{-mmultiple} on little
16667 endian PowerPC systems, since those instructions do not work when the
16668 processor is in little endian mode. The exceptions are PPC740 and
16669 PPC750 which permit the instructions usage in little endian mode.
16674 @opindex mno-string
16675 Generate code that uses (does not use) the load string instructions
16676 and the store string word instructions to save multiple registers and
16677 do small block moves. These instructions are generated by default on
16678 POWER systems, and not generated on PowerPC systems. Do not use
16679 @option{-mstring} on little endian PowerPC systems, since those
16680 instructions do not work when the processor is in little endian mode.
16681 The exceptions are PPC740 and PPC750 which permit the instructions
16682 usage in little endian mode.
16687 @opindex mno-update
16688 Generate code that uses (does not use) the load or store instructions
16689 that update the base register to the address of the calculated memory
16690 location. These instructions are generated by default. If you use
16691 @option{-mno-update}, there is a small window between the time that the
16692 stack pointer is updated and the address of the previous frame is
16693 stored, which means code that walks the stack frame across interrupts or
16694 signals may get corrupted data.
16696 @item -mavoid-indexed-addresses
16697 @itemx -mno-avoid-indexed-addresses
16698 @opindex mavoid-indexed-addresses
16699 @opindex mno-avoid-indexed-addresses
16700 Generate code that tries to avoid (not avoid) the use of indexed load
16701 or store instructions. These instructions can incur a performance
16702 penalty on Power6 processors in certain situations, such as when
16703 stepping through large arrays that cross a 16M boundary. This option
16704 is enabled by default when targetting Power6 and disabled otherwise.
16707 @itemx -mno-fused-madd
16708 @opindex mfused-madd
16709 @opindex mno-fused-madd
16710 Generate code that uses (does not use) the floating-point multiply and
16711 accumulate instructions. These instructions are generated by default
16712 if hardware floating point is used. The machine-dependent
16713 @option{-mfused-madd} option is now mapped to the machine-independent
16714 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16715 mapped to @option{-ffp-contract=off}.
16721 Generate code that uses (does not use) the half-word multiply and
16722 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
16723 These instructions are generated by default when targetting those
16730 Generate code that uses (does not use) the string-search @samp{dlmzb}
16731 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
16732 generated by default when targetting those processors.
16734 @item -mno-bit-align
16736 @opindex mno-bit-align
16737 @opindex mbit-align
16738 On System V.4 and embedded PowerPC systems do not (do) force structures
16739 and unions that contain bit-fields to be aligned to the base type of the
16742 For example, by default a structure containing nothing but 8
16743 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
16744 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
16745 the structure would be aligned to a 1 byte boundary and be one byte in
16748 @item -mno-strict-align
16749 @itemx -mstrict-align
16750 @opindex mno-strict-align
16751 @opindex mstrict-align
16752 On System V.4 and embedded PowerPC systems do not (do) assume that
16753 unaligned memory references will be handled by the system.
16755 @item -mrelocatable
16756 @itemx -mno-relocatable
16757 @opindex mrelocatable
16758 @opindex mno-relocatable
16759 Generate code that allows (does not allow) a static executable to be
16760 relocated to a different address at run time. A simple embedded
16761 PowerPC system loader should relocate the entire contents of
16762 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16763 a table of 32-bit addresses generated by this option. For this to
16764 work, all objects linked together must be compiled with
16765 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16766 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
16768 @item -mrelocatable-lib
16769 @itemx -mno-relocatable-lib
16770 @opindex mrelocatable-lib
16771 @opindex mno-relocatable-lib
16772 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16773 @code{.fixup} section to allow static executables to be relocated at
16774 run time, but @option{-mrelocatable-lib} does not use the smaller stack
16775 alignment of @option{-mrelocatable}. Objects compiled with
16776 @option{-mrelocatable-lib} may be linked with objects compiled with
16777 any combination of the @option{-mrelocatable} options.
16783 On System V.4 and embedded PowerPC systems do not (do) assume that
16784 register 2 contains a pointer to a global area pointing to the addresses
16785 used in the program.
16788 @itemx -mlittle-endian
16790 @opindex mlittle-endian
16791 On System V.4 and embedded PowerPC systems compile code for the
16792 processor in little endian mode. The @option{-mlittle-endian} option is
16793 the same as @option{-mlittle}.
16796 @itemx -mbig-endian
16798 @opindex mbig-endian
16799 On System V.4 and embedded PowerPC systems compile code for the
16800 processor in big endian mode. The @option{-mbig-endian} option is
16801 the same as @option{-mbig}.
16803 @item -mdynamic-no-pic
16804 @opindex mdynamic-no-pic
16805 On Darwin and Mac OS X systems, compile code so that it is not
16806 relocatable, but that its external references are relocatable. The
16807 resulting code is suitable for applications, but not shared
16810 @item -msingle-pic-base
16811 @opindex msingle-pic-base
16812 Treat the register used for PIC addressing as read-only, rather than
16813 loading it in the prologue for each function. The runtime system is
16814 responsible for initializing this register with an appropriate value
16815 before execution begins.
16817 @item -mprioritize-restricted-insns=@var{priority}
16818 @opindex mprioritize-restricted-insns
16819 This option controls the priority that is assigned to
16820 dispatch-slot restricted instructions during the second scheduling
16821 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16822 @var{no/highest/second-highest} priority to dispatch slot restricted
16825 @item -msched-costly-dep=@var{dependence_type}
16826 @opindex msched-costly-dep
16827 This option controls which dependences are considered costly
16828 by the target during instruction scheduling. The argument
16829 @var{dependence_type} takes one of the following values:
16830 @var{no}: no dependence is costly,
16831 @var{all}: all dependences are costly,
16832 @var{true_store_to_load}: a true dependence from store to load is costly,
16833 @var{store_to_load}: any dependence from store to load is costly,
16834 @var{number}: any dependence which latency >= @var{number} is costly.
16836 @item -minsert-sched-nops=@var{scheme}
16837 @opindex minsert-sched-nops
16838 This option controls which nop insertion scheme will be used during
16839 the second scheduling pass. The argument @var{scheme} takes one of the
16841 @var{no}: Don't insert nops.
16842 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16843 according to the scheduler's grouping.
16844 @var{regroup_exact}: Insert nops to force costly dependent insns into
16845 separate groups. Insert exactly as many nops as needed to force an insn
16846 to a new group, according to the estimated processor grouping.
16847 @var{number}: Insert nops to force costly dependent insns into
16848 separate groups. Insert @var{number} nops to force an insn to a new group.
16851 @opindex mcall-sysv
16852 On System V.4 and embedded PowerPC systems compile code using calling
16853 conventions that adheres to the March 1995 draft of the System V
16854 Application Binary Interface, PowerPC processor supplement. This is the
16855 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16857 @item -mcall-sysv-eabi
16859 @opindex mcall-sysv-eabi
16860 @opindex mcall-eabi
16861 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16863 @item -mcall-sysv-noeabi
16864 @opindex mcall-sysv-noeabi
16865 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16867 @item -mcall-aixdesc
16869 On System V.4 and embedded PowerPC systems compile code for the AIX
16873 @opindex mcall-linux
16874 On System V.4 and embedded PowerPC systems compile code for the
16875 Linux-based GNU system.
16877 @item -mcall-freebsd
16878 @opindex mcall-freebsd
16879 On System V.4 and embedded PowerPC systems compile code for the
16880 FreeBSD operating system.
16882 @item -mcall-netbsd
16883 @opindex mcall-netbsd
16884 On System V.4 and embedded PowerPC systems compile code for the
16885 NetBSD operating system.
16887 @item -mcall-openbsd
16888 @opindex mcall-netbsd
16889 On System V.4 and embedded PowerPC systems compile code for the
16890 OpenBSD operating system.
16892 @item -maix-struct-return
16893 @opindex maix-struct-return
16894 Return all structures in memory (as specified by the AIX ABI)@.
16896 @item -msvr4-struct-return
16897 @opindex msvr4-struct-return
16898 Return structures smaller than 8 bytes in registers (as specified by the
16901 @item -mabi=@var{abi-type}
16903 Extend the current ABI with a particular extension, or remove such extension.
16904 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16905 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16909 Extend the current ABI with SPE ABI extensions. This does not change
16910 the default ABI, instead it adds the SPE ABI extensions to the current
16914 @opindex mabi=no-spe
16915 Disable Booke SPE ABI extensions for the current ABI@.
16917 @item -mabi=ibmlongdouble
16918 @opindex mabi=ibmlongdouble
16919 Change the current ABI to use IBM extended-precision long double.
16920 This is a PowerPC 32-bit SYSV ABI option.
16922 @item -mabi=ieeelongdouble
16923 @opindex mabi=ieeelongdouble
16924 Change the current ABI to use IEEE extended-precision long double.
16925 This is a PowerPC 32-bit Linux ABI option.
16928 @itemx -mno-prototype
16929 @opindex mprototype
16930 @opindex mno-prototype
16931 On System V.4 and embedded PowerPC systems assume that all calls to
16932 variable argument functions are properly prototyped. Otherwise, the
16933 compiler must insert an instruction before every non prototyped call to
16934 set or clear bit 6 of the condition code register (@var{CR}) to
16935 indicate whether floating-point values were passed in the floating-point
16936 registers in case the function takes variable arguments. With
16937 @option{-mprototype}, only calls to prototyped variable argument functions
16938 will set or clear the bit.
16942 On embedded PowerPC systems, assume that the startup module is called
16943 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16944 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16949 On embedded PowerPC systems, assume that the startup module is called
16950 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16955 On embedded PowerPC systems, assume that the startup module is called
16956 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16959 @item -myellowknife
16960 @opindex myellowknife
16961 On embedded PowerPC systems, assume that the startup module is called
16962 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16967 On System V.4 and embedded PowerPC systems, specify that you are
16968 compiling for a VxWorks system.
16972 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16973 header to indicate that @samp{eabi} extended relocations are used.
16979 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16980 Embedded Applications Binary Interface (eabi) which is a set of
16981 modifications to the System V.4 specifications. Selecting @option{-meabi}
16982 means that the stack is aligned to an 8 byte boundary, a function
16983 @code{__eabi} is called to from @code{main} to set up the eabi
16984 environment, and the @option{-msdata} option can use both @code{r2} and
16985 @code{r13} to point to two separate small data areas. Selecting
16986 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16987 do not call an initialization function from @code{main}, and the
16988 @option{-msdata} option will only use @code{r13} to point to a single
16989 small data area. The @option{-meabi} option is on by default if you
16990 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16993 @opindex msdata=eabi
16994 On System V.4 and embedded PowerPC systems, put small initialized
16995 @code{const} global and static data in the @samp{.sdata2} section, which
16996 is pointed to by register @code{r2}. Put small initialized
16997 non-@code{const} global and static data in the @samp{.sdata} section,
16998 which is pointed to by register @code{r13}. Put small uninitialized
16999 global and static data in the @samp{.sbss} section, which is adjacent to
17000 the @samp{.sdata} section. The @option{-msdata=eabi} option is
17001 incompatible with the @option{-mrelocatable} option. The
17002 @option{-msdata=eabi} option also sets the @option{-memb} option.
17005 @opindex msdata=sysv
17006 On System V.4 and embedded PowerPC systems, put small global and static
17007 data in the @samp{.sdata} section, which is pointed to by register
17008 @code{r13}. Put small uninitialized global and static data in the
17009 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
17010 The @option{-msdata=sysv} option is incompatible with the
17011 @option{-mrelocatable} option.
17013 @item -msdata=default
17015 @opindex msdata=default
17017 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
17018 compile code the same as @option{-msdata=eabi}, otherwise compile code the
17019 same as @option{-msdata=sysv}.
17022 @opindex msdata=data
17023 On System V.4 and embedded PowerPC systems, put small global
17024 data in the @samp{.sdata} section. Put small uninitialized global
17025 data in the @samp{.sbss} section. Do not use register @code{r13}
17026 to address small data however. This is the default behavior unless
17027 other @option{-msdata} options are used.
17031 @opindex msdata=none
17033 On embedded PowerPC systems, put all initialized global and static data
17034 in the @samp{.data} section, and all uninitialized data in the
17035 @samp{.bss} section.
17037 @item -mblock-move-inline-limit=@var{num}
17038 @opindex mblock-move-inline-limit
17039 Inline all block moves (such as calls to @code{memcpy} or structure
17040 copies) less than or equal to @var{num} bytes. The minimum value for
17041 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
17042 targets. The default value is target-specific.
