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}}
926 @emph{TILE-Gx Options}
927 @gccoptlist{-mcpu=CPU -m32 -m64}
929 @emph{TILEPro Options}
930 @gccoptlist{-mcpu=CPU -m32}
933 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
934 -mprolog-function -mno-prolog-function -mspace @gol
935 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
936 -mapp-regs -mno-app-regs @gol
937 -mdisable-callt -mno-disable-callt @gol
940 -mv850e1 -mv850es @gol
945 @gccoptlist{-mg -mgnu -munix}
947 @emph{VxWorks Options}
948 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
949 -Xbind-lazy -Xbind-now}
951 @emph{x86-64 Options}
952 See i386 and x86-64 Options.
954 @emph{Xstormy16 Options}
957 @emph{Xtensa Options}
958 @gccoptlist{-mconst16 -mno-const16 @gol
959 -mfused-madd -mno-fused-madd @gol
961 -mserialize-volatile -mno-serialize-volatile @gol
962 -mtext-section-literals -mno-text-section-literals @gol
963 -mtarget-align -mno-target-align @gol
964 -mlongcalls -mno-longcalls}
966 @emph{zSeries Options}
967 See S/390 and zSeries Options.
969 @item Code Generation Options
970 @xref{Code Gen Options,,Options for Code Generation Conventions}.
971 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
972 -ffixed-@var{reg} -fexceptions @gol
973 -fnon-call-exceptions -funwind-tables @gol
974 -fasynchronous-unwind-tables @gol
975 -finhibit-size-directive -finstrument-functions @gol
976 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
977 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
978 -fno-common -fno-ident @gol
979 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
980 -fno-jump-tables @gol
981 -frecord-gcc-switches @gol
982 -freg-struct-return -fshort-enums @gol
983 -fshort-double -fshort-wchar @gol
984 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
985 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
986 -fno-stack-limit -fsplit-stack @gol
987 -fleading-underscore -ftls-model=@var{model} @gol
988 -ftrapv -fwrapv -fbounds-check @gol
989 -fvisibility -fstrict-volatile-bitfields}
993 * Overall Options:: Controlling the kind of output:
994 an executable, object files, assembler files,
995 or preprocessed source.
996 * C Dialect Options:: Controlling the variant of C language compiled.
997 * C++ Dialect Options:: Variations on C++.
998 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
1000 * Language Independent Options:: Controlling how diagnostics should be
1002 * Warning Options:: How picky should the compiler be?
1003 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
1004 * Optimize Options:: How much optimization?
1005 * Preprocessor Options:: Controlling header files and macro definitions.
1006 Also, getting dependency information for Make.
1007 * Assembler Options:: Passing options to the assembler.
1008 * Link Options:: Specifying libraries and so on.
1009 * Directory Options:: Where to find header files and libraries.
1010 Where to find the compiler executable files.
1011 * Spec Files:: How to pass switches to sub-processes.
1012 * Target Options:: Running a cross-compiler, or an old version of GCC.
1015 @node Overall Options
1016 @section Options Controlling the Kind of Output
1018 Compilation can involve up to four stages: preprocessing, compilation
1019 proper, assembly and linking, always in that order. GCC is capable of
1020 preprocessing and compiling several files either into several
1021 assembler input files, or into one assembler input file; then each
1022 assembler input file produces an object file, and linking combines all
1023 the object files (those newly compiled, and those specified as input)
1024 into an executable file.
1026 @cindex file name suffix
1027 For any given input file, the file name suffix determines what kind of
1028 compilation is done:
1032 C source code that must be preprocessed.
1035 C source code that should not be preprocessed.
1038 C++ source code that should not be preprocessed.
1041 Objective-C source code. Note that you must link with the @file{libobjc}
1042 library to make an Objective-C program work.
1045 Objective-C source code that should not be preprocessed.
1049 Objective-C++ source code. Note that you must link with the @file{libobjc}
1050 library to make an Objective-C++ program work. Note that @samp{.M} refers
1051 to a literal capital M@.
1053 @item @var{file}.mii
1054 Objective-C++ source code that should not be preprocessed.
1057 C, C++, Objective-C or Objective-C++ header file to be turned into a
1058 precompiled header (default), or C, C++ header file to be turned into an
1059 Ada spec (via the @option{-fdump-ada-spec} switch).
1062 @itemx @var{file}.cp
1063 @itemx @var{file}.cxx
1064 @itemx @var{file}.cpp
1065 @itemx @var{file}.CPP
1066 @itemx @var{file}.c++
1068 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1069 the last two letters must both be literally @samp{x}. Likewise,
1070 @samp{.C} refers to a literal capital C@.
1074 Objective-C++ source code that must be preprocessed.
1076 @item @var{file}.mii
1077 Objective-C++ source code that should not be preprocessed.
1081 @itemx @var{file}.hp
1082 @itemx @var{file}.hxx
1083 @itemx @var{file}.hpp
1084 @itemx @var{file}.HPP
1085 @itemx @var{file}.h++
1086 @itemx @var{file}.tcc
1087 C++ header file to be turned into a precompiled header or Ada spec.
1090 @itemx @var{file}.for
1091 @itemx @var{file}.ftn
1092 Fixed form Fortran source code that should not be preprocessed.
1095 @itemx @var{file}.FOR
1096 @itemx @var{file}.fpp
1097 @itemx @var{file}.FPP
1098 @itemx @var{file}.FTN
1099 Fixed form Fortran source code that must be preprocessed (with the traditional
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 that should not be preprocessed.
1108 @item @var{file}.F90
1109 @itemx @var{file}.F95
1110 @itemx @var{file}.F03
1111 @itemx @var{file}.F08
1112 Free form Fortran source code that must be preprocessed (with the
1113 traditional preprocessor).
1118 @c FIXME: Descriptions of Java file types.
1124 @item @var{file}.ads
1125 Ada source code file that contains a library unit declaration (a
1126 declaration of a package, subprogram, or generic, or a generic
1127 instantiation), or a library unit renaming declaration (a package,
1128 generic, or subprogram renaming declaration). Such files are also
1131 @item @var{file}.adb
1132 Ada source code file containing a library unit body (a subprogram or
1133 package body). Such files are also called @dfn{bodies}.
1135 @c GCC also knows about some suffixes for languages not yet included:
1146 @itemx @var{file}.sx
1147 Assembler code that must be preprocessed.
1150 An object file to be fed straight into linking.
1151 Any file name with no recognized suffix is treated this way.
1155 You can specify the input language explicitly with the @option{-x} option:
1158 @item -x @var{language}
1159 Specify explicitly the @var{language} for the following input files
1160 (rather than letting the compiler choose a default based on the file
1161 name suffix). This option applies to all following input files until
1162 the next @option{-x} option. Possible values for @var{language} are:
1164 c c-header cpp-output
1165 c++ c++-header c++-cpp-output
1166 objective-c objective-c-header objective-c-cpp-output
1167 objective-c++ objective-c++-header objective-c++-cpp-output
1168 assembler assembler-with-cpp
1170 f77 f77-cpp-input f95 f95-cpp-input
1176 Turn off any specification of a language, so that subsequent files are
1177 handled according to their file name suffixes (as they are if @option{-x}
1178 has not been used at all).
1180 @item -pass-exit-codes
1181 @opindex pass-exit-codes
1182 Normally the @command{gcc} program will exit with the code of 1 if any
1183 phase of the compiler returns a non-success return code. If you specify
1184 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1185 numerically highest error produced by any phase that returned an error
1186 indication. The C, C++, and Fortran frontends return 4, if an internal
1187 compiler error is encountered.
1190 If you only want some of the stages of compilation, you can use
1191 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1192 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1193 @command{gcc} is to stop. Note that some combinations (for example,
1194 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1199 Compile or assemble the source files, but do not link. The linking
1200 stage simply is not done. The ultimate output is in the form of an
1201 object file for each source file.
1203 By default, the object file name for a source file is made by replacing
1204 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1206 Unrecognized input files, not requiring compilation or assembly, are
1211 Stop after the stage of compilation proper; do not assemble. The output
1212 is in the form of an assembler code file for each non-assembler input
1215 By default, the assembler file name for a source file is made by
1216 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1218 Input files that don't require compilation are ignored.
1222 Stop after the preprocessing stage; do not run the compiler proper. The
1223 output is in the form of preprocessed source code, which is sent to the
1226 Input files that don't require preprocessing are ignored.
1228 @cindex output file option
1231 Place output in file @var{file}. This applies regardless to whatever
1232 sort of output is being produced, whether it be an executable file,
1233 an object file, an assembler file or preprocessed C code.
1235 If @option{-o} is not specified, the default is to put an executable
1236 file in @file{a.out}, the object file for
1237 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1238 assembler file in @file{@var{source}.s}, a precompiled header file in
1239 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1244 Print (on standard error output) the commands executed to run the stages
1245 of compilation. Also print the version number of the compiler driver
1246 program and of the preprocessor and the compiler proper.
1250 Like @option{-v} except the commands are not executed and arguments
1251 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1252 This is useful for shell scripts to capture the driver-generated command lines.
1256 Use pipes rather than temporary files for communication between the
1257 various stages of compilation. This fails to work on some systems where
1258 the assembler is unable to read from a pipe; but the GNU assembler has
1263 Print (on the standard output) a description of the command-line options
1264 understood by @command{gcc}. If the @option{-v} option is also specified
1265 then @option{--help} will also be passed on to the various processes
1266 invoked by @command{gcc}, so that they can display the command-line options
1267 they accept. If the @option{-Wextra} option has also been specified
1268 (prior to the @option{--help} option), then command-line options that
1269 have no documentation associated with them will also be displayed.
1272 @opindex target-help
1273 Print (on the standard output) a description of target-specific command-line
1274 options for each tool. For some targets extra target-specific
1275 information may also be printed.
1277 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1278 Print (on the standard output) a description of the command-line
1279 options understood by the compiler that fit into all specified classes
1280 and qualifiers. These are the supported classes:
1283 @item @samp{optimizers}
1284 This will display all of the optimization options supported by the
1287 @item @samp{warnings}
1288 This will display all of the options controlling warning messages
1289 produced by the compiler.
1292 This will display target-specific options. Unlike the
1293 @option{--target-help} option however, target-specific options of the
1294 linker and assembler will not be displayed. This is because those
1295 tools do not currently support the extended @option{--help=} syntax.
1298 This will display the values recognized by the @option{--param}
1301 @item @var{language}
1302 This will display the options supported for @var{language}, where
1303 @var{language} is the name of one of the languages supported in this
1307 This will display the options that are common to all languages.
1310 These are the supported qualifiers:
1313 @item @samp{undocumented}
1314 Display only those options that are undocumented.
1317 Display options taking an argument that appears after an equal
1318 sign in the same continuous piece of text, such as:
1319 @samp{--help=target}.
1321 @item @samp{separate}
1322 Display options taking an argument that appears as a separate word
1323 following the original option, such as: @samp{-o output-file}.
1326 Thus for example to display all the undocumented target-specific
1327 switches supported by the compiler the following can be used:
1330 --help=target,undocumented
1333 The sense of a qualifier can be inverted by prefixing it with the
1334 @samp{^} character, so for example to display all binary warning
1335 options (i.e., ones that are either on or off and that do not take an
1336 argument) that have a description, use:
1339 --help=warnings,^joined,^undocumented
1342 The argument to @option{--help=} should not consist solely of inverted
1345 Combining several classes is possible, although this usually
1346 restricts the output by so much that there is nothing to display. One
1347 case where it does work however is when one of the classes is
1348 @var{target}. So for example to display all the target-specific
1349 optimization options the following can be used:
1352 --help=target,optimizers
1355 The @option{--help=} option can be repeated on the command line. Each
1356 successive use will display its requested class of options, skipping
1357 those that have already been displayed.
1359 If the @option{-Q} option appears on the command line before the
1360 @option{--help=} option, then the descriptive text displayed by
1361 @option{--help=} is changed. Instead of describing the displayed
1362 options, an indication is given as to whether the option is enabled,
1363 disabled or set to a specific value (assuming that the compiler
1364 knows this at the point where the @option{--help=} option is used).
1366 Here is a truncated example from the ARM port of @command{gcc}:
1369 % gcc -Q -mabi=2 --help=target -c
1370 The following options are target specific:
1372 -mabort-on-noreturn [disabled]
1376 The output is sensitive to the effects of previous command-line
1377 options, so for example it is possible to find out which optimizations
1378 are enabled at @option{-O2} by using:
1381 -Q -O2 --help=optimizers
1384 Alternatively you can discover which binary optimizations are enabled
1385 by @option{-O3} by using:
1388 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1389 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1390 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1393 @item -no-canonical-prefixes
1394 @opindex no-canonical-prefixes
1395 Do not expand any symbolic links, resolve references to @samp{/../}
1396 or @samp{/./}, or make the path absolute when generating a relative
1401 Display the version number and copyrights of the invoked GCC@.
1405 Invoke all subcommands under a wrapper program. The name of the
1406 wrapper program and its parameters are passed as a comma separated
1410 gcc -c t.c -wrapper gdb,--args
1413 This will invoke all subprograms of @command{gcc} under
1414 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1415 @samp{gdb --args cc1 @dots{}}.
1417 @item -fplugin=@var{name}.so
1418 Load the plugin code in file @var{name}.so, assumed to be a
1419 shared object to be dlopen'd by the compiler. The base name of
1420 the shared object file is used to identify the plugin for the
1421 purposes of argument parsing (See
1422 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1423 Each plugin should define the callback functions specified in the
1426 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1427 Define an argument called @var{key} with a value of @var{value}
1428 for the plugin called @var{name}.
1430 @item -fdump-ada-spec@r{[}-slim@r{]}
1431 For C and C++ source and include files, generate corresponding Ada
1432 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1433 GNAT User's Guide}, which provides detailed documentation on this feature.
1435 @item -fdump-go-spec=@var{file}
1436 For input files in any language, generate corresponding Go
1437 declarations in @var{file}. This generates Go @code{const},
1438 @code{type}, @code{var}, and @code{func} declarations which may be a
1439 useful way to start writing a Go interface to code written in some
1442 @include @value{srcdir}/../libiberty/at-file.texi
1446 @section Compiling C++ Programs
1448 @cindex suffixes for C++ source
1449 @cindex C++ source file suffixes
1450 C++ source files conventionally use one of the suffixes @samp{.C},
1451 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1452 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1453 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1454 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1455 files with these names and compiles them as C++ programs even if you
1456 call the compiler the same way as for compiling C programs (usually
1457 with the name @command{gcc}).
1461 However, the use of @command{gcc} does not add the C++ library.
1462 @command{g++} is a program that calls GCC and treats @samp{.c},
1463 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1464 files unless @option{-x} is used, and automatically specifies linking
1465 against the C++ library. This program is also useful when
1466 precompiling a C header file with a @samp{.h} extension for use in C++
1467 compilations. On many systems, @command{g++} is also installed with
1468 the name @command{c++}.
1470 @cindex invoking @command{g++}
1471 When you compile C++ programs, you may specify many of the same
1472 command-line options that you use for compiling programs in any
1473 language; or command-line options meaningful for C and related
1474 languages; or options that are meaningful only for C++ programs.
1475 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1476 explanations of options for languages related to C@.
1477 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1478 explanations of options that are meaningful only for C++ programs.
1480 @node C Dialect Options
1481 @section Options Controlling C Dialect
1482 @cindex dialect options
1483 @cindex language dialect options
1484 @cindex options, dialect
1486 The following options control the dialect of C (or languages derived
1487 from C, such as C++, Objective-C and Objective-C++) that the compiler
1491 @cindex ANSI support
1495 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1496 equivalent to @samp{-std=c++98}.
1498 This turns off certain features of GCC that are incompatible with ISO
1499 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1500 such as the @code{asm} and @code{typeof} keywords, and
1501 predefined macros such as @code{unix} and @code{vax} that identify the
1502 type of system you are using. It also enables the undesirable and
1503 rarely used ISO trigraph feature. For the C compiler,
1504 it disables recognition of C++ style @samp{//} comments as well as
1505 the @code{inline} keyword.
1507 The alternate keywords @code{__asm__}, @code{__extension__},
1508 @code{__inline__} and @code{__typeof__} continue to work despite
1509 @option{-ansi}. You would not want to use them in an ISO C program, of
1510 course, but it is useful to put them in header files that might be included
1511 in compilations done with @option{-ansi}. Alternate predefined macros
1512 such as @code{__unix__} and @code{__vax__} are also available, with or
1513 without @option{-ansi}.
1515 The @option{-ansi} option does not cause non-ISO programs to be
1516 rejected gratuitously. For that, @option{-pedantic} is required in
1517 addition to @option{-ansi}. @xref{Warning Options}.
1519 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1520 option is used. Some header files may notice this macro and refrain
1521 from declaring certain functions or defining certain macros that the
1522 ISO standard doesn't call for; this is to avoid interfering with any
1523 programs that might use these names for other things.
1525 Functions that would normally be built in but do not have semantics
1526 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1527 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1528 built-in functions provided by GCC}, for details of the functions
1533 Determine the language standard. @xref{Standards,,Language Standards
1534 Supported by GCC}, for details of these standard versions. This option
1535 is currently only supported when compiling C or C++.
1537 The compiler can accept several base standards, such as @samp{c90} or
1538 @samp{c++98}, and GNU dialects of those standards, such as
1539 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1540 compiler will accept all programs following that standard and those
1541 using GNU extensions that do not contradict it. For example,
1542 @samp{-std=c90} turns off certain features of GCC that are
1543 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1544 keywords, but not other GNU extensions that do not have a meaning in
1545 ISO C90, such as omitting the middle term of a @code{?:}
1546 expression. On the other hand, by specifying a GNU dialect of a
1547 standard, all features the compiler support are enabled, even when
1548 those features change the meaning of the base standard and some
1549 strict-conforming programs may be rejected. The particular standard
1550 is used by @option{-pedantic} to identify which features are GNU
1551 extensions given that version of the standard. For example
1552 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1553 comments, while @samp{-std=gnu99 -pedantic} would not.
1555 A value for this option must be provided; possible values are
1561 Support all ISO C90 programs (certain GNU extensions that conflict
1562 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1564 @item iso9899:199409
1565 ISO C90 as modified in amendment 1.
1571 ISO C99. Note that this standard is not yet fully supported; see
1572 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1573 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1578 ISO C11, the 2011 revision of the ISO C standard.
1579 Support is incomplete and experimental. The name @samp{c1x} is
1584 GNU dialect of ISO C90 (including some C99 features). This
1585 is the default for C code.
1589 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1590 this will become the default. The name @samp{gnu9x} is deprecated.
1594 GNU dialect of ISO C11. Support is incomplete and experimental. The
1595 name @samp{gnu1x} is deprecated.
1598 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1602 GNU dialect of @option{-std=c++98}. This is the default for
1606 The 2011 ISO C++ standard plus amendments. Support for C++11 is still
1607 experimental, and may change in incompatible ways in future releases.
1610 GNU dialect of @option{-std=c++11}. Support for C++11 is still
1611 experimental, and may change in incompatible ways in future releases.
1614 @item -fgnu89-inline
1615 @opindex fgnu89-inline
1616 The option @option{-fgnu89-inline} tells GCC to use the traditional
1617 GNU semantics for @code{inline} functions when in C99 mode.
1618 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1619 is accepted and ignored by GCC versions 4.1.3 up to but not including
1620 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1621 C99 mode. Using this option is roughly equivalent to adding the
1622 @code{gnu_inline} function attribute to all inline functions
1623 (@pxref{Function Attributes}).
1625 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1626 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1627 specifies the default behavior). This option was first supported in
1628 GCC 4.3. This option is not supported in @option{-std=c90} or
1629 @option{-std=gnu90} mode.
1631 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1632 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1633 in effect for @code{inline} functions. @xref{Common Predefined
1634 Macros,,,cpp,The C Preprocessor}.
1636 @item -aux-info @var{filename}
1638 Output to the given filename prototyped declarations for all functions
1639 declared and/or defined in a translation unit, including those in header
1640 files. This option is silently ignored in any language other than C@.
1642 Besides declarations, the file indicates, in comments, the origin of
1643 each declaration (source file and line), whether the declaration was
1644 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1645 @samp{O} for old, respectively, in the first character after the line
1646 number and the colon), and whether it came from a declaration or a
1647 definition (@samp{C} or @samp{F}, respectively, in the following
1648 character). In the case of function definitions, a K&R-style list of
1649 arguments followed by their declarations is also provided, inside
1650 comments, after the declaration.
1652 @item -fallow-parameterless-variadic-functions
1653 Accept variadic functions without named parameters.
1655 Although it is possible to define such a function, this is not very
1656 useful as it is not possible to read the arguments. This is only
1657 supported for C as this construct is allowed by C++.
1661 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1662 keyword, so that code can use these words as identifiers. You can use
1663 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1664 instead. @option{-ansi} implies @option{-fno-asm}.
1666 In C++, this switch only affects the @code{typeof} keyword, since
1667 @code{asm} and @code{inline} are standard keywords. You may want to
1668 use the @option{-fno-gnu-keywords} flag instead, which has the same
1669 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1670 switch only affects the @code{asm} and @code{typeof} keywords, since
1671 @code{inline} is a standard keyword in ISO C99.
1674 @itemx -fno-builtin-@var{function}
1675 @opindex fno-builtin
1676 @cindex built-in functions
1677 Don't recognize built-in functions that do not begin with
1678 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1679 functions provided by GCC}, for details of the functions affected,
1680 including those which are not built-in functions when @option{-ansi} or
1681 @option{-std} options for strict ISO C conformance are used because they
1682 do not have an ISO standard meaning.
1684 GCC normally generates special code to handle certain built-in functions
1685 more efficiently; for instance, calls to @code{alloca} may become single
1686 instructions which adjust the stack directly, and calls to @code{memcpy}
1687 may become inline copy loops. The resulting code is often both smaller
1688 and faster, but since the function calls no longer appear as such, you
1689 cannot set a breakpoint on those calls, nor can you change the behavior
1690 of the functions by linking with a different library. In addition,
1691 when a function is recognized as a built-in function, GCC may use
1692 information about that function to warn about problems with calls to
1693 that function, or to generate more efficient code, even if the
1694 resulting code still contains calls to that function. For example,
1695 warnings are given with @option{-Wformat} for bad calls to
1696 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1697 known not to modify global memory.
1699 With the @option{-fno-builtin-@var{function}} option
1700 only the built-in function @var{function} is
1701 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1702 function is named that is not built-in in this version of GCC, this
1703 option is ignored. There is no corresponding
1704 @option{-fbuiltin-@var{function}} option; if you wish to enable
1705 built-in functions selectively when using @option{-fno-builtin} or
1706 @option{-ffreestanding}, you may define macros such as:
1709 #define abs(n) __builtin_abs ((n))
1710 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1715 @cindex hosted environment
1717 Assert that compilation takes place in a hosted environment. This implies
1718 @option{-fbuiltin}. A hosted environment is one in which the
1719 entire standard library is available, and in which @code{main} has a return
1720 type of @code{int}. Examples are nearly everything except a kernel.
1721 This is equivalent to @option{-fno-freestanding}.
1723 @item -ffreestanding
1724 @opindex ffreestanding
1725 @cindex hosted environment
1727 Assert that compilation takes place in a freestanding environment. This
1728 implies @option{-fno-builtin}. A freestanding environment
1729 is one in which the standard library may not exist, and program startup may
1730 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1731 This is equivalent to @option{-fno-hosted}.
1733 @xref{Standards,,Language Standards Supported by GCC}, for details of
1734 freestanding and hosted environments.
1738 @cindex OpenMP parallel
1739 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1740 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1741 compiler generates parallel code according to the OpenMP Application
1742 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1743 implies @option{-pthread}, and thus is only supported on targets that
1744 have support for @option{-pthread}.
1748 When the option @option{-fgnu-tm} is specified, the compiler will
1749 generate code for the Linux variant of Intel's current Transactional
1750 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1751 an experimental feature whose interface may change in future versions
1752 of GCC, as the official specification changes. Please note that not
1753 all architectures are supported for this feature.
1755 For more information on GCC's support for transactional memory,
1756 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1757 Transactional Memory Library}.
1759 Note that the transactional memory feature is not supported with
1760 non-call exceptions (@option{-fnon-call-exceptions}).
1762 @item -fms-extensions
1763 @opindex fms-extensions
1764 Accept some non-standard constructs used in Microsoft header files.
1766 In C++ code, this allows member names in structures to be similar
1767 to previous types declarations.
1776 Some cases of unnamed fields in structures and unions are only
1777 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1778 fields within structs/unions}, for details.
1780 @item -fplan9-extensions
1781 Accept some non-standard constructs used in Plan 9 code.
1783 This enables @option{-fms-extensions}, permits passing pointers to
1784 structures with anonymous fields to functions that expect pointers to
1785 elements of the type of the field, and permits referring to anonymous
1786 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1787 struct/union fields within structs/unions}, for details. This is only
1788 supported for C, not C++.
1792 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1793 options for strict ISO C conformance) implies @option{-trigraphs}.
1795 @item -no-integrated-cpp
1796 @opindex no-integrated-cpp
1797 Performs a compilation in two passes: preprocessing and compiling. This
1798 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1799 @option{-B} option. The user supplied compilation step can then add in
1800 an additional preprocessing step after normal preprocessing but before
1801 compiling. The default is to use the integrated cpp (internal cpp)
1803 The semantics of this option will change if "cc1", "cc1plus", and
1804 "cc1obj" are merged.
1806 @cindex traditional C language
1807 @cindex C language, traditional
1809 @itemx -traditional-cpp
1810 @opindex traditional-cpp
1811 @opindex traditional
1812 Formerly, these options caused GCC to attempt to emulate a pre-standard
1813 C compiler. They are now only supported with the @option{-E} switch.
1814 The preprocessor continues to support a pre-standard mode. See the GNU
1815 CPP manual for details.
1817 @item -fcond-mismatch
1818 @opindex fcond-mismatch
1819 Allow conditional expressions with mismatched types in the second and
1820 third arguments. The value of such an expression is void. This option
1821 is not supported for C++.
1823 @item -flax-vector-conversions
1824 @opindex flax-vector-conversions
1825 Allow implicit conversions between vectors with differing numbers of
1826 elements and/or incompatible element types. This option should not be
1829 @item -funsigned-char
1830 @opindex funsigned-char
1831 Let the type @code{char} be unsigned, like @code{unsigned char}.
1833 Each kind of machine has a default for what @code{char} should
1834 be. It is either like @code{unsigned char} by default or like
1835 @code{signed char} by default.
1837 Ideally, a portable program should always use @code{signed char} or
1838 @code{unsigned char} when it depends on the signedness of an object.
1839 But many programs have been written to use plain @code{char} and
1840 expect it to be signed, or expect it to be unsigned, depending on the
1841 machines they were written for. This option, and its inverse, let you
1842 make such a program work with the opposite default.
1844 The type @code{char} is always a distinct type from each of
1845 @code{signed char} or @code{unsigned char}, even though its behavior
1846 is always just like one of those two.
1849 @opindex fsigned-char
1850 Let the type @code{char} be signed, like @code{signed char}.
1852 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1853 the negative form of @option{-funsigned-char}. Likewise, the option
1854 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1856 @item -fsigned-bitfields
1857 @itemx -funsigned-bitfields
1858 @itemx -fno-signed-bitfields
1859 @itemx -fno-unsigned-bitfields
1860 @opindex fsigned-bitfields
1861 @opindex funsigned-bitfields
1862 @opindex fno-signed-bitfields
1863 @opindex fno-unsigned-bitfields
1864 These options control whether a bit-field is signed or unsigned, when the
1865 declaration does not use either @code{signed} or @code{unsigned}. By
1866 default, such a bit-field is signed, because this is consistent: the
1867 basic integer types such as @code{int} are signed types.
1870 @node C++ Dialect Options
1871 @section Options Controlling C++ Dialect
1873 @cindex compiler options, C++
1874 @cindex C++ options, command-line
1875 @cindex options, C++
1876 This section describes the command-line options that are only meaningful
1877 for C++ programs; but you can also use most of the GNU compiler options
1878 regardless of what language your program is in. For example, you
1879 might compile a file @code{firstClass.C} like this:
1882 g++ -g -frepo -O -c firstClass.C
1886 In this example, only @option{-frepo} is an option meant
1887 only for C++ programs; you can use the other options with any
1888 language supported by GCC@.
1890 Here is a list of options that are @emph{only} for compiling C++ programs:
1894 @item -fabi-version=@var{n}
1895 @opindex fabi-version
1896 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1897 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1898 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1899 the version that conforms most closely to the C++ ABI specification.
1900 Therefore, the ABI obtained using version 0 will change as ABI bugs
1903 The default is version 2.
1905 Version 3 corrects an error in mangling a constant address as a
1908 Version 4, which first appeared in G++ 4.5, implements a standard
1909 mangling for vector types.
1911 Version 5, which first appeared in G++ 4.6, corrects the mangling of
1912 attribute const/volatile on function pointer types, decltype of a
1913 plain decl, and use of a function parameter in the declaration of
1916 Version 6, which first appeared in G++ 4.7, corrects the promotion
1917 behavior of C++11 scoped enums and the mangling of template argument
1918 packs, const/static_cast, prefix ++ and --, and a class scope function
1919 used as a template argument.
1921 See also @option{-Wabi}.
1923 @item -fno-access-control
1924 @opindex fno-access-control
1925 Turn off all access checking. This switch is mainly useful for working
1926 around bugs in the access control code.
1930 Check that the pointer returned by @code{operator new} is non-null
1931 before attempting to modify the storage allocated. This check is
1932 normally unnecessary because the C++ standard specifies that
1933 @code{operator new} will only return @code{0} if it is declared
1934 @samp{throw()}, in which case the compiler will always check the
1935 return value even without this option. In all other cases, when
1936 @code{operator new} has a non-empty exception specification, memory
1937 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1938 @samp{new (nothrow)}.
1940 @item -fconserve-space
1941 @opindex fconserve-space
1942 Put uninitialized or run-time-initialized global variables into the
1943 common segment, as C does. This saves space in the executable at the
1944 cost of not diagnosing duplicate definitions. If you compile with this
1945 flag and your program mysteriously crashes after @code{main()} has
1946 completed, you may have an object that is being destroyed twice because
1947 two definitions were merged.
1949 This option is no longer useful on most targets, now that support has
1950 been added for putting variables into BSS without making them common.
1952 @item -fconstexpr-depth=@var{n}
1953 @opindex fconstexpr-depth
1954 Set the maximum nested evaluation depth for C++11 constexpr functions
1955 to @var{n}. A limit is needed to detect endless recursion during
1956 constant expression evaluation. The minimum specified by the standard
1959 @item -fdeduce-init-list
1960 @opindex fdeduce-init-list
1961 Enable deduction of a template type parameter as
1962 std::initializer_list from a brace-enclosed initializer list, i.e.
1965 template <class T> auto forward(T t) -> decltype (realfn (t))
1972 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1976 This deduction was implemented as a possible extension to the
1977 originally proposed semantics for the C++11 standard, but was not part
1978 of the final standard, so it is disabled by default. This option is
1979 deprecated, and may be removed in a future version of G++.
1981 @item -ffriend-injection
1982 @opindex ffriend-injection
1983 Inject friend functions into the enclosing namespace, so that they are
1984 visible outside the scope of the class in which they are declared.
1985 Friend functions were documented to work this way in the old Annotated
1986 C++ Reference Manual, and versions of G++ before 4.1 always worked
1987 that way. However, in ISO C++ a friend function that is not declared
1988 in an enclosing scope can only be found using argument dependent
1989 lookup. This option causes friends to be injected as they were in
1992 This option is for compatibility, and may be removed in a future
1995 @item -fno-elide-constructors
1996 @opindex fno-elide-constructors
1997 The C++ standard allows an implementation to omit creating a temporary
1998 that is only used to initialize another object of the same type.
1999 Specifying this option disables that optimization, and forces G++ to
2000 call the copy constructor in all cases.
2002 @item -fno-enforce-eh-specs
2003 @opindex fno-enforce-eh-specs
2004 Don't generate code to check for violation of exception specifications
2005 at run time. This option violates the C++ standard, but may be useful
2006 for reducing code size in production builds, much like defining
2007 @samp{NDEBUG}. This does not give user code permission to throw
2008 exceptions in violation of the exception specifications; the compiler
2009 will still optimize based on the specifications, so throwing an
2010 unexpected exception will result in undefined behavior.
2013 @itemx -fno-for-scope
2015 @opindex fno-for-scope
2016 If @option{-ffor-scope} is specified, the scope of variables declared in
2017 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2018 as specified by the C++ standard.
2019 If @option{-fno-for-scope} is specified, the scope of variables declared in
2020 a @i{for-init-statement} extends to the end of the enclosing scope,
2021 as was the case in old versions of G++, and other (traditional)
2022 implementations of C++.
2024 The default if neither flag is given to follow the standard,
2025 but to allow and give a warning for old-style code that would
2026 otherwise be invalid, or have different behavior.
2028 @item -fno-gnu-keywords
2029 @opindex fno-gnu-keywords
2030 Do not recognize @code{typeof} as a keyword, so that code can use this
2031 word as an identifier. You can use the keyword @code{__typeof__} instead.
2032 @option{-ansi} implies @option{-fno-gnu-keywords}.
2034 @item -fno-implicit-templates
2035 @opindex fno-implicit-templates
2036 Never emit code for non-inline templates that are instantiated
2037 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2038 @xref{Template Instantiation}, for more information.
2040 @item -fno-implicit-inline-templates
2041 @opindex fno-implicit-inline-templates
2042 Don't emit code for implicit instantiations of inline templates, either.
2043 The default is to handle inlines differently so that compiles with and
2044 without optimization will need the same set of explicit instantiations.
2046 @item -fno-implement-inlines
2047 @opindex fno-implement-inlines
2048 To save space, do not emit out-of-line copies of inline functions
2049 controlled by @samp{#pragma implementation}. This will cause linker
2050 errors if these functions are not inlined everywhere they are called.
2052 @item -fms-extensions
2053 @opindex fms-extensions
2054 Disable pedantic warnings about constructs used in MFC, such as implicit
2055 int and getting a pointer to member function via non-standard syntax.
2057 @item -fno-nonansi-builtins
2058 @opindex fno-nonansi-builtins
2059 Disable built-in declarations of functions that are not mandated by
2060 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2061 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2064 @opindex fnothrow-opt
2065 Treat a @code{throw()} exception specification as though it were a
2066 @code{noexcept} specification to reduce or eliminate the text size
2067 overhead relative to a function with no exception specification. If
2068 the function has local variables of types with non-trivial
2069 destructors, the exception specification will actually make the
2070 function smaller because the EH cleanups for those variables can be
2071 optimized away. The semantic effect is that an exception thrown out of
2072 a function with such an exception specification will result in a call
2073 to @code{terminate} rather than @code{unexpected}.
2075 @item -fno-operator-names
2076 @opindex fno-operator-names
2077 Do not treat the operator name keywords @code{and}, @code{bitand},
2078 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2079 synonyms as keywords.
2081 @item -fno-optional-diags
2082 @opindex fno-optional-diags
2083 Disable diagnostics that the standard says a compiler does not need to
2084 issue. Currently, the only such diagnostic issued by G++ is the one for
2085 a name having multiple meanings within a class.
2088 @opindex fpermissive
2089 Downgrade some diagnostics about nonconformant code from errors to
2090 warnings. Thus, using @option{-fpermissive} will allow some
2091 nonconforming code to compile.
2093 @item -fno-pretty-templates
2094 @opindex fno-pretty-templates
2095 When an error message refers to a specialization of a function
2096 template, the compiler will normally print the signature of the
2097 template followed by the template arguments and any typedefs or
2098 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2099 rather than @code{void f(int)}) so that it's clear which template is
2100 involved. When an error message refers to a specialization of a class
2101 template, the compiler will omit any template arguments that match
2102 the default template arguments for that template. If either of these
2103 behaviors make it harder to understand the error message rather than
2104 easier, using @option{-fno-pretty-templates} will disable them.
2108 Enable automatic template instantiation at link time. This option also
2109 implies @option{-fno-implicit-templates}. @xref{Template
2110 Instantiation}, for more information.
2114 Disable generation of information about every class with virtual
2115 functions for use by the C++ run-time type identification features
2116 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2117 of the language, you can save some space by using this flag. Note that
2118 exception handling uses the same information, but it will generate it as
2119 needed. The @samp{dynamic_cast} operator can still be used for casts that
2120 do not require run-time type information, i.e.@: casts to @code{void *} or to
2121 unambiguous base classes.
2125 Emit statistics about front-end processing at the end of the compilation.
2126 This information is generally only useful to the G++ development team.
2128 @item -fstrict-enums
2129 @opindex fstrict-enums
2130 Allow the compiler to optimize using the assumption that a value of
2131 enumeration type can only be one of the values of the enumeration (as
2132 defined in the C++ standard; basically, a value that can be
2133 represented in the minimum number of bits needed to represent all the
2134 enumerators). This assumption may not be valid if the program uses a
2135 cast to convert an arbitrary integer value to the enumeration type.
2137 @item -ftemplate-depth=@var{n}
2138 @opindex ftemplate-depth
2139 Set the maximum instantiation depth for template classes to @var{n}.
2140 A limit on the template instantiation depth is needed to detect
2141 endless recursions during template class instantiation. ANSI/ISO C++
2142 conforming programs must not rely on a maximum depth greater than 17
2143 (changed to 1024 in C++11). The default value is 900, as the compiler
2144 can run out of stack space before hitting 1024 in some situations.
2146 @item -fno-threadsafe-statics
2147 @opindex fno-threadsafe-statics
2148 Do not emit the extra code to use the routines specified in the C++
2149 ABI for thread-safe initialization of local statics. You can use this
2150 option to reduce code size slightly in code that doesn't need to be
2153 @item -fuse-cxa-atexit
2154 @opindex fuse-cxa-atexit
2155 Register destructors for objects with static storage duration with the
2156 @code{__cxa_atexit} function rather than the @code{atexit} function.
2157 This option is required for fully standards-compliant handling of static
2158 destructors, but will only work if your C library supports
2159 @code{__cxa_atexit}.
2161 @item -fno-use-cxa-get-exception-ptr
2162 @opindex fno-use-cxa-get-exception-ptr
2163 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2164 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2165 if the runtime routine is not available.
2167 @item -fvisibility-inlines-hidden
2168 @opindex fvisibility-inlines-hidden
2169 This switch declares that the user does not attempt to compare
2170 pointers to inline functions or methods where the addresses of the two functions
2171 were taken in different shared objects.
2173 The effect of this is that GCC may, effectively, mark inline methods with
2174 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2175 appear in the export table of a DSO and do not require a PLT indirection
2176 when used within the DSO@. Enabling this option can have a dramatic effect
2177 on load and link times of a DSO as it massively reduces the size of the
2178 dynamic export table when the library makes heavy use of templates.
2180 The behavior of this switch is not quite the same as marking the
2181 methods as hidden directly, because it does not affect static variables
2182 local to the function or cause the compiler to deduce that
2183 the function is defined in only one shared object.
2185 You may mark a method as having a visibility explicitly to negate the
2186 effect of the switch for that method. For example, if you do want to
2187 compare pointers to a particular inline method, you might mark it as
2188 having default visibility. Marking the enclosing class with explicit
2189 visibility will have no effect.
2191 Explicitly instantiated inline methods are unaffected by this option
2192 as their linkage might otherwise cross a shared library boundary.
2193 @xref{Template Instantiation}.
2195 @item -fvisibility-ms-compat
2196 @opindex fvisibility-ms-compat
2197 This flag attempts to use visibility settings to make GCC's C++
2198 linkage model compatible with that of Microsoft Visual Studio.
2200 The flag makes these changes to GCC's linkage model:
2204 It sets the default visibility to @code{hidden}, like
2205 @option{-fvisibility=hidden}.
2208 Types, but not their members, are not hidden by default.
2211 The One Definition Rule is relaxed for types without explicit
2212 visibility specifications that are defined in more than one different
2213 shared object: those declarations are permitted if they would have
2214 been permitted when this option was not used.
2217 In new code it is better to use @option{-fvisibility=hidden} and
2218 export those classes that are intended to be externally visible.
2219 Unfortunately it is possible for code to rely, perhaps accidentally,
2220 on the Visual Studio behavior.
2222 Among the consequences of these changes are that static data members
2223 of the same type with the same name but defined in different shared
2224 objects will be different, so changing one will not change the other;
2225 and that pointers to function members defined in different shared
2226 objects may not compare equal. When this flag is given, it is a
2227 violation of the ODR to define types with the same name differently.
2231 Do not use weak symbol support, even if it is provided by the linker.
2232 By default, G++ will use weak symbols if they are available. This
2233 option exists only for testing, and should not be used by end-users;
2234 it will result in inferior code and has no benefits. This option may
2235 be removed in a future release of G++.
2239 Do not search for header files in the standard directories specific to
2240 C++, but do still search the other standard directories. (This option
2241 is used when building the C++ library.)
2244 In addition, these optimization, warning, and code generation options
2245 have meanings only for C++ programs:
2248 @item -fno-default-inline
2249 @opindex fno-default-inline
2250 Do not assume @samp{inline} for functions defined inside a class scope.
2251 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2252 functions will have linkage like inline functions; they just won't be
2255 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2258 Warn when G++ generates code that is probably not compatible with the
2259 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2260 all such cases, there are probably some cases that are not warned about,
2261 even though G++ is generating incompatible code. There may also be
2262 cases where warnings are emitted even though the code that is generated
2265 You should rewrite your code to avoid these warnings if you are
2266 concerned about the fact that code generated by G++ may not be binary
2267 compatible with code generated by other compilers.
2269 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2274 A template with a non-type template parameter of reference type is
2275 mangled incorrectly:
2278 template <int &> struct S @{@};
2282 This is fixed in @option{-fabi-version=3}.
2285 SIMD vector types declared using @code{__attribute ((vector_size))} are
2286 mangled in a non-standard way that does not allow for overloading of
2287 functions taking vectors of different sizes.
2289 The mangling is changed in @option{-fabi-version=4}.
2292 The known incompatibilities in @option{-fabi-version=1} include:
2297 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2298 pack data into the same byte as a base class. For example:
2301 struct A @{ virtual void f(); int f1 : 1; @};
2302 struct B : public A @{ int f2 : 1; @};
2306 In this case, G++ will place @code{B::f2} into the same byte
2307 as@code{A::f1}; other compilers will not. You can avoid this problem
2308 by explicitly padding @code{A} so that its size is a multiple of the
2309 byte size on your platform; that will cause G++ and other compilers to
2310 layout @code{B} identically.
2313 Incorrect handling of tail-padding for virtual bases. G++ does not use
2314 tail padding when laying out virtual bases. For example:
2317 struct A @{ virtual void f(); char c1; @};
2318 struct B @{ B(); char c2; @};
2319 struct C : public A, public virtual B @{@};
2323 In this case, G++ will not place @code{B} into the tail-padding for
2324 @code{A}; other compilers will. You can avoid this problem by
2325 explicitly padding @code{A} so that its size is a multiple of its
2326 alignment (ignoring virtual base classes); that will cause G++ and other
2327 compilers to layout @code{C} identically.
2330 Incorrect handling of bit-fields with declared widths greater than that
2331 of their underlying types, when the bit-fields appear in a union. For
2335 union U @{ int i : 4096; @};
2339 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2340 union too small by the number of bits in an @code{int}.
2343 Empty classes can be placed at incorrect offsets. For example:
2353 struct C : public B, public A @{@};
2357 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2358 it should be placed at offset zero. G++ mistakenly believes that the
2359 @code{A} data member of @code{B} is already at offset zero.
2362 Names of template functions whose types involve @code{typename} or
2363 template template parameters can be mangled incorrectly.
2366 template <typename Q>
2367 void f(typename Q::X) @{@}
2369 template <template <typename> class Q>
2370 void f(typename Q<int>::X) @{@}
2374 Instantiations of these templates may be mangled incorrectly.
2378 It also warns psABI related changes. The known psABI changes at this
2384 For SYSV/x86-64, when passing union with long double, it is changed to
2385 pass in memory as specified in psABI. For example:
2395 @code{union U} will always be passed in memory.
2399 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2400 @opindex Wctor-dtor-privacy
2401 @opindex Wno-ctor-dtor-privacy
2402 Warn when a class seems unusable because all the constructors or
2403 destructors in that class are private, and it has neither friends nor
2404 public static member functions.
2406 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2407 @opindex Wdelete-non-virtual-dtor
2408 @opindex Wno-delete-non-virtual-dtor
2409 Warn when @samp{delete} is used to destroy an instance of a class that
2410 has virtual functions and non-virtual destructor. It is unsafe to delete
2411 an instance of a derived class through a pointer to a base class if the
2412 base class does not have a virtual destructor. This warning is enabled
2415 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2417 @opindex Wno-narrowing
2418 Warn when a narrowing conversion prohibited by C++11 occurs within
2422 int i = @{ 2.2 @}; // error: narrowing from double to int
2425 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2427 With -std=c++11, @option{-Wno-narrowing} suppresses the diagnostic
2428 required by the standard. Note that this does not affect the meaning
2429 of well-formed code; narrowing conversions are still considered
2430 ill-formed in SFINAE context.
2432 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2434 @opindex Wno-noexcept
2435 Warn when a noexcept-expression evaluates to false because of a call
2436 to a function that does not have a non-throwing exception
2437 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2438 the compiler to never throw an exception.
2440 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2441 @opindex Wnon-virtual-dtor
2442 @opindex Wno-non-virtual-dtor
2443 Warn when a class has virtual functions and accessible non-virtual
2444 destructor, in which case it would be possible but unsafe to delete
2445 an instance of a derived class through a pointer to the base class.
2446 This warning is also enabled if @option{-Weffc++} is specified.
2448 @item -Wreorder @r{(C++ and Objective-C++ only)}
2450 @opindex Wno-reorder
2451 @cindex reordering, warning
2452 @cindex warning for reordering of member initializers
2453 Warn when the order of member initializers given in the code does not
2454 match the order in which they must be executed. For instance:
2460 A(): j (0), i (1) @{ @}
2464 The compiler will rearrange the member initializers for @samp{i}
2465 and @samp{j} to match the declaration order of the members, emitting
2466 a warning to that effect. This warning is enabled by @option{-Wall}.
2469 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2472 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2475 Warn about violations of the following style guidelines from Scott Meyers'
2476 @cite{Effective C++, Second Edition} book:
2480 Item 11: Define a copy constructor and an assignment operator for classes
2481 with dynamically allocated memory.
2484 Item 12: Prefer initialization to assignment in constructors.
2487 Item 14: Make destructors virtual in base classes.
2490 Item 15: Have @code{operator=} return a reference to @code{*this}.
2493 Item 23: Don't try to return a reference when you must return an object.
2497 Also warn about violations of the following style guidelines from
2498 Scott Meyers' @cite{More Effective C++} book:
2502 Item 6: Distinguish between prefix and postfix forms of increment and
2503 decrement operators.
2506 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2510 When selecting this option, be aware that the standard library
2511 headers do not obey all of these guidelines; use @samp{grep -v}
2512 to filter out those warnings.
2514 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2515 @opindex Wstrict-null-sentinel
2516 @opindex Wno-strict-null-sentinel
2517 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2518 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2519 to @code{__null}. Although it is a null pointer constant not a null pointer,
2520 it is guaranteed to be of the same size as a pointer. But this use is
2521 not portable across different compilers.
2523 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2524 @opindex Wno-non-template-friend
2525 @opindex Wnon-template-friend
2526 Disable warnings when non-templatized friend functions are declared
2527 within a template. Since the advent of explicit template specification
2528 support in G++, if the name of the friend is an unqualified-id (i.e.,
2529 @samp{friend foo(int)}), the C++ language specification demands that the
2530 friend declare or define an ordinary, nontemplate function. (Section
2531 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2532 could be interpreted as a particular specialization of a templatized
2533 function. Because this non-conforming behavior is no longer the default
2534 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2535 check existing code for potential trouble spots and is on by default.
2536 This new compiler behavior can be turned off with
2537 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2538 but disables the helpful warning.
2540 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2541 @opindex Wold-style-cast
2542 @opindex Wno-old-style-cast
2543 Warn if an old-style (C-style) cast to a non-void type is used within
2544 a C++ program. The new-style casts (@samp{dynamic_cast},
2545 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2546 less vulnerable to unintended effects and much easier to search for.
2548 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2549 @opindex Woverloaded-virtual
2550 @opindex Wno-overloaded-virtual
2551 @cindex overloaded virtual function, warning
2552 @cindex warning for overloaded virtual function
2553 Warn when a function declaration hides virtual functions from a
2554 base class. For example, in:
2561 struct B: public A @{
2566 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2574 will fail to compile.
2576 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2577 @opindex Wno-pmf-conversions
2578 @opindex Wpmf-conversions
2579 Disable the diagnostic for converting a bound pointer to member function
2582 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2583 @opindex Wsign-promo
2584 @opindex Wno-sign-promo
2585 Warn when overload resolution chooses a promotion from unsigned or
2586 enumerated type to a signed type, over a conversion to an unsigned type of
2587 the same size. Previous versions of G++ would try to preserve
2588 unsignedness, but the standard mandates the current behavior.
2593 A& operator = (int);
2603 In this example, G++ will synthesize a default @samp{A& operator =
2604 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2607 @node Objective-C and Objective-C++ Dialect Options
2608 @section Options Controlling Objective-C and Objective-C++ Dialects
2610 @cindex compiler options, Objective-C and Objective-C++
2611 @cindex Objective-C and Objective-C++ options, command-line
2612 @cindex options, Objective-C and Objective-C++
2613 (NOTE: This manual does not describe the Objective-C and Objective-C++
2614 languages themselves. @xref{Standards,,Language Standards
2615 Supported by GCC}, for references.)
2617 This section describes the command-line options that are only meaningful
2618 for Objective-C and Objective-C++ programs, but you can also use most of
2619 the language-independent GNU compiler options.
2620 For example, you might compile a file @code{some_class.m} like this:
2623 gcc -g -fgnu-runtime -O -c some_class.m
2627 In this example, @option{-fgnu-runtime} is an option meant only for
2628 Objective-C and Objective-C++ programs; you can use the other options with
2629 any language supported by GCC@.
2631 Note that since Objective-C is an extension of the C language, Objective-C
2632 compilations may also use options specific to the C front-end (e.g.,
2633 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2634 C++-specific options (e.g., @option{-Wabi}).
2636 Here is a list of options that are @emph{only} for compiling Objective-C
2637 and Objective-C++ programs:
2640 @item -fconstant-string-class=@var{class-name}
2641 @opindex fconstant-string-class
2642 Use @var{class-name} as the name of the class to instantiate for each
2643 literal string specified with the syntax @code{@@"@dots{}"}. The default
2644 class name is @code{NXConstantString} if the GNU runtime is being used, and
2645 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2646 @option{-fconstant-cfstrings} option, if also present, will override the
2647 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2648 to be laid out as constant CoreFoundation strings.
2651 @opindex fgnu-runtime
2652 Generate object code compatible with the standard GNU Objective-C
2653 runtime. This is the default for most types of systems.
2655 @item -fnext-runtime
2656 @opindex fnext-runtime
2657 Generate output compatible with the NeXT runtime. This is the default
2658 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2659 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2662 @item -fno-nil-receivers
2663 @opindex fno-nil-receivers
2664 Assume that all Objective-C message dispatches (@code{[receiver
2665 message:arg]}) in this translation unit ensure that the receiver is
2666 not @code{nil}. This allows for more efficient entry points in the
2667 runtime to be used. This option is only available in conjunction with
2668 the NeXT runtime and ABI version 0 or 1.
2670 @item -fobjc-abi-version=@var{n}
2671 @opindex fobjc-abi-version
2672 Use version @var{n} of the Objective-C ABI for the selected runtime.
2673 This option is currently supported only for the NeXT runtime. In that
2674 case, Version 0 is the traditional (32-bit) ABI without support for
2675 properties and other Objective-C 2.0 additions. Version 1 is the
2676 traditional (32-bit) ABI with support for properties and other
2677 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2678 nothing is specified, the default is Version 0 on 32-bit target
2679 machines, and Version 2 on 64-bit target machines.
2681 @item -fobjc-call-cxx-cdtors
2682 @opindex fobjc-call-cxx-cdtors
2683 For each Objective-C class, check if any of its instance variables is a
2684 C++ object with a non-trivial default constructor. If so, synthesize a
2685 special @code{- (id) .cxx_construct} instance method which will run
2686 non-trivial default constructors on any such instance variables, in order,
2687 and then return @code{self}. Similarly, check if any instance variable
2688 is a C++ object with a non-trivial destructor, and if so, synthesize a
2689 special @code{- (void) .cxx_destruct} method which will run
2690 all such default destructors, in reverse order.
2692 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2693 methods thusly generated will only operate on instance variables
2694 declared in the current Objective-C class, and not those inherited
2695 from superclasses. It is the responsibility of the Objective-C
2696 runtime to invoke all such methods in an object's inheritance
2697 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2698 by the runtime immediately after a new object instance is allocated;
2699 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2700 before the runtime deallocates an object instance.
2702 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2703 support for invoking the @code{- (id) .cxx_construct} and
2704 @code{- (void) .cxx_destruct} methods.
2706 @item -fobjc-direct-dispatch
2707 @opindex fobjc-direct-dispatch
2708 Allow fast jumps to the message dispatcher. On Darwin this is
2709 accomplished via the comm page.
2711 @item -fobjc-exceptions
2712 @opindex fobjc-exceptions
2713 Enable syntactic support for structured exception handling in
2714 Objective-C, similar to what is offered by C++ and Java. This option
2715 is required to use the Objective-C keywords @code{@@try},
2716 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2717 @code{@@synchronized}. This option is available with both the GNU
2718 runtime and the NeXT runtime (but not available in conjunction with
2719 the NeXT runtime on Mac OS X 10.2 and earlier).
2723 Enable garbage collection (GC) in Objective-C and Objective-C++
2724 programs. This option is only available with the NeXT runtime; the
2725 GNU runtime has a different garbage collection implementation that
2726 does not require special compiler flags.
2728 @item -fobjc-nilcheck
2729 @opindex fobjc-nilcheck
2730 For the NeXT runtime with version 2 of the ABI, check for a nil
2731 receiver in method invocations before doing the actual method call.
2732 This is the default and can be disabled using
2733 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2734 checked for nil in this way no matter what this flag is set to.
2735 Currently this flag does nothing when the GNU runtime, or an older
2736 version of the NeXT runtime ABI, is used.
2738 @item -fobjc-std=objc1
2740 Conform to the language syntax of Objective-C 1.0, the language
2741 recognized by GCC 4.0. This only affects the Objective-C additions to
2742 the C/C++ language; it does not affect conformance to C/C++ standards,
2743 which is controlled by the separate C/C++ dialect option flags. When
2744 this option is used with the Objective-C or Objective-C++ compiler,
2745 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2746 This is useful if you need to make sure that your Objective-C code can
2747 be compiled with older versions of GCC.
2749 @item -freplace-objc-classes
2750 @opindex freplace-objc-classes
2751 Emit a special marker instructing @command{ld(1)} not to statically link in
2752 the resulting object file, and allow @command{dyld(1)} to load it in at
2753 run time instead. This is used in conjunction with the Fix-and-Continue
2754 debugging mode, where the object file in question may be recompiled and
2755 dynamically reloaded in the course of program execution, without the need
2756 to restart the program itself. Currently, Fix-and-Continue functionality
2757 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2762 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2763 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2764 compile time) with static class references that get initialized at load time,
2765 which improves run-time performance. Specifying the @option{-fzero-link} flag
2766 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2767 to be retained. This is useful in Zero-Link debugging mode, since it allows
2768 for individual class implementations to be modified during program execution.
2769 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2770 regardless of command-line options.
2774 Dump interface declarations for all classes seen in the source file to a
2775 file named @file{@var{sourcename}.decl}.
2777 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2778 @opindex Wassign-intercept
2779 @opindex Wno-assign-intercept
2780 Warn whenever an Objective-C assignment is being intercepted by the
2783 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2784 @opindex Wno-protocol
2786 If a class is declared to implement a protocol, a warning is issued for
2787 every method in the protocol that is not implemented by the class. The
2788 default behavior is to issue a warning for every method not explicitly
2789 implemented in the class, even if a method implementation is inherited
2790 from the superclass. If you use the @option{-Wno-protocol} option, then
2791 methods inherited from the superclass are considered to be implemented,
2792 and no warning is issued for them.
2794 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2796 @opindex Wno-selector
2797 Warn if multiple methods of different types for the same selector are
2798 found during compilation. The check is performed on the list of methods
2799 in the final stage of compilation. Additionally, a check is performed
2800 for each selector appearing in a @code{@@selector(@dots{})}
2801 expression, and a corresponding method for that selector has been found
2802 during compilation. Because these checks scan the method table only at
2803 the end of compilation, these warnings are not produced if the final
2804 stage of compilation is not reached, for example because an error is
2805 found during compilation, or because the @option{-fsyntax-only} option is
2808 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2809 @opindex Wstrict-selector-match
2810 @opindex Wno-strict-selector-match
2811 Warn if multiple methods with differing argument and/or return types are
2812 found for a given selector when attempting to send a message using this
2813 selector to a receiver of type @code{id} or @code{Class}. When this flag
2814 is off (which is the default behavior), the compiler will omit such warnings
2815 if any differences found are confined to types that share the same size
2818 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2819 @opindex Wundeclared-selector
2820 @opindex Wno-undeclared-selector
2821 Warn if a @code{@@selector(@dots{})} expression referring to an
2822 undeclared selector is found. A selector is considered undeclared if no
2823 method with that name has been declared before the
2824 @code{@@selector(@dots{})} expression, either explicitly in an
2825 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2826 an @code{@@implementation} section. This option always performs its
2827 checks as soon as a @code{@@selector(@dots{})} expression is found,
2828 while @option{-Wselector} only performs its checks in the final stage of
2829 compilation. This also enforces the coding style convention
2830 that methods and selectors must be declared before being used.
2832 @item -print-objc-runtime-info
2833 @opindex print-objc-runtime-info
2834 Generate C header describing the largest structure that is passed by
2839 @node Language Independent Options
2840 @section Options to Control Diagnostic Messages Formatting
2841 @cindex options to control diagnostics formatting
2842 @cindex diagnostic messages
2843 @cindex message formatting
2845 Traditionally, diagnostic messages have been formatted irrespective of
2846 the output device's aspect (e.g.@: its width, @dots{}). The options described
2847 below can be used to control the diagnostic messages formatting
2848 algorithm, e.g.@: how many characters per line, how often source location
2849 information should be reported. Right now, only the C++ front end can
2850 honor these options. However it is expected, in the near future, that
2851 the remaining front ends would be able to digest them correctly.
2854 @item -fmessage-length=@var{n}
2855 @opindex fmessage-length
2856 Try to format error messages so that they fit on lines of about @var{n}
2857 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2858 the front ends supported by GCC@. If @var{n} is zero, then no
2859 line-wrapping will be done; each error message will appear on a single
2862 @opindex fdiagnostics-show-location
2863 @item -fdiagnostics-show-location=once
2864 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2865 reporter to emit @emph{once} source location information; that is, in
2866 case the message is too long to fit on a single physical line and has to
2867 be wrapped, the source location won't be emitted (as prefix) again,
2868 over and over, in subsequent continuation lines. This is the default
2871 @item -fdiagnostics-show-location=every-line
2872 Only meaningful in line-wrapping mode. Instructs the diagnostic
2873 messages reporter to emit the same source location information (as
2874 prefix) for physical lines that result from the process of breaking
2875 a message which is too long to fit on a single line.
2877 @item -fno-diagnostics-show-option
2878 @opindex fno-diagnostics-show-option
2879 @opindex fdiagnostics-show-option
2880 By default, each diagnostic emitted includes text indicating the
2881 command-line option that directly controls the diagnostic (if such an
2882 option is known to the diagnostic machinery). Specifying the
2883 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2887 @node Warning Options
2888 @section Options to Request or Suppress Warnings
2889 @cindex options to control warnings
2890 @cindex warning messages
2891 @cindex messages, warning
2892 @cindex suppressing warnings
2894 Warnings are diagnostic messages that report constructions that
2895 are not inherently erroneous but that are risky or suggest there
2896 may have been an error.
2898 The following language-independent options do not enable specific
2899 warnings but control the kinds of diagnostics produced by GCC.
2902 @cindex syntax checking
2904 @opindex fsyntax-only
2905 Check the code for syntax errors, but don't do anything beyond that.
2907 @item -fmax-errors=@var{n}
2908 @opindex fmax-errors
2909 Limits the maximum number of error messages to @var{n}, at which point
2910 GCC bails out rather than attempting to continue processing the source
2911 code. If @var{n} is 0 (the default), there is no limit on the number
2912 of error messages produced. If @option{-Wfatal-errors} is also
2913 specified, then @option{-Wfatal-errors} takes precedence over this
2918 Inhibit all warning messages.
2923 Make all warnings into errors.
2928 Make the specified warning into an error. The specifier for a warning
2929 is appended, for example @option{-Werror=switch} turns the warnings
2930 controlled by @option{-Wswitch} into errors. This switch takes a
2931 negative form, to be used to negate @option{-Werror} for specific
2932 warnings, for example @option{-Wno-error=switch} makes
2933 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2936 The warning message for each controllable warning includes the
2937 option that controls the warning. That option can then be used with
2938 @option{-Werror=} and @option{-Wno-error=} as described above.
2939 (Printing of the option in the warning message can be disabled using the
2940 @option{-fno-diagnostics-show-option} flag.)
2942 Note that specifying @option{-Werror=}@var{foo} automatically implies
2943 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2946 @item -Wfatal-errors
2947 @opindex Wfatal-errors
2948 @opindex Wno-fatal-errors
2949 This option causes the compiler to abort compilation on the first error
2950 occurred rather than trying to keep going and printing further error
2955 You can request many specific warnings with options beginning
2956 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2957 implicit declarations. Each of these specific warning options also
2958 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2959 example, @option{-Wno-implicit}. This manual lists only one of the
2960 two forms, whichever is not the default. For further,
2961 language-specific options also refer to @ref{C++ Dialect Options} and
2962 @ref{Objective-C and Objective-C++ Dialect Options}.
2964 When an unrecognized warning option is requested (e.g.,
2965 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2966 that the option is not recognized. However, if the @option{-Wno-} form
2967 is used, the behavior is slightly different: No diagnostic will be
2968 produced for @option{-Wno-unknown-warning} unless other diagnostics
2969 are being produced. This allows the use of new @option{-Wno-} options
2970 with old compilers, but if something goes wrong, the compiler will
2971 warn that an unrecognized option was used.
2976 Issue all the warnings demanded by strict ISO C and ISO C++;
2977 reject all programs that use forbidden extensions, and some other
2978 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2979 version of the ISO C standard specified by any @option{-std} option used.
2981 Valid ISO C and ISO C++ programs should compile properly with or without
2982 this option (though a rare few will require @option{-ansi} or a
2983 @option{-std} option specifying the required version of ISO C)@. However,
2984 without this option, certain GNU extensions and traditional C and C++
2985 features are supported as well. With this option, they are rejected.
2987 @option{-pedantic} does not cause warning messages for use of the
2988 alternate keywords whose names begin and end with @samp{__}. Pedantic
2989 warnings are also disabled in the expression that follows
2990 @code{__extension__}. However, only system header files should use
2991 these escape routes; application programs should avoid them.
2992 @xref{Alternate Keywords}.
2994 Some users try to use @option{-pedantic} to check programs for strict ISO
2995 C conformance. They soon find that it does not do quite what they want:
2996 it finds some non-ISO practices, but not all---only those for which
2997 ISO C @emph{requires} a diagnostic, and some others for which
2998 diagnostics have been added.
3000 A feature to report any failure to conform to ISO C might be useful in
3001 some instances, but would require considerable additional work and would
3002 be quite different from @option{-pedantic}. We don't have plans to
3003 support such a feature in the near future.
3005 Where the standard specified with @option{-std} represents a GNU
3006 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3007 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3008 extended dialect is based. Warnings from @option{-pedantic} are given
3009 where they are required by the base standard. (It would not make sense
3010 for such warnings to be given only for features not in the specified GNU
3011 C dialect, since by definition the GNU dialects of C include all
3012 features the compiler supports with the given option, and there would be
3013 nothing to warn about.)
3015 @item -pedantic-errors
3016 @opindex pedantic-errors
3017 Like @option{-pedantic}, except that errors are produced rather than
3023 This enables all the warnings about constructions that some users
3024 consider questionable, and that are easy to avoid (or modify to
3025 prevent the warning), even in conjunction with macros. This also
3026 enables some language-specific warnings described in @ref{C++ Dialect
3027 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3029 @option{-Wall} turns on the following warning flags:
3031 @gccoptlist{-Waddress @gol
3032 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3034 -Wchar-subscripts @gol
3035 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
3036 -Wimplicit-int @r{(C and Objective-C only)} @gol
3037 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3040 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3041 -Wmaybe-uninitialized @gol
3042 -Wmissing-braces @gol
3048 -Wsequence-point @gol
3049 -Wsign-compare @r{(only in C++)} @gol
3050 -Wstrict-aliasing @gol
3051 -Wstrict-overflow=1 @gol
3054 -Wuninitialized @gol
3055 -Wunknown-pragmas @gol
3056 -Wunused-function @gol
3059 -Wunused-variable @gol
3060 -Wvolatile-register-var @gol
3063 Note that some warning flags are not implied by @option{-Wall}. Some of
3064 them warn about constructions that users generally do not consider
3065 questionable, but which occasionally you might wish to check for;
3066 others warn about constructions that are necessary or hard to avoid in
3067 some cases, and there is no simple way to modify the code to suppress
3068 the warning. Some of them are enabled by @option{-Wextra} but many of
3069 them must be enabled individually.
3075 This enables some extra warning flags that are not enabled by
3076 @option{-Wall}. (This option used to be called @option{-W}. The older
3077 name is still supported, but the newer name is more descriptive.)
3079 @gccoptlist{-Wclobbered @gol
3081 -Wignored-qualifiers @gol
3082 -Wmissing-field-initializers @gol
3083 -Wmissing-parameter-type @r{(C only)} @gol
3084 -Wold-style-declaration @r{(C only)} @gol
3085 -Woverride-init @gol
3088 -Wuninitialized @gol
3089 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3090 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3093 The option @option{-Wextra} also prints warning messages for the
3099 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3100 @samp{>}, or @samp{>=}.
3103 (C++ only) An enumerator and a non-enumerator both appear in a
3104 conditional expression.
3107 (C++ only) Ambiguous virtual bases.
3110 (C++ only) Subscripting an array that has been declared @samp{register}.
3113 (C++ only) Taking the address of a variable that has been declared
3117 (C++ only) A base class is not initialized in a derived class' copy
3122 @item -Wchar-subscripts
3123 @opindex Wchar-subscripts
3124 @opindex Wno-char-subscripts
3125 Warn if an array subscript has type @code{char}. This is a common cause
3126 of error, as programmers often forget that this type is signed on some
3128 This warning is enabled by @option{-Wall}.
3132 @opindex Wno-comment
3133 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3134 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3135 This warning is enabled by @option{-Wall}.