17046 @cindex smaller data references (PowerPC)
17047 @cindex .sdata/.sdata2 references (PowerPC)
17048 On embedded PowerPC systems, put global and static items less than or
17049 equal to @var{num} bytes into the small data or bss sections instead of
17050 the normal data or bss section. By default, @var{num} is 8. The
17051 @option{-G @var{num}} switch is also passed to the linker.
17052 All modules should be compiled with the same @option{-G @var{num}} value.
17055 @itemx -mno-regnames
17057 @opindex mno-regnames
17058 On System V.4 and embedded PowerPC systems do (do not) emit register
17059 names in the assembly language output using symbolic forms.
17062 @itemx -mno-longcall
17064 @opindex mno-longcall
17065 By default assume that all calls are far away so that a longer more
17066 expensive calling sequence is required. This is required for calls
17067 further than 32 megabytes (33,554,432 bytes) from the current location.
17068 A short call will be generated if the compiler knows
17069 the call cannot be that far away. This setting can be overridden by
17070 the @code{shortcall} function attribute, or by @code{#pragma
17073 Some linkers are capable of detecting out-of-range calls and generating
17074 glue code on the fly. On these systems, long calls are unnecessary and
17075 generate slower code. As of this writing, the AIX linker can do this,
17076 as can the GNU linker for PowerPC/64. It is planned to add this feature
17077 to the GNU linker for 32-bit PowerPC systems as well.
17079 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
17080 callee, L42'', plus a ``branch island'' (glue code). The two target
17081 addresses represent the callee and the ``branch island''. The
17082 Darwin/PPC linker will prefer the first address and generate a ``bl
17083 callee'' if the PPC ``bl'' instruction will reach the callee directly;
17084 otherwise, the linker will generate ``bl L42'' to call the ``branch
17085 island''. The ``branch island'' is appended to the body of the
17086 calling function; it computes the full 32-bit address of the callee
17089 On Mach-O (Darwin) systems, this option directs the compiler emit to
17090 the glue for every direct call, and the Darwin linker decides whether
17091 to use or discard it.
17093 In the future, we may cause GCC to ignore all longcall specifications
17094 when the linker is known to generate glue.
17096 @item -mtls-markers
17097 @itemx -mno-tls-markers
17098 @opindex mtls-markers
17099 @opindex mno-tls-markers
17100 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
17101 specifying the function argument. The relocation allows ld to
17102 reliably associate function call with argument setup instructions for
17103 TLS optimization, which in turn allows gcc to better schedule the
17108 Adds support for multithreading with the @dfn{pthreads} library.
17109 This option sets flags for both the preprocessor and linker.
17114 This option will enable GCC to use the reciprocal estimate and
17115 reciprocal square root estimate instructions with additional
17116 Newton-Raphson steps to increase precision instead of doing a divide or
17117 square root and divide for floating-point arguments. You should use
17118 the @option{-ffast-math} option when using @option{-mrecip} (or at
17119 least @option{-funsafe-math-optimizations},
17120 @option{-finite-math-only}, @option{-freciprocal-math} and
17121 @option{-fno-trapping-math}). Note that while the throughput of the
17122 sequence is generally higher than the throughput of the non-reciprocal
17123 instruction, the precision of the sequence can be decreased by up to 2
17124 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
17127 @item -mrecip=@var{opt}
17128 @opindex mrecip=opt
17129 This option allows to control which reciprocal estimate instructions
17130 may be used. @var{opt} is a comma separated list of options, that may
17131 be preceded by a @code{!} to invert the option:
17132 @code{all}: enable all estimate instructions,
17133 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
17134 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
17135 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
17136 @code{divf}: enable the single-precision reciprocal approximation instructions;
17137 @code{divd}: enable the double-precision reciprocal approximation instructions;
17138 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
17139 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
17140 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
17142 So for example, @option{-mrecip=all,!rsqrtd} would enable the
17143 all of the reciprocal estimate instructions, except for the
17144 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
17145 which handle the double-precision reciprocal square root calculations.
17147 @item -mrecip-precision
17148 @itemx -mno-recip-precision
17149 @opindex mrecip-precision
17150 Assume (do not assume) that the reciprocal estimate instructions
17151 provide higher-precision estimates than is mandated by the PowerPC
17152 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
17153 automatically selects @option{-mrecip-precision}. The double-precision
17154 square root estimate instructions are not generated by
17155 default on low-precision machines, since they do not provide an
17156 estimate that converges after three steps.
17158 @item -mveclibabi=@var{type}
17159 @opindex mveclibabi
17160 Specifies the ABI type to use for vectorizing intrinsics using an
17161 external library. The only type supported at present is @code{mass},
17162 which specifies to use IBM's Mathematical Acceleration Subsystem
17163 (MASS) libraries for vectorizing intrinsics using external libraries.
17164 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
17165 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
17166 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
17167 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
17168 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
17169 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
17170 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
17171 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
17172 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
17173 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
17174 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
17175 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
17176 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
17177 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
17178 for power7. Both @option{-ftree-vectorize} and
17179 @option{-funsafe-math-optimizations} have to be enabled. The MASS
17180 libraries will have to be specified at link time.
17185 Generate (do not generate) the @code{friz} instruction when the
17186 @option{-funsafe-math-optimizations} option is used to optimize
17187 rounding of floating-point values to 64-bit integer and back to floating
17188 point. The @code{friz} instruction does not return the same value if
17189 the floating-point number is too large to fit in an integer.
17191 @item -mpointers-to-nested-functions
17192 @itemx -mno-pointers-to-nested-functions
17193 @opindex mpointers-to-nested-functions
17194 Generate (do not generate) code to load up the static chain register
17195 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
17196 systems where a function pointer points to a 3 word descriptor giving
17197 the function address, TOC value to be loaded in register @var{r2}, and
17198 static chain value to be loaded in register @var{r11}. The
17199 @option{-mpointers-to-nested-functions} is on by default. You will
17200 not be able to call through pointers to nested functions or pointers
17201 to functions compiled in other languages that use the static chain if
17202 you use the @option{-mno-pointers-to-nested-functions}.
17204 @item -msave-toc-indirect
17205 @itemx -mno-save-toc-indirect
17206 @opindex msave-toc-indirect
17207 Generate (do not generate) code to save the TOC value in the reserved
17208 stack location in the function prologue if the function calls through
17209 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
17210 saved in the prologue, it is saved just before the call through the
17211 pointer. The @option{-mno-save-toc-indirect} option is the default.
17215 @subsection RX Options
17218 These command-line options are defined for RX targets:
17221 @item -m64bit-doubles
17222 @itemx -m32bit-doubles
17223 @opindex m64bit-doubles
17224 @opindex m32bit-doubles
17225 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
17226 or 32 bits (@option{-m32bit-doubles}) in size. The default is
17227 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
17228 works on 32-bit values, which is why the default is
17229 @option{-m32bit-doubles}.
17235 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
17236 floating-point hardware. The default is enabled for the @var{RX600}
17237 series and disabled for the @var{RX200} series.
17239 Floating-point instructions will only be generated for 32-bit floating-point
17240 values however, so if the @option{-m64bit-doubles} option is in
17241 use then the FPU hardware will not be used for doubles.
17243 @emph{Note} If the @option{-fpu} option is enabled then
17244 @option{-funsafe-math-optimizations} is also enabled automatically.
17245 This is because the RX FPU instructions are themselves unsafe.
17247 @item -mcpu=@var{name}
17249 Selects the type of RX CPU to be targeted. Currently three types are
17250 supported, the generic @var{RX600} and @var{RX200} series hardware and
17251 the specific @var{RX610} CPU. The default is @var{RX600}.
17253 The only difference between @var{RX600} and @var{RX610} is that the
17254 @var{RX610} does not support the @code{MVTIPL} instruction.
17256 The @var{RX200} series does not have a hardware floating-point unit
17257 and so @option{-nofpu} is enabled by default when this type is
17260 @item -mbig-endian-data
17261 @itemx -mlittle-endian-data
17262 @opindex mbig-endian-data
17263 @opindex mlittle-endian-data
17264 Store data (but not code) in the big-endian format. The default is
17265 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
17268 @item -msmall-data-limit=@var{N}
17269 @opindex msmall-data-limit
17270 Specifies the maximum size in bytes of global and static variables
17271 which can be placed into the small data area. Using the small data
17272 area can lead to smaller and faster code, but the size of area is
17273 limited and it is up to the programmer to ensure that the area does
17274 not overflow. Also when the small data area is used one of the RX's
17275 registers (usually @code{r13}) is reserved for use pointing to this
17276 area, so it is no longer available for use by the compiler. This
17277 could result in slower and/or larger code if variables which once
17278 could have been held in the reserved register are now pushed onto the
17281 Note, common variables (variables which have not been initialised) and
17282 constants are not placed into the small data area as they are assigned
17283 to other sections in the output executable.
17285 The default value is zero, which disables this feature. Note, this
17286 feature is not enabled by default with higher optimization levels
17287 (@option{-O2} etc) because of the potentially detrimental effects of
17288 reserving a register. It is up to the programmer to experiment and
17289 discover whether this feature is of benefit to their program. See the
17290 description of the @option{-mpid} option for a description of how the
17291 actual register to hold the small data area pointer is chosen.
17297 Use the simulator runtime. The default is to use the libgloss board
17300 @item -mas100-syntax
17301 @itemx -mno-as100-syntax
17302 @opindex mas100-syntax
17303 @opindex mno-as100-syntax
17304 When generating assembler output use a syntax that is compatible with
17305 Renesas's AS100 assembler. This syntax can also be handled by the GAS
17306 assembler but it has some restrictions so generating it is not the
17309 @item -mmax-constant-size=@var{N}
17310 @opindex mmax-constant-size
17311 Specifies the maximum size, in bytes, of a constant that can be used as
17312 an operand in a RX instruction. Although the RX instruction set does
17313 allow constants of up to 4 bytes in length to be used in instructions,
17314 a longer value equates to a longer instruction. Thus in some
17315 circumstances it can be beneficial to restrict the size of constants
17316 that are used in instructions. Constants that are too big are instead
17317 placed into a constant pool and referenced via register indirection.
17319 The value @var{N} can be between 0 and 4. A value of 0 (the default)
17320 or 4 means that constants of any size are allowed.
17324 Enable linker relaxation. Linker relaxation is a process whereby the
17325 linker will attempt to reduce the size of a program by finding shorter
17326 versions of various instructions. Disabled by default.
17328 @item -mint-register=@var{N}
17329 @opindex mint-register
17330 Specify the number of registers to reserve for fast interrupt handler
17331 functions. The value @var{N} can be between 0 and 4. A value of 1
17332 means that register @code{r13} will be reserved for the exclusive use
17333 of fast interrupt handlers. A value of 2 reserves @code{r13} and
17334 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
17335 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
17336 A value of 0, the default, does not reserve any registers.
17338 @item -msave-acc-in-interrupts
17339 @opindex msave-acc-in-interrupts
17340 Specifies that interrupt handler functions should preserve the
17341 accumulator register. This is only necessary if normal code might use
17342 the accumulator register, for example because it performs 64-bit
17343 multiplications. The default is to ignore the accumulator as this
17344 makes the interrupt handlers faster.
17350 Enables the generation of position independent data. When enabled any
17351 access to constant data will done via an offset from a base address
17352 held in a register. This allows the location of constant data to be
17353 determined at run time without requiring the executable to be
17354 relocated, which is a benefit to embedded applications with tight
17355 memory constraints. Data that can be modified is not affected by this
17358 Note, using this feature reserves a register, usually @code{r13}, for
17359 the constant data base address. This can result in slower and/or
17360 larger code, especially in complicated functions.
17362 The actual register chosen to hold the constant data base address
17363 depends upon whether the @option{-msmall-data-limit} and/or the
17364 @option{-mint-register} command-line options are enabled. Starting
17365 with register @code{r13} and proceeding downwards, registers are
17366 allocated first to satisfy the requirements of @option{-mint-register},
17367 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
17368 is possible for the small data area register to be @code{r8} if both
17369 @option{-mint-register=4} and @option{-mpid} are specified on the
17372 By default this feature is not enabled. The default can be restored
17373 via the @option{-mno-pid} command-line option.
17377 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
17378 has special significance to the RX port when used with the
17379 @code{interrupt} function attribute. This attribute indicates a
17380 function intended to process fast interrupts. GCC will will ensure
17381 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
17382 and/or @code{r13} and only provided that the normal use of the
17383 corresponding registers have been restricted via the
17384 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
17387 @node S/390 and zSeries Options
17388 @subsection S/390 and zSeries Options
17389 @cindex S/390 and zSeries Options
17391 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
17395 @itemx -msoft-float
17396 @opindex mhard-float
17397 @opindex msoft-float
17398 Use (do not use) the hardware floating-point instructions and registers
17399 for floating-point operations. When @option{-msoft-float} is specified,
17400 functions in @file{libgcc.a} will be used to perform floating-point
17401 operations. When @option{-mhard-float} is specified, the compiler
17402 generates IEEE floating-point instructions. This is the default.
17405 @itemx -mno-hard-dfp
17407 @opindex mno-hard-dfp
17408 Use (do not use) the hardware decimal-floating-point instructions for
17409 decimal-floating-point operations. When @option{-mno-hard-dfp} is
17410 specified, functions in @file{libgcc.a} will be used to perform
17411 decimal-floating-point operations. When @option{-mhard-dfp} is
17412 specified, the compiler generates decimal-floating-point hardware
17413 instructions. This is the default for @option{-march=z9-ec} or higher.
17415 @item -mlong-double-64
17416 @itemx -mlong-double-128
17417 @opindex mlong-double-64
17418 @opindex mlong-double-128
17419 These switches control the size of @code{long double} type. A size
17420 of 64bit makes the @code{long double} type equivalent to the @code{double}
17421 type. This is the default.
17424 @itemx -mno-backchain
17425 @opindex mbackchain
17426 @opindex mno-backchain
17427 Store (do not store) the address of the caller's frame as backchain pointer
17428 into the callee's stack frame.
17429 A backchain may be needed to allow debugging using tools that do not understand
17430 DWARF-2 call frame information.
17431 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
17432 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
17433 the backchain is placed into the topmost word of the 96/160 byte register
17436 In general, code compiled with @option{-mbackchain} is call-compatible with
17437 code compiled with @option{-mmo-backchain}; however, use of the backchain
17438 for debugging purposes usually requires that the whole binary is built with
17439 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
17440 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17441 to build a linux kernel use @option{-msoft-float}.
17443 The default is to not maintain the backchain.
17445 @item -mpacked-stack
17446 @itemx -mno-packed-stack
17447 @opindex mpacked-stack
17448 @opindex mno-packed-stack
17449 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
17450 specified, the compiler uses the all fields of the 96/160 byte register save
17451 area only for their default purpose; unused fields still take up stack space.
17452 When @option{-mpacked-stack} is specified, register save slots are densely
17453 packed at the top of the register save area; unused space is reused for other
17454 purposes, allowing for more efficient use of the available stack space.
17455 However, when @option{-mbackchain} is also in effect, the topmost word of
17456 the save area is always used to store the backchain, and the return address
17457 register is always saved two words below the backchain.
17459 As long as the stack frame backchain is not used, code generated with
17460 @option{-mpacked-stack} is call-compatible with code generated with
17461 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
17462 S/390 or zSeries generated code that uses the stack frame backchain at run
17463 time, not just for debugging purposes. Such code is not call-compatible
17464 with code compiled with @option{-mpacked-stack}. Also, note that the
17465 combination of @option{-mbackchain},
17466 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17467 to build a linux kernel use @option{-msoft-float}.
17469 The default is to not use the packed stack layout.
17472 @itemx -mno-small-exec
17473 @opindex msmall-exec
17474 @opindex mno-small-exec
17475 Generate (or do not generate) code using the @code{bras} instruction
17476 to do subroutine calls.
17477 This only works reliably if the total executable size does not
17478 exceed 64k. The default is to use the @code{basr} instruction instead,
17479 which does not have this limitation.
17485 When @option{-m31} is specified, generate code compliant to the
17486 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
17487 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
17488 particular to generate 64-bit instructions. For the @samp{s390}
17489 targets, the default is @option{-m31}, while the @samp{s390x}
17490 targets default to @option{-m64}.
17496 When @option{-mzarch} is specified, generate code using the
17497 instructions available on z/Architecture.
17498 When @option{-mesa} is specified, generate code using the
17499 instructions available on ESA/390. Note that @option{-mesa} is
17500 not possible with @option{-m64}.
17501 When generating code compliant to the GNU/Linux for S/390 ABI,
17502 the default is @option{-mesa}. When generating code compliant
17503 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
17509 Generate (or do not generate) code using the @code{mvcle} instruction
17510 to perform block moves. When @option{-mno-mvcle} is specified,
17511 use a @code{mvc} loop instead. This is the default unless optimizing for
17518 Print (or do not print) additional debug information when compiling.
17519 The default is to not print debug information.
17521 @item -march=@var{cpu-type}
17523 Generate code that will run on @var{cpu-type}, which is the name of a system
17524 representing a certain processor type. Possible values for
17525 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
17526 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
17527 When generating code using the instructions available on z/Architecture,
17528 the default is @option{-march=z900}. Otherwise, the default is
17529 @option{-march=g5}.
17531 @item -mtune=@var{cpu-type}
17533 Tune to @var{cpu-type} everything applicable about the generated code,
17534 except for the ABI and the set of available instructions.
17535 The list of @var{cpu-type} values is the same as for @option{-march}.
17536 The default is the value used for @option{-march}.
17539 @itemx -mno-tpf-trace
17540 @opindex mtpf-trace
17541 @opindex mno-tpf-trace
17542 Generate code that adds (does not add) in TPF OS specific branches to trace
17543 routines in the operating system. This option is off by default, even
17544 when compiling for the TPF OS@.
17547 @itemx -mno-fused-madd
17548 @opindex mfused-madd
17549 @opindex mno-fused-madd
17550 Generate code that uses (does not use) the floating-point multiply and
17551 accumulate instructions. These instructions are generated by default if
17552 hardware floating point is used.
17554 @item -mwarn-framesize=@var{framesize}
17555 @opindex mwarn-framesize
17556 Emit a warning if the current function exceeds the given frame size. Because
17557 this is a compile-time check it doesn't need to be a real problem when the program
17558 runs. It is intended to identify functions which most probably cause
17559 a stack overflow. It is useful to be used in an environment with limited stack
17560 size e.g.@: the linux kernel.
17562 @item -mwarn-dynamicstack
17563 @opindex mwarn-dynamicstack
17564 Emit a warning if the function calls alloca or uses dynamically
17565 sized arrays. This is generally a bad idea with a limited stack size.
17567 @item -mstack-guard=@var{stack-guard}
17568 @itemx -mstack-size=@var{stack-size}
17569 @opindex mstack-guard
17570 @opindex mstack-size
17571 If these options are provided the s390 back end emits additional instructions in
17572 the function prologue which trigger a trap if the stack size is @var{stack-guard}
17573 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
17574 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
17575 the frame size of the compiled function is chosen.
17576 These options are intended to be used to help debugging stack overflow problems.
17577 The additionally emitted code causes only little overhead and hence can also be
17578 used in production like systems without greater performance degradation. The given
17579 values have to be exact powers of 2 and @var{stack-size} has to be greater than
17580 @var{stack-guard} without exceeding 64k.
17581 In order to be efficient the extra code makes the assumption that the stack starts
17582 at an address aligned to the value given by @var{stack-size}.
17583 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
17586 @node Score Options
17587 @subsection Score Options
17588 @cindex Score Options
17590 These options are defined for Score implementations:
17595 Compile code for big endian mode. This is the default.
17599 Compile code for little endian mode.
17603 Disable generate bcnz instruction.
17607 Enable generate unaligned load and store instruction.
17611 Enable the use of multiply-accumulate instructions. Disabled by default.
17615 Specify the SCORE5 as the target architecture.
17619 Specify the SCORE5U of the target architecture.
17623 Specify the SCORE7 as the target architecture. This is the default.
17627 Specify the SCORE7D as the target architecture.
17631 @subsection SH Options
17633 These @samp{-m} options are defined for the SH implementations:
17638 Generate code for the SH1.
17642 Generate code for the SH2.
17645 Generate code for the SH2e.
17649 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
17650 that the floating-point unit is not used.
17652 @item -m2a-single-only
17653 @opindex m2a-single-only
17654 Generate code for the SH2a-FPU, in such a way that no double-precision
17655 floating-point operations are used.
17658 @opindex m2a-single
17659 Generate code for the SH2a-FPU assuming the floating-point unit is in
17660 single-precision mode by default.
17664 Generate code for the SH2a-FPU assuming the floating-point unit is in
17665 double-precision mode by default.
17669 Generate code for the SH3.
17673 Generate code for the SH3e.
17677 Generate code for the SH4 without a floating-point unit.
17679 @item -m4-single-only
17680 @opindex m4-single-only
17681 Generate code for the SH4 with a floating-point unit that only
17682 supports single-precision arithmetic.
17686 Generate code for the SH4 assuming the floating-point unit is in
17687 single-precision mode by default.
17691 Generate code for the SH4.
17695 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
17696 floating-point unit is not used.
17698 @item -m4a-single-only
17699 @opindex m4a-single-only
17700 Generate code for the SH4a, in such a way that no double-precision
17701 floating-point operations are used.
17704 @opindex m4a-single
17705 Generate code for the SH4a assuming the floating-point unit is in
17706 single-precision mode by default.
17710 Generate code for the SH4a.
17714 Same as @option{-m4a-nofpu}, except that it implicitly passes
17715 @option{-dsp} to the assembler. GCC doesn't generate any DSP
17716 instructions at the moment.
17720 Compile code for the processor in big endian mode.
17724 Compile code for the processor in little endian mode.
17728 Align doubles at 64-bit boundaries. Note that this changes the calling
17729 conventions, and thus some functions from the standard C library will
17730 not work unless you recompile it first with @option{-mdalign}.
17734 Shorten some address references at link time, when possible; uses the
17735 linker option @option{-relax}.
17739 Use 32-bit offsets in @code{switch} tables. The default is to use
17744 Enable the use of bit manipulation instructions on SH2A.
17748 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
17749 alignment constraints.
17753 Comply with the calling conventions defined by Renesas.
17757 Comply with the calling conventions defined by Renesas.
17761 Comply with the calling conventions defined for GCC before the Renesas
17762 conventions were available. This option is the default for all
17763 targets of the SH toolchain.
17766 @opindex mnomacsave
17767 Mark the @code{MAC} register as call-clobbered, even if
17768 @option{-mhitachi} is given.
17772 Increase IEEE compliance of floating-point code.
17773 At the moment, this is equivalent to @option{-fno-finite-math-only}.
17774 When generating 16 bit SH opcodes, getting IEEE-conforming results for
17775 comparisons of NANs / infinities incurs extra overhead in every
17776 floating-point comparison, therefore the default is set to
17777 @option{-ffinite-math-only}.
17779 @item -minline-ic_invalidate
17780 @opindex minline-ic_invalidate
17781 Inline code to invalidate instruction cache entries after setting up
17782 nested function trampolines.
17783 This option has no effect if -musermode is in effect and the selected
17784 code generation option (e.g. -m4) does not allow the use of the icbi
17786 If the selected code generation option does not allow the use of the icbi
17787 instruction, and -musermode is not in effect, the inlined code will
17788 manipulate the instruction cache address array directly with an associative
17789 write. This not only requires privileged mode, but it will also
17790 fail if the cache line had been mapped via the TLB and has become unmapped.
17794 Dump instruction size and location in the assembly code.
17797 @opindex mpadstruct
17798 This option is deprecated. It pads structures to multiple of 4 bytes,
17799 which is incompatible with the SH ABI@.
17801 @item -msoft-atomic
17802 @opindex msoft-atomic
17803 Generate software atomic sequences for the atomic operations.
17804 This is the default when the target is @code{sh-*-linux*}.
17808 Optimize for space instead of speed. Implied by @option{-Os}.
17811 @opindex mprefergot
17812 When generating position-independent code, emit function calls using
17813 the Global Offset Table instead of the Procedure Linkage Table.
17817 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17818 if the inlined code would not work in user mode.