3137 @item -Wno-coverage-mismatch
3138 @opindex Wno-coverage-mismatch
3139 Warn if feedback profiles do not match when using the
3140 @option{-fprofile-use} option.
3141 If a source file was changed between @option{-fprofile-gen} and
3142 @option{-fprofile-use}, the files with the profile feedback can fail
3143 to match the source file and GCC cannot use the profile feedback
3144 information. By default, this warning is enabled and is treated as an
3145 error. @option{-Wno-coverage-mismatch} can be used to disable the
3146 warning or @option{-Wno-error=coverage-mismatch} can be used to
3147 disable the error. Disabling the error for this warning can result in
3148 poorly optimized code and is useful only in the
3149 case of very minor changes such as bug fixes to an existing code-base.
3150 Completely disabling the warning is not recommended.
3153 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3155 Suppress warning messages emitted by @code{#warning} directives.
3157 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3158 @opindex Wdouble-promotion
3159 @opindex Wno-double-promotion
3160 Give a warning when a value of type @code{float} is implicitly
3161 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3162 floating-point unit implement @code{float} in hardware, but emulate
3163 @code{double} in software. On such a machine, doing computations
3164 using @code{double} values is much more expensive because of the
3165 overhead required for software emulation.
3167 It is easy to accidentally do computations with @code{double} because
3168 floating-point literals are implicitly of type @code{double}. For
3172 float area(float radius)
3174 return 3.14159 * radius * radius;
3178 the compiler will perform the entire computation with @code{double}
3179 because the floating-point literal is a @code{double}.
3184 @opindex ffreestanding
3185 @opindex fno-builtin
3186 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3187 the arguments supplied have types appropriate to the format string
3188 specified, and that the conversions specified in the format string make
3189 sense. This includes standard functions, and others specified by format
3190 attributes (@pxref{Function Attributes}), in the @code{printf},
3191 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3192 not in the C standard) families (or other target-specific families).
3193 Which functions are checked without format attributes having been
3194 specified depends on the standard version selected, and such checks of
3195 functions without the attribute specified are disabled by
3196 @option{-ffreestanding} or @option{-fno-builtin}.
3198 The formats are checked against the format features supported by GNU
3199 libc version 2.2. These include all ISO C90 and C99 features, as well
3200 as features from the Single Unix Specification and some BSD and GNU
3201 extensions. Other library implementations may not support all these
3202 features; GCC does not support warning about features that go beyond a
3203 particular library's limitations. However, if @option{-pedantic} is used
3204 with @option{-Wformat}, warnings will be given about format features not
3205 in the selected standard version (but not for @code{strfmon} formats,
3206 since those are not in any version of the C standard). @xref{C Dialect
3207 Options,,Options Controlling C Dialect}.
3209 Since @option{-Wformat} also checks for null format arguments for
3210 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3212 @option{-Wformat} is included in @option{-Wall}. For more control over some
3213 aspects of format checking, the options @option{-Wformat-y2k},
3214 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3215 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3216 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3219 @opindex Wformat-y2k
3220 @opindex Wno-format-y2k
3221 If @option{-Wformat} is specified, also warn about @code{strftime}
3222 formats that may yield only a two-digit year.
3224 @item -Wno-format-contains-nul
3225 @opindex Wno-format-contains-nul
3226 @opindex Wformat-contains-nul
3227 If @option{-Wformat} is specified, do not warn about format strings that
3230 @item -Wno-format-extra-args
3231 @opindex Wno-format-extra-args
3232 @opindex Wformat-extra-args
3233 If @option{-Wformat} is specified, do not warn about excess arguments to a
3234 @code{printf} or @code{scanf} format function. The C standard specifies
3235 that such arguments are ignored.
3237 Where the unused arguments lie between used arguments that are
3238 specified with @samp{$} operand number specifications, normally
3239 warnings are still given, since the implementation could not know what
3240 type to pass to @code{va_arg} to skip the unused arguments. However,
3241 in the case of @code{scanf} formats, this option will suppress the
3242 warning if the unused arguments are all pointers, since the Single
3243 Unix Specification says that such unused arguments are allowed.
3245 @item -Wno-format-zero-length
3246 @opindex Wno-format-zero-length
3247 @opindex Wformat-zero-length
3248 If @option{-Wformat} is specified, do not warn about zero-length formats.
3249 The C standard specifies that zero-length formats are allowed.
3251 @item -Wformat-nonliteral
3252 @opindex Wformat-nonliteral
3253 @opindex Wno-format-nonliteral
3254 If @option{-Wformat} is specified, also warn if the format string is not a
3255 string literal and so cannot be checked, unless the format function
3256 takes its format arguments as a @code{va_list}.
3258 @item -Wformat-security
3259 @opindex Wformat-security
3260 @opindex Wno-format-security
3261 If @option{-Wformat} is specified, also warn about uses of format
3262 functions that represent possible security problems. At present, this
3263 warns about calls to @code{printf} and @code{scanf} functions where the
3264 format string is not a string literal and there are no format arguments,
3265 as in @code{printf (foo);}. This may be a security hole if the format
3266 string came from untrusted input and contains @samp{%n}. (This is
3267 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3268 in future warnings may be added to @option{-Wformat-security} that are not
3269 included in @option{-Wformat-nonliteral}.)
3273 @opindex Wno-format=2
3274 Enable @option{-Wformat} plus format checks not included in
3275 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3276 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3280 @opindex Wno-nonnull
3281 Warn about passing a null pointer for arguments marked as
3282 requiring a non-null value by the @code{nonnull} function attribute.
3284 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3285 can be disabled with the @option{-Wno-nonnull} option.
3287 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3289 @opindex Wno-init-self
3290 Warn about uninitialized variables that are initialized with themselves.
3291 Note this option can only be used with the @option{-Wuninitialized} option.
3293 For example, GCC will warn about @code{i} being uninitialized in the
3294 following snippet only when @option{-Winit-self} has been specified:
3305 @item -Wimplicit-int @r{(C and Objective-C only)}
3306 @opindex Wimplicit-int
3307 @opindex Wno-implicit-int
3308 Warn when a declaration does not specify a type.
3309 This warning is enabled by @option{-Wall}.
3311 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3312 @opindex Wimplicit-function-declaration
3313 @opindex Wno-implicit-function-declaration
3314 Give a warning whenever a function is used before being declared. In
3315 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3316 enabled by default and it is made into an error by
3317 @option{-pedantic-errors}. This warning is also enabled by
3320 @item -Wimplicit @r{(C and Objective-C only)}
3322 @opindex Wno-implicit
3323 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3324 This warning is enabled by @option{-Wall}.
3326 @item -Wignored-qualifiers @r{(C and C++ only)}
3327 @opindex Wignored-qualifiers
3328 @opindex Wno-ignored-qualifiers
3329 Warn if the return type of a function has a type qualifier
3330 such as @code{const}. For ISO C such a type qualifier has no effect,
3331 since the value returned by a function is not an lvalue.
3332 For C++, the warning is only emitted for scalar types or @code{void}.
3333 ISO C prohibits qualified @code{void} return types on function
3334 definitions, so such return types always receive a warning
3335 even without this option.
3337 This warning is also enabled by @option{-Wextra}.
3342 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3343 a function with external linkage, returning int, taking either zero
3344 arguments, two, or three arguments of appropriate types. This warning
3345 is enabled by default in C++ and is enabled by either @option{-Wall}
3346 or @option{-pedantic}.
3348 @item -Wmissing-braces
3349 @opindex Wmissing-braces
3350 @opindex Wno-missing-braces
3351 Warn if an aggregate or union initializer is not fully bracketed. In
3352 the following example, the initializer for @samp{a} is not fully
3353 bracketed, but that for @samp{b} is fully bracketed.
3356 int a[2][2] = @{ 0, 1, 2, 3 @};
3357 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3360 This warning is enabled by @option{-Wall}.
3362 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3363 @opindex Wmissing-include-dirs
3364 @opindex Wno-missing-include-dirs
3365 Warn if a user-supplied include directory does not exist.
3368 @opindex Wparentheses
3369 @opindex Wno-parentheses
3370 Warn if parentheses are omitted in certain contexts, such
3371 as when there is an assignment in a context where a truth value
3372 is expected, or when operators are nested whose precedence people
3373 often get confused about.
3375 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3376 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3377 interpretation from that of ordinary mathematical notation.
3379 Also warn about constructions where there may be confusion to which
3380 @code{if} statement an @code{else} branch belongs. Here is an example of
3395 In C/C++, every @code{else} branch belongs to the innermost possible
3396 @code{if} statement, which in this example is @code{if (b)}. This is
3397 often not what the programmer expected, as illustrated in the above
3398 example by indentation the programmer chose. When there is the
3399 potential for this confusion, GCC will issue a warning when this flag
3400 is specified. To eliminate the warning, add explicit braces around
3401 the innermost @code{if} statement so there is no way the @code{else}
3402 could belong to the enclosing @code{if}. The resulting code would
3419 Also warn for dangerous uses of the
3420 ?: with omitted middle operand GNU extension. When the condition
3421 in the ?: operator is a boolean expression the omitted value will
3422 be always 1. Often the user expects it to be a value computed
3423 inside the conditional expression instead.
3425 This warning is enabled by @option{-Wall}.
3427 @item -Wsequence-point
3428 @opindex Wsequence-point
3429 @opindex Wno-sequence-point
3430 Warn about code that may have undefined semantics because of violations
3431 of sequence point rules in the C and C++ standards.
3433 The C and C++ standards defines the order in which expressions in a C/C++
3434 program are evaluated in terms of @dfn{sequence points}, which represent
3435 a partial ordering between the execution of parts of the program: those
3436 executed before the sequence point, and those executed after it. These
3437 occur after the evaluation of a full expression (one which is not part
3438 of a larger expression), after the evaluation of the first operand of a
3439 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3440 function is called (but after the evaluation of its arguments and the
3441 expression denoting the called function), and in certain other places.
3442 Other than as expressed by the sequence point rules, the order of
3443 evaluation of subexpressions of an expression is not specified. All
3444 these rules describe only a partial order rather than a total order,
3445 since, for example, if two functions are called within one expression
3446 with no sequence point between them, the order in which the functions
3447 are called is not specified. However, the standards committee have
3448 ruled that function calls do not overlap.
3450 It is not specified when between sequence points modifications to the
3451 values of objects take effect. Programs whose behavior depends on this
3452 have undefined behavior; the C and C++ standards specify that ``Between
3453 the previous and next sequence point an object shall have its stored
3454 value modified at most once by the evaluation of an expression.
3455 Furthermore, the prior value shall be read only to determine the value
3456 to be stored.''. If a program breaks these rules, the results on any
3457 particular implementation are entirely unpredictable.
3459 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3460 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3461 diagnosed by this option, and it may give an occasional false positive
3462 result, but in general it has been found fairly effective at detecting
3463 this sort of problem in programs.
3465 The standard is worded confusingly, therefore there is some debate
3466 over the precise meaning of the sequence point rules in subtle cases.
3467 Links to discussions of the problem, including proposed formal
3468 definitions, may be found on the GCC readings page, at
3469 @uref{http://gcc.gnu.org/@/readings.html}.
3471 This warning is enabled by @option{-Wall} for C and C++.
3474 @opindex Wreturn-type
3475 @opindex Wno-return-type
3476 Warn whenever a function is defined with a return-type that defaults
3477 to @code{int}. Also warn about any @code{return} statement with no
3478 return-value in a function whose return-type is not @code{void}
3479 (falling off the end of the function body is considered returning
3480 without a value), and about a @code{return} statement with an
3481 expression in a function whose return-type is @code{void}.
3483 For C++, a function without return type always produces a diagnostic
3484 message, even when @option{-Wno-return-type} is specified. The only
3485 exceptions are @samp{main} and functions defined in system headers.
3487 This warning is enabled by @option{-Wall}.
3492 Warn whenever a @code{switch} statement has an index of enumerated type
3493 and lacks a @code{case} for one or more of the named codes of that
3494 enumeration. (The presence of a @code{default} label prevents this
3495 warning.) @code{case} labels outside the enumeration range also
3496 provoke warnings when this option is used (even if there is a
3497 @code{default} label).
3498 This warning is enabled by @option{-Wall}.
3500 @item -Wswitch-default
3501 @opindex Wswitch-default
3502 @opindex Wno-switch-default
3503 Warn whenever a @code{switch} statement does not have a @code{default}
3507 @opindex Wswitch-enum
3508 @opindex Wno-switch-enum
3509 Warn whenever a @code{switch} statement has an index of enumerated type
3510 and lacks a @code{case} for one or more of the named codes of that
3511 enumeration. @code{case} labels outside the enumeration range also
3512 provoke warnings when this option is used. The only difference
3513 between @option{-Wswitch} and this option is that this option gives a
3514 warning about an omitted enumeration code even if there is a
3515 @code{default} label.
3517 @item -Wsync-nand @r{(C and C++ only)}
3519 @opindex Wno-sync-nand
3520 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3521 built-in functions are used. These functions changed semantics in GCC 4.4.
3525 @opindex Wno-trigraphs
3526 Warn if any trigraphs are encountered that might change the meaning of
3527 the program (trigraphs within comments are not warned about).
3528 This warning is enabled by @option{-Wall}.
3530 @item -Wunused-but-set-parameter
3531 @opindex Wunused-but-set-parameter
3532 @opindex Wno-unused-but-set-parameter
3533 Warn whenever a function parameter is assigned to, but otherwise unused
3534 (aside from its declaration).
3536 To suppress this warning use the @samp{unused} attribute
3537 (@pxref{Variable Attributes}).
3539 This warning is also enabled by @option{-Wunused} together with
3542 @item -Wunused-but-set-variable
3543 @opindex Wunused-but-set-variable
3544 @opindex Wno-unused-but-set-variable
3545 Warn whenever a local variable is assigned to, but otherwise unused
3546 (aside from its declaration).
3547 This warning is enabled by @option{-Wall}.
3549 To suppress this warning use the @samp{unused} attribute
3550 (@pxref{Variable Attributes}).
3552 This warning is also enabled by @option{-Wunused}, which is enabled
3555 @item -Wunused-function
3556 @opindex Wunused-function
3557 @opindex Wno-unused-function
3558 Warn whenever a static function is declared but not defined or a
3559 non-inline static function is unused.
3560 This warning is enabled by @option{-Wall}.
3562 @item -Wunused-label
3563 @opindex Wunused-label
3564 @opindex Wno-unused-label
3565 Warn whenever a label is declared but not used.
3566 This warning is enabled by @option{-Wall}.
3568 To suppress this warning use the @samp{unused} attribute
3569 (@pxref{Variable Attributes}).
3571 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3572 @opindex Wunused-local-typedefs
3573 Warn when a typedef locally defined in a function is not used.
3575 @item -Wunused-parameter
3576 @opindex Wunused-parameter
3577 @opindex Wno-unused-parameter
3578 Warn whenever a function parameter is unused aside from its declaration.
3580 To suppress this warning use the @samp{unused} attribute
3581 (@pxref{Variable Attributes}).
3583 @item -Wno-unused-result
3584 @opindex Wunused-result
3585 @opindex Wno-unused-result
3586 Do not warn if a caller of a function marked with attribute
3587 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3588 its return value. The default is @option{-Wunused-result}.
3590 @item -Wunused-variable
3591 @opindex Wunused-variable
3592 @opindex Wno-unused-variable
3593 Warn whenever a local variable or non-constant static variable is unused
3594 aside from its declaration.
3595 This warning is enabled by @option{-Wall}.
3597 To suppress this warning use the @samp{unused} attribute
3598 (@pxref{Variable Attributes}).
3600 @item -Wunused-value
3601 @opindex Wunused-value
3602 @opindex Wno-unused-value
3603 Warn whenever a statement computes a result that is explicitly not
3604 used. To suppress this warning cast the unused expression to
3605 @samp{void}. This includes an expression-statement or the left-hand
3606 side of a comma expression that contains no side effects. For example,
3607 an expression such as @samp{x[i,j]} will cause a warning, while
3608 @samp{x[(void)i,j]} will not.
3610 This warning is enabled by @option{-Wall}.
3615 All the above @option{-Wunused} options combined.
3617 In order to get a warning about an unused function parameter, you must
3618 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3619 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3621 @item -Wuninitialized
3622 @opindex Wuninitialized
3623 @opindex Wno-uninitialized
3624 Warn if an automatic variable is used without first being initialized
3625 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3626 warn if a non-static reference or non-static @samp{const} member
3627 appears in a class without constructors.
3629 If you want to warn about code that uses the uninitialized value of the
3630 variable in its own initializer, use the @option{-Winit-self} option.
3632 These warnings occur for individual uninitialized or clobbered
3633 elements of structure, union or array variables as well as for
3634 variables that are uninitialized or clobbered as a whole. They do
3635 not occur for variables or elements declared @code{volatile}. Because
3636 these warnings depend on optimization, the exact variables or elements
3637 for which there are warnings will depend on the precise optimization
3638 options and version of GCC used.
3640 Note that there may be no warning about a variable that is used only
3641 to compute a value that itself is never used, because such
3642 computations may be deleted by data flow analysis before the warnings
3645 @item -Wmaybe-uninitialized
3646 @opindex Wmaybe-uninitialized
3647 @opindex Wno-maybe-uninitialized
3648 For an automatic variable, if there exists a path from the function
3649 entry to a use of the variable that is initialized, but there exist
3650 some other paths the variable is not initialized, the compiler will
3651 emit a warning if it can not prove the uninitialized paths do not
3652 happen at run time. These warnings are made optional because GCC is
3653 not smart enough to see all the reasons why the code might be correct
3654 despite appearing to have an error. Here is one example of how
3675 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3676 always initialized, but GCC doesn't know this. To suppress the
3677 warning, the user needs to provide a default case with assert(0) or
3680 @cindex @code{longjmp} warnings
3681 This option also warns when a non-volatile automatic variable might be
3682 changed by a call to @code{longjmp}. These warnings as well are possible
3683 only in optimizing compilation.
3685 The compiler sees only the calls to @code{setjmp}. It cannot know
3686 where @code{longjmp} will be called; in fact, a signal handler could
3687 call it at any point in the code. As a result, you may get a warning
3688 even when there is in fact no problem because @code{longjmp} cannot
3689 in fact be called at the place that would cause a problem.
3691 Some spurious warnings can be avoided if you declare all the functions
3692 you use that never return as @code{noreturn}. @xref{Function
3695 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3697 @item -Wunknown-pragmas
3698 @opindex Wunknown-pragmas
3699 @opindex Wno-unknown-pragmas
3700 @cindex warning for unknown pragmas
3701 @cindex unknown pragmas, warning
3702 @cindex pragmas, warning of unknown
3703 Warn when a @code{#pragma} directive is encountered that is not understood by
3704 GCC@. If this command-line option is used, warnings will even be issued
3705 for unknown pragmas in system header files. This is not the case if
3706 the warnings were only enabled by the @option{-Wall} command-line option.
3709 @opindex Wno-pragmas
3711 Do not warn about misuses of pragmas, such as incorrect parameters,
3712 invalid syntax, or conflicts between pragmas. See also
3713 @samp{-Wunknown-pragmas}.
3715 @item -Wstrict-aliasing
3716 @opindex Wstrict-aliasing
3717 @opindex Wno-strict-aliasing
3718 This option is only active when @option{-fstrict-aliasing} is active.
3719 It warns about code that might break the strict aliasing rules that the
3720 compiler is using for optimization. The warning does not catch all
3721 cases, but does attempt to catch the more common pitfalls. It is
3722 included in @option{-Wall}.
3723 It is equivalent to @option{-Wstrict-aliasing=3}
3725 @item -Wstrict-aliasing=n
3726 @opindex Wstrict-aliasing=n
3727 @opindex Wno-strict-aliasing=n
3728 This option is only active when @option{-fstrict-aliasing} is active.
3729 It warns about code that might break the strict aliasing rules that the
3730 compiler is using for optimization.
3731 Higher levels correspond to higher accuracy (fewer false positives).
3732 Higher levels also correspond to more effort, similar to the way -O works.
3733 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3736 Level 1: Most aggressive, quick, least accurate.
3737 Possibly useful when higher levels
3738 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3739 false negatives. However, it has many false positives.
3740 Warns for all pointer conversions between possibly incompatible types,
3741 even if never dereferenced. Runs in the frontend only.
3743 Level 2: Aggressive, quick, not too precise.
3744 May still have many false positives (not as many as level 1 though),
3745 and few false negatives (but possibly more than level 1).
3746 Unlike level 1, it only warns when an address is taken. Warns about
3747 incomplete types. Runs in the frontend only.
3749 Level 3 (default for @option{-Wstrict-aliasing}):
3750 Should have very few false positives and few false
3751 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3752 Takes care of the common pun+dereference pattern in the frontend:
3753 @code{*(int*)&some_float}.
3754 If optimization is enabled, it also runs in the backend, where it deals
3755 with multiple statement cases using flow-sensitive points-to information.
3756 Only warns when the converted pointer is dereferenced.
3757 Does not warn about incomplete types.
3759 @item -Wstrict-overflow
3760 @itemx -Wstrict-overflow=@var{n}
3761 @opindex Wstrict-overflow
3762 @opindex Wno-strict-overflow
3763 This option is only active when @option{-fstrict-overflow} is active.
3764 It warns about cases where the compiler optimizes based on the
3765 assumption that signed overflow does not occur. Note that it does not
3766 warn about all cases where the code might overflow: it only warns
3767 about cases where the compiler implements some optimization. Thus
3768 this warning depends on the optimization level.
3770 An optimization that assumes that signed overflow does not occur is
3771 perfectly safe if the values of the variables involved are such that
3772 overflow never does, in fact, occur. Therefore this warning can
3773 easily give a false positive: a warning about code that is not
3774 actually a problem. To help focus on important issues, several
3775 warning levels are defined. No warnings are issued for the use of
3776 undefined signed overflow when estimating how many iterations a loop
3777 will require, in particular when determining whether a loop will be
3781 @item -Wstrict-overflow=1
3782 Warn about cases that are both questionable and easy to avoid. For
3783 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3784 compiler will simplify this to @code{1}. This level of
3785 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3786 are not, and must be explicitly requested.
3788 @item -Wstrict-overflow=2
3789 Also warn about other cases where a comparison is simplified to a
3790 constant. For example: @code{abs (x) >= 0}. This can only be
3791 simplified when @option{-fstrict-overflow} is in effect, because
3792 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3793 zero. @option{-Wstrict-overflow} (with no level) is the same as
3794 @option{-Wstrict-overflow=2}.
3796 @item -Wstrict-overflow=3
3797 Also warn about other cases where a comparison is simplified. For
3798 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3800 @item -Wstrict-overflow=4
3801 Also warn about other simplifications not covered by the above cases.
3802 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3804 @item -Wstrict-overflow=5
3805 Also warn about cases where the compiler reduces the magnitude of a
3806 constant involved in a comparison. For example: @code{x + 2 > y} will
3807 be simplified to @code{x + 1 >= y}. This is reported only at the
3808 highest warning level because this simplification applies to many
3809 comparisons, so this warning level will give a very large number of
3813 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3814 @opindex Wsuggest-attribute=
3815 @opindex Wno-suggest-attribute=
3816 Warn for cases where adding an attribute may be beneficial. The
3817 attributes currently supported are listed below.
3820 @item -Wsuggest-attribute=pure
3821 @itemx -Wsuggest-attribute=const
3822 @itemx -Wsuggest-attribute=noreturn
3823 @opindex Wsuggest-attribute=pure
3824 @opindex Wno-suggest-attribute=pure
3825 @opindex Wsuggest-attribute=const
3826 @opindex Wno-suggest-attribute=const
3827 @opindex Wsuggest-attribute=noreturn
3828 @opindex Wno-suggest-attribute=noreturn
3830 Warn about functions that might be candidates for attributes
3831 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3832 functions visible in other compilation units or (in the case of @code{pure} and
3833 @code{const}) if it cannot prove that the function returns normally. A function
3834 returns normally if it doesn't contain an infinite loop nor returns abnormally
3835 by throwing, calling @code{abort()} or trapping. This analysis requires option
3836 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3837 higher. Higher optimization levels improve the accuracy of the analysis.
3840 @item -Warray-bounds
3841 @opindex Wno-array-bounds
3842 @opindex Warray-bounds
3843 This option is only active when @option{-ftree-vrp} is active
3844 (default for @option{-O2} and above). It warns about subscripts to arrays
3845 that are always out of bounds. This warning is enabled by @option{-Wall}.
3847 @item -Wno-div-by-zero
3848 @opindex Wno-div-by-zero
3849 @opindex Wdiv-by-zero
3850 Do not warn about compile-time integer division by zero. Floating-point
3851 division by zero is not warned about, as it can be a legitimate way of
3852 obtaining infinities and NaNs.
3854 @item -Wsystem-headers
3855 @opindex Wsystem-headers
3856 @opindex Wno-system-headers
3857 @cindex warnings from system headers
3858 @cindex system headers, warnings from
3859 Print warning messages for constructs found in system header files.
3860 Warnings from system headers are normally suppressed, on the assumption
3861 that they usually do not indicate real problems and would only make the
3862 compiler output harder to read. Using this command-line option tells
3863 GCC to emit warnings from system headers as if they occurred in user
3864 code. However, note that using @option{-Wall} in conjunction with this
3865 option will @emph{not} warn about unknown pragmas in system
3866 headers---for that, @option{-Wunknown-pragmas} must also be used.
3869 @opindex Wtrampolines
3870 @opindex Wno-trampolines
3871 Warn about trampolines generated for pointers to nested functions.
3873 A trampoline is a small piece of data or code that is created at run
3874 time on the stack when the address of a nested function is taken, and
3875 is used to call the nested function indirectly. For some targets, it
3876 is made up of data only and thus requires no special treatment. But,
3877 for most targets, it is made up of code and thus requires the stack
3878 to be made executable in order for the program to work properly.
3881 @opindex Wfloat-equal
3882 @opindex Wno-float-equal
3883 Warn if floating-point values are used in equality comparisons.
3885 The idea behind this is that sometimes it is convenient (for the
3886 programmer) to consider floating-point values as approximations to
3887 infinitely precise real numbers. If you are doing this, then you need
3888 to compute (by analyzing the code, or in some other way) the maximum or
3889 likely maximum error that the computation introduces, and allow for it
3890 when performing comparisons (and when producing output, but that's a
3891 different problem). In particular, instead of testing for equality, you
3892 would check to see whether the two values have ranges that overlap; and
3893 this is done with the relational operators, so equality comparisons are
3896 @item -Wtraditional @r{(C and Objective-C only)}
3897 @opindex Wtraditional
3898 @opindex Wno-traditional
3899 Warn about certain constructs that behave differently in traditional and
3900 ISO C@. Also warn about ISO C constructs that have no traditional C
3901 equivalent, and/or problematic constructs that should be avoided.
3905 Macro parameters that appear within string literals in the macro body.
3906 In traditional C macro replacement takes place within string literals,
3907 but does not in ISO C@.
3910 In traditional C, some preprocessor directives did not exist.
3911 Traditional preprocessors would only consider a line to be a directive
3912 if the @samp{#} appeared in column 1 on the line. Therefore
3913 @option{-Wtraditional} warns about directives that traditional C
3914 understands but would ignore because the @samp{#} does not appear as the
3915 first character on the line. It also suggests you hide directives like
3916 @samp{#pragma} not understood by traditional C by indenting them. Some
3917 traditional implementations would not recognize @samp{#elif}, so it
3918 suggests avoiding it altogether.
3921 A function-like macro that appears without arguments.
3924 The unary plus operator.
3927 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
3928 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3929 constants.) Note, these suffixes appear in macros defined in the system
3930 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3931 Use of these macros in user code might normally lead to spurious
3932 warnings, however GCC's integrated preprocessor has enough context to
3933 avoid warning in these cases.
3936 A function declared external in one block and then used after the end of
3940 A @code{switch} statement has an operand of type @code{long}.
3943 A non-@code{static} function declaration follows a @code{static} one.
3944 This construct is not accepted by some traditional C compilers.
3947 The ISO type of an integer constant has a different width or
3948 signedness from its traditional type. This warning is only issued if
3949 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3950 typically represent bit patterns, are not warned about.
3953 Usage of ISO string concatenation is detected.
3956 Initialization of automatic aggregates.
3959 Identifier conflicts with labels. Traditional C lacks a separate
3960 namespace for labels.
3963 Initialization of unions. If the initializer is zero, the warning is
3964 omitted. This is done under the assumption that the zero initializer in
3965 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3966 initializer warnings and relies on default initialization to zero in the
3970 Conversions by prototypes between fixed/floating-point values and vice
3971 versa. The absence of these prototypes when compiling with traditional
3972 C would cause serious problems. This is a subset of the possible
3973 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3976 Use of ISO C style function definitions. This warning intentionally is
3977 @emph{not} issued for prototype declarations or variadic functions
3978 because these ISO C features will appear in your code when using
3979 libiberty's traditional C compatibility macros, @code{PARAMS} and
3980 @code{VPARAMS}. This warning is also bypassed for nested functions
3981 because that feature is already a GCC extension and thus not relevant to
3982 traditional C compatibility.
3985 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3986 @opindex Wtraditional-conversion
3987 @opindex Wno-traditional-conversion
3988 Warn if a prototype causes a type conversion that is different from what
3989 would happen to the same argument in the absence of a prototype. This
3990 includes conversions of fixed point to floating and vice versa, and
3991 conversions changing the width or signedness of a fixed-point argument
3992 except when the same as the default promotion.
3994 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3995 @opindex Wdeclaration-after-statement
3996 @opindex Wno-declaration-after-statement
3997 Warn when a declaration is found after a statement in a block. This
3998 construct, known from C++, was introduced with ISO C99 and is by default
3999 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4000 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4005 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4007 @item -Wno-endif-labels
4008 @opindex Wno-endif-labels
4009 @opindex Wendif-labels
4010 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4015 Warn whenever a local variable or type declaration shadows another variable,
4016 parameter, type, or class member (in C++), or whenever a built-in function
4017 is shadowed. Note that in C++, the compiler will not warn if a local variable
4018 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
4020 @item -Wlarger-than=@var{len}
4021 @opindex Wlarger-than=@var{len}
4022 @opindex Wlarger-than-@var{len}
4023 Warn whenever an object of larger than @var{len} bytes is defined.
4025 @item -Wframe-larger-than=@var{len}
4026 @opindex Wframe-larger-than
4027 Warn if the size of a function frame is larger than @var{len} bytes.
4028 The computation done to determine the stack frame size is approximate
4029 and not conservative.
4030 The actual requirements may be somewhat greater than @var{len}
4031 even if you do not get a warning. In addition, any space allocated
4032 via @code{alloca}, variable-length arrays, or related constructs
4033 is not included by the compiler when determining
4034 whether or not to issue a warning.
4036 @item -Wno-free-nonheap-object
4037 @opindex Wno-free-nonheap-object
4038 @opindex Wfree-nonheap-object
4039 Do not warn when attempting to free an object that was not allocated
4042 @item -Wstack-usage=@var{len}
4043 @opindex Wstack-usage
4044 Warn if the stack usage of a function might be larger than @var{len} bytes.
4045 The computation done to determine the stack usage is conservative.
4046 Any space allocated via @code{alloca}, variable-length arrays, or related
4047 constructs is included by the compiler when determining whether or not to
4050 The message is in keeping with the output of @option{-fstack-usage}.
4054 If the stack usage is fully static but exceeds the specified amount, it's:
4057 warning: stack usage is 1120 bytes
4060 If the stack usage is (partly) dynamic but bounded, it's:
4063 warning: stack usage might be 1648 bytes
4066 If the stack usage is (partly) dynamic and not bounded, it's:
4069 warning: stack usage might be unbounded
4073 @item -Wunsafe-loop-optimizations
4074 @opindex Wunsafe-loop-optimizations
4075 @opindex Wno-unsafe-loop-optimizations
4076 Warn if the loop cannot be optimized because the compiler could not
4077 assume anything on the bounds of the loop indices. With
4078 @option{-funsafe-loop-optimizations} warn if the compiler made
4081 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4082 @opindex Wno-pedantic-ms-format
4083 @opindex Wpedantic-ms-format
4084 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4085 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4086 depending on the MS runtime, when you are using the options @option{-Wformat}
4087 and @option{-pedantic} without gnu-extensions.
4089 @item -Wpointer-arith
4090 @opindex Wpointer-arith
4091 @opindex Wno-pointer-arith
4092 Warn about anything that depends on the ``size of'' a function type or
4093 of @code{void}. GNU C assigns these types a size of 1, for
4094 convenience in calculations with @code{void *} pointers and pointers
4095 to functions. In C++, warn also when an arithmetic operation involves
4096 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4099 @opindex Wtype-limits
4100 @opindex Wno-type-limits
4101 Warn if a comparison is always true or always false due to the limited
4102 range of the data type, but do not warn for constant expressions. For
4103 example, warn if an unsigned variable is compared against zero with
4104 @samp{<} or @samp{>=}. This warning is also enabled by
4107 @item -Wbad-function-cast @r{(C and Objective-C only)}
4108 @opindex Wbad-function-cast
4109 @opindex Wno-bad-function-cast
4110 Warn whenever a function call is cast to a non-matching type.
4111 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4113 @item -Wc++-compat @r{(C and Objective-C only)}
4114 Warn about ISO C constructs that are outside of the common subset of
4115 ISO C and ISO C++, e.g.@: request for implicit conversion from
4116 @code{void *} to a pointer to non-@code{void} type.
4118 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4119 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4120 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4121 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4122 enabled by @option{-Wall}.
4126 @opindex Wno-cast-qual
4127 Warn whenever a pointer is cast so as to remove a type qualifier from
4128 the target type. For example, warn if a @code{const char *} is cast
4129 to an ordinary @code{char *}.
4131 Also warn when making a cast that introduces a type qualifier in an
4132 unsafe way. For example, casting @code{char **} to @code{const char **}
4133 is unsafe, as in this example:
4136 /* p is char ** value. */
4137 const char **q = (const char **) p;
4138 /* Assignment of readonly string to const char * is OK. */
4140 /* Now char** pointer points to read-only memory. */
4145 @opindex Wcast-align
4146 @opindex Wno-cast-align
4147 Warn whenever a pointer is cast such that the required alignment of the
4148 target is increased. For example, warn if a @code{char *} is cast to
4149 an @code{int *} on machines where integers can only be accessed at
4150 two- or four-byte boundaries.
4152 @item -Wwrite-strings
4153 @opindex Wwrite-strings
4154 @opindex Wno-write-strings
4155 When compiling C, give string constants the type @code{const
4156 char[@var{length}]} so that copying the address of one into a
4157 non-@code{const} @code{char *} pointer will get a warning. These
4158 warnings will help you find at compile time code that can try to write
4159 into a string constant, but only if you have been very careful about
4160 using @code{const} in declarations and prototypes. Otherwise, it will
4161 just be a nuisance. This is why we did not make @option{-Wall} request
4164 When compiling C++, warn about the deprecated conversion from string
4165 literals to @code{char *}. This warning is enabled by default for C++
4170 @opindex Wno-clobbered
4171 Warn for variables that might be changed by @samp{longjmp} or
4172 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4175 @opindex Wconversion
4176 @opindex Wno-conversion
4177 Warn for implicit conversions that may alter a value. This includes
4178 conversions between real and integer, like @code{abs (x)} when
4179 @code{x} is @code{double}; conversions between signed and unsigned,
4180 like @code{unsigned ui = -1}; and conversions to smaller types, like
4181 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4182 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4183 changed by the conversion like in @code{abs (2.0)}. Warnings about
4184 conversions between signed and unsigned integers can be disabled by
4185 using @option{-Wno-sign-conversion}.
4187 For C++, also warn for confusing overload resolution for user-defined
4188 conversions; and conversions that will never use a type conversion
4189 operator: conversions to @code{void}, the same type, a base class or a
4190 reference to them. Warnings about conversions between signed and
4191 unsigned integers are disabled by default in C++ unless
4192 @option{-Wsign-conversion} is explicitly enabled.
4194 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4195 @opindex Wconversion-null
4196 @opindex Wno-conversion-null
4197 Do not warn for conversions between @code{NULL} and non-pointer
4198 types. @option{-Wconversion-null} is enabled by default.
4200 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4201 @opindex Wzero-as-null-pointer-constant
4202 @opindex Wno-zero-as-null-pointer-constant
4203 Warn when a literal '0' is used as null pointer constant. This can
4204 be useful to facilitate the conversion to @code{nullptr} in C++11.
4207 @opindex Wempty-body
4208 @opindex Wno-empty-body
4209 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4210 while} statement. This warning is also enabled by @option{-Wextra}.
4212 @item -Wenum-compare
4213 @opindex Wenum-compare
4214 @opindex Wno-enum-compare
4215 Warn about a comparison between values of different enum types. In C++
4216 this warning is enabled by default. In C this warning is enabled by
4219 @item -Wjump-misses-init @r{(C, Objective-C only)}
4220 @opindex Wjump-misses-init
4221 @opindex Wno-jump-misses-init
4222 Warn if a @code{goto} statement or a @code{switch} statement jumps
4223 forward across the initialization of a variable, or jumps backward to a
4224 label after the variable has been initialized. This only warns about
4225 variables that are initialized when they are declared. This warning is
4226 only supported for C and Objective C; in C++ this sort of branch is an
4229 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4230 can be disabled with the @option{-Wno-jump-misses-init} option.
4232 @item -Wsign-compare
4233 @opindex Wsign-compare
4234 @opindex Wno-sign-compare
4235 @cindex warning for comparison of signed and unsigned values
4236 @cindex comparison of signed and unsigned values, warning
4237 @cindex signed and unsigned values, comparison warning
4238 Warn when a comparison between signed and unsigned values could produce
4239 an incorrect result when the signed value is converted to unsigned.
4240 This warning is also enabled by @option{-Wextra}; to get the other warnings
4241 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4243 @item -Wsign-conversion
4244 @opindex Wsign-conversion
4245 @opindex Wno-sign-conversion
4246 Warn for implicit conversions that may change the sign of an integer
4247 value, like assigning a signed integer expression to an unsigned
4248 integer variable. An explicit cast silences the warning. In C, this
4249 option is enabled also by @option{-Wconversion}.
4253 @opindex Wno-address
4254 Warn about suspicious uses of memory addresses. These include using
4255 the address of a function in a conditional expression, such as
4256 @code{void func(void); if (func)}, and comparisons against the memory
4257 address of a string literal, such as @code{if (x == "abc")}. Such
4258 uses typically indicate a programmer error: the address of a function
4259 always evaluates to true, so their use in a conditional usually
4260 indicate that the programmer forgot the parentheses in a function
4261 call; and comparisons against string literals result in unspecified
4262 behavior and are not portable in C, so they usually indicate that the
4263 programmer intended to use @code{strcmp}. This warning is enabled by
4267 @opindex Wlogical-op
4268 @opindex Wno-logical-op
4269 Warn about suspicious uses of logical operators in expressions.
4270 This includes using logical operators in contexts where a
4271 bit-wise operator is likely to be expected.
4273 @item -Waggregate-return
4274 @opindex Waggregate-return
4275 @opindex Wno-aggregate-return
4276 Warn if any functions that return structures or unions are defined or
4277 called. (In languages where you can return an array, this also elicits
4280 @item -Wno-attributes
4281 @opindex Wno-attributes
4282 @opindex Wattributes
4283 Do not warn if an unexpected @code{__attribute__} is used, such as
4284 unrecognized attributes, function attributes applied to variables,
4285 etc. This will not stop errors for incorrect use of supported
4288 @item -Wno-builtin-macro-redefined
4289 @opindex Wno-builtin-macro-redefined
4290 @opindex Wbuiltin-macro-redefined
4291 Do not warn if certain built-in macros are redefined. This suppresses
4292 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4293 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4295 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4296 @opindex Wstrict-prototypes
4297 @opindex Wno-strict-prototypes
4298 Warn if a function is declared or defined without specifying the
4299 argument types. (An old-style function definition is permitted without
4300 a warning if preceded by a declaration that specifies the argument
4303 @item -Wold-style-declaration @r{(C and Objective-C only)}
4304 @opindex Wold-style-declaration
4305 @opindex Wno-old-style-declaration
4306 Warn for obsolescent usages, according to the C Standard, in a
4307 declaration. For example, warn if storage-class specifiers like
4308 @code{static} are not the first things in a declaration. This warning
4309 is also enabled by @option{-Wextra}.
4311 @item -Wold-style-definition @r{(C and Objective-C only)}
4312 @opindex Wold-style-definition
4313 @opindex Wno-old-style-definition
4314 Warn if an old-style function definition is used. A warning is given
4315 even if there is a previous prototype.
4317 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4318 @opindex Wmissing-parameter-type
4319 @opindex Wno-missing-parameter-type
4320 A function parameter is declared without a type specifier in K&R-style
4327 This warning is also enabled by @option{-Wextra}.
4329 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4330 @opindex Wmissing-prototypes
4331 @opindex Wno-missing-prototypes
4332 Warn if a global function is defined without a previous prototype
4333 declaration. This warning is issued even if the definition itself
4334 provides a prototype. The aim is to detect global functions that
4335 are not declared in header files.
4337 @item -Wmissing-declarations
4338 @opindex Wmissing-declarations
4339 @opindex Wno-missing-declarations
4340 Warn if a global function is defined without a previous declaration.
4341 Do so even if the definition itself provides a prototype.
4342 Use this option to detect global functions that are not declared in
4343 header files. In C++, no warnings are issued for function templates,
4344 or for inline functions, or for functions in anonymous namespaces.
4346 @item -Wmissing-field-initializers
4347 @opindex Wmissing-field-initializers
4348 @opindex Wno-missing-field-initializers
4352 Warn if a structure's initializer has some fields missing. For
4353 example, the following code would cause such a warning, because
4354 @code{x.h} is implicitly zero:
4357 struct s @{ int f, g, h; @};
4358 struct s x = @{ 3, 4 @};
4361 This option does not warn about designated initializers, so the following
4362 modification would not trigger a warning:
4365 struct s @{ int f, g, h; @};
4366 struct s x = @{ .f = 3, .g = 4 @};
4369 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4370 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4372 @item -Wmissing-format-attribute
4373 @opindex Wmissing-format-attribute
4374 @opindex Wno-missing-format-attribute
4377 Warn about function pointers that might be candidates for @code{format}
4378 attributes. Note these are only possible candidates, not absolute ones.
4379 GCC will guess that function pointers with @code{format} attributes that
4380 are used in assignment, initialization, parameter passing or return
4381 statements should have a corresponding @code{format} attribute in the
4382 resulting type. I.e.@: the left-hand side of the assignment or
4383 initialization, the type of the parameter variable, or the return type
4384 of the containing function respectively should also have a @code{format}
4385 attribute to avoid the warning.
4387 GCC will also warn about function definitions that might be
4388 candidates for @code{format} attributes. Again, these are only
4389 possible candidates. GCC will guess that @code{format} attributes
4390 might be appropriate for any function that calls a function like
4391 @code{vprintf} or @code{vscanf}, but this might not always be the
4392 case, and some functions for which @code{format} attributes are
4393 appropriate may not be detected.
4395 @item -Wno-multichar
4396 @opindex Wno-multichar
4398 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4399 Usually they indicate a typo in the user's code, as they have
4400 implementation-defined values, and should not be used in portable code.
4402 @item -Wnormalized=<none|id|nfc|nfkc>
4403 @opindex Wnormalized=
4406 @cindex character set, input normalization
4407 In ISO C and ISO C++, two identifiers are different if they are
4408 different sequences of characters. However, sometimes when characters
4409 outside the basic ASCII character set are used, you can have two
4410 different character sequences that look the same. To avoid confusion,
4411 the ISO 10646 standard sets out some @dfn{normalization rules} which
4412 when applied ensure that two sequences that look the same are turned into
4413 the same sequence. GCC can warn you if you are using identifiers that
4414 have not been normalized; this option controls that warning.
4416 There are four levels of warning supported by GCC. The default is
4417 @option{-Wnormalized=nfc}, which warns about any identifier that is
4418 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4419 recommended form for most uses.
4421 Unfortunately, there are some characters allowed in identifiers by
4422 ISO C and ISO C++ that, when turned into NFC, are not allowed in
4423 identifiers. That is, there's no way to use these symbols in portable
4424 ISO C or C++ and have all your identifiers in NFC@.
4425 @option{-Wnormalized=id} suppresses the warning for these characters.
4426 It is hoped that future versions of the standards involved will correct
4427 this, which is why this option is not the default.
4429 You can switch the warning off for all characters by writing
4430 @option{-Wnormalized=none}. You would only want to do this if you
4431 were using some other normalization scheme (like ``D''), because
4432 otherwise you can easily create bugs that are literally impossible to see.
4434 Some characters in ISO 10646 have distinct meanings but look identical
4435 in some fonts or display methodologies, especially once formatting has
4436 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4437 LETTER N'', will display just like a regular @code{n} that has been
4438 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4439 normalization scheme to convert all these into a standard form as
4440 well, and GCC will warn if your code is not in NFKC if you use
4441 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4442 about every identifier that contains the letter O because it might be
4443 confused with the digit 0, and so is not the default, but may be
4444 useful as a local coding convention if the programming environment is
4445 unable to be fixed to display these characters distinctly.
4447 @item -Wno-deprecated
4448 @opindex Wno-deprecated
4449 @opindex Wdeprecated
4450 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4452 @item -Wno-deprecated-declarations
4453 @opindex Wno-deprecated-declarations
4454 @opindex Wdeprecated-declarations
4455 Do not warn about uses of functions (@pxref{Function Attributes}),
4456 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4457 Attributes}) marked as deprecated by using the @code{deprecated}
4461 @opindex Wno-overflow
4463 Do not warn about compile-time overflow in constant expressions.
4465 @item -Woverride-init @r{(C and Objective-C only)}
4466 @opindex Woverride-init
4467 @opindex Wno-override-init
4471 Warn if an initialized field without side effects is overridden when
4472 using designated initializers (@pxref{Designated Inits, , Designated
4475 This warning is included in @option{-Wextra}. To get other
4476 @option{-Wextra} warnings without this one, use @samp{-Wextra
4477 -Wno-override-init}.
4482 Warn if a structure is given the packed attribute, but the packed
4483 attribute has no effect on the layout or size of the structure.
4484 Such structures may be mis-aligned for little benefit. For
4485 instance, in this code, the variable @code{f.x} in @code{struct bar}
4486 will be misaligned even though @code{struct bar} does not itself
4487 have the packed attribute:
4494 @} __attribute__((packed));
4502 @item -Wpacked-bitfield-compat
4503 @opindex Wpacked-bitfield-compat
4504 @opindex Wno-packed-bitfield-compat
4505 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4506 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4507 the change can lead to differences in the structure layout. GCC
4508 informs you when the offset of such a field has changed in GCC 4.4.
4509 For example there is no longer a 4-bit padding between field @code{a}
4510 and @code{b} in this structure:
4517 @} __attribute__ ((packed));
4520 This warning is enabled by default. Use
4521 @option{-Wno-packed-bitfield-compat} to disable this warning.
4526 Warn if padding is included in a structure, either to align an element
4527 of the structure or to align the whole structure. Sometimes when this
4528 happens it is possible to rearrange the fields of the structure to
4529 reduce the padding and so make the structure smaller.
4531 @item -Wredundant-decls
4532 @opindex Wredundant-decls
4533 @opindex Wno-redundant-decls
4534 Warn if anything is declared more than once in the same scope, even in
4535 cases where multiple declaration is valid and changes nothing.
4537 @item -Wnested-externs @r{(C and Objective-C only)}
4538 @opindex Wnested-externs
4539 @opindex Wno-nested-externs
4540 Warn if an @code{extern} declaration is encountered within a function.
4545 Warn if a function can not be inlined and it was declared as inline.
4546 Even with this option, the compiler will not warn about failures to
4547 inline functions declared in system headers.
4549 The compiler uses a variety of heuristics to determine whether or not
4550 to inline a function. For example, the compiler takes into account
4551 the size of the function being inlined and the amount of inlining
4552 that has already been done in the current function. Therefore,
4553 seemingly insignificant changes in the source program can cause the
4554 warnings produced by @option{-Winline} to appear or disappear.
4556 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4557 @opindex Wno-invalid-offsetof
4558 @opindex Winvalid-offsetof
4559 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4560 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4561 to a non-POD type is undefined. In existing C++ implementations,
4562 however, @samp{offsetof} typically gives meaningful results even when
4563 applied to certain kinds of non-POD types. (Such as a simple
4564 @samp{struct} that fails to be a POD type only by virtue of having a
4565 constructor.) This flag is for users who are aware that they are
4566 writing nonportable code and who have deliberately chosen to ignore the
4569 The restrictions on @samp{offsetof} may be relaxed in a future version
4570 of the C++ standard.
4572 @item -Wno-int-to-pointer-cast
4573 @opindex Wno-int-to-pointer-cast
4574 @opindex Wint-to-pointer-cast
4575 Suppress warnings from casts to pointer type of an integer of a
4576 different size. In C++, casting to a pointer type of smaller size is
4577 an error. @option{Wint-to-pointer-cast} is enabled by default.
4580 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4581 @opindex Wno-pointer-to-int-cast
4582 @opindex Wpointer-to-int-cast
4583 Suppress warnings from casts from a pointer to an integer type of a
4587 @opindex Winvalid-pch
4588 @opindex Wno-invalid-pch
4589 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4590 the search path but can't be used.
4594 @opindex Wno-long-long
4595 Warn if @samp{long long} type is used. This is enabled by either
4596 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4597 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4599 @item -Wvariadic-macros
4600 @opindex Wvariadic-macros
4601 @opindex Wno-variadic-macros
4602 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4603 alternate syntax when in pedantic ISO C99 mode. This is default.
4604 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4606 @item -Wvector-operation-performance
4607 @opindex Wvector-operation-performance
4608 @opindex Wno-vector-operation-performance
4609 Warn if vector operation is not implemented via SIMD capabilities of the
4610 architecture. Mainly useful for the performance tuning.
4611 Vector operation can be implemented @code{piecewise}, which means that the
4612 scalar operation is performed on every vector element;
4613 @code{in parallel}, which means that the vector operation is implemented
4614 using scalars of wider type, which normally is more performance efficient;
4615 and @code{as a single scalar}, which means that vector fits into a
4621 Warn if variable length array is used in the code.
4622 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4623 the variable length array.
4625 @item -Wvolatile-register-var
4626 @opindex Wvolatile-register-var
4627 @opindex Wno-volatile-register-var
4628 Warn if a register variable is declared volatile. The volatile
4629 modifier does not inhibit all optimizations that may eliminate reads
4630 and/or writes to register variables. This warning is enabled by
4633 @item -Wdisabled-optimization
4634 @opindex Wdisabled-optimization
4635 @opindex Wno-disabled-optimization
4636 Warn if a requested optimization pass is disabled. This warning does
4637 not generally indicate that there is anything wrong with your code; it
4638 merely indicates that GCC's optimizers were unable to handle the code
4639 effectively. Often, the problem is that your code is too big or too
4640 complex; GCC will refuse to optimize programs when the optimization
4641 itself is likely to take inordinate amounts of time.
4643 @item -Wpointer-sign @r{(C and Objective-C only)}
4644 @opindex Wpointer-sign
4645 @opindex Wno-pointer-sign
4646 Warn for pointer argument passing or assignment with different signedness.
4647 This option is only supported for C and Objective-C@. It is implied by
4648 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4649 @option{-Wno-pointer-sign}.
4651 @item -Wstack-protector
4652 @opindex Wstack-protector
4653 @opindex Wno-stack-protector
4654 This option is only active when @option{-fstack-protector} is active. It
4655 warns about functions that will not be protected against stack smashing.
4658 @opindex Wno-mudflap
4659 Suppress warnings about constructs that cannot be instrumented by
4662 @item -Woverlength-strings
4663 @opindex Woverlength-strings
4664 @opindex Wno-overlength-strings
4665 Warn about string constants that are longer than the ``minimum
4666 maximum'' length specified in the C standard. Modern compilers
4667 generally allow string constants that are much longer than the
4668 standard's minimum limit, but very portable programs should avoid
4669 using longer strings.
4671 The limit applies @emph{after} string constant concatenation, and does
4672 not count the trailing NUL@. In C90, the limit was 509 characters; in
4673 C99, it was raised to 4095. C++98 does not specify a normative
4674 minimum maximum, so we do not diagnose overlength strings in C++@.
4676 This option is implied by @option{-pedantic}, and can be disabled with
4677 @option{-Wno-overlength-strings}.
4679 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4680 @opindex Wunsuffixed-float-constants
4682 GCC will issue a warning for any floating constant that does not have
4683 a suffix. When used together with @option{-Wsystem-headers} it will
4684 warn about such constants in system header files. This can be useful
4685 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4686 from the decimal floating-point extension to C99.
4689 @node Debugging Options
4690 @section Options for Debugging Your Program or GCC
4691 @cindex options, debugging
4692 @cindex debugging information options
4694 GCC has various special options that are used for debugging
4695 either your program or GCC:
4700 Produce debugging information in the operating system's native format
4701 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4704 On most systems that use stabs format, @option{-g} enables use of extra
4705 debugging information that only GDB can use; this extra information
4706 makes debugging work better in GDB but will probably make other debuggers
4708 refuse to read the program. If you want to control for certain whether
4709 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4710 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4712 GCC allows you to use @option{-g} with
4713 @option{-O}. The shortcuts taken by optimized code may occasionally
4714 produce surprising results: some variables you declared may not exist
4715 at all; flow of control may briefly move where you did not expect it;
4716 some statements may not be executed because they compute constant
4717 results or their values were already at hand; some statements may
4718 execute in different places because they were moved out of loops.
4720 Nevertheless it proves possible to debug optimized output. This makes
4721 it reasonable to use the optimizer for programs that might have bugs.
4723 The following options are useful when GCC is generated with the
4724 capability for more than one debugging format.
4728 Produce debugging information for use by GDB@. This means to use the
4729 most expressive format available (DWARF 2, stabs, or the native format
4730 if neither of those are supported), including GDB extensions if at all
4735 Produce debugging information in stabs format (if that is supported),
4736 without GDB extensions. This is the format used by DBX on most BSD
4737 systems. On MIPS, Alpha and System V Release 4 systems this option
4738 produces stabs debugging output that is not understood by DBX or SDB@.
4739 On System V Release 4 systems this option requires the GNU assembler.
4741 @item -feliminate-unused-debug-symbols
4742 @opindex feliminate-unused-debug-symbols
4743 Produce debugging information in stabs format (if that is supported),
4744 for only symbols that are actually used.
4746 @item -femit-class-debug-always
4747 Instead of emitting debugging information for a C++ class in only one
4748 object file, emit it in all object files using the class. This option
4749 should be used only with debuggers that are unable to handle the way GCC
4750 normally emits debugging information for classes because using this
4751 option will increase the size of debugging information by as much as a
4754 @item -fno-debug-types-section
4755 @opindex fno-debug-types-section
4756 @opindex fdebug-types-section
4757 By default when using DWARF v4 or higher type DIEs will be put into
4758 their own .debug_types section instead of making them part of the
4759 .debug_info section. It is more efficient to put them in a separate
4760 comdat sections since the linker will then be able to remove duplicates.
4761 But not all DWARF consumers support .debug_types sections yet.
4765 Produce debugging information in stabs format (if that is supported),
4766 using GNU extensions understood only by the GNU debugger (GDB)@. The
4767 use of these extensions is likely to make other debuggers crash or
4768 refuse to read the program.
4772 Produce debugging information in COFF format (if that is supported).
4773 This is the format used by SDB on most System V systems prior to
4778 Produce debugging information in XCOFF format (if that is supported).
4779 This is the format used by the DBX debugger on IBM RS/6000 systems.
4783 Produce debugging information in XCOFF format (if that is supported),
4784 using GNU extensions understood only by the GNU debugger (GDB)@. The
4785 use of these extensions is likely to make other debuggers crash or
4786 refuse to read the program, and may cause assemblers other than the GNU
4787 assembler (GAS) to fail with an error.
4789 @item -gdwarf-@var{version}
4790 @opindex gdwarf-@var{version}
4791 Produce debugging information in DWARF format (if that is
4792 supported). This is the format used by DBX on IRIX 6. The value
4793 of @var{version} may be either 2, 3 or 4; the default version is 2.
4795 Note that with DWARF version 2 some ports require, and will always
4796 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4798 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4799 for maximum benefit.
4801 @item -grecord-gcc-switches
4802 @opindex grecord-gcc-switches
4803 This switch causes the command-line options used to invoke the
4804 compiler that may affect code generation to be appended to the
4805 DW_AT_producer attribute in DWARF debugging information. The options
4806 are concatenated with spaces separating them from each other and from
4807 the compiler version. See also @option{-frecord-gcc-switches} for another
4808 way of storing compiler options into the object file.
4810 @item -gno-record-gcc-switches
4811 @opindex gno-record-gcc-switches
4812 Disallow appending command-line options to the DW_AT_producer attribute
4813 in DWARF debugging information. This is the default.
4815 @item -gstrict-dwarf
4816 @opindex gstrict-dwarf
4817 Disallow using extensions of later DWARF standard version than selected
4818 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4819 DWARF extensions from later standard versions is allowed.
4821 @item -gno-strict-dwarf
4822 @opindex gno-strict-dwarf
4823 Allow using extensions of later DWARF standard version than selected with
4824 @option{-gdwarf-@var{version}}.
4828 Produce debugging information in VMS debug format (if that is
4829 supported). This is the format used by DEBUG on VMS systems.
4832 @itemx -ggdb@var{level}
4833 @itemx -gstabs@var{level}
4834 @itemx -gcoff@var{level}
4835 @itemx -gxcoff@var{level}
4836 @itemx -gvms@var{level}
4837 Request debugging information and also use @var{level} to specify how
4838 much information. The default level is 2.
4840 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4843 Level 1 produces minimal information, enough for making backtraces in
4844 parts of the program that you don't plan to debug. This includes
4845 descriptions of functions and external variables, but no information
4846 about local variables and no line numbers.
4848 Level 3 includes extra information, such as all the macro definitions
4849 present in the program. Some debuggers support macro expansion when
4850 you use @option{-g3}.
4852 @option{-gdwarf-2} does not accept a concatenated debug level, because
4853 GCC used to support an option @option{-gdwarf} that meant to generate
4854 debug information in version 1 of the DWARF format (which is very
4855 different from version 2), and it would have been too confusing. That
4856 debug format is long obsolete, but the option cannot be changed now.
4857 Instead use an additional @option{-g@var{level}} option to change the
4858 debug level for DWARF.
4862 Turn off generation of debug info, if leaving out this option would have
4863 generated it, or turn it on at level 2 otherwise. The position of this
4864 argument in the command line does not matter, it takes effect after all
4865 other options are processed, and it does so only once, no matter how
4866 many times it is given. This is mainly intended to be used with
4867 @option{-fcompare-debug}.
4869 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4870 @opindex fdump-final-insns
4871 Dump the final internal representation (RTL) to @var{file}. If the
4872 optional argument is omitted (or if @var{file} is @code{.}), the name
4873 of the dump file will be determined by appending @code{.gkd} to the
4874 compilation output file name.
4876 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4877 @opindex fcompare-debug
4878 @opindex fno-compare-debug
4879 If no error occurs during compilation, run the compiler a second time,
4880 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4881 passed to the second compilation. Dump the final internal
4882 representation in both compilations, and print an error if they differ.
4884 If the equal sign is omitted, the default @option{-gtoggle} is used.
4886 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4887 and nonzero, implicitly enables @option{-fcompare-debug}. If
4888 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4889 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4892 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4893 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4894 of the final representation and the second compilation, preventing even
4895 @env{GCC_COMPARE_DEBUG} from taking effect.
4897 To verify full coverage during @option{-fcompare-debug} testing, set
4898 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4899 which GCC will reject as an invalid option in any actual compilation
4900 (rather than preprocessing, assembly or linking). To get just a
4901 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4902 not overridden} will do.
4904 @item -fcompare-debug-second
4905 @opindex fcompare-debug-second
4906 This option is implicitly passed to the compiler for the second
4907 compilation requested by @option{-fcompare-debug}, along with options to
4908 silence warnings, and omitting other options that would cause
4909 side-effect compiler outputs to files or to the standard output. Dump
4910 files and preserved temporary files are renamed so as to contain the
4911 @code{.gk} additional extension during the second compilation, to avoid
4912 overwriting those generated by the first.
4914 When this option is passed to the compiler driver, it causes the
4915 @emph{first} compilation to be skipped, which makes it useful for little
4916 other than debugging the compiler proper.
4918 @item -feliminate-dwarf2-dups
4919 @opindex feliminate-dwarf2-dups
4920 Compress DWARF2 debugging information by eliminating duplicated
4921 information about each symbol. This option only makes sense when
4922 generating DWARF2 debugging information with @option{-gdwarf-2}.
4924 @item -femit-struct-debug-baseonly
4925 Emit debug information for struct-like types
4926 only when the base name of the compilation source file
4927 matches the base name of file in which the struct was defined.
4929 This option substantially reduces the size of debugging information,
4930 but at significant potential loss in type information to the debugger.
4931 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4932 See @option{-femit-struct-debug-detailed} for more detailed control.
4934 This option works only with DWARF 2.
4936 @item -femit-struct-debug-reduced
4937 Emit debug information for struct-like types
4938 only when the base name of the compilation source file
4939 matches the base name of file in which the type was defined,
4940 unless the struct is a template or defined in a system header.
4942 This option significantly reduces the size of debugging information,
4943 with some potential loss in type information to the debugger.
4944 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4945 See @option{-femit-struct-debug-detailed} for more detailed control.
4947 This option works only with DWARF 2.
4949 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4950 Specify the struct-like types
4951 for which the compiler will generate debug information.
4952 The intent is to reduce duplicate struct debug information
4953 between different object files within the same program.
4955 This option is a detailed version of
4956 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4957 which will serve for most needs.
4959 A specification has the syntax@*
4960 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4962 The optional first word limits the specification to
4963 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4964 A struct type is used directly when it is the type of a variable, member.
4965 Indirect uses arise through pointers to structs.
4966 That is, when use of an incomplete struct would be legal, the use is indirect.
4968 @samp{struct one direct; struct two * indirect;}.
4970 The optional second word limits the specification to
4971 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4972 Generic structs are a bit complicated to explain.
4973 For C++, these are non-explicit specializations of template classes,
4974 or non-template classes within the above.
4975 Other programming languages have generics,
4976 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4978 The third word specifies the source files for those
4979 structs for which the compiler will emit debug information.
4980 The values @samp{none} and @samp{any} have the normal meaning.
4981 The value @samp{base} means that
4982 the base of name of the file in which the type declaration appears
4983 must match the base of the name of the main compilation file.
4984 In practice, this means that
4985 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4986 but types declared in other header will not.
4987 The value @samp{sys} means those types satisfying @samp{base}
4988 or declared in system or compiler headers.
4990 You may need to experiment to determine the best settings for your application.
4992 The default is @samp{-femit-struct-debug-detailed=all}.
4994 This option works only with DWARF 2.
4996 @item -fno-merge-debug-strings
4997 @opindex fmerge-debug-strings
4998 @opindex fno-merge-debug-strings
4999 Direct the linker to not merge together strings in the debugging
5000 information that are identical in different object files. Merging is
5001 not supported by all assemblers or linkers. Merging decreases the size
5002 of the debug information in the output file at the cost of increasing
5003 link processing time. Merging is enabled by default.
5005 @item -fdebug-prefix-map=@var{old}=@var{new}
5006 @opindex fdebug-prefix-map
5007 When compiling files in directory @file{@var{old}}, record debugging
5008 information describing them as in @file{@var{new}} instead.
5010 @item -fno-dwarf2-cfi-asm
5011 @opindex fdwarf2-cfi-asm
5012 @opindex fno-dwarf2-cfi-asm
5013 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5014 instead of using GAS @code{.cfi_*} directives.
5016 @cindex @command{prof}
5019 Generate extra code to write profile information suitable for the
5020 analysis program @command{prof}. You must use this option when compiling
5021 the source files you want data about, and you must also use it when
5024 @cindex @command{gprof}
5027 Generate extra code to write profile information suitable for the
5028 analysis program @command{gprof}. You must use this option when compiling
5029 the source files you want data about, and you must also use it when
5034 Makes the compiler print out each function name as it is compiled, and
5035 print some statistics about each pass when it finishes.
5038 @opindex ftime-report
5039 Makes the compiler print some statistics about the time consumed by each
5040 pass when it finishes.
5043 @opindex fmem-report
5044 Makes the compiler print some statistics about permanent memory
5045 allocation when it finishes.
5047 @item -fpre-ipa-mem-report
5048 @opindex fpre-ipa-mem-report
5049 @item -fpost-ipa-mem-report
5050 @opindex fpost-ipa-mem-report
5051 Makes the compiler print some statistics about permanent memory
5052 allocation before or after interprocedural optimization.
5055 @opindex fstack-usage
5056 Makes the compiler output stack usage information for the program, on a
5057 per-function basis. The filename for the dump is made by appending
5058 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5059 the output file, if explicitly specified and it is not an executable,
5060 otherwise it is the basename of the source file. An entry is made up
5065 The name of the function.
5069 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5072 The qualifier @code{static} means that the function manipulates the stack
5073 statically: a fixed number of bytes are allocated for the frame on function
5074 entry and released on function exit; no stack adjustments are otherwise made
5075 in the function. The second field is this fixed number of bytes.
5077 The qualifier @code{dynamic} means that the function manipulates the stack
5078 dynamically: in addition to the static allocation described above, stack
5079 adjustments are made in the body of the function, for example to push/pop
5080 arguments around function calls. If the qualifier @code{bounded} is also
5081 present, the amount of these adjustments is bounded at compile time and
5082 the second field is an upper bound of the total amount of stack used by
5083 the function. If it is not present, the amount of these adjustments is
5084 not bounded at compile time and the second field only represents the
5087 @item -fprofile-arcs
5088 @opindex fprofile-arcs
5089 Add code so that program flow @dfn{arcs} are instrumented. During
5090 execution the program records how many times each branch and call is
5091 executed and how many times it is taken or returns. When the compiled
5092 program exits it saves this data to a file called
5093 @file{@var{auxname}.gcda} for each source file. The data may be used for
5094 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5095 test coverage analysis (@option{-ftest-coverage}). Each object file's
5096 @var{auxname} is generated from the name of the output file, if
5097 explicitly specified and it is not the final executable, otherwise it is
5098 the basename of the source file. In both cases any suffix is removed
5099 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5100 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5101 @xref{Cross-profiling}.
5103 @cindex @command{gcov}
5107 This option is used to compile and link code instrumented for coverage
5108 analysis. The option is a synonym for @option{-fprofile-arcs}
5109 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5110 linking). See the documentation for those options for more details.
5115 Compile the source files with @option{-fprofile-arcs} plus optimization
5116 and code generation options. For test coverage analysis, use the
5117 additional @option{-ftest-coverage} option. You do not need to profile
5118 every source file in a program.