17819 This is the default when the target is @code{sh-*-linux*}.
17821 @item -multcost=@var{number}
17822 @opindex multcost=@var{number}
17823 Set the cost to assume for a multiply insn.
17825 @item -mdiv=@var{strategy}
17826 @opindex mdiv=@var{strategy}
17827 Set the division strategy to use for SHmedia code. @var{strategy} must be
17828 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17829 inv:call2, inv:fp .
17830 "fp" performs the operation in floating point. This has a very high latency,
17831 but needs only a few instructions, so it might be a good choice if
17832 your code has enough easily-exploitable ILP to allow the compiler to
17833 schedule the floating-point instructions together with other instructions.
17834 Division by zero causes a floating-point exception.
17835 "inv" uses integer operations to calculate the inverse of the divisor,
17836 and then multiplies the dividend with the inverse. This strategy allows
17837 cse and hoisting of the inverse calculation. Division by zero calculates
17838 an unspecified result, but does not trap.
17839 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17840 have been found, or if the entire operation has been hoisted to the same
17841 place, the last stages of the inverse calculation are intertwined with the
17842 final multiply to reduce the overall latency, at the expense of using a few
17843 more instructions, and thus offering fewer scheduling opportunities with
17845 "call" calls a library function that usually implements the inv:minlat
17847 This gives high code density for m5-*media-nofpu compilations.
17848 "call2" uses a different entry point of the same library function, where it
17849 assumes that a pointer to a lookup table has already been set up, which
17850 exposes the pointer load to cse / code hoisting optimizations.
17851 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17852 code generation, but if the code stays unoptimized, revert to the "call",
17853 "call2", or "fp" strategies, respectively. Note that the
17854 potentially-trapping side effect of division by zero is carried by a
17855 separate instruction, so it is possible that all the integer instructions
17856 are hoisted out, but the marker for the side effect stays where it is.
17857 A recombination to fp operations or a call is not possible in that case.
17858 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17859 that the inverse calculation was nor separated from the multiply, they speed
17860 up division where the dividend fits into 20 bits (plus sign where applicable),
17861 by inserting a test to skip a number of operations in this case; this test
17862 slows down the case of larger dividends. inv20u assumes the case of a such
17863 a small dividend to be unlikely, and inv20l assumes it to be likely.
17865 @item -maccumulate-outgoing-args
17866 @opindex maccumulate-outgoing-args
17867 Reserve space once for outgoing arguments in the function prologue rather
17868 than around each call. Generally beneficial for performance and size. Also
17869 needed for unwinding to avoid changing the stack frame around conditional code.
17871 @item -mdivsi3_libfunc=@var{name}
17872 @opindex mdivsi3_libfunc=@var{name}
17873 Set the name of the library function used for 32 bit signed division to
17874 @var{name}. This only affect the name used in the call and inv:call
17875 division strategies, and the compiler will still expect the same
17876 sets of input/output/clobbered registers as if this option was not present.
17878 @item -mfixed-range=@var{register-range}
17879 @opindex mfixed-range
17880 Generate code treating the given register range as fixed registers.
17881 A fixed register is one that the register allocator can not use. This is
17882 useful when compiling kernel code. A register range is specified as
17883 two registers separated by a dash. Multiple register ranges can be
17884 specified separated by a comma.
17886 @item -madjust-unroll
17887 @opindex madjust-unroll
17888 Throttle unrolling to avoid thrashing target registers.
17889 This option only has an effect if the gcc code base supports the
17890 TARGET_ADJUST_UNROLL_MAX target hook.
17892 @item -mindexed-addressing
17893 @opindex mindexed-addressing
17894 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17895 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17896 semantics for the indexed addressing mode. The architecture allows the
17897 implementation of processors with 64 bit MMU, which the OS could use to
17898 get 32 bit addressing, but since no current hardware implementation supports
17899 this or any other way to make the indexed addressing mode safe to use in
17900 the 32 bit ABI, the default is -mno-indexed-addressing.
17902 @item -mgettrcost=@var{number}
17903 @opindex mgettrcost=@var{number}
17904 Set the cost assumed for the gettr instruction to @var{number}.
17905 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17909 Assume pt* instructions won't trap. This will generally generate better
17910 scheduled code, but is unsafe on current hardware. The current architecture
17911 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17912 This has the unintentional effect of making it unsafe to schedule ptabs /
17913 ptrel before a branch, or hoist it out of a loop. For example,
17914 __do_global_ctors, a part of libgcc that runs constructors at program
17915 startup, calls functions in a list which is delimited by @minus{}1. With the
17916 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17917 That means that all the constructors will be run a bit quicker, but when
17918 the loop comes to the end of the list, the program crashes because ptabs
17919 loads @minus{}1 into a target register. Since this option is unsafe for any
17920 hardware implementing the current architecture specification, the default
17921 is -mno-pt-fixed. Unless the user specifies a specific cost with
17922 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17923 this deters register allocation using target registers for storing
17926 @item -minvalid-symbols
17927 @opindex minvalid-symbols
17928 Assume symbols might be invalid. Ordinary function symbols generated by
17929 the compiler will always be valid to load with movi/shori/ptabs or
17930 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17931 to generate symbols that will cause ptabs / ptrel to trap.
17932 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17933 It will then prevent cross-basic-block cse, hoisting and most scheduling
17934 of symbol loads. The default is @option{-mno-invalid-symbols}.
17937 @node Solaris 2 Options
17938 @subsection Solaris 2 Options
17939 @cindex Solaris 2 options
17941 These @samp{-m} options are supported on Solaris 2:
17944 @item -mimpure-text
17945 @opindex mimpure-text
17946 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17947 the compiler to not pass @option{-z text} to the linker when linking a
17948 shared object. Using this option, you can link position-dependent
17949 code into a shared object.
17951 @option{-mimpure-text} suppresses the ``relocations remain against
17952 allocatable but non-writable sections'' linker error message.
17953 However, the necessary relocations will trigger copy-on-write, and the
17954 shared object is not actually shared across processes. Instead of
17955 using @option{-mimpure-text}, you should compile all source code with
17956 @option{-fpic} or @option{-fPIC}.
17960 These switches are supported in addition to the above on Solaris 2:
17965 Add support for multithreading using the POSIX threads library. This
17966 option sets flags for both the preprocessor and linker. This option does
17967 not affect the thread safety of object code produced by the compiler or
17968 that of libraries supplied with it.
17972 This is a synonym for @option{-pthreads}.
17975 @node SPARC Options
17976 @subsection SPARC Options
17977 @cindex SPARC options
17979 These @samp{-m} options are supported on the SPARC:
17982 @item -mno-app-regs
17984 @opindex mno-app-regs
17986 Specify @option{-mapp-regs} to generate output using the global registers
17987 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17990 To be fully SVR4 ABI compliant at the cost of some performance loss,
17991 specify @option{-mno-app-regs}. You should compile libraries and system
17992 software with this option.
17998 With @option{-mflat}, the compiler does not generate save/restore instructions
17999 and uses a ``flat'' or single register window model. This model is compatible
18000 with the regular register window model. The local registers and the input
18001 registers (0--5) are still treated as ``call-saved'' registers and will be
18002 saved on the stack as needed.
18004 With @option{-mno-flat} (the default), the compiler generates save/restore
18005 instructions (except for leaf functions). This is the normal operating mode.
18008 @itemx -mhard-float
18010 @opindex mhard-float
18011 Generate output containing floating-point instructions. This is the
18015 @itemx -msoft-float
18017 @opindex msoft-float
18018 Generate output containing library calls for floating point.
18019 @strong{Warning:} the requisite libraries are not available for all SPARC
18020 targets. Normally the facilities of the machine's usual C compiler are
18021 used, but this cannot be done directly in cross-compilation. You must make
18022 your own arrangements to provide suitable library functions for
18023 cross-compilation. The embedded targets @samp{sparc-*-aout} and
18024 @samp{sparclite-*-*} do provide software floating-point support.
18026 @option{-msoft-float} changes the calling convention in the output file;
18027 therefore, it is only useful if you compile @emph{all} of a program with
18028 this option. In particular, you need to compile @file{libgcc.a}, the
18029 library that comes with GCC, with @option{-msoft-float} in order for
18032 @item -mhard-quad-float
18033 @opindex mhard-quad-float
18034 Generate output containing quad-word (long double) floating-point
18037 @item -msoft-quad-float
18038 @opindex msoft-quad-float
18039 Generate output containing library calls for quad-word (long double)
18040 floating-point instructions. The functions called are those specified
18041 in the SPARC ABI@. This is the default.
18043 As of this writing, there are no SPARC implementations that have hardware
18044 support for the quad-word floating-point instructions. They all invoke
18045 a trap handler for one of these instructions, and then the trap handler
18046 emulates the effect of the instruction. Because of the trap handler overhead,
18047 this is much slower than calling the ABI library routines. Thus the
18048 @option{-msoft-quad-float} option is the default.
18050 @item -mno-unaligned-doubles
18051 @itemx -munaligned-doubles
18052 @opindex mno-unaligned-doubles
18053 @opindex munaligned-doubles
18054 Assume that doubles have 8 byte alignment. This is the default.
18056 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
18057 alignment only if they are contained in another type, or if they have an
18058 absolute address. Otherwise, it assumes they have 4 byte alignment.
18059 Specifying this option avoids some rare compatibility problems with code
18060 generated by other compilers. It is not the default because it results
18061 in a performance loss, especially for floating-point code.
18063 @item -mno-faster-structs
18064 @itemx -mfaster-structs
18065 @opindex mno-faster-structs
18066 @opindex mfaster-structs
18067 With @option{-mfaster-structs}, the compiler assumes that structures
18068 should have 8 byte alignment. This enables the use of pairs of
18069 @code{ldd} and @code{std} instructions for copies in structure
18070 assignment, in place of twice as many @code{ld} and @code{st} pairs.
18071 However, the use of this changed alignment directly violates the SPARC
18072 ABI@. Thus, it's intended only for use on targets where the developer
18073 acknowledges that their resulting code will not be directly in line with
18074 the rules of the ABI@.
18076 @item -mcpu=@var{cpu_type}
18078 Set the instruction set, register set, and instruction scheduling parameters
18079 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
18080 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
18081 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
18082 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
18083 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
18084 and @samp{niagara4}.
18086 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
18087 which selects the best architecture option for the host processor.
18088 @option{-mcpu=native} has no effect if GCC does not recognize
18091 Default instruction scheduling parameters are used for values that select
18092 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
18093 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
18095 Here is a list of each supported architecture and their supported
18103 supersparc, hypersparc, leon
18106 f930, f934, sparclite86x
18112 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
18115 By default (unless configured otherwise), GCC generates code for the V7
18116 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
18117 additionally optimizes it for the Cypress CY7C602 chip, as used in the
18118 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
18119 SPARCStation 1, 2, IPX etc.
18121 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
18122 architecture. The only difference from V7 code is that the compiler emits
18123 the integer multiply and integer divide instructions which exist in SPARC-V8
18124 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
18125 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
18128 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
18129 the SPARC architecture. This adds the integer multiply, integer divide step
18130 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
18131 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
18132 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
18133 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
18134 MB86934 chip, which is the more recent SPARClite with FPU@.
18136 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
18137 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
18138 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
18139 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
18140 optimizes it for the TEMIC SPARClet chip.
18142 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
18143 architecture. This adds 64-bit integer and floating-point move instructions,
18144 3 additional floating-point condition code registers and conditional move
18145 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
18146 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
18147 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
18148 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
18149 @option{-mcpu=niagara}, the compiler additionally optimizes it for
18150 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
18151 additionally optimizes it for Sun UltraSPARC T2 chips. With
18152 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
18153 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
18154 additionally optimizes it for Sun UltraSPARC T4 chips.
18156 @item -mtune=@var{cpu_type}
18158 Set the instruction scheduling parameters for machine type
18159 @var{cpu_type}, but do not set the instruction set or register set that the
18160 option @option{-mcpu=@var{cpu_type}} would.
18162 The same values for @option{-mcpu=@var{cpu_type}} can be used for
18163 @option{-mtune=@var{cpu_type}}, but the only useful values are those
18164 that select a particular CPU implementation. Those are @samp{cypress},
18165 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
18166 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
18167 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
18168 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
18173 @opindex mno-v8plus
18174 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
18175 difference from the V8 ABI is that the global and out registers are
18176 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
18177 mode for all SPARC-V9 processors.
18183 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
18184 Visual Instruction Set extensions. The default is @option{-mno-vis}.
18190 With @option{-mvis2}, GCC generates code that takes advantage of
18191 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
18192 default is @option{-mvis2} when targetting a cpu that supports such
18193 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
18194 also sets @option{-mvis}.
18200 With @option{-mvis3}, GCC generates code that takes advantage of
18201 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
18202 default is @option{-mvis3} when targetting a cpu that supports such
18203 instructions, such as niagara-3 and later. Setting @option{-mvis3}
18204 also sets @option{-mvis2} and @option{-mvis}.
18210 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
18211 population count instruction. The default is @option{-mpopc}
18212 when targetting a cpu that supports such instructions, such as Niagara-2 and
18219 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
18220 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
18221 when targetting a cpu that supports such instructions, such as Niagara-3 and
18225 @opindex mfix-at697f
18226 Enable the documented workaround for the single erratum of the Atmel AT697F
18227 processor (which corresponds to erratum #13 of the AT697E processor).
18230 These @samp{-m} options are supported in addition to the above
18231 on SPARC-V9 processors in 64-bit environments:
18234 @item -mlittle-endian
18235 @opindex mlittle-endian
18236 Generate code for a processor running in little-endian mode. It is only
18237 available for a few configurations and most notably not on Solaris and Linux.
18243 Generate code for a 32-bit or 64-bit environment.
18244 The 32-bit environment sets int, long and pointer to 32 bits.
18245 The 64-bit environment sets int to 32 bits and long and pointer
18248 @item -mcmodel=@var{which}
18250 Set the code model to one of
18254 The Medium/Low code model: 64-bit addresses, programs
18255 must be linked in the low 32 bits of memory. Programs can be statically
18256 or dynamically linked.
18259 The Medium/Middle code model: 64-bit addresses, programs
18260 must be linked in the low 44 bits of memory, the text and data segments must
18261 be less than 2GB in size and the data segment must be located within 2GB of
18265 The Medium/Anywhere code model: 64-bit addresses, programs
18266 may be linked anywhere in memory, the text and data segments must be less
18267 than 2GB in size and the data segment must be located within 2GB of the
18271 The Medium/Anywhere code model for embedded systems:
18272 64-bit addresses, the text and data segments must be less than 2GB in
18273 size, both starting anywhere in memory (determined at link time). The
18274 global register %g4 points to the base of the data segment. Programs
18275 are statically linked and PIC is not supported.
18278 @item -mmemory-model=@var{mem-model}
18279 @opindex mmemory-model
18280 Set the memory model in force on the processor to one of
18284 The default memory model for the processor and operating system.
18287 Relaxed Memory Order
18290 Partial Store Order
18296 Sequential Consistency
18299 These memory models are formally defined in Appendix D of the Sparc V9
18300 architecture manual, as set in the processor's @code{PSTATE.MM} field.
18303 @itemx -mno-stack-bias
18304 @opindex mstack-bias
18305 @opindex mno-stack-bias
18306 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
18307 frame pointer if present, are offset by @minus{}2047 which must be added back
18308 when making stack frame references. This is the default in 64-bit mode.
18309 Otherwise, assume no such offset is present.
18313 @subsection SPU Options
18314 @cindex SPU options
18316 These @samp{-m} options are supported on the SPU:
18320 @itemx -merror-reloc
18321 @opindex mwarn-reloc
18322 @opindex merror-reloc
18324 The loader for SPU does not handle dynamic relocations. By default, GCC
18325 will give an error when it generates code that requires a dynamic
18326 relocation. @option{-mno-error-reloc} disables the error,
18327 @option{-mwarn-reloc} will generate a warning instead.
18330 @itemx -munsafe-dma
18332 @opindex munsafe-dma
18334 Instructions which initiate or test completion of DMA must not be
18335 reordered with respect to loads and stores of the memory which is being
18336 accessed. Users typically address this problem using the volatile
18337 keyword, but that can lead to inefficient code in places where the
18338 memory is known to not change. Rather than mark the memory as volatile
18339 we treat the DMA instructions as potentially effecting all memory. With
18340 @option{-munsafe-dma} users must use the volatile keyword to protect
18343 @item -mbranch-hints
18344 @opindex mbranch-hints
18346 By default, GCC will generate a branch hint instruction to avoid
18347 pipeline stalls for always taken or probably taken branches. A hint
18348 will not be generated closer than 8 instructions away from its branch.
18349 There is little reason to disable them, except for debugging purposes,
18350 or to make an object a little bit smaller.
18354 @opindex msmall-mem
18355 @opindex mlarge-mem
18357 By default, GCC generates code assuming that addresses are never larger
18358 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
18359 a full 32 bit address.
18364 By default, GCC links against startup code that assumes the SPU-style
18365 main function interface (which has an unconventional parameter list).
18366 With @option{-mstdmain}, GCC will link your program against startup
18367 code that assumes a C99-style interface to @code{main}, including a
18368 local copy of @code{argv} strings.
18370 @item -mfixed-range=@var{register-range}
18371 @opindex mfixed-range
18372 Generate code treating the given register range as fixed registers.
18373 A fixed register is one that the register allocator can not use. This is
18374 useful when compiling kernel code. A register range is specified as
18375 two registers separated by a dash. Multiple register ranges can be
18376 specified separated by a comma.
18382 Compile code assuming that pointers to the PPU address space accessed
18383 via the @code{__ea} named address space qualifier are either 32 or 64
18384 bits wide. The default is 32 bits. As this is an ABI changing option,
18385 all object code in an executable must be compiled with the same setting.
18387 @item -maddress-space-conversion
18388 @itemx -mno-address-space-conversion
18389 @opindex maddress-space-conversion
18390 @opindex mno-address-space-conversion
18391 Allow/disallow treating the @code{__ea} address space as superset
18392 of the generic address space. This enables explicit type casts
18393 between @code{__ea} and generic pointer as well as implicit
18394 conversions of generic pointers to @code{__ea} pointers. The
18395 default is to allow address space pointer conversions.
18397 @item -mcache-size=@var{cache-size}
18398 @opindex mcache-size
18399 This option controls the version of libgcc that the compiler links to an
18400 executable and selects a software-managed cache for accessing variables
18401 in the @code{__ea} address space with a particular cache size. Possible
18402 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
18403 and @samp{128}. The default cache size is 64KB.
18405 @item -matomic-updates
18406 @itemx -mno-atomic-updates
18407 @opindex matomic-updates
18408 @opindex mno-atomic-updates
18409 This option controls the version of libgcc that the compiler links to an
18410 executable and selects whether atomic updates to the software-managed
18411 cache of PPU-side variables are used. If you use atomic updates, changes
18412 to a PPU variable from SPU code using the @code{__ea} named address space
18413 qualifier will not interfere with changes to other PPU variables residing
18414 in the same cache line from PPU code. If you do not use atomic updates,
18415 such interference may occur; however, writing back cache lines will be
18416 more efficient. The default behavior is to use atomic updates.
18419 @itemx -mdual-nops=@var{n}
18420 @opindex mdual-nops
18421 By default, GCC will insert nops to increase dual issue when it expects
18422 it to increase performance. @var{n} can be a value from 0 to 10. A
18423 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
18424 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
18426 @item -mhint-max-nops=@var{n}
18427 @opindex mhint-max-nops
18428 Maximum number of nops to insert for a branch hint. A branch hint must
18429 be at least 8 instructions away from the branch it is effecting. GCC
18430 will insert up to @var{n} nops to enforce this, otherwise it will not
18431 generate the branch hint.
18433 @item -mhint-max-distance=@var{n}
18434 @opindex mhint-max-distance
18435 The encoding of the branch hint instruction limits the hint to be within
18436 256 instructions of the branch it is effecting. By default, GCC makes
18437 sure it is within 125.
18440 @opindex msafe-hints
18441 Work around a hardware bug which causes the SPU to stall indefinitely.
18442 By default, GCC will insert the @code{hbrp} instruction to make sure
18443 this stall won't happen.
18447 @node System V Options
18448 @subsection Options for System V
18450 These additional options are available on System V Release 4 for
18451 compatibility with other compilers on those systems:
18456 Create a shared object.
18457 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
18461 Identify the versions of each tool used by the compiler, in a
18462 @code{.ident} assembler directive in the output.
18466 Refrain from adding @code{.ident} directives to the output file (this is
18469 @item -YP,@var{dirs}
18471 Search the directories @var{dirs}, and no others, for libraries
18472 specified with @option{-l}.
18474 @item -Ym,@var{dir}
18476 Look in the directory @var{dir} to find the M4 preprocessor.
18477 The assembler uses this option.
18478 @c This is supposed to go with a -Yd for predefined M4 macro files, but
18479 @c the generic assembler that comes with Solaris takes just -Ym.
18483 @subsection V850 Options
18484 @cindex V850 Options
18486 These @samp{-m} options are defined for V850 implementations:
18490 @itemx -mno-long-calls
18491 @opindex mlong-calls
18492 @opindex mno-long-calls
18493 Treat all calls as being far away (near). If calls are assumed to be
18494 far away, the compiler will always load the functions address up into a
18495 register, and call indirect through the pointer.
18501 Do not optimize (do optimize) basic blocks that use the same index
18502 pointer 4 or more times to copy pointer into the @code{ep} register, and
18503 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
18504 option is on by default if you optimize.
18506 @item -mno-prolog-function
18507 @itemx -mprolog-function
18508 @opindex mno-prolog-function
18509 @opindex mprolog-function
18510 Do not use (do use) external functions to save and restore registers
18511 at the prologue and epilogue of a function. The external functions
18512 are slower, but use less code space if more than one function saves
18513 the same number of registers. The @option{-mprolog-function} option
18514 is on by default if you optimize.
18518 Try to make the code as small as possible. At present, this just turns
18519 on the @option{-mep} and @option{-mprolog-function} options.
18521 @item -mtda=@var{n}
18523 Put static or global variables whose size is @var{n} bytes or less into
18524 the tiny data area that register @code{ep} points to. The tiny data
18525 area can hold up to 256 bytes in total (128 bytes for byte references).
18527 @item -msda=@var{n}
18529 Put static or global variables whose size is @var{n} bytes or less into
18530 the small data area that register @code{gp} points to. The small data
18531 area can hold up to 64 kilobytes.
18533 @item -mzda=@var{n}
18535 Put static or global variables whose size is @var{n} bytes or less into
18536 the first 32 kilobytes of memory.
18540 Specify that the target processor is the V850.
18543 @opindex mbig-switch
18544 Generate code suitable for big switch tables. Use this option only if
18545 the assembler/linker complain about out of range branches within a switch
18550 This option will cause r2 and r5 to be used in the code generated by
18551 the compiler. This setting is the default.
18553 @item -mno-app-regs
18554 @opindex mno-app-regs
18555 This option will cause r2 and r5 to be treated as fixed registers.
18559 Specify that the target processor is the V850E2V3. The preprocessor
18560 constants @samp{__v850e2v3__} will be defined if
18561 this option is used.
18565 Specify that the target processor is the V850E2. The preprocessor
18566 constants @samp{__v850e2__} will be defined if this option is used.
18570 Specify that the target processor is the V850E1. The preprocessor
18571 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
18572 this option is used.
18576 Specify that the target processor is the V850ES. This is an alias for
18577 the @option{-mv850e1} option.
18581 Specify that the target processor is the V850E@. The preprocessor
18582 constant @samp{__v850e__} will be defined if this option is used.
18584 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
18585 nor @option{-mv850e2} nor @option{-mv850e2v3}
18586 are defined then a default target processor will be chosen and the
18587 relevant @samp{__v850*__} preprocessor constant will be defined.
18589 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
18590 defined, regardless of which processor variant is the target.
18592 @item -mdisable-callt
18593 @opindex mdisable-callt
18594 This option will suppress generation of the CALLT instruction for the
18595 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
18596 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
18601 @subsection VAX Options
18602 @cindex VAX options
18604 These @samp{-m} options are defined for the VAX:
18609 Do not output certain jump instructions (@code{aobleq} and so on)
18610 that the Unix assembler for the VAX cannot handle across long
18615 Do output those jump instructions, on the assumption that you
18616 will assemble with the GNU assembler.
18620 Output code for G-format floating-point numbers instead of D-format.
18623 @node VxWorks Options
18624 @subsection VxWorks Options
18625 @cindex VxWorks Options
18627 The options in this section are defined for all VxWorks targets.
18628 Options specific to the target hardware are listed with the other
18629 options for that target.