5121 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5122 (the latter implies the former).
5125 Run the program on a representative workload to generate the arc profile
5126 information. This may be repeated any number of times. You can run
5127 concurrent instances of your program, and provided that the file system
5128 supports locking, the data files will be correctly updated. Also
5129 @code{fork} calls are detected and correctly handled (double counting
5133 For profile-directed optimizations, compile the source files again with
5134 the same optimization and code generation options plus
5135 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5136 Control Optimization}).
5139 For test coverage analysis, use @command{gcov} to produce human readable
5140 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5141 @command{gcov} documentation for further information.
5145 With @option{-fprofile-arcs}, for each function of your program GCC
5146 creates a program flow graph, then finds a spanning tree for the graph.
5147 Only arcs that are not on the spanning tree have to be instrumented: the
5148 compiler adds code to count the number of times that these arcs are
5149 executed. When an arc is the only exit or only entrance to a block, the
5150 instrumentation code can be added to the block; otherwise, a new basic
5151 block must be created to hold the instrumentation code.
5154 @item -ftest-coverage
5155 @opindex ftest-coverage
5156 Produce a notes file that the @command{gcov} code-coverage utility
5157 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5158 show program coverage. Each source file's note file is called
5159 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5160 above for a description of @var{auxname} and instructions on how to
5161 generate test coverage data. Coverage data will match the source files
5162 more closely, if you do not optimize.
5164 @item -fdbg-cnt-list
5165 @opindex fdbg-cnt-list
5166 Print the name and the counter upper bound for all debug counters.
5169 @item -fdbg-cnt=@var{counter-value-list}
5171 Set the internal debug counter upper bound. @var{counter-value-list}
5172 is a comma-separated list of @var{name}:@var{value} pairs
5173 which sets the upper bound of each debug counter @var{name} to @var{value}.
5174 All debug counters have the initial upper bound of @var{UINT_MAX},
5175 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5176 e.g. With -fdbg-cnt=dce:10,tail_call:0
5177 dbg_cnt(dce) will return true only for first 10 invocations
5179 @itemx -fenable-@var{kind}-@var{pass}
5180 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5184 This is a set of debugging options that are used to explicitly disable/enable
5185 optimization passes. For compiler users, regular options for enabling/disabling
5186 passes should be used instead.
5190 @item -fdisable-ipa-@var{pass}
5191 Disable ipa 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.
5195 @item -fdisable-rtl-@var{pass}
5196 @item -fdisable-rtl-@var{pass}=@var{range-list}
5197 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5198 statically invoked in the compiler multiple times, the pass name should be
5199 appended with a sequential number starting from 1. @var{range-list} is a comma
5200 seperated list of function ranges or assembler names. Each range is a number
5201 pair seperated by a colon. The range is inclusive in both ends. If the range
5202 is trivial, the number pair can be simplified as a single number. If the
5203 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5204 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5205 function header of a dump file, and the pass names can be dumped by using
5206 option @option{-fdump-passes}.
5208 @item -fdisable-tree-@var{pass}
5209 @item -fdisable-tree-@var{pass}=@var{range-list}
5210 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5213 @item -fenable-ipa-@var{pass}
5214 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5215 statically invoked in the compiler multiple times, the pass name should be
5216 appended with a sequential number starting from 1.
5218 @item -fenable-rtl-@var{pass}
5219 @item -fenable-rtl-@var{pass}=@var{range-list}
5220 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5221 description and examples.
5223 @item -fenable-tree-@var{pass}
5224 @item -fenable-tree-@var{pass}=@var{range-list}
5225 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5226 of option arguments.
5230 # disable ccp1 for all functions
5232 # disable complete unroll for function whose cgraph node uid is 1
5233 -fenable-tree-cunroll=1
5234 # disable gcse2 for functions at the following ranges [1,1],
5235 # [300,400], and [400,1000]
5236 # disable gcse2 for functions foo and foo2
5237 -fdisable-rtl-gcse2=foo,foo2
5238 # disable early inlining
5239 -fdisable-tree-einline
5240 # disable ipa inlining
5241 -fdisable-ipa-inline
5242 # enable tree full unroll
5243 -fenable-tree-unroll
5249 @item -d@var{letters}
5250 @itemx -fdump-rtl-@var{pass}
5252 Says to make debugging dumps during compilation at times specified by
5253 @var{letters}. This is used for debugging the RTL-based passes of the
5254 compiler. The file names for most of the dumps are made by appending
5255 a pass number and a word to the @var{dumpname}, and the files are
5256 created in the directory of the output file. Note that the pass
5257 number is computed statically as passes get registered into the pass
5258 manager. Thus the numbering is not related to the dynamic order of
5259 execution of passes. In particular, a pass installed by a plugin
5260 could have a number over 200 even if it executed quite early.
5261 @var{dumpname} is generated from the name of the output file, if
5262 explicitly specified and it is not an executable, otherwise it is the
5263 basename of the source file. These switches may have different effects
5264 when @option{-E} is used for preprocessing.
5266 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5267 @option{-d} option @var{letters}. Here are the possible
5268 letters for use in @var{pass} and @var{letters}, and their meanings:
5272 @item -fdump-rtl-alignments
5273 @opindex fdump-rtl-alignments
5274 Dump after branch alignments have been computed.
5276 @item -fdump-rtl-asmcons
5277 @opindex fdump-rtl-asmcons
5278 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5280 @item -fdump-rtl-auto_inc_dec
5281 @opindex fdump-rtl-auto_inc_dec
5282 Dump after auto-inc-dec discovery. This pass is only run on
5283 architectures that have auto inc or auto dec instructions.
5285 @item -fdump-rtl-barriers
5286 @opindex fdump-rtl-barriers
5287 Dump after cleaning up the barrier instructions.
5289 @item -fdump-rtl-bbpart
5290 @opindex fdump-rtl-bbpart
5291 Dump after partitioning hot and cold basic blocks.
5293 @item -fdump-rtl-bbro
5294 @opindex fdump-rtl-bbro
5295 Dump after block reordering.
5297 @item -fdump-rtl-btl1
5298 @itemx -fdump-rtl-btl2
5299 @opindex fdump-rtl-btl2
5300 @opindex fdump-rtl-btl2
5301 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5302 after the two branch
5303 target load optimization passes.
5305 @item -fdump-rtl-bypass
5306 @opindex fdump-rtl-bypass
5307 Dump after jump bypassing and control flow optimizations.
5309 @item -fdump-rtl-combine
5310 @opindex fdump-rtl-combine
5311 Dump after the RTL instruction combination pass.
5313 @item -fdump-rtl-compgotos
5314 @opindex fdump-rtl-compgotos
5315 Dump after duplicating the computed gotos.
5317 @item -fdump-rtl-ce1
5318 @itemx -fdump-rtl-ce2
5319 @itemx -fdump-rtl-ce3
5320 @opindex fdump-rtl-ce1
5321 @opindex fdump-rtl-ce2
5322 @opindex fdump-rtl-ce3
5323 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5324 @option{-fdump-rtl-ce3} enable dumping after the three
5325 if conversion passes.
5327 @itemx -fdump-rtl-cprop_hardreg
5328 @opindex fdump-rtl-cprop_hardreg
5329 Dump after hard register copy propagation.
5331 @itemx -fdump-rtl-csa
5332 @opindex fdump-rtl-csa
5333 Dump after combining stack adjustments.
5335 @item -fdump-rtl-cse1
5336 @itemx -fdump-rtl-cse2
5337 @opindex fdump-rtl-cse1
5338 @opindex fdump-rtl-cse2
5339 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5340 the two common sub-expression elimination passes.
5342 @itemx -fdump-rtl-dce
5343 @opindex fdump-rtl-dce
5344 Dump after the standalone dead code elimination passes.
5346 @itemx -fdump-rtl-dbr
5347 @opindex fdump-rtl-dbr
5348 Dump after delayed branch scheduling.
5350 @item -fdump-rtl-dce1
5351 @itemx -fdump-rtl-dce2
5352 @opindex fdump-rtl-dce1
5353 @opindex fdump-rtl-dce2
5354 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5355 the two dead store elimination passes.
5358 @opindex fdump-rtl-eh
5359 Dump after finalization of EH handling code.
5361 @item -fdump-rtl-eh_ranges
5362 @opindex fdump-rtl-eh_ranges
5363 Dump after conversion of EH handling range regions.
5365 @item -fdump-rtl-expand
5366 @opindex fdump-rtl-expand
5367 Dump after RTL generation.
5369 @item -fdump-rtl-fwprop1
5370 @itemx -fdump-rtl-fwprop2
5371 @opindex fdump-rtl-fwprop1
5372 @opindex fdump-rtl-fwprop2
5373 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5374 dumping after the two forward propagation passes.
5376 @item -fdump-rtl-gcse1
5377 @itemx -fdump-rtl-gcse2
5378 @opindex fdump-rtl-gcse1
5379 @opindex fdump-rtl-gcse2
5380 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5381 after global common subexpression elimination.
5383 @item -fdump-rtl-init-regs
5384 @opindex fdump-rtl-init-regs
5385 Dump after the initialization of the registers.
5387 @item -fdump-rtl-initvals
5388 @opindex fdump-rtl-initvals
5389 Dump after the computation of the initial value sets.
5391 @itemx -fdump-rtl-into_cfglayout
5392 @opindex fdump-rtl-into_cfglayout
5393 Dump after converting to cfglayout mode.
5395 @item -fdump-rtl-ira
5396 @opindex fdump-rtl-ira
5397 Dump after iterated register allocation.
5399 @item -fdump-rtl-jump
5400 @opindex fdump-rtl-jump
5401 Dump after the second jump optimization.
5403 @item -fdump-rtl-loop2
5404 @opindex fdump-rtl-loop2
5405 @option{-fdump-rtl-loop2} enables dumping after the rtl
5406 loop optimization passes.
5408 @item -fdump-rtl-mach
5409 @opindex fdump-rtl-mach
5410 Dump after performing the machine dependent reorganization pass, if that
5413 @item -fdump-rtl-mode_sw
5414 @opindex fdump-rtl-mode_sw
5415 Dump after removing redundant mode switches.
5417 @item -fdump-rtl-rnreg
5418 @opindex fdump-rtl-rnreg
5419 Dump after register renumbering.
5421 @itemx -fdump-rtl-outof_cfglayout
5422 @opindex fdump-rtl-outof_cfglayout
5423 Dump after converting from cfglayout mode.
5425 @item -fdump-rtl-peephole2
5426 @opindex fdump-rtl-peephole2
5427 Dump after the peephole pass.
5429 @item -fdump-rtl-postreload
5430 @opindex fdump-rtl-postreload
5431 Dump after post-reload optimizations.
5433 @itemx -fdump-rtl-pro_and_epilogue
5434 @opindex fdump-rtl-pro_and_epilogue
5435 Dump after generating the function pro and epilogues.
5437 @item -fdump-rtl-regmove
5438 @opindex fdump-rtl-regmove
5439 Dump after the register move pass.
5441 @item -fdump-rtl-sched1
5442 @itemx -fdump-rtl-sched2
5443 @opindex fdump-rtl-sched1
5444 @opindex fdump-rtl-sched2
5445 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5446 after the basic block scheduling passes.
5448 @item -fdump-rtl-see
5449 @opindex fdump-rtl-see
5450 Dump after sign extension elimination.
5452 @item -fdump-rtl-seqabstr
5453 @opindex fdump-rtl-seqabstr
5454 Dump after common sequence discovery.
5456 @item -fdump-rtl-shorten
5457 @opindex fdump-rtl-shorten
5458 Dump after shortening branches.
5460 @item -fdump-rtl-sibling
5461 @opindex fdump-rtl-sibling
5462 Dump after sibling call optimizations.
5464 @item -fdump-rtl-split1
5465 @itemx -fdump-rtl-split2
5466 @itemx -fdump-rtl-split3
5467 @itemx -fdump-rtl-split4
5468 @itemx -fdump-rtl-split5
5469 @opindex fdump-rtl-split1
5470 @opindex fdump-rtl-split2
5471 @opindex fdump-rtl-split3
5472 @opindex fdump-rtl-split4
5473 @opindex fdump-rtl-split5
5474 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5475 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5476 @option{-fdump-rtl-split5} enable dumping after five rounds of
5477 instruction splitting.
5479 @item -fdump-rtl-sms
5480 @opindex fdump-rtl-sms
5481 Dump after modulo scheduling. This pass is only run on some
5484 @item -fdump-rtl-stack
5485 @opindex fdump-rtl-stack
5486 Dump after conversion from GCC's "flat register file" registers to the
5487 x87's stack-like registers. This pass is only run on x86 variants.
5489 @item -fdump-rtl-subreg1
5490 @itemx -fdump-rtl-subreg2
5491 @opindex fdump-rtl-subreg1
5492 @opindex fdump-rtl-subreg2
5493 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5494 the two subreg expansion passes.
5496 @item -fdump-rtl-unshare
5497 @opindex fdump-rtl-unshare
5498 Dump after all rtl has been unshared.
5500 @item -fdump-rtl-vartrack
5501 @opindex fdump-rtl-vartrack
5502 Dump after variable tracking.
5504 @item -fdump-rtl-vregs
5505 @opindex fdump-rtl-vregs
5506 Dump after converting virtual registers to hard registers.
5508 @item -fdump-rtl-web
5509 @opindex fdump-rtl-web
5510 Dump after live range splitting.
5512 @item -fdump-rtl-regclass
5513 @itemx -fdump-rtl-subregs_of_mode_init
5514 @itemx -fdump-rtl-subregs_of_mode_finish
5515 @itemx -fdump-rtl-dfinit
5516 @itemx -fdump-rtl-dfinish
5517 @opindex fdump-rtl-regclass
5518 @opindex fdump-rtl-subregs_of_mode_init
5519 @opindex fdump-rtl-subregs_of_mode_finish
5520 @opindex fdump-rtl-dfinit
5521 @opindex fdump-rtl-dfinish
5522 These dumps are defined but always produce empty files.
5524 @item -fdump-rtl-all
5525 @opindex fdump-rtl-all
5526 Produce all the dumps listed above.
5530 Annotate the assembler output with miscellaneous debugging information.
5534 Dump all macro definitions, at the end of preprocessing, in addition to
5539 Produce a core dump whenever an error occurs.
5543 Print statistics on memory usage, at the end of the run, to
5548 Annotate the assembler output with a comment indicating which
5549 pattern and alternative was used. The length of each instruction is
5554 Dump the RTL in the assembler output as a comment before each instruction.
5555 Also turns on @option{-dp} annotation.
5559 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5560 dump a representation of the control flow graph suitable for viewing with VCG
5561 to @file{@var{file}.@var{pass}.vcg}.
5565 Just generate RTL for a function instead of compiling it. Usually used
5566 with @option{-fdump-rtl-expand}.
5570 @opindex fdump-noaddr
5571 When doing debugging dumps, suppress address output. This makes it more
5572 feasible to use diff on debugging dumps for compiler invocations with
5573 different compiler binaries and/or different
5574 text / bss / data / heap / stack / dso start locations.
5576 @item -fdump-unnumbered
5577 @opindex fdump-unnumbered
5578 When doing debugging dumps, suppress instruction numbers and address output.
5579 This makes it more feasible to use diff on debugging dumps for compiler
5580 invocations with different options, in particular with and without
5583 @item -fdump-unnumbered-links
5584 @opindex fdump-unnumbered-links
5585 When doing debugging dumps (see @option{-d} option above), suppress
5586 instruction numbers for the links to the previous and next instructions
5589 @item -fdump-translation-unit @r{(C++ only)}
5590 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5591 @opindex fdump-translation-unit
5592 Dump a representation of the tree structure for the entire translation
5593 unit to a file. The file name is made by appending @file{.tu} to the
5594 source file name, and the file is created in the same directory as the
5595 output file. If the @samp{-@var{options}} form is used, @var{options}
5596 controls the details of the dump as described for the
5597 @option{-fdump-tree} options.
5599 @item -fdump-class-hierarchy @r{(C++ only)}
5600 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5601 @opindex fdump-class-hierarchy
5602 Dump a representation of each class's hierarchy and virtual function
5603 table layout to a file. The file name is made by appending
5604 @file{.class} to the source file name, and the file is created in the
5605 same directory as the output file. If the @samp{-@var{options}} form
5606 is used, @var{options} controls the details of the dump as described
5607 for the @option{-fdump-tree} options.
5609 @item -fdump-ipa-@var{switch}
5611 Control the dumping at various stages of inter-procedural analysis
5612 language tree to a file. The file name is generated by appending a
5613 switch specific suffix to the source file name, and the file is created
5614 in the same directory as the output file. The following dumps are
5619 Enables all inter-procedural analysis dumps.
5622 Dumps information about call-graph optimization, unused function removal,
5623 and inlining decisions.
5626 Dump after function inlining.
5631 @opindex fdump-passes
5632 Dump the list of optimization passes that are turned on and off by
5633 the current command-line options.
5635 @item -fdump-statistics-@var{option}
5636 @opindex fdump-statistics
5637 Enable and control dumping of pass statistics in a separate file. The
5638 file name is generated by appending a suffix ending in
5639 @samp{.statistics} to the source file name, and the file is created in
5640 the same directory as the output file. If the @samp{-@var{option}}
5641 form is used, @samp{-stats} will cause counters to be summed over the
5642 whole compilation unit while @samp{-details} will dump every event as
5643 the passes generate them. The default with no option is to sum
5644 counters for each function compiled.
5646 @item -fdump-tree-@var{switch}
5647 @itemx -fdump-tree-@var{switch}-@var{options}
5649 Control the dumping at various stages of processing the intermediate
5650 language tree to a file. The file name is generated by appending a
5651 switch specific suffix to the source file name, and the file is
5652 created in the same directory as the output file. If the
5653 @samp{-@var{options}} form is used, @var{options} is a list of
5654 @samp{-} separated options which control the details of the dump. Not
5655 all options are applicable to all dumps; those that are not
5656 meaningful will be ignored. The following options are available
5660 Print the address of each node. Usually this is not meaningful as it
5661 changes according to the environment and source file. Its primary use
5662 is for tying up a dump file with a debug environment.
5664 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5665 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5666 use working backward from mangled names in the assembly file.
5668 Inhibit dumping of members of a scope or body of a function merely
5669 because that scope has been reached. Only dump such items when they
5670 are directly reachable by some other path. When dumping pretty-printed
5671 trees, this option inhibits dumping the bodies of control structures.
5673 Print a raw representation of the tree. By default, trees are
5674 pretty-printed into a C-like representation.
5676 Enable more detailed dumps (not honored by every dump option).
5678 Enable dumping various statistics about the pass (not honored by every dump
5681 Enable showing basic block boundaries (disabled in raw dumps).
5683 Enable showing virtual operands for every statement.
5685 Enable showing line numbers for statements.
5687 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5689 Enable showing the tree dump for each statement.
5691 Enable showing the EH region number holding each statement.
5693 Enable showing scalar evolution analysis details.
5695 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5696 and @option{lineno}.
5699 The following tree dumps are possible:
5703 @opindex fdump-tree-original
5704 Dump before any tree based optimization, to @file{@var{file}.original}.
5707 @opindex fdump-tree-optimized
5708 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5711 @opindex fdump-tree-gimple
5712 Dump each function before and after the gimplification pass to a file. The
5713 file name is made by appending @file{.gimple} to the source file name.
5716 @opindex fdump-tree-cfg
5717 Dump the control flow graph of each function to a file. The file name is
5718 made by appending @file{.cfg} to the source file name.
5721 @opindex fdump-tree-vcg
5722 Dump the control flow graph of each function to a file in VCG format. The
5723 file name is made by appending @file{.vcg} to the source file name. Note
5724 that if the file contains more than one function, the generated file cannot
5725 be used directly by VCG@. You will need to cut and paste each function's
5726 graph into its own separate file first.
5729 @opindex fdump-tree-ch
5730 Dump each function after copying loop headers. The file name is made by
5731 appending @file{.ch} to the source file name.
5734 @opindex fdump-tree-ssa
5735 Dump SSA related information to a file. The file name is made by appending
5736 @file{.ssa} to the source file name.
5739 @opindex fdump-tree-alias
5740 Dump aliasing information for each function. The file name is made by
5741 appending @file{.alias} to the source file name.
5744 @opindex fdump-tree-ccp
5745 Dump each function after CCP@. The file name is made by appending
5746 @file{.ccp} to the source file name.
5749 @opindex fdump-tree-storeccp
5750 Dump each function after STORE-CCP@. The file name is made by appending
5751 @file{.storeccp} to the source file name.
5754 @opindex fdump-tree-pre
5755 Dump trees after partial redundancy elimination. The file name is made
5756 by appending @file{.pre} to the source file name.
5759 @opindex fdump-tree-fre
5760 Dump trees after full redundancy elimination. The file name is made
5761 by appending @file{.fre} to the source file name.
5764 @opindex fdump-tree-copyprop
5765 Dump trees after copy propagation. The file name is made
5766 by appending @file{.copyprop} to the source file name.
5768 @item store_copyprop
5769 @opindex fdump-tree-store_copyprop
5770 Dump trees after store copy-propagation. The file name is made
5771 by appending @file{.store_copyprop} to the source file name.
5774 @opindex fdump-tree-dce
5775 Dump each function after dead code elimination. The file name is made by
5776 appending @file{.dce} to the source file name.
5779 @opindex fdump-tree-mudflap
5780 Dump each function after adding mudflap instrumentation. The file name is
5781 made by appending @file{.mudflap} to the source file name.
5784 @opindex fdump-tree-sra
5785 Dump each function after performing scalar replacement of aggregates. The
5786 file name is made by appending @file{.sra} to the source file name.
5789 @opindex fdump-tree-sink
5790 Dump each function after performing code sinking. The file name is made
5791 by appending @file{.sink} to the source file name.
5794 @opindex fdump-tree-dom
5795 Dump each function after applying dominator tree optimizations. The file
5796 name is made by appending @file{.dom} to the source file name.
5799 @opindex fdump-tree-dse
5800 Dump each function after applying dead store elimination. The file
5801 name is made by appending @file{.dse} to the source file name.
5804 @opindex fdump-tree-phiopt
5805 Dump each function after optimizing PHI nodes into straightline code. The file
5806 name is made by appending @file{.phiopt} to the source file name.
5809 @opindex fdump-tree-forwprop
5810 Dump each function after forward propagating single use variables. The file
5811 name is made by appending @file{.forwprop} to the source file name.
5814 @opindex fdump-tree-copyrename
5815 Dump each function after applying the copy rename optimization. The file
5816 name is made by appending @file{.copyrename} to the source file name.
5819 @opindex fdump-tree-nrv
5820 Dump each function after applying the named return value optimization on
5821 generic trees. The file name is made by appending @file{.nrv} to the source
5825 @opindex fdump-tree-vect
5826 Dump each function after applying vectorization of loops. The file name is
5827 made by appending @file{.vect} to the source file name.
5830 @opindex fdump-tree-slp
5831 Dump each function after applying vectorization of basic blocks. The file name
5832 is made by appending @file{.slp} to the source file name.
5835 @opindex fdump-tree-vrp
5836 Dump each function after Value Range Propagation (VRP). The file name
5837 is made by appending @file{.vrp} to the source file name.
5840 @opindex fdump-tree-all
5841 Enable all the available tree dumps with the flags provided in this option.
5844 @item -ftree-vectorizer-verbose=@var{n}
5845 @opindex ftree-vectorizer-verbose
5846 This option controls the amount of debugging output the vectorizer prints.
5847 This information is written to standard error, unless
5848 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5849 in which case it is output to the usual dump listing file, @file{.vect}.
5850 For @var{n}=0 no diagnostic information is reported.
5851 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5852 and the total number of loops that got vectorized.
5853 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5854 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5855 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5856 level that @option{-fdump-tree-vect-stats} uses.
5857 Higher verbosity levels mean either more information dumped for each
5858 reported loop, or same amount of information reported for more loops:
5859 if @var{n}=3, vectorizer cost model information is reported.
5860 If @var{n}=4, alignment related information is added to the reports.
5861 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5862 memory access-patterns) is added to the reports.
5863 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5864 that did not pass the first analysis phase (i.e., may not be countable, or
5865 may have complicated control-flow).
5866 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5867 If @var{n}=8, SLP related information is added to the reports.
5868 For @var{n}=9, all the information the vectorizer generates during its
5869 analysis and transformation is reported. This is the same verbosity level
5870 that @option{-fdump-tree-vect-details} uses.
5872 @item -frandom-seed=@var{string}
5873 @opindex frandom-seed
5874 This option provides a seed that GCC uses when it would otherwise use
5875 random numbers. It is used to generate certain symbol names
5876 that have to be different in every compiled file. It is also used to
5877 place unique stamps in coverage data files and the object files that
5878 produce them. You can use the @option{-frandom-seed} option to produce
5879 reproducibly identical object files.
5881 The @var{string} should be different for every file you compile.
5883 @item -fsched-verbose=@var{n}
5884 @opindex fsched-verbose
5885 On targets that use instruction scheduling, this option controls the
5886 amount of debugging output the scheduler prints. This information is
5887 written to standard error, unless @option{-fdump-rtl-sched1} or
5888 @option{-fdump-rtl-sched2} is specified, in which case it is output
5889 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5890 respectively. However for @var{n} greater than nine, the output is
5891 always printed to standard error.
5893 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5894 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5895 For @var{n} greater than one, it also output basic block probabilities,
5896 detailed ready list information and unit/insn info. For @var{n} greater
5897 than two, it includes RTL at abort point, control-flow and regions info.
5898 And for @var{n} over four, @option{-fsched-verbose} also includes
5902 @itemx -save-temps=cwd
5904 Store the usual ``temporary'' intermediate files permanently; place them
5905 in the current directory and name them based on the source file. Thus,
5906 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5907 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5908 preprocessed @file{foo.i} output file even though the compiler now
5909 normally uses an integrated preprocessor.
5911 When used in combination with the @option{-x} command-line option,
5912 @option{-save-temps} is sensible enough to avoid over writing an
5913 input source file with the same extension as an intermediate file.
5914 The corresponding intermediate file may be obtained by renaming the
5915 source file before using @option{-save-temps}.
5917 If you invoke GCC in parallel, compiling several different source
5918 files that share a common base name in different subdirectories or the
5919 same source file compiled for multiple output destinations, it is
5920 likely that the different parallel compilers will interfere with each
5921 other, and overwrite the temporary files. For instance:
5924 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5925 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5928 may result in @file{foo.i} and @file{foo.o} being written to
5929 simultaneously by both compilers.
5931 @item -save-temps=obj
5932 @opindex save-temps=obj
5933 Store the usual ``temporary'' intermediate files permanently. If the
5934 @option{-o} option is used, the temporary files are based on the
5935 object file. If the @option{-o} option is not used, the
5936 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5941 gcc -save-temps=obj -c foo.c
5942 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5943 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5946 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5947 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5948 @file{dir2/yfoobar.o}.
5950 @item -time@r{[}=@var{file}@r{]}
5952 Report the CPU time taken by each subprocess in the compilation
5953 sequence. For C source files, this is the compiler proper and assembler
5954 (plus the linker if linking is done).
5956 Without the specification of an output file, the output looks like this:
5963 The first number on each line is the ``user time'', that is time spent
5964 executing the program itself. The second number is ``system time'',
5965 time spent executing operating system routines on behalf of the program.
5966 Both numbers are in seconds.
5968 With the specification of an output file, the output is appended to the
5969 named file, and it looks like this:
5972 0.12 0.01 cc1 @var{options}
5973 0.00 0.01 as @var{options}
5976 The ``user time'' and the ``system time'' are moved before the program
5977 name, and the options passed to the program are displayed, so that one
5978 can later tell what file was being compiled, and with which options.
5980 @item -fvar-tracking
5981 @opindex fvar-tracking
5982 Run variable tracking pass. It computes where variables are stored at each
5983 position in code. Better debugging information is then generated
5984 (if the debugging information format supports this information).
5986 It is enabled by default when compiling with optimization (@option{-Os},
5987 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5988 the debug info format supports it.
5990 @item -fvar-tracking-assignments
5991 @opindex fvar-tracking-assignments
5992 @opindex fno-var-tracking-assignments
5993 Annotate assignments to user variables early in the compilation and
5994 attempt to carry the annotations over throughout the compilation all the
5995 way to the end, in an attempt to improve debug information while
5996 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5998 It can be enabled even if var-tracking is disabled, in which case
5999 annotations will be created and maintained, but discarded at the end.
6001 @item -fvar-tracking-assignments-toggle
6002 @opindex fvar-tracking-assignments-toggle
6003 @opindex fno-var-tracking-assignments-toggle
6004 Toggle @option{-fvar-tracking-assignments}, in the same way that
6005 @option{-gtoggle} toggles @option{-g}.
6007 @item -print-file-name=@var{library}
6008 @opindex print-file-name
6009 Print the full absolute name of the library file @var{library} that
6010 would be used when linking---and don't do anything else. With this
6011 option, GCC does not compile or link anything; it just prints the
6014 @item -print-multi-directory
6015 @opindex print-multi-directory
6016 Print the directory name corresponding to the multilib selected by any
6017 other switches present in the command line. This directory is supposed
6018 to exist in @env{GCC_EXEC_PREFIX}.
6020 @item -print-multi-lib
6021 @opindex print-multi-lib
6022 Print the mapping from multilib directory names to compiler switches
6023 that enable them. The directory name is separated from the switches by
6024 @samp{;}, and each switch starts with an @samp{@@} instead of the
6025 @samp{-}, without spaces between multiple switches. This is supposed to
6026 ease shell-processing.
6028 @item -print-multi-os-directory
6029 @opindex print-multi-os-directory
6030 Print the path to OS libraries for the selected
6031 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6032 present in the @file{lib} subdirectory and no multilibs are used, this is
6033 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6034 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6035 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6036 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6038 @item -print-prog-name=@var{program}
6039 @opindex print-prog-name
6040 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6042 @item -print-libgcc-file-name
6043 @opindex print-libgcc-file-name
6044 Same as @option{-print-file-name=libgcc.a}.
6046 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6047 but you do want to link with @file{libgcc.a}. You can do
6050 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6053 @item -print-search-dirs
6054 @opindex print-search-dirs
6055 Print the name of the configured installation directory and a list of
6056 program and library directories @command{gcc} will search---and don't do anything else.
6058 This is useful when @command{gcc} prints the error message
6059 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6060 To resolve this you either need to put @file{cpp0} and the other compiler
6061 components where @command{gcc} expects to find them, or you can set the environment
6062 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6063 Don't forget the trailing @samp{/}.
6064 @xref{Environment Variables}.
6066 @item -print-sysroot
6067 @opindex print-sysroot
6068 Print the target sysroot directory that will be used during
6069 compilation. This is the target sysroot specified either at configure
6070 time or using the @option{--sysroot} option, possibly with an extra
6071 suffix that depends on compilation options. If no target sysroot is
6072 specified, the option prints nothing.
6074 @item -print-sysroot-headers-suffix
6075 @opindex print-sysroot-headers-suffix
6076 Print the suffix added to the target sysroot when searching for
6077 headers, or give an error if the compiler is not configured with such
6078 a suffix---and don't do anything else.
6081 @opindex dumpmachine
6082 Print the compiler's target machine (for example,
6083 @samp{i686-pc-linux-gnu})---and don't do anything else.
6086 @opindex dumpversion
6087 Print the compiler version (for example, @samp{3.0})---and don't do
6092 Print the compiler's built-in specs---and don't do anything else. (This
6093 is used when GCC itself is being built.) @xref{Spec Files}.
6095 @item -feliminate-unused-debug-types
6096 @opindex feliminate-unused-debug-types
6097 Normally, when producing DWARF2 output, GCC will emit debugging
6098 information for all types declared in a compilation
6099 unit, regardless of whether or not they are actually used
6100 in that compilation unit. Sometimes this is useful, such as
6101 if, in the debugger, you want to cast a value to a type that is
6102 not actually used in your program (but is declared). More often,
6103 however, this results in a significant amount of wasted space.
6104 With this option, GCC will avoid producing debug symbol output
6105 for types that are nowhere used in the source file being compiled.
6108 @node Optimize Options
6109 @section Options That Control Optimization
6110 @cindex optimize options
6111 @cindex options, optimization
6113 These options control various sorts of optimizations.
6115 Without any optimization option, the compiler's goal is to reduce the
6116 cost of compilation and to make debugging produce the expected
6117 results. Statements are independent: if you stop the program with a
6118 breakpoint between statements, you can then assign a new value to any
6119 variable or change the program counter to any other statement in the
6120 function and get exactly the results you would expect from the source
6123 Turning on optimization flags makes the compiler attempt to improve
6124 the performance and/or code size at the expense of compilation time
6125 and possibly the ability to debug the program.
6127 The compiler performs optimization based on the knowledge it has of the
6128 program. Compiling multiple files at once to a single output file mode allows
6129 the compiler to use information gained from all of the files when compiling
6132 Not all optimizations are controlled directly by a flag. Only
6133 optimizations that have a flag are listed in this section.
6135 Most optimizations are only enabled if an @option{-O} level is set on
6136 the command line. Otherwise they are disabled, even if individual
6137 optimization flags are specified.
6139 Depending on the target and how GCC was configured, a slightly different
6140 set of optimizations may be enabled at each @option{-O} level than
6141 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6142 to find out the exact set of optimizations that are enabled at each level.
6143 @xref{Overall Options}, for examples.
6150 Optimize. Optimizing compilation takes somewhat more time, and a lot
6151 more memory for a large function.
6153 With @option{-O}, the compiler tries to reduce code size and execution
6154 time, without performing any optimizations that take a great deal of
6157 @option{-O} turns on the following optimization flags:
6161 -fcprop-registers @gol
6164 -fdelayed-branch @gol
6166 -fguess-branch-probability @gol
6167 -fif-conversion2 @gol
6168 -fif-conversion @gol
6169 -fipa-pure-const @gol
6171 -fipa-reference @gol
6173 -fsplit-wide-types @gol
6175 -ftree-builtin-call-dce @gol
6178 -ftree-copyrename @gol
6180 -ftree-dominator-opts @gol
6182 -ftree-forwprop @gol
6190 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6191 where doing so does not interfere with debugging.
6195 Optimize even more. GCC performs nearly all supported optimizations
6196 that do not involve a space-speed tradeoff.
6197 As compared to @option{-O}, this option increases both compilation time
6198 and the performance of the generated code.
6200 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6201 also turns on the following optimization flags:
6202 @gccoptlist{-fthread-jumps @gol
6203 -falign-functions -falign-jumps @gol
6204 -falign-loops -falign-labels @gol
6207 -fcse-follow-jumps -fcse-skip-blocks @gol
6208 -fdelete-null-pointer-checks @gol
6210 -fexpensive-optimizations @gol
6211 -fgcse -fgcse-lm @gol
6212 -finline-small-functions @gol
6213 -findirect-inlining @gol
6215 -foptimize-sibling-calls @gol
6216 -fpartial-inlining @gol
6219 -freorder-blocks -freorder-functions @gol
6220 -frerun-cse-after-loop @gol
6221 -fsched-interblock -fsched-spec @gol
6222 -fschedule-insns -fschedule-insns2 @gol
6223 -fstrict-aliasing -fstrict-overflow @gol
6224 -ftree-switch-conversion -ftree-tail-merge @gol
6228 Please note the warning under @option{-fgcse} about
6229 invoking @option{-O2} on programs that use computed gotos.
6233 Optimize yet more. @option{-O3} turns on all optimizations specified
6234 by @option{-O2} and also turns on the @option{-finline-functions},
6235 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6236 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6237 @option{-fipa-cp-clone} options.
6241 Reduce compilation time and make debugging produce the expected
6242 results. This is the default.
6246 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6247 do not typically increase code size. It also performs further
6248 optimizations designed to reduce code size.
6250 @option{-Os} disables the following optimization flags:
6251 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6252 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6253 -fprefetch-loop-arrays -ftree-vect-loop-version}
6257 Disregard strict standards compliance. @option{-Ofast} enables all
6258 @option{-O3} optimizations. It also enables optimizations that are not
6259 valid for all standard compliant programs.
6260 It turns on @option{-ffast-math} and the Fortran-specific
6261 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6263 If you use multiple @option{-O} options, with or without level numbers,
6264 the last such option is the one that is effective.
6267 Options of the form @option{-f@var{flag}} specify machine-independent
6268 flags. Most flags have both positive and negative forms; the negative
6269 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6270 below, only one of the forms is listed---the one you typically will
6271 use. You can figure out the other form by either removing @samp{no-}
6274 The following options control specific optimizations. They are either
6275 activated by @option{-O} options or are related to ones that are. You
6276 can use the following flags in the rare cases when ``fine-tuning'' of
6277 optimizations to be performed is desired.
6280 @item -fno-default-inline
6281 @opindex fno-default-inline
6282 Do not make member functions inline by default merely because they are
6283 defined inside the class scope (C++ only). Otherwise, when you specify
6284 @w{@option{-O}}, member functions defined inside class scope are compiled
6285 inline by default; i.e., you don't need to add @samp{inline} in front of
6286 the member function name.
6288 @item -fno-defer-pop
6289 @opindex fno-defer-pop
6290 Always pop the arguments to each function call as soon as that function
6291 returns. For machines that must pop arguments after a function call,
6292 the compiler normally lets arguments accumulate on the stack for several
6293 function calls and pops them all at once.
6295 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6297 @item -fforward-propagate
6298 @opindex fforward-propagate
6299 Perform a forward propagation pass on RTL@. The pass tries to combine two
6300 instructions and checks if the result can be simplified. If loop unrolling
6301 is active, two passes are performed and the second is scheduled after
6304 This option is enabled by default at optimization levels @option{-O},
6305 @option{-O2}, @option{-O3}, @option{-Os}.
6307 @item -ffp-contract=@var{style}
6308 @opindex ffp-contract
6309 @option{-ffp-contract=off} disables floating-point expression contraction.
6310 @option{-ffp-contract=fast} enables floating-point expression contraction
6311 such as forming of fused multiply-add operations if the target has
6312 native support for them.
6313 @option{-ffp-contract=on} enables floating-point expression contraction
6314 if allowed by the language standard. This is currently not implemented
6315 and treated equal to @option{-ffp-contract=off}.
6317 The default is @option{-ffp-contract=fast}.
6319 @item -fomit-frame-pointer
6320 @opindex fomit-frame-pointer
6321 Don't keep the frame pointer in a register for functions that
6322 don't need one. This avoids the instructions to save, set up and
6323 restore frame pointers; it also makes an extra register available
6324 in many functions. @strong{It also makes debugging impossible on
6327 On some machines, such as the VAX, this flag has no effect, because
6328 the standard calling sequence automatically handles the frame pointer
6329 and nothing is saved by pretending it doesn't exist. The
6330 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6331 whether a target machine supports this flag. @xref{Registers,,Register
6332 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6334 Starting with GCC version 4.6, the default setting (when not optimizing for
6335 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6336 @option{-fomit-frame-pointer}. The default can be reverted to
6337 @option{-fno-omit-frame-pointer} by configuring GCC with the
6338 @option{--enable-frame-pointer} configure option.
6340 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6342 @item -foptimize-sibling-calls
6343 @opindex foptimize-sibling-calls
6344 Optimize sibling and tail recursive calls.
6346 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6350 Do not expand any functions inline apart from those marked with
6351 the @code{always_inline} attribute. This is the default when not
6354 Single functions can be exempted from inlining by marking them
6355 with the @code{noinline} attribute.
6357 @item -finline-small-functions
6358 @opindex finline-small-functions
6359 Integrate functions into their callers when their body is smaller than expected
6360 function call code (so overall size of program gets smaller). The compiler
6361 heuristically decides which functions are simple enough to be worth integrating
6362 in this way. This inlining applies to all functions, even those not declared
6365 Enabled at level @option{-O2}.
6367 @item -findirect-inlining
6368 @opindex findirect-inlining
6369 Inline also indirect calls that are discovered to be known at compile
6370 time thanks to previous inlining. This option has any effect only
6371 when inlining itself is turned on by the @option{-finline-functions}
6372 or @option{-finline-small-functions} options.
6374 Enabled at level @option{-O2}.
6376 @item -finline-functions
6377 @opindex finline-functions
6378 Consider all functions for inlining, even if they are not declared inline.
6379 The compiler heuristically decides which functions are worth integrating
6382 If all calls to a given function are integrated, and the function is
6383 declared @code{static}, then the function is normally not output as
6384 assembler code in its own right.
6386 Enabled at level @option{-O3}.
6388 @item -finline-functions-called-once
6389 @opindex finline-functions-called-once
6390 Consider all @code{static} functions called once for inlining into their
6391 caller even if they are not marked @code{inline}. If a call to a given
6392 function is integrated, then the function is not output as assembler code
6395 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6397 @item -fearly-inlining
6398 @opindex fearly-inlining
6399 Inline functions marked by @code{always_inline} and functions whose body seems
6400 smaller than the function call overhead early before doing
6401 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6402 makes profiling significantly cheaper and usually inlining faster on programs
6403 having large chains of nested wrapper functions.
6409 Perform interprocedural scalar replacement of aggregates, removal of
6410 unused parameters and replacement of parameters passed by reference
6411 by parameters passed by value.
6413 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6415 @item -finline-limit=@var{n}
6416 @opindex finline-limit
6417 By default, GCC limits the size of functions that can be inlined. This flag
6418 allows coarse control of this limit. @var{n} is the size of functions that
6419 can be inlined in number of pseudo instructions.
6421 Inlining is actually controlled by a number of parameters, which may be
6422 specified individually by using @option{--param @var{name}=@var{value}}.
6423 The @option{-finline-limit=@var{n}} option sets some of these parameters
6427 @item max-inline-insns-single
6428 is set to @var{n}/2.
6429 @item max-inline-insns-auto
6430 is set to @var{n}/2.
6433 See below for a documentation of the individual
6434 parameters controlling inlining and for the defaults of these parameters.
6436 @emph{Note:} there may be no value to @option{-finline-limit} that results
6437 in default behavior.
6439 @emph{Note:} pseudo instruction represents, in this particular context, an
6440 abstract measurement of function's size. In no way does it represent a count
6441 of assembly instructions and as such its exact meaning might change from one
6442 release to an another.
6444 @item -fno-keep-inline-dllexport
6445 @opindex -fno-keep-inline-dllexport
6446 This is a more fine-grained version of @option{-fkeep-inline-functions},
6447 which applies only to functions that are declared using the @code{dllexport}
6448 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6451 @item -fkeep-inline-functions
6452 @opindex fkeep-inline-functions
6453 In C, emit @code{static} functions that are declared @code{inline}
6454 into the object file, even if the function has been inlined into all
6455 of its callers. This switch does not affect functions using the
6456 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6457 inline functions into the object file.
6459 @item -fkeep-static-consts
6460 @opindex fkeep-static-consts
6461 Emit variables declared @code{static const} when optimization isn't turned
6462 on, even if the variables aren't referenced.
6464 GCC enables this option by default. If you want to force the compiler to
6465 check if the variable was referenced, regardless of whether or not
6466 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6468 @item -fmerge-constants
6469 @opindex fmerge-constants
6470 Attempt to merge identical constants (string constants and floating-point
6471 constants) across compilation units.
6473 This option is the default for optimized compilation if the assembler and
6474 linker support it. Use @option{-fno-merge-constants} to inhibit this
6477 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6479 @item -fmerge-all-constants
6480 @opindex fmerge-all-constants
6481 Attempt to merge identical constants and identical variables.
6483 This option implies @option{-fmerge-constants}. In addition to
6484 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6485 arrays or initialized constant variables with integral or floating-point
6486 types. Languages like C or C++ require each variable, including multiple
6487 instances of the same variable in recursive calls, to have distinct locations,
6488 so using this option will result in non-conforming
6491 @item -fmodulo-sched
6492 @opindex fmodulo-sched
6493 Perform swing modulo scheduling immediately before the first scheduling
6494 pass. This pass looks at innermost loops and reorders their
6495 instructions by overlapping different iterations.
6497 @item -fmodulo-sched-allow-regmoves
6498 @opindex fmodulo-sched-allow-regmoves
6499 Perform more aggressive SMS based modulo scheduling with register moves
6500 allowed. By setting this flag certain anti-dependences edges will be
6501 deleted which will trigger the generation of reg-moves based on the
6502 life-range analysis. This option is effective only with
6503 @option{-fmodulo-sched} enabled.
6505 @item -fno-branch-count-reg
6506 @opindex fno-branch-count-reg
6507 Do not use ``decrement and branch'' instructions on a count register,
6508 but instead generate a sequence of instructions that decrement a
6509 register, compare it against zero, then branch based upon the result.
6510 This option is only meaningful on architectures that support such
6511 instructions, which include x86, PowerPC, IA-64 and S/390.
6513 The default is @option{-fbranch-count-reg}.
6515 @item -fno-function-cse
6516 @opindex fno-function-cse
6517 Do not put function addresses in registers; make each instruction that
6518 calls a constant function contain the function's address explicitly.
6520 This option results in less efficient code, but some strange hacks
6521 that alter the assembler output may be confused by the optimizations
6522 performed when this option is not used.
6524 The default is @option{-ffunction-cse}
6526 @item -fno-zero-initialized-in-bss
6527 @opindex fno-zero-initialized-in-bss
6528 If the target supports a BSS section, GCC by default puts variables that
6529 are initialized to zero into BSS@. This can save space in the resulting
6532 This option turns off this behavior because some programs explicitly
6533 rely on variables going to the data section. E.g., so that the
6534 resulting executable can find the beginning of that section and/or make
6535 assumptions based on that.
6537 The default is @option{-fzero-initialized-in-bss}.
6539 @item -fmudflap -fmudflapth -fmudflapir
6543 @cindex bounds checking
6545 For front-ends that support it (C and C++), instrument all risky
6546 pointer/array dereferencing operations, some standard library
6547 string/heap functions, and some other associated constructs with
6548 range/validity tests. Modules so instrumented should be immune to
6549 buffer overflows, invalid heap use, and some other classes of C/C++
6550 programming errors. The instrumentation relies on a separate runtime
6551 library (@file{libmudflap}), which will be linked into a program if
6552 @option{-fmudflap} is given at link time. Run-time behavior of the
6553 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6554 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6557 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6558 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6559 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6560 instrumentation should ignore pointer reads. This produces less
6561 instrumentation (and therefore faster execution) and still provides
6562 some protection against outright memory corrupting writes, but allows
6563 erroneously read data to propagate within a program.
6565 @item -fthread-jumps
6566 @opindex fthread-jumps
6567 Perform optimizations where we check to see if a jump branches to a
6568 location where another comparison subsumed by the first is found. If
6569 so, the first branch is redirected to either the destination of the
6570 second branch or a point immediately following it, depending on whether
6571 the condition is known to be true or false.
6573 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6575 @item -fsplit-wide-types
6576 @opindex fsplit-wide-types
6577 When using a type that occupies multiple registers, such as @code{long
6578 long} on a 32-bit system, split the registers apart and allocate them
6579 independently. This normally generates better code for those types,
6580 but may make debugging more difficult.
6582 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6585 @item -fcse-follow-jumps
6586 @opindex fcse-follow-jumps
6587 In common subexpression elimination (CSE), scan through jump instructions
6588 when the target of the jump is not reached by any other path. For
6589 example, when CSE encounters an @code{if} statement with an
6590 @code{else} clause, CSE will follow the jump when the condition
6593 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6595 @item -fcse-skip-blocks
6596 @opindex fcse-skip-blocks
6597 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6598 follow jumps that conditionally skip over blocks. When CSE
6599 encounters a simple @code{if} statement with no else clause,
6600 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6601 body of the @code{if}.
6603 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6605 @item -frerun-cse-after-loop
6606 @opindex frerun-cse-after-loop
6607 Re-run common subexpression elimination after loop optimizations has been
6610 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6614 Perform a global common subexpression elimination pass.
6615 This pass also performs global constant and copy propagation.
6617 @emph{Note:} When compiling a program using computed gotos, a GCC
6618 extension, you may get better run-time performance if you disable
6619 the global common subexpression elimination pass by adding
6620 @option{-fno-gcse} to the command line.
6622 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6626 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6627 attempt to move loads that are only killed by stores into themselves. This
6628 allows a loop containing a load/store sequence to be changed to a load outside
6629 the loop, and a copy/store within the loop.
6631 Enabled by default when gcse is enabled.
6635 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6636 global common subexpression elimination. This pass will attempt to move
6637 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6638 loops containing a load/store sequence can be changed to a load before
6639 the loop and a store after the loop.
6641 Not enabled at any optimization level.
6645 When @option{-fgcse-las} is enabled, the global common subexpression
6646 elimination pass eliminates redundant loads that come after stores to the
6647 same memory location (both partial and full redundancies).
6649 Not enabled at any optimization level.
6651 @item -fgcse-after-reload
6652 @opindex fgcse-after-reload
6653 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6654 pass is performed after reload. The purpose of this pass is to cleanup
6657 @item -funsafe-loop-optimizations
6658 @opindex funsafe-loop-optimizations
6659 If given, the loop optimizer will assume that loop indices do not
6660 overflow, and that the loops with nontrivial exit condition are not
6661 infinite. This enables a wider range of loop optimizations even if
6662 the loop optimizer itself cannot prove that these assumptions are valid.
6663 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6664 if it finds this kind of loop.
6666 @item -fcrossjumping
6667 @opindex fcrossjumping
6668 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6669 resulting code may or may not perform better than without cross-jumping.
6671 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6673 @item -fauto-inc-dec
6674 @opindex fauto-inc-dec
6675 Combine increments or decrements of addresses with memory accesses.
6676 This pass is always skipped on architectures that do not have
6677 instructions to support this. Enabled by default at @option{-O} and
6678 higher on architectures that support this.
6682 Perform dead code elimination (DCE) on RTL@.
6683 Enabled by default at @option{-O} and higher.
6687 Perform dead store elimination (DSE) on RTL@.
6688 Enabled by default at @option{-O} and higher.
6690 @item -fif-conversion
6691 @opindex fif-conversion
6692 Attempt to transform conditional jumps into branch-less equivalents. This
6693 include use of conditional moves, min, max, set flags and abs instructions, and
6694 some tricks doable by standard arithmetics. The use of conditional execution
6695 on chips where it is available is controlled by @code{if-conversion2}.
6697 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6699 @item -fif-conversion2
6700 @opindex fif-conversion2
6701 Use conditional execution (where available) to transform conditional jumps into
6702 branch-less equivalents.
6704 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6706 @item -fdelete-null-pointer-checks
6707 @opindex fdelete-null-pointer-checks
6708 Assume that programs cannot safely dereference null pointers, and that
6709 no code or data element resides there. This enables simple constant
6710 folding optimizations at all optimization levels. In addition, other
6711 optimization passes in GCC use this flag to control global dataflow
6712 analyses that eliminate useless checks for null pointers; these assume
6713 that if a pointer is checked after it has already been dereferenced,
6716 Note however that in some environments this assumption is not true.
6717 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6718 for programs that depend on that behavior.
6720 Some targets, especially embedded ones, disable this option at all levels.
6721 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6722 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6723 are enabled independently at different optimization levels.
6725 @item -fdevirtualize
6726 @opindex fdevirtualize
6727 Attempt to convert calls to virtual functions to direct calls. This
6728 is done both within a procedure and interprocedurally as part of
6729 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6730 propagation (@option{-fipa-cp}).
6731 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6733 @item -fexpensive-optimizations
6734 @opindex fexpensive-optimizations
6735 Perform a number of minor optimizations that are relatively expensive.
6737 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6741 Attempt to remove redundant extension instructions. This is especially
6742 helpful for the x86-64 architecture which implicitly zero-extends in 64-bit
6743 registers after writing to their lower 32-bit half.
6745 Enabled for x86 at levels @option{-O2}, @option{-O3}.
6747 @item -foptimize-register-move
6749 @opindex foptimize-register-move
6751 Attempt to reassign register numbers in move instructions and as
6752 operands of other simple instructions in order to maximize the amount of
6753 register tying. This is especially helpful on machines with two-operand
6756 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6759 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6761 @item -fira-algorithm=@var{algorithm}
6762 Use specified coloring algorithm for the integrated register
6763 allocator. The @var{algorithm} argument should be @code{priority} or
6764 @code{CB}. The first algorithm specifies Chow's priority coloring,
6765 the second one specifies Chaitin-Briggs coloring. The second
6766 algorithm can be unimplemented for some architectures. If it is
6767 implemented, it is the default because Chaitin-Briggs coloring as a
6768 rule generates a better code.
6770 @item -fira-region=@var{region}
6771 Use specified regions for the integrated register allocator. The
6772 @var{region} argument should be one of @code{all}, @code{mixed}, or
6773 @code{one}. The first value means using all loops as register
6774 allocation regions, the second value which is enabled by default when
6775 compiling with optimization for speed (@option{-O}, @option{-O2},
6776 @dots{}) means using all loops except for loops with small register
6777 pressure as the regions, and third one which is enabled by default for
6778 @option{-Os} or @option{-O0} means using all function as a single
6779 region. The first value can give best result for machines with small
6780 size and irregular register set, the third one results in faster and
6781 generates decent code and the smallest size code, and the second value
6782 usually give the best results in most cases and for most
6785 @item -fira-loop-pressure
6786 @opindex fira-loop-pressure
6787 Use IRA to evaluate register pressure in loops for decision to move
6788 loop invariants. Usage of this option usually results in generation
6789 of faster and smaller code on machines with big register files (>= 32
6790 registers) but it can slow compiler down.
6792 This option is enabled at level @option{-O3} for some targets.
6794 @item -fno-ira-share-save-slots
6795 @opindex fno-ira-share-save-slots
6796 Switch off sharing stack slots used for saving call used hard
6797 registers living through a call. Each hard register will get a
6798 separate stack slot and as a result function stack frame will be
6801 @item -fno-ira-share-spill-slots
6802 @opindex fno-ira-share-spill-slots
6803 Switch off sharing stack slots allocated for pseudo-registers. Each
6804 pseudo-register that did not get a hard register will get a separate
6805 stack slot and as a result function stack frame will be bigger.
6807 @item -fira-verbose=@var{n}
6808 @opindex fira-verbose
6809 Set up how verbose dump file for the integrated register allocator
6810 will be. Default value is 5. If the value is greater or equal to 10,
6811 the dump file will be stderr as if the value were @var{n} minus 10.
6813 @item -fdelayed-branch
6814 @opindex fdelayed-branch
6815 If supported for the target machine, attempt to reorder instructions
6816 to exploit instruction slots available after delayed branch
6819 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6821 @item -fschedule-insns
6822 @opindex fschedule-insns
6823 If supported for the target machine, attempt to reorder instructions to
6824 eliminate execution stalls due to required data being unavailable. This
6825 helps machines that have slow floating point or memory load instructions
6826 by allowing other instructions to be issued until the result of the load
6827 or floating-point instruction is required.
6829 Enabled at levels @option{-O2}, @option{-O3}.
6831 @item -fschedule-insns2
6832 @opindex fschedule-insns2
6833 Similar to @option{-fschedule-insns}, but requests an additional pass of
6834 instruction scheduling after register allocation has been done. This is
6835 especially useful on machines with a relatively small number of
6836 registers and where memory load instructions take more than one cycle.
6838 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6840 @item -fno-sched-interblock
6841 @opindex fno-sched-interblock
6842 Don't schedule instructions across basic blocks. 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 -fno-sched-spec
6847 @opindex fno-sched-spec
6848 Don't allow speculative motion of non-load instructions. This is normally
6849 enabled by default when scheduling before register allocation, i.e.@:
6850 with @option{-fschedule-insns} or at @option{-O2} or higher.
6852 @item -fsched-pressure
6853 @opindex fsched-pressure
6854 Enable register pressure sensitive insn scheduling before the register
6855 allocation. This only makes sense when scheduling before register
6856 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6857 @option{-O2} or higher. Usage of this option can improve the
6858 generated code and decrease its size by preventing register pressure
6859 increase above the number of available hard registers and as a
6860 consequence register spills in the register allocation.
6862 @item -fsched-spec-load
6863 @opindex fsched-spec-load
6864 Allow speculative motion of some 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-spec-load-dangerous
6869 @opindex fsched-spec-load-dangerous
6870 Allow speculative motion of more load instructions. This only makes
6871 sense when scheduling before register allocation, i.e.@: with
6872 @option{-fschedule-insns} or at @option{-O2} or higher.
6874 @item -fsched-stalled-insns
6875 @itemx -fsched-stalled-insns=@var{n}
6876 @opindex fsched-stalled-insns
6877 Define how many insns (if any) can be moved prematurely from the queue
6878 of stalled insns into the ready list, during the second scheduling pass.
6879 @option{-fno-sched-stalled-insns} means that no insns will be moved
6880 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6881 on how many queued insns can be moved prematurely.
6882 @option{-fsched-stalled-insns} without a value is equivalent to
6883 @option{-fsched-stalled-insns=1}.
6885 @item -fsched-stalled-insns-dep
6886 @itemx -fsched-stalled-insns-dep=@var{n}
6887 @opindex fsched-stalled-insns-dep
6888 Define how many insn groups (cycles) will be examined for a dependency
6889 on a stalled insn that is candidate for premature removal from the queue
6890 of stalled insns. This has an effect only during the second scheduling pass,
6891 and only if @option{-fsched-stalled-insns} is used.
6892 @option{-fno-sched-stalled-insns-dep} is equivalent to
6893 @option{-fsched-stalled-insns-dep=0}.
6894 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6895 @option{-fsched-stalled-insns-dep=1}.
6897 @item -fsched2-use-superblocks
6898 @opindex fsched2-use-superblocks
6899 When scheduling after register allocation, do use superblock scheduling
6900 algorithm. Superblock scheduling allows motion across basic block boundaries
6901 resulting on faster schedules. This option is experimental, as not all machine
6902 descriptions used by GCC model the CPU closely enough to avoid unreliable
6903 results from the algorithm.
6905 This only makes sense when scheduling after register allocation, i.e.@: with
6906 @option{-fschedule-insns2} or at @option{-O2} or higher.
6908 @item -fsched-group-heuristic
6909 @opindex fsched-group-heuristic
6910 Enable the group heuristic in the scheduler. This heuristic favors
6911 the instruction that belongs to a schedule group. This is enabled
6912 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6913 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6915 @item -fsched-critical-path-heuristic
6916 @opindex fsched-critical-path-heuristic
6917 Enable the critical-path heuristic in the scheduler. This heuristic favors
6918 instructions on the critical path. This is enabled by default when
6919 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6920 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6922 @item -fsched-spec-insn-heuristic
6923 @opindex fsched-spec-insn-heuristic
6924 Enable the speculative instruction heuristic in the scheduler. This
6925 heuristic favors speculative instructions with greater dependency weakness.
6926 This is enabled by default when scheduling is enabled, i.e.@:
6927 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6928 or at @option{-O2} or higher.
6930 @item -fsched-rank-heuristic
6931 @opindex fsched-rank-heuristic
6932 Enable the rank heuristic in the scheduler. This heuristic favors
6933 the instruction belonging to a basic block with greater size or frequency.
6934 This is enabled by default when scheduling is enabled, i.e.@:
6935 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6936 at @option{-O2} or higher.
6938 @item -fsched-last-insn-heuristic
6939 @opindex fsched-last-insn-heuristic
6940 Enable the last-instruction heuristic in the scheduler. This heuristic
6941 favors the instruction that is less dependent on the last instruction
6942 scheduled. This is enabled by default when scheduling is enabled,
6943 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6944 at @option{-O2} or higher.
6946 @item -fsched-dep-count-heuristic
6947 @opindex fsched-dep-count-heuristic
6948 Enable the dependent-count heuristic in the scheduler. This heuristic
6949 favors the instruction that has more instructions depending on it.
6950 This is enabled by default when scheduling is enabled, i.e.@:
6951 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6952 at @option{-O2} or higher.
6954 @item -freschedule-modulo-scheduled-loops
6955 @opindex freschedule-modulo-scheduled-loops
6956 The modulo scheduling comes before the traditional scheduling, if a loop
6957 was modulo scheduled we may want to prevent the later scheduling passes
6958 from changing its schedule, we use this option to control that.
6960 @item -fselective-scheduling
6961 @opindex fselective-scheduling
6962 Schedule instructions using selective scheduling algorithm. Selective
6963 scheduling runs instead of the first scheduler pass.
6965 @item -fselective-scheduling2
6966 @opindex fselective-scheduling2
6967 Schedule instructions using selective scheduling algorithm. Selective
6968 scheduling runs instead of the second scheduler pass.
6970 @item -fsel-sched-pipelining
6971 @opindex fsel-sched-pipelining
6972 Enable software pipelining of innermost loops during selective scheduling.
6973 This option has no effect until one of @option{-fselective-scheduling} or
6974 @option{-fselective-scheduling2} is turned on.
6976 @item -fsel-sched-pipelining-outer-loops
6977 @opindex fsel-sched-pipelining-outer-loops
6978 When pipelining loops during selective scheduling, also pipeline outer loops.
6979 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6982 @opindex fshrink-wrap
6983 Emit function prologues only before parts of the function that need it,
6984 rather than at the top of the function. This flag is enabled by default at
6985 @option{-O} and higher.
6987 @item -fcaller-saves
6988 @opindex fcaller-saves
6989 Enable values to be allocated in registers that will be clobbered by
6990 function calls, by emitting extra instructions to save and restore the
6991 registers around such calls. Such allocation is done only when it
6992 seems to result in better code than would otherwise be produced.
6994 This option is always enabled by default on certain machines, usually
6995 those which have no call-preserved registers to use instead.
6997 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6999 @item -fcombine-stack-adjustments
7000 @opindex fcombine-stack-adjustments
7001 Tracks stack adjustments (pushes and pops) and stack memory references
7002 and then tries to find ways to combine them.
7004 Enabled by default at @option{-O1} and higher.
7006 @item -fconserve-stack
7007 @opindex fconserve-stack
7008 Attempt to minimize stack usage. The compiler will attempt to use less
7009 stack space, even if that makes the program slower. This option
7010 implies setting the @option{large-stack-frame} parameter to 100
7011 and the @option{large-stack-frame-growth} parameter to 400.
7013 @item -ftree-reassoc
7014 @opindex ftree-reassoc
7015 Perform reassociation on trees. This flag is enabled by default
7016 at @option{-O} and higher.
7020 Perform partial redundancy elimination (PRE) on trees. This flag is
7021 enabled by default at @option{-O2} and @option{-O3}.
7023 @item -ftree-forwprop
7024 @opindex ftree-forwprop
7025 Perform forward propagation on trees. This flag is enabled by default
7026 at @option{-O} and higher.
7030 Perform full redundancy elimination (FRE) on trees. The difference
7031 between FRE and PRE is that FRE only considers expressions
7032 that are computed on all paths leading to the redundant computation.
7033 This analysis is faster than PRE, though it exposes fewer redundancies.
7034 This flag is enabled by default at @option{-O} and higher.
7036 @item -ftree-phiprop
7037 @opindex ftree-phiprop
7038 Perform hoisting of loads from conditional pointers on trees. This
7039 pass is enabled by default at @option{-O} and higher.
7041 @item -ftree-copy-prop
7042 @opindex ftree-copy-prop
7043 Perform copy propagation on trees. This pass eliminates unnecessary
7044 copy operations. This flag is enabled by default at @option{-O} and
7047 @item -fipa-pure-const
7048 @opindex fipa-pure-const
7049 Discover which functions are pure or constant.
7050 Enabled by default at @option{-O} and higher.
7052 @item -fipa-reference
7053 @opindex fipa-reference
7054 Discover which static variables do not escape cannot escape the
7056 Enabled by default at @option{-O} and higher.
7060 Perform interprocedural pointer analysis and interprocedural modification
7061 and reference analysis. This option can cause excessive memory and
7062 compile-time usage on large compilation units. It is not enabled by
7063 default at any optimization level.
7066 @opindex fipa-profile
7067 Perform interprocedural profile propagation. The functions called only from
7068 cold functions are marked as cold. Also functions executed once (such as
7069 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7070 functions and loop less parts of functions executed once are then optimized for
7072 Enabled by default at @option{-O} and higher.
7076 Perform interprocedural constant propagation.
7077 This optimization analyzes the program to determine when values passed
7078 to functions are constants and then optimizes accordingly.
7079 This optimization can substantially increase performance
7080 if the application has constants passed to functions.
7081 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7083 @item -fipa-cp-clone
7084 @opindex fipa-cp-clone
7085 Perform function cloning to make interprocedural constant propagation stronger.
7086 When enabled, interprocedural constant propagation will perform function cloning
7087 when externally visible function can be called with constant arguments.
7088 Because this optimization can create multiple copies of functions,
7089 it may significantly increase code size
7090 (see @option{--param ipcp-unit-growth=@var{value}}).
7091 This flag is enabled by default at @option{-O3}.
7093 @item -fipa-matrix-reorg
7094 @opindex fipa-matrix-reorg
7095 Perform matrix flattening and transposing.
7096 Matrix flattening tries to replace an @math{m}-dimensional matrix
7097 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7098 This reduces the level of indirection needed for accessing the elements
7099 of the matrix. The second optimization is matrix transposing, which
7100 attempts to change the order of the matrix's dimensions in order to
7101 improve cache locality.
7102 Both optimizations need the @option{-fwhole-program} flag.
7103 Transposing is enabled only if profiling information is available.
7107 Perform forward store motion on trees. This flag is
7108 enabled by default at @option{-O} and higher.
7110 @item -ftree-bit-ccp
7111 @opindex ftree-bit-ccp
7112 Perform sparse conditional bit constant propagation on trees and propagate
7113 pointer alignment information.
7114 This pass only operates on local scalar variables and is enabled by default
7115 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7119 Perform sparse conditional constant propagation (CCP) on trees. This
7120 pass only operates on local scalar variables and is enabled by default
7121 at @option{-O} and higher.
7123 @item -ftree-switch-conversion
7124 Perform conversion of simple initializations in a switch to
7125 initializations from a scalar array. This flag is enabled by default
7126 at @option{-O2} and higher.
7128 @item -ftree-tail-merge
7129 Look for identical code sequences. When found, replace one with a jump to the
7130 other. This optimization is known as tail merging or cross jumping. This flag
7131 is enabled by default at @option{-O2} and higher. The compilation time
7133 be limited using @option{max-tail-merge-comparisons} parameter and
7134 @option{max-tail-merge-iterations} parameter.
7138 Perform dead code elimination (DCE) on trees. This flag is enabled by
7139 default at @option{-O} and higher.
7141 @item -ftree-builtin-call-dce
7142 @opindex ftree-builtin-call-dce
7143 Perform conditional dead code elimination (DCE) for calls to builtin functions
7144 that may set @code{errno} but are otherwise side-effect free. This flag is
7145 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7148 @item -ftree-dominator-opts
7149 @opindex ftree-dominator-opts
7150 Perform a variety of simple scalar cleanups (constant/copy
7151 propagation, redundancy elimination, range propagation and expression
7152 simplification) based on a dominator tree traversal. This also
7153 performs jump threading (to reduce jumps to jumps). This flag is
7154 enabled by default at @option{-O} and higher.
7158 Perform dead store elimination (DSE) on trees. A dead store is a store into
7159 a memory location that is later overwritten by another store without
7160 any intervening loads. In this case the earlier store can be deleted. This
7161 flag is enabled by default at @option{-O} and higher.
7165 Perform loop header copying on trees. This is beneficial since it increases
7166 effectiveness of code motion optimizations. It also saves one jump. This flag
7167 is enabled by default at @option{-O} and higher. It is not enabled
7168 for @option{-Os}, since it usually increases code size.
7170 @item -ftree-loop-optimize
7171 @opindex ftree-loop-optimize
7172 Perform loop optimizations on trees. This flag is enabled by default
7173 at @option{-O} and higher.
7175 @item -ftree-loop-linear
7176 @opindex ftree-loop-linear
7177 Perform loop interchange transformations on tree. Same as
7178 @option{-floop-interchange}. To use this code transformation, GCC has
7179 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7180 enable the Graphite loop transformation infrastructure.
7182 @item -floop-interchange
7183 @opindex floop-interchange
7184 Perform loop interchange transformations on loops. Interchanging two
7185 nested loops switches the inner and outer loops. For example, given a
7190 A(J, I) = A(J, I) * C
7194 loop interchange will transform the loop as if the user had written:
7198 A(J, I) = A(J, I) * C
7202 which can be beneficial when @code{N} is larger than the caches,
7203 because in Fortran, the elements of an array are stored in memory
7204 contiguously by column, and the original loop iterates over rows,
7205 potentially creating at each access a cache miss. This optimization
7206 applies to all the languages supported by GCC and is not limited to
7207 Fortran. To use this code transformation, GCC has to be configured
7208 with @option{--with-ppl} and @option{--with-cloog} to enable the
7209 Graphite loop transformation infrastructure.
7211 @item -floop-strip-mine
7212 @opindex floop-strip-mine
7213 Perform loop strip mining transformations on loops. Strip mining
7214 splits a loop into two nested loops. The outer loop has strides
7215 equal to the strip size and the inner loop has strides of the
7216 original loop within a strip. The strip length can be changed
7217 using the @option{loop-block-tile-size} parameter. For example,
7224 loop strip mining will transform the loop as if the user had written:
7227 DO I = II, min (II + 50, N)
7232 This optimization applies to all the languages supported by GCC and is
7233 not limited to Fortran. To use this code transformation, GCC has to
7234 be configured with @option{--with-ppl} and @option{--with-cloog} to
7235 enable the Graphite loop transformation infrastructure.
7238 @opindex floop-block
7239 Perform loop blocking transformations on loops. Blocking strip mines
7240 each loop in the loop nest such that the memory accesses of the
7241 element loops fit inside caches. The strip length can be changed
7242 using the @option{loop-block-tile-size} parameter. For example, given
7247 A(J, I) = B(I) + C(J)
7251 loop blocking will transform the loop as if the user had written:
7255 DO I = II, min (II + 50, N)
7256 DO J = JJ, min (JJ + 50, M)
7257 A(J, I) = B(I) + C(J)
7263 which can be beneficial when @code{M} is larger than the caches,
7264 because the innermost loop will iterate over a smaller amount of data
7265 which can be kept in the caches. This optimization applies to all the
7266 languages supported by GCC and is not limited to Fortran. To use this
7267 code transformation, GCC has to be configured with @option{--with-ppl}
7268 and @option{--with-cloog} to enable the Graphite loop transformation
7271 @item -fgraphite-identity
7272 @opindex fgraphite-identity
7273 Enable the identity transformation for graphite. For every SCoP we generate
7274 the polyhedral representation and transform it back to gimple. Using
7275 @option{-fgraphite-identity} we can check the costs or benefits of the
7276 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7277 are also performed by the code generator CLooG, like index splitting and
7278 dead code elimination in loops.
7280 @item -floop-flatten
7281 @opindex floop-flatten
7282 Removes the loop nesting structure: transforms the loop nest into a
7283 single loop. This transformation can be useful to vectorize all the
7284 levels of the loop nest.
7286 @item -floop-parallelize-all
7287 @opindex floop-parallelize-all
7288 Use the Graphite data dependence analysis to identify loops that can
7289 be parallelized. Parallelize all the loops that can be analyzed to
7290 not contain loop carried dependences without checking that it is
7291 profitable to parallelize the loops.
7293 @item -fcheck-data-deps
7294 @opindex fcheck-data-deps
7295 Compare the results of several data dependence analyzers. This option
7296 is used for debugging the data dependence analyzers.
7298 @item -ftree-loop-if-convert
7299 Attempt to transform conditional jumps in the innermost loops to
7300 branch-less equivalents. The intent is to remove control-flow from
7301 the innermost loops in order to improve the ability of the
7302 vectorization pass to handle these loops. This is enabled by default
7303 if vectorization is enabled.
7305 @item -ftree-loop-if-convert-stores
7306 Attempt to also if-convert conditional jumps containing memory writes.
7307 This transformation can be unsafe for multi-threaded programs as it
7308 transforms conditional memory writes into unconditional memory writes.
7311 for (i = 0; i < N; i++)
7315 would be transformed to
7317 for (i = 0; i < N; i++)
7318 A[i] = cond ? expr : A[i];
7320 potentially producing data races.
7322 @item -ftree-loop-distribution
7323 Perform loop distribution. This flag can improve cache performance on
7324 big loop bodies and allow further loop optimizations, like
7325 parallelization or vectorization, to take place. For example, the loop
7342 @item -ftree-loop-distribute-patterns
7343 Perform loop distribution of patterns that can be code generated with
7344 calls to a library. This flag is enabled by default at @option{-O3}.
7346 This pass distributes the initialization loops and generates a call to
7347 memset zero. For example, the loop
7363 and the initialization loop is transformed into a call to memset zero.
7365 @item -ftree-loop-im
7366 @opindex ftree-loop-im
7367 Perform loop invariant motion on trees. This pass moves only invariants that
7368 would be hard to handle at RTL level (function calls, operations that expand to
7369 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7370 operands of conditions that are invariant out of the loop, so that we can use
7371 just trivial invariantness analysis in loop unswitching. The pass also includes
7374 @item -ftree-loop-ivcanon
7375 @opindex ftree-loop-ivcanon
7376 Create a canonical counter for number of iterations in loops for which
7377 determining number of iterations requires complicated analysis. Later
7378 optimizations then may determine the number easily. Useful especially
7379 in connection with unrolling.
7383 Perform induction variable optimizations (strength reduction, induction
7384 variable merging and induction variable elimination) on trees.
7386 @item -ftree-parallelize-loops=n
7387 @opindex ftree-parallelize-loops
7388 Parallelize loops, i.e., split their iteration space to run in n threads.
7389 This is only possible for loops whose iterations are independent
7390 and can be arbitrarily reordered. The optimization is only
7391 profitable on multiprocessor machines, for loops that are CPU-intensive,
7392 rather than constrained e.g.@: by memory bandwidth. This option
7393 implies @option{-pthread}, and thus is only supported on targets
7394 that have support for @option{-pthread}.
7398 Perform function-local points-to analysis on trees. This flag is
7399 enabled by default at @option{-O} and higher.
7403 Perform scalar replacement of aggregates. This pass replaces structure
7404 references with scalars to prevent committing structures to memory too
7405 early. This flag is enabled by default at @option{-O} and higher.
7407 @item -ftree-copyrename
7408 @opindex ftree-copyrename
7409 Perform copy renaming on trees. This pass attempts to rename compiler
7410 temporaries to other variables at copy locations, usually resulting in
7411 variable names which more closely resemble the original variables. This flag
7412 is enabled by default at @option{-O} and higher.
7416 Perform temporary expression replacement during the SSA->normal phase. Single
7417 use/single def temporaries are replaced at their use location with their
7418 defining expression. This results in non-GIMPLE code, but gives the expanders
7419 much more complex trees to work on resulting in better RTL generation. This is
7420 enabled by default at @option{-O} and higher.
7422 @item -ftree-vectorize
7423 @opindex ftree-vectorize
7424 Perform loop vectorization on trees. This flag is enabled by default at
7427 @item -ftree-slp-vectorize
7428 @opindex ftree-slp-vectorize
7429 Perform basic block vectorization on trees. This flag is enabled by default at
7430 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7432 @item -ftree-vect-loop-version
7433 @opindex ftree-vect-loop-version
7434 Perform loop versioning when doing loop vectorization on trees. When a loop
7435 appears to be vectorizable except that data alignment or data dependence cannot
7436 be determined at compile time, then vectorized and non-vectorized versions of
7437 the loop are generated along with run-time checks for alignment or dependence
7438 to control which version is executed. This option is enabled by default
7439 except at level @option{-Os} where it is disabled.
7441 @item -fvect-cost-model
7442 @opindex fvect-cost-model
7443 Enable cost model for vectorization.
7447 Perform Value Range Propagation on trees. This is similar to the
7448 constant propagation pass, but instead of values, ranges of values are
7449 propagated. This allows the optimizers to remove unnecessary range
7450 checks like array bound checks and null pointer checks. This is
7451 enabled by default at @option{-O2} and higher. Null pointer check
7452 elimination is only done if @option{-fdelete-null-pointer-checks} is
7457 Perform tail duplication to enlarge superblock size. This transformation
7458 simplifies the control flow of the function allowing other optimizations to do
7461 @item -funroll-loops
7462 @opindex funroll-loops
7463 Unroll loops whose number of iterations can be determined at compile
7464 time or upon entry to the loop. @option{-funroll-loops} implies
7465 @option{-frerun-cse-after-loop}. This option makes code larger,
7466 and may or may not make it run faster.
7468 @item -funroll-all-loops
7469 @opindex funroll-all-loops
7470 Unroll all loops, even if their number of iterations is uncertain when
7471 the loop is entered. This usually makes programs run more slowly.
7472 @option{-funroll-all-loops} implies the same options as
7473 @option{-funroll-loops},
7475 @item -fsplit-ivs-in-unroller
7476 @opindex fsplit-ivs-in-unroller
7477 Enables expressing of values of induction variables in later iterations
7478 of the unrolled loop using the value in the first iteration. This breaks
7479 long dependency chains, thus improving efficiency of the scheduling passes.
7481 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7482 same effect. However in cases the loop body is more complicated than
7483 a single basic block, this is not reliable. It also does not work at all
7484 on some of the architectures due to restrictions in the CSE pass.
7486 This optimization is enabled by default.
7488 @item -fvariable-expansion-in-unroller
7489 @opindex fvariable-expansion-in-unroller
7490 With this option, the compiler will create multiple copies of some
7491 local variables when unrolling a loop which can result in superior code.
7493 @item -fpartial-inlining
7494 @opindex fpartial-inlining
7495 Inline parts of functions. This option has any effect only
7496 when inlining itself is turned on by the @option{-finline-functions}
7497 or @option{-finline-small-functions} options.
7499 Enabled at level @option{-O2}.
7501 @item -fpredictive-commoning
7502 @opindex fpredictive-commoning
7503 Perform predictive commoning optimization, i.e., reusing computations
7504 (especially memory loads and stores) performed in previous
7505 iterations of loops.
7507 This option is enabled at level @option{-O3}.
7509 @item -fprefetch-loop-arrays
7510 @opindex fprefetch-loop-arrays
7511 If supported by the target machine, generate instructions to prefetch
7512 memory to improve the performance of loops that access large arrays.
7514 This option may generate better or worse code; results are highly
7515 dependent on the structure of loops within the source code.
7517 Disabled at level @option{-Os}.
7520 @itemx -fno-peephole2
7521 @opindex fno-peephole
7522 @opindex fno-peephole2
7523 Disable any machine-specific peephole optimizations. The difference
7524 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7525 are implemented in the compiler; some targets use one, some use the
7526 other, a few use both.
7528 @option{-fpeephole} is enabled by default.
7529 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7531 @item -fno-guess-branch-probability
7532 @opindex fno-guess-branch-probability
7533 Do not guess branch probabilities using heuristics.
7535 GCC will use heuristics to guess branch probabilities if they are
7536 not provided by profiling feedback (@option{-fprofile-arcs}). These
7537 heuristics are based on the control flow graph. If some branch probabilities
7538 are specified by @samp{__builtin_expect}, then the heuristics will be
7539 used to guess branch probabilities for the rest of the control flow graph,
7540 taking the @samp{__builtin_expect} info into account. The interactions
7541 between the heuristics and @samp{__builtin_expect} can be complex, and in
7542 some cases, it may be useful to disable the heuristics so that the effects
7543 of @samp{__builtin_expect} are easier to understand.
7545 The default is @option{-fguess-branch-probability} at levels
7546 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7548 @item -freorder-blocks
7549 @opindex freorder-blocks
7550 Reorder basic blocks in the compiled function in order to reduce number of
7551 taken branches and improve code locality.
7553 Enabled at levels @option{-O2}, @option{-O3}.
7555 @item -freorder-blocks-and-partition
7556 @opindex freorder-blocks-and-partition
7557 In addition to reordering basic blocks in the compiled function, in order
7558 to reduce number of taken branches, partitions hot and cold basic blocks
7559 into separate sections of the assembly and .o files, to improve
7560 paging and cache locality performance.
7562 This optimization is automatically turned off in the presence of
7563 exception handling, for linkonce sections, for functions with a user-defined
7564 section attribute and on any architecture that does not support named
7567 @item -freorder-functions
7568 @opindex freorder-functions
7569 Reorder functions in the object file in order to
7570 improve code locality. This is implemented by using special
7571 subsections @code{.text.hot} for most frequently executed functions and
7572 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7573 the linker so object file format must support named sections and linker must
7574 place them in a reasonable way.
7576 Also profile feedback must be available in to make this option effective. See
7577 @option{-fprofile-arcs} for details.
7579 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7581 @item -fstrict-aliasing
7582 @opindex fstrict-aliasing
7583 Allow the compiler to assume the strictest aliasing rules applicable to
7584 the language being compiled. For C (and C++), this activates
7585 optimizations based on the type of expressions. In particular, an
7586 object of one type is assumed never to reside at the same address as an
7587 object of a different type, unless the types are almost the same. For
7588 example, an @code{unsigned int} can alias an @code{int}, but not a
7589 @code{void*} or a @code{double}. A character type may alias any other
7592 @anchor{Type-punning}Pay special attention to code like this:
7605 The practice of reading from a different union member than the one most
7606 recently written to (called ``type-punning'') is common. Even with
7607 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7608 is accessed through the union type. So, the code above will work as
7609 expected. @xref{Structures unions enumerations and bit-fields
7610 implementation}. However, this code might not:
7621 Similarly, access by taking the address, casting the resulting pointer
7622 and dereferencing the result has undefined behavior, even if the cast
7623 uses a union type, e.g.:
7627 return ((union a_union *) &d)->i;
7631 The @option{-fstrict-aliasing} option is enabled at levels
7632 @option{-O2}, @option{-O3}, @option{-Os}.
7634 @item -fstrict-overflow
7635 @opindex fstrict-overflow
7636 Allow the compiler to assume strict signed overflow rules, depending
7637 on the language being compiled. For C (and C++) this means that
7638 overflow when doing arithmetic with signed numbers is undefined, which
7639 means that the compiler may assume that it will not happen. This
7640 permits various optimizations. For example, the compiler will assume
7641 that an expression like @code{i + 10 > i} will always be true for
7642 signed @code{i}. This assumption is only valid if signed overflow is
7643 undefined, as the expression is false if @code{i + 10} overflows when
7644 using twos complement arithmetic. When this option is in effect any
7645 attempt to determine whether an operation on signed numbers will
7646 overflow must be written carefully to not actually involve overflow.
7648 This option also allows the compiler to assume strict pointer
7649 semantics: given a pointer to an object, if adding an offset to that
7650 pointer does not produce a pointer to the same object, the addition is
7651 undefined. This permits the compiler to conclude that @code{p + u >
7652 p} is always true for a pointer @code{p} and unsigned integer
7653 @code{u}. This assumption is only valid because pointer wraparound is
7654 undefined, as the expression is false if @code{p + u} overflows using
7655 twos complement arithmetic.
7657 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7658 that integer signed overflow is fully defined: it wraps. When
7659 @option{-fwrapv} is used, there is no difference between
7660 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7661 integers. With @option{-fwrapv} certain types of overflow are
7662 permitted. For example, if the compiler gets an overflow when doing
7663 arithmetic on constants, the overflowed value can still be used with
7664 @option{-fwrapv}, but not otherwise.
7666 The @option{-fstrict-overflow} option is enabled at levels
7667 @option{-O2}, @option{-O3}, @option{-Os}.
7669 @item -falign-functions
7670 @itemx -falign-functions=@var{n}
7671 @opindex falign-functions
7672 Align the start of functions to the next power-of-two greater than
7673 @var{n}, skipping up to @var{n} bytes. For instance,
7674 @option{-falign-functions=32} aligns functions to the next 32-byte
7675 boundary, but @option{-falign-functions=24} would align to the next
7676 32-byte boundary only if this can be done by skipping 23 bytes or less.
7678 @option{-fno-align-functions} and @option{-falign-functions=1} are
7679 equivalent and mean that functions will not be aligned.
7681 Some assemblers only support this flag when @var{n} is a power of two;
7682 in that case, it is rounded up.
7684 If @var{n} is not specified or is zero, use a machine-dependent default.
7686 Enabled at levels @option{-O2}, @option{-O3}.
7688 @item -falign-labels
7689 @itemx -falign-labels=@var{n}
7690 @opindex falign-labels
7691 Align all branch targets to a power-of-two boundary, skipping up to
7692 @var{n} bytes like @option{-falign-functions}. This option can easily
7693 make code slower, because it must insert dummy operations for when the
7694 branch target is reached in the usual flow of the code.
7696 @option{-fno-align-labels} and @option{-falign-labels=1} are
7697 equivalent and mean that labels will not be aligned.
7699 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7700 are greater than this value, then their values are used instead.
7702 If @var{n} is not specified or is zero, use a machine-dependent default
7703 which is very likely to be @samp{1}, meaning no alignment.
7705 Enabled at levels @option{-O2}, @option{-O3}.
7708 @itemx -falign-loops=@var{n}
7709 @opindex falign-loops
7710 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7711 like @option{-falign-functions}. The hope is that the loop will be
7712 executed many times, which will make up for any execution of the dummy
7715 @option{-fno-align-loops} and @option{-falign-loops=1} are
7716 equivalent and mean that loops will not be aligned.
7718 If @var{n} is not specified or is zero, use a machine-dependent default.
7720 Enabled at levels @option{-O2}, @option{-O3}.
7723 @itemx -falign-jumps=@var{n}
7724 @opindex falign-jumps
7725 Align branch targets to a power-of-two boundary, for branch targets
7726 where the targets can only be reached by jumping, skipping up to @var{n}
7727 bytes like @option{-falign-functions}. In this case, no dummy operations
7730 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7731 equivalent and mean that loops will not be aligned.
7733 If @var{n} is not specified or is zero, use a machine-dependent default.
7735 Enabled at levels @option{-O2}, @option{-O3}.
7737 @item -funit-at-a-time
7738 @opindex funit-at-a-time
7739 This option is left for compatibility reasons. @option{-funit-at-a-time}
7740 has no effect, while @option{-fno-unit-at-a-time} implies
7741 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7745 @item -fno-toplevel-reorder
7746 @opindex fno-toplevel-reorder
7747 Do not reorder top-level functions, variables, and @code{asm}
7748 statements. Output them in the same order that they appear in the
7749 input file. When this option is used, unreferenced static variables
7750 will not be removed. This option is intended to support existing code
7751 that relies on a particular ordering. For new code, it is better to
7754 Enabled at level @option{-O0}. When disabled explicitly, it also implies
7755 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
7760 Constructs webs as commonly used for register allocation purposes and assign
7761 each web individual pseudo register. This allows the register allocation pass
7762 to operate on pseudos directly, but also strengthens several other optimization
7763 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7764 however, make debugging impossible, since variables will no longer stay in a
7767 Enabled by default with @option{-funroll-loops}.
7769 @item -fwhole-program
7770 @opindex fwhole-program
7771 Assume that the current compilation unit represents the whole program being
7772 compiled. All public functions and variables with the exception of @code{main}
7773 and those merged by attribute @code{externally_visible} become static functions
7774 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.
7775 While this option is equivalent to proper use of the @code{static} keyword for
7776 programs consisting of a single file, in combination with option
7777 @option{-flto} this flag can be used to
7778 compile many smaller scale programs since the functions and variables become
7779 local for the whole combined compilation unit, not for the single source file
7782 This option implies @option{-fwhole-file} for Fortran programs.
7784 @item -flto[=@var{n}]
7786 This option runs the standard link-time optimizer. When invoked
7787 with source code, it generates GIMPLE (one of GCC's internal
7788 representations) and writes it to special ELF sections in the object
7789 file. When the object files are linked together, all the function
7790 bodies are read from these ELF sections and instantiated as if they
7791 had been part of the same translation unit.
7793 To use the link-time optimizer, @option{-flto} needs to be specified at
7794 compile time and during the final link. For example:
7797 gcc -c -O2 -flto foo.c
7798 gcc -c -O2 -flto bar.c
7799 gcc -o myprog -flto -O2 foo.o bar.o
7802 The first two invocations to GCC save a bytecode representation
7803 of GIMPLE into special ELF sections inside @file{foo.o} and
7804 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
7805 @file{foo.o} and @file{bar.o}, merges the two files into a single
7806 internal image, and compiles the result as usual. Since both
7807 @file{foo.o} and @file{bar.o} are merged into a single image, this
7808 causes all the interprocedural analyses and optimizations in GCC to
7809 work across the two files as if they were a single one. This means,
7810 for example, that the inliner is able to inline functions in
7811 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7813 Another (simpler) way to enable link-time optimization is:
7816 gcc -o myprog -flto -O2 foo.c bar.c
7819 The above generates bytecode for @file{foo.c} and @file{bar.c},
7820 merges them together into a single GIMPLE representation and optimizes
7821 them as usual to produce @file{myprog}.
7823 The only important thing to keep in mind is that to enable link-time
7824 optimizations the @option{-flto} flag needs to be passed to both the
7825 compile and the link commands.
7827 To make whole program optimization effective, it is necessary to make
7828 certain whole program assumptions. The compiler needs to know
7829 what functions and variables can be accessed by libraries and runtime
7830 outside of the link-time optimized unit. When supported by the linker,
7831 the linker plugin (see @option{-fuse-linker-plugin}) passes information
7832 to the compiler about used and externally visible symbols. When
7833 the linker plugin is not available, @option{-fwhole-program} should be
7834 used to allow the compiler to make these assumptions, which leads
7835 to more aggressive optimization decisions.
7837 Note that when a file is compiled with @option{-flto}, the generated
7838 object file is larger than a regular object file because it
7839 contains GIMPLE bytecodes and the usual final code. This means that
7840 object files with LTO information can be linked as normal object
7841 files; if @option{-flto} is not passed to the linker, no
7842 interprocedural optimizations are applied.
7844 Additionally, the optimization flags used to compile individual files
7845 are not necessarily related to those used at link time. For instance,
7848 gcc -c -O0 -flto foo.c
7849 gcc -c -O0 -flto bar.c
7850 gcc -o myprog -flto -O3 foo.o bar.o
7853 This produces individual object files with unoptimized assembler
7854 code, but the resulting binary @file{myprog} is optimized at
7855 @option{-O3}. If, instead, the final binary is generated without
7856 @option{-flto}, then @file{myprog} is not optimized.
7858 When producing the final binary with @option{-flto}, GCC only
7859 applies link-time optimizations to those files that contain bytecode.
7860 Therefore, you can mix and match object files and libraries with
7861 GIMPLE bytecodes and final object code. GCC automatically selects
7862 which files to optimize in LTO mode and which files to link without
7865 There are some code generation flags preserved by GCC when
7866 generating bytecodes, as they need to be used during the final link
7867 stage. Currently, the following options are saved into the GIMPLE
7868 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7869 @option{-m} target flags.
7871 At link time, these options are read in and reapplied. Note that the
7872 current implementation makes no attempt to recognize conflicting
7873 values for these options. If different files have conflicting option
7874 values (e.g., one file is compiled with @option{-fPIC} and another
7875 isn't), the compiler simply uses the last value read from the
7876 bytecode files. It is recommended, then, that you compile all the files
7877 participating in the same link with the same options.
7879 If LTO encounters objects with C linkage declared with incompatible
7880 types in separate translation units to be linked together (undefined
7881 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7882 issued. The behavior is still undefined at run time.
7884 Another feature of LTO is that it is possible to apply interprocedural
7885 optimizations on files written in different languages. This requires
7886 support in the language front end. Currently, the C, C++ and
7887 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7888 something like this should work:
7893 gfortran -c -flto baz.f90
7894 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7897 Notice that the final link is done with @command{g++} to get the C++
7898 runtime libraries and @option{-lgfortran} is added to get the Fortran
7899 runtime libraries. In general, when mixing languages in LTO mode, you
7900 should use the same link command options as when mixing languages in a
7901 regular (non-LTO) compilation; all you need to add is @option{-flto} to
7902 all the compile and link commands.
7904 If object files containing GIMPLE bytecode are stored in a library archive, say
7905 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7906 are using a linker with plugin support. To enable this feature, use
7907 the flag @option{-fuse-linker-plugin} at link time:
7910 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7913 With the linker plugin enabled, the linker extracts the needed
7914 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
7915 to make them part of the aggregated GIMPLE image to be optimized.
7917 If you are not using a linker with plugin support and/or do not
7918 enable the linker plugin, then the objects inside @file{libfoo.a}
7919 are extracted and linked as usual, but they do not participate
7920 in the LTO optimization process.
7922 Link-time optimizations do not require the presence of the whole program to
7923 operate. If the program does not require any symbols to be exported, it is
7924 possible to combine @option{-flto} and @option{-fwhole-program} to allow
7925 the interprocedural optimizers to use more aggressive assumptions which may
7926 lead to improved optimization opportunities.
7927 Use of @option{-fwhole-program} is not needed when linker plugin is
7928 active (see @option{-fuse-linker-plugin}).
7930 The current implementation of LTO makes no
7931 attempt to generate bytecode that is portable between different
7932 types of hosts. The bytecode files are versioned and there is a
7933 strict version check, so bytecode files generated in one version of
7934 GCC will not work with an older/newer version of GCC.
7936 Link-time optimization does not work well with generation of debugging
7937 information. Combining @option{-flto} with
7938 @option{-g} is currently experimental and expected to produce wrong
7941 If you specify the optional @var{n}, the optimization and code
7942 generation done at link time is executed in parallel using @var{n}
7943 parallel jobs by utilizing an installed @command{make} program. The
7944 environment variable @env{MAKE} may be used to override the program
7945 used. The default value for @var{n} is 1.
7947 You can also specify @option{-flto=jobserver} to use GNU make's
7948 job server mode to determine the number of parallel jobs. This
7949 is useful when the Makefile calling GCC is already executing in parallel.
7950 You must prepend a @samp{+} to the command recipe in the parent Makefile
7951 for this to work. This option likely only works if @env{MAKE} is
7954 This option is disabled by default
7956 @item -flto-partition=@var{alg}
7957 @opindex flto-partition
7958 Specify the partitioning algorithm used by the link-time optimizer.
7959 The value is either @code{1to1} to specify a partitioning mirroring
7960 the original source files or @code{balanced} to specify partitioning
7961 into equally sized chunks (whenever possible). Specifying @code{none}
7962 as an algorithm disables partitioning and streaming completely. The
7963 default value is @code{balanced}.
7965 @item -flto-compression-level=@var{n}
7966 This option specifies the level of compression used for intermediate
7967 language written to LTO object files, and is only meaningful in
7968 conjunction with LTO mode (@option{-flto}). Valid
7969 values are 0 (no compression) to 9 (maximum compression). Values
7970 outside this range are clamped to either 0 or 9. If the option is not
7971 given, a default balanced compression setting is used.
7974 Prints a report with internal details on the workings of the link-time
7975 optimizer. The contents of this report vary from version to version.
7976 It is meant to be useful to GCC developers when processing object
7977 files in LTO mode (via @option{-flto}).
7979 Disabled by default.
7981 @item -fuse-linker-plugin
7982 Enables the use of a linker plugin during link-time optimization. This
7983 option relies on plugin support in the linker, which is available in gold
7984 or in GNU ld 2.21 or newer.
7986 This option enables the extraction of object files with GIMPLE bytecode out
7987 of library archives. This improves the quality of optimization by exposing
7988 more code to the link-time optimizer. This information specifies what
7989 symbols can be accessed externally (by non-LTO object or during dynamic
7990 linking). Resulting code quality improvements on binaries (and shared
7991 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7992 See @option{-flto} for a description of the effect of this flag and how to
7995 This option is enabled by default when LTO support in GCC is enabled
7996 and GCC was configured for use with
7997 a linker supporting plugins (GNU ld 2.21 or newer or gold).
7999 @item -ffat-lto-objects
8000 @opindex ffat-lto-objects
8001 Fat LTO objects are object files that contain both the intermediate language
8002 and the object code. This makes them usable for both LTO linking and normal
8003 linking. This option is effective only when compiling with @option{-flto}
8004 and is ignored at link time.
8006 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8007 requires the complete toolchain to be aware of LTO. It requires a linker with
8008 linker plugin support for basic functionality. Additionally, nm, ar and ranlib
8009 need to support linker plugins to allow a full-featured build environment
8010 (capable of building static libraries etc).
8012 The default is @option{-ffat-lto-objects} but this default is intended to
8013 change in future releases when linker plugin enabled environments become more
8016 @item -fcompare-elim
8017 @opindex fcompare-elim
8018 After register allocation and post-register allocation instruction splitting,
8019 identify arithmetic instructions that compute processor flags similar to a
8020 comparison operation based on that arithmetic. If possible, eliminate the
8021 explicit comparison operation.
8023 This pass only applies to certain targets that cannot explicitly represent
8024 the comparison operation before register allocation is complete.
8026 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8028 @item -fcprop-registers
8029 @opindex fcprop-registers
8030 After register allocation and post-register allocation instruction splitting,
8031 we perform a copy-propagation pass to try to reduce scheduling dependencies
8032 and occasionally eliminate the copy.
8034 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8036 @item -fprofile-correction
8037 @opindex fprofile-correction
8038 Profiles collected using an instrumented binary for multi-threaded programs may
8039 be inconsistent due to missed counter updates. When this option is specified,
8040 GCC will use heuristics to correct or smooth out such inconsistencies. By
8041 default, GCC will emit an error message when an inconsistent profile is detected.
8043 @item -fprofile-dir=@var{path}
8044 @opindex fprofile-dir
8046 Set the directory to search for the profile data files in to @var{path}.
8047 This option affects only the profile data generated by
8048 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8049 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8050 and its related options. Both absolute and relative paths can be used.
8051 By default, GCC will use the current directory as @var{path}, thus the
8052 profile data file will appear in the same directory as the object file.
8054 @item -fprofile-generate
8055 @itemx -fprofile-generate=@var{path}
8056 @opindex fprofile-generate
8058 Enable options usually used for instrumenting application to produce
8059 profile useful for later recompilation with profile feedback based
8060 optimization. You must use @option{-fprofile-generate} both when
8061 compiling and when linking your program.
8063 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
8065 If @var{path} is specified, GCC will look at the @var{path} to find
8066 the profile feedback data files. See @option{-fprofile-dir}.
8069 @itemx -fprofile-use=@var{path}
8070 @opindex fprofile-use
8071 Enable profile feedback directed optimizations, and optimizations
8072 generally profitable only with profile feedback available.
8074 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8075 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
8077 By default, GCC emits an error message if the feedback profiles do not
8078 match the source code. This error can be turned into a warning by using
8079 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8082 If @var{path} is specified, GCC will look at the @var{path} to find
8083 the profile feedback data files. See @option{-fprofile-dir}.
8086 The following options control compiler behavior regarding floating-point
8087 arithmetic. These options trade off between speed and
8088 correctness. All must be specifically enabled.
8092 @opindex ffloat-store
8093 Do not store floating-point variables in registers, and inhibit other
8094 options that might change whether a floating-point value is taken from a
8097 @cindex floating-point precision
8098 This option prevents undesirable excess precision on machines such as
8099 the 68000 where the floating registers (of the 68881) keep more
8100 precision than a @code{double} is supposed to have. Similarly for the
8101 x86 architecture. For most programs, the excess precision does only
8102 good, but a few programs rely on the precise definition of IEEE floating
8103 point. Use @option{-ffloat-store} for such programs, after modifying
8104 them to store all pertinent intermediate computations into variables.
8106 @item -fexcess-precision=@var{style}
8107 @opindex fexcess-precision
8108 This option allows further control over excess precision on machines
8109 where floating-point registers have more precision than the IEEE
8110 @code{float} and @code{double} types and the processor does not
8111 support operations rounding to those types. By default,
8112 @option{-fexcess-precision=fast} is in effect; this means that
8113 operations are carried out in the precision of the registers and that
8114 it is unpredictable when rounding to the types specified in the source
8115 code takes place. When compiling C, if
8116 @option{-fexcess-precision=standard} is specified then excess
8117 precision will follow the rules specified in ISO C99; in particular,
8118 both casts and assignments cause values to be rounded to their
8119 semantic types (whereas @option{-ffloat-store} only affects
8120 assignments). This option is enabled by default for C if a strict
8121 conformance option such as @option{-std=c99} is used.
8124 @option{-fexcess-precision=standard} is not implemented for languages
8125 other than C, and has no effect if
8126 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8127 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8128 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8129 semantics apply without excess precision, and in the latter, rounding
8134 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8135 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8136 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8138 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8140 This option is not turned on by any @option{-O} option besides
8141 @option{-Ofast} since it can result in incorrect output for programs
8142 that depend on an exact implementation of IEEE or ISO rules/specifications
8143 for math functions. It may, however, yield faster code for programs
8144 that do not require the guarantees of these specifications.
8146 @item -fno-math-errno
8147 @opindex fno-math-errno
8148 Do not set ERRNO after calling math functions that are executed
8149 with a single instruction, e.g., sqrt. A program that relies on
8150 IEEE exceptions for math error handling may want to use this flag
8151 for speed while maintaining IEEE arithmetic compatibility.
8153 This option is not turned on by any @option{-O} option since
8154 it can result in incorrect output for programs that depend on
8155 an exact implementation of IEEE or ISO rules/specifications for
8156 math functions. It may, however, yield faster code for programs
8157 that do not require the guarantees of these specifications.
8159 The default is @option{-fmath-errno}.
8161 On Darwin systems, the math library never sets @code{errno}. There is
8162 therefore no reason for the compiler to consider the possibility that
8163 it might, and @option{-fno-math-errno} is the default.
8165 @item -funsafe-math-optimizations
8166 @opindex funsafe-math-optimizations
8168 Allow optimizations for floating-point arithmetic that (a) assume
8169 that arguments and results are valid and (b) may violate IEEE or
8170 ANSI standards. When used at link-time, it may include libraries
8171 or startup files that change the default FPU control word or other
8172 similar optimizations.
8174 This option is not turned on by any @option{-O} option since
8175 it can result in incorrect output for programs that depend on
8176 an exact implementation of IEEE or ISO rules/specifications for
8177 math functions. It may, however, yield faster code for programs
8178 that do not require the guarantees of these specifications.
8179 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8180 @option{-fassociative-math} and @option{-freciprocal-math}.
8182 The default is @option{-fno-unsafe-math-optimizations}.
8184 @item -fassociative-math
8185 @opindex fassociative-math
8187 Allow re-association of operands in series of floating-point operations.
8188 This violates the ISO C and C++ language standard by possibly changing
8189 computation result. NOTE: re-ordering may change the sign of zero as
8190 well as ignore NaNs and inhibit or create underflow or overflow (and
8191 thus cannot be used on code that relies on rounding behavior like
8192 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8193 and thus may not be used when ordered comparisons are required.
8194 This option requires that both @option{-fno-signed-zeros} and
8195 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8196 much sense with @option{-frounding-math}. For Fortran the option
8197 is automatically enabled when both @option{-fno-signed-zeros} and
8198 @option{-fno-trapping-math} are in effect.
8200 The default is @option{-fno-associative-math}.
8202 @item -freciprocal-math
8203 @opindex freciprocal-math
8205 Allow the reciprocal of a value to be used instead of dividing by
8206 the value if this enables optimizations. For example @code{x / y}
8207 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
8208 is subject to common subexpression elimination. Note that this loses
8209 precision and increases the number of flops operating on the value.
8211 The default is @option{-fno-reciprocal-math}.
8213 @item -ffinite-math-only
8214 @opindex ffinite-math-only
8215 Allow optimizations for floating-point arithmetic that assume
8216 that arguments and results are not NaNs or +-Infs.
8218 This option is not turned on by any @option{-O} option since
8219 it can result in incorrect output for programs that depend on
8220 an exact implementation of IEEE or ISO rules/specifications for
8221 math functions. It may, however, yield faster code for programs
8222 that do not require the guarantees of these specifications.
8224 The default is @option{-fno-finite-math-only}.
8226 @item -fno-signed-zeros
8227 @opindex fno-signed-zeros
8228 Allow optimizations for floating-point arithmetic that ignore the
8229 signedness of zero. IEEE arithmetic specifies the behavior of
8230 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8231 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8232 This option implies that the sign of a zero result isn't significant.
8234 The default is @option{-fsigned-zeros}.
8236 @item -fno-trapping-math
8237 @opindex fno-trapping-math
8238 Compile code assuming that floating-point operations cannot generate
8239 user-visible traps. These traps include division by zero, overflow,
8240 underflow, inexact result and invalid operation. This option requires
8241 that @option{-fno-signaling-nans} be in effect. Setting this option may
8242 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8244 This option should never be turned on by any @option{-O} option since
8245 it can result in incorrect output for programs that depend on
8246 an exact implementation of IEEE or ISO rules/specifications for
8249 The default is @option{-ftrapping-math}.
8251 @item -frounding-math
8252 @opindex frounding-math
8253 Disable transformations and optimizations that assume default floating-point
8254 rounding behavior. This is round-to-zero for all floating point
8255 to integer conversions, and round-to-nearest for all other arithmetic
8256 truncations. This option should be specified for programs that change
8257 the FP rounding mode dynamically, or that may be executed with a
8258 non-default rounding mode. This option disables constant folding of
8259 floating-point expressions at compile time (which may be affected by
8260 rounding mode) and arithmetic transformations that are unsafe in the
8261 presence of sign-dependent rounding modes.
8263 The default is @option{-fno-rounding-math}.
8265 This option is experimental and does not currently guarantee to
8266 disable all GCC optimizations that are affected by rounding mode.
8267 Future versions of GCC may provide finer control of this setting
8268 using C99's @code{FENV_ACCESS} pragma. This command-line option
8269 will be used to specify the default state for @code{FENV_ACCESS}.
8271 @item -fsignaling-nans
8272 @opindex fsignaling-nans
8273 Compile code assuming that IEEE signaling NaNs may generate user-visible
8274 traps during floating-point operations. Setting this option disables
8275 optimizations that may change the number of exceptions visible with
8276 signaling NaNs. This option implies @option{-ftrapping-math}.
8278 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8281 The default is @option{-fno-signaling-nans}.
8283 This option is experimental and does not currently guarantee to
8284 disable all GCC optimizations that affect signaling NaN behavior.
8286 @item -fsingle-precision-constant
8287 @opindex fsingle-precision-constant
8288 Treat floating-point constants as single precision instead of
8289 implicitly converting them to double-precision constants.
8291 @item -fcx-limited-range
8292 @opindex fcx-limited-range
8293 When enabled, this option states that a range reduction step is not
8294 needed when performing complex division. Also, there is no checking
8295 whether the result of a complex multiplication or division is @code{NaN
8296 + I*NaN}, with an attempt to rescue the situation in that case. The
8297 default is @option{-fno-cx-limited-range}, but is enabled by
8298 @option{-ffast-math}.
8300 This option controls the default setting of the ISO C99
8301 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8304 @item -fcx-fortran-rules
8305 @opindex fcx-fortran-rules
8306 Complex multiplication and division follow Fortran rules. Range
8307 reduction is done as part of complex division, but there is no checking
8308 whether the result of a complex multiplication or division is @code{NaN
8309 + I*NaN}, with an attempt to rescue the situation in that case.
8311 The default is @option{-fno-cx-fortran-rules}.
8315 The following options control optimizations that may improve
8316 performance, but are not enabled by any @option{-O} options. This
8317 section includes experimental options that may produce broken code.
8320 @item -fbranch-probabilities
8321 @opindex fbranch-probabilities
8322 After running a program compiled with @option{-fprofile-arcs}
8323 (@pxref{Debugging Options,, Options for Debugging Your Program or
8324 @command{gcc}}), you can compile it a second time using
8325 @option{-fbranch-probabilities}, to improve optimizations based on
8326 the number of times each branch was taken. When the program
8327 compiled with @option{-fprofile-arcs} exits it saves arc execution
8328 counts to a file called @file{@var{sourcename}.gcda} for each source
8329 file. The information in this data file is very dependent on the
8330 structure of the generated code, so you must use the same source code
8331 and the same optimization options for both compilations.
8333 With @option{-fbranch-probabilities}, GCC puts a
8334 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8335 These can be used to improve optimization. Currently, they are only
8336 used in one place: in @file{reorg.c}, instead of guessing which path a
8337 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8338 exactly determine which path is taken more often.
8340 @item -fprofile-values
8341 @opindex fprofile-values
8342 If combined with @option{-fprofile-arcs}, it adds code so that some
8343 data about values of expressions in the program is gathered.
8345 With @option{-fbranch-probabilities}, it reads back the data gathered
8346 from profiling values of expressions for usage in optimizations.
8348 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8352 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8353 a code to gather information about values of expressions.
8355 With @option{-fbranch-probabilities}, it reads back the data gathered
8356 and actually performs the optimizations based on them.
8357 Currently the optimizations include specialization of division operation
8358 using the knowledge about the value of the denominator.
8360 @item -frename-registers
8361 @opindex frename-registers
8362 Attempt to avoid false dependencies in scheduled code by making use
8363 of registers left over after register allocation. This optimization
8364 will most benefit processors with lots of registers. Depending on the
8365 debug information format adopted by the target, however, it can
8366 make debugging impossible, since variables will no longer stay in
8367 a ``home register''.
8369 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8373 Perform tail duplication to enlarge superblock size. This transformation
8374 simplifies the control flow of the function allowing other optimizations to do
8377 Enabled with @option{-fprofile-use}.
8379 @item -funroll-loops
8380 @opindex funroll-loops
8381 Unroll loops whose number of iterations can be determined at compile time or
8382 upon entry to the loop. @option{-funroll-loops} implies
8383 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8384 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8385 small constant number of iterations). This option makes code larger, and may
8386 or may not make it run faster.
8388 Enabled with @option{-fprofile-use}.
8390 @item -funroll-all-loops
8391 @opindex funroll-all-loops
8392 Unroll all loops, even if their number of iterations is uncertain when
8393 the loop is entered. This usually makes programs run more slowly.
8394 @option{-funroll-all-loops} implies the same options as
8395 @option{-funroll-loops}.
8398 @opindex fpeel-loops
8399 Peels loops for which there is enough information that they do not
8400 roll much (from profile feedback). It also turns on complete loop peeling
8401 (i.e.@: complete removal of loops with small constant number of iterations).
8403 Enabled with @option{-fprofile-use}.
8405 @item -fmove-loop-invariants
8406 @opindex fmove-loop-invariants
8407 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8408 at level @option{-O1}
8410 @item -funswitch-loops
8411 @opindex funswitch-loops
8412 Move branches with loop invariant conditions out of the loop, with duplicates
8413 of the loop on both branches (modified according to result of the condition).
8415 @item -ffunction-sections
8416 @itemx -fdata-sections
8417 @opindex ffunction-sections
8418 @opindex fdata-sections
8419 Place each function or data item into its own section in the output
8420 file if the target supports arbitrary sections. The name of the
8421 function or the name of the data item determines the section's name
8424 Use these options on systems where the linker can perform optimizations
8425 to improve locality of reference in the instruction space. Most systems
8426 using the ELF object format and SPARC processors running Solaris 2 have
8427 linkers with such optimizations. AIX may have these optimizations in
8430 Only use these options when there are significant benefits from doing
8431 so. When you specify these options, the assembler and linker will
8432 create larger object and executable files and will also be slower.
8433 You will not be able to use @code{gprof} on all systems if you
8434 specify this option and you may have problems with debugging if
8435 you specify both this option and @option{-g}.
8437 @item -fbranch-target-load-optimize
8438 @opindex fbranch-target-load-optimize
8439 Perform branch target register load optimization before prologue / epilogue
8441 The use of target registers can typically be exposed only during reload,
8442 thus hoisting loads out of loops and doing inter-block scheduling needs
8443 a separate optimization pass.
8445 @item -fbranch-target-load-optimize2
8446 @opindex fbranch-target-load-optimize2
8447 Perform branch target register load optimization after prologue / epilogue
8450 @item -fbtr-bb-exclusive
8451 @opindex fbtr-bb-exclusive
8452 When performing branch target register load optimization, don't reuse
8453 branch target registers in within any basic block.
8455 @item -fstack-protector
8456 @opindex fstack-protector
8457 Emit extra code to check for buffer overflows, such as stack smashing
8458 attacks. This is done by adding a guard variable to functions with
8459 vulnerable objects. This includes functions that call alloca, and
8460 functions with buffers larger than 8 bytes. The guards are initialized
8461 when a function is entered and then checked when the function exits.
8462 If a guard check fails, an error message is printed and the program exits.
8464 @item -fstack-protector-all
8465 @opindex fstack-protector-all
8466 Like @option{-fstack-protector} except that all functions are protected.
8468 @item -fsection-anchors
8469 @opindex fsection-anchors
8470 Try to reduce the number of symbolic address calculations by using
8471 shared ``anchor'' symbols to address nearby objects. This transformation
8472 can help to reduce the number of GOT entries and GOT accesses on some
8475 For example, the implementation of the following function @code{foo}:
8479 int foo (void) @{ return a + b + c; @}
8482 would usually calculate the addresses of all three variables, but if you
8483 compile it with @option{-fsection-anchors}, it will access the variables
8484 from a common anchor point instead. The effect is similar to the
8485 following pseudocode (which isn't valid C):
8490 register int *xr = &x;
8491 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8495 Not all targets support this option.
8497 @item --param @var{name}=@var{value}
8499 In some places, GCC uses various constants to control the amount of
8500 optimization that is done. For example, GCC will not inline functions
8501 that contain more than a certain number of instructions. You can
8502 control some of these constants on the command line using the
8503 @option{--param} option.
8505 The names of specific parameters, and the meaning of the values, are
8506 tied to the internals of the compiler, and are subject to change
8507 without notice in future releases.
8509 In each case, the @var{value} is an integer. The allowable choices for
8510 @var{name} are given in the following table:
8513 @item predictable-branch-outcome
8514 When branch is predicted to be taken with probability lower than this threshold
8515 (in percent), then it is considered well predictable. The default is 10.
8517 @item max-crossjump-edges
8518 The maximum number of incoming edges to consider for crossjumping.
8519 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8520 the number of edges incoming to each block. Increasing values mean
8521 more aggressive optimization, making the compilation time increase with
8522 probably small improvement in executable size.
8524 @item min-crossjump-insns
8525 The minimum number of instructions that must be matched at the end
8526 of two blocks before crossjumping will be performed on them. This
8527 value is ignored in the case where all instructions in the block being
8528 crossjumped from are matched. The default value is 5.
8530 @item max-grow-copy-bb-insns
8531 The maximum code size expansion factor when copying basic blocks
8532 instead of jumping. The expansion is relative to a jump instruction.
8533 The default value is 8.
8535 @item max-goto-duplication-insns
8536 The maximum number of instructions to duplicate to a block that jumps
8537 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8538 passes, GCC factors computed gotos early in the compilation process,
8539 and unfactors them as late as possible. Only computed jumps at the
8540 end of a basic blocks with no more than max-goto-duplication-insns are
8541 unfactored. The default value is 8.
8543 @item max-delay-slot-insn-search
8544 The maximum number of instructions to consider when looking for an
8545 instruction to fill a delay slot. If more than this arbitrary number of
8546 instructions is searched, the time savings from filling the delay slot
8547 will be minimal so stop searching. Increasing values mean more
8548 aggressive optimization, making the compilation time increase with probably
8549 small improvement in execution time.
8551 @item max-delay-slot-live-search
8552 When trying to fill delay slots, the maximum number of instructions to
8553 consider when searching for a block with valid live register
8554 information. Increasing this arbitrarily chosen value means more
8555 aggressive optimization, increasing the compilation time. This parameter
8556 should be removed when the delay slot code is rewritten to maintain the
8559 @item max-gcse-memory
8560 The approximate maximum amount of memory that will be allocated in
8561 order to perform the global common subexpression elimination
8562 optimization. If more memory than specified is required, the
8563 optimization will not be done.
8565 @item max-gcse-insertion-ratio
8566 If the ratio of expression insertions to deletions is larger than this value
8567 for any expression, then RTL PRE will insert or remove the expression and thus
8568 leave partially redundant computations in the instruction stream. The default value is 20.
8570 @item max-pending-list-length
8571 The maximum number of pending dependencies scheduling will allow
8572 before flushing the current state and starting over. Large functions
8573 with few branches or calls can create excessively large lists which
8574 needlessly consume memory and resources.
8576 @item max-modulo-backtrack-attempts
8577 The maximum number of backtrack attempts the scheduler should make
8578 when modulo scheduling a loop. Larger values can exponentially increase
8581 @item max-inline-insns-single
8582 Several parameters control the tree inliner used in gcc.
8583 This number sets the maximum number of instructions (counted in GCC's
8584 internal representation) in a single function that the tree inliner
8585 will consider for inlining. This only affects functions declared
8586 inline and methods implemented in a class declaration (C++).
8587 The default value is 400.
8589 @item max-inline-insns-auto
8590 When you use @option{-finline-functions} (included in @option{-O3}),
8591 a lot of functions that would otherwise not be considered for inlining
8592 by the compiler will be investigated. To those functions, a different
8593 (more restrictive) limit compared to functions declared inline can
8595 The default value is 40.
8597 @item large-function-insns
8598 The limit specifying really large functions. For functions larger than this
8599 limit after inlining, inlining is constrained by
8600 @option{--param large-function-growth}. This parameter is useful primarily
8601 to avoid extreme compilation time caused by non-linear algorithms used by the
8603 The default value is 2700.
8605 @item large-function-growth
8606 Specifies maximal growth of large function caused by inlining in percents.
8607 The default value is 100 which limits large function growth to 2.0 times
8610 @item large-unit-insns
8611 The limit specifying large translation unit. Growth caused by inlining of
8612 units larger than this limit is limited by @option{--param inline-unit-growth}.
8613 For small units this might be too tight (consider unit consisting of function A
8614 that is inline and B that just calls A three time. If B is small relative to
8615 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8616 large units consisting of small inlineable functions however the overall unit
8617 growth limit is needed to avoid exponential explosion of code size. Thus for
8618 smaller units, the size is increased to @option{--param large-unit-insns}
8619 before applying @option{--param inline-unit-growth}. The default is 10000
8621 @item inline-unit-growth
8622 Specifies maximal overall growth of the compilation unit caused by inlining.
8623 The default value is 30 which limits unit growth to 1.3 times the original
8626 @item ipcp-unit-growth
8627 Specifies maximal overall growth of the compilation unit caused by
8628 interprocedural constant propagation. The default value is 10 which limits
8629 unit growth to 1.1 times the original size.
8631 @item large-stack-frame
8632 The limit specifying large stack frames. While inlining the algorithm is trying
8633 to not grow past this limit too much. Default value is 256 bytes.
8635 @item large-stack-frame-growth
8636 Specifies maximal growth of large stack frames caused by inlining in percents.
8637 The default value is 1000 which limits large stack frame growth to 11 times
8640 @item max-inline-insns-recursive
8641 @itemx max-inline-insns-recursive-auto
8642 Specifies maximum number of instructions out-of-line copy of self recursive inline
8643 function can grow into by performing recursive inlining.
8645 For functions declared inline @option{--param max-inline-insns-recursive} is
8646 taken into account. For function not declared inline, recursive inlining
8647 happens only when @option{-finline-functions} (included in @option{-O3}) is
8648 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8649 default value is 450.
8651 @item max-inline-recursive-depth
8652 @itemx max-inline-recursive-depth-auto
8653 Specifies maximum recursion depth used by the recursive inlining.
8655 For functions declared inline @option{--param max-inline-recursive-depth} is
8656 taken into account. For function not declared inline, recursive inlining
8657 happens only when @option{-finline-functions} (included in @option{-O3}) is
8658 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8661 @item min-inline-recursive-probability
8662 Recursive inlining is profitable only for function having deep recursion
8663 in average and can hurt for function having little recursion depth by
8664 increasing the prologue size or complexity of function body to other
8667 When profile feedback is available (see @option{-fprofile-generate}) the actual
8668 recursion depth can be guessed from probability that function will recurse via
8669 given call expression. This parameter limits inlining only to call expression
8670 whose probability exceeds given threshold (in percents). The default value is
8673 @item early-inlining-insns
8674 Specify growth that early inliner can make. In effect it increases amount of
8675 inlining for code having large abstraction penalty. The default value is 10.
8677 @item max-early-inliner-iterations
8678 @itemx max-early-inliner-iterations
8679 Limit of iterations of early inliner. This basically bounds number of nested
8680 indirect calls early inliner can resolve. Deeper chains are still handled by
8683 @item comdat-sharing-probability
8684 @itemx comdat-sharing-probability
8685 Probability (in percent) that C++ inline function with comdat visibility
8686 will be shared across multiple compilation units. The default value is 20.
8688 @item min-vect-loop-bound
8689 The minimum number of iterations under which a loop will not get vectorized
8690 when @option{-ftree-vectorize} is used. The number of iterations after
8691 vectorization needs to be greater than the value specified by this option
8692 to allow vectorization. The default value is 0.
8694 @item gcse-cost-distance-ratio
8695 Scaling factor in calculation of maximum distance an expression
8696 can be moved by GCSE optimizations. This is currently supported only in the
8697 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8698 will be with simple expressions, i.e., the expressions that have cost
8699 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8700 hoisting of simple expressions. The default value is 10.
8702 @item gcse-unrestricted-cost
8703 Cost, roughly measured as the cost of a single typical machine
8704 instruction, at which GCSE optimizations will not constrain
8705 the distance an expression can travel. This is currently
8706 supported only in the code hoisting pass. The lesser the cost,
8707 the more aggressive code hoisting will be. Specifying 0 will
8708 allow all expressions to travel unrestricted distances.
8709 The default value is 3.
8711 @item max-hoist-depth
8712 The depth of search in the dominator tree for expressions to hoist.
8713 This is used to avoid quadratic behavior in hoisting algorithm.
8714 The value of 0 will avoid limiting the search, but may slow down compilation
8715 of huge functions. The default value is 30.
8717 @item max-tail-merge-comparisons
8718 The maximum amount of similar bbs to compare a bb with. This is used to
8719 avoid quadratic behaviour in tree tail merging. The default value is 10.
8721 @item max-tail-merge-iterations
8722 The maximum amount of iterations of the pass over the function. This is used to
8723 limit compilation time in tree tail merging. The default value is 2.
8725 @item max-unrolled-insns
8726 The maximum number of instructions that a loop should have if that loop
8727 is unrolled, and if the loop is unrolled, it determines how many times
8728 the loop code is unrolled.
8730 @item max-average-unrolled-insns
8731 The maximum number of instructions biased by probabilities of their execution
8732 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8733 it determines how many times the loop code is unrolled.
8735 @item max-unroll-times
8736 The maximum number of unrollings of a single loop.
8738 @item max-peeled-insns
8739 The maximum number of instructions that a loop should have if that loop
8740 is peeled, and if the loop is peeled, it determines how many times
8741 the loop code is peeled.
8743 @item max-peel-times
8744 The maximum number of peelings of a single loop.
8746 @item max-completely-peeled-insns
8747 The maximum number of insns of a completely peeled loop.
8749 @item max-completely-peel-times
8750 The maximum number of iterations of a loop to be suitable for complete peeling.
8752 @item max-completely-peel-loop-nest-depth
8753 The maximum depth of a loop nest suitable for complete peeling.
8755 @item max-unswitch-insns
8756 The maximum number of insns of an unswitched loop.
8758 @item max-unswitch-level
8759 The maximum number of branches unswitched in a single loop.
8762 The minimum cost of an expensive expression in the loop invariant motion.
8764 @item iv-consider-all-candidates-bound
8765 Bound on number of candidates for induction variables below that
8766 all candidates are considered for each use in induction variable
8767 optimizations. Only the most relevant candidates are considered
8768 if there are more candidates, to avoid quadratic time complexity.
8770 @item iv-max-considered-uses
8771 The induction variable optimizations give up on loops that contain more
8772 induction variable uses.
8774 @item iv-always-prune-cand-set-bound
8775 If number of candidates in the set is smaller than this value,
8776 we always try to remove unnecessary ivs from the set during its
8777 optimization when a new iv is added to the set.
8779 @item scev-max-expr-size
8780 Bound on size of expressions used in the scalar evolutions analyzer.
8781 Large expressions slow the analyzer.
8783 @item scev-max-expr-complexity
8784 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8785 Complex expressions slow the analyzer.
8787 @item omega-max-vars
8788 The maximum number of variables in an Omega constraint system.
8789 The default value is 128.
8791 @item omega-max-geqs
8792 The maximum number of inequalities in an Omega constraint system.
8793 The default value is 256.
8796 The maximum number of equalities in an Omega constraint system.
8797 The default value is 128.
8799 @item omega-max-wild-cards
8800 The maximum number of wildcard variables that the Omega solver will
8801 be able to insert. The default value is 18.
8803 @item omega-hash-table-size
8804 The size of the hash table in the Omega solver. The default value is
8807 @item omega-max-keys
8808 The maximal number of keys used by the Omega solver. The default
8811 @item omega-eliminate-redundant-constraints
8812 When set to 1, use expensive methods to eliminate all redundant
8813 constraints. The default value is 0.
8815 @item vect-max-version-for-alignment-checks
8816 The maximum number of run-time checks that can be performed when
8817 doing loop versioning for alignment in the vectorizer. See option
8818 ftree-vect-loop-version for more information.
8820 @item vect-max-version-for-alias-checks
8821 The maximum number of run-time checks that can be performed when
8822 doing loop versioning for alias in the vectorizer. See option
8823 ftree-vect-loop-version for more information.
8825 @item max-iterations-to-track
8827 The maximum number of iterations of a loop the brute force algorithm
8828 for analysis of # of iterations of the loop tries to evaluate.
8830 @item hot-bb-count-fraction
8831 Select fraction of the maximal count of repetitions of basic block in program
8832 given basic block needs to have to be considered hot.
8834 @item hot-bb-frequency-fraction
8835 Select fraction of the entry block frequency of executions of basic block in
8836 function given basic block needs to have to be considered hot.
8838 @item max-predicted-iterations
8839 The maximum number of loop iterations we predict statically. This is useful
8840 in cases where function contain single loop with known bound and other loop
8841 with unknown. We predict the known number of iterations correctly, while
8842 the unknown number of iterations average to roughly 10. This means that the
8843 loop without bounds would appear artificially cold relative to the other one.
8845 @item align-threshold
8847 Select fraction of the maximal frequency of executions of basic block in
8848 function given basic block will get aligned.
8850 @item align-loop-iterations
8852 A loop expected to iterate at lest the selected number of iterations will get
8855 @item tracer-dynamic-coverage
8856 @itemx tracer-dynamic-coverage-feedback
8858 This value is used to limit superblock formation once the given percentage of
8859 executed instructions is covered. This limits unnecessary code size
8862 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8863 feedback is available. The real profiles (as opposed to statically estimated
8864 ones) are much less balanced allowing the threshold to be larger value.
8866 @item tracer-max-code-growth
8867 Stop tail duplication once code growth has reached given percentage. This is
8868 rather hokey argument, as most of the duplicates will be eliminated later in
8869 cross jumping, so it may be set to much higher values than is the desired code
8872 @item tracer-min-branch-ratio
8874 Stop reverse growth when the reverse probability of best edge is less than this
8875 threshold (in percent).
8877 @item tracer-min-branch-ratio
8878 @itemx tracer-min-branch-ratio-feedback
8880 Stop forward growth if the best edge do have probability lower than this
8883 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8884 compilation for profile feedback and one for compilation without. The value
8885 for compilation with profile feedback needs to be more conservative (higher) in
8886 order to make tracer effective.
8888 @item max-cse-path-length
8890 Maximum number of basic blocks on path that cse considers. The default is 10.
8893 The maximum instructions CSE process before flushing. The default is 1000.
8895 @item ggc-min-expand
8897 GCC uses a garbage collector to manage its own memory allocation. This
8898 parameter specifies the minimum percentage by which the garbage
8899 collector's heap should be allowed to expand between collections.
8900 Tuning this may improve compilation speed; it has no effect on code
8903 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8904 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8905 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8906 GCC is not able to calculate RAM on a particular platform, the lower
8907 bound of 30% is used. Setting this parameter and
8908 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8909 every opportunity. This is extremely slow, but can be useful for
8912 @item ggc-min-heapsize
8914 Minimum size of the garbage collector's heap before it begins bothering
8915 to collect garbage. The first collection occurs after the heap expands
8916 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8917 tuning this may improve compilation speed, and has no effect on code
8920 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
8921 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8922 with a lower bound of 4096 (four megabytes) and an upper bound of
8923 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8924 particular platform, the lower bound is used. Setting this parameter
8925 very large effectively disables garbage collection. Setting this
8926 parameter and @option{ggc-min-expand} to zero causes a full collection
8927 to occur at every opportunity.
8929 @item max-reload-search-insns
8930 The maximum number of instruction reload should look backward for equivalent
8931 register. Increasing values mean more aggressive optimization, making the
8932 compilation time increase with probably slightly better performance.
8933 The default value is 100.
8935 @item max-cselib-memory-locations
8936 The maximum number of memory locations cselib should take into account.
8937 Increasing values mean more aggressive optimization, making the compilation time
8938 increase with probably slightly better performance. The default value is 500.
8940 @item reorder-blocks-duplicate
8941 @itemx reorder-blocks-duplicate-feedback
8943 Used by basic block reordering pass to decide whether to use unconditional
8944 branch or duplicate the code on its destination. Code is duplicated when its
8945 estimated size is smaller than this value multiplied by the estimated size of
8946 unconditional jump in the hot spots of the program.
8948 The @option{reorder-block-duplicate-feedback} is used only when profile
8949 feedback is available and may be set to higher values than
8950 @option{reorder-block-duplicate} since information about the hot spots is more
8953 @item max-sched-ready-insns
8954 The maximum number of instructions ready to be issued the scheduler should
8955 consider at any given time during the first scheduling pass. Increasing
8956 values mean more thorough searches, making the compilation time increase
8957 with probably little benefit. The default value is 100.
8959 @item max-sched-region-blocks
8960 The maximum number of blocks in a region to be considered for
8961 interblock scheduling. The default value is 10.
8963 @item max-pipeline-region-blocks
8964 The maximum number of blocks in a region to be considered for
8965 pipelining in the selective scheduler. The default value is 15.
8967 @item max-sched-region-insns
8968 The maximum number of insns in a region to be considered for
8969 interblock scheduling. The default value is 100.
8971 @item max-pipeline-region-insns
8972 The maximum number of insns in a region to be considered for
8973 pipelining in the selective scheduler. The default value is 200.
8976 The minimum probability (in percents) of reaching a source block
8977 for interblock speculative scheduling. The default value is 40.
8979 @item max-sched-extend-regions-iters
8980 The maximum number of iterations through CFG to extend regions.
8981 0 - disable region extension,
8982 N - do at most N iterations.
8983 The default value is 0.
8985 @item max-sched-insn-conflict-delay
8986 The maximum conflict delay for an insn to be considered for speculative motion.
8987 The default value is 3.
8989 @item sched-spec-prob-cutoff
8990 The minimal probability of speculation success (in percents), so that
8991 speculative insn will be scheduled.
8992 The default value is 40.
8994 @item sched-mem-true-dep-cost
8995 Minimal distance (in CPU cycles) between store and load targeting same
8996 memory locations. The default value is 1.
8998 @item selsched-max-lookahead
8999 The maximum size of the lookahead window of selective scheduling. It is a
9000 depth of search for available instructions.
9001 The default value is 50.
9003 @item selsched-max-sched-times
9004 The maximum number of times that an instruction will be scheduled during
9005 selective scheduling. This is the limit on the number of iterations
9006 through which the instruction may be pipelined. The default value is 2.
9008 @item selsched-max-insns-to-rename
9009 The maximum number of best instructions in the ready list that are considered
9010 for renaming in the selective scheduler. The default value is 2.
9013 The minimum value of stage count that swing modulo scheduler will
9014 generate. The default value is 2.
9016 @item max-last-value-rtl
9017 The maximum size measured as number of RTLs that can be recorded in an expression
9018 in combiner for a pseudo register as last known value of that register. The default
9021 @item integer-share-limit
9022 Small integer constants can use a shared data structure, reducing the
9023 compiler's memory usage and increasing its speed. This sets the maximum
9024 value of a shared integer constant. The default value is 256.
9026 @item min-virtual-mappings
9027 Specifies the minimum number of virtual mappings in the incremental
9028 SSA updater that should be registered to trigger the virtual mappings
9029 heuristic defined by virtual-mappings-ratio. The default value is
9032 @item virtual-mappings-ratio
9033 If the number of virtual mappings is virtual-mappings-ratio bigger
9034 than the number of virtual symbols to be updated, then the incremental
9035 SSA updater switches to a full update for those symbols. The default
9038 @item ssp-buffer-size
9039 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
9040 protection when @option{-fstack-protection} is used.
9042 @item max-jump-thread-duplication-stmts
9043 Maximum number of statements allowed in a block that needs to be
9044 duplicated when threading jumps.
9046 @item max-fields-for-field-sensitive
9047 Maximum number of fields in a structure we will treat in
9048 a field sensitive manner during pointer analysis. The default is zero
9049 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
9051 @item prefetch-latency
9052 Estimate on average number of instructions that are executed before
9053 prefetch finishes. The distance we prefetch ahead is proportional
9054 to this constant. Increasing this number may also lead to less
9055 streams being prefetched (see @option{simultaneous-prefetches}).
9057 @item simultaneous-prefetches
9058 Maximum number of prefetches that can run at the same time.
9060 @item l1-cache-line-size
9061 The size of cache line in L1 cache, in bytes.
9064 The size of L1 cache, in kilobytes.
9067 The size of L2 cache, in kilobytes.
9069 @item min-insn-to-prefetch-ratio
9070 The minimum ratio between the number of instructions and the
9071 number of prefetches to enable prefetching in a loop.
9073 @item prefetch-min-insn-to-mem-ratio
9074 The minimum ratio between the number of instructions and the
9075 number of memory references to enable prefetching in a loop.