18634 GCC can generate code for both VxWorks kernels and real time processes
18635 (RTPs). This option switches from the former to the latter. It also
18636 defines the preprocessor macro @code{__RTP__}.
18639 @opindex non-static
18640 Link an RTP executable against shared libraries rather than static
18641 libraries. The options @option{-static} and @option{-shared} can
18642 also be used for RTPs (@pxref{Link Options}); @option{-static}
18649 These options are passed down to the linker. They are defined for
18650 compatibility with Diab.
18653 @opindex Xbind-lazy
18654 Enable lazy binding of function calls. This option is equivalent to
18655 @option{-Wl,-z,now} and is defined for compatibility with Diab.
18659 Disable lazy binding of function calls. This option is the default and
18660 is defined for compatibility with Diab.
18663 @node x86-64 Options
18664 @subsection x86-64 Options
18665 @cindex x86-64 options
18667 These are listed under @xref{i386 and x86-64 Options}.
18669 @node Xstormy16 Options
18670 @subsection Xstormy16 Options
18671 @cindex Xstormy16 Options
18673 These options are defined for Xstormy16:
18678 Choose startup files and linker script suitable for the simulator.
18681 @node Xtensa Options
18682 @subsection Xtensa Options
18683 @cindex Xtensa Options
18685 These options are supported for Xtensa targets:
18689 @itemx -mno-const16
18691 @opindex mno-const16
18692 Enable or disable use of @code{CONST16} instructions for loading
18693 constant values. The @code{CONST16} instruction is currently not a
18694 standard option from Tensilica. When enabled, @code{CONST16}
18695 instructions are always used in place of the standard @code{L32R}
18696 instructions. The use of @code{CONST16} is enabled by default only if
18697 the @code{L32R} instruction is not available.
18700 @itemx -mno-fused-madd
18701 @opindex mfused-madd
18702 @opindex mno-fused-madd
18703 Enable or disable use of fused multiply/add and multiply/subtract
18704 instructions in the floating-point option. This has no effect if the
18705 floating-point option is not also enabled. Disabling fused multiply/add
18706 and multiply/subtract instructions forces the compiler to use separate
18707 instructions for the multiply and add/subtract operations. This may be
18708 desirable in some cases where strict IEEE 754-compliant results are
18709 required: the fused multiply add/subtract instructions do not round the
18710 intermediate result, thereby producing results with @emph{more} bits of
18711 precision than specified by the IEEE standard. Disabling fused multiply
18712 add/subtract instructions also ensures that the program output is not
18713 sensitive to the compiler's ability to combine multiply and add/subtract
18716 @item -mserialize-volatile
18717 @itemx -mno-serialize-volatile
18718 @opindex mserialize-volatile
18719 @opindex mno-serialize-volatile
18720 When this option is enabled, GCC inserts @code{MEMW} instructions before
18721 @code{volatile} memory references to guarantee sequential consistency.
18722 The default is @option{-mserialize-volatile}. Use
18723 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
18725 @item -mforce-no-pic
18726 @opindex mforce-no-pic
18727 For targets, like GNU/Linux, where all user-mode Xtensa code must be
18728 position-independent code (PIC), this option disables PIC for compiling
18731 @item -mtext-section-literals
18732 @itemx -mno-text-section-literals
18733 @opindex mtext-section-literals
18734 @opindex mno-text-section-literals
18735 Control the treatment of literal pools. The default is
18736 @option{-mno-text-section-literals}, which places literals in a separate
18737 section in the output file. This allows the literal pool to be placed
18738 in a data RAM/ROM, and it also allows the linker to combine literal
18739 pools from separate object files to remove redundant literals and
18740 improve code size. With @option{-mtext-section-literals}, the literals
18741 are interspersed in the text section in order to keep them as close as
18742 possible to their references. This may be necessary for large assembly
18745 @item -mtarget-align
18746 @itemx -mno-target-align
18747 @opindex mtarget-align
18748 @opindex mno-target-align
18749 When this option is enabled, GCC instructs the assembler to
18750 automatically align instructions to reduce branch penalties at the
18751 expense of some code density. The assembler attempts to widen density
18752 instructions to align branch targets and the instructions following call
18753 instructions. If there are not enough preceding safe density
18754 instructions to align a target, no widening will be performed. The
18755 default is @option{-mtarget-align}. These options do not affect the
18756 treatment of auto-aligned instructions like @code{LOOP}, which the
18757 assembler will always align, either by widening density instructions or
18758 by inserting no-op instructions.
18761 @itemx -mno-longcalls
18762 @opindex mlongcalls
18763 @opindex mno-longcalls
18764 When this option is enabled, GCC instructs the assembler to translate
18765 direct calls to indirect calls unless it can determine that the target
18766 of a direct call is in the range allowed by the call instruction. This
18767 translation typically occurs for calls to functions in other source
18768 files. Specifically, the assembler translates a direct @code{CALL}
18769 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
18770 The default is @option{-mno-longcalls}. This option should be used in
18771 programs where the call target can potentially be out of range. This
18772 option is implemented in the assembler, not the compiler, so the
18773 assembly code generated by GCC will still show direct call
18774 instructions---look at the disassembled object code to see the actual
18775 instructions. Note that the assembler will use an indirect call for
18776 every cross-file call, not just those that really will be out of range.
18779 @node zSeries Options
18780 @subsection zSeries Options
18781 @cindex zSeries options
18783 These are listed under @xref{S/390 and zSeries Options}.
18785 @node Code Gen Options
18786 @section Options for Code Generation Conventions
18787 @cindex code generation conventions
18788 @cindex options, code generation
18789 @cindex run-time options
18791 These machine-independent options control the interface conventions
18792 used in code generation.
18794 Most of them have both positive and negative forms; the negative form
18795 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
18796 one of the forms is listed---the one which is not the default. You
18797 can figure out the other form by either removing @samp{no-} or adding
18801 @item -fbounds-check
18802 @opindex fbounds-check
18803 For front-ends that support it, generate additional code to check that
18804 indices used to access arrays are within the declared range. This is
18805 currently only supported by the Java and Fortran front-ends, where
18806 this option defaults to true and false respectively.
18810 This option generates traps for signed overflow on addition, subtraction,
18811 multiplication operations.
18815 This option instructs the compiler to assume that signed arithmetic
18816 overflow of addition, subtraction and multiplication wraps around
18817 using twos-complement representation. This flag enables some optimizations
18818 and disables others. This option is enabled by default for the Java
18819 front-end, as required by the Java language specification.
18822 @opindex fexceptions
18823 Enable exception handling. Generates extra code needed to propagate
18824 exceptions. For some targets, this implies GCC will generate frame
18825 unwind information for all functions, which can produce significant data
18826 size overhead, although it does not affect execution. If you do not
18827 specify this option, GCC will enable it by default for languages like
18828 C++ which normally require exception handling, and disable it for
18829 languages like C that do not normally require it. However, you may need
18830 to enable this option when compiling C code that needs to interoperate
18831 properly with exception handlers written in C++. You may also wish to
18832 disable this option if you are compiling older C++ programs that don't
18833 use exception handling.
18835 @item -fnon-call-exceptions
18836 @opindex fnon-call-exceptions
18837 Generate code that allows trapping instructions to throw exceptions.
18838 Note that this requires platform-specific runtime support that does
18839 not exist everywhere. Moreover, it only allows @emph{trapping}
18840 instructions to throw exceptions, i.e.@: memory references or floating-point
18841 instructions. It does not allow exceptions to be thrown from
18842 arbitrary signal handlers such as @code{SIGALRM}.
18844 @item -funwind-tables
18845 @opindex funwind-tables
18846 Similar to @option{-fexceptions}, except that it will just generate any needed
18847 static data, but will not affect the generated code in any other way.
18848 You will normally not enable this option; instead, a language processor
18849 that needs this handling would enable it on your behalf.
18851 @item -fasynchronous-unwind-tables
18852 @opindex fasynchronous-unwind-tables
18853 Generate unwind table in dwarf2 format, if supported by target machine. The
18854 table is exact at each instruction boundary, so it can be used for stack
18855 unwinding from asynchronous events (such as debugger or garbage collector).
18857 @item -fpcc-struct-return
18858 @opindex fpcc-struct-return
18859 Return ``short'' @code{struct} and @code{union} values in memory like
18860 longer ones, rather than in registers. This convention is less
18861 efficient, but it has the advantage of allowing intercallability between
18862 GCC-compiled files and files compiled with other compilers, particularly
18863 the Portable C Compiler (pcc).
18865 The precise convention for returning structures in memory depends
18866 on the target configuration macros.
18868 Short structures and unions are those whose size and alignment match
18869 that of some integer type.
18871 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
18872 switch is not binary compatible with code compiled with the
18873 @option{-freg-struct-return} switch.
18874 Use it to conform to a non-default application binary interface.
18876 @item -freg-struct-return
18877 @opindex freg-struct-return
18878 Return @code{struct} and @code{union} values in registers when possible.
18879 This is more efficient for small structures than
18880 @option{-fpcc-struct-return}.
18882 If you specify neither @option{-fpcc-struct-return} nor
18883 @option{-freg-struct-return}, GCC defaults to whichever convention is
18884 standard for the target. If there is no standard convention, GCC
18885 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18886 the principal compiler. In those cases, we can choose the standard, and
18887 we chose the more efficient register return alternative.
18889 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18890 switch is not binary compatible with code compiled with the
18891 @option{-fpcc-struct-return} switch.
18892 Use it to conform to a non-default application binary interface.
18894 @item -fshort-enums
18895 @opindex fshort-enums
18896 Allocate to an @code{enum} type only as many bytes as it needs for the
18897 declared range of possible values. Specifically, the @code{enum} type
18898 will be equivalent to the smallest integer type which has enough room.
18900 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18901 code that is not binary compatible with code generated without that switch.
18902 Use it to conform to a non-default application binary interface.
18904 @item -fshort-double
18905 @opindex fshort-double
18906 Use the same size for @code{double} as for @code{float}.
18908 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18909 code that is not binary compatible with code generated without that switch.
18910 Use it to conform to a non-default application binary interface.
18912 @item -fshort-wchar
18913 @opindex fshort-wchar
18914 Override the underlying type for @samp{wchar_t} to be @samp{short
18915 unsigned int} instead of the default for the target. This option is
18916 useful for building programs to run under WINE@.
18918 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18919 code that is not binary compatible with code generated without that switch.
18920 Use it to conform to a non-default application binary interface.
18923 @opindex fno-common
18924 In C code, controls the placement of uninitialized global variables.
18925 Unix C compilers have traditionally permitted multiple definitions of
18926 such variables in different compilation units by placing the variables
18928 This is the behavior specified by @option{-fcommon}, and is the default
18929 for GCC on most targets.
18930 On the other hand, this behavior is not required by ISO C, and on some
18931 targets may carry a speed or code size penalty on variable references.
18932 The @option{-fno-common} option specifies that the compiler should place
18933 uninitialized global variables in the data section of the object file,
18934 rather than generating them as common blocks.
18935 This has the effect that if the same variable is declared
18936 (without @code{extern}) in two different compilations,
18937 you will get a multiple-definition error when you link them.
18938 In this case, you must compile with @option{-fcommon} instead.
18939 Compiling with @option{-fno-common} is useful on targets for which
18940 it provides better performance, or if you wish to verify that the
18941 program will work on other systems which always treat uninitialized
18942 variable declarations this way.
18946 Ignore the @samp{#ident} directive.
18948 @item -finhibit-size-directive
18949 @opindex finhibit-size-directive
18950 Don't output a @code{.size} assembler directive, or anything else that
18951 would cause trouble if the function is split in the middle, and the
18952 two halves are placed at locations far apart in memory. This option is
18953 used when compiling @file{crtstuff.c}; you should not need to use it
18956 @item -fverbose-asm
18957 @opindex fverbose-asm
18958 Put extra commentary information in the generated assembly code to
18959 make it more readable. This option is generally only of use to those
18960 who actually need to read the generated assembly code (perhaps while
18961 debugging the compiler itself).
18963 @option{-fno-verbose-asm}, the default, causes the
18964 extra information to be omitted and is useful when comparing two assembler
18967 @item -frecord-gcc-switches
18968 @opindex frecord-gcc-switches
18969 This switch causes the command line that was used to invoke the
18970 compiler to be recorded into the object file that is being created.