9077 @item use-canonical-types
9078 Whether the compiler should use the ``canonical'' type system. By
9079 default, this should always be 1, which uses a more efficient internal
9080 mechanism for comparing types in C++ and Objective-C++. However, if
9081 bugs in the canonical type system are causing compilation failures,
9082 set this value to 0 to disable canonical types.
9084 @item switch-conversion-max-branch-ratio
9085 Switch initialization conversion will refuse to create arrays that are
9086 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9087 branches in the switch.
9089 @item max-partial-antic-length
9090 Maximum length of the partial antic set computed during the tree
9091 partial redundancy elimination optimization (@option{-ftree-pre}) when
9092 optimizing at @option{-O3} and above. For some sorts of source code
9093 the enhanced partial redundancy elimination optimization can run away,
9094 consuming all of the memory available on the host machine. This
9095 parameter sets a limit on the length of the sets that are computed,
9096 which prevents the runaway behavior. Setting a value of 0 for
9097 this parameter will allow an unlimited set length.
9099 @item sccvn-max-scc-size
9100 Maximum size of a strongly connected component (SCC) during SCCVN
9101 processing. If this limit is hit, SCCVN processing for the whole
9102 function will not be done and optimizations depending on it will
9103 be disabled. The default maximum SCC size is 10000.
9105 @item ira-max-loops-num
9106 IRA uses a regional register allocation by default. If a function
9107 contains loops more than number given by the parameter, only at most
9108 given number of the most frequently executed loops will form regions
9109 for the regional register allocation. The default value of the
9112 @item ira-max-conflict-table-size
9113 Although IRA uses a sophisticated algorithm of compression conflict
9114 table, the table can be still big for huge functions. If the conflict
9115 table for a function could be more than size in MB given by the
9116 parameter, the conflict table is not built and faster, simpler, and
9117 lower quality register allocation algorithm will be used. The
9118 algorithm do not use pseudo-register conflicts. The default value of
9119 the parameter is 2000.
9121 @item ira-loop-reserved-regs
9122 IRA can be used to evaluate more accurate register pressure in loops
9123 for decision to move loop invariants (see @option{-O3}). The number
9124 of available registers reserved for some other purposes is described
9125 by this parameter. The default value of the parameter is 2 which is
9126 minimal number of registers needed for execution of typical
9127 instruction. This value is the best found from numerous experiments.
9129 @item loop-invariant-max-bbs-in-loop
9130 Loop invariant motion can be very expensive, both in compilation time and
9131 in amount of needed compile-time memory, with very large loops. Loops
9132 with more basic blocks than this parameter won't have loop invariant
9133 motion optimization performed on them. The default value of the
9134 parameter is 1000 for -O1 and 10000 for -O2 and above.
9136 @item loop-max-datarefs-for-datadeps
9137 Building data dapendencies is expensive for very large loops. This
9138 parameter limits the number of data references in loops that are
9139 considered for data dependence analysis. These large loops will not
9140 be handled then by the optimizations using loop data dependencies.
9141 The default value is 1000.
9143 @item max-vartrack-size
9144 Sets a maximum number of hash table slots to use during variable
9145 tracking dataflow analysis of any function. If this limit is exceeded
9146 with variable tracking at assignments enabled, analysis for that
9147 function is retried without it, after removing all debug insns from
9148 the function. If the limit is exceeded even without debug insns, var
9149 tracking analysis is completely disabled for the function. Setting
9150 the parameter to zero makes it unlimited.
9152 @item max-vartrack-expr-depth
9153 Sets a maximum number of recursion levels when attempting to map
9154 variable names or debug temporaries to value expressions. This trades
9155 compilation time for more complete debug information. If this is set too
9156 low, value expressions that are available and could be represented in
9157 debug information may end up not being used; setting this higher may
9158 enable the compiler to find more complex debug expressions, but compile
9159 time and memory use may grow. The default is 12.
9161 @item min-nondebug-insn-uid
9162 Use uids starting at this parameter for nondebug insns. The range below
9163 the parameter is reserved exclusively for debug insns created by
9164 @option{-fvar-tracking-assignments}, but debug insns may get
9165 (non-overlapping) uids above it if the reserved range is exhausted.
9167 @item ipa-sra-ptr-growth-factor
9168 IPA-SRA will replace a pointer to an aggregate with one or more new
9169 parameters only when their cumulative size is less or equal to
9170 @option{ipa-sra-ptr-growth-factor} times the size of the original
9173 @item tm-max-aggregate-size
9174 When making copies of thread-local variables in a transaction, this
9175 parameter specifies the size in bytes after which variables will be
9176 saved with the logging functions as opposed to save/restore code
9177 sequence pairs. This option only applies when using
9180 @item graphite-max-nb-scop-params
9181 To avoid exponential effects in the Graphite loop transforms, the
9182 number of parameters in a Static Control Part (SCoP) is bounded. The
9183 default value is 10 parameters. A variable whose value is unknown at
9184 compilation time and defined outside a SCoP is a parameter of the SCoP.
9186 @item graphite-max-bbs-per-function
9187 To avoid exponential effects in the detection of SCoPs, the size of
9188 the functions analyzed by Graphite is bounded. The default value is
9191 @item loop-block-tile-size
9192 Loop blocking or strip mining transforms, enabled with
9193 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9194 loop in the loop nest by a given number of iterations. The strip
9195 length can be changed using the @option{loop-block-tile-size}
9196 parameter. The default value is 51 iterations.
9198 @item ipa-cp-value-list-size
9199 IPA-CP attempts to track all possible values and types passed to a function's
9200 parameter in order to propagate them and perform devirtualization.
9201 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9202 stores per one formal parameter of a function.
9204 @item lto-partitions
9205 Specify desired number of partitions produced during WHOPR compilation.
9206 The number of partitions should exceed the number of CPUs used for compilation.
9207 The default value is 32.
9209 @item lto-minpartition
9210 Size of minimal partition for WHOPR (in estimated instructions).
9211 This prevents expenses of splitting very small programs into too many
9214 @item cxx-max-namespaces-for-diagnostic-help
9215 The maximum number of namespaces to consult for suggestions when C++
9216 name lookup fails for an identifier. The default is 1000.
9218 @item sink-frequency-threshold
9219 The maximum relative execution frequency (in percents) of the target block
9220 relative to a statement's original block to allow statement sinking of a
9221 statement. Larger numbers result in more aggressive statement sinking.
9222 The default value is 75. A small positive adjustment is applied for
9223 statements with memory operands as those are even more profitable so sink.
9225 @item max-stores-to-sink
9226 The maximum number of conditional stores paires that can be sunk. Set to 0
9227 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9228 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9230 @item allow-load-data-races
9231 Allow optimizers to introduce new data races on loads.
9232 Set to 1 to allow, otherwise to 0. This option is enabled by default
9233 unless implicitly set by the @option{-fmemory-model=} option.
9235 @item allow-store-data-races
9236 Allow optimizers to introduce new data races on stores.
9237 Set to 1 to allow, otherwise to 0. This option is enabled by default
9238 unless implicitly set by the @option{-fmemory-model=} option.
9240 @item allow-packed-load-data-races
9241 Allow optimizers to introduce new data races on packed data loads.
9242 Set to 1 to allow, otherwise to 0. This option is enabled by default
9243 unless implicitly set by the @option{-fmemory-model=} option.
9245 @item allow-packed-store-data-races
9246 Allow optimizers to introduce new data races on packed data stores.
9247 Set to 1 to allow, otherwise to 0. This option is enabled by default
9248 unless implicitly set by the @option{-fmemory-model=} option.
9250 @item case-values-threshold
9251 The smallest number of different values for which it is best to use a
9252 jump-table instead of a tree of conditional branches. If the value is
9253 0, use the default for the machine. The default is 0.
9255 @item tree-reassoc-width
9256 Set the maximum number of instructions executed in parallel in
9257 reassociated tree. This parameter overrides target dependent
9258 heuristics used by default if has non zero value.
9263 @node Preprocessor Options
9264 @section Options Controlling the Preprocessor
9265 @cindex preprocessor options
9266 @cindex options, preprocessor
9268 These options control the C preprocessor, which is run on each C source
9269 file before actual compilation.
9271 If you use the @option{-E} option, nothing is done except preprocessing.
9272 Some of these options make sense only together with @option{-E} because
9273 they cause the preprocessor output to be unsuitable for actual
9277 @item -Wp,@var{option}
9279 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9280 and pass @var{option} directly through to the preprocessor. If
9281 @var{option} contains commas, it is split into multiple options at the
9282 commas. However, many options are modified, translated or interpreted
9283 by the compiler driver before being passed to the preprocessor, and
9284 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9285 interface is undocumented and subject to change, so whenever possible
9286 you should avoid using @option{-Wp} and let the driver handle the
9289 @item -Xpreprocessor @var{option}
9290 @opindex Xpreprocessor
9291 Pass @var{option} as an option to the preprocessor. You can use this to
9292 supply system-specific preprocessor options that GCC does not know how to
9295 If you want to pass an option that takes an argument, you must use
9296 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9299 @include cppopts.texi
9301 @node Assembler Options
9302 @section Passing Options to the Assembler
9304 @c prevent bad page break with this line
9305 You can pass options to the assembler.
9308 @item -Wa,@var{option}
9310 Pass @var{option} as an option to the assembler. If @var{option}
9311 contains commas, it is split into multiple options at the commas.
9313 @item -Xassembler @var{option}
9315 Pass @var{option} as an option to the assembler. You can use this to
9316 supply system-specific assembler options that GCC does not know how to
9319 If you want to pass an option that takes an argument, you must use
9320 @option{-Xassembler} twice, once for the option and once for the argument.
9325 @section Options for Linking
9326 @cindex link options
9327 @cindex options, linking
9329 These options come into play when the compiler links object files into
9330 an executable output file. They are meaningless if the compiler is
9331 not doing a link step.
9335 @item @var{object-file-name}
9336 A file name that does not end in a special recognized suffix is
9337 considered to name an object file or library. (Object files are
9338 distinguished from libraries by the linker according to the file
9339 contents.) If linking is done, these object files are used as input
9348 If any of these options is used, then the linker is not run, and
9349 object file names should not be used as arguments. @xref{Overall
9353 @item -l@var{library}
9354 @itemx -l @var{library}
9356 Search the library named @var{library} when linking. (The second
9357 alternative with the library as a separate argument is only for
9358 POSIX compliance and is not recommended.)
9360 It makes a difference where in the command you write this option; the
9361 linker searches and processes libraries and object files in the order they
9362 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9363 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9364 to functions in @samp{z}, those functions may not be loaded.
9366 The linker searches a standard list of directories for the library,
9367 which is actually a file named @file{lib@var{library}.a}. The linker
9368 then uses this file as if it had been specified precisely by name.
9370 The directories searched include several standard system directories
9371 plus any that you specify with @option{-L}.
9373 Normally the files found this way are library files---archive files
9374 whose members are object files. The linker handles an archive file by
9375 scanning through it for members which define symbols that have so far
9376 been referenced but not defined. But if the file that is found is an
9377 ordinary object file, it is linked in the usual fashion. The only
9378 difference between using an @option{-l} option and specifying a file name
9379 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9380 and searches several directories.
9384 You need this special case of the @option{-l} option in order to
9385 link an Objective-C or Objective-C++ program.
9388 @opindex nostartfiles
9389 Do not use the standard system startup files when linking.
9390 The standard system libraries are used normally, unless @option{-nostdlib}
9391 or @option{-nodefaultlibs} is used.
9393 @item -nodefaultlibs
9394 @opindex nodefaultlibs
9395 Do not use the standard system libraries when linking.
9396 Only the libraries you specify will be passed to the linker, options
9397 specifying linkage of the system libraries, such as @code{-static-libgcc}
9398 or @code{-shared-libgcc}, will be ignored.
9399 The standard startup files are used normally, unless @option{-nostartfiles}
9400 is used. The compiler may generate calls to @code{memcmp},
9401 @code{memset}, @code{memcpy} and @code{memmove}.
9402 These entries are usually resolved by entries in
9403 libc. These entry points should be supplied through some other
9404 mechanism when this option is specified.
9408 Do not use the standard system startup files or libraries when linking.
9409 No startup files and only the libraries you specify will be passed to
9410 the linker, options specifying linkage of the system libraries, such as
9411 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9412 The compiler may generate calls to @code{memcmp}, @code{memset},
9413 @code{memcpy} and @code{memmove}.
9414 These entries are usually resolved by entries in
9415 libc. These entry points should be supplied through some other
9416 mechanism when this option is specified.
9418 @cindex @option{-lgcc}, use with @option{-nostdlib}
9419 @cindex @option{-nostdlib} and unresolved references
9420 @cindex unresolved references and @option{-nostdlib}
9421 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9422 @cindex @option{-nodefaultlibs} and unresolved references
9423 @cindex unresolved references and @option{-nodefaultlibs}
9424 One of the standard libraries bypassed by @option{-nostdlib} and
9425 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9426 which GCC uses to overcome shortcomings of particular machines, or special
9427 needs for some languages.
9428 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9429 Collection (GCC) Internals},
9430 for more discussion of @file{libgcc.a}.)
9431 In most cases, you need @file{libgcc.a} even when you want to avoid
9432 other standard libraries. In other words, when you specify @option{-nostdlib}
9433 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9434 This ensures that you have no unresolved references to internal GCC
9435 library subroutines. (For example, @samp{__main}, used to ensure C++
9436 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9437 GNU Compiler Collection (GCC) Internals}.)
9441 Produce a position independent executable on targets that support it.
9442 For predictable results, you must also specify the same set of options
9443 that were used to generate code (@option{-fpie}, @option{-fPIE},
9444 or model suboptions) when you specify this option.
9448 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9449 that support it. This instructs the linker to add all symbols, not
9450 only used ones, to the dynamic symbol table. This option is needed
9451 for some uses of @code{dlopen} or to allow obtaining backtraces
9452 from within a program.
9456 Remove all symbol table and relocation information from the executable.
9460 On systems that support dynamic linking, this prevents linking with the shared
9461 libraries. On other systems, this option has no effect.
9465 Produce a shared object which can then be linked with other objects to
9466 form an executable. Not all systems support this option. For predictable
9467 results, you must also specify the same set of options that were used to
9468 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9469 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9470 needs to build supplementary stub code for constructors to work. On
9471 multi-libbed systems, @samp{gcc -shared} must select the correct support
9472 libraries to link against. Failing to supply the correct flags may lead
9473 to subtle defects. Supplying them in cases where they are not necessary
9476 @item -shared-libgcc
9477 @itemx -static-libgcc
9478 @opindex shared-libgcc
9479 @opindex static-libgcc
9480 On systems that provide @file{libgcc} as a shared library, these options
9481 force the use of either the shared or static version respectively.
9482 If no shared version of @file{libgcc} was built when the compiler was
9483 configured, these options have no effect.
9485 There are several situations in which an application should use the
9486 shared @file{libgcc} instead of the static version. The most common
9487 of these is when the application wishes to throw and catch exceptions
9488 across different shared libraries. In that case, each of the libraries
9489 as well as the application itself should use the shared @file{libgcc}.
9491 Therefore, the G++ and GCJ drivers automatically add
9492 @option{-shared-libgcc} whenever you build a shared library or a main
9493 executable, because C++ and Java programs typically use exceptions, so
9494 this is the right thing to do.
9496 If, instead, you use the GCC driver to create shared libraries, you may
9497 find that they will not always be linked with the shared @file{libgcc}.
9498 If GCC finds, at its configuration time, that you have a non-GNU linker
9499 or a GNU linker that does not support option @option{--eh-frame-hdr},
9500 it will link the shared version of @file{libgcc} into shared libraries
9501 by default. Otherwise, it will take advantage of the linker and optimize
9502 away the linking with the shared version of @file{libgcc}, linking with
9503 the static version of libgcc by default. This allows exceptions to
9504 propagate through such shared libraries, without incurring relocation
9505 costs at library load time.
9507 However, if a library or main executable is supposed to throw or catch
9508 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9509 for the languages used in the program, or using the option
9510 @option{-shared-libgcc}, such that it is linked with the shared
9513 @item -static-libstdc++
9514 When the @command{g++} program is used to link a C++ program, it will
9515 normally automatically link against @option{libstdc++}. If
9516 @file{libstdc++} is available as a shared library, and the
9517 @option{-static} option is not used, then this will link against the
9518 shared version of @file{libstdc++}. That is normally fine. However, it
9519 is sometimes useful to freeze the version of @file{libstdc++} used by
9520 the program without going all the way to a fully static link. The
9521 @option{-static-libstdc++} option directs the @command{g++} driver to
9522 link @file{libstdc++} statically, without necessarily linking other
9523 libraries statically.
9527 Bind references to global symbols when building a shared object. Warn
9528 about any unresolved references (unless overridden by the link editor
9529 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9532 @item -T @var{script}
9534 @cindex linker script
9535 Use @var{script} as the linker script. This option is supported by most
9536 systems using the GNU linker. On some targets, such as bare-board
9537 targets without an operating system, the @option{-T} option may be required
9538 when linking to avoid references to undefined symbols.
9540 @item -Xlinker @var{option}
9542 Pass @var{option} as an option to the linker. You can use this to
9543 supply system-specific linker options that GCC does not recognize.
9545 If you want to pass an option that takes a separate argument, you must use
9546 @option{-Xlinker} twice, once for the option and once for the argument.
9547 For example, to pass @option{-assert definitions}, you must write
9548 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9549 @option{-Xlinker "-assert definitions"}, because this passes the entire
9550 string as a single argument, which is not what the linker expects.
9552 When using the GNU linker, it is usually more convenient to pass
9553 arguments to linker options using the @option{@var{option}=@var{value}}
9554 syntax than as separate arguments. For example, you can specify
9555 @samp{-Xlinker -Map=output.map} rather than
9556 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9557 this syntax for command-line options.
9559 @item -Wl,@var{option}
9561 Pass @var{option} as an option to the linker. If @var{option} contains
9562 commas, it is split into multiple options at the commas. You can use this
9563 syntax to pass an argument to the option.
9564 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9565 linker. When using the GNU linker, you can also get the same effect with
9566 @samp{-Wl,-Map=output.map}.
9568 @item -u @var{symbol}
9570 Pretend the symbol @var{symbol} is undefined, to force linking of
9571 library modules to define it. You can use @option{-u} multiple times with
9572 different symbols to force loading of additional library modules.
9575 @node Directory Options
9576 @section Options for Directory Search
9577 @cindex directory options
9578 @cindex options, directory search
9581 These options specify directories to search for header files, for
9582 libraries and for parts of the compiler:
9587 Add the directory @var{dir} to the head of the list of directories to be
9588 searched for header files. This can be used to override a system header
9589 file, substituting your own version, since these directories are
9590 searched before the system header file directories. However, you should
9591 not use this option to add directories that contain vendor-supplied
9592 system header files (use @option{-isystem} for that). If you use more than
9593 one @option{-I} option, the directories are scanned in left-to-right
9594 order; the standard system directories come after.
9596 If a standard system include directory, or a directory specified with
9597 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9598 option will be ignored. The directory will still be searched but as a
9599 system directory at its normal position in the system include chain.
9600 This is to ensure that GCC's procedure to fix buggy system headers and
9601 the ordering for the include_next directive are not inadvertently changed.
9602 If you really need to change the search order for system directories,
9603 use the @option{-nostdinc} and/or @option{-isystem} options.
9605 @item -iplugindir=@var{dir}
9606 Set the directory to search for plugins that are passed
9607 by @option{-fplugin=@var{name}} instead of
9608 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9609 to be used by the user, but only passed by the driver.
9611 @item -iquote@var{dir}
9613 Add the directory @var{dir} to the head of the list of directories to
9614 be searched for header files only for the case of @samp{#include
9615 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9616 otherwise just like @option{-I}.
9620 Add directory @var{dir} to the list of directories to be searched
9623 @item -B@var{prefix}
9625 This option specifies where to find the executables, libraries,
9626 include files, and data files of the compiler itself.
9628 The compiler driver program runs one or more of the subprograms
9629 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9630 @var{prefix} as a prefix for each program it tries to run, both with and
9631 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9633 For each subprogram to be run, the compiler driver first tries the
9634 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9635 was not specified, the driver tries two standard prefixes,
9636 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9637 those results in a file name that is found, the unmodified program
9638 name is searched for using the directories specified in your
9639 @env{PATH} environment variable.
9641 The compiler will check to see if the path provided by the @option{-B}
9642 refers to a directory, and if necessary it will add a directory
9643 separator character at the end of the path.
9645 @option{-B} prefixes that effectively specify directory names also apply
9646 to libraries in the linker, because the compiler translates these
9647 options into @option{-L} options for the linker. They also apply to
9648 includes files in the preprocessor, because the compiler translates these
9649 options into @option{-isystem} options for the preprocessor. In this case,
9650 the compiler appends @samp{include} to the prefix.
9652 The runtime support file @file{libgcc.a} can also be searched for using
9653 the @option{-B} prefix, if needed. If it is not found there, the two
9654 standard prefixes above are tried, and that is all. The file is left
9655 out of the link if it is not found by those means.
9657 Another way to specify a prefix much like the @option{-B} prefix is to use
9658 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9661 As a special kludge, if the path provided by @option{-B} is
9662 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9663 9, then it will be replaced by @file{[dir/]include}. This is to help
9664 with boot-strapping the compiler.
9666 @item -specs=@var{file}
9668 Process @var{file} after the compiler reads in the standard @file{specs}
9669 file, in order to override the defaults which the @file{gcc} driver
9670 program uses when determining what switches to pass to @file{cc1},
9671 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9672 @option{-specs=@var{file}} can be specified on the command line, and they
9673 are processed in order, from left to right.
9675 @item --sysroot=@var{dir}
9677 Use @var{dir} as the logical root directory for headers and libraries.
9678 For example, if the compiler would normally search for headers in
9679 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9680 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9682 If you use both this option and the @option{-isysroot} option, then
9683 the @option{--sysroot} option will apply to libraries, but the
9684 @option{-isysroot} option will apply to header files.
9686 The GNU linker (beginning with version 2.16) has the necessary support
9687 for this option. If your linker does not support this option, the
9688 header file aspect of @option{--sysroot} will still work, but the
9689 library aspect will not.
9693 This option has been deprecated. Please use @option{-iquote} instead for
9694 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9695 Any directories you specify with @option{-I} options before the @option{-I-}
9696 option are searched only for the case of @samp{#include "@var{file}"};
9697 they are not searched for @samp{#include <@var{file}>}.
9699 If additional directories are specified with @option{-I} options after
9700 the @option{-I-}, these directories are searched for all @samp{#include}
9701 directives. (Ordinarily @emph{all} @option{-I} directories are used
9704 In addition, the @option{-I-} option inhibits the use of the current
9705 directory (where the current input file came from) as the first search
9706 directory for @samp{#include "@var{file}"}. There is no way to
9707 override this effect of @option{-I-}. With @option{-I.} you can specify
9708 searching the directory that was current when the compiler was
9709 invoked. That is not exactly the same as what the preprocessor does
9710 by default, but it is often satisfactory.
9712 @option{-I-} does not inhibit the use of the standard system directories
9713 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9720 @section Specifying subprocesses and the switches to pass to them
9723 @command{gcc} is a driver program. It performs its job by invoking a
9724 sequence of other programs to do the work of compiling, assembling and
9725 linking. GCC interprets its command-line parameters and uses these to
9726 deduce which programs it should invoke, and which command-line options
9727 it ought to place on their command lines. This behavior is controlled
9728 by @dfn{spec strings}. In most cases there is one spec string for each
9729 program that GCC can invoke, but a few programs have multiple spec
9730 strings to control their behavior. The spec strings built into GCC can
9731 be overridden by using the @option{-specs=} command-line switch to specify
9734 @dfn{Spec files} are plaintext files that are used to construct spec
9735 strings. They consist of a sequence of directives separated by blank
9736 lines. The type of directive is determined by the first non-whitespace
9737 character on the line, which can be one of the following:
9740 @item %@var{command}
9741 Issues a @var{command} to the spec file processor. The commands that can
9745 @item %include <@var{file}>
9746 @cindex @code{%include}
9747 Search for @var{file} and insert its text at the current point in the
9750 @item %include_noerr <@var{file}>
9751 @cindex @code{%include_noerr}
9752 Just like @samp{%include}, but do not generate an error message if the include
9753 file cannot be found.
9755 @item %rename @var{old_name} @var{new_name}
9756 @cindex @code{%rename}
9757 Rename the spec string @var{old_name} to @var{new_name}.
9761 @item *[@var{spec_name}]:
9762 This tells the compiler to create, override or delete the named spec
9763 string. All lines after this directive up to the next directive or
9764 blank line are considered to be the text for the spec string. If this
9765 results in an empty string then the spec will be deleted. (Or, if the
9766 spec did not exist, then nothing will happen.) Otherwise, if the spec
9767 does not currently exist a new spec will be created. If the spec does
9768 exist then its contents will be overridden by the text of this
9769 directive, unless the first character of that text is the @samp{+}
9770 character, in which case the text will be appended to the spec.
9772 @item [@var{suffix}]:
9773 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9774 and up to the next directive or blank line are considered to make up the
9775 spec string for the indicated suffix. When the compiler encounters an
9776 input file with the named suffix, it will processes the spec string in
9777 order to work out how to compile that file. For example:
9784 This says that any input file whose name ends in @samp{.ZZ} should be
9785 passed to the program @samp{z-compile}, which should be invoked with the
9786 command-line switch @option{-input} and with the result of performing the
9787 @samp{%i} substitution. (See below.)
9789 As an alternative to providing a spec string, the text that follows a
9790 suffix directive can be one of the following:
9793 @item @@@var{language}
9794 This says that the suffix is an alias for a known @var{language}. This is
9795 similar to using the @option{-x} command-line switch to GCC to specify a
9796 language explicitly. For example:
9803 Says that .ZZ files are, in fact, C++ source files.
9806 This causes an error messages saying:
9809 @var{name} compiler not installed on this system.
9813 GCC already has an extensive list of suffixes built into it.
9814 This directive will add an entry to the end of the list of suffixes, but
9815 since the list is searched from the end backwards, it is effectively
9816 possible to override earlier entries using this technique.
9820 GCC has the following spec strings built into it. Spec files can
9821 override these strings or create their own. Note that individual
9822 targets can also add their own spec strings to this list.
9825 asm Options to pass to the assembler
9826 asm_final Options to pass to the assembler post-processor
9827 cpp Options to pass to the C preprocessor
9828 cc1 Options to pass to the C compiler
9829 cc1plus Options to pass to the C++ compiler
9830 endfile Object files to include at the end of the link
9831 link Options to pass to the linker
9832 lib Libraries to include on the command line to the linker
9833 libgcc Decides which GCC support library to pass to the linker
9834 linker Sets the name of the linker
9835 predefines Defines to be passed to the C preprocessor
9836 signed_char Defines to pass to CPP to say whether @code{char} is signed
9838 startfile Object files to include at the start of the link
9841 Here is a small example of a spec file:
9847 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9850 This example renames the spec called @samp{lib} to @samp{old_lib} and
9851 then overrides the previous definition of @samp{lib} with a new one.
9852 The new definition adds in some extra command-line options before
9853 including the text of the old definition.
9855 @dfn{Spec strings} are a list of command-line options to be passed to their
9856 corresponding program. In addition, the spec strings can contain
9857 @samp{%}-prefixed sequences to substitute variable text or to
9858 conditionally insert text into the command line. Using these constructs
9859 it is possible to generate quite complex command lines.
9861 Here is a table of all defined @samp{%}-sequences for spec
9862 strings. Note that spaces are not generated automatically around the
9863 results of expanding these sequences. Therefore you can concatenate them
9864 together or combine them with constant text in a single argument.
9868 Substitute one @samp{%} into the program name or argument.
9871 Substitute the name of the input file being processed.
9874 Substitute the basename of the input file being processed.
9875 This is the substring up to (and not including) the last period
9876 and not including the directory.
9879 This is the same as @samp{%b}, but include the file suffix (text after
9883 Marks the argument containing or following the @samp{%d} as a
9884 temporary file name, so that that file will be deleted if GCC exits
9885 successfully. Unlike @samp{%g}, this contributes no text to the
9888 @item %g@var{suffix}
9889 Substitute a file name that has suffix @var{suffix} and is chosen
9890 once per compilation, and mark the argument in the same way as
9891 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9892 name is now chosen in a way that is hard to predict even when previously
9893 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9894 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9895 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9896 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9897 was simply substituted with a file name chosen once per compilation,
9898 without regard to any appended suffix (which was therefore treated
9899 just like ordinary text), making such attacks more likely to succeed.
9901 @item %u@var{suffix}
9902 Like @samp{%g}, but generates a new temporary file name even if
9903 @samp{%u@var{suffix}} was already seen.
9905 @item %U@var{suffix}
9906 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9907 new one if there is no such last file name. In the absence of any
9908 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9909 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9910 would involve the generation of two distinct file names, one
9911 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9912 simply substituted with a file name chosen for the previous @samp{%u},
9913 without regard to any appended suffix.
9915 @item %j@var{suffix}
9916 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9917 writable, and if save-temps is off; otherwise, substitute the name
9918 of a temporary file, just like @samp{%u}. This temporary file is not
9919 meant for communication between processes, but rather as a junk
9922 @item %|@var{suffix}
9923 @itemx %m@var{suffix}
9924 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9925 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9926 all. These are the two most common ways to instruct a program that it
9927 should read from standard input or write to standard output. If you
9928 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9929 construct: see for example @file{f/lang-specs.h}.
9931 @item %.@var{SUFFIX}
9932 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9933 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9934 terminated by the next space or %.
9937 Marks the argument containing or following the @samp{%w} as the
9938 designated output file of this compilation. This puts the argument
9939 into the sequence of arguments that @samp{%o} will substitute later.
9942 Substitutes the names of all the output files, with spaces
9943 automatically placed around them. You should write spaces
9944 around the @samp{%o} as well or the results are undefined.
9945 @samp{%o} is for use in the specs for running the linker.
9946 Input files whose names have no recognized suffix are not compiled
9947 at all, but they are included among the output files, so they will
9951 Substitutes the suffix for object files. Note that this is
9952 handled specially when it immediately follows @samp{%g, %u, or %U},
9953 because of the need for those to form complete file names. The
9954 handling is such that @samp{%O} is treated exactly as if it had already
9955 been substituted, except that @samp{%g, %u, and %U} do not currently
9956 support additional @var{suffix} characters following @samp{%O} as they would
9957 following, for example, @samp{.o}.
9960 Substitutes the standard macro predefinitions for the
9961 current target machine. Use this when running @code{cpp}.
9964 Like @samp{%p}, but puts @samp{__} before and after the name of each
9965 predefined macro, except for macros that start with @samp{__} or with
9966 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9970 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9971 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9972 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9973 and @option{-imultilib} as necessary.
9976 Current argument is the name of a library or startup file of some sort.
9977 Search for that file in a standard list of directories and substitute
9978 the full name found. The current working directory is included in the
9979 list of directories scanned.
9982 Current argument is the name of a linker script. Search for that file
9983 in the current list of directories to scan for libraries. If the file
9984 is located insert a @option{--script} option into the command line
9985 followed by the full path name found. If the file is not found then
9986 generate an error message. Note: the current working directory is not
9990 Print @var{str} as an error message. @var{str} is terminated by a newline.
9991 Use this when inconsistent options are detected.
9994 Substitute the contents of spec string @var{name} at this point.
9996 @item %x@{@var{option}@}
9997 Accumulate an option for @samp{%X}.
10000 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
10004 Output the accumulated assembler options specified by @option{-Wa}.
10007 Output the accumulated preprocessor options specified by @option{-Wp}.
10010 Process the @code{asm} spec. This is used to compute the
10011 switches to be passed to the assembler.
10014 Process the @code{asm_final} spec. This is a spec string for
10015 passing switches to an assembler post-processor, if such a program is
10019 Process the @code{link} spec. This is the spec for computing the
10020 command line passed to the linker. Typically it will make use of the
10021 @samp{%L %G %S %D and %E} sequences.
10024 Dump out a @option{-L} option for each directory that GCC believes might
10025 contain startup files. If the target supports multilibs then the
10026 current multilib directory will be prepended to each of these paths.
10029 Process the @code{lib} spec. This is a spec string for deciding which
10030 libraries should be included on the command line to the linker.
10033 Process the @code{libgcc} spec. This is a spec string for deciding
10034 which GCC support library should be included on the command line to the linker.
10037 Process the @code{startfile} spec. This is a spec for deciding which
10038 object files should be the first ones passed to the linker. Typically
10039 this might be a file named @file{crt0.o}.
10042 Process the @code{endfile} spec. This is a spec string that specifies
10043 the last object files that will be passed to the linker.
10046 Process the @code{cpp} spec. This is used to construct the arguments
10047 to be passed to the C preprocessor.
10050 Process the @code{cc1} spec. This is used to construct the options to be
10051 passed to the actual C compiler (@samp{cc1}).
10054 Process the @code{cc1plus} spec. This is used to construct the options to be
10055 passed to the actual C++ compiler (@samp{cc1plus}).
10058 Substitute the variable part of a matched option. See below.
10059 Note that each comma in the substituted string is replaced by
10063 Remove all occurrences of @code{-S} from the command line. Note---this
10064 command is position dependent. @samp{%} commands in the spec string
10065 before this one will see @code{-S}, @samp{%} commands in the spec string
10066 after this one will not.
10068 @item %:@var{function}(@var{args})
10069 Call the named function @var{function}, passing it @var{args}.
10070 @var{args} is first processed as a nested spec string, then split
10071 into an argument vector in the usual fashion. The function returns
10072 a string which is processed as if it had appeared literally as part
10073 of the current spec.
10075 The following built-in spec functions are provided:
10078 @item @code{getenv}
10079 The @code{getenv} spec function takes two arguments: an environment
10080 variable name and a string. If the environment variable is not
10081 defined, a fatal error is issued. Otherwise, the return value is the
10082 value of the environment variable concatenated with the string. For
10083 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
10086 %:getenv(TOPDIR /include)
10089 expands to @file{/path/to/top/include}.
10091 @item @code{if-exists}
10092 The @code{if-exists} spec function takes one argument, an absolute
10093 pathname to a file. If the file exists, @code{if-exists} returns the
10094 pathname. Here is a small example of its usage:
10098 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
10101 @item @code{if-exists-else}
10102 The @code{if-exists-else} spec function is similar to the @code{if-exists}
10103 spec function, except that it takes two arguments. The first argument is
10104 an absolute pathname to a file. If the file exists, @code{if-exists-else}
10105 returns the pathname. If it does not exist, it returns the second argument.
10106 This way, @code{if-exists-else} can be used to select one file or another,
10107 based on the existence of the first. Here is a small example of its usage:
10111 crt0%O%s %:if-exists(crti%O%s) \
10112 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
10115 @item @code{replace-outfile}
10116 The @code{replace-outfile} spec function takes two arguments. It looks for the
10117 first argument in the outfiles array and replaces it with the second argument. Here
10118 is a small example of its usage:
10121 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10124 @item @code{remove-outfile}
10125 The @code{remove-outfile} spec function takes one argument. It looks for the
10126 first argument in the outfiles array and removes it. Here is a small example
10130 %:remove-outfile(-lm)
10133 @item @code{pass-through-libs}
10134 The @code{pass-through-libs} spec function takes any number of arguments. It
10135 finds any @option{-l} options and any non-options ending in ".a" (which it
10136 assumes are the names of linker input library archive files) and returns a
10137 result containing all the found arguments each prepended by
10138 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10139 intended to be passed to the LTO linker plugin.
10142 %:pass-through-libs(%G %L %G)
10145 @item @code{print-asm-header}
10146 The @code{print-asm-header} function takes no arguments and simply
10147 prints a banner like:
10153 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10156 It is used to separate compiler options from assembler options
10157 in the @option{--target-help} output.
10160 @item %@{@code{S}@}
10161 Substitutes the @code{-S} switch, if that switch was given to GCC@.
10162 If that switch was not specified, this substitutes nothing. Note that
10163 the leading dash is omitted when specifying this option, and it is
10164 automatically inserted if the substitution is performed. Thus the spec
10165 string @samp{%@{foo@}} would match the command-line option @option{-foo}
10166 and would output the command-line option @option{-foo}.
10168 @item %W@{@code{S}@}
10169 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10170 deleted on failure.
10172 @item %@{@code{S}*@}
10173 Substitutes all the switches specified to GCC whose names start
10174 with @code{-S}, but which also take an argument. This is used for
10175 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10176 GCC considers @option{-o foo} as being
10177 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10178 text, including the space. Thus two arguments would be generated.
10180 @item %@{@code{S}*&@code{T}*@}
10181 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10182 (the order of @code{S} and @code{T} in the spec is not significant).
10183 There can be any number of ampersand-separated variables; for each the
10184 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10186 @item %@{@code{S}:@code{X}@}
10187 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10189 @item %@{!@code{S}:@code{X}@}
10190 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10192 @item %@{@code{S}*:@code{X}@}
10193 Substitutes @code{X} if one or more switches whose names start with
10194 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10195 once, no matter how many such switches appeared. However, if @code{%*}
10196 appears somewhere in @code{X}, then @code{X} will be substituted once
10197 for each matching switch, with the @code{%*} replaced by the part of
10198 that switch that matched the @code{*}.
10200 @item %@{.@code{S}:@code{X}@}
10201 Substitutes @code{X}, if processing a file with suffix @code{S}.
10203 @item %@{!.@code{S}:@code{X}@}
10204 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10206 @item %@{,@code{S}:@code{X}@}
10207 Substitutes @code{X}, if processing a file for language @code{S}.
10209 @item %@{!,@code{S}:@code{X}@}
10210 Substitutes @code{X}, if not processing a file for language @code{S}.
10212 @item %@{@code{S}|@code{P}:@code{X}@}
10213 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10214 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10215 @code{*} sequences as well, although they have a stronger binding than
10216 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10217 alternatives must be starred, and only the first matching alternative
10220 For example, a spec string like this:
10223 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10226 will output the following command-line options from the following input
10227 command-line options:
10232 -d fred.c -foo -baz -boggle
10233 -d jim.d -bar -baz -boggle
10236 @item %@{S:X; T:Y; :D@}
10238 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10239 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10240 be as many clauses as you need. This may be combined with @code{.},
10241 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10246 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10247 construct may contain other nested @samp{%} constructs or spaces, or
10248 even newlines. They are processed as usual, as described above.
10249 Trailing white space in @code{X} is ignored. White space may also
10250 appear anywhere on the left side of the colon in these constructs,
10251 except between @code{.} or @code{*} and the corresponding word.
10253 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10254 handled specifically in these constructs. If another value of
10255 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10256 @option{-W} switch is found later in the command line, the earlier
10257 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10258 just one letter, which passes all matching options.
10260 The character @samp{|} at the beginning of the predicate text is used to
10261 indicate that a command should be piped to the following command, but
10262 only if @option{-pipe} is specified.
10264 It is built into GCC which switches take arguments and which do not.
10265 (You might think it would be useful to generalize this to allow each
10266 compiler's spec to say which switches take arguments. But this cannot
10267 be done in a consistent fashion. GCC cannot even decide which input
10268 files have been specified without knowing which switches take arguments,
10269 and it must know which input files to compile in order to tell which
10272 GCC also knows implicitly that arguments starting in @option{-l} are to be
10273 treated as compiler output files, and passed to the linker in their
10274 proper position among the other output files.
10276 @c man begin OPTIONS
10278 @node Target Options
10279 @section Specifying Target Machine and Compiler Version
10280 @cindex target options
10281 @cindex cross compiling
10282 @cindex specifying machine version
10283 @cindex specifying compiler version and target machine
10284 @cindex compiler version, specifying
10285 @cindex target machine, specifying
10287 The usual way to run GCC is to run the executable called @command{gcc}, or
10288 @command{@var{machine}-gcc} when cross-compiling, or
10289 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10290 one that was installed last.
10292 @node Submodel Options
10293 @section Hardware Models and Configurations
10294 @cindex submodel options
10295 @cindex specifying hardware config
10296 @cindex hardware models and configurations, specifying
10297 @cindex machine dependent options
10299 Each target machine types can have its own
10300 special options, starting with @samp{-m}, to choose among various
10301 hardware models or configurations---for example, 68010 vs 68020,
10302 floating coprocessor or none. A single installed version of the
10303 compiler can compile for any model or configuration, according to the
10306 Some configurations of the compiler also support additional special
10307 options, usually for compatibility with other compilers on the same
10310 @c This list is ordered alphanumerically by subsection name.
10311 @c It should be the same order and spelling as these options are listed
10312 @c in Machine Dependent Options
10315 * Adapteva Epiphany Options::
10318 * Blackfin Options::
10323 * DEC Alpha Options::
10324 * DEC Alpha/VMS Options::
10327 * GNU/Linux Options::
10330 * i386 and x86-64 Options::
10331 * i386 and x86-64 Windows Options::
10333 * IA-64/VMS Options::
10340 * MicroBlaze Options::
10343 * MN10300 Options::
10345 * picoChip Options::
10346 * PowerPC Options::
10348 * RS/6000 and PowerPC Options::
10350 * S/390 and zSeries Options::
10353 * Solaris 2 Options::
10356 * System V Options::
10357 * TILE-Gx Options::
10358 * TILEPro Options::
10361 * VxWorks Options::
10363 * Xstormy16 Options::
10365 * zSeries Options::
10368 @node Adapteva Epiphany Options
10369 @subsection Adapteva Epiphany Options
10371 These @samp{-m} options are defined for Adapteva Epiphany:
10374 @item -mhalf-reg-file
10375 @opindex mhalf-reg-file
10376 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
10377 That allows code to run on hardware variants that lack these registers.
10379 @item -mprefer-short-insn-regs
10380 @opindex mprefer-short-insn-regs
10381 Preferrentially allocate registers that allow short instruction generation.
10382 This can result in increasesd instruction count, so if this reduces or
10383 increases code size might vary from case to case.
10385 @item -mbranch-cost=@var{num}
10386 @opindex mbranch-cost
10387 Set the cost of branches to roughly @var{num} ``simple'' instructions.
10388 This cost is only a heuristic and is not guaranteed to produce
10389 consistent results across releases.
10393 Enable the generation of conditional moves.
10395 @item -mnops=@var{num}
10397 Emit @var{num} nops before every other generated instruction.
10399 @item -mno-soft-cmpsf
10400 @opindex mno-soft-cmpsf
10401 For single-precision floating-point comparisons, emit an fsub instruction
10402 and test the flags. This is faster than a software comparison, but can
10403 get incorrect results in the presence of NaNs, or when two different small
10404 numbers are compared such that their difference is calculated as zero.
10405 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
10406 software comparisons.
10408 @item -mstack-offset=@var{num}
10409 @opindex mstack-offset
10410 Set the offset between the top of the stack and the stack pointer.
10411 E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7
10412 can be used by leaf functions without stack allocation.
10413 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
10414 Note also that this option changes the ABI, compiling a program with a
10415 different stack offset than the libraries have been compiled with
10416 will generally not work.
10417 This option can be useful if you want to evaluate if a different stack
10418 offset would give you better code, but to actually use a different stack
10419 offset to build working programs, it is recommended to configure the
10420 toolchain with the appropriate @samp{--with-stack-offset=@var{num}} option.
10422 @item -mno-round-nearest
10423 @opindex mno-round-nearest
10424 Make the scheduler assume that the rounding mode has been set to
10425 truncating. The default is @option{-mround-nearest}.
10428 @opindex mlong-calls
10429 If not otherwise specified by an attribute, assume all calls might be beyond
10430 the offset range of the b / bl instructions, and therefore load the
10431 function address into a register before performing a (otherwise direct) call.
10432 This is the default.
10434 @item -mshort-calls
10435 @opindex short-calls
10436 If not otherwise specified by an attribute, assume all direct calls are
10437 in the range of the b / bl instructions, so use these instructions
10438 for direct calls. The default is @option{-mlong-calls}.
10442 Assume addresses can be loaded as 16-bit unsigned values. This does not
10443 apply to function addresses for which @option{-mlong-calls} semantics
10446 @item -mfp-mode=@var{mode}
10448 Set the prevailing mode of the floating-point unit.
10449 This determines the floating-point mode that is provided and expected
10450 at function call and return time. Making this mode match the mode you
10451 predominantly need at function start can make your programs smaller and
10452 faster by avoiding unnecessary mode switches.
10454 @var{mode} can be set to one the following values:
10458 Any mode at function entry is valid, and retained or restored when
10459 the function returns, and when it calls other functions.
10460 This mode is useful for compiling libraries or other compilation units
10461 you might want to incorporate into different programs with different
10462 prevailing FPU modes, and the convenience of being able to use a single
10463 object file outweighs the size and speed overhead for any extra
10464 mode switching that might be needed, compared with what would be needed
10465 with a more specific choice of prevailing FPU mode.
10468 This is the mode used for floating-point calculations with
10469 truncating (i.e.@: round towards zero) rounding mode. That includes
10470 conversion from floating point to integer.
10472 @item round-nearest
10473 This is the mode used for floating-point calculations with
10474 round-to-nearest-or-even rounding mode.
10477 This is the mode used to perform integer calculations in the FPU, e.g.@:
10478 integer multiply, or integer multiply-and-accumulate.
10481 The default is @option{-mfp-mode=caller}
10483 @item -mnosplit-lohi
10484 @opindex mnosplit-lohi
10486 @opindex mno-postinc
10487 @item -mno-postmodify
10488 @opindex mno-postmodify
10489 Code generation tweaks that disable, respectively, splitting of 32-bit
10490 loads, generation of post-increment addresses, and generation of
10491 post-modify addresses. The defaults are @option{msplit-lohi},
10492 @option{-mpost-inc}, and @option{-mpost-modify}.
10494 @item -mnovect-double
10495 @opindex mno-vect-double
10496 Change the preferred SIMD mode to SImode. The default is
10497 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
10499 @item -max-vect-align=@var{num}
10500 @opindex max-vect-align
10501 The maximum alignment for SIMD vector mode types.
10502 @var{num} may be 4 or 8. The default is 8.
10503 Note that this is an ABI change, even though many library function
10504 interfaces will be unaffected, if they don't use SIMD vector modes
10505 in places where they affect size and/or alignment of relevant types.
10507 @item -msplit-vecmove-early
10508 @opindex msplit-vecmove-early
10509 Split vector moves into single word moves before reload. In theory this
10510 could give better register allocation, but so far the reverse seems to be
10511 generally the case.
10513 @item -m1reg-@var{reg}
10515 Specify a register to hold the constant @minus{}1, which makes loading small negative
10516 constants and certain bitmasks faster.
10517 Allowable values for reg are r43 and r63, which specify to use that register
10518 as a fixed register, and none, which means that no register is used for this
10519 purpose. The default is @option{-m1reg-none}.
10524 @subsection ARM Options
10525 @cindex ARM options
10527 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10531 @item -mabi=@var{name}
10533 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10534 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10537 @opindex mapcs-frame
10538 Generate a stack frame that is compliant with the ARM Procedure Call
10539 Standard for all functions, even if this is not strictly necessary for
10540 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10541 with this option will cause the stack frames not to be generated for
10542 leaf functions. The default is @option{-mno-apcs-frame}.
10546 This is a synonym for @option{-mapcs-frame}.
10549 @c not currently implemented
10550 @item -mapcs-stack-check
10551 @opindex mapcs-stack-check
10552 Generate code to check the amount of stack space available upon entry to
10553 every function (that actually uses some stack space). If there is
10554 insufficient space available then either the function
10555 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10556 called, depending upon the amount of stack space required. The runtime
10557 system is required to provide these functions. The default is
10558 @option{-mno-apcs-stack-check}, since this produces smaller code.
10560 @c not currently implemented
10562 @opindex mapcs-float
10563 Pass floating-point arguments using the floating-point registers. This is
10564 one of the variants of the APCS@. This option is recommended if the
10565 target hardware has a floating-point unit or if a lot of floating-point
10566 arithmetic is going to be performed by the code. The default is
10567 @option{-mno-apcs-float}, since integer only code is slightly increased in
10568 size if @option{-mapcs-float} is used.
10570 @c not currently implemented
10571 @item -mapcs-reentrant
10572 @opindex mapcs-reentrant
10573 Generate reentrant, position independent code. The default is
10574 @option{-mno-apcs-reentrant}.
10577 @item -mthumb-interwork
10578 @opindex mthumb-interwork
10579 Generate code that supports calling between the ARM and Thumb
10580 instruction sets. Without this option, on pre-v5 architectures, the
10581 two instruction sets cannot be reliably used inside one program. The
10582 default is @option{-mno-thumb-interwork}, since slightly larger code
10583 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10584 configurations this option is meaningless.
10586 @item -mno-sched-prolog
10587 @opindex mno-sched-prolog
10588 Prevent the reordering of instructions in the function prolog, or the
10589 merging of those instruction with the instructions in the function's
10590 body. This means that all functions will start with a recognizable set
10591 of instructions (or in fact one of a choice from a small set of
10592 different function prologues), and this information can be used to
10593 locate the start if functions inside an executable piece of code. The
10594 default is @option{-msched-prolog}.
10596 @item -mfloat-abi=@var{name}
10597 @opindex mfloat-abi
10598 Specifies which floating-point ABI to use. Permissible values
10599 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10601 Specifying @samp{soft} causes GCC to generate output containing
10602 library calls for floating-point operations.
10603 @samp{softfp} allows the generation of code using hardware floating-point
10604 instructions, but still uses the soft-float calling conventions.
10605 @samp{hard} allows generation of floating-point instructions
10606 and uses FPU-specific calling conventions.
10608 The default depends on the specific target configuration. Note that
10609 the hard-float and soft-float ABIs are not link-compatible; you must
10610 compile your entire program with the same ABI, and link with a
10611 compatible set of libraries.
10613 @item -mlittle-endian
10614 @opindex mlittle-endian
10615 Generate code for a processor running in little-endian mode. This is
10616 the default for all standard configurations.
10619 @opindex mbig-endian
10620 Generate code for a processor running in big-endian mode; the default is
10621 to compile code for a little-endian processor.
10623 @item -mwords-little-endian
10624 @opindex mwords-little-endian
10625 This option only applies when generating code for big-endian processors.
10626 Generate code for a little-endian word order but a big-endian byte
10627 order. That is, a byte order of the form @samp{32107654}. Note: this
10628 option should only be used if you require compatibility with code for
10629 big-endian ARM processors generated by versions of the compiler prior to
10630 2.8. This option is now deprecated.
10632 @item -mcpu=@var{name}
10634 This specifies the name of the target ARM processor. GCC uses this name
10635 to determine what kind of instructions it can emit when generating
10636 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10637 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10638 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10639 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10640 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10642 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10643 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10644 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10645 @samp{strongarm1110},
10646 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10647 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10648 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10649 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10650 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10651 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10652 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10653 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
10654 @samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10655 @samp{cortex-m4}, @samp{cortex-m3},
10658 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10659 @samp{fa526}, @samp{fa626},
10660 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10663 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10664 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10665 See @option{-mtune} for more information.
10667 @option{-mcpu=native} causes the compiler to auto-detect the CPU
10668 of the build computer. At present, this feature is only supported on
10669 Linux, and not all architectures are recognised. If the auto-detect is
10670 unsuccessful the option has no effect.
10672 @item -mtune=@var{name}
10674 This option is very similar to the @option{-mcpu=} option, except that
10675 instead of specifying the actual target processor type, and hence
10676 restricting which instructions can be used, it specifies that GCC should
10677 tune the performance of the code as if the target were of the type
10678 specified in this option, but still choosing the instructions that it
10679 will generate based on the CPU specified by a @option{-mcpu=} option.
10680 For some ARM implementations better performance can be obtained by using
10683 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10684 performance for a blend of processors within architecture @var{arch}.
10685 The aim is to generate code that run well on the current most popular
10686 processors, balancing between optimizations that benefit some CPUs in the
10687 range, and avoiding performance pitfalls of other CPUs. The effects of
10688 this option may change in future GCC versions as CPU models come and go.
10690 @option{-mtune=native} causes the compiler to auto-detect the CPU
10691 of the build computer. At present, this feature is only supported on
10692 Linux, and not all architectures are recognised. If the auto-detect is
10693 unsuccessful the option has no effect.
10695 @item -march=@var{name}
10697 This specifies the name of the target ARM architecture. GCC uses this
10698 name to determine what kind of instructions it can emit when generating
10699 assembly code. This option can be used in conjunction with or instead
10700 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10701 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10702 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10703 @samp{armv6}, @samp{armv6j},
10704 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10705 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10706 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10708 @option{-march=native} causes the compiler to auto-detect the architecture
10709 of the build computer. At present, this feature is only supported on
10710 Linux, and not all architectures are recognised. If the auto-detect is
10711 unsuccessful the option has no effect.
10713 @item -mfpu=@var{name}
10714 @itemx -mfpe=@var{number}
10715 @itemx -mfp=@var{number}
10719 This specifies what floating-point hardware (or hardware emulation) is
10720 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10721 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10722 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10723 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10724 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10725 @option{-mfp} and @option{-mfpe} are synonyms for
10726 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10729 If @option{-msoft-float} is specified this specifies the format of
10730 floating-point values.
10732 If the selected floating-point hardware includes the NEON extension
10733 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10734 operations will not be used by GCC's auto-vectorization pass unless
10735 @option{-funsafe-math-optimizations} is also specified. This is
10736 because NEON hardware does not fully implement the IEEE 754 standard for
10737 floating-point arithmetic (in particular denormal values are treated as
10738 zero), so the use of NEON instructions may lead to a loss of precision.
10740 @item -mfp16-format=@var{name}
10741 @opindex mfp16-format
10742 Specify the format of the @code{__fp16} half-precision floating-point type.
10743 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10744 the default is @samp{none}, in which case the @code{__fp16} type is not
10745 defined. @xref{Half-Precision}, for more information.
10747 @item -mstructure-size-boundary=@var{n}
10748 @opindex mstructure-size-boundary
10749 The size of all structures and unions will be rounded up to a multiple
10750 of the number of bits set by this option. Permissible values are 8, 32
10751 and 64. The default value varies for different toolchains. For the COFF
10752 targeted toolchain the default value is 8. A value of 64 is only allowed
10753 if the underlying ABI supports it.
10755 Specifying the larger number can produce faster, more efficient code, but
10756 can also increase the size of the program. Different values are potentially
10757 incompatible. Code compiled with one value cannot necessarily expect to
10758 work with code or libraries compiled with another value, if they exchange
10759 information using structures or unions.
10761 @item -mabort-on-noreturn
10762 @opindex mabort-on-noreturn
10763 Generate a call to the function @code{abort} at the end of a
10764 @code{noreturn} function. It will be executed if the function tries to
10768 @itemx -mno-long-calls
10769 @opindex mlong-calls
10770 @opindex mno-long-calls
10771 Tells the compiler to perform function calls by first loading the
10772 address of the function into a register and then performing a subroutine
10773 call on this register. This switch is needed if the target function
10774 will lie outside of the 64 megabyte addressing range of the offset based
10775 version of subroutine call instruction.
10777 Even if this switch is enabled, not all function calls will be turned
10778 into long calls. The heuristic is that static functions, functions
10779 that have the @samp{short-call} attribute, functions that are inside
10780 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10781 definitions have already been compiled within the current compilation
10782 unit, will not be turned into long calls. The exception to this rule is
10783 that weak function definitions, functions with the @samp{long-call}
10784 attribute or the @samp{section} attribute, and functions that are within
10785 the scope of a @samp{#pragma long_calls} directive, will always be
10786 turned into long calls.
10788 This feature is not enabled by default. Specifying
10789 @option{-mno-long-calls} will restore the default behavior, as will
10790 placing the function calls within the scope of a @samp{#pragma
10791 long_calls_off} directive. Note these switches have no effect on how
10792 the compiler generates code to handle function calls via function
10795 @item -msingle-pic-base
10796 @opindex msingle-pic-base
10797 Treat the register used for PIC addressing as read-only, rather than
10798 loading it in the prologue for each function. The runtime system is
10799 responsible for initializing this register with an appropriate value
10800 before execution begins.
10802 @item -mpic-register=@var{reg}
10803 @opindex mpic-register
10804 Specify the register to be used for PIC addressing. The default is R10
10805 unless stack-checking is enabled, when R9 is used.
10807 @item -mcirrus-fix-invalid-insns
10808 @opindex mcirrus-fix-invalid-insns
10809 @opindex mno-cirrus-fix-invalid-insns
10810 Insert NOPs into the instruction stream to in order to work around
10811 problems with invalid Maverick instruction combinations. This option
10812 is only valid if the @option{-mcpu=ep9312} option has been used to
10813 enable generation of instructions for the Cirrus Maverick floating-point
10814 co-processor. This option is not enabled by default, since the
10815 problem is only present in older Maverick implementations. The default
10816 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10819 @item -mpoke-function-name
10820 @opindex mpoke-function-name
10821 Write the name of each function into the text section, directly
10822 preceding the function prologue. The generated code is similar to this:
10826 .ascii "arm_poke_function_name", 0
10829 .word 0xff000000 + (t1 - t0)
10830 arm_poke_function_name
10832 stmfd sp!, @{fp, ip, lr, pc@}
10836 When performing a stack backtrace, code can inspect the value of
10837 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10838 location @code{pc - 12} and the top 8 bits are set, then we know that
10839 there is a function name embedded immediately preceding this location
10840 and has length @code{((pc[-3]) & 0xff000000)}.
10847 Select between generating code that executes in ARM and Thumb
10848 states. The default for most configurations is to generate code
10849 that executes in ARM state, but the default can be changed by
10850 configuring GCC with the @option{--with-mode=}@var{state}
10854 @opindex mtpcs-frame
10855 Generate a stack frame that is compliant with the Thumb Procedure Call
10856 Standard for all non-leaf functions. (A leaf function is one that does
10857 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10859 @item -mtpcs-leaf-frame
10860 @opindex mtpcs-leaf-frame
10861 Generate a stack frame that is compliant with the Thumb Procedure Call
10862 Standard for all leaf functions. (A leaf function is one that does
10863 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10865 @item -mcallee-super-interworking
10866 @opindex mcallee-super-interworking
10867 Gives all externally visible functions in the file being compiled an ARM
10868 instruction set header which switches to Thumb mode before executing the
10869 rest of the function. This allows these functions to be called from
10870 non-interworking code. This option is not valid in AAPCS configurations
10871 because interworking is enabled by default.
10873 @item -mcaller-super-interworking
10874 @opindex mcaller-super-interworking
10875 Allows calls via function pointers (including virtual functions) to
10876 execute correctly regardless of whether the target code has been
10877 compiled for interworking or not. There is a small overhead in the cost
10878 of executing a function pointer if this option is enabled. This option
10879 is not valid in AAPCS configurations because interworking is enabled
10882 @item -mtp=@var{name}
10884 Specify the access model for the thread local storage pointer. The valid
10885 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10886 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10887 (supported in the arm6k architecture), and @option{auto}, which uses the
10888 best available method for the selected processor. The default setting is
10891 @item -mtls-dialect=@var{dialect}
10892 @opindex mtls-dialect
10893 Specify the dialect to use for accessing thread local storage. Two
10894 dialects are supported --- @option{gnu} and @option{gnu2}. The
10895 @option{gnu} dialect selects the original GNU scheme for supporting
10896 local and global dynamic TLS models. The @option{gnu2} dialect
10897 selects the GNU descriptor scheme, which provides better performance
10898 for shared libraries. The GNU descriptor scheme is compatible with
10899 the original scheme, but does require new assembler, linker and
10900 library support. Initial and local exec TLS models are unaffected by
10901 this option and always use the original scheme.
10903 @item -mword-relocations
10904 @opindex mword-relocations
10905 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
10906 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10907 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10910 @item -mfix-cortex-m3-ldrd
10911 @opindex mfix-cortex-m3-ldrd
10912 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10913 with overlapping destination and base registers are used. This option avoids
10914 generating these instructions. This option is enabled by default when
10915 @option{-mcpu=cortex-m3} is specified.
10920 @subsection AVR Options
10921 @cindex AVR Options
10923 These options are defined for AVR implementations:
10926 @item -mmcu=@var{mcu}
10928 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
10930 For a complete list of @var{mcu} values that are supported by avr-gcc,
10931 see the compiler output when called with the @code{--help=target}
10932 command line option.
10933 The default for this option is@tie{}@code{avr2}.
10935 avr-gcc supports the following AVR devices and ISAs:
10940 This ISA is implemented by the minimal AVR core and supported
10941 for assembler only.
10942 @*@var{mcu}@tie{}= @code{at90s1200},
10943 @code{attiny10}, @code{attiny11}, @code{attiny12}, @code{attiny15},
10947 ``Classic'' devices with up to 8@tie{}KiB of program memory.
10948 @*@var{mcu}@tie{}= @code{at90s2313}, @code{attiny26}, @code{at90c8534},
10952 ``Classic'' devices with up to 8@tie{}KiB of program memory and with
10953 the @code{MOVW} instruction.
10954 @*@var{mcu}@tie{}= @code{attiny2313}, @code{attiny261}, @code{attiny24},
10958 ``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory.
10959 @*@var{mcu}@tie{}= @code{at43usb355}, @code{at76c711}.
10962 ``Classic'' devices with 128@tie{}KiB of program memory.
10963 @*@var{mcu}@tie{}= @code{atmega103}, @code{at43usb320}.
10966 ``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program
10967 memory and with the @code{MOVW} instruction.
10968 @*@var{mcu}@tie{}= @code{at90usb162}, @code{atmega8u2},
10969 @code{attiny167}, @dots{}
10972 ``Enhanced'' devices with up to 8@tie{}KiB of program memory.
10973 @*@var{mcu}@tie{}= @code{atmega8}, @code{atmega88}, @code{at90pwm81},
10977 ``Enhanced'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory.
10978 @*@var{mcu}@tie{}= @code{atmega16}, @code{atmega6490}, @code{at90can64},
10982 ``Enhanced'' devices with 128@tie{}KiB of program memory.
10983 @*@var{mcu}@tie{}= @code{atmega128}, @code{at90can128}, @code{at90usb1287},
10987 ``Enhanced'' devices with 3-byte PC, i.e.@: with at least 256@tie{}KiB
10989 @*@var{mcu}@tie{}= @code{atmega2560}, @code{atmega2561}.
10994 @item -maccumulate-args
10995 @opindex maccumulate-args
10996 Accumulate outgoing function arguments and acquire/release the needed
10997 stack space for outgoing function arguments once in function
10998 prologue/epilogue. Without this option, outgoing arguments are pushed
10999 before calling a function and popped afterwards.
11001 Popping the arguments after the function call can be expensive on
11002 AVR so that accumulating the stack space might lead to smaller
11003 executables because arguments need not to be removed from the
11004 stack after such a function call.
11006 This option can lead to reduced code size for functions that perform
11007 several calls to functions that get their arguments on the stack like
11008 calls to printf-like functions.
11010 @item -mbranch-cost=@var{cost}
11011 @opindex mbranch-cost
11012 Set the branch costs for conditional branch instructions to
11013 @var{cost}. Reasonable values for @var{cost} are small, non-negative
11014 integers. The default branch cost is 0.
11016 @item -mcall-prologues
11017 @opindex mcall-prologues
11018 Functions prologues/epilogues expanded as call to appropriate
11019 subroutines. Code size will be smaller.
11023 Assume int to be 8-bit integer. This affects the sizes of all types: a
11024 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
11025 and long long will be 4 bytes. Please note that this option does not
11026 comply to the C standards, but it will provide you with smaller code
11029 @item -mno-interrupts
11030 @opindex mno-interrupts
11031 Generated code is not compatible with hardware interrupts.
11032 Code size will be smaller.
11036 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
11037 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
11038 Setting @code{-mrelax} just adds the @code{--relax} option to the
11039 linker command line when the linker is called.
11041 Jump relaxing is performed by the linker because jump offsets are not
11042 known before code is located. Therefore, the assembler code generated by the
11043 compiler will be the same, but the instructions in the executable may
11044 differ from instructions in the assembler code.
11046 @item -mshort-calls
11047 @opindex mshort-calls
11048 Use @code{RCALL}/@code{RJMP} instructions even on devices with
11049 16@tie{}KiB or more of program memory, i.e.@: on devices that
11050 have the @code{CALL} and @code{JMP} instructions.
11051 See also the @code{-mrelax} command line option.
11055 Use address register @code{X} in a way proposed by the hardware. This means
11056 that @code{X} will only be used in indirect, post-increment or
11057 pre-decrement addressing.
11059 Without this option, the @code{X} register may be used in the same way
11060 as @code{Y} or @code{Z} which then is emulated by additional
11062 For example, loading a value with @code{X+const} addressing with a
11063 small non-negative @code{const < 64} to a register @var{Rn} will be
11067 adiw r26, const ; X += const
11068 ld @var{Rn}, X ; @var{Rn} = *X
11069 sbiw r26, const ; X -= const
11073 @opindex mtiny-stack
11074 Only use the lower 8@tie{}bits of the stack pointer and assume that the high
11075 byte of SP is always zero.
11078 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
11080 Pointers in the implementation are 16@tie{}bits wide.
11081 The address of a function or label is represented as word address so
11082 that indirect jumps and calls can target any code address in the
11083 range of 64@tie{}Ki words.
11085 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
11086 bytes of program memory space, there is a special function register called
11087 @code{EIND} that serves as most significant part of the target address
11088 when @code{EICALL} or @code{EIJMP} instructions are used.
11090 Indirect jumps and calls on these devices are handled as follows by
11091 the compiler and are subject to some limitations:
11096 The compiler never sets @code{EIND}.
11099 The startup code from libgcc never sets @code{EIND}.
11100 Notice that startup code is a blend of code from libgcc and avr-libc.
11101 For the impact of avr-libc on @code{EIND}, see the
11102 @w{@uref{http://nongnu.org/avr-libc/user-manual,avr-libc user manual}}.
11105 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
11106 instructions or might read @code{EIND} directly in order to emulate an
11107 indirect call/jump by means of a @code{RET} instruction.
11110 The compiler assumes that @code{EIND} never changes during the startup
11111 code or during the application. In particular, @code{EIND} is not
11112 saved/restored in function or interrupt service routine
11116 It is legitimate for user-specific startup code to set up @code{EIND}
11117 early, for example by means of initialization code located in
11118 section @code{.init3}. Such code runs prior to general startup code
11119 that initializes RAM and calls constructors.
11122 For indirect calls to functions and computed goto, the linker will
11123 generate @emph{stubs}. Stubs are jump pads sometimes also called
11124 @emph{trampolines}. Thus, the indirect call/jump will jump to such a stub.
11125 The stub contains a direct jump to the desired address.
11128 Stubs will be generated automatically by the linker if
11129 the following two conditions are met:
11132 @item The address of a label is taken by means of the @code{gs} modifier
11133 (short for @emph{generate stubs}) like so:
11135 LDI r24, lo8(gs(@var{func}))
11136 LDI r25, hi8(gs(@var{func}))
11138 @item The final location of that label is in a code segment
11139 @emph{outside} the segment where the stubs are located.