18971 This switch is only implemented on some targets and the exact format
18972 of the recording is target and binary file format dependent, but it
18973 usually takes the form of a section containing ASCII text. This
18974 switch is related to the @option{-fverbose-asm} switch, but that
18975 switch only records information in the assembler output file as
18976 comments, so it never reaches the object file.
18977 See also @option{-grecord-gcc-switches} for another
18978 way of storing compiler options into the object file.
18982 @cindex global offset table
18984 Generate position-independent code (PIC) suitable for use in a shared
18985 library, if supported for the target machine. Such code accesses all
18986 constant addresses through a global offset table (GOT)@. The dynamic
18987 loader resolves the GOT entries when the program starts (the dynamic
18988 loader is not part of GCC; it is part of the operating system). If
18989 the GOT size for the linked executable exceeds a machine-specific
18990 maximum size, you get an error message from the linker indicating that
18991 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18992 instead. (These maximums are 8k on the SPARC and 32k
18993 on the m68k and RS/6000. The 386 has no such limit.)
18995 Position-independent code requires special support, and therefore works
18996 only on certain machines. For the 386, GCC supports PIC for System V
18997 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18998 position-independent.
19000 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19005 If supported for the target machine, emit position-independent code,
19006 suitable for dynamic linking and avoiding any limit on the size of the
19007 global offset table. This option makes a difference on the m68k,
19008 PowerPC and SPARC@.
19010 Position-independent code requires special support, and therefore works
19011 only on certain machines.
19013 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19020 These options are similar to @option{-fpic} and @option{-fPIC}, but
19021 generated position independent code can be only linked into executables.
19022 Usually these options are used when @option{-pie} GCC option will be
19023 used during linking.
19025 @option{-fpie} and @option{-fPIE} both define the macros
19026 @code{__pie__} and @code{__PIE__}. The macros have the value 1
19027 for @option{-fpie} and 2 for @option{-fPIE}.
19029 @item -fno-jump-tables
19030 @opindex fno-jump-tables
19031 Do not use jump tables for switch statements even where it would be
19032 more efficient than other code generation strategies. This option is
19033 of use in conjunction with @option{-fpic} or @option{-fPIC} for
19034 building code which forms part of a dynamic linker and cannot
19035 reference the address of a jump table. On some targets, jump tables
19036 do not require a GOT and this option is not needed.
19038 @item -ffixed-@var{reg}
19040 Treat the register named @var{reg} as a fixed register; generated code
19041 should never refer to it (except perhaps as a stack pointer, frame
19042 pointer or in some other fixed role).
19044 @var{reg} must be the name of a register. The register names accepted
19045 are machine-specific and are defined in the @code{REGISTER_NAMES}
19046 macro in the machine description macro file.
19048 This flag does not have a negative form, because it specifies a
19051 @item -fcall-used-@var{reg}
19052 @opindex fcall-used
19053 Treat the register named @var{reg} as an allocable register that is
19054 clobbered by function calls. It may be allocated for temporaries or
19055 variables that do not live across a call. Functions compiled this way
19056 will not save and restore the register @var{reg}.
19058 It is an error to used this flag with the frame pointer or stack pointer.
19059 Use of this flag for other registers that have fixed pervasive roles in
19060 the machine's execution model will produce disastrous results.
19062 This flag does not have a negative form, because it specifies a
19065 @item -fcall-saved-@var{reg}
19066 @opindex fcall-saved
19067 Treat the register named @var{reg} as an allocable register saved by
19068 functions. It may be allocated even for temporaries or variables that
19069 live across a call. Functions compiled this way will save and restore
19070 the register @var{reg} if they use it.
19072 It is an error to used this flag with the frame pointer or stack pointer.
19073 Use of this flag for other registers that have fixed pervasive roles in
19074 the machine's execution model will produce disastrous results.
19076 A different sort of disaster will result from the use of this flag for
19077 a register in which function values may be returned.
19079 This flag does not have a negative form, because it specifies a
19082 @item -fpack-struct[=@var{n}]
19083 @opindex fpack-struct
19084 Without a value specified, pack all structure members together without
19085 holes. When a value is specified (which must be a small power of two), pack
19086 structure members according to this value, representing the maximum
19087 alignment (that is, objects with default alignment requirements larger than
19088 this will be output potentially unaligned at the next fitting location.
19090 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
19091 code that is not binary compatible with code generated without that switch.
19092 Additionally, it makes the code suboptimal.
19093 Use it to conform to a non-default application binary interface.
19095 @item -finstrument-functions
19096 @opindex finstrument-functions
19097 Generate instrumentation calls for entry and exit to functions. Just
19098 after function entry and just before function exit, the following
19099 profiling functions will be called with the address of the current
19100 function and its call site. (On some platforms,
19101 @code{__builtin_return_address} does not work beyond the current
19102 function, so the call site information may not be available to the
19103 profiling functions otherwise.)
19106 void __cyg_profile_func_enter (void *this_fn,
19108 void __cyg_profile_func_exit (void *this_fn,
19112 The first argument is the address of the start of the current function,
19113 which may be looked up exactly in the symbol table.
19115 This instrumentation is also done for functions expanded inline in other
19116 functions. The profiling calls will indicate where, conceptually, the
19117 inline function is entered and exited. This means that addressable
19118 versions of such functions must be available. If all your uses of a
19119 function are expanded inline, this may mean an additional expansion of
19120 code size. If you use @samp{extern inline} in your C code, an
19121 addressable version of such functions must be provided. (This is
19122 normally the case anyways, but if you get lucky and the optimizer always
19123 expands the functions inline, you might have gotten away without
19124 providing static copies.)
19126 A function may be given the attribute @code{no_instrument_function}, in
19127 which case this instrumentation will not be done. This can be used, for
19128 example, for the profiling functions listed above, high-priority
19129 interrupt routines, and any functions from which the profiling functions
19130 cannot safely be called (perhaps signal handlers, if the profiling
19131 routines generate output or allocate memory).
19133 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
19134 @opindex finstrument-functions-exclude-file-list
19136 Set the list of functions that are excluded from instrumentation (see
19137 the description of @code{-finstrument-functions}). If the file that
19138 contains a function definition matches with one of @var{file}, then
19139 that function is not instrumented. The match is done on substrings:
19140 if the @var{file} parameter is a substring of the file name, it is
19141 considered to be a match.
19146 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
19150 will exclude any inline function defined in files whose pathnames
19151 contain @code{/bits/stl} or @code{include/sys}.
19153 If, for some reason, you want to include letter @code{','} in one of
19154 @var{sym}, write @code{'\,'}. For example,
19155 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
19156 (note the single quote surrounding the option).
19158 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
19159 @opindex finstrument-functions-exclude-function-list
19161 This is similar to @code{-finstrument-functions-exclude-file-list},
19162 but this option sets the list of function names to be excluded from
19163 instrumentation. The function name to be matched is its user-visible
19164 name, such as @code{vector<int> blah(const vector<int> &)}, not the
19165 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
19166 match is done on substrings: if the @var{sym} parameter is a substring
19167 of the function name, it is considered to be a match. For C99 and C++
19168 extended identifiers, the function name must be given in UTF-8, not
19169 using universal character names.
19171 @item -fstack-check
19172 @opindex fstack-check
19173 Generate code to verify that you do not go beyond the boundary of the
19174 stack. You should specify this flag if you are running in an
19175 environment with multiple threads, but only rarely need to specify it in
19176 a single-threaded environment since stack overflow is automatically
19177 detected on nearly all systems if there is only one stack.
19179 Note that this switch does not actually cause checking to be done; the
19180 operating system or the language runtime must do that. The switch causes
19181 generation of code to ensure that they see the stack being extended.
19183 You can additionally specify a string parameter: @code{no} means no
19184 checking, @code{generic} means force the use of old-style checking,
19185 @code{specific} means use the best checking method and is equivalent
19186 to bare @option{-fstack-check}.
19188 Old-style checking is a generic mechanism that requires no specific
19189 target support in the compiler but comes with the following drawbacks:
19193 Modified allocation strategy for large objects: they will always be
19194 allocated dynamically if their size exceeds a fixed threshold.
19197 Fixed limit on the size of the static frame of functions: when it is
19198 topped by a particular function, stack checking is not reliable and
19199 a warning is issued by the compiler.
19202 Inefficiency: because of both the modified allocation strategy and the
19203 generic implementation, the performances of the code are hampered.
19206 Note that old-style stack checking is also the fallback method for
19207 @code{specific} if no target support has been added in the compiler.
19209 @item -fstack-limit-register=@var{reg}
19210 @itemx -fstack-limit-symbol=@var{sym}
19211 @itemx -fno-stack-limit
19212 @opindex fstack-limit-register
19213 @opindex fstack-limit-symbol
19214 @opindex fno-stack-limit
19215 Generate code to ensure that the stack does not grow beyond a certain value,
19216 either the value of a register or the address of a symbol. If the stack
19217 would grow beyond the value, a signal is raised. For most targets,
19218 the signal is raised before the stack overruns the boundary, so
19219 it is possible to catch the signal without taking special precautions.
19221 For instance, if the stack starts at absolute address @samp{0x80000000}
19222 and grows downwards, you can use the flags
19223 @option{-fstack-limit-symbol=__stack_limit} and
19224 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
19225 of 128KB@. Note that this may only work with the GNU linker.
19227 @item -fsplit-stack
19228 @opindex fsplit-stack
19229 Generate code to automatically split the stack before it overflows.
19230 The resulting program has a discontiguous stack which can only
19231 overflow if the program is unable to allocate any more memory. This
19232 is most useful when running threaded programs, as it is no longer
19233 necessary to calculate a good stack size to use for each thread. This
19234 is currently only implemented for the i386 and x86_64 backends running
19237 When code compiled with @option{-fsplit-stack} calls code compiled
19238 without @option{-fsplit-stack}, there may not be much stack space
19239 available for the latter code to run. If compiling all code,
19240 including library code, with @option{-fsplit-stack} is not an option,
19241 then the linker can fix up these calls so that the code compiled
19242 without @option{-fsplit-stack} always has a large stack. Support for
19243 this is implemented in the gold linker in GNU binutils release 2.21
19246 @item -fleading-underscore
19247 @opindex fleading-underscore
19248 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
19249 change the way C symbols are represented in the object file. One use
19250 is to help link with legacy assembly code.
19252 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
19253 generate code that is not binary compatible with code generated without that
19254 switch. Use it to conform to a non-default application binary interface.
19255 Not all targets provide complete support for this switch.
19257 @item -ftls-model=@var{model}
19258 @opindex ftls-model
19259 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
19260 The @var{model} argument should be one of @code{global-dynamic},
19261 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
19263 The default without @option{-fpic} is @code{initial-exec}; with
19264 @option{-fpic} the default is @code{global-dynamic}.
19266 @item -fvisibility=@var{default|internal|hidden|protected}
19267 @opindex fvisibility
19268 Set the default ELF image symbol visibility to the specified option---all
19269 symbols will be marked with this unless overridden within the code.
19270 Using this feature can very substantially improve linking and
19271 load times of shared object libraries, produce more optimized
19272 code, provide near-perfect API export and prevent symbol clashes.
19273 It is @strong{strongly} recommended that you use this in any shared objects
19276 Despite the nomenclature, @code{default} always means public; i.e.,
19277 available to be linked against from outside the shared object.
19278 @code{protected} and @code{internal} are pretty useless in real-world
19279 usage so the only other commonly used option will be @code{hidden}.
19280 The default if @option{-fvisibility} isn't specified is
19281 @code{default}, i.e., make every
19282 symbol public---this causes the same behavior as previous versions of
19285 A good explanation of the benefits offered by ensuring ELF
19286 symbols have the correct visibility is given by ``How To Write
19287 Shared Libraries'' by Ulrich Drepper (which can be found at
19288 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
19289 solution made possible by this option to marking things hidden when
19290 the default is public is to make the default hidden and mark things
19291 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
19292 and @code{__attribute__ ((visibility("default")))} instead of
19293 @code{__declspec(dllexport)} you get almost identical semantics with
19294 identical syntax. This is a great boon to those working with
19295 cross-platform projects.
19297 For those adding visibility support to existing code, you may find
19298 @samp{#pragma GCC visibility} of use. This works by you enclosing
19299 the declarations you wish to set visibility for with (for example)
19300 @samp{#pragma GCC visibility push(hidden)} and
19301 @samp{#pragma GCC visibility pop}.