11143 The compiler will emit such @code{gs} modifiers for code labels in the
11144 following situations:
11146 @item Taking address of a function or code label.
11147 @item Computed goto.
11148 @item If prologue-save function is used, see @option{-mcall-prologues}
11149 command-line option.
11150 @item Switch/case dispatch tables. If you do not want such dispatch
11151 tables you can specify the @option{-fno-jump-tables} command-line option.
11152 @item C and C++ constructors/destructors called during startup/shutdown.
11153 @item If the tools hit a @code{gs()} modifier explained above.
11157 The default linker script is arranged for code with @code{EIND = 0}.
11158 If code is supposed to work for a setup with @code{EIND != 0}, a custom
11159 linker script has to be used in order to place the sections whose
11160 name start with @code{.trampolines} into the segment where @code{EIND}
11164 Jumping to non-symbolic addresses like so is @emph{not} supported:
11169 /* Call function at word address 0x2 */
11170 return ((int(*)(void)) 0x2)();
11174 Instead, a stub has to be set up, i.e.@: the function has to be called
11175 through a symbol (@code{func_4} in the example):
11180 extern int func_4 (void);
11182 /* Call function at byte address 0x4 */
11187 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
11188 Alternatively, @code{func_4} can be defined in the linker script.
11191 @subsubsection AVR Built-in Macros
11193 avr-gcc defines several built-in macros so that the user code can test
11194 for presence of absence of features. Almost any of the following
11195 built-in macros are deduced from device capabilities and thus
11196 triggered by the @code{-mmcu=} command-line option.
11198 For even more AVR-specific built-in macros see
11199 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
11203 @item __AVR_@var{Device}__
11204 Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects
11205 the device's name. For example, @code{-mmcu=atmega8} will define the
11206 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
11207 @code{__AVR_ATtiny261A__}, etc.
11209 The built-in macros' names follow
11210 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
11211 the device name as from the AVR user manual. The difference between
11212 @var{Device} in the built-in macro and @var{device} in
11213 @code{-mmcu=@var{device}} is that the latter is always lower case.
11215 @item __AVR_HAVE_RAMPZ__
11216 @item __AVR_HAVE_ELPM__
11217 The device has the @code{RAMPZ} special function register and thus the
11218 @code{ELPM} instruction.
11220 @item __AVR_HAVE_ELPMX__
11221 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
11222 R@var{n},Z+} instructions.
11224 @item __AVR_HAVE_MOVW__
11225 The device has the @code{MOVW} instruction to perform 16-bit
11226 register-register moves.
11228 @item __AVR_HAVE_LPMX__
11229 The device has the @code{LPM R@var{n},Z} and @code{LPM
11230 R@var{n},Z+} instructions.
11232 @item __AVR_HAVE_MUL__
11233 The device has a hardware multiplier.
11235 @item __AVR_HAVE_JMP_CALL__
11236 The device has the @code{JMP} and @code{CALL} instructions.
11237 This is the case for devices with at least 16@tie{}KiB of program
11238 memory and if @code{-mshort-calls} is not set.
11240 @item __AVR_HAVE_EIJMP_EICALL__
11241 @item __AVR_3_BYTE_PC__
11242 The device has the @code{EIJMP} and @code{EICALL} instructions.
11243 This is the case for devices with at least 256@tie{}KiB of program memory.
11244 This also means that the program counter
11245 (PC) is 3@tie{}bytes wide.
11247 @item __AVR_2_BYTE_PC__
11248 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
11249 with up to 128@tie{}KiB of program memory.
11251 @item __AVR_HAVE_8BIT_SP__
11252 @item __AVR_HAVE_16BIT_SP__
11253 The stack pointer (SP) is respectively 8 or 16 bits wide.
11254 The definition of these macros is affected by @code{-mtiny-stack}.
11256 @item __NO_INTERRUPTS__
11257 This macro reflects the @code{-mno-interrupts} command line option.
11259 @item __AVR_ERRATA_SKIP__
11260 @item __AVR_ERRATA_SKIP_JMP_CALL__
11261 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
11262 instructions because of a hardware erratum. Skip instructions are
11263 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
11264 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
11267 @item __AVR_SFR_OFFSET__=@var{offset}
11268 Instructions that can address I/O special function registers directly
11269 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
11270 address as if addressed by an instruction to access RAM like @code{LD}
11271 or @code{STS}. This offset depends on the device architecture and has
11272 to be subtracted from the RAM address in order to get the
11273 respective I/O@tie{}address.
11277 @node Blackfin Options
11278 @subsection Blackfin Options
11279 @cindex Blackfin Options
11282 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
11284 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
11285 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
11286 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
11287 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
11288 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
11289 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
11290 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
11291 @samp{bf561}, @samp{bf592}.
11292 The optional @var{sirevision} specifies the silicon revision of the target
11293 Blackfin processor. Any workarounds available for the targeted silicon revision
11294 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
11295 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
11296 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
11297 hexadecimal digits representing the major and minor numbers in the silicon
11298 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
11299 is not defined. If @var{sirevision} is @samp{any}, the
11300 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
11301 If this optional @var{sirevision} is not used, GCC assumes the latest known
11302 silicon revision of the targeted Blackfin processor.
11304 Support for @samp{bf561} is incomplete. For @samp{bf561},
11305 Only the processor macro is defined.
11306 Without this option, @samp{bf532} is used as the processor by default.
11307 The corresponding predefined processor macros for @var{cpu} is to
11308 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
11309 provided by libgloss to be linked in if @option{-msim} is not given.
11313 Specifies that the program will be run on the simulator. This causes
11314 the simulator BSP provided by libgloss to be linked in. This option
11315 has effect only for @samp{bfin-elf} toolchain.
11316 Certain other options, such as @option{-mid-shared-library} and
11317 @option{-mfdpic}, imply @option{-msim}.
11319 @item -momit-leaf-frame-pointer
11320 @opindex momit-leaf-frame-pointer
11321 Don't keep the frame pointer in a register for leaf functions. This
11322 avoids the instructions to save, set up and restore frame pointers and
11323 makes an extra register available in leaf functions. The option
11324 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
11325 which might make debugging harder.
11327 @item -mspecld-anomaly
11328 @opindex mspecld-anomaly
11329 When enabled, the compiler will ensure that the generated code does not
11330 contain speculative loads after jump instructions. If this option is used,
11331 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
11333 @item -mno-specld-anomaly
11334 @opindex mno-specld-anomaly
11335 Don't generate extra code to prevent speculative loads from occurring.
11337 @item -mcsync-anomaly
11338 @opindex mcsync-anomaly
11339 When enabled, the compiler will ensure that the generated code does not
11340 contain CSYNC or SSYNC instructions too soon after conditional branches.
11341 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
11343 @item -mno-csync-anomaly
11344 @opindex mno-csync-anomaly
11345 Don't generate extra code to prevent CSYNC or SSYNC instructions from
11346 occurring too soon after a conditional branch.
11350 When enabled, the compiler is free to take advantage of the knowledge that
11351 the entire program fits into the low 64k of memory.
11354 @opindex mno-low-64k
11355 Assume that the program is arbitrarily large. This is the default.
11357 @item -mstack-check-l1
11358 @opindex mstack-check-l1
11359 Do stack checking using information placed into L1 scratchpad memory by the
11362 @item -mid-shared-library
11363 @opindex mid-shared-library
11364 Generate code that supports shared libraries via the library ID method.
11365 This allows for execute in place and shared libraries in an environment
11366 without virtual memory management. This option implies @option{-fPIC}.
11367 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11369 @item -mno-id-shared-library
11370 @opindex mno-id-shared-library
11371 Generate code that doesn't assume ID based shared libraries are being used.
11372 This is the default.
11374 @item -mleaf-id-shared-library
11375 @opindex mleaf-id-shared-library
11376 Generate code that supports shared libraries via the library ID method,
11377 but assumes that this library or executable won't link against any other
11378 ID shared libraries. That allows the compiler to use faster code for jumps
11381 @item -mno-leaf-id-shared-library
11382 @opindex mno-leaf-id-shared-library
11383 Do not assume that the code being compiled won't link against any ID shared
11384 libraries. Slower code will be generated for jump and call insns.
11386 @item -mshared-library-id=n
11387 @opindex mshared-library-id
11388 Specified the identification number of the ID based shared library being
11389 compiled. Specifying a value of 0 will generate more compact code, specifying
11390 other values will force the allocation of that number to the current
11391 library but is no more space or time efficient than omitting this option.
11395 Generate code that allows the data segment to be located in a different
11396 area of memory from the text segment. This allows for execute in place in
11397 an environment without virtual memory management by eliminating relocations
11398 against the text section.
11400 @item -mno-sep-data
11401 @opindex mno-sep-data
11402 Generate code that assumes that the data segment follows the text segment.
11403 This is the default.
11406 @itemx -mno-long-calls
11407 @opindex mlong-calls
11408 @opindex mno-long-calls
11409 Tells the compiler to perform function calls by first loading the
11410 address of the function into a register and then performing a subroutine
11411 call on this register. This switch is needed if the target function
11412 lies outside of the 24-bit addressing range of the offset-based
11413 version of subroutine call instruction.
11415 This feature is not enabled by default. Specifying
11416 @option{-mno-long-calls} will restore the default behavior. Note these
11417 switches have no effect on how the compiler generates code to handle
11418 function calls via function pointers.
11422 Link with the fast floating-point library. This library relaxes some of
11423 the IEEE floating-point standard's rules for checking inputs against
11424 Not-a-Number (NAN), in the interest of performance.
11427 @opindex minline-plt
11428 Enable inlining of PLT entries in function calls to functions that are
11429 not known to bind locally. It has no effect without @option{-mfdpic}.
11432 @opindex mmulticore
11433 Build standalone application for multicore Blackfin processor. Proper
11434 start files and link scripts will be used to support multicore.
11435 This option defines @code{__BFIN_MULTICORE}. It can only be used with
11436 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
11437 @option{-mcorea} or @option{-mcoreb}. If it's used without
11438 @option{-mcorea} or @option{-mcoreb}, single application/dual core
11439 programming model is used. In this model, the main function of Core B
11440 should be named as coreb_main. If it's used with @option{-mcorea} or
11441 @option{-mcoreb}, one application per core programming model is used.
11442 If this option is not used, single core application programming
11447 Build standalone application for Core A of BF561 when using
11448 one application per core programming model. Proper start files
11449 and link scripts will be used to support Core A. This option
11450 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
11454 Build standalone application for Core B of BF561 when using
11455 one application per core programming model. Proper start files
11456 and link scripts will be used to support Core B. This option
11457 defines @code{__BFIN_COREB}. When this option is used, coreb_main
11458 should be used instead of main. It must be used with
11459 @option{-mmulticore}.
11463 Build standalone application for SDRAM. Proper start files and
11464 link scripts will be used to put the application into SDRAM.
11465 Loader should initialize SDRAM before loading the application
11466 into SDRAM. This option defines @code{__BFIN_SDRAM}.
11470 Assume that ICPLBs are enabled at run time. This has an effect on certain
11471 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
11472 are enabled; for standalone applications the default is off.
11476 @subsection C6X Options
11477 @cindex C6X Options
11480 @item -march=@var{name}
11482 This specifies the name of the target architecture. GCC uses this
11483 name to determine what kind of instructions it can emit when generating
11484 assembly code. Permissible names are: @samp{c62x},
11485 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
11488 @opindex mbig-endian
11489 Generate code for a big-endian target.
11491 @item -mlittle-endian
11492 @opindex mlittle-endian
11493 Generate code for a little-endian target. This is the default.
11497 Choose startup files and linker script suitable for the simulator.
11499 @item -msdata=default
11500 @opindex msdata=default
11501 Put small global and static data in the @samp{.neardata} section,
11502 which is pointed to by register @code{B14}. Put small uninitialized
11503 global and static data in the @samp{.bss} section, which is adjacent
11504 to the @samp{.neardata} section. Put small read-only data into the
11505 @samp{.rodata} section. The corresponding sections used for large
11506 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
11509 @opindex msdata=all
11510 Put all data, not just small objets, into the sections reserved for
11511 small data, and use addressing relative to the @code{B14} register to
11515 @opindex msdata=none
11516 Make no use of the sections reserved for small data, and use absolute
11517 addresses to access all data. Put all initialized global and static
11518 data in the @samp{.fardata} section, and all uninitialized data in the
11519 @samp{.far} section. Put all constant data into the @samp{.const}
11524 @subsection CRIS Options
11525 @cindex CRIS Options
11527 These options are defined specifically for the CRIS ports.
11530 @item -march=@var{architecture-type}
11531 @itemx -mcpu=@var{architecture-type}
11534 Generate code for the specified architecture. The choices for
11535 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
11536 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
11537 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
11540 @item -mtune=@var{architecture-type}
11542 Tune to @var{architecture-type} everything applicable about the generated
11543 code, except for the ABI and the set of available instructions. The
11544 choices for @var{architecture-type} are the same as for
11545 @option{-march=@var{architecture-type}}.
11547 @item -mmax-stack-frame=@var{n}
11548 @opindex mmax-stack-frame
11549 Warn when the stack frame of a function exceeds @var{n} bytes.
11555 The options @option{-metrax4} and @option{-metrax100} are synonyms for
11556 @option{-march=v3} and @option{-march=v8} respectively.
11558 @item -mmul-bug-workaround
11559 @itemx -mno-mul-bug-workaround
11560 @opindex mmul-bug-workaround
11561 @opindex mno-mul-bug-workaround
11562 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
11563 models where it applies. This option is active by default.
11567 Enable CRIS-specific verbose debug-related information in the assembly
11568 code. This option also has the effect to turn off the @samp{#NO_APP}
11569 formatted-code indicator to the assembler at the beginning of the
11574 Do not use condition-code results from previous instruction; always emit
11575 compare and test instructions before use of condition codes.
11577 @item -mno-side-effects
11578 @opindex mno-side-effects
11579 Do not emit instructions with side-effects in addressing modes other than
11582 @item -mstack-align
11583 @itemx -mno-stack-align
11584 @itemx -mdata-align
11585 @itemx -mno-data-align
11586 @itemx -mconst-align
11587 @itemx -mno-const-align
11588 @opindex mstack-align
11589 @opindex mno-stack-align
11590 @opindex mdata-align
11591 @opindex mno-data-align
11592 @opindex mconst-align
11593 @opindex mno-const-align
11594 These options (no-options) arranges (eliminate arrangements) for the
11595 stack-frame, individual data and constants to be aligned for the maximum
11596 single data access size for the chosen CPU model. The default is to
11597 arrange for 32-bit alignment. ABI details such as structure layout are
11598 not affected by these options.
11606 Similar to the stack- data- and const-align options above, these options
11607 arrange for stack-frame, writable data and constants to all be 32-bit,
11608 16-bit or 8-bit aligned. The default is 32-bit alignment.
11610 @item -mno-prologue-epilogue
11611 @itemx -mprologue-epilogue
11612 @opindex mno-prologue-epilogue
11613 @opindex mprologue-epilogue
11614 With @option{-mno-prologue-epilogue}, the normal function prologue and
11615 epilogue which set up the stack frame are omitted and no return
11616 instructions or return sequences are generated in the code. Use this
11617 option only together with visual inspection of the compiled code: no
11618 warnings or errors are generated when call-saved registers must be saved,
11619 or storage for local variable needs to be allocated.
11623 @opindex mno-gotplt
11625 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11626 instruction sequences that load addresses for functions from the PLT part
11627 of the GOT rather than (traditional on other architectures) calls to the
11628 PLT@. The default is @option{-mgotplt}.
11632 Legacy no-op option only recognized with the cris-axis-elf and
11633 cris-axis-linux-gnu targets.
11637 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11641 This option, recognized for the cris-axis-elf arranges
11642 to link with input-output functions from a simulator library. Code,
11643 initialized data and zero-initialized data are allocated consecutively.
11647 Like @option{-sim}, but pass linker options to locate initialized data at
11648 0x40000000 and zero-initialized data at 0x80000000.
11652 @subsection CR16 Options
11653 @cindex CR16 Options
11655 These options are defined specifically for the CR16 ports.
11661 Enable the use of multiply-accumulate instructions. Disabled by default.
11665 @opindex mcr16cplus
11667 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
11672 Links the library libsim.a which is in compatible with simulator. Applicable
11673 to elf compiler only.
11677 Choose integer type as 32-bit wide.
11681 Generates sbit/cbit instructions for bit manipulations.
11683 @item -mdata-model=@var{model}
11684 @opindex mdata-model
11685 Choose a data model. The choices for @var{model} are @samp{near},
11686 @samp{far} or @samp{medium}. @samp{medium} is default.
11687 However, @samp{far} is not valid when -mcr16c option is chosen as
11688 CR16C architecture does not support far data model.
11691 @node Darwin Options
11692 @subsection Darwin Options
11693 @cindex Darwin options
11695 These options are defined for all architectures running the Darwin operating
11698 FSF GCC on Darwin does not create ``fat'' object files; it will create
11699 an object file for the single architecture that it was built to
11700 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11701 @option{-arch} options are used; it does so by running the compiler or
11702 linker multiple times and joining the results together with
11705 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11706 @samp{i686}) is determined by the flags that specify the ISA
11707 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11708 @option{-force_cpusubtype_ALL} option can be used to override this.
11710 The Darwin tools vary in their behavior when presented with an ISA
11711 mismatch. The assembler, @file{as}, will only permit instructions to
11712 be used that are valid for the subtype of the file it is generating,
11713 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11714 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11715 and print an error if asked to create a shared library with a less
11716 restrictive subtype than its input files (for instance, trying to put
11717 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11718 for executables, @file{ld}, will quietly give the executable the most
11719 restrictive subtype of any of its input files.
11724 Add the framework directory @var{dir} to the head of the list of
11725 directories to be searched for header files. These directories are
11726 interleaved with those specified by @option{-I} options and are
11727 scanned in a left-to-right order.
11729 A framework directory is a directory with frameworks in it. A
11730 framework is a directory with a @samp{"Headers"} and/or
11731 @samp{"PrivateHeaders"} directory contained directly in it that ends
11732 in @samp{".framework"}. The name of a framework is the name of this
11733 directory excluding the @samp{".framework"}. Headers associated with
11734 the framework are found in one of those two directories, with
11735 @samp{"Headers"} being searched first. A subframework is a framework
11736 directory that is in a framework's @samp{"Frameworks"} directory.
11737 Includes of subframework headers can only appear in a header of a
11738 framework that contains the subframework, or in a sibling subframework
11739 header. Two subframeworks are siblings if they occur in the same
11740 framework. A subframework should not have the same name as a
11741 framework, a warning will be issued if this is violated. Currently a
11742 subframework cannot have subframeworks, in the future, the mechanism
11743 may be extended to support this. The standard frameworks can be found
11744 in @samp{"/System/Library/Frameworks"} and
11745 @samp{"/Library/Frameworks"}. An example include looks like
11746 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11747 the name of the framework and header.h is found in the
11748 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11750 @item -iframework@var{dir}
11751 @opindex iframework
11752 Like @option{-F} except the directory is a treated as a system
11753 directory. The main difference between this @option{-iframework} and
11754 @option{-F} is that with @option{-iframework} the compiler does not
11755 warn about constructs contained within header files found via
11756 @var{dir}. This option is valid only for the C family of languages.
11760 Emit debugging information for symbols that are used. For STABS
11761 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11762 This is by default ON@.
11766 Emit debugging information for all symbols and types.
11768 @item -mmacosx-version-min=@var{version}
11769 The earliest version of MacOS X that this executable will run on
11770 is @var{version}. Typical values of @var{version} include @code{10.1},
11771 @code{10.2}, and @code{10.3.9}.
11773 If the compiler was built to use the system's headers by default,
11774 then the default for this option is the system version on which the
11775 compiler is running, otherwise the default is to make choices that
11776 are compatible with as many systems and code bases as possible.
11780 Enable kernel development mode. The @option{-mkernel} option sets
11781 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11782 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11783 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11784 applicable. This mode also sets @option{-mno-altivec},
11785 @option{-msoft-float}, @option{-fno-builtin} and
11786 @option{-mlong-branch} for PowerPC targets.
11788 @item -mone-byte-bool
11789 @opindex mone-byte-bool
11790 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11791 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11792 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11793 option has no effect on x86.
11795 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11796 to generate code that is not binary compatible with code generated
11797 without that switch. Using this switch may require recompiling all
11798 other modules in a program, including system libraries. Use this
11799 switch to conform to a non-default data model.
11801 @item -mfix-and-continue
11802 @itemx -ffix-and-continue
11803 @itemx -findirect-data
11804 @opindex mfix-and-continue
11805 @opindex ffix-and-continue
11806 @opindex findirect-data
11807 Generate code suitable for fast turn around development. Needed to
11808 enable gdb to dynamically load @code{.o} files into already running
11809 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11810 are provided for backwards compatibility.
11814 Loads all members of static archive libraries.
11815 See man ld(1) for more information.
11817 @item -arch_errors_fatal
11818 @opindex arch_errors_fatal
11819 Cause the errors having to do with files that have the wrong architecture
11822 @item -bind_at_load
11823 @opindex bind_at_load
11824 Causes the output file to be marked such that the dynamic linker will
11825 bind all undefined references when the file is loaded or launched.
11829 Produce a Mach-o bundle format file.
11830 See man ld(1) for more information.
11832 @item -bundle_loader @var{executable}
11833 @opindex bundle_loader
11834 This option specifies the @var{executable} that will be loading the build
11835 output file being linked. See man ld(1) for more information.
11838 @opindex dynamiclib
11839 When passed this option, GCC will produce a dynamic library instead of
11840 an executable when linking, using the Darwin @file{libtool} command.
11842 @item -force_cpusubtype_ALL
11843 @opindex force_cpusubtype_ALL
11844 This causes GCC's output file to have the @var{ALL} subtype, instead of
11845 one controlled by the @option{-mcpu} or @option{-march} option.
11847 @item -allowable_client @var{client_name}
11848 @itemx -client_name
11849 @itemx -compatibility_version
11850 @itemx -current_version
11852 @itemx -dependency-file
11854 @itemx -dylinker_install_name
11856 @itemx -exported_symbols_list
11859 @itemx -flat_namespace
11860 @itemx -force_flat_namespace
11861 @itemx -headerpad_max_install_names
11864 @itemx -install_name
11865 @itemx -keep_private_externs
11866 @itemx -multi_module
11867 @itemx -multiply_defined
11868 @itemx -multiply_defined_unused
11871 @itemx -no_dead_strip_inits_and_terms
11872 @itemx -nofixprebinding
11873 @itemx -nomultidefs
11875 @itemx -noseglinkedit
11876 @itemx -pagezero_size
11878 @itemx -prebind_all_twolevel_modules
11879 @itemx -private_bundle
11881 @itemx -read_only_relocs
11883 @itemx -sectobjectsymbols
11887 @itemx -sectobjectsymbols
11890 @itemx -segs_read_only_addr
11892 @itemx -segs_read_write_addr
11893 @itemx -seg_addr_table
11894 @itemx -seg_addr_table_filename
11895 @itemx -seglinkedit
11897 @itemx -segs_read_only_addr
11898 @itemx -segs_read_write_addr
11899 @itemx -single_module
11901 @itemx -sub_library
11903 @itemx -sub_umbrella
11904 @itemx -twolevel_namespace
11907 @itemx -unexported_symbols_list
11908 @itemx -weak_reference_mismatches
11909 @itemx -whatsloaded
11910 @opindex allowable_client
11911 @opindex client_name
11912 @opindex compatibility_version
11913 @opindex current_version
11914 @opindex dead_strip
11915 @opindex dependency-file
11916 @opindex dylib_file
11917 @opindex dylinker_install_name
11919 @opindex exported_symbols_list
11921 @opindex flat_namespace
11922 @opindex force_flat_namespace
11923 @opindex headerpad_max_install_names
11924 @opindex image_base
11926 @opindex install_name
11927 @opindex keep_private_externs
11928 @opindex multi_module
11929 @opindex multiply_defined
11930 @opindex multiply_defined_unused
11931 @opindex noall_load
11932 @opindex no_dead_strip_inits_and_terms
11933 @opindex nofixprebinding
11934 @opindex nomultidefs
11936 @opindex noseglinkedit
11937 @opindex pagezero_size
11939 @opindex prebind_all_twolevel_modules
11940 @opindex private_bundle
11941 @opindex read_only_relocs
11943 @opindex sectobjectsymbols
11946 @opindex sectcreate
11947 @opindex sectobjectsymbols
11950 @opindex segs_read_only_addr
11951 @opindex segs_read_write_addr
11952 @opindex seg_addr_table
11953 @opindex seg_addr_table_filename
11954 @opindex seglinkedit
11956 @opindex segs_read_only_addr
11957 @opindex segs_read_write_addr
11958 @opindex single_module
11960 @opindex sub_library
11961 @opindex sub_umbrella
11962 @opindex twolevel_namespace
11965 @opindex unexported_symbols_list
11966 @opindex weak_reference_mismatches
11967 @opindex whatsloaded
11968 These options are passed to the Darwin linker. The Darwin linker man page
11969 describes them in detail.
11972 @node DEC Alpha Options
11973 @subsection DEC Alpha Options
11975 These @samp{-m} options are defined for the DEC Alpha implementations:
11978 @item -mno-soft-float
11979 @itemx -msoft-float
11980 @opindex mno-soft-float
11981 @opindex msoft-float
11982 Use (do not use) the hardware floating-point instructions for
11983 floating-point operations. When @option{-msoft-float} is specified,
11984 functions in @file{libgcc.a} will be used to perform floating-point
11985 operations. Unless they are replaced by routines that emulate the
11986 floating-point operations, or compiled in such a way as to call such
11987 emulations routines, these routines will issue floating-point
11988 operations. If you are compiling for an Alpha without floating-point
11989 operations, you must ensure that the library is built so as not to call
11992 Note that Alpha implementations without floating-point operations are
11993 required to have floating-point registers.
11996 @itemx -mno-fp-regs
11998 @opindex mno-fp-regs
11999 Generate code that uses (does not use) the floating-point register set.
12000 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
12001 register set is not used, floating-point operands are passed in integer
12002 registers as if they were integers and floating-point results are passed
12003 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
12004 so any function with a floating-point argument or return value called by code
12005 compiled with @option{-mno-fp-regs} must also be compiled with that
12008 A typical use of this option is building a kernel that does not use,
12009 and hence need not save and restore, any floating-point registers.
12013 The Alpha architecture implements floating-point hardware optimized for
12014 maximum performance. It is mostly compliant with the IEEE floating-point
12015 standard. However, for full compliance, software assistance is
12016 required. This option generates code fully IEEE-compliant code
12017 @emph{except} that the @var{inexact-flag} is not maintained (see below).
12018 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
12019 defined during compilation. The resulting code is less efficient but is
12020 able to correctly support denormalized numbers and exceptional IEEE
12021 values such as not-a-number and plus/minus infinity. Other Alpha
12022 compilers call this option @option{-ieee_with_no_inexact}.
12024 @item -mieee-with-inexact
12025 @opindex mieee-with-inexact
12026 This is like @option{-mieee} except the generated code also maintains
12027 the IEEE @var{inexact-flag}. Turning on this option causes the
12028 generated code to implement fully-compliant IEEE math. In addition to
12029 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
12030 macro. On some Alpha implementations the resulting code may execute
12031 significantly slower than the code generated by default. Since there is
12032 very little code that depends on the @var{inexact-flag}, you should
12033 normally not specify this option. Other Alpha compilers call this
12034 option @option{-ieee_with_inexact}.
12036 @item -mfp-trap-mode=@var{trap-mode}
12037 @opindex mfp-trap-mode
12038 This option controls what floating-point related traps are enabled.
12039 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
12040 The trap mode can be set to one of four values:
12044 This is the default (normal) setting. The only traps that are enabled
12045 are the ones that cannot be disabled in software (e.g., division by zero
12049 In addition to the traps enabled by @samp{n}, underflow traps are enabled
12053 Like @samp{u}, but the instructions are marked to be safe for software
12054 completion (see Alpha architecture manual for details).
12057 Like @samp{su}, but inexact traps are enabled as well.
12060 @item -mfp-rounding-mode=@var{rounding-mode}
12061 @opindex mfp-rounding-mode
12062 Selects the IEEE rounding mode. Other Alpha compilers call this option
12063 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
12068 Normal IEEE rounding mode. Floating-point numbers are rounded towards
12069 the nearest machine number or towards the even machine number in case
12073 Round towards minus infinity.
12076 Chopped rounding mode. Floating-point numbers are rounded towards zero.
12079 Dynamic rounding mode. A field in the floating-point control register
12080 (@var{fpcr}, see Alpha architecture reference manual) controls the
12081 rounding mode in effect. The C library initializes this register for
12082 rounding towards plus infinity. Thus, unless your program modifies the
12083 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
12086 @item -mtrap-precision=@var{trap-precision}
12087 @opindex mtrap-precision
12088 In the Alpha architecture, floating-point traps are imprecise. This
12089 means without software assistance it is impossible to recover from a
12090 floating trap and program execution normally needs to be terminated.
12091 GCC can generate code that can assist operating system trap handlers
12092 in determining the exact location that caused a floating-point trap.
12093 Depending on the requirements of an application, different levels of
12094 precisions can be selected:
12098 Program precision. This option is the default and means a trap handler
12099 can only identify which program caused a floating-point exception.
12102 Function precision. The trap handler can determine the function that
12103 caused a floating-point exception.
12106 Instruction precision. The trap handler can determine the exact
12107 instruction that caused a floating-point exception.
12110 Other Alpha compilers provide the equivalent options called
12111 @option{-scope_safe} and @option{-resumption_safe}.
12113 @item -mieee-conformant
12114 @opindex mieee-conformant
12115 This option marks the generated code as IEEE conformant. You must not
12116 use this option unless you also specify @option{-mtrap-precision=i} and either
12117 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
12118 is to emit the line @samp{.eflag 48} in the function prologue of the
12119 generated assembly file. Under DEC Unix, this has the effect that
12120 IEEE-conformant math library routines will be linked in.
12122 @item -mbuild-constants
12123 @opindex mbuild-constants
12124 Normally GCC examines a 32- or 64-bit integer constant to
12125 see if it can construct it from smaller constants in two or three
12126 instructions. If it cannot, it will output the constant as a literal and
12127 generate code to load it from the data segment at run time.
12129 Use this option to require GCC to construct @emph{all} integer constants
12130 using code, even if it takes more instructions (the maximum is six).
12132 You would typically use this option to build a shared library dynamic
12133 loader. Itself a shared library, it must relocate itself in memory
12134 before it can find the variables and constants in its own data segment.
12140 Select whether to generate code to be assembled by the vendor-supplied
12141 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
12159 Indicate whether GCC should generate code to use the optional BWX,
12160 CIX, FIX and MAX instruction sets. The default is to use the instruction
12161 sets supported by the CPU type specified via @option{-mcpu=} option or that
12162 of the CPU on which GCC was built if none was specified.
12165 @itemx -mfloat-ieee
12166 @opindex mfloat-vax
12167 @opindex mfloat-ieee
12168 Generate code that uses (does not use) VAX F and G floating-point
12169 arithmetic instead of IEEE single and double precision.
12171 @item -mexplicit-relocs
12172 @itemx -mno-explicit-relocs
12173 @opindex mexplicit-relocs
12174 @opindex mno-explicit-relocs
12175 Older Alpha assemblers provided no way to generate symbol relocations
12176 except via assembler macros. Use of these macros does not allow
12177 optimal instruction scheduling. GNU binutils as of version 2.12
12178 supports a new syntax that allows the compiler to explicitly mark
12179 which relocations should apply to which instructions. This option
12180 is mostly useful for debugging, as GCC detects the capabilities of
12181 the assembler when it is built and sets the default accordingly.
12184 @itemx -mlarge-data
12185 @opindex msmall-data
12186 @opindex mlarge-data
12187 When @option{-mexplicit-relocs} is in effect, static data is
12188 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
12189 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
12190 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
12191 16-bit relocations off of the @code{$gp} register. This limits the
12192 size of the small data area to 64KB, but allows the variables to be
12193 directly accessed via a single instruction.
12195 The default is @option{-mlarge-data}. With this option the data area
12196 is limited to just below 2GB@. Programs that require more than 2GB of
12197 data must use @code{malloc} or @code{mmap} to allocate the data in the
12198 heap instead of in the program's data segment.
12200 When generating code for shared libraries, @option{-fpic} implies
12201 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
12204 @itemx -mlarge-text
12205 @opindex msmall-text
12206 @opindex mlarge-text
12207 When @option{-msmall-text} is used, the compiler assumes that the
12208 code of the entire program (or shared library) fits in 4MB, and is
12209 thus reachable with a branch instruction. When @option{-msmall-data}
12210 is used, the compiler can assume that all local symbols share the
12211 same @code{$gp} value, and thus reduce the number of instructions
12212 required for a function call from 4 to 1.
12214 The default is @option{-mlarge-text}.
12216 @item -mcpu=@var{cpu_type}
12218 Set the instruction set and instruction scheduling parameters for
12219 machine type @var{cpu_type}. You can specify either the @samp{EV}
12220 style name or the corresponding chip number. GCC supports scheduling
12221 parameters for the EV4, EV5 and EV6 family of processors and will
12222 choose the default values for the instruction set from the processor
12223 you specify. If you do not specify a processor type, GCC will default
12224 to the processor on which the compiler was built.
12226 Supported values for @var{cpu_type} are
12232 Schedules as an EV4 and has no instruction set extensions.
12236 Schedules as an EV5 and has no instruction set extensions.
12240 Schedules as an EV5 and supports the BWX extension.
12245 Schedules as an EV5 and supports the BWX and MAX extensions.
12249 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
12253 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
12256 Native toolchains also support the value @samp{native},
12257 which selects the best architecture option for the host processor.
12258 @option{-mcpu=native} has no effect if GCC does not recognize
12261 @item -mtune=@var{cpu_type}
12263 Set only the instruction scheduling parameters for machine type
12264 @var{cpu_type}. The instruction set is not changed.
12266 Native toolchains also support the value @samp{native},
12267 which selects the best architecture option for the host processor.
12268 @option{-mtune=native} has no effect if GCC does not recognize
12271 @item -mmemory-latency=@var{time}
12272 @opindex mmemory-latency
12273 Sets the latency the scheduler should assume for typical memory
12274 references as seen by the application. This number is highly
12275 dependent on the memory access patterns used by the application
12276 and the size of the external cache on the machine.
12278 Valid options for @var{time} are
12282 A decimal number representing clock cycles.
12288 The compiler contains estimates of the number of clock cycles for
12289 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
12290 (also called Dcache, Scache, and Bcache), as well as to main memory.
12291 Note that L3 is only valid for EV5.
12296 @node DEC Alpha/VMS Options
12297 @subsection DEC Alpha/VMS Options
12299 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
12302 @item -mvms-return-codes
12303 @opindex mvms-return-codes
12304 Return VMS condition codes from main. The default is to return POSIX
12305 style condition (e.g.@: error) codes.
12307 @item -mdebug-main=@var{prefix}
12308 @opindex mdebug-main=@var{prefix}
12309 Flag the first routine whose name starts with @var{prefix} as the main
12310 routine for the debugger.
12314 Default to 64-bit memory allocation routines.
12318 @subsection FR30 Options
12319 @cindex FR30 Options
12321 These options are defined specifically for the FR30 port.
12325 @item -msmall-model
12326 @opindex msmall-model
12327 Use the small address space model. This can produce smaller code, but
12328 it does assume that all symbolic values and addresses will fit into a
12333 Assume that runtime support has been provided and so there is no need
12334 to include the simulator library (@file{libsim.a}) on the linker
12340 @subsection FRV Options
12341 @cindex FRV Options
12347 Only use the first 32 general-purpose registers.
12352 Use all 64 general-purpose registers.
12357 Use only the first 32 floating-point registers.
12362 Use all 64 floating-point registers.
12365 @opindex mhard-float
12367 Use hardware instructions for floating-point operations.
12370 @opindex msoft-float
12372 Use library routines for floating-point operations.
12377 Dynamically allocate condition code registers.
12382 Do not try to dynamically allocate condition code registers, only
12383 use @code{icc0} and @code{fcc0}.
12388 Change ABI to use double word insns.
12393 Do not use double word instructions.
12398 Use floating-point double instructions.
12401 @opindex mno-double
12403 Do not use floating-point double instructions.
12408 Use media instructions.
12413 Do not use media instructions.
12418 Use multiply and add/subtract instructions.
12421 @opindex mno-muladd
12423 Do not use multiply and add/subtract instructions.
12428 Select the FDPIC ABI, which uses function descriptors to represent
12429 pointers to functions. Without any PIC/PIE-related options, it
12430 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
12431 assumes GOT entries and small data are within a 12-bit range from the
12432 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
12433 are computed with 32 bits.
12434 With a @samp{bfin-elf} target, this option implies @option{-msim}.
12437 @opindex minline-plt
12439 Enable inlining of PLT entries in function calls to functions that are
12440 not known to bind locally. It has no effect without @option{-mfdpic}.
12441 It's enabled by default if optimizing for speed and compiling for
12442 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
12443 optimization option such as @option{-O3} or above is present in the
12449 Assume a large TLS segment when generating thread-local code.
12454 Do not assume a large TLS segment when generating thread-local code.
12459 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
12460 that is known to be in read-only sections. It's enabled by default,
12461 except for @option{-fpic} or @option{-fpie}: even though it may help
12462 make the global offset table smaller, it trades 1 instruction for 4.
12463 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
12464 one of which may be shared by multiple symbols, and it avoids the need
12465 for a GOT entry for the referenced symbol, so it's more likely to be a
12466 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
12468 @item -multilib-library-pic
12469 @opindex multilib-library-pic
12471 Link with the (library, not FD) pic libraries. It's implied by
12472 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
12473 @option{-fpic} without @option{-mfdpic}. You should never have to use
12477 @opindex mlinked-fp
12479 Follow the EABI requirement of always creating a frame pointer whenever
12480 a stack frame is allocated. This option is enabled by default and can
12481 be disabled with @option{-mno-linked-fp}.
12484 @opindex mlong-calls
12486 Use indirect addressing to call functions outside the current
12487 compilation unit. This allows the functions to be placed anywhere
12488 within the 32-bit address space.
12490 @item -malign-labels
12491 @opindex malign-labels
12493 Try to align labels to an 8-byte boundary by inserting nops into the
12494 previous packet. This option only has an effect when VLIW packing
12495 is enabled. It doesn't create new packets; it merely adds nops to
12498 @item -mlibrary-pic
12499 @opindex mlibrary-pic
12501 Generate position-independent EABI code.
12506 Use only the first four media accumulator registers.
12511 Use all eight media accumulator registers.
12516 Pack VLIW instructions.
12521 Do not pack VLIW instructions.
12524 @opindex mno-eflags
12526 Do not mark ABI switches in e_flags.
12529 @opindex mcond-move
12531 Enable the use of conditional-move instructions (default).
12533 This switch is mainly for debugging the compiler and will likely be removed
12534 in a future version.
12536 @item -mno-cond-move
12537 @opindex mno-cond-move
12539 Disable the use of conditional-move instructions.
12541 This switch is mainly for debugging the compiler and will likely be removed
12542 in a future version.
12547 Enable the use of conditional set instructions (default).
12549 This switch is mainly for debugging the compiler and will likely be removed
12550 in a future version.
12555 Disable the use of conditional set instructions.
12557 This switch is mainly for debugging the compiler and will likely be removed
12558 in a future version.
12561 @opindex mcond-exec
12563 Enable the use of conditional execution (default).
12565 This switch is mainly for debugging the compiler and will likely be removed
12566 in a future version.
12568 @item -mno-cond-exec
12569 @opindex mno-cond-exec
12571 Disable the use of conditional execution.
12573 This switch is mainly for debugging the compiler and will likely be removed
12574 in a future version.
12576 @item -mvliw-branch
12577 @opindex mvliw-branch
12579 Run a pass to pack branches into VLIW instructions (default).
12581 This switch is mainly for debugging the compiler and will likely be removed
12582 in a future version.
12584 @item -mno-vliw-branch
12585 @opindex mno-vliw-branch
12587 Do not run a pass to pack branches into VLIW instructions.
12589 This switch is mainly for debugging the compiler and will likely be removed
12590 in a future version.
12592 @item -mmulti-cond-exec
12593 @opindex mmulti-cond-exec
12595 Enable optimization of @code{&&} and @code{||} in conditional execution
12598 This switch is mainly for debugging the compiler and will likely be removed
12599 in a future version.
12601 @item -mno-multi-cond-exec
12602 @opindex mno-multi-cond-exec
12604 Disable optimization of @code{&&} and @code{||} in conditional execution.
12606 This switch is mainly for debugging the compiler and will likely be removed
12607 in a future version.
12609 @item -mnested-cond-exec
12610 @opindex mnested-cond-exec
12612 Enable nested conditional execution optimizations (default).
12614 This switch is mainly for debugging the compiler and will likely be removed
12615 in a future version.
12617 @item -mno-nested-cond-exec
12618 @opindex mno-nested-cond-exec
12620 Disable nested conditional execution optimizations.
12622 This switch is mainly for debugging the compiler and will likely be removed
12623 in a future version.
12625 @item -moptimize-membar
12626 @opindex moptimize-membar
12628 This switch removes redundant @code{membar} instructions from the
12629 compiler generated code. It is enabled by default.
12631 @item -mno-optimize-membar
12632 @opindex mno-optimize-membar
12634 This switch disables the automatic removal of redundant @code{membar}
12635 instructions from the generated code.
12637 @item -mtomcat-stats
12638 @opindex mtomcat-stats
12640 Cause gas to print out tomcat statistics.
12642 @item -mcpu=@var{cpu}
12645 Select the processor type for which to generate code. Possible values are
12646 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12647 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12651 @node GNU/Linux Options
12652 @subsection GNU/Linux Options
12654 These @samp{-m} options are defined for GNU/Linux targets:
12659 Use the GNU C library. This is the default except
12660 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12664 Use uClibc C library. This is the default on
12665 @samp{*-*-linux-*uclibc*} targets.
12669 Use Bionic C library. This is the default on
12670 @samp{*-*-linux-*android*} targets.
12674 Compile code compatible with Android platform. This is the default on
12675 @samp{*-*-linux-*android*} targets.
12677 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12678 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12679 this option makes the GCC driver pass Android-specific options to the linker.
12680 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12683 @item -tno-android-cc
12684 @opindex tno-android-cc
12685 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12686 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12687 @option{-fno-rtti} by default.
12689 @item -tno-android-ld
12690 @opindex tno-android-ld
12691 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12692 linking options to the linker.
12696 @node H8/300 Options
12697 @subsection H8/300 Options
12699 These @samp{-m} options are defined for the H8/300 implementations:
12704 Shorten some address references at link time, when possible; uses the
12705 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12706 ld, Using ld}, for a fuller description.
12710 Generate code for the H8/300H@.
12714 Generate code for the H8S@.
12718 Generate code for the H8S and H8/300H in the normal mode. This switch
12719 must be used either with @option{-mh} or @option{-ms}.
12723 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12727 Make @code{int} data 32 bits by default.
12730 @opindex malign-300
12731 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12732 The default for the H8/300H and H8S is to align longs and floats on
12734 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
12735 This option has no effect on the H8/300.
12739 @subsection HPPA Options
12740 @cindex HPPA Options
12742 These @samp{-m} options are defined for the HPPA family of computers:
12745 @item -march=@var{architecture-type}
12747 Generate code for the specified architecture. The choices for
12748 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12749 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12750 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12751 architecture option for your machine. Code compiled for lower numbered
12752 architectures will run on higher numbered architectures, but not the
12755 @item -mpa-risc-1-0
12756 @itemx -mpa-risc-1-1
12757 @itemx -mpa-risc-2-0
12758 @opindex mpa-risc-1-0
12759 @opindex mpa-risc-1-1
12760 @opindex mpa-risc-2-0
12761 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12764 @opindex mbig-switch
12765 Generate code suitable for big switch tables. Use this option only if
12766 the assembler/linker complain about out of range branches within a switch
12769 @item -mjump-in-delay
12770 @opindex mjump-in-delay
12771 Fill delay slots of function calls with unconditional jump instructions
12772 by modifying the return pointer for the function call to be the target
12773 of the conditional jump.
12775 @item -mdisable-fpregs
12776 @opindex mdisable-fpregs
12777 Prevent floating-point registers from being used in any manner. This is
12778 necessary for compiling kernels that perform lazy context switching of
12779 floating-point registers. If you use this option and attempt to perform
12780 floating-point operations, the compiler aborts.
12782 @item -mdisable-indexing
12783 @opindex mdisable-indexing
12784 Prevent the compiler from using indexing address modes. This avoids some
12785 rather obscure problems when compiling MIG generated code under MACH@.
12787 @item -mno-space-regs
12788 @opindex mno-space-regs
12789 Generate code that assumes the target has no space registers. This allows
12790 GCC to generate faster indirect calls and use unscaled index address modes.
12792 Such code is suitable for level 0 PA systems and kernels.
12794 @item -mfast-indirect-calls
12795 @opindex mfast-indirect-calls
12796 Generate code that assumes calls never cross space boundaries. This
12797 allows GCC to emit code that performs faster indirect calls.
12799 This option will not work in the presence of shared libraries or nested
12802 @item -mfixed-range=@var{register-range}
12803 @opindex mfixed-range
12804 Generate code treating the given register range as fixed registers.
12805 A fixed register is one that the register allocator can not use. This is
12806 useful when compiling kernel code. A register range is specified as
12807 two registers separated by a dash. Multiple register ranges can be
12808 specified separated by a comma.
12810 @item -mlong-load-store
12811 @opindex mlong-load-store
12812 Generate 3-instruction load and store sequences as sometimes required by
12813 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12816 @item -mportable-runtime
12817 @opindex mportable-runtime
12818 Use the portable calling conventions proposed by HP for ELF systems.
12822 Enable the use of assembler directives only GAS understands.
12824 @item -mschedule=@var{cpu-type}
12826 Schedule code according to the constraints for the machine type
12827 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12828 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12829 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12830 proper scheduling option for your machine. The default scheduling is
12834 @opindex mlinker-opt
12835 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12836 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12837 linkers in which they give bogus error messages when linking some programs.
12840 @opindex msoft-float
12841 Generate output containing library calls for floating point.
12842 @strong{Warning:} the requisite libraries are not available for all HPPA
12843 targets. Normally the facilities of the machine's usual C compiler are
12844 used, but this cannot be done directly in cross-compilation. You must make
12845 your own arrangements to provide suitable library functions for
12848 @option{-msoft-float} changes the calling convention in the output file;
12849 therefore, it is only useful if you compile @emph{all} of a program with
12850 this option. In particular, you need to compile @file{libgcc.a}, the
12851 library that comes with GCC, with @option{-msoft-float} in order for
12856 Generate the predefine, @code{_SIO}, for server IO@. The default is
12857 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12858 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12859 options are available under HP-UX and HI-UX@.
12863 Use GNU ld specific options. This passes @option{-shared} to ld when
12864 building a shared library. It is the default when GCC is configured,
12865 explicitly or implicitly, with the GNU linker. This option does not
12866 have any affect on which ld is called, it only changes what parameters
12867 are passed to that ld. The ld that is called is determined by the
12868 @option{--with-ld} configure option, GCC's program search path, and
12869 finally by the user's @env{PATH}. The linker used by GCC can be printed
12870 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12871 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12875 Use HP ld specific options. This passes @option{-b} to ld when building
12876 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12877 links. It is the default when GCC is configured, explicitly or
12878 implicitly, with the HP linker. This option does not have any affect on
12879 which ld is called, it only changes what parameters are passed to that
12880 ld. The ld that is called is determined by the @option{--with-ld}
12881 configure option, GCC's program search path, and finally by the user's
12882 @env{PATH}. The linker used by GCC can be printed using @samp{which
12883 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
12884 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12887 @opindex mno-long-calls
12888 Generate code that uses long call sequences. This ensures that a call
12889 is always able to reach linker generated stubs. The default is to generate
12890 long calls only when the distance from the call site to the beginning
12891 of the function or translation unit, as the case may be, exceeds a
12892 predefined limit set by the branch type being used. The limits for
12893 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12894 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12897 Distances are measured from the beginning of functions when using the
12898 @option{-ffunction-sections} option, or when using the @option{-mgas}
12899 and @option{-mno-portable-runtime} options together under HP-UX with
12902 It is normally not desirable to use this option as it will degrade
12903 performance. However, it may be useful in large applications,
12904 particularly when partial linking is used to build the application.
12906 The types of long calls used depends on the capabilities of the
12907 assembler and linker, and the type of code being generated. The
12908 impact on systems that support long absolute calls, and long pic
12909 symbol-difference or pc-relative calls should be relatively small.
12910 However, an indirect call is used on 32-bit ELF systems in pic code
12911 and it is quite long.
12913 @item -munix=@var{unix-std}
12915 Generate compiler predefines and select a startfile for the specified
12916 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12917 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12918 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12919 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12920 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12923 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12924 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12925 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12926 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12927 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12928 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12930 It is @emph{important} to note that this option changes the interfaces
12931 for various library routines. It also affects the operational behavior
12932 of the C library. Thus, @emph{extreme} care is needed in using this
12935 Library code that is intended to operate with more than one UNIX
12936 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12937 as appropriate. Most GNU software doesn't provide this capability.
12941 Suppress the generation of link options to search libdld.sl when the
12942 @option{-static} option is specified on HP-UX 10 and later.
12946 The HP-UX implementation of setlocale in libc has a dependency on
12947 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12948 when the @option{-static} option is specified, special link options
12949 are needed to resolve this dependency.
12951 On HP-UX 10 and later, the GCC driver adds the necessary options to
12952 link with libdld.sl when the @option{-static} option is specified.
12953 This causes the resulting binary to be dynamic. On the 64-bit port,
12954 the linkers generate dynamic binaries by default in any case. The
12955 @option{-nolibdld} option can be used to prevent the GCC driver from
12956 adding these link options.
12960 Add support for multithreading with the @dfn{dce thread} library
12961 under HP-UX@. This option sets flags for both the preprocessor and
12965 @node i386 and x86-64 Options
12966 @subsection Intel 386 and AMD x86-64 Options
12967 @cindex i386 Options
12968 @cindex x86-64 Options
12969 @cindex Intel 386 Options
12970 @cindex AMD x86-64 Options
12972 These @samp{-m} options are defined for the i386 and x86-64 family of
12976 @item -mtune=@var{cpu-type}
12978 Tune to @var{cpu-type} everything applicable about the generated code, except
12979 for the ABI and the set of available instructions. The choices for
12980 @var{cpu-type} are:
12983 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12984 If you know the CPU on which your code will run, then you should use
12985 the corresponding @option{-mtune} option instead of
12986 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12987 of your application will have, then you should use this option.
12989 As new processors are deployed in the marketplace, the behavior of this
12990 option will change. Therefore, if you upgrade to a newer version of
12991 GCC, the code generated option will change to reflect the processors
12992 that were most common when that version of GCC was released.
12994 There is no @option{-march=generic} option because @option{-march}
12995 indicates the instruction set the compiler can use, and there is no
12996 generic instruction set applicable to all processors. In contrast,
12997 @option{-mtune} indicates the processor (or, in this case, collection of
12998 processors) for which the code is optimized.
13000 This selects the CPU to tune for at compilation time by determining
13001 the processor type of the compiling machine. Using @option{-mtune=native}
13002 will produce code optimized for the local machine under the constraints
13003 of the selected instruction set. Using @option{-march=native} will
13004 enable all instruction subsets supported by the local machine (hence
13005 the result might not run on different machines).
13007 Original Intel's i386 CPU@.
13009 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
13010 @item i586, pentium
13011 Intel Pentium CPU with no MMX support.
13013 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
13015 Intel PentiumPro CPU@.
13017 Same as @code{generic}, but when used as @code{march} option, PentiumPro
13018 instruction set will be used, so the code will run on all i686 family chips.
13020 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
13021 @item pentium3, pentium3m
13022 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
13025 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
13026 support. Used by Centrino notebooks.
13027 @item pentium4, pentium4m
13028 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
13030 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
13033 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
13034 SSE2 and SSE3 instruction set support.
13036 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13037 instruction set support.
13039 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
13040 and SSE4.2 instruction set support.
13042 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13043 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
13045 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13046 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
13049 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13050 instruction set support.
13052 AMD K6 CPU with MMX instruction set support.
13054 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
13055 @item athlon, athlon-tbird
13056 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
13058 @item athlon-4, athlon-xp, athlon-mp
13059 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
13060 instruction set support.
13061 @item k8, opteron, athlon64, athlon-fx
13062 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
13063 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
13064 @item k8-sse3, opteron-sse3, athlon64-sse3
13065 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
13066 @item amdfam10, barcelona
13067 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
13068 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
13069 instruction set extensions.)
13071 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
13072 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
13073 SSSE3, SSE4.1, SSE4.2, 3DNow!, enhanced 3DNow!, ABM and 64-bit
13074 instruction set extensions.)
13076 AMD Family 14h core based CPUs with x86-64 instruction set support. (This
13077 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
13078 instruction set extensions.)
13080 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
13083 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
13084 instruction set support.
13086 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
13087 implemented for this chip.)
13089 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
13090 implemented for this chip.)
13092 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
13095 While picking a specific @var{cpu-type} will schedule things appropriately
13096 for that particular chip, the compiler will not generate any code that
13097 does not run on the default machine type without the @option{-march=@var{cpu-type}}
13098 option being used. For example, if GCC is configured for i686-pc-linux-gnu
13099 then @option{-mtune=pentium4} will generate code that is tuned for Pentium4
13100 but will still run on i686 machines.
13102 @item -march=@var{cpu-type}
13104 Generate instructions for the machine type @var{cpu-type}. The choices
13105 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
13106 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
13108 @item -mcpu=@var{cpu-type}
13110 A deprecated synonym for @option{-mtune}.
13112 @item -mfpmath=@var{unit}
13114 Generate floating-point arithmetic for selected unit @var{unit}. The choices
13115 for @var{unit} are:
13119 Use the standard 387 floating-point coprocessor present on the majority of chips and
13120 emulated otherwise. Code compiled with this option runs almost everywhere.
13121 The temporary results are computed in 80-bit precision instead of the precision
13122 specified by the type, resulting in slightly different results compared to most
13123 of other chips. See @option{-ffloat-store} for more detailed description.
13125 This is the default choice for i386 compiler.
13128 Use scalar floating-point instructions present in the SSE instruction set.
13129 This instruction set is supported by Pentium3 and newer chips, in the AMD line
13130 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
13131 instruction set supports only single-precision arithmetic, thus the double and
13132 extended-precision arithmetic are still done using 387. A later version, present
13133 only in Pentium4 and the future AMD x86-64 chips, supports double-precision
13136 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
13137 or @option{-msse2} switches to enable SSE extensions and make this option
13138 effective. For the x86-64 compiler, these extensions are enabled by default.
13140 The resulting code should be considerably faster in the majority of cases and avoid
13141 the numerical instability problems of 387 code, but may break some existing
13142 code that expects temporaries to be 80 bits.
13144 This is the default choice for the x86-64 compiler.
13149 Attempt to utilize both instruction sets at once. This effectively double the
13150 amount of available registers and on chips with separate execution units for
13151 387 and SSE the execution resources too. Use this option with care, as it is
13152 still experimental, because the GCC register allocator does not model separate
13153 functional units well resulting in instable performance.
13156 @item -masm=@var{dialect}
13157 @opindex masm=@var{dialect}
13158 Output asm instructions using selected @var{dialect}. Supported
13159 choices are @samp{intel} or @samp{att} (the default one). Darwin does
13160 not support @samp{intel}.
13163 @itemx -mno-ieee-fp
13165 @opindex mno-ieee-fp
13166 Control whether or not the compiler uses IEEE floating-point
13167 comparisons. These handle correctly the case where the result of a
13168 comparison is unordered.
13171 @opindex msoft-float
13172 Generate output containing library calls for floating point.
13173 @strong{Warning:} the requisite libraries are not part of GCC@.
13174 Normally the facilities of the machine's usual C compiler are used, but
13175 this can't be done directly in cross-compilation. You must make your
13176 own arrangements to provide suitable library functions for
13179 On machines where a function returns floating-point results in the 80387
13180 register stack, some floating-point opcodes may be emitted even if
13181 @option{-msoft-float} is used.
13183 @item -mno-fp-ret-in-387
13184 @opindex mno-fp-ret-in-387
13185 Do not use the FPU registers for return values of functions.
13187 The usual calling convention has functions return values of types
13188 @code{float} and @code{double} in an FPU register, even if there
13189 is no FPU@. The idea is that the operating system should emulate
13192 The option @option{-mno-fp-ret-in-387} causes such values to be returned
13193 in ordinary CPU registers instead.
13195 @item -mno-fancy-math-387
13196 @opindex mno-fancy-math-387
13197 Some 387 emulators do not support the @code{sin}, @code{cos} and
13198 @code{sqrt} instructions for the 387. Specify this option to avoid
13199 generating those instructions. This option is the default on FreeBSD,
13200 OpenBSD and NetBSD@. This option is overridden when @option{-march}
13201 indicates that the target CPU will always have an FPU and so the
13202 instruction will not need emulation. As of revision 2.6.1, these
13203 instructions are not generated unless you also use the
13204 @option{-funsafe-math-optimizations} switch.
13206 @item -malign-double
13207 @itemx -mno-align-double
13208 @opindex malign-double
13209 @opindex mno-align-double
13210 Control whether GCC aligns @code{double}, @code{long double}, and
13211 @code{long long} variables on a two-word boundary or a one-word
13212 boundary. Aligning @code{double} variables on a two-word boundary
13213 produces code that runs somewhat faster on a @samp{Pentium} at the
13214 expense of more memory.
13216 On x86-64, @option{-malign-double} is enabled by default.
13218 @strong{Warning:} if you use the @option{-malign-double} switch,
13219 structures containing the above types will be aligned differently than
13220 the published application binary interface specifications for the 386
13221 and will not be binary compatible with structures in code compiled
13222 without that switch.
13224 @item -m96bit-long-double
13225 @itemx -m128bit-long-double
13226 @opindex m96bit-long-double
13227 @opindex m128bit-long-double
13228 These switches control the size of @code{long double} type. The i386
13229 application binary interface specifies the size to be 96 bits,
13230 so @option{-m96bit-long-double} is the default in 32-bit mode.
13232 Modern architectures (Pentium and newer) prefer @code{long double}
13233 to be aligned to an 8- or 16-byte boundary. In arrays or structures
13234 conforming to the ABI, this is not possible. So specifying
13235 @option{-m128bit-long-double} aligns @code{long double}
13236 to a 16-byte boundary by padding the @code{long double} with an additional
13239 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
13240 its ABI specifies that @code{long double} is to be aligned on 16-byte boundary.
13242 Notice that neither of these options enable any extra precision over the x87
13243 standard of 80 bits for a @code{long double}.
13245 @strong{Warning:} if you override the default value for your target ABI, the
13246 structures and arrays containing @code{long double} variables will change
13247 their size as well as function calling convention for function taking
13248 @code{long double} will be modified. Hence they will not be binary
13249 compatible with arrays or structures in code compiled without that switch.
13251 @item -mlarge-data-threshold=@var{number}
13252 @opindex mlarge-data-threshold=@var{number}
13253 When @option{-mcmodel=medium} is specified, the data greater than
13254 @var{threshold} are placed in large data section. This value must be the
13255 same across all object linked into the binary and defaults to 65535.
13259 Use a different function-calling convention, in which functions that
13260 take a fixed number of arguments return with the @code{ret} @var{num}
13261 instruction, which pops their arguments while returning. This saves one
13262 instruction in the caller since there is no need to pop the arguments
13265 You can specify that an individual function is called with this calling
13266 sequence with the function attribute @samp{stdcall}. You can also
13267 override the @option{-mrtd} option by using the function attribute
13268 @samp{cdecl}. @xref{Function Attributes}.
13270 @strong{Warning:} this calling convention is incompatible with the one
13271 normally used on Unix, so you cannot use it if you need to call
13272 libraries compiled with the Unix compiler.
13274 Also, you must provide function prototypes for all functions that
13275 take variable numbers of arguments (including @code{printf});
13276 otherwise incorrect code will be generated for calls to those
13279 In addition, seriously incorrect code will result if you call a
13280 function with too many arguments. (Normally, extra arguments are
13281 harmlessly ignored.)
13283 @item -mregparm=@var{num}
13285 Control how many registers are used to pass integer arguments. By
13286 default, no registers are used to pass arguments, and at most 3
13287 registers can be used. You can control this behavior for a specific
13288 function by using the function attribute @samp{regparm}.
13289 @xref{Function Attributes}.
13291 @strong{Warning:} if you use this switch, and
13292 @var{num} is nonzero, then you must build all modules with the same
13293 value, including any libraries. This includes the system libraries and
13297 @opindex msseregparm
13298 Use SSE register passing conventions for float and double arguments
13299 and return values. You can control this behavior for a specific
13300 function by using the function attribute @samp{sseregparm}.
13301 @xref{Function Attributes}.
13303 @strong{Warning:} if you use this switch then you must build all
13304 modules with the same value, including any libraries. This includes
13305 the system libraries and startup modules.
13307 @item -mvect8-ret-in-mem
13308 @opindex mvect8-ret-in-mem
13309 Return 8-byte vectors in memory instead of MMX registers. This is the
13310 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
13311 Studio compilers until version 12. Later compiler versions (starting
13312 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
13313 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
13314 you need to remain compatible with existing code produced by those
13315 previous compiler versions or older versions of GCC.
13324 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
13325 is specified, the significands of results of floating-point operations are
13326 rounded to 24 bits (single precision); @option{-mpc64} rounds the
13327 significands of results of floating-point operations to 53 bits (double
13328 precision) and @option{-mpc80} rounds the significands of results of
13329 floating-point operations to 64 bits (extended double precision), which is
13330 the default. When this option is used, floating-point operations in higher
13331 precisions are not available to the programmer without setting the FPU
13332 control word explicitly.
13334 Setting the rounding of floating-point operations to less than the default
13335 80 bits can speed some programs by 2% or more. Note that some mathematical
13336 libraries assume that extended-precision (80-bit) floating-point operations
13337 are enabled by default; routines in such libraries could suffer significant
13338 loss of accuracy, typically through so-called "catastrophic cancellation",
13339 when this option is used to set the precision to less than extended precision.
13341 @item -mstackrealign
13342 @opindex mstackrealign
13343 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
13344 option will generate an alternate prologue and epilogue that realigns the
13345 run-time stack if necessary. This supports mixing legacy codes that keep
13346 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
13347 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
13348 applicable to individual functions.
13350 @item -mpreferred-stack-boundary=@var{num}
13351 @opindex mpreferred-stack-boundary
13352 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
13353 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
13354 the default is 4 (16 bytes or 128 bits).
13356 @item -mincoming-stack-boundary=@var{num}
13357 @opindex mincoming-stack-boundary
13358 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
13359 boundary. If @option{-mincoming-stack-boundary} is not specified,
13360 the one specified by @option{-mpreferred-stack-boundary} will be used.
13362 On Pentium and PentiumPro, @code{double} and @code{long double} values
13363 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
13364 suffer significant run time performance penalties. On Pentium III, the
13365 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
13366 properly if it is not 16-byte aligned.
13368 To ensure proper alignment of this values on the stack, the stack boundary
13369 must be as aligned as that required by any value stored on the stack.
13370 Further, every function must be generated such that it keeps the stack
13371 aligned. Thus calling a function compiled with a higher preferred
13372 stack boundary from a function compiled with a lower preferred stack
13373 boundary will most likely misalign the stack. It is recommended that
13374 libraries that use callbacks always use the default setting.
13376 This extra alignment does consume extra stack space, and generally
13377 increases code size. Code that is sensitive to stack space usage, such
13378 as embedded systems and operating system kernels, may want to reduce the
13379 preferred alignment to @option{-mpreferred-stack-boundary=2}.
13408 @itemx -mno-fsgsbase
13444 These switches enable or disable the use of instructions in the MMX, SSE,
13445 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
13446 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
13447 @: extended instruction sets.
13448 These extensions are also available as built-in functions: see
13449 @ref{X86 Built-in Functions}, for details of the functions enabled and
13450 disabled by these switches.
13452 To have SSE/SSE2 instructions generated automatically from floating-point
13453 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
13455 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
13456 generates new AVX instructions or AVX equivalence for all SSEx instructions
13459 These options will enable GCC to use these extended instructions in
13460 generated code, even without @option{-mfpmath=sse}. Applications that
13461 perform run-time CPU detection must compile separate files for each
13462 supported architecture, using the appropriate flags. In particular,
13463 the file containing the CPU detection code should be compiled without
13468 This option instructs GCC to emit a @code{cld} instruction in the prologue
13469 of functions that use string instructions. String instructions depend on
13470 the DF flag to select between autoincrement or autodecrement mode. While the
13471 ABI specifies the DF flag to be cleared on function entry, some operating
13472 systems violate this specification by not clearing the DF flag in their
13473 exception dispatchers. The exception handler can be invoked with the DF flag
13474 set, which leads to wrong direction mode when string instructions are used.
13475 This option can be enabled by default on 32-bit x86 targets by configuring
13476 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
13477 instructions can be suppressed with the @option{-mno-cld} compiler option
13481 @opindex mvzeroupper
13482 This option instructs GCC to emit a @code{vzeroupper} instruction
13483 before a transfer of control flow out of the function to minimize
13484 AVX to SSE transition penalty as well as remove unnecessary zeroupper
13489 This option will enable GCC to use CMPXCHG16B instruction in generated code.
13490 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
13491 data types. This is useful for high resolution counters that could be updated
13492 by multiple processors (or cores). This instruction is generated as part of
13493 atomic built-in functions: see @ref{__sync Builtins} or
13494 @ref{__atomic Builtins} for details.
13498 This option will enable GCC to use SAHF instruction in generated 64-bit code.
13499 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
13500 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
13501 SAHF are load and store instructions, respectively, for certain status flags.
13502 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
13503 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
13507 This option will enable GCC to use movbe instruction to implement
13508 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
13512 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
13513 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
13514 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
13518 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
13519 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
13520 to increase precision instead of DIVSS and SQRTSS (and their vectorized
13521 variants) for single-precision floating-point arguments. These instructions
13522 are generated only when @option{-funsafe-math-optimizations} is enabled
13523 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
13524 Note that while the throughput of the sequence is higher than the throughput
13525 of the non-reciprocal instruction, the precision of the sequence can be
13526 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
13528 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS
13529 (or RSQRTPS) already with @option{-ffast-math} (or the above option
13530 combination), and doesn't need @option{-mrecip}.
13532 Also note that GCC emits the above sequence with additional Newton-Raphson step
13533 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
13534 already with @option{-ffast-math} (or the above option combination), and
13535 doesn't need @option{-mrecip}.