19302 Bear in mind that symbol visibility should be viewed @strong{as
19303 part of the API interface contract} and thus all new code should
19304 always specify visibility when it is not the default; i.e., declarations
19305 only for use within the local DSO should @strong{always} be marked explicitly
19306 as hidden as so to avoid PLT indirection overheads---making this
19307 abundantly clear also aids readability and self-documentation of the code.
19308 Note that due to ISO C++ specification requirements, operator new and
19309 operator delete must always be of default visibility.
19311 Be aware that headers from outside your project, in particular system
19312 headers and headers from any other library you use, may not be
19313 expecting to be compiled with visibility other than the default. You
19314 may need to explicitly say @samp{#pragma GCC visibility push(default)}
19315 before including any such headers.
19317 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
19318 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
19319 no modifications. However, this means that calls to @samp{extern}
19320 functions with no explicit visibility will use the PLT, so it is more
19321 effective to use @samp{__attribute ((visibility))} and/or
19322 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
19323 declarations should be treated as hidden.
19325 Note that @samp{-fvisibility} does affect C++ vague linkage
19326 entities. This means that, for instance, an exception class that will
19327 be thrown between DSOs must be explicitly marked with default
19328 visibility so that the @samp{type_info} nodes will be unified between
19331 An overview of these techniques, their benefits and how to use them
19332 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
19334 @item -fstrict-volatile-bitfields
19335 @opindex fstrict-volatile-bitfields
19336 This option should be used if accesses to volatile bitfields (or other
19337 structure fields, although the compiler usually honors those types
19338 anyway) should use a single access of the width of the
19339 field's type, aligned to a natural alignment if possible. For
19340 example, targets with memory-mapped peripheral registers might require
19341 all such accesses to be 16 bits wide; with this flag the user could
19342 declare all peripheral bitfields as ``unsigned short'' (assuming short
19343 is 16 bits on these targets) to force GCC to use 16 bit accesses
19344 instead of, perhaps, a more efficient 32 bit access.
19346 If this option is disabled, the compiler will use the most efficient
19347 instruction. In the previous example, that might be a 32-bit load
19348 instruction, even though that will access bytes that do not contain
19349 any portion of the bitfield, or memory-mapped registers unrelated to
19350 the one being updated.
19352 If the target requires strict alignment, and honoring the field
19353 type would require violating this alignment, a warning is issued.
19354 If the field has @code{packed} attribute, the access is done without
19355 honoring the field type. If the field doesn't have @code{packed}
19356 attribute, the access is done honoring the field type. In both cases,
19357 GCC assumes that the user knows something about the target hardware
19358 that it is unaware of.
19360 The default value of this option is determined by the application binary
19361 interface for the target processor.
19367 @node Environment Variables
19368 @section Environment Variables Affecting GCC
19369 @cindex environment variables
19371 @c man begin ENVIRONMENT
19372 This section describes several environment variables that affect how GCC
19373 operates. Some of them work by specifying directories or prefixes to use
19374 when searching for various kinds of files. Some are used to specify other
19375 aspects of the compilation environment.
19377 Note that you can also specify places to search using options such as
19378 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
19379 take precedence over places specified using environment variables, which
19380 in turn take precedence over those specified by the configuration of GCC@.
19381 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
19382 GNU Compiler Collection (GCC) Internals}.
19387 @c @itemx LC_COLLATE
19389 @c @itemx LC_MONETARY
19390 @c @itemx LC_NUMERIC
19395 @c @findex LC_COLLATE
19396 @findex LC_MESSAGES
19397 @c @findex LC_MONETARY
19398 @c @findex LC_NUMERIC
19402 These environment variables control the way that GCC uses
19403 localization information that allow GCC to work with different
19404 national conventions. GCC inspects the locale categories
19405 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
19406 so. These locale categories can be set to any value supported by your
19407 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
19408 Kingdom encoded in UTF-8.
19410 The @env{LC_CTYPE} environment variable specifies character
19411 classification. GCC uses it to determine the character boundaries in
19412 a string; this is needed for some multibyte encodings that contain quote
19413 and escape characters that would otherwise be interpreted as a string
19416 The @env{LC_MESSAGES} environment variable specifies the language to
19417 use in diagnostic messages.
19419 If the @env{LC_ALL} environment variable is set, it overrides the value
19420 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
19421 and @env{LC_MESSAGES} default to the value of the @env{LANG}
19422 environment variable. If none of these variables are set, GCC
19423 defaults to traditional C English behavior.
19427 If @env{TMPDIR} is set, it specifies the directory to use for temporary
19428 files. GCC uses temporary files to hold the output of one stage of
19429 compilation which is to be used as input to the next stage: for example,
19430 the output of the preprocessor, which is the input to the compiler
19433 @item GCC_COMPARE_DEBUG
19434 @findex GCC_COMPARE_DEBUG
19435 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
19436 @option{-fcompare-debug} to the compiler driver. See the documentation
19437 of this option for more details.
19439 @item GCC_EXEC_PREFIX
19440 @findex GCC_EXEC_PREFIX
19441 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
19442 names of the subprograms executed by the compiler. No slash is added
19443 when this prefix is combined with the name of a subprogram, but you can
19444 specify a prefix that ends with a slash if you wish.
19446 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
19447 an appropriate prefix to use based on the pathname it was invoked with.
19449 If GCC cannot find the subprogram using the specified prefix, it
19450 tries looking in the usual places for the subprogram.
19452 The default value of @env{GCC_EXEC_PREFIX} is
19453 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
19454 the installed compiler. In many cases @var{prefix} is the value
19455 of @code{prefix} when you ran the @file{configure} script.
19457 Other prefixes specified with @option{-B} take precedence over this prefix.
19459 This prefix is also used for finding files such as @file{crt0.o} that are
19462 In addition, the prefix is used in an unusual way in finding the
19463 directories to search for header files. For each of the standard
19464 directories whose name normally begins with @samp{/usr/local/lib/gcc}
19465 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
19466 replacing that beginning with the specified prefix to produce an
19467 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
19468 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
19469 These alternate directories are searched first; the standard directories
19470 come next. If a standard directory begins with the configured
19471 @var{prefix} then the value of @var{prefix} is replaced by
19472 @env{GCC_EXEC_PREFIX} when looking for header files.
19474 @item COMPILER_PATH
19475 @findex COMPILER_PATH
19476 The value of @env{COMPILER_PATH} is a colon-separated list of
19477 directories, much like @env{PATH}. GCC tries the directories thus
19478 specified when searching for subprograms, if it can't find the
19479 subprograms using @env{GCC_EXEC_PREFIX}.
19482 @findex LIBRARY_PATH
19483 The value of @env{LIBRARY_PATH} is a colon-separated list of
19484 directories, much like @env{PATH}. When configured as a native compiler,
19485 GCC tries the directories thus specified when searching for special
19486 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
19487 using GCC also uses these directories when searching for ordinary
19488 libraries for the @option{-l} option (but directories specified with
19489 @option{-L} come first).
19493 @cindex locale definition
19494 This variable is used to pass locale information to the compiler. One way in
19495 which this information is used is to determine the character set to be used
19496 when character literals, string literals and comments are parsed in C and C++.
19497 When the compiler is configured to allow multibyte characters,
19498 the following values for @env{LANG} are recognized:
19502 Recognize JIS characters.
19504 Recognize SJIS characters.
19506 Recognize EUCJP characters.
19509 If @env{LANG} is not defined, or if it has some other value, then the
19510 compiler will use mblen and mbtowc as defined by the default locale to
19511 recognize and translate multibyte characters.
19515 Some additional environments variables affect the behavior of the
19518 @include cppenv.texi
19522 @node Precompiled Headers
19523 @section Using Precompiled Headers
19524 @cindex precompiled headers
19525 @cindex speed of compilation
19527 Often large projects have many header files that are included in every
19528 source file. The time the compiler takes to process these header files
19529 over and over again can account for nearly all of the time required to
19530 build the project. To make builds faster, GCC allows users to
19531 `precompile' a header file; then, if builds can use the precompiled
19532 header file they will be much faster.
19534 To create a precompiled header file, simply compile it as you would any
19535 other file, if necessary using the @option{-x} option to make the driver
19536 treat it as a C or C++ header file. You will probably want to use a
19537 tool like @command{make} to keep the precompiled header up-to-date when
19538 the headers it contains change.
19540 A precompiled header file will be searched for when @code{#include} is
19541 seen in the compilation. As it searches for the included file
19542 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
19543 compiler looks for a precompiled header in each directory just before it
19544 looks for the include file in that directory. The name searched for is
19545 the name specified in the @code{#include} with @samp{.gch} appended. If
19546 the precompiled header file can't be used, it is ignored.
19548 For instance, if you have @code{#include "all.h"}, and you have
19549 @file{all.h.gch} in the same directory as @file{all.h}, then the
19550 precompiled header file will be used if possible, and the original
19551 header will be used otherwise.
19553 Alternatively, you might decide to put the precompiled header file in a
19554 directory and use @option{-I} to ensure that directory is searched
19555 before (or instead of) the directory containing the original header.
19556 Then, if you want to check that the precompiled header file is always
19557 used, you can put a file of the same name as the original header in this
19558 directory containing an @code{#error} command.
19560 This also works with @option{-include}. So yet another way to use
19561 precompiled headers, good for projects not designed with precompiled
19562 header files in mind, is to simply take most of the header files used by
19563 a project, include them from another header file, precompile that header
19564 file, and @option{-include} the precompiled header. If the header files
19565 have guards against multiple inclusion, they will be skipped because
19566 they've already been included (in the precompiled header).
19568 If you need to precompile the same header file for different
19569 languages, targets, or compiler options, you can instead make a
19570 @emph{directory} named like @file{all.h.gch}, and put each precompiled
19571 header in the directory, perhaps using @option{-o}. It doesn't matter
19572 what you call the files in the directory, every precompiled header in
19573 the directory will be considered. The first precompiled header
19574 encountered in the directory that is valid for this compilation will
19575 be used; they're searched in no particular order.
19577 There are many other possibilities, limited only by your imagination,
19578 good sense, and the constraints of your build system.
19580 A precompiled header file can be used only when these conditions apply:
19584 Only one precompiled header can be used in a particular compilation.
19587 A precompiled header can't be used once the first C token is seen. You
19588 can have preprocessor directives before a precompiled header; you can
19589 even include a precompiled header from inside another header, so long as
19590 there are no C tokens before the @code{#include}.
19593 The precompiled header file must be produced for the same language as
19594 the current compilation. You can't use a C precompiled header for a C++
19598 The precompiled header file must have been produced by the same compiler
19599 binary as the current compilation is using.
19602 Any macros defined before the precompiled header is included must
19603 either be defined in the same way as when the precompiled header was
19604 generated, or must not affect the precompiled header, which usually
19605 means that they don't appear in the precompiled header at all.
19607 The @option{-D} option is one way to define a macro before a
19608 precompiled header is included; using a @code{#define} can also do it.
19609 There are also some options that define macros implicitly, like
19610 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
19613 @item If debugging information is output when using the precompiled
19614 header, using @option{-g} or similar, the same kind of debugging information
19615 must have been output when building the precompiled header. However,
19616 a precompiled header built using @option{-g} can be used in a compilation
19617 when no debugging information is being output.
19619 @item The same @option{-m} options must generally be used when building
19620 and using the precompiled header. @xref{Submodel Options},
19621 for any cases where this rule is relaxed.
19623 @item Each of the following options must be the same when building and using
19624 the precompiled header:
19626 @gccoptlist{-fexceptions}
19629 Some other command-line options starting with @option{-f},
19630 @option{-p}, or @option{-O} must be defined in the same way as when
19631 the precompiled header was generated. At present, it's not clear
19632 which options are safe to change and which are not; the safest choice
19633 is to use exactly the same options when generating and using the
19634 precompiled header. The following are known to be safe:
19636 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
19637 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
19638 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
19643 For all of these except the last, the compiler will automatically
19644 ignore the precompiled header if the conditions aren't met. If you
19645 find an option combination that doesn't work and doesn't cause the
19646 precompiled header to be ignored, please consider filing a bug report,
19649 If you do use differing options when generating and using the
19650 precompiled header, the actual behavior will be a mixture of the
19651 behavior for the options. For instance, if you use @option{-g} to
19652 generate the precompiled header but not when using it, you may or may
19653 not get debugging information for routines in the precompiled header.