13537 @item -mrecip=@var{opt}
13538 @opindex mrecip=opt
13539 This option allows to control which reciprocal estimate instructions
13540 may be used. @var{opt} is a comma separated list of options, which may
13541 be preceded by a @code{!} to invert the option:
13542 @code{all}: enable all estimate instructions,
13543 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
13544 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip},
13545 @code{div}: enable the approximation for scalar division,
13546 @code{vec-div}: enable the approximation for vectorized division,
13547 @code{sqrt}: enable the approximation for scalar square root,
13548 @code{vec-sqrt}: enable the approximation for vectorized square root.
13550 So for example, @option{-mrecip=all,!sqrt} would enable
13551 all of the reciprocal approximations, except for square root.
13553 @item -mveclibabi=@var{type}
13554 @opindex mveclibabi
13555 Specifies the ABI type to use for vectorizing intrinsics using an
13556 external library. Supported types are @code{svml} for the Intel short
13557 vector math library and @code{acml} for the AMD math core library style
13558 of interfacing. GCC will currently emit calls to @code{vmldExp2},
13559 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
13560 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
13561 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
13562 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
13563 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
13564 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
13565 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
13566 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
13567 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
13568 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
13569 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
13570 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
13571 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
13572 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
13573 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
13574 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
13575 compatible library will have to be specified at link time.
13577 @item -mabi=@var{name}
13579 Generate code for the specified calling convention. Permissible values
13580 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
13581 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
13582 ABI when targeting Windows. On all other systems, the default is the
13583 SYSV ABI. You can control this behavior for a specific function by
13584 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
13585 @xref{Function Attributes}.
13587 @item -mtls-dialect=@var{type}
13588 @opindex mtls-dialect
13589 Generate code to access thread-local storage using the @samp{gnu} or
13590 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
13591 @samp{gnu2} is more efficient, but it may add compile- and run-time
13592 requirements that cannot be satisfied on all systems.
13595 @itemx -mno-push-args
13596 @opindex mpush-args
13597 @opindex mno-push-args
13598 Use PUSH operations to store outgoing parameters. This method is shorter
13599 and usually equally fast as method using SUB/MOV operations and is enabled
13600 by default. In some cases disabling it may improve performance because of
13601 improved scheduling and reduced dependencies.
13603 @item -maccumulate-outgoing-args
13604 @opindex maccumulate-outgoing-args
13605 If enabled, the maximum amount of space required for outgoing arguments will be
13606 computed in the function prologue. This is faster on most modern CPUs
13607 because of reduced dependencies, improved scheduling and reduced stack usage
13608 when preferred stack boundary is not equal to 2. The drawback is a notable
13609 increase in code size. This switch implies @option{-mno-push-args}.
13613 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
13614 on thread-safe exception handling must compile and link all code with the
13615 @option{-mthreads} option. When compiling, @option{-mthreads} defines
13616 @option{-D_MT}; when linking, it links in a special thread helper library
13617 @option{-lmingwthrd} which cleans up per thread exception handling data.
13619 @item -mno-align-stringops
13620 @opindex mno-align-stringops
13621 Do not align destination of inlined string operations. This switch reduces
13622 code size and improves performance in case the destination is already aligned,
13623 but GCC doesn't know about it.
13625 @item -minline-all-stringops
13626 @opindex minline-all-stringops
13627 By default GCC inlines string operations only when the destination is
13628 known to be aligned to least a 4-byte boundary.
13629 This enables more inlining, increase code
13630 size, but may improve performance of code that depends on fast memcpy, strlen
13631 and memset for short lengths.
13633 @item -minline-stringops-dynamically
13634 @opindex minline-stringops-dynamically
13635 For string operations of unknown size, use run-time checks with
13636 inline code for small blocks and a library call for large blocks.
13638 @item -mstringop-strategy=@var{alg}
13639 @opindex mstringop-strategy=@var{alg}
13640 Overwrite internal decision heuristic about particular algorithm to inline
13641 string operation with. The allowed values are @code{rep_byte},
13642 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
13643 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
13644 expanding inline loop, @code{libcall} for always expanding library call.
13646 @item -momit-leaf-frame-pointer
13647 @opindex momit-leaf-frame-pointer
13648 Don't keep the frame pointer in a register for leaf functions. This
13649 avoids the instructions to save, set up and restore frame pointers and
13650 makes an extra register available in leaf functions. The option
13651 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13652 which might make debugging harder.
13654 @item -mtls-direct-seg-refs
13655 @itemx -mno-tls-direct-seg-refs
13656 @opindex mtls-direct-seg-refs
13657 Controls whether TLS variables may be accessed with offsets from the
13658 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
13659 or whether the thread base pointer must be added. Whether or not this
13660 is legal depends on the operating system, and whether it maps the
13661 segment to cover the entire TLS area.
13663 For systems that use GNU libc, the default is on.
13666 @itemx -mno-sse2avx
13668 Specify that the assembler should encode SSE instructions with VEX
13669 prefix. The option @option{-mavx} turns this on by default.
13674 If profiling is active @option{-pg} put the profiling
13675 counter call before prologue.
13676 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13677 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13680 @itemx -mno-8bit-idiv
13682 On some processors, like Intel Atom, 8-bit unsigned integer divide is
13683 much faster than 32-bit/64-bit integer divide. This option generates a
13684 run-time check. If both dividend and divisor are within range of 0
13685 to 255, 8-bit unsigned integer divide is used instead of
13686 32-bit/64-bit integer divide.
13688 @item -mavx256-split-unaligned-load
13689 @item -mavx256-split-unaligned-store
13690 @opindex avx256-split-unaligned-load
13691 @opindex avx256-split-unaligned-store
13692 Split 32-byte AVX unaligned load and store.
13696 These @samp{-m} switches are supported in addition to the above
13697 on AMD x86-64 processors in 64-bit environments.
13706 Generate code for a 32-bit or 64-bit environment.
13707 The @option{-m32} option sets int, long and pointer to 32 bits and
13708 generates code that runs on any i386 system.
13709 The @option{-m64} option sets int to 32 bits and long and pointer
13710 to 64 bits and generates code for AMD's x86-64 architecture.
13711 The @option{-mx32} option sets int, long and pointer to 32 bits and
13712 generates code for AMD's x86-64 architecture.
13713 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
13714 and @option{-mdynamic-no-pic} options.
13716 @item -mno-red-zone
13717 @opindex mno-red-zone
13718 Do not use a so called red zone for x86-64 code. The red zone is mandated
13719 by the x86-64 ABI, it is a 128-byte area beyond the location of the
13720 stack pointer that will not be modified by signal or interrupt handlers
13721 and therefore can be used for temporary data without adjusting the stack
13722 pointer. The flag @option{-mno-red-zone} disables this red zone.
13724 @item -mcmodel=small
13725 @opindex mcmodel=small
13726 Generate code for the small code model: the program and its symbols must
13727 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13728 Programs can be statically or dynamically linked. This is the default
13731 @item -mcmodel=kernel
13732 @opindex mcmodel=kernel
13733 Generate code for the kernel code model. The kernel runs in the
13734 negative 2 GB of the address space.
13735 This model has to be used for Linux kernel code.
13737 @item -mcmodel=medium
13738 @opindex mcmodel=medium
13739 Generate code for the medium model: The program is linked in the lower 2
13740 GB of the address space. Small symbols are also placed there. Symbols
13741 with sizes larger than @option{-mlarge-data-threshold} are put into
13742 large data or bss sections and can be located above 2GB. Programs can
13743 be statically or dynamically linked.
13745 @item -mcmodel=large
13746 @opindex mcmodel=large
13747 Generate code for the large model: This model makes no assumptions
13748 about addresses and sizes of sections.
13751 @node i386 and x86-64 Windows Options
13752 @subsection i386 and x86-64 Windows Options
13753 @cindex i386 and x86-64 Windows Options
13755 These additional options are available for Windows targets:
13760 This option is available for Cygwin and MinGW targets. It
13761 specifies that a console application is to be generated, by
13762 instructing the linker to set the PE header subsystem type
13763 required for console applications.
13764 This is the default behavior for Cygwin and MinGW targets.
13768 This option is available for Cygwin and MinGW targets. It
13769 specifies that a DLL - a dynamic link library - is to be
13770 generated, enabling the selection of the required runtime
13771 startup object and entry point.
13773 @item -mnop-fun-dllimport
13774 @opindex mnop-fun-dllimport
13775 This option is available for Cygwin and MinGW targets. It
13776 specifies that the dllimport attribute should be ignored.
13780 This option is available for MinGW targets. It specifies
13781 that MinGW-specific thread support is to be used.
13785 This option is available for mingw-w64 targets. It specifies
13786 that the UNICODE macro is getting pre-defined and that the
13787 unicode capable runtime startup code is chosen.
13791 This option is available for Cygwin and MinGW targets. It
13792 specifies that the typical Windows pre-defined macros are to
13793 be set in the pre-processor, but does not influence the choice
13794 of runtime library/startup code.
13798 This option is available for Cygwin and MinGW targets. It
13799 specifies that a GUI application is to be generated by
13800 instructing the linker to set the PE header subsystem type
13803 @item -fno-set-stack-executable
13804 @opindex fno-set-stack-executable
13805 This option is available for MinGW targets. It specifies that
13806 the executable flag for stack used by nested functions isn't
13807 set. This is necessary for binaries running in kernel mode of
13808 Windows, as there the user32 API, which is used to set executable
13809 privileges, isn't available.
13811 @item -mpe-aligned-commons
13812 @opindex mpe-aligned-commons
13813 This option is available for Cygwin and MinGW targets. It
13814 specifies that the GNU extension to the PE file format that
13815 permits the correct alignment of COMMON variables should be
13816 used when generating code. It will be enabled by default if
13817 GCC detects that the target assembler found during configuration
13818 supports the feature.
13821 See also under @ref{i386 and x86-64 Options} for standard options.
13823 @node IA-64 Options
13824 @subsection IA-64 Options
13825 @cindex IA-64 Options
13827 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13831 @opindex mbig-endian
13832 Generate code for a big-endian target. This is the default for HP-UX@.
13834 @item -mlittle-endian
13835 @opindex mlittle-endian
13836 Generate code for a little-endian target. This is the default for AIX5
13842 @opindex mno-gnu-as
13843 Generate (or don't) code for the GNU assembler. This is the default.
13844 @c Also, this is the default if the configure option @option{--with-gnu-as}
13850 @opindex mno-gnu-ld
13851 Generate (or don't) code for the GNU linker. This is the default.
13852 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13857 Generate code that does not use a global pointer register. The result
13858 is not position independent code, and violates the IA-64 ABI@.
13860 @item -mvolatile-asm-stop
13861 @itemx -mno-volatile-asm-stop
13862 @opindex mvolatile-asm-stop
13863 @opindex mno-volatile-asm-stop
13864 Generate (or don't) a stop bit immediately before and after volatile asm
13867 @item -mregister-names
13868 @itemx -mno-register-names
13869 @opindex mregister-names
13870 @opindex mno-register-names
13871 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13872 the stacked registers. This may make assembler output more readable.
13878 Disable (or enable) optimizations that use the small data section. This may
13879 be useful for working around optimizer bugs.
13881 @item -mconstant-gp
13882 @opindex mconstant-gp
13883 Generate code that uses a single constant global pointer value. This is
13884 useful when compiling kernel code.
13888 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13889 This is useful when compiling firmware code.
13891 @item -minline-float-divide-min-latency
13892 @opindex minline-float-divide-min-latency
13893 Generate code for inline divides of floating-point values
13894 using the minimum latency algorithm.
13896 @item -minline-float-divide-max-throughput
13897 @opindex minline-float-divide-max-throughput
13898 Generate code for inline divides of floating-point values
13899 using the maximum throughput algorithm.
13901 @item -mno-inline-float-divide
13902 @opindex mno-inline-float-divide
13903 Do not generate inline code for divides of floating-point values.
13905 @item -minline-int-divide-min-latency
13906 @opindex minline-int-divide-min-latency
13907 Generate code for inline divides of integer values
13908 using the minimum latency algorithm.
13910 @item -minline-int-divide-max-throughput
13911 @opindex minline-int-divide-max-throughput
13912 Generate code for inline divides of integer values
13913 using the maximum throughput algorithm.
13915 @item -mno-inline-int-divide
13916 @opindex mno-inline-int-divide
13917 Do not generate inline code for divides of integer values.
13919 @item -minline-sqrt-min-latency
13920 @opindex minline-sqrt-min-latency
13921 Generate code for inline square roots
13922 using the minimum latency algorithm.
13924 @item -minline-sqrt-max-throughput
13925 @opindex minline-sqrt-max-throughput
13926 Generate code for inline square roots
13927 using the maximum throughput algorithm.
13929 @item -mno-inline-sqrt
13930 @opindex mno-inline-sqrt
13931 Do not generate inline code for sqrt.
13934 @itemx -mno-fused-madd
13935 @opindex mfused-madd
13936 @opindex mno-fused-madd
13937 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13938 instructions. The default is to use these instructions.
13940 @item -mno-dwarf2-asm
13941 @itemx -mdwarf2-asm
13942 @opindex mno-dwarf2-asm
13943 @opindex mdwarf2-asm
13944 Don't (or do) generate assembler code for the DWARF2 line number debugging
13945 info. This may be useful when not using the GNU assembler.
13947 @item -mearly-stop-bits
13948 @itemx -mno-early-stop-bits
13949 @opindex mearly-stop-bits
13950 @opindex mno-early-stop-bits
13951 Allow stop bits to be placed earlier than immediately preceding the
13952 instruction that triggered the stop bit. This can improve instruction
13953 scheduling, but does not always do so.
13955 @item -mfixed-range=@var{register-range}
13956 @opindex mfixed-range
13957 Generate code treating the given register range as fixed registers.
13958 A fixed register is one that the register allocator can not use. This is
13959 useful when compiling kernel code. A register range is specified as
13960 two registers separated by a dash. Multiple register ranges can be
13961 specified separated by a comma.
13963 @item -mtls-size=@var{tls-size}
13965 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13968 @item -mtune=@var{cpu-type}
13970 Tune the instruction scheduling for a particular CPU, Valid values are
13971 itanium, itanium1, merced, itanium2, and mckinley.
13977 Generate code for a 32-bit or 64-bit environment.
13978 The 32-bit environment sets int, long and pointer to 32 bits.
13979 The 64-bit environment sets int to 32 bits and long and pointer
13980 to 64 bits. These are HP-UX specific flags.
13982 @item -mno-sched-br-data-spec
13983 @itemx -msched-br-data-spec
13984 @opindex mno-sched-br-data-spec
13985 @opindex msched-br-data-spec
13986 (Dis/En)able data speculative scheduling before reload.
13987 This will result in generation of the ld.a instructions and
13988 the corresponding check instructions (ld.c / chk.a).
13989 The default is 'disable'.
13991 @item -msched-ar-data-spec
13992 @itemx -mno-sched-ar-data-spec
13993 @opindex msched-ar-data-spec
13994 @opindex mno-sched-ar-data-spec
13995 (En/Dis)able data speculative scheduling after reload.
13996 This will result in generation of the ld.a instructions and
13997 the corresponding check instructions (ld.c / chk.a).
13998 The default is 'enable'.
14000 @item -mno-sched-control-spec
14001 @itemx -msched-control-spec
14002 @opindex mno-sched-control-spec
14003 @opindex msched-control-spec
14004 (Dis/En)able control speculative scheduling. This feature is
14005 available only during region scheduling (i.e.@: before reload).
14006 This will result in generation of the ld.s instructions and
14007 the corresponding check instructions chk.s .
14008 The default is 'disable'.
14010 @item -msched-br-in-data-spec
14011 @itemx -mno-sched-br-in-data-spec
14012 @opindex msched-br-in-data-spec
14013 @opindex mno-sched-br-in-data-spec
14014 (En/Dis)able speculative scheduling of the instructions that
14015 are dependent on the data speculative loads before reload.
14016 This is effective only with @option{-msched-br-data-spec} enabled.
14017 The default is 'enable'.
14019 @item -msched-ar-in-data-spec
14020 @itemx -mno-sched-ar-in-data-spec
14021 @opindex msched-ar-in-data-spec
14022 @opindex mno-sched-ar-in-data-spec
14023 (En/Dis)able speculative scheduling of the instructions that
14024 are dependent on the data speculative loads after reload.
14025 This is effective only with @option{-msched-ar-data-spec} enabled.
14026 The default is 'enable'.
14028 @item -msched-in-control-spec
14029 @itemx -mno-sched-in-control-spec
14030 @opindex msched-in-control-spec
14031 @opindex mno-sched-in-control-spec
14032 (En/Dis)able speculative scheduling of the instructions that
14033 are dependent on the control speculative loads.
14034 This is effective only with @option{-msched-control-spec} enabled.
14035 The default is 'enable'.
14037 @item -mno-sched-prefer-non-data-spec-insns
14038 @itemx -msched-prefer-non-data-spec-insns
14039 @opindex mno-sched-prefer-non-data-spec-insns
14040 @opindex msched-prefer-non-data-spec-insns
14041 If enabled, data speculative instructions will be chosen for schedule
14042 only if there are no other choices at the moment. This will make
14043 the use of the data speculation much more conservative.
14044 The default is 'disable'.
14046 @item -mno-sched-prefer-non-control-spec-insns
14047 @itemx -msched-prefer-non-control-spec-insns
14048 @opindex mno-sched-prefer-non-control-spec-insns
14049 @opindex msched-prefer-non-control-spec-insns
14050 If enabled, control speculative instructions will be chosen for schedule
14051 only if there are no other choices at the moment. This will make
14052 the use of the control speculation much more conservative.
14053 The default is 'disable'.
14055 @item -mno-sched-count-spec-in-critical-path
14056 @itemx -msched-count-spec-in-critical-path
14057 @opindex mno-sched-count-spec-in-critical-path
14058 @opindex msched-count-spec-in-critical-path
14059 If enabled, speculative dependencies will be considered during
14060 computation of the instructions priorities. This will make the use of the
14061 speculation a bit more conservative.
14062 The default is 'disable'.
14064 @item -msched-spec-ldc
14065 @opindex msched-spec-ldc
14066 Use a simple data speculation check. This option is on by default.
14068 @item -msched-control-spec-ldc
14069 @opindex msched-spec-ldc
14070 Use a simple check for control speculation. This option is on by default.
14072 @item -msched-stop-bits-after-every-cycle
14073 @opindex msched-stop-bits-after-every-cycle
14074 Place a stop bit after every cycle when scheduling. This option is on
14077 @item -msched-fp-mem-deps-zero-cost
14078 @opindex msched-fp-mem-deps-zero-cost
14079 Assume that floating-point stores and loads are not likely to cause a conflict
14080 when placed into the same instruction group. This option is disabled by
14083 @item -msel-sched-dont-check-control-spec
14084 @opindex msel-sched-dont-check-control-spec
14085 Generate checks for control speculation in selective scheduling.
14086 This flag is disabled by default.
14088 @item -msched-max-memory-insns=@var{max-insns}
14089 @opindex msched-max-memory-insns
14090 Limit on the number of memory insns per instruction group, giving lower
14091 priority to subsequent memory insns attempting to schedule in the same
14092 instruction group. Frequently useful to prevent cache bank conflicts.
14093 The default value is 1.
14095 @item -msched-max-memory-insns-hard-limit
14096 @opindex msched-max-memory-insns-hard-limit
14097 Disallow more than `msched-max-memory-insns' in instruction group.
14098 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
14099 when limit is reached but may still schedule memory operations.
14103 @node IA-64/VMS Options
14104 @subsection IA-64/VMS Options
14106 These @samp{-m} options are defined for the IA-64/VMS implementations:
14109 @item -mvms-return-codes
14110 @opindex mvms-return-codes
14111 Return VMS condition codes from main. The default is to return POSIX
14112 style condition (e.g.@ error) codes.
14114 @item -mdebug-main=@var{prefix}
14115 @opindex mdebug-main=@var{prefix}
14116 Flag the first routine whose name starts with @var{prefix} as the main
14117 routine for the debugger.
14121 Default to 64-bit memory allocation routines.
14125 @subsection LM32 Options
14126 @cindex LM32 options
14128 These @option{-m} options are defined for the Lattice Mico32 architecture:
14131 @item -mbarrel-shift-enabled
14132 @opindex mbarrel-shift-enabled
14133 Enable barrel-shift instructions.
14135 @item -mdivide-enabled
14136 @opindex mdivide-enabled
14137 Enable divide and modulus instructions.
14139 @item -mmultiply-enabled
14140 @opindex multiply-enabled
14141 Enable multiply instructions.
14143 @item -msign-extend-enabled
14144 @opindex msign-extend-enabled
14145 Enable sign extend instructions.
14147 @item -muser-enabled
14148 @opindex muser-enabled
14149 Enable user-defined instructions.
14154 @subsection M32C Options
14155 @cindex M32C options
14158 @item -mcpu=@var{name}
14160 Select the CPU for which code is generated. @var{name} may be one of
14161 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
14162 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
14163 the M32C/80 series.
14167 Specifies that the program will be run on the simulator. This causes
14168 an alternate runtime library to be linked in which supports, for
14169 example, file I/O@. You must not use this option when generating
14170 programs that will run on real hardware; you must provide your own
14171 runtime library for whatever I/O functions are needed.
14173 @item -memregs=@var{number}
14175 Specifies the number of memory-based pseudo-registers GCC will use
14176 during code generation. These pseudo-registers will be used like real
14177 registers, so there is a tradeoff between GCC's ability to fit the
14178 code into available registers, and the performance penalty of using
14179 memory instead of registers. Note that all modules in a program must
14180 be compiled with the same value for this option. Because of that, you
14181 must not use this option with the default runtime libraries gcc
14186 @node M32R/D Options
14187 @subsection M32R/D Options
14188 @cindex M32R/D options
14190 These @option{-m} options are defined for Renesas M32R/D architectures:
14195 Generate code for the M32R/2@.
14199 Generate code for the M32R/X@.
14203 Generate code for the M32R@. This is the default.
14205 @item -mmodel=small
14206 @opindex mmodel=small
14207 Assume all objects live in the lower 16MB of memory (so that their addresses
14208 can be loaded with the @code{ld24} instruction), and assume all subroutines
14209 are reachable with the @code{bl} instruction.
14210 This is the default.
14212 The addressability of a particular object can be set with the
14213 @code{model} attribute.
14215 @item -mmodel=medium
14216 @opindex mmodel=medium
14217 Assume objects may be anywhere in the 32-bit address space (the compiler
14218 will generate @code{seth/add3} instructions to load their addresses), and
14219 assume all subroutines are reachable with the @code{bl} instruction.
14221 @item -mmodel=large
14222 @opindex mmodel=large
14223 Assume objects may be anywhere in the 32-bit address space (the compiler
14224 will generate @code{seth/add3} instructions to load their addresses), and
14225 assume subroutines may not be reachable with the @code{bl} instruction
14226 (the compiler will generate the much slower @code{seth/add3/jl}
14227 instruction sequence).
14230 @opindex msdata=none
14231 Disable use of the small data area. Variables will be put into
14232 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
14233 @code{section} attribute has been specified).
14234 This is the default.
14236 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
14237 Objects may be explicitly put in the small data area with the
14238 @code{section} attribute using one of these sections.
14240 @item -msdata=sdata
14241 @opindex msdata=sdata
14242 Put small global and static data in the small data area, but do not
14243 generate special code to reference them.
14246 @opindex msdata=use
14247 Put small global and static data in the small data area, and generate
14248 special instructions to reference them.
14252 @cindex smaller data references
14253 Put global and static objects less than or equal to @var{num} bytes
14254 into the small data or bss sections instead of the normal data or bss
14255 sections. The default value of @var{num} is 8.
14256 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
14257 for this option to have any effect.
14259 All modules should be compiled with the same @option{-G @var{num}} value.
14260 Compiling with different values of @var{num} may or may not work; if it
14261 doesn't the linker will give an error message---incorrect code will not be
14266 Makes the M32R specific code in the compiler display some statistics
14267 that might help in debugging programs.
14269 @item -malign-loops
14270 @opindex malign-loops
14271 Align all loops to a 32-byte boundary.
14273 @item -mno-align-loops
14274 @opindex mno-align-loops
14275 Do not enforce a 32-byte alignment for loops. This is the default.
14277 @item -missue-rate=@var{number}
14278 @opindex missue-rate=@var{number}
14279 Issue @var{number} instructions per cycle. @var{number} can only be 1
14282 @item -mbranch-cost=@var{number}
14283 @opindex mbranch-cost=@var{number}
14284 @var{number} can only be 1 or 2. If it is 1 then branches will be
14285 preferred over conditional code, if it is 2, then the opposite will
14288 @item -mflush-trap=@var{number}
14289 @opindex mflush-trap=@var{number}
14290 Specifies the trap number to use to flush the cache. The default is
14291 12. Valid numbers are between 0 and 15 inclusive.
14293 @item -mno-flush-trap
14294 @opindex mno-flush-trap
14295 Specifies that the cache cannot be flushed by using a trap.
14297 @item -mflush-func=@var{name}
14298 @opindex mflush-func=@var{name}
14299 Specifies the name of the operating system function to call to flush
14300 the cache. The default is @emph{_flush_cache}, but a function call
14301 will only be used if a trap is not available.
14303 @item -mno-flush-func
14304 @opindex mno-flush-func
14305 Indicates that there is no OS function for flushing the cache.
14309 @node M680x0 Options
14310 @subsection M680x0 Options
14311 @cindex M680x0 options
14313 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
14314 The default settings depend on which architecture was selected when
14315 the compiler was configured; the defaults for the most common choices
14319 @item -march=@var{arch}
14321 Generate code for a specific M680x0 or ColdFire instruction set
14322 architecture. Permissible values of @var{arch} for M680x0
14323 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
14324 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
14325 architectures are selected according to Freescale's ISA classification
14326 and the permissible values are: @samp{isaa}, @samp{isaaplus},
14327 @samp{isab} and @samp{isac}.
14329 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
14330 code for a ColdFire target. The @var{arch} in this macro is one of the
14331 @option{-march} arguments given above.
14333 When used together, @option{-march} and @option{-mtune} select code
14334 that runs on a family of similar processors but that is optimized
14335 for a particular microarchitecture.
14337 @item -mcpu=@var{cpu}
14339 Generate code for a specific M680x0 or ColdFire processor.
14340 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
14341 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
14342 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
14343 below, which also classifies the CPUs into families:
14345 @multitable @columnfractions 0.20 0.80
14346 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
14347 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
14348 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
14349 @item @samp{5206e} @tab @samp{5206e}
14350 @item @samp{5208} @tab @samp{5207} @samp{5208}
14351 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
14352 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
14353 @item @samp{5216} @tab @samp{5214} @samp{5216}
14354 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
14355 @item @samp{5225} @tab @samp{5224} @samp{5225}
14356 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
14357 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
14358 @item @samp{5249} @tab @samp{5249}
14359 @item @samp{5250} @tab @samp{5250}
14360 @item @samp{5271} @tab @samp{5270} @samp{5271}
14361 @item @samp{5272} @tab @samp{5272}
14362 @item @samp{5275} @tab @samp{5274} @samp{5275}
14363 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
14364 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
14365 @item @samp{5307} @tab @samp{5307}
14366 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
14367 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
14368 @item @samp{5407} @tab @samp{5407}
14369 @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}
14372 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
14373 @var{arch} is compatible with @var{cpu}. Other combinations of
14374 @option{-mcpu} and @option{-march} are rejected.
14376 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
14377 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
14378 where the value of @var{family} is given by the table above.
14380 @item -mtune=@var{tune}
14382 Tune the code for a particular microarchitecture, within the
14383 constraints set by @option{-march} and @option{-mcpu}.
14384 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
14385 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
14386 and @samp{cpu32}. The ColdFire microarchitectures
14387 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
14389 You can also use @option{-mtune=68020-40} for code that needs
14390 to run relatively well on 68020, 68030 and 68040 targets.
14391 @option{-mtune=68020-60} is similar but includes 68060 targets
14392 as well. These two options select the same tuning decisions as
14393 @option{-m68020-40} and @option{-m68020-60} respectively.
14395 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
14396 when tuning for 680x0 architecture @var{arch}. It also defines
14397 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
14398 option is used. If gcc is tuning for a range of architectures,
14399 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
14400 it defines the macros for every architecture in the range.
14402 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
14403 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
14404 of the arguments given above.
14410 Generate output for a 68000. This is the default
14411 when the compiler is configured for 68000-based systems.
14412 It is equivalent to @option{-march=68000}.
14414 Use this option for microcontrollers with a 68000 or EC000 core,
14415 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
14419 Generate output for a 68010. This is the default
14420 when the compiler is configured for 68010-based systems.
14421 It is equivalent to @option{-march=68010}.
14427 Generate output for a 68020. This is the default
14428 when the compiler is configured for 68020-based systems.
14429 It is equivalent to @option{-march=68020}.
14433 Generate output for a 68030. This is the default when the compiler is
14434 configured for 68030-based systems. It is equivalent to
14435 @option{-march=68030}.
14439 Generate output for a 68040. This is the default when the compiler is
14440 configured for 68040-based systems. It is equivalent to
14441 @option{-march=68040}.
14443 This option inhibits the use of 68881/68882 instructions that have to be
14444 emulated by software on the 68040. Use this option if your 68040 does not
14445 have code to emulate those instructions.
14449 Generate output for a 68060. This is the default when the compiler is
14450 configured for 68060-based systems. It is equivalent to
14451 @option{-march=68060}.
14453 This option inhibits the use of 68020 and 68881/68882 instructions that
14454 have to be emulated by software on the 68060. Use this option if your 68060
14455 does not have code to emulate those instructions.
14459 Generate output for a CPU32. This is the default
14460 when the compiler is configured for CPU32-based systems.
14461 It is equivalent to @option{-march=cpu32}.
14463 Use this option for microcontrollers with a
14464 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
14465 68336, 68340, 68341, 68349 and 68360.
14469 Generate output for a 520X ColdFire CPU@. This is the default
14470 when the compiler is configured for 520X-based systems.
14471 It is equivalent to @option{-mcpu=5206}, and is now deprecated
14472 in favor of that option.
14474 Use this option for microcontroller with a 5200 core, including
14475 the MCF5202, MCF5203, MCF5204 and MCF5206.
14479 Generate output for a 5206e ColdFire CPU@. The option is now
14480 deprecated in favor of the equivalent @option{-mcpu=5206e}.
14484 Generate output for a member of the ColdFire 528X family.
14485 The option is now deprecated in favor of the equivalent
14486 @option{-mcpu=528x}.
14490 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
14491 in favor of the equivalent @option{-mcpu=5307}.
14495 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
14496 in favor of the equivalent @option{-mcpu=5407}.
14500 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
14501 This includes use of hardware floating-point instructions.
14502 The option is equivalent to @option{-mcpu=547x}, and is now
14503 deprecated in favor of that option.
14507 Generate output for a 68040, without using any of the new instructions.
14508 This results in code that can run relatively efficiently on either a
14509 68020/68881 or a 68030 or a 68040. The generated code does use the
14510 68881 instructions that are emulated on the 68040.
14512 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
14516 Generate output for a 68060, without using any of the new instructions.
14517 This results in code that can run relatively efficiently on either a
14518 68020/68881 or a 68030 or a 68040. The generated code does use the
14519 68881 instructions that are emulated on the 68060.
14521 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
14525 @opindex mhard-float
14527 Generate floating-point instructions. This is the default for 68020
14528 and above, and for ColdFire devices that have an FPU@. It defines the
14529 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
14530 on ColdFire targets.
14533 @opindex msoft-float
14534 Do not generate floating-point instructions; use library calls instead.
14535 This is the default for 68000, 68010, and 68832 targets. It is also
14536 the default for ColdFire devices that have no FPU.
14542 Generate (do not generate) ColdFire hardware divide and remainder
14543 instructions. If @option{-march} is used without @option{-mcpu},
14544 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
14545 architectures. Otherwise, the default is taken from the target CPU
14546 (either the default CPU, or the one specified by @option{-mcpu}). For
14547 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
14548 @option{-mcpu=5206e}.
14550 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
14554 Consider type @code{int} to be 16 bits wide, like @code{short int}.
14555 Additionally, parameters passed on the stack are also aligned to a
14556 16-bit boundary even on targets whose API mandates promotion to 32-bit.
14560 Do not consider type @code{int} to be 16 bits wide. This is the default.
14563 @itemx -mno-bitfield
14564 @opindex mnobitfield
14565 @opindex mno-bitfield
14566 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
14567 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
14571 Do use the bit-field instructions. The @option{-m68020} option implies
14572 @option{-mbitfield}. This is the default if you use a configuration
14573 designed for a 68020.
14577 Use a different function-calling convention, in which functions
14578 that take a fixed number of arguments return with the @code{rtd}
14579 instruction, which pops their arguments while returning. This
14580 saves one instruction in the caller since there is no need to pop
14581 the arguments there.
14583 This calling convention is incompatible with the one normally
14584 used on Unix, so you cannot use it if you need to call libraries
14585 compiled with the Unix compiler.
14587 Also, you must provide function prototypes for all functions that
14588 take variable numbers of arguments (including @code{printf});
14589 otherwise incorrect code will be generated for calls to those
14592 In addition, seriously incorrect code will result if you call a
14593 function with too many arguments. (Normally, extra arguments are
14594 harmlessly ignored.)
14596 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
14597 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
14601 Do not use the calling conventions selected by @option{-mrtd}.
14602 This is the default.
14605 @itemx -mno-align-int
14606 @opindex malign-int
14607 @opindex mno-align-int
14608 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
14609 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
14610 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
14611 Aligning variables on 32-bit boundaries produces code that runs somewhat
14612 faster on processors with 32-bit busses at the expense of more memory.
14614 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
14615 align structures containing the above types differently than
14616 most published application binary interface specifications for the m68k.
14620 Use the pc-relative addressing mode of the 68000 directly, instead of
14621 using a global offset table. At present, this option implies @option{-fpic},
14622 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
14623 not presently supported with @option{-mpcrel}, though this could be supported for
14624 68020 and higher processors.
14626 @item -mno-strict-align
14627 @itemx -mstrict-align
14628 @opindex mno-strict-align
14629 @opindex mstrict-align
14630 Do not (do) assume that unaligned memory references will be handled by
14634 Generate code that allows the data segment to be located in a different
14635 area of memory from the text segment. This allows for execute in place in
14636 an environment without virtual memory management. This option implies
14639 @item -mno-sep-data
14640 Generate code that assumes that the data segment follows the text segment.
14641 This is the default.
14643 @item -mid-shared-library
14644 Generate code that supports shared libraries via the library ID method.
14645 This allows for execute in place and shared libraries in an environment
14646 without virtual memory management. This option implies @option{-fPIC}.
14648 @item -mno-id-shared-library
14649 Generate code that doesn't assume ID based shared libraries are being used.
14650 This is the default.
14652 @item -mshared-library-id=n
14653 Specified the identification number of the ID based shared library being
14654 compiled. Specifying a value of 0 will generate more compact code, specifying
14655 other values will force the allocation of that number to the current
14656 library but is no more space or time efficient than omitting this option.
14662 When generating position-independent code for ColdFire, generate code
14663 that works if the GOT has more than 8192 entries. This code is
14664 larger and slower than code generated without this option. On M680x0
14665 processors, this option is not needed; @option{-fPIC} suffices.
14667 GCC normally uses a single instruction to load values from the GOT@.
14668 While this is relatively efficient, it only works if the GOT
14669 is smaller than about 64k. Anything larger causes the linker
14670 to report an error such as:
14672 @cindex relocation truncated to fit (ColdFire)
14674 relocation truncated to fit: R_68K_GOT16O foobar
14677 If this happens, you should recompile your code with @option{-mxgot}.
14678 It should then work with very large GOTs. However, code generated with
14679 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14680 the value of a global symbol.
14682 Note that some linkers, including newer versions of the GNU linker,
14683 can create multiple GOTs and sort GOT entries. If you have such a linker,
14684 you should only need to use @option{-mxgot} when compiling a single
14685 object file that accesses more than 8192 GOT entries. Very few do.
14687 These options have no effect unless GCC is generating
14688 position-independent code.
14692 @node MCore Options
14693 @subsection MCore Options
14694 @cindex MCore options
14696 These are the @samp{-m} options defined for the Motorola M*Core
14702 @itemx -mno-hardlit
14704 @opindex mno-hardlit
14705 Inline constants into the code stream if it can be done in two
14706 instructions or less.
14712 Use the divide instruction. (Enabled by default).
14714 @item -mrelax-immediate
14715 @itemx -mno-relax-immediate
14716 @opindex mrelax-immediate
14717 @opindex mno-relax-immediate
14718 Allow arbitrary sized immediates in bit operations.
14720 @item -mwide-bitfields
14721 @itemx -mno-wide-bitfields
14722 @opindex mwide-bitfields
14723 @opindex mno-wide-bitfields
14724 Always treat bit-fields as int-sized.
14726 @item -m4byte-functions
14727 @itemx -mno-4byte-functions
14728 @opindex m4byte-functions
14729 @opindex mno-4byte-functions
14730 Force all functions to be aligned to a 4-byte boundary.
14732 @item -mcallgraph-data
14733 @itemx -mno-callgraph-data
14734 @opindex mcallgraph-data
14735 @opindex mno-callgraph-data
14736 Emit callgraph information.
14739 @itemx -mno-slow-bytes
14740 @opindex mslow-bytes
14741 @opindex mno-slow-bytes
14742 Prefer word access when reading byte quantities.
14744 @item -mlittle-endian
14745 @itemx -mbig-endian
14746 @opindex mlittle-endian
14747 @opindex mbig-endian
14748 Generate code for a little-endian target.
14754 Generate code for the 210 processor.
14758 Assume that runtime support has been provided and so omit the
14759 simulator library (@file{libsim.a)} from the linker command line.
14761 @item -mstack-increment=@var{size}
14762 @opindex mstack-increment
14763 Set the maximum amount for a single stack increment operation. Large
14764 values can increase the speed of programs that contain functions
14765 that need a large amount of stack space, but they can also trigger a
14766 segmentation fault if the stack is extended too much. The default
14772 @subsection MeP Options
14773 @cindex MeP options
14779 Enables the @code{abs} instruction, which is the absolute difference
14780 between two registers.
14784 Enables all the optional instructions - average, multiply, divide, bit
14785 operations, leading zero, absolute difference, min/max, clip, and
14791 Enables the @code{ave} instruction, which computes the average of two
14794 @item -mbased=@var{n}
14796 Variables of size @var{n} bytes or smaller will be placed in the
14797 @code{.based} section by default. Based variables use the @code{$tp}
14798 register as a base register, and there is a 128-byte limit to the
14799 @code{.based} section.
14803 Enables the bit operation instructions - bit test (@code{btstm}), set
14804 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14805 test-and-set (@code{tas}).
14807 @item -mc=@var{name}
14809 Selects which section constant data will be placed in. @var{name} may
14810 be @code{tiny}, @code{near}, or @code{far}.
14814 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14815 useful unless you also provide @code{-mminmax}.
14817 @item -mconfig=@var{name}
14819 Selects one of the build-in core configurations. Each MeP chip has
14820 one or more modules in it; each module has a core CPU and a variety of
14821 coprocessors, optional instructions, and peripherals. The
14822 @code{MeP-Integrator} tool, not part of GCC, provides these
14823 configurations through this option; using this option is the same as
14824 using all the corresponding command-line options. The default
14825 configuration is @code{default}.
14829 Enables the coprocessor instructions. By default, this is a 32-bit
14830 coprocessor. Note that the coprocessor is normally enabled via the
14831 @code{-mconfig=} option.
14835 Enables the 32-bit coprocessor's instructions.
14839 Enables the 64-bit coprocessor's instructions.
14843 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14847 Causes constant variables to be placed in the @code{.near} section.
14851 Enables the @code{div} and @code{divu} instructions.
14855 Generate big-endian code.
14859 Generate little-endian code.
14861 @item -mio-volatile
14862 @opindex mio-volatile
14863 Tells the compiler that any variable marked with the @code{io}
14864 attribute is to be considered volatile.
14868 Causes variables to be assigned to the @code{.far} section by default.
14872 Enables the @code{leadz} (leading zero) instruction.
14876 Causes variables to be assigned to the @code{.near} section by default.
14880 Enables the @code{min} and @code{max} instructions.
14884 Enables the multiplication and multiply-accumulate instructions.
14888 Disables all the optional instructions enabled by @code{-mall-opts}.
14892 Enables the @code{repeat} and @code{erepeat} instructions, used for
14893 low-overhead looping.
14897 Causes all variables to default to the @code{.tiny} section. Note
14898 that there is a 65536-byte limit to this section. Accesses to these
14899 variables use the @code{%gp} base register.
14903 Enables the saturation instructions. Note that the compiler does not
14904 currently generate these itself, but this option is included for
14905 compatibility with other tools, like @code{as}.
14909 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14913 Link the simulator runtime libraries.
14917 Link the simulator runtime libraries, excluding built-in support
14918 for reset and exception vectors and tables.
14922 Causes all functions to default to the @code{.far} section. Without
14923 this option, functions default to the @code{.near} section.
14925 @item -mtiny=@var{n}
14927 Variables that are @var{n} bytes or smaller will be allocated to the
14928 @code{.tiny} section. These variables use the @code{$gp} base
14929 register. The default for this option is 4, but note that there's a
14930 65536-byte limit to the @code{.tiny} section.
14934 @node MicroBlaze Options
14935 @subsection MicroBlaze Options
14936 @cindex MicroBlaze Options
14941 @opindex msoft-float
14942 Use software emulation for floating point (default).
14945 @opindex mhard-float
14946 Use hardware floating-point instructions.
14950 Do not optimize block moves, use @code{memcpy}.
14952 @item -mno-clearbss
14953 @opindex mno-clearbss
14954 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14956 @item -mcpu=@var{cpu-type}
14958 Use features of and schedule code for given CPU.
14959 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14960 where @var{X} is a major version, @var{YY} is the minor version, and
14961 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14962 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14964 @item -mxl-soft-mul
14965 @opindex mxl-soft-mul
14966 Use software multiply emulation (default).
14968 @item -mxl-soft-div
14969 @opindex mxl-soft-div
14970 Use software emulation for divides (default).
14972 @item -mxl-barrel-shift
14973 @opindex mxl-barrel-shift
14974 Use the hardware barrel shifter.
14976 @item -mxl-pattern-compare
14977 @opindex mxl-pattern-compare
14978 Use pattern compare instructions.
14980 @item -msmall-divides
14981 @opindex msmall-divides
14982 Use table lookup optimization for small signed integer divisions.
14984 @item -mxl-stack-check
14985 @opindex mxl-stack-check
14986 This option is deprecated. Use -fstack-check instead.
14989 @opindex mxl-gp-opt
14990 Use GP relative sdata/sbss sections.
14992 @item -mxl-multiply-high
14993 @opindex mxl-multiply-high
14994 Use multiply high instructions for high part of 32x32 multiply.
14996 @item -mxl-float-convert
14997 @opindex mxl-float-convert
14998 Use hardware floating-point conversion instructions.
15000 @item -mxl-float-sqrt
15001 @opindex mxl-float-sqrt
15002 Use hardware floating-point square root instruction.
15004 @item -mxl-mode-@var{app-model}
15005 Select application model @var{app-model}. Valid models are
15008 normal executable (default), uses startup code @file{crt0.o}.
15011 for use with Xilinx Microprocessor Debugger (XMD) based
15012 software intrusive debug agent called xmdstub. This uses startup file
15013 @file{crt1.o} and sets the start address of the program to be 0x800.
15016 for applications that are loaded using a bootloader.
15017 This model uses startup file @file{crt2.o} which does not contain a processor
15018 reset vector handler. This is suitable for transferring control on a
15019 processor reset to the bootloader rather than the application.
15022 for applications that do not require any of the
15023 MicroBlaze vectors. This option may be useful for applications running
15024 within a monitoring application. This model uses @file{crt3.o} as a startup file.
15027 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
15028 @option{-mxl-mode-@var{app-model}}.
15033 @subsection MIPS Options
15034 @cindex MIPS options
15040 Generate big-endian code.
15044 Generate little-endian code. This is the default for @samp{mips*el-*-*}
15047 @item -march=@var{arch}
15049 Generate code that will run on @var{arch}, which can be the name of a
15050 generic MIPS ISA, or the name of a particular processor.
15052 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
15053 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
15054 The processor names are:
15055 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
15056 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
15057 @samp{5kc}, @samp{5kf},
15059 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
15060 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
15061 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
15062 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
15063 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
15064 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
15066 @samp{octeon}, @samp{octeon+}, @samp{octeon2},
15068 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
15069 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
15070 @samp{rm7000}, @samp{rm9000},
15071 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
15074 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
15075 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
15077 The special value @samp{from-abi} selects the
15078 most compatible architecture for the selected ABI (that is,
15079 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
15081 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
15082 which selects the best architecture option for the host processor.
15083 @option{-march=native} has no effect if GCC does not recognize
15086 In processor names, a final @samp{000} can be abbreviated as @samp{k}
15087 (for example, @samp{-march=r2k}). Prefixes are optional, and
15088 @samp{vr} may be written @samp{r}.
15090 Names of the form @samp{@var{n}f2_1} refer to processors with
15091 FPUs clocked at half the rate of the core, names of the form
15092 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
15093 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
15094 processors with FPUs clocked a ratio of 3:2 with respect to the core.
15095 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
15096 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
15097 accepted as synonyms for @samp{@var{n}f1_1}.
15099 GCC defines two macros based on the value of this option. The first
15100 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
15101 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
15102 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
15103 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
15104 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
15106 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
15107 above. In other words, it will have the full prefix and will not
15108 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
15109 the macro names the resolved architecture (either @samp{"mips1"} or
15110 @samp{"mips3"}). It names the default architecture when no
15111 @option{-march} option is given.
15113 @item -mtune=@var{arch}
15115 Optimize for @var{arch}. Among other things, this option controls
15116 the way instructions are scheduled, and the perceived cost of arithmetic
15117 operations. The list of @var{arch} values is the same as for
15120 When this option is not used, GCC will optimize for the processor
15121 specified by @option{-march}. By using @option{-march} and
15122 @option{-mtune} together, it is possible to generate code that will
15123 run on a family of processors, but optimize the code for one
15124 particular member of that family.
15126 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
15127 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
15128 @samp{-march} ones described above.
15132 Equivalent to @samp{-march=mips1}.
15136 Equivalent to @samp{-march=mips2}.
15140 Equivalent to @samp{-march=mips3}.
15144 Equivalent to @samp{-march=mips4}.
15148 Equivalent to @samp{-march=mips32}.
15152 Equivalent to @samp{-march=mips32r2}.
15156 Equivalent to @samp{-march=mips64}.
15160 Equivalent to @samp{-march=mips64r2}.
15165 @opindex mno-mips16
15166 Generate (do not generate) MIPS16 code. If GCC is targetting a
15167 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
15169 MIPS16 code generation can also be controlled on a per-function basis
15170 by means of @code{mips16} and @code{nomips16} attributes.
15171 @xref{Function Attributes}, for more information.
15173 @item -mflip-mips16
15174 @opindex mflip-mips16
15175 Generate MIPS16 code on alternating functions. This option is provided
15176 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
15177 not intended for ordinary use in compiling user code.
15179 @item -minterlink-mips16
15180 @itemx -mno-interlink-mips16
15181 @opindex minterlink-mips16
15182 @opindex mno-interlink-mips16
15183 Require (do not require) that non-MIPS16 code be link-compatible with
15186 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
15187 it must either use a call or an indirect jump. @option{-minterlink-mips16}
15188 therefore disables direct jumps unless GCC knows that the target of the
15189 jump is not MIPS16.
15201 Generate code for the given ABI@.
15203 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
15204 generates 64-bit code when you select a 64-bit architecture, but you
15205 can use @option{-mgp32} to get 32-bit code instead.
15207 For information about the O64 ABI, see
15208 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
15210 GCC supports a variant of the o32 ABI in which floating-point registers
15211 are 64 rather than 32 bits wide. You can select this combination with
15212 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
15213 and @samp{mfhc1} instructions and is therefore only supported for
15214 MIPS32R2 processors.
15216 The register assignments for arguments and return values remain the
15217 same, but each scalar value is passed in a single 64-bit register
15218 rather than a pair of 32-bit registers. For example, scalar
15219 floating-point values are returned in @samp{$f0} only, not a
15220 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
15221 remains the same, but all 64 bits are saved.
15224 @itemx -mno-abicalls
15226 @opindex mno-abicalls
15227 Generate (do not generate) code that is suitable for SVR4-style
15228 dynamic objects. @option{-mabicalls} is the default for SVR4-based
15233 Generate (do not generate) code that is fully position-independent,
15234 and that can therefore be linked into shared libraries. This option
15235 only affects @option{-mabicalls}.
15237 All @option{-mabicalls} code has traditionally been position-independent,
15238 regardless of options like @option{-fPIC} and @option{-fpic}. However,
15239 as an extension, the GNU toolchain allows executables to use absolute
15240 accesses for locally-binding symbols. It can also use shorter GP
15241 initialization sequences and generate direct calls to locally-defined
15242 functions. This mode is selected by @option{-mno-shared}.
15244 @option{-mno-shared} depends on binutils 2.16 or higher and generates
15245 objects that can only be linked by the GNU linker. However, the option
15246 does not affect the ABI of the final executable; it only affects the ABI
15247 of relocatable objects. Using @option{-mno-shared} will generally make
15248 executables both smaller and quicker.
15250 @option{-mshared} is the default.
15256 Assume (do not assume) that the static and dynamic linkers
15257 support PLTs and copy relocations. This option only affects
15258 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
15259 has no effect without @samp{-msym32}.
15261 You can make @option{-mplt} the default by configuring
15262 GCC with @option{--with-mips-plt}. The default is
15263 @option{-mno-plt} otherwise.
15269 Lift (do not lift) the usual restrictions on the size of the global
15272 GCC normally uses a single instruction to load values from the GOT@.
15273 While this is relatively efficient, it will only work if the GOT
15274 is smaller than about 64k. Anything larger will cause the linker
15275 to report an error such as:
15277 @cindex relocation truncated to fit (MIPS)
15279 relocation truncated to fit: R_MIPS_GOT16 foobar
15282 If this happens, you should recompile your code with @option{-mxgot}.
15283 It should then work with very large GOTs, although it will also be
15284 less efficient, since it will take three instructions to fetch the
15285 value of a global symbol.
15287 Note that some linkers can create multiple GOTs. If you have such a
15288 linker, you should only need to use @option{-mxgot} when a single object
15289 file accesses more than 64k's worth of GOT entries. Very few do.
15291 These options have no effect unless GCC is generating position
15296 Assume that general-purpose registers are 32 bits wide.
15300 Assume that general-purpose registers are 64 bits wide.
15304 Assume that floating-point registers are 32 bits wide.
15308 Assume that floating-point registers are 64 bits wide.
15311 @opindex mhard-float
15312 Use floating-point coprocessor instructions.
15315 @opindex msoft-float
15316 Do not use floating-point coprocessor instructions. Implement
15317 floating-point calculations using library calls instead.
15319 @item -msingle-float
15320 @opindex msingle-float
15321 Assume that the floating-point coprocessor only supports single-precision
15324 @item -mdouble-float
15325 @opindex mdouble-float
15326 Assume that the floating-point coprocessor supports double-precision
15327 operations. This is the default.
15333 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
15334 implement atomic memory built-in functions. When neither option is
15335 specified, GCC will use the instructions if the target architecture
15338 @option{-mllsc} is useful if the runtime environment can emulate the
15339 instructions and @option{-mno-llsc} can be useful when compiling for
15340 nonstandard ISAs. You can make either option the default by
15341 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
15342 respectively. @option{--with-llsc} is the default for some
15343 configurations; see the installation documentation for details.
15349 Use (do not use) revision 1 of the MIPS DSP ASE@.
15350 @xref{MIPS DSP Built-in Functions}. This option defines the
15351 preprocessor macro @samp{__mips_dsp}. It also defines
15352 @samp{__mips_dsp_rev} to 1.
15358 Use (do not use) revision 2 of the MIPS DSP ASE@.
15359 @xref{MIPS DSP Built-in Functions}. This option defines the
15360 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
15361 It also defines @samp{__mips_dsp_rev} to 2.
15364 @itemx -mno-smartmips
15365 @opindex msmartmips
15366 @opindex mno-smartmips
15367 Use (do not use) the MIPS SmartMIPS ASE.
15369 @item -mpaired-single
15370 @itemx -mno-paired-single
15371 @opindex mpaired-single
15372 @opindex mno-paired-single
15373 Use (do not use) paired-single floating-point instructions.
15374 @xref{MIPS Paired-Single Support}. This option requires
15375 hardware floating-point support to be enabled.
15381 Use (do not use) MIPS Digital Media Extension instructions.
15382 This option can only be used when generating 64-bit code and requires
15383 hardware floating-point support to be enabled.
15388 @opindex mno-mips3d
15389 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
15390 The option @option{-mips3d} implies @option{-mpaired-single}.
15396 Use (do not use) MT Multithreading instructions.
15400 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
15401 an explanation of the default and the way that the pointer size is
15406 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
15408 The default size of @code{int}s, @code{long}s and pointers depends on
15409 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
15410 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
15411 32-bit @code{long}s. Pointers are the same size as @code{long}s,
15412 or the same size as integer registers, whichever is smaller.
15418 Assume (do not assume) that all symbols have 32-bit values, regardless
15419 of the selected ABI@. This option is useful in combination with
15420 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
15421 to generate shorter and faster references to symbolic addresses.
15425 Put definitions of externally-visible data in a small data section
15426 if that data is no bigger than @var{num} bytes. GCC can then access
15427 the data more efficiently; see @option{-mgpopt} for details.
15429 The default @option{-G} option depends on the configuration.
15431 @item -mlocal-sdata
15432 @itemx -mno-local-sdata
15433 @opindex mlocal-sdata
15434 @opindex mno-local-sdata
15435 Extend (do not extend) the @option{-G} behavior to local data too,
15436 such as to static variables in C@. @option{-mlocal-sdata} is the
15437 default for all configurations.
15439 If the linker complains that an application is using too much small data,
15440 you might want to try rebuilding the less performance-critical parts with
15441 @option{-mno-local-sdata}. You might also want to build large
15442 libraries with @option{-mno-local-sdata}, so that the libraries leave
15443 more room for the main program.
15445 @item -mextern-sdata
15446 @itemx -mno-extern-sdata
15447 @opindex mextern-sdata
15448 @opindex mno-extern-sdata
15449 Assume (do not assume) that externally-defined data will be in
15450 a small data section if that data is within the @option{-G} limit.
15451 @option{-mextern-sdata} is the default for all configurations.
15453 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
15454 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
15455 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
15456 is placed in a small data section. If @var{Var} is defined by another
15457 module, you must either compile that module with a high-enough
15458 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
15459 definition. If @var{Var} is common, you must link the application
15460 with a high-enough @option{-G} setting.
15462 The easiest way of satisfying these restrictions is to compile
15463 and link every module with the same @option{-G} option. However,
15464 you may wish to build a library that supports several different
15465 small data limits. You can do this by compiling the library with
15466 the highest supported @option{-G} setting and additionally using
15467 @option{-mno-extern-sdata} to stop the library from making assumptions
15468 about externally-defined data.
15474 Use (do not use) GP-relative accesses for symbols that are known to be
15475 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
15476 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
15479 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
15480 might not hold the value of @code{_gp}. For example, if the code is
15481 part of a library that might be used in a boot monitor, programs that
15482 call boot monitor routines will pass an unknown value in @code{$gp}.
15483 (In such situations, the boot monitor itself would usually be compiled
15484 with @option{-G0}.)
15486 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
15487 @option{-mno-extern-sdata}.
15489 @item -membedded-data
15490 @itemx -mno-embedded-data
15491 @opindex membedded-data
15492 @opindex mno-embedded-data
15493 Allocate variables to the read-only data section first if possible, then
15494 next in the small data section if possible, otherwise in data. This gives
15495 slightly slower code than the default, but reduces the amount of RAM required
15496 when executing, and thus may be preferred for some embedded systems.
15498 @item -muninit-const-in-rodata
15499 @itemx -mno-uninit-const-in-rodata
15500 @opindex muninit-const-in-rodata
15501 @opindex mno-uninit-const-in-rodata
15502 Put uninitialized @code{const} variables in the read-only data section.
15503 This option is only meaningful in conjunction with @option{-membedded-data}.
15505 @item -mcode-readable=@var{setting}
15506 @opindex mcode-readable
15507 Specify whether GCC may generate code that reads from executable sections.
15508 There are three possible settings:
15511 @item -mcode-readable=yes
15512 Instructions may freely access executable sections. This is the
15515 @item -mcode-readable=pcrel
15516 MIPS16 PC-relative load instructions can access executable sections,
15517 but other instructions must not do so. This option is useful on 4KSc
15518 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
15519 It is also useful on processors that can be configured to have a dual
15520 instruction/data SRAM interface and that, like the M4K, automatically
15521 redirect PC-relative loads to the instruction RAM.
15523 @item -mcode-readable=no
15524 Instructions must not access executable sections. This option can be
15525 useful on targets that are configured to have a dual instruction/data
15526 SRAM interface but that (unlike the M4K) do not automatically redirect
15527 PC-relative loads to the instruction RAM.
15530 @item -msplit-addresses
15531 @itemx -mno-split-addresses
15532 @opindex msplit-addresses
15533 @opindex mno-split-addresses
15534 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
15535 relocation operators. This option has been superseded by
15536 @option{-mexplicit-relocs} but is retained for backwards compatibility.
15538 @item -mexplicit-relocs
15539 @itemx -mno-explicit-relocs
15540 @opindex mexplicit-relocs
15541 @opindex mno-explicit-relocs
15542 Use (do not use) assembler relocation operators when dealing with symbolic
15543 addresses. The alternative, selected by @option{-mno-explicit-relocs},
15544 is to use assembler macros instead.
15546 @option{-mexplicit-relocs} is the default if GCC was configured
15547 to use an assembler that supports relocation operators.
15549 @item -mcheck-zero-division
15550 @itemx -mno-check-zero-division
15551 @opindex mcheck-zero-division
15552 @opindex mno-check-zero-division
15553 Trap (do not trap) on integer division by zero.
15555 The default is @option{-mcheck-zero-division}.
15557 @item -mdivide-traps
15558 @itemx -mdivide-breaks
15559 @opindex mdivide-traps
15560 @opindex mdivide-breaks
15561 MIPS systems check for division by zero by generating either a
15562 conditional trap or a break instruction. Using traps results in
15563 smaller code, but is only supported on MIPS II and later. Also, some
15564 versions of the Linux kernel have a bug that prevents trap from
15565 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
15566 allow conditional traps on architectures that support them and
15567 @option{-mdivide-breaks} to force the use of breaks.
15569 The default is usually @option{-mdivide-traps}, but this can be
15570 overridden at configure time using @option{--with-divide=breaks}.
15571 Divide-by-zero checks can be completely disabled using
15572 @option{-mno-check-zero-division}.
15577 @opindex mno-memcpy
15578 Force (do not force) the use of @code{memcpy()} for non-trivial block
15579 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
15580 most constant-sized copies.
15583 @itemx -mno-long-calls
15584 @opindex mlong-calls
15585 @opindex mno-long-calls
15586 Disable (do not disable) use of the @code{jal} instruction. Calling
15587 functions using @code{jal} is more efficient but requires the caller
15588 and callee to be in the same 256 megabyte segment.
15590 This option has no effect on abicalls code. The default is
15591 @option{-mno-long-calls}.
15597 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
15598 instructions, as provided by the R4650 ISA@.
15601 @itemx -mno-fused-madd
15602 @opindex mfused-madd
15603 @opindex mno-fused-madd
15604 Enable (disable) use of the floating-point multiply-accumulate
15605 instructions, when they are available. The default is
15606 @option{-mfused-madd}.
15608 When multiply-accumulate instructions are used, the intermediate
15609 product is calculated to infinite precision and is not subject to
15610 the FCSR Flush to Zero bit. This may be undesirable in some
15615 Tell the MIPS assembler to not run its preprocessor over user
15616 assembler files (with a @samp{.s} suffix) when assembling them.
15621 @opindex mno-fix-24k
15622 Work around the 24K E48 (lost data on stores during refill) errata.
15623 The workarounds are implemented by the assembler rather than by GCC.
15626 @itemx -mno-fix-r4000
15627 @opindex mfix-r4000
15628 @opindex mno-fix-r4000
15629 Work around certain R4000 CPU errata:
15632 A double-word or a variable shift may give an incorrect result if executed
15633 immediately after starting an integer division.
15635 A double-word or a variable shift may give an incorrect result if executed
15636 while an integer multiplication is in progress.
15638 An integer division may give an incorrect result if started in a delay slot
15639 of a taken branch or a jump.
15643 @itemx -mno-fix-r4400
15644 @opindex mfix-r4400
15645 @opindex mno-fix-r4400
15646 Work around certain R4400 CPU errata:
15649 A double-word or a variable shift may give an incorrect result if executed
15650 immediately after starting an integer division.
15654 @itemx -mno-fix-r10000
15655 @opindex mfix-r10000
15656 @opindex mno-fix-r10000
15657 Work around certain R10000 errata:
15660 @code{ll}/@code{sc} sequences may not behave atomically on revisions
15661 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15664 This option can only be used if the target architecture supports
15665 branch-likely instructions. @option{-mfix-r10000} is the default when
15666 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15670 @itemx -mno-fix-vr4120
15671 @opindex mfix-vr4120
15672 Work around certain VR4120 errata:
15675 @code{dmultu} does not always produce the correct result.
15677 @code{div} and @code{ddiv} do not always produce the correct result if one
15678 of the operands is negative.
15680 The workarounds for the division errata rely on special functions in
15681 @file{libgcc.a}. At present, these functions are only provided by
15682 the @code{mips64vr*-elf} configurations.
15684 Other VR4120 errata require a nop to be inserted between certain pairs of
15685 instructions. These errata are handled by the assembler, not by GCC itself.
15688 @opindex mfix-vr4130
15689 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15690 workarounds are implemented by the assembler rather than by GCC,
15691 although GCC will avoid using @code{mflo} and @code{mfhi} if the
15692 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15693 instructions are available instead.
15696 @itemx -mno-fix-sb1
15698 Work around certain SB-1 CPU core errata.
15699 (This flag currently works around the SB-1 revision 2
15700 ``F1'' and ``F2'' floating-point errata.)
15702 @item -mr10k-cache-barrier=@var{setting}
15703 @opindex mr10k-cache-barrier
15704 Specify whether GCC should insert cache barriers to avoid the
15705 side-effects of speculation on R10K processors.
15707 In common with many processors, the R10K tries to predict the outcome
15708 of a conditional branch and speculatively executes instructions from
15709 the ``taken'' branch. It later aborts these instructions if the
15710 predicted outcome was wrong. However, on the R10K, even aborted
15711 instructions can have side effects.
15713 This problem only affects kernel stores and, depending on the system,
15714 kernel loads. As an example, a speculatively-executed store may load
15715 the target memory into cache and mark the cache line as dirty, even if
15716 the store itself is later aborted. If a DMA operation writes to the
15717 same area of memory before the ``dirty'' line is flushed, the cached
15718 data will overwrite the DMA-ed data. See the R10K processor manual
15719 for a full description, including other potential problems.
15721 One workaround is to insert cache barrier instructions before every memory
15722 access that might be speculatively executed and that might have side
15723 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15724 controls GCC's implementation of this workaround. It assumes that
15725 aborted accesses to any byte in the following regions will not have
15730 the memory occupied by the current function's stack frame;
15733 the memory occupied by an incoming stack argument;
15736 the memory occupied by an object with a link-time-constant address.
15739 It is the kernel's responsibility to ensure that speculative
15740 accesses to these regions are indeed safe.
15742 If the input program contains a function declaration such as:
15748 then the implementation of @code{foo} must allow @code{j foo} and
15749 @code{jal foo} to be executed speculatively. GCC honors this
15750 restriction for functions it compiles itself. It expects non-GCC
15751 functions (such as hand-written assembly code) to do the same.
15753 The option has three forms:
15756 @item -mr10k-cache-barrier=load-store
15757 Insert a cache barrier before a load or store that might be
15758 speculatively executed and that might have side effects even
15761 @item -mr10k-cache-barrier=store
15762 Insert a cache barrier before a store that might be speculatively
15763 executed and that might have side effects even if aborted.
15765 @item -mr10k-cache-barrier=none
15766 Disable the insertion of cache barriers. This is the default setting.
15769 @item -mflush-func=@var{func}
15770 @itemx -mno-flush-func
15771 @opindex mflush-func
15772 Specifies the function to call to flush the I and D caches, or to not
15773 call any such function. If called, the function must take the same
15774 arguments as the common @code{_flush_func()}, that is, the address of the
15775 memory range for which the cache is being flushed, the size of the
15776 memory range, and the number 3 (to flush both caches). The default
15777 depends on the target GCC was configured for, but commonly is either
15778 @samp{_flush_func} or @samp{__cpu_flush}.
15780 @item mbranch-cost=@var{num}
15781 @opindex mbranch-cost
15782 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15783 This cost is only a heuristic and is not guaranteed to produce
15784 consistent results across releases. A zero cost redundantly selects
15785 the default, which is based on the @option{-mtune} setting.
15787 @item -mbranch-likely
15788 @itemx -mno-branch-likely
15789 @opindex mbranch-likely
15790 @opindex mno-branch-likely
15791 Enable or disable use of Branch Likely instructions, regardless of the
15792 default for the selected architecture. By default, Branch Likely
15793 instructions may be generated if they are supported by the selected
15794 architecture. An exception is for the MIPS32 and MIPS64 architectures
15795 and processors that implement those architectures; for those, Branch
15796 Likely instructions will not be generated by default because the MIPS32
15797 and MIPS64 architectures specifically deprecate their use.
15799 @item -mfp-exceptions
15800 @itemx -mno-fp-exceptions
15801 @opindex mfp-exceptions
15802 Specifies whether FP exceptions are enabled. This affects how we schedule
15803 FP instructions for some processors. The default is that FP exceptions are
15806 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15807 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15810 @item -mvr4130-align
15811 @itemx -mno-vr4130-align
15812 @opindex mvr4130-align
15813 The VR4130 pipeline is two-way superscalar, but can only issue two
15814 instructions together if the first one is 8-byte aligned. When this
15815 option is enabled, GCC will align pairs of instructions that it
15816 thinks should execute in parallel.
15818 This option only has an effect when optimizing for the VR4130.
15819 It normally makes code faster, but at the expense of making it bigger.
15820 It is enabled by default at optimization level @option{-O3}.
15825 Enable (disable) generation of @code{synci} instructions on
15826 architectures that support it. The @code{synci} instructions (if
15827 enabled) will be generated when @code{__builtin___clear_cache()} is
15830 This option defaults to @code{-mno-synci}, but the default can be
15831 overridden by configuring with @code{--with-synci}.
15833 When compiling code for single processor systems, it is generally safe
15834 to use @code{synci}. However, on many multi-core (SMP) systems, it
15835 will not invalidate the instruction caches on all cores and may lead
15836 to undefined behavior.
15838 @item -mrelax-pic-calls
15839 @itemx -mno-relax-pic-calls
15840 @opindex mrelax-pic-calls
15841 Try to turn PIC calls that are normally dispatched via register
15842 @code{$25} into direct calls. This is only possible if the linker can
15843 resolve the destination at link-time and if the destination is within
15844 range for a direct call.
15846 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15847 an assembler and a linker that supports the @code{.reloc} assembly
15848 directive and @code{-mexplicit-relocs} is in effect. With
15849 @code{-mno-explicit-relocs}, this optimization can be performed by the
15850 assembler and the linker alone without help from the compiler.
15852 @item -mmcount-ra-address
15853 @itemx -mno-mcount-ra-address
15854 @opindex mmcount-ra-address
15855 @opindex mno-mcount-ra-address
15856 Emit (do not emit) code that allows @code{_mcount} to modify the
15857 calling function's return address. When enabled, this option extends
15858 the usual @code{_mcount} interface with a new @var{ra-address}
15859 parameter, which has type @code{intptr_t *} and is passed in register
15860 @code{$12}. @code{_mcount} can then modify the return address by
15861 doing both of the following:
15864 Returning the new address in register @code{$31}.
15866 Storing the new address in @code{*@var{ra-address}},
15867 if @var{ra-address} is nonnull.
15870 The default is @option{-mno-mcount-ra-address}.
15875 @subsection MMIX Options
15876 @cindex MMIX Options
15878 These options are defined for the MMIX:
15882 @itemx -mno-libfuncs
15884 @opindex mno-libfuncs
15885 Specify that intrinsic library functions are being compiled, passing all
15886 values in registers, no matter the size.
15889 @itemx -mno-epsilon
15891 @opindex mno-epsilon
15892 Generate floating-point comparison instructions that compare with respect
15893 to the @code{rE} epsilon register.
15895 @item -mabi=mmixware
15897 @opindex mabi=mmixware
15899 Generate code that passes function parameters and return values that (in
15900 the called function) are seen as registers @code{$0} and up, as opposed to
15901 the GNU ABI which uses global registers @code{$231} and up.
15903 @item -mzero-extend
15904 @itemx -mno-zero-extend
15905 @opindex mzero-extend
15906 @opindex mno-zero-extend
15907 When reading data from memory in sizes shorter than 64 bits, use (do not
15908 use) zero-extending load instructions by default, rather than
15909 sign-extending ones.
15912 @itemx -mno-knuthdiv
15914 @opindex mno-knuthdiv
15915 Make the result of a division yielding a remainder have the same sign as
15916 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15917 remainder follows the sign of the dividend. Both methods are
15918 arithmetically valid, the latter being almost exclusively used.
15920 @item -mtoplevel-symbols
15921 @itemx -mno-toplevel-symbols
15922 @opindex mtoplevel-symbols
15923 @opindex mno-toplevel-symbols
15924 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15925 code can be used with the @code{PREFIX} assembly directive.
15929 Generate an executable in the ELF format, rather than the default
15930 @samp{mmo} format used by the @command{mmix} simulator.
15932 @item -mbranch-predict
15933 @itemx -mno-branch-predict
15934 @opindex mbranch-predict
15935 @opindex mno-branch-predict
15936 Use (do not use) the probable-branch instructions, when static branch
15937 prediction indicates a probable branch.
15939 @item -mbase-addresses
15940 @itemx -mno-base-addresses
15941 @opindex mbase-addresses
15942 @opindex mno-base-addresses
15943 Generate (do not generate) code that uses @emph{base addresses}. Using a
15944 base address automatically generates a request (handled by the assembler
15945 and the linker) for a constant to be set up in a global register. The
15946 register is used for one or more base address requests within the range 0
15947 to 255 from the value held in the register. The generally leads to short
15948 and fast code, but the number of different data items that can be
15949 addressed is limited. This means that a program that uses lots of static
15950 data may require @option{-mno-base-addresses}.
15952 @item -msingle-exit
15953 @itemx -mno-single-exit
15954 @opindex msingle-exit
15955 @opindex mno-single-exit
15956 Force (do not force) generated code to have a single exit point in each
15960 @node MN10300 Options
15961 @subsection MN10300 Options
15962 @cindex MN10300 options
15964 These @option{-m} options are defined for Matsushita MN10300 architectures:
15969 Generate code to avoid bugs in the multiply instructions for the MN10300
15970 processors. This is the default.
15972 @item -mno-mult-bug
15973 @opindex mno-mult-bug
15974 Do not generate code to avoid bugs in the multiply instructions for the
15975 MN10300 processors.
15979 Generate code using features specific to the AM33 processor.
15983 Do not generate code using features specific to the AM33 processor. This
15988 Generate code using features specific to the AM33/2.0 processor.
15992 Generate code using features specific to the AM34 processor.
15994 @item -mtune=@var{cpu-type}
15996 Use the timing characteristics of the indicated CPU type when
15997 scheduling instructions. This does not change the targeted processor
15998 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15999 @samp{am33-2} or @samp{am34}.
16001 @item -mreturn-pointer-on-d0
16002 @opindex mreturn-pointer-on-d0
16003 When generating a function that returns a pointer, return the pointer
16004 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
16005 only in a0, and attempts to call such functions without a prototype
16006 would result in errors. Note that this option is on by default; use
16007 @option{-mno-return-pointer-on-d0} to disable it.
16011 Do not link in the C run-time initialization object file.
16015 Indicate to the linker that it should perform a relaxation optimization pass
16016 to shorten branches, calls and absolute memory addresses. This option only
16017 has an effect when used on the command line for the final link step.
16019 This option makes symbolic debugging impossible.
16023 Allow the compiler to generate @emph{Long Instruction Word}
16024 instructions if the target is the @samp{AM33} or later. This is the
16025 default. This option defines the preprocessor macro @samp{__LIW__}.
16029 Do not allow the compiler to generate @emph{Long Instruction Word}
16030 instructions. This option defines the preprocessor macro
16035 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
16036 instructions if the target is the @samp{AM33} or later. This is the
16037 default. This option defines the preprocessor macro @samp{__SETLB__}.
16041 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
16042 instructions. This option defines the preprocessor macro
16043 @samp{__NO_SETLB__}.
16047 @node PDP-11 Options
16048 @subsection PDP-11 Options
16049 @cindex PDP-11 Options
16051 These options are defined for the PDP-11:
16056 Use hardware FPP floating point. This is the default. (FIS floating
16057 point on the PDP-11/40 is not supported.)
16060 @opindex msoft-float
16061 Do not use hardware floating point.
16065 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
16069 Return floating-point results in memory. This is the default.
16073 Generate code for a PDP-11/40.
16077 Generate code for a PDP-11/45. This is the default.
16081 Generate code for a PDP-11/10.
16083 @item -mbcopy-builtin
16084 @opindex mbcopy-builtin
16085 Use inline @code{movmemhi} patterns for copying memory. This is the
16090 Do not use inline @code{movmemhi} patterns for copying memory.
16096 Use 16-bit @code{int}. This is the default.
16102 Use 32-bit @code{int}.
16105 @itemx -mno-float32
16107 @opindex mno-float32
16108 Use 64-bit @code{float}. This is the default.
16111 @itemx -mno-float64
16113 @opindex mno-float64
16114 Use 32-bit @code{float}.
16118 Use @code{abshi2} pattern. This is the default.
16122 Do not use @code{abshi2} pattern.
16124 @item -mbranch-expensive
16125 @opindex mbranch-expensive
16126 Pretend that branches are expensive. This is for experimenting with
16127 code generation only.
16129 @item -mbranch-cheap
16130 @opindex mbranch-cheap
16131 Do not pretend that branches are expensive. This is the default.
16135 Use Unix assembler syntax. This is the default when configured for
16136 @samp{pdp11-*-bsd}.
16140 Use DEC assembler syntax. This is the default when configured for any
16141 PDP-11 target other than @samp{pdp11-*-bsd}.
16144 @node picoChip Options
16145 @subsection picoChip Options
16146 @cindex picoChip options
16148 These @samp{-m} options are defined for picoChip implementations:
16152 @item -mae=@var{ae_type}
16154 Set the instruction set, register set, and instruction scheduling
16155 parameters for array element type @var{ae_type}. Supported values
16156 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
16158 @option{-mae=ANY} selects a completely generic AE type. Code
16159 generated with this option will run on any of the other AE types. The
16160 code will not be as efficient as it would be if compiled for a specific
16161 AE type, and some types of operation (e.g., multiplication) will not
16162 work properly on all types of AE.
16164 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
16165 for compiled code, and is the default.
16167 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
16168 option may suffer from poor performance of byte (char) manipulation,
16169 since the DSP AE does not provide hardware support for byte load/stores.
16171 @item -msymbol-as-address
16172 Enable the compiler to directly use a symbol name as an address in a
16173 load/store instruction, without first loading it into a
16174 register. Typically, the use of this option will generate larger
16175 programs, which run faster than when the option isn't used. However, the
16176 results vary from program to program, so it is left as a user option,
16177 rather than being permanently enabled.
16179 @item -mno-inefficient-warnings
16180 Disables warnings about the generation of inefficient code. These
16181 warnings can be generated, for example, when compiling code that
16182 performs byte-level memory operations on the MAC AE type. The MAC AE has
16183 no hardware support for byte-level memory operations, so all byte
16184 load/stores must be synthesized from word load/store operations. This is
16185 inefficient and a warning will be generated indicating to the programmer
16186 that they should rewrite the code to avoid byte operations, or to target
16187 an AE type that has the necessary hardware support. This option enables
16188 the warning to be turned off.
16192 @node PowerPC Options
16193 @subsection PowerPC Options
16194 @cindex PowerPC options
16196 These are listed under @xref{RS/6000 and PowerPC Options}.
16199 @subsection RL78 Options
16200 @cindex RL78 Options
16206 Links in additional target libraries to support operation within a
16213 Specifies the type of hardware multiplication support to be used. The
16214 default is @code{none}, which uses software multiplication functions.
16215 The @code{g13} option is for the hardware multiply/divide peripheral
16216 only on the RL78/G13 targets. The @code{rl78} option is for the
16217 standard hardware multiplication defined in the RL78 software manual.
16221 @node RS/6000 and PowerPC Options
16222 @subsection IBM RS/6000 and PowerPC Options
16223 @cindex RS/6000 and PowerPC Options
16224 @cindex IBM RS/6000 and PowerPC Options
16226 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
16233 @itemx -mno-powerpc
16234 @itemx -mpowerpc-gpopt
16235 @itemx -mno-powerpc-gpopt
16236 @itemx -mpowerpc-gfxopt
16237 @itemx -mno-powerpc-gfxopt
16240 @itemx -mno-powerpc64
16244 @itemx -mno-popcntb
16246 @itemx -mno-popcntd
16255 @itemx -mno-hard-dfp
16259 @opindex mno-power2
16261 @opindex mno-powerpc
16262 @opindex mpowerpc-gpopt
16263 @opindex mno-powerpc-gpopt
16264 @opindex mpowerpc-gfxopt
16265 @opindex mno-powerpc-gfxopt
16266 @opindex mpowerpc64
16267 @opindex mno-powerpc64
16271 @opindex mno-popcntb
16273 @opindex mno-popcntd
16279 @opindex mno-mfpgpr
16281 @opindex mno-hard-dfp
16282 GCC supports two related instruction set architectures for the
16283 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
16284 instructions supported by the @samp{rios} chip set used in the original
16285 RS/6000 systems and the @dfn{PowerPC} instruction set is the
16286 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
16287 the IBM 4xx, 6xx, and follow-on microprocessors.
16289 Neither architecture is a subset of the other. However there is a
16290 large common subset of instructions supported by both. An MQ
16291 register is included in processors supporting the POWER architecture.
16293 You use these options to specify which instructions are available on the
16294 processor you are using. The default value of these options is
16295 determined when configuring GCC@. Specifying the
16296 @option{-mcpu=@var{cpu_type}} overrides the specification of these
16297 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
16298 rather than the options listed above.
16300 The @option{-mpower} option allows GCC to generate instructions that
16301 are found only in the POWER architecture and to use the MQ register.
16302 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
16303 to generate instructions that are present in the POWER2 architecture but
16304 not the original POWER architecture.
16306 The @option{-mpowerpc} option allows GCC to generate instructions that
16307 are found only in the 32-bit subset of the PowerPC architecture.
16308 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
16309 GCC to use the optional PowerPC architecture instructions in the
16310 General Purpose group, including floating-point square root. Specifying
16311 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
16312 use the optional PowerPC architecture instructions in the Graphics
16313 group, including floating-point select.
16315 The @option{-mmfcrf} option allows GCC to generate the move from
16316 condition register field instruction implemented on the POWER4
16317 processor and other processors that support the PowerPC V2.01
16319 The @option{-mpopcntb} option allows GCC to generate the popcount and
16320 double-precision FP reciprocal estimate instruction implemented on the
16321 POWER5 processor and other processors that support the PowerPC V2.02
16323 The @option{-mpopcntd} option allows GCC to generate the popcount
16324 instruction implemented on the POWER7 processor and other processors
16325 that support the PowerPC V2.06 architecture.
16326 The @option{-mfprnd} option allows GCC to generate the FP round to
16327 integer instructions implemented on the POWER5+ processor and other
16328 processors that support the PowerPC V2.03 architecture.
16329 The @option{-mcmpb} option allows GCC to generate the compare bytes
16330 instruction implemented on the POWER6 processor and other processors
16331 that support the PowerPC V2.05 architecture.
16332 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
16333 general-purpose register instructions implemented on the POWER6X
16334 processor and other processors that support the extended PowerPC V2.05
16336 The @option{-mhard-dfp} option allows GCC to generate the decimal
16337 floating-point instructions implemented on some POWER processors.
16339 The @option{-mpowerpc64} option allows GCC to generate the additional
16340 64-bit instructions that are found in the full PowerPC64 architecture
16341 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
16342 @option{-mno-powerpc64}.
16344 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
16345 will use only the instructions in the common subset of both
16346 architectures plus some special AIX common-mode calls, and will not use
16347 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
16348 permits GCC to use any instruction from either architecture and to
16349 allow use of the MQ register; specify this for the Motorola MPC601.
16351 @item -mnew-mnemonics
16352 @itemx -mold-mnemonics
16353 @opindex mnew-mnemonics
16354 @opindex mold-mnemonics
16355 Select which mnemonics to use in the generated assembler code. With
16356 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
16357 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
16358 assembler mnemonics defined for the POWER architecture. Instructions
16359 defined in only one architecture have only one mnemonic; GCC uses that
16360 mnemonic irrespective of which of these options is specified.
16362 GCC defaults to the mnemonics appropriate for the architecture in
16363 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
16364 value of these option. Unless you are building a cross-compiler, you
16365 should normally not specify either @option{-mnew-mnemonics} or
16366 @option{-mold-mnemonics}, but should instead accept the default.
16368 @item -mcpu=@var{cpu_type}
16370 Set architecture type, register usage, choice of mnemonics, and
16371 instruction scheduling parameters for machine type @var{cpu_type}.
16372 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
16373 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
16374 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
16375 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
16376 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
16377 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
16378 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
16379 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
16380 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
16381 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
16382 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
16384 @option{-mcpu=common} selects a completely generic processor. Code
16385 generated under this option will run on any POWER or PowerPC processor.
16386 GCC will use only the instructions in the common subset of both
16387 architectures, and will not use the MQ register. GCC assumes a generic
16388 processor model for scheduling purposes.
16390 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
16391 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
16392 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
16393 types, with an appropriate, generic processor model assumed for
16394 scheduling purposes.
16396 The other options specify a specific processor. Code generated under
16397 those options will run best on that processor, and may not run at all on
16400 The @option{-mcpu} options automatically enable or disable the
16403 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
16404 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
16405 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
16406 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
16408 The particular options set for any particular CPU will vary between
16409 compiler versions, depending on what setting seems to produce optimal
16410 code for that CPU; it doesn't necessarily reflect the actual hardware's
16411 capabilities. If you wish to set an individual option to a particular
16412 value, you may specify it after the @option{-mcpu} option, like
16413 @samp{-mcpu=970 -mno-altivec}.
16415 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
16416 not enabled or disabled by the @option{-mcpu} option at present because
16417 AIX does not have full support for these options. You may still
16418 enable or disable them individually if you're sure it'll work in your
16421 @item -mtune=@var{cpu_type}
16423 Set the instruction scheduling parameters for machine type
16424 @var{cpu_type}, but do not set the architecture type, register usage, or
16425 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
16426 values for @var{cpu_type} are used for @option{-mtune} as for
16427 @option{-mcpu}. If both are specified, the code generated will use the
16428 architecture, registers, and mnemonics set by @option{-mcpu}, but the
16429 scheduling parameters set by @option{-mtune}.
16431 @item -mcmodel=small
16432 @opindex mcmodel=small
16433 Generate PowerPC64 code for the small model: The TOC is limited to
16436 @item -mcmodel=medium
16437 @opindex mcmodel=medium
16438 Generate PowerPC64 code for the medium model: The TOC and other static
16439 data may be up to a total of 4G in size.
16441 @item -mcmodel=large
16442 @opindex mcmodel=large
16443 Generate PowerPC64 code for the large model: The TOC may be up to 4G
16444 in size. Other data and code is only limited by the 64-bit address
16448 @itemx -mno-altivec
16450 @opindex mno-altivec
16451 Generate code that uses (does not use) AltiVec instructions, and also
16452 enable the use of built-in functions that allow more direct access to
16453 the AltiVec instruction set. You may also need to set
16454 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
16460 @opindex mno-vrsave
16461 Generate VRSAVE instructions when generating AltiVec code.
16463 @item -mgen-cell-microcode
16464 @opindex mgen-cell-microcode
16465 Generate Cell microcode instructions
16467 @item -mwarn-cell-microcode
16468 @opindex mwarn-cell-microcode
16469 Warning when a Cell microcode instruction is going to emitted. An example
16470 of a Cell microcode instruction is a variable shift.
16473 @opindex msecure-plt
16474 Generate code that allows ld and ld.so to build executables and shared
16475 libraries with non-exec .plt and .got sections. This is a PowerPC
16476 32-bit SYSV ABI option.
16480 Generate code that uses a BSS .plt section that ld.so fills in, and
16481 requires .plt and .got sections that are both writable and executable.
16482 This is a PowerPC 32-bit SYSV ABI option.
16488 This switch enables or disables the generation of ISEL instructions.
16490 @item -misel=@var{yes/no}
16491 This switch has been deprecated. Use @option{-misel} and
16492 @option{-mno-isel} instead.
16498 This switch enables or disables the generation of SPE simd
16504 @opindex mno-paired
16505 This switch enables or disables the generation of PAIRED simd
16508 @item -mspe=@var{yes/no}
16509 This option has been deprecated. Use @option{-mspe} and
16510 @option{-mno-spe} instead.
16516 Generate code that uses (does not use) vector/scalar (VSX)
16517 instructions, and also enable the use of built-in functions that allow
16518 more direct access to the VSX instruction set.
16520 @item -mfloat-gprs=@var{yes/single/double/no}
16521 @itemx -mfloat-gprs
16522 @opindex mfloat-gprs
16523 This switch enables or disables the generation of floating-point
16524 operations on the general-purpose registers for architectures that
16527 The argument @var{yes} or @var{single} enables the use of
16528 single-precision floating-point operations.
16530 The argument @var{double} enables the use of single and
16531 double-precision floating-point operations.
16533 The argument @var{no} disables floating-point operations on the
16534 general-purpose registers.
16536 This option is currently only available on the MPC854x.
16542 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
16543 targets (including GNU/Linux). The 32-bit environment sets int, long
16544 and pointer to 32 bits and generates code that runs on any PowerPC
16545 variant. The 64-bit environment sets int to 32 bits and long and
16546 pointer to 64 bits, and generates code for PowerPC64, as for
16547 @option{-mpowerpc64}.
16550 @itemx -mno-fp-in-toc
16551 @itemx -mno-sum-in-toc
16552 @itemx -mminimal-toc
16554 @opindex mno-fp-in-toc
16555 @opindex mno-sum-in-toc
16556 @opindex mminimal-toc
16557 Modify generation of the TOC (Table Of Contents), which is created for
16558 every executable file. The @option{-mfull-toc} option is selected by
16559 default. In that case, GCC will allocate at least one TOC entry for
16560 each unique non-automatic variable reference in your program. GCC
16561 will also place floating-point constants in the TOC@. However, only
16562 16,384 entries are available in the TOC@.
16564 If you receive a linker error message that saying you have overflowed
16565 the available TOC space, you can reduce the amount of TOC space used
16566 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
16567 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
16568 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
16569 generate code to calculate the sum of an address and a constant at
16570 run time instead of putting that sum into the TOC@. You may specify one
16571 or both of these options. Each causes GCC to produce very slightly
16572 slower and larger code at the expense of conserving TOC space.
16574 If you still run out of space in the TOC even when you specify both of
16575 these options, specify @option{-mminimal-toc} instead. This option causes
16576 GCC to make only one TOC entry for every file. When you specify this
16577 option, GCC will produce code that is slower and larger but which
16578 uses extremely little TOC space. You may wish to use this option
16579 only on files that contain less frequently executed code.
16585 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
16586 @code{long} type, and the infrastructure needed to support them.
16587 Specifying @option{-maix64} implies @option{-mpowerpc64} and
16588 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
16589 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
16592 @itemx -mno-xl-compat
16593 @opindex mxl-compat
16594 @opindex mno-xl-compat
16595 Produce code that conforms more closely to IBM XL compiler semantics
16596 when using AIX-compatible ABI@. Pass floating-point arguments to
16597 prototyped functions beyond the register save area (RSA) on the stack
16598 in addition to argument FPRs. Do not assume that most significant
16599 double in 128-bit long double value is properly rounded when comparing
16600 values and converting to double. Use XL symbol names for long double
16603 The AIX calling convention was extended but not initially documented to
16604 handle an obscure K&R C case of calling a function that takes the
16605 address of its arguments with fewer arguments than declared. IBM XL
16606 compilers access floating-point arguments that do not fit in the
16607 RSA from the stack when a subroutine is compiled without
16608 optimization. Because always storing floating-point arguments on the
16609 stack is inefficient and rarely needed, this option is not enabled by
16610 default and only is necessary when calling subroutines compiled by IBM
16611 XL compilers without optimization.
16615 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
16616 application written to use message passing with special startup code to
16617 enable the application to run. The system must have PE installed in the
16618 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
16619 must be overridden with the @option{-specs=} option to specify the
16620 appropriate directory location. The Parallel Environment does not
16621 support threads, so the @option{-mpe} option and the @option{-pthread}
16622 option are incompatible.
16624 @item -malign-natural
16625 @itemx -malign-power
16626 @opindex malign-natural
16627 @opindex malign-power
16628 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
16629 @option{-malign-natural} overrides the ABI-defined alignment of larger
16630 types, such as floating-point doubles, on their natural size-based boundary.
16631 The option @option{-malign-power} instructs GCC to follow the ABI-specified
16632 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
16634 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
16638 @itemx -mhard-float
16639 @opindex msoft-float
16640 @opindex mhard-float
16641 Generate code that does not use (uses) the floating-point register set.
16642 Software floating-point emulation is provided if you use the
16643 @option{-msoft-float} option, and pass the option to GCC when linking.
16645 @item -msingle-float
16646 @itemx -mdouble-float
16647 @opindex msingle-float
16648 @opindex mdouble-float
16649 Generate code for single- or double-precision floating-point operations.
16650 @option{-mdouble-float} implies @option{-msingle-float}.
16653 @opindex msimple-fpu
16654 Do not generate sqrt and div instructions for hardware floating-point unit.
16658 Specify type of floating-point unit. Valid values are @var{sp_lite}
16659 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
16660 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
16661 and @var{dp_full} (equivalent to -mdouble-float).
16664 @opindex mxilinx-fpu
16665 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
16668 @itemx -mno-multiple
16670 @opindex mno-multiple
16671 Generate code that uses (does not use) the load multiple word
16672 instructions and the store multiple word instructions. These
16673 instructions are generated by default on POWER systems, and not
16674 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
16675 PowerPC systems, since those instructions do not work when the
16676 processor is in little-endian mode. The exceptions are PPC740 and
16677 PPC750 which permit these instructions in little-endian mode.
16682 @opindex mno-string
16683 Generate code that uses (does not use) the load string instructions
16684 and the store string word instructions to save multiple registers and
16685 do small block moves. These instructions are generated by default on
16686 POWER systems, and not generated on PowerPC systems. Do not use
16687 @option{-mstring} on little-endian PowerPC systems, since those
16688 instructions do not work when the processor is in little-endian mode.
16689 The exceptions are PPC740 and PPC750 which permit these instructions
16690 in little-endian mode.
16695 @opindex mno-update
16696 Generate code that uses (does not use) the load or store instructions
16697 that update the base register to the address of the calculated memory
16698 location. These instructions are generated by default. If you use
16699 @option{-mno-update}, there is a small window between the time that the
16700 stack pointer is updated and the address of the previous frame is
16701 stored, which means code that walks the stack frame across interrupts or
16702 signals may get corrupted data.
16704 @item -mavoid-indexed-addresses
16705 @itemx -mno-avoid-indexed-addresses
16706 @opindex mavoid-indexed-addresses
16707 @opindex mno-avoid-indexed-addresses
16708 Generate code that tries to avoid (not avoid) the use of indexed load
16709 or store instructions. These instructions can incur a performance
16710 penalty on Power6 processors in certain situations, such as when
16711 stepping through large arrays that cross a 16M boundary. This option
16712 is enabled by default when targetting Power6 and disabled otherwise.
16715 @itemx -mno-fused-madd
16716 @opindex mfused-madd
16717 @opindex mno-fused-madd
16718 Generate code that uses (does not use) the floating-point multiply and
16719 accumulate instructions. These instructions are generated by default
16720 if hardware floating point is used. The machine-dependent
16721 @option{-mfused-madd} option is now mapped to the machine-independent
16722 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16723 mapped to @option{-ffp-contract=off}.
16729 Generate code that uses (does not use) the half-word multiply and
16730 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
16731 These instructions are generated by default when targetting those
16738 Generate code that uses (does not use) the string-search @samp{dlmzb}
16739 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
16740 generated by default when targetting those processors.
16742 @item -mno-bit-align
16744 @opindex mno-bit-align
16745 @opindex mbit-align
16746 On System V.4 and embedded PowerPC systems do not (do) force structures
16747 and unions that contain bit-fields to be aligned to the base type of the
16750 For example, by default a structure containing nothing but 8
16751 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
16752 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
16753 the structure is aligned to a 1-byte boundary and is 1 byte in
16756 @item -mno-strict-align
16757 @itemx -mstrict-align
16758 @opindex mno-strict-align
16759 @opindex mstrict-align
16760 On System V.4 and embedded PowerPC systems do not (do) assume that
16761 unaligned memory references will be handled by the system.
16763 @item -mrelocatable
16764 @itemx -mno-relocatable
16765 @opindex mrelocatable
16766 @opindex mno-relocatable
16767 Generate code that allows (does not allow) a static executable to be
16768 relocated to a different address at run time. A simple embedded
16769 PowerPC system loader should relocate the entire contents of
16770 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16771 a table of 32-bit addresses generated by this option. For this to
16772 work, all objects linked together must be compiled with
16773 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16774 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
16776 @item -mrelocatable-lib
16777 @itemx -mno-relocatable-lib
16778 @opindex mrelocatable-lib
16779 @opindex mno-relocatable-lib
16780 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16781 @code{.fixup} section to allow static executables to be relocated at
16782 run time, but @option{-mrelocatable-lib} does not use the smaller stack
16783 alignment of @option{-mrelocatable}. Objects compiled with
16784 @option{-mrelocatable-lib} may be linked with objects compiled with
16785 any combination of the @option{-mrelocatable} options.
16791 On System V.4 and embedded PowerPC systems do not (do) assume that
16792 register 2 contains a pointer to a global area pointing to the addresses
16793 used in the program.
16796 @itemx -mlittle-endian
16798 @opindex mlittle-endian
16799 On System V.4 and embedded PowerPC systems compile code for the
16800 processor in little-endian mode. The @option{-mlittle-endian} option is
16801 the same as @option{-mlittle}.
16804 @itemx -mbig-endian
16806 @opindex mbig-endian
16807 On System V.4 and embedded PowerPC systems compile code for the
16808 processor in big-endian mode. The @option{-mbig-endian} option is
16809 the same as @option{-mbig}.
16811 @item -mdynamic-no-pic
16812 @opindex mdynamic-no-pic
16813 On Darwin and Mac OS X systems, compile code so that it is not
16814 relocatable, but that its external references are relocatable. The
16815 resulting code is suitable for applications, but not shared
16818 @item -msingle-pic-base
16819 @opindex msingle-pic-base
16820 Treat the register used for PIC addressing as read-only, rather than
16821 loading it in the prologue for each function. The runtime system is
16822 responsible for initializing this register with an appropriate value
16823 before execution begins.
16825 @item -mprioritize-restricted-insns=@var{priority}
16826 @opindex mprioritize-restricted-insns
16827 This option controls the priority that is assigned to
16828 dispatch-slot restricted instructions during the second scheduling
16829 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16830 @var{no/highest/second-highest} priority to dispatch slot restricted
16833 @item -msched-costly-dep=@var{dependence_type}
16834 @opindex msched-costly-dep
16835 This option controls which dependences are considered costly
16836 by the target during instruction scheduling. The argument
16837 @var{dependence_type} takes one of the following values:
16838 @var{no}: no dependence is costly,
16839 @var{all}: all dependences are costly,
16840 @var{true_store_to_load}: a true dependence from store to load is costly,
16841 @var{store_to_load}: any dependence from store to load is costly,
16842 @var{number}: any dependence for which latency >= @var{number} is costly.
16844 @item -minsert-sched-nops=@var{scheme}
16845 @opindex minsert-sched-nops
16846 This option controls which nop insertion scheme will be used during
16847 the second scheduling pass. The argument @var{scheme} takes one of the
16849 @var{no}: Don't insert nops.
16850 @var{pad}: Pad with nops any dispatch group that has vacant issue slots,
16851 according to the scheduler's grouping.
16852 @var{regroup_exact}: Insert nops to force costly dependent insns into
16853 separate groups. Insert exactly as many nops as needed to force an insn
16854 to a new group, according to the estimated processor grouping.
16855 @var{number}: Insert nops to force costly dependent insns into
16856 separate groups. Insert @var{number} nops to force an insn to a new group.
16859 @opindex mcall-sysv
16860 On System V.4 and embedded PowerPC systems compile code using calling
16861 conventions that adheres to the March 1995 draft of the System V
16862 Application Binary Interface, PowerPC processor supplement. This is the
16863 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16865 @item -mcall-sysv-eabi
16867 @opindex mcall-sysv-eabi
16868 @opindex mcall-eabi
16869 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16871 @item -mcall-sysv-noeabi
16872 @opindex mcall-sysv-noeabi
16873 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16875 @item -mcall-aixdesc
16877 On System V.4 and embedded PowerPC systems compile code for the AIX
16881 @opindex mcall-linux
16882 On System V.4 and embedded PowerPC systems compile code for the
16883 Linux-based GNU system.
16885 @item -mcall-freebsd
16886 @opindex mcall-freebsd
16887 On System V.4 and embedded PowerPC systems compile code for the
16888 FreeBSD operating system.
16890 @item -mcall-netbsd
16891 @opindex mcall-netbsd
16892 On System V.4 and embedded PowerPC systems compile code for the
16893 NetBSD operating system.
16895 @item -mcall-openbsd
16896 @opindex mcall-netbsd
16897 On System V.4 and embedded PowerPC systems compile code for the
16898 OpenBSD operating system.
16900 @item -maix-struct-return
16901 @opindex maix-struct-return
16902 Return all structures in memory (as specified by the AIX ABI)@.
16904 @item -msvr4-struct-return
16905 @opindex msvr4-struct-return
16906 Return structures smaller than 8 bytes in registers (as specified by the
16909 @item -mabi=@var{abi-type}
16911 Extend the current ABI with a particular extension, or remove such extension.
16912 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16913 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16917 Extend the current ABI with SPE ABI extensions. This does not change
16918 the default ABI, instead it adds the SPE ABI extensions to the current
16922 @opindex mabi=no-spe
16923 Disable Booke SPE ABI extensions for the current ABI@.
16925 @item -mabi=ibmlongdouble
16926 @opindex mabi=ibmlongdouble
16927 Change the current ABI to use IBM extended-precision long double.
16928 This is a PowerPC 32-bit SYSV ABI option.
16930 @item -mabi=ieeelongdouble
16931 @opindex mabi=ieeelongdouble
16932 Change the current ABI to use IEEE extended-precision long double.
16933 This is a PowerPC 32-bit Linux ABI option.
16936 @itemx -mno-prototype
16937 @opindex mprototype
16938 @opindex mno-prototype
16939 On System V.4 and embedded PowerPC systems assume that all calls to
16940 variable argument functions are properly prototyped. Otherwise, the
16941 compiler must insert an instruction before every non prototyped call to
16942 set or clear bit 6 of the condition code register (@var{CR}) to
16943 indicate whether floating-point values were passed in the floating-point
16944 registers in case the function takes variable arguments. With
16945 @option{-mprototype}, only calls to prototyped variable argument functions
16946 will set or clear the bit.
16950 On embedded PowerPC systems, assume that the startup module is called
16951 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16952 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16957 On embedded PowerPC systems, assume that the startup module is called
16958 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16963 On embedded PowerPC systems, assume that the startup module is called
16964 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16967 @item -myellowknife
16968 @opindex myellowknife
16969 On embedded PowerPC systems, assume that the startup module is called
16970 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16975 On System V.4 and embedded PowerPC systems, specify that you are
16976 compiling for a VxWorks system.
16980 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16981 header to indicate that @samp{eabi} extended relocations are used.
16987 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16988 Embedded Applications Binary Interface (eabi) which is a set of
16989 modifications to the System V.4 specifications. Selecting @option{-meabi}
16990 means that the stack is aligned to an 8-byte boundary, a function
16991 @code{__eabi} is called to from @code{main} to set up the eabi
16992 environment, and the @option{-msdata} option can use both @code{r2} and
16993 @code{r13} to point to two separate small data areas. Selecting
16994 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
16995 do not call an initialization function from @code{main}, and the
16996 @option{-msdata} option will only use @code{r13} to point to a single
16997 small data area. The @option{-meabi} option is on by default if you
16998 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
17001 @opindex msdata=eabi
17002 On System V.4 and embedded PowerPC systems, put small initialized
17003 @code{const} global and static data in the @samp{.sdata2} section, which
17004 is pointed to by register @code{r2}. Put small initialized
17005 non-@code{const} global and static data in the @samp{.sdata} section,
17006 which is pointed to by register @code{r13}. Put small uninitialized
17007 global and static data in the @samp{.sbss} section, which is adjacent to
17008 the @samp{.sdata} section. The @option{-msdata=eabi} option is
17009 incompatible with the @option{-mrelocatable} option. The
17010 @option{-msdata=eabi} option also sets the @option{-memb} option.
17013 @opindex msdata=sysv
17014 On System V.4 and embedded PowerPC systems, put small global and static
17015 data in the @samp{.sdata} section, which is pointed to by register
17016 @code{r13}. Put small uninitialized global and static data in the
17017 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
17018 The @option{-msdata=sysv} option is incompatible with the
17019 @option{-mrelocatable} option.
17021 @item -msdata=default
17023 @opindex msdata=default
17025 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
17026 compile code the same as @option{-msdata=eabi}, otherwise compile code the
17027 same as @option{-msdata=sysv}.
17030 @opindex msdata=data
17031 On System V.4 and embedded PowerPC systems, put small global
17032 data in the @samp{.sdata} section. Put small uninitialized global
17033 data in the @samp{.sbss} section. Do not use register @code{r13}
17034 to address small data however. This is the default behavior unless
17035 other @option{-msdata} options are used.
17039 @opindex msdata=none
17041 On embedded PowerPC systems, put all initialized global and static data
17042 in the @samp{.data} section, and all uninitialized data in the
17043 @samp{.bss} section.
17045 @item -mblock-move-inline-limit=@var{num}
17046 @opindex mblock-move-inline-limit
17047 Inline all block moves (such as calls to @code{memcpy} or structure
17048 copies) less than or equal to @var{num} bytes. The minimum value for
17049 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
17050 targets. The default value is target-specific.
17054 @cindex smaller data references (PowerPC)
17055 @cindex .sdata/.sdata2 references (PowerPC)
17056 On embedded PowerPC systems, put global and static items less than or
17057 equal to @var{num} bytes into the small data or bss sections instead of
17058 the normal data or bss section. By default, @var{num} is 8. The
17059 @option{-G @var{num}} switch is also passed to the linker.
17060 All modules should be compiled with the same @option{-G @var{num}} value.
17063 @itemx -mno-regnames
17065 @opindex mno-regnames
17066 On System V.4 and embedded PowerPC systems do (do not) emit register
17067 names in the assembly language output using symbolic forms.
17070 @itemx -mno-longcall
17072 @opindex mno-longcall
17073 By default assume that all calls are far away so that a longer more
17074 expensive calling sequence is required. This is required for calls
17075 further than 32 megabytes (33,554,432 bytes) from the current location.
17076 A short call will be generated if the compiler knows
17077 the call cannot be that far away. This setting can be overridden by
17078 the @code{shortcall} function attribute, or by @code{#pragma
17081 Some linkers are capable of detecting out-of-range calls and generating
17082 glue code on the fly. On these systems, long calls are unnecessary and
17083 generate slower code. As of this writing, the AIX linker can do this,
17084 as can the GNU linker for PowerPC/64. It is planned to add this feature
17085 to the GNU linker for 32-bit PowerPC systems as well.
17087 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
17088 callee, L42'', plus a ``branch island'' (glue code). The two target
17089 addresses represent the callee and the ``branch island''. The
17090 Darwin/PPC linker will prefer the first address and generate a ``bl
17091 callee'' if the PPC ``bl'' instruction will reach the callee directly;
17092 otherwise, the linker will generate ``bl L42'' to call the ``branch
17093 island''. The ``branch island'' is appended to the body of the
17094 calling function; it computes the full 32-bit address of the callee
17097 On Mach-O (Darwin) systems, this option directs the compiler emit to
17098 the glue for every direct call, and the Darwin linker decides whether
17099 to use or discard it.
17101 In the future, we may cause GCC to ignore all longcall specifications
17102 when the linker is known to generate glue.
17104 @item -mtls-markers
17105 @itemx -mno-tls-markers
17106 @opindex mtls-markers
17107 @opindex mno-tls-markers
17108 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
17109 specifying the function argument. The relocation allows ld to
17110 reliably associate function call with argument setup instructions for
17111 TLS optimization, which in turn allows gcc to better schedule the
17116 Adds support for multithreading with the @dfn{pthreads} library.
17117 This option sets flags for both the preprocessor and linker.
17122 This option will enable GCC to use the reciprocal estimate and
17123 reciprocal square root estimate instructions with additional
17124 Newton-Raphson steps to increase precision instead of doing a divide or
17125 square root and divide for floating-point arguments. You should use
17126 the @option{-ffast-math} option when using @option{-mrecip} (or at
17127 least @option{-funsafe-math-optimizations},
17128 @option{-finite-math-only}, @option{-freciprocal-math} and
17129 @option{-fno-trapping-math}). Note that while the throughput of the
17130 sequence is generally higher than the throughput of the non-reciprocal
17131 instruction, the precision of the sequence can be decreased by up to 2
17132 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
17135 @item -mrecip=@var{opt}
17136 @opindex mrecip=opt
17137 This option allows to control which reciprocal estimate instructions
17138 may be used. @var{opt} is a comma separated list of options, which may
17139 be preceded by a @code{!} to invert the option:
17140 @code{all}: enable all estimate instructions,
17141 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
17142 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
17143 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
17144 @code{divf}: enable the single-precision reciprocal approximation instructions;
17145 @code{divd}: enable the double-precision reciprocal approximation instructions;
17146 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
17147 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
17148 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
17150 So for example, @option{-mrecip=all,!rsqrtd} would enable the
17151 all of the reciprocal estimate instructions, except for the
17152 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
17153 which handle the double-precision reciprocal square root calculations.
17155 @item -mrecip-precision
17156 @itemx -mno-recip-precision
17157 @opindex mrecip-precision
17158 Assume (do not assume) that the reciprocal estimate instructions
17159 provide higher-precision estimates than is mandated by the PowerPC
17160 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
17161 automatically selects @option{-mrecip-precision}. The double-precision
17162 square root estimate instructions are not generated by
17163 default on low-precision machines, since they do not provide an
17164 estimate that converges after three steps.
17166 @item -mveclibabi=@var{type}
17167 @opindex mveclibabi
17168 Specifies the ABI type to use for vectorizing intrinsics using an
17169 external library. The only type supported at present is @code{mass},
17170 which specifies to use IBM's Mathematical Acceleration Subsystem
17171 (MASS) libraries for vectorizing intrinsics using external libraries.
17172 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
17173 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
17174 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
17175 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
17176 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
17177 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
17178 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
17179 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
17180 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
17181 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
17182 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
17183 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
17184 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
17185 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
17186 for power7. Both @option{-ftree-vectorize} and
17187 @option{-funsafe-math-optimizations} have to be enabled. The MASS
17188 libraries will have to be specified at link time.
17193 Generate (do not generate) the @code{friz} instruction when the
17194 @option{-funsafe-math-optimizations} option is used to optimize
17195 rounding of floating-point values to 64-bit integer and back to floating
17196 point. The @code{friz} instruction does not return the same value if
17197 the floating-point number is too large to fit in an integer.
17199 @item -mpointers-to-nested-functions
17200 @itemx -mno-pointers-to-nested-functions
17201 @opindex mpointers-to-nested-functions
17202 Generate (do not generate) code to load up the static chain register
17203 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
17204 systems where a function pointer points to a 3-word descriptor giving
17205 the function address, TOC value to be loaded in register @var{r2}, and
17206 static chain value to be loaded in register @var{r11}. The
17207 @option{-mpointers-to-nested-functions} is on by default. You will
17208 not be able to call through pointers to nested functions or pointers
17209 to functions compiled in other languages that use the static chain if
17210 you use the @option{-mno-pointers-to-nested-functions}.
17212 @item -msave-toc-indirect
17213 @itemx -mno-save-toc-indirect
17214 @opindex msave-toc-indirect
17215 Generate (do not generate) code to save the TOC value in the reserved
17216 stack location in the function prologue if the function calls through
17217 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
17218 saved in the prologue, it is saved just before the call through the
17219 pointer. The @option{-mno-save-toc-indirect} option is the default.
17223 @subsection RX Options
17226 These command-line options are defined for RX targets:
17229 @item -m64bit-doubles
17230 @itemx -m32bit-doubles
17231 @opindex m64bit-doubles
17232 @opindex m32bit-doubles
17233 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
17234 or 32 bits (@option{-m32bit-doubles}) in size. The default is
17235 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
17236 works on 32-bit values, which is why the default is
17237 @option{-m32bit-doubles}.
17243 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
17244 floating-point hardware. The default is enabled for the @var{RX600}
17245 series and disabled for the @var{RX200} series.
17247 Floating-point instructions will only be generated for 32-bit floating-point
17248 values however, so if the @option{-m64bit-doubles} option is in
17249 use then the FPU hardware will not be used for doubles.
17251 @emph{Note} If the @option{-fpu} option is enabled then
17252 @option{-funsafe-math-optimizations} is also enabled automatically.
17253 This is because the RX FPU instructions are themselves unsafe.
17255 @item -mcpu=@var{name}
17257 Selects the type of RX CPU to be targeted. Currently three types are
17258 supported, the generic @var{RX600} and @var{RX200} series hardware and
17259 the specific @var{RX610} CPU. The default is @var{RX600}.
17261 The only difference between @var{RX600} and @var{RX610} is that the
17262 @var{RX610} does not support the @code{MVTIPL} instruction.
17264 The @var{RX200} series does not have a hardware floating-point unit
17265 and so @option{-nofpu} is enabled by default when this type is
17268 @item -mbig-endian-data
17269 @itemx -mlittle-endian-data
17270 @opindex mbig-endian-data
17271 @opindex mlittle-endian-data
17272 Store data (but not code) in the big-endian format. The default is
17273 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
17276 @item -msmall-data-limit=@var{N}
17277 @opindex msmall-data-limit
17278 Specifies the maximum size in bytes of global and static variables
17279 which can be placed into the small data area. Using the small data
17280 area can lead to smaller and faster code, but the size of area is
17281 limited and it is up to the programmer to ensure that the area does
17282 not overflow. Also when the small data area is used one of the RX's
17283 registers (usually @code{r13}) is reserved for use pointing to this
17284 area, so it is no longer available for use by the compiler. This
17285 could result in slower and/or larger code if variables which once
17286 could have been held in the reserved register are now pushed onto the
17289 Note, common variables (variables that have not been initialised) and
17290 constants are not placed into the small data area as they are assigned
17291 to other sections in the output executable.
17293 The default value is zero, which disables this feature. Note, this
17294 feature is not enabled by default with higher optimization levels
17295 (@option{-O2} etc) because of the potentially detrimental effects of
17296 reserving a register. It is up to the programmer to experiment and
17297 discover whether this feature is of benefit to their program. See the
17298 description of the @option{-mpid} option for a description of how the
17299 actual register to hold the small data area pointer is chosen.
17305 Use the simulator runtime. The default is to use the libgloss board
17308 @item -mas100-syntax
17309 @itemx -mno-as100-syntax
17310 @opindex mas100-syntax
17311 @opindex mno-as100-syntax
17312 When generating assembler output use a syntax that is compatible with
17313 Renesas's AS100 assembler. This syntax can also be handled by the GAS
17314 assembler but it has some restrictions so generating it is not the
17317 @item -mmax-constant-size=@var{N}
17318 @opindex mmax-constant-size
17319 Specifies the maximum size, in bytes, of a constant that can be used as
17320 an operand in a RX instruction. Although the RX instruction set does
17321 allow constants of up to 4 bytes in length to be used in instructions,
17322 a longer value equates to a longer instruction. Thus in some
17323 circumstances it can be beneficial to restrict the size of constants
17324 that are used in instructions. Constants that are too big are instead
17325 placed into a constant pool and referenced via register indirection.
17327 The value @var{N} can be between 0 and 4. A value of 0 (the default)
17328 or 4 means that constants of any size are allowed.
17332 Enable linker relaxation. Linker relaxation is a process whereby the
17333 linker will attempt to reduce the size of a program by finding shorter
17334 versions of various instructions. Disabled by default.
17336 @item -mint-register=@var{N}
17337 @opindex mint-register
17338 Specify the number of registers to reserve for fast interrupt handler
17339 functions. The value @var{N} can be between 0 and 4. A value of 1
17340 means that register @code{r13} will be reserved for the exclusive use
17341 of fast interrupt handlers. A value of 2 reserves @code{r13} and
17342 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
17343 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
17344 A value of 0, the default, does not reserve any registers.
17346 @item -msave-acc-in-interrupts
17347 @opindex msave-acc-in-interrupts
17348 Specifies that interrupt handler functions should preserve the
17349 accumulator register. This is only necessary if normal code might use
17350 the accumulator register, for example because it performs 64-bit
17351 multiplications. The default is to ignore the accumulator as this
17352 makes the interrupt handlers faster.
17358 Enables the generation of position independent data. When enabled any
17359 access to constant data will done via an offset from a base address
17360 held in a register. This allows the location of constant data to be
17361 determined at run time without requiring the executable to be
17362 relocated, which is a benefit to embedded applications with tight
17363 memory constraints. Data that can be modified is not affected by this
17366 Note, using this feature reserves a register, usually @code{r13}, for
17367 the constant data base address. This can result in slower and/or
17368 larger code, especially in complicated functions.
17370 The actual register chosen to hold the constant data base address
17371 depends upon whether the @option{-msmall-data-limit} and/or the
17372 @option{-mint-register} command-line options are enabled. Starting
17373 with register @code{r13} and proceeding downwards, registers are
17374 allocated first to satisfy the requirements of @option{-mint-register},
17375 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
17376 is possible for the small data area register to be @code{r8} if both
17377 @option{-mint-register=4} and @option{-mpid} are specified on the
17380 By default this feature is not enabled. The default can be restored
17381 via the @option{-mno-pid} command-line option.
17385 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
17386 has special significance to the RX port when used with the
17387 @code{interrupt} function attribute. This attribute indicates a
17388 function intended to process fast interrupts. GCC will will ensure
17389 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
17390 and/or @code{r13} and only provided that the normal use of the
17391 corresponding registers have been restricted via the
17392 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
17395 @node S/390 and zSeries Options
17396 @subsection S/390 and zSeries Options
17397 @cindex S/390 and zSeries Options
17399 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
17403 @itemx -msoft-float
17404 @opindex mhard-float
17405 @opindex msoft-float
17406 Use (do not use) the hardware floating-point instructions and registers
17407 for floating-point operations. When @option{-msoft-float} is specified,
17408 functions in @file{libgcc.a} will be used to perform floating-point
17409 operations. When @option{-mhard-float} is specified, the compiler
17410 generates IEEE floating-point instructions. This is the default.
17413 @itemx -mno-hard-dfp
17415 @opindex mno-hard-dfp
17416 Use (do not use) the hardware decimal-floating-point instructions for
17417 decimal-floating-point operations. When @option{-mno-hard-dfp} is
17418 specified, functions in @file{libgcc.a} will be used to perform
17419 decimal-floating-point operations. When @option{-mhard-dfp} is
17420 specified, the compiler generates decimal-floating-point hardware
17421 instructions. This is the default for @option{-march=z9-ec} or higher.
17423 @item -mlong-double-64
17424 @itemx -mlong-double-128
17425 @opindex mlong-double-64
17426 @opindex mlong-double-128
17427 These switches control the size of @code{long double} type. A size
17428 of 64 bits makes the @code{long double} type equivalent to the @code{double}
17429 type. This is the default.
17432 @itemx -mno-backchain
17433 @opindex mbackchain
17434 @opindex mno-backchain
17435 Store (do not store) the address of the caller's frame as backchain pointer
17436 into the callee's stack frame.
17437 A backchain may be needed to allow debugging using tools that do not understand
17438 DWARF-2 call frame information.
17439 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
17440 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
17441 the backchain is placed into the topmost word of the 96/160 byte register
17444 In general, code compiled with @option{-mbackchain} is call-compatible with
17445 code compiled with @option{-mmo-backchain}; however, use of the backchain
17446 for debugging purposes usually requires that the whole binary is built with
17447 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
17448 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17449 to build a linux kernel use @option{-msoft-float}.
17451 The default is to not maintain the backchain.
17453 @item -mpacked-stack
17454 @itemx -mno-packed-stack
17455 @opindex mpacked-stack
17456 @opindex mno-packed-stack
17457 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
17458 specified, the compiler uses the all fields of the 96/160 byte register save
17459 area only for their default purpose; unused fields still take up stack space.
17460 When @option{-mpacked-stack} is specified, register save slots are densely
17461 packed at the top of the register save area; unused space is reused for other
17462 purposes, allowing for more efficient use of the available stack space.
17463 However, when @option{-mbackchain} is also in effect, the topmost word of
17464 the save area is always used to store the backchain, and the return address
17465 register is always saved two words below the backchain.
17467 As long as the stack frame backchain is not used, code generated with
17468 @option{-mpacked-stack} is call-compatible with code generated with
17469 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
17470 S/390 or zSeries generated code that uses the stack frame backchain at run
17471 time, not just for debugging purposes. Such code is not call-compatible
17472 with code compiled with @option{-mpacked-stack}. Also, note that the
17473 combination of @option{-mbackchain},
17474 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17475 to build a linux kernel use @option{-msoft-float}.
17477 The default is to not use the packed stack layout.
17480 @itemx -mno-small-exec
17481 @opindex msmall-exec
17482 @opindex mno-small-exec
17483 Generate (or do not generate) code using the @code{bras} instruction
17484 to do subroutine calls.
17485 This only works reliably if the total executable size does not
17486 exceed 64k. The default is to use the @code{basr} instruction instead,
17487 which does not have this limitation.
17493 When @option{-m31} is specified, generate code compliant to the
17494 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
17495 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
17496 particular to generate 64-bit instructions. For the @samp{s390}
17497 targets, the default is @option{-m31}, while the @samp{s390x}
17498 targets default to @option{-m64}.
17504 When @option{-mzarch} is specified, generate code using the
17505 instructions available on z/Architecture.
17506 When @option{-mesa} is specified, generate code using the
17507 instructions available on ESA/390. Note that @option{-mesa} is
17508 not possible with @option{-m64}.
17509 When generating code compliant to the GNU/Linux for S/390 ABI,
17510 the default is @option{-mesa}. When generating code compliant
17511 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
17517 Generate (or do not generate) code using the @code{mvcle} instruction
17518 to perform block moves. When @option{-mno-mvcle} is specified,
17519 use a @code{mvc} loop instead. This is the default unless optimizing for
17526 Print (or do not print) additional debug information when compiling.
17527 The default is to not print debug information.
17529 @item -march=@var{cpu-type}
17531 Generate code that will run on @var{cpu-type}, which is the name of a system
17532 representing a certain processor type. Possible values for
17533 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
17534 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
17535 When generating code using the instructions available on z/Architecture,
17536 the default is @option{-march=z900}. Otherwise, the default is
17537 @option{-march=g5}.
17539 @item -mtune=@var{cpu-type}
17541 Tune to @var{cpu-type} everything applicable about the generated code,
17542 except for the ABI and the set of available instructions.
17543 The list of @var{cpu-type} values is the same as for @option{-march}.
17544 The default is the value used for @option{-march}.
17547 @itemx -mno-tpf-trace
17548 @opindex mtpf-trace
17549 @opindex mno-tpf-trace
17550 Generate code that adds (does not add) in TPF OS specific branches to trace
17551 routines in the operating system. This option is off by default, even
17552 when compiling for the TPF OS@.
17555 @itemx -mno-fused-madd
17556 @opindex mfused-madd
17557 @opindex mno-fused-madd
17558 Generate code that uses (does not use) the floating-point multiply and
17559 accumulate instructions. These instructions are generated by default if
17560 hardware floating point is used.
17562 @item -mwarn-framesize=@var{framesize}
17563 @opindex mwarn-framesize
17564 Emit a warning if the current function exceeds the given frame size. Because
17565 this is a compile-time check it doesn't need to be a real problem when the program
17566 runs. It is intended to identify functions that most probably cause
17567 a stack overflow. It is useful to be used in an environment with limited stack
17568 size e.g.@: the linux kernel.
17570 @item -mwarn-dynamicstack
17571 @opindex mwarn-dynamicstack
17572 Emit a warning if the function calls alloca or uses dynamically
17573 sized arrays. This is generally a bad idea with a limited stack size.
17575 @item -mstack-guard=@var{stack-guard}
17576 @itemx -mstack-size=@var{stack-size}
17577 @opindex mstack-guard
17578 @opindex mstack-size
17579 If these options are provided the s390 back end emits additional instructions in
17580 the function prologue which trigger a trap if the stack size is @var{stack-guard}
17581 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
17582 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
17583 the frame size of the compiled function is chosen.
17584 These options are intended to be used to help debugging stack overflow problems.
17585 The additionally emitted code causes only little overhead and hence can also be
17586 used in production like systems without greater performance degradation. The given
17587 values have to be exact powers of 2 and @var{stack-size} has to be greater than
17588 @var{stack-guard} without exceeding 64k.
17589 In order to be efficient the extra code makes the assumption that the stack starts
17590 at an address aligned to the value given by @var{stack-size}.
17591 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
17594 @node Score Options
17595 @subsection Score Options
17596 @cindex Score Options
17598 These options are defined for Score implementations:
17603 Compile code for big-endian mode. This is the default.
17607 Compile code for little-endian mode.
17611 Disable generate bcnz instruction.
17615 Enable generate unaligned load and store instruction.
17619 Enable the use of multiply-accumulate instructions. Disabled by default.
17623 Specify the SCORE5 as the target architecture.
17627 Specify the SCORE5U of the target architecture.
17631 Specify the SCORE7 as the target architecture. This is the default.
17635 Specify the SCORE7D as the target architecture.
17639 @subsection SH Options
17641 These @samp{-m} options are defined for the SH implementations:
17646 Generate code for the SH1.
17650 Generate code for the SH2.
17653 Generate code for the SH2e.
17657 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
17658 that the floating-point unit is not used.
17660 @item -m2a-single-only
17661 @opindex m2a-single-only
17662 Generate code for the SH2a-FPU, in such a way that no double-precision
17663 floating-point operations are used.
17666 @opindex m2a-single
17667 Generate code for the SH2a-FPU assuming the floating-point unit is in
17668 single-precision mode by default.
17672 Generate code for the SH2a-FPU assuming the floating-point unit is in
17673 double-precision mode by default.
17677 Generate code for the SH3.
17681 Generate code for the SH3e.
17685 Generate code for the SH4 without a floating-point unit.
17687 @item -m4-single-only
17688 @opindex m4-single-only
17689 Generate code for the SH4 with a floating-point unit that only
17690 supports single-precision arithmetic.
17694 Generate code for the SH4 assuming the floating-point unit is in
17695 single-precision mode by default.
17699 Generate code for the SH4.
17703 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
17704 floating-point unit is not used.
17706 @item -m4a-single-only
17707 @opindex m4a-single-only
17708 Generate code for the SH4a, in such a way that no double-precision
17709 floating-point operations are used.
17712 @opindex m4a-single
17713 Generate code for the SH4a assuming the floating-point unit is in
17714 single-precision mode by default.
17718 Generate code for the SH4a.
17722 Same as @option{-m4a-nofpu}, except that it implicitly passes
17723 @option{-dsp} to the assembler. GCC doesn't generate any DSP
17724 instructions at the moment.
17728 Compile code for the processor in big-endian mode.
17732 Compile code for the processor in little-endian mode.
17736 Align doubles at 64-bit boundaries. Note that this changes the calling
17737 conventions, and thus some functions from the standard C library will
17738 not work unless you recompile it first with @option{-mdalign}.
17742 Shorten some address references at link time, when possible; uses the
17743 linker option @option{-relax}.
17747 Use 32-bit offsets in @code{switch} tables. The default is to use
17752 Enable the use of bit manipulation instructions on SH2A.
17756 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
17757 alignment constraints.
17761 Comply with the calling conventions defined by Renesas.
17765 Comply with the calling conventions defined by Renesas.
17769 Comply with the calling conventions defined for GCC before the Renesas
17770 conventions were available. This option is the default for all
17771 targets of the SH toolchain.
17774 @opindex mnomacsave
17775 Mark the @code{MAC} register as call-clobbered, even if
17776 @option{-mhitachi} is given.
17780 Increase IEEE compliance of floating-point code.
17781 At the moment, this is equivalent to @option{-fno-finite-math-only}.
17782 When generating 16-bit SH opcodes, getting IEEE-conforming results for
17783 comparisons of NANs / infinities incurs extra overhead in every
17784 floating-point comparison, therefore the default is set to
17785 @option{-ffinite-math-only}.
17787 @item -minline-ic_invalidate
17788 @opindex minline-ic_invalidate
17789 Inline code to invalidate instruction cache entries after setting up
17790 nested function trampolines.
17791 This option has no effect if -musermode is in effect and the selected
17792 code generation option (e.g. -m4) does not allow the use of the icbi
17794 If the selected code generation option does not allow the use of the icbi
17795 instruction, and -musermode is not in effect, the inlined code will
17796 manipulate the instruction cache address array directly with an associative
17797 write. This not only requires privileged mode, but it will also
17798 fail if the cache line had been mapped via the TLB and has become unmapped.
17802 Dump instruction size and location in the assembly code.
17805 @opindex mpadstruct
17806 This option is deprecated. It pads structures to multiple of 4 bytes,
17807 which is incompatible with the SH ABI@.
17809 @item -msoft-atomic
17810 @opindex msoft-atomic
17811 Generate software atomic sequences for the atomic operations.
17812 This is the default when the target is @code{sh-*-linux*}.
17816 Optimize for space instead of speed. Implied by @option{-Os}.
17819 @opindex mprefergot
17820 When generating position-independent code, emit function calls using
17821 the Global Offset Table instead of the Procedure Linkage Table.
17825 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17826 if the inlined code would not work in user mode.
17827 This is the default when the target is @code{sh-*-linux*}.
17829 @item -multcost=@var{number}
17830 @opindex multcost=@var{number}
17831 Set the cost to assume for a multiply insn.
17833 @item -mdiv=@var{strategy}
17834 @opindex mdiv=@var{strategy}
17835 Set the division strategy to use for SHmedia code. @var{strategy} must be
17836 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17837 inv:call2, inv:fp .
17838 "fp" performs the operation in floating point. This has a very high latency,
17839 but needs only a few instructions, so it might be a good choice if
17840 your code has enough easily-exploitable ILP to allow the compiler to
17841 schedule the floating-point instructions together with other instructions.
17842 Division by zero causes a floating-point exception.
17843 "inv" uses integer operations to calculate the inverse of the divisor,
17844 and then multiplies the dividend with the inverse. This strategy allows
17845 cse and hoisting of the inverse calculation. Division by zero calculates
17846 an unspecified result, but does not trap.
17847 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17848 have been found, or if the entire operation has been hoisted to the same
17849 place, the last stages of the inverse calculation are intertwined with the
17850 final multiply to reduce the overall latency, at the expense of using a few
17851 more instructions, and thus offering fewer scheduling opportunities with
17853 "call" calls a library function that usually implements the inv:minlat
17855 This gives high code density for m5-*media-nofpu compilations.
17856 "call2" uses a different entry point of the same library function, where it
17857 assumes that a pointer to a lookup table has already been set up, which
17858 exposes the pointer load to cse / code hoisting optimizations.
17859 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17860 code generation, but if the code stays unoptimized, revert to the "call",
17861 "call2", or "fp" strategies, respectively. Note that the
17862 potentially-trapping side effect of division by zero is carried by a
17863 separate instruction, so it is possible that all the integer instructions
17864 are hoisted out, but the marker for the side effect stays where it is.
17865 A recombination to fp operations or a call is not possible in that case.
17866 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17867 that the inverse calculation was nor separated from the multiply, they speed
17868 up division where the dividend fits into 20 bits (plus sign where applicable),
17869 by inserting a test to skip a number of operations in this case; this test
17870 slows down the case of larger dividends. inv20u assumes the case of a such
17871 a small dividend to be unlikely, and inv20l assumes it to be likely.
17873 @item -maccumulate-outgoing-args
17874 @opindex maccumulate-outgoing-args
17875 Reserve space once for outgoing arguments in the function prologue rather
17876 than around each call. Generally beneficial for performance and size. Also
17877 needed for unwinding to avoid changing the stack frame around conditional code.
17879 @item -mdivsi3_libfunc=@var{name}
17880 @opindex mdivsi3_libfunc=@var{name}
17881 Set the name of the library function used for 32-bit signed division to
17882 @var{name}. This only affect the name used in the call and inv:call
17883 division strategies, and the compiler will still expect the same
17884 sets of input/output/clobbered registers as if this option was not present.
17886 @item -mfixed-range=@var{register-range}
17887 @opindex mfixed-range
17888 Generate code treating the given register range as fixed registers.
17889 A fixed register is one that the register allocator can not use. This is
17890 useful when compiling kernel code. A register range is specified as
17891 two registers separated by a dash. Multiple register ranges can be
17892 specified separated by a comma.
17894 @item -madjust-unroll
17895 @opindex madjust-unroll
17896 Throttle unrolling to avoid thrashing target registers.
17897 This option only has an effect if the gcc code base supports the
17898 TARGET_ADJUST_UNROLL_MAX target hook.
17900 @item -mindexed-addressing
17901 @opindex mindexed-addressing
17902 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17903 This is only safe if the hardware and/or OS implement 32-bit wrap-around
17904 semantics for the indexed addressing mode. The architecture allows the
17905 implementation of processors with 64-bit MMU, which the OS could use to
17906 get 32-bit addressing, but since no current hardware implementation supports
17907 this or any other way to make the indexed addressing mode safe to use in
17908 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
17910 @item -mgettrcost=@var{number}
17911 @opindex mgettrcost=@var{number}
17912 Set the cost assumed for the gettr instruction to @var{number}.
17913 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17917 Assume pt* instructions won't trap. This will generally generate better
17918 scheduled code, but is unsafe on current hardware. The current architecture
17919 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17920 This has the unintentional effect of making it unsafe to schedule ptabs /
17921 ptrel before a branch, or hoist it out of a loop. For example,
17922 __do_global_ctors, a part of libgcc that runs constructors at program
17923 startup, calls functions in a list which is delimited by @minus{}1. With the
17924 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17925 That means that all the constructors will be run a bit quicker, but when
17926 the loop comes to the end of the list, the program crashes because ptabs
17927 loads @minus{}1 into a target register. Since this option is unsafe for any
17928 hardware implementing the current architecture specification, the default
17929 is -mno-pt-fixed. Unless the user specifies a specific cost with
17930 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17931 this deters register allocation using target registers for storing
17934 @item -minvalid-symbols
17935 @opindex minvalid-symbols
17936 Assume symbols might be invalid. Ordinary function symbols generated by
17937 the compiler will always be valid to load with movi/shori/ptabs or
17938 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17939 to generate symbols that will cause ptabs / ptrel to trap.
17940 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17941 It will then prevent cross-basic-block cse, hoisting and most scheduling
17942 of symbol loads. The default is @option{-mno-invalid-symbols}.
17945 @node Solaris 2 Options
17946 @subsection Solaris 2 Options
17947 @cindex Solaris 2 options
17949 These @samp{-m} options are supported on Solaris 2:
17952 @item -mimpure-text
17953 @opindex mimpure-text
17954 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17955 the compiler to not pass @option{-z text} to the linker when linking a
17956 shared object. Using this option, you can link position-dependent
17957 code into a shared object.
17959 @option{-mimpure-text} suppresses the ``relocations remain against
17960 allocatable but non-writable sections'' linker error message.
17961 However, the necessary relocations will trigger copy-on-write, and the
17962 shared object is not actually shared across processes. Instead of
17963 using @option{-mimpure-text}, you should compile all source code with
17964 @option{-fpic} or @option{-fPIC}.
17968 These switches are supported in addition to the above on Solaris 2:
17973 Add support for multithreading using the POSIX threads library. This
17974 option sets flags for both the preprocessor and linker. This option does
17975 not affect the thread safety of object code produced by the compiler or
17976 that of libraries supplied with it.
17980 This is a synonym for @option{-pthreads}.
17983 @node SPARC Options
17984 @subsection SPARC Options
17985 @cindex SPARC options
17987 These @samp{-m} options are supported on the SPARC:
17990 @item -mno-app-regs
17992 @opindex mno-app-regs
17994 Specify @option{-mapp-regs} to generate output using the global registers
17995 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17998 To be fully SVR4 ABI compliant at the cost of some performance loss,
17999 specify @option{-mno-app-regs}. You should compile libraries and system
18000 software with this option.
18006 With @option{-mflat}, the compiler does not generate save/restore instructions
18007 and uses a ``flat'' or single register window model. This model is compatible
18008 with the regular register window model. The local registers and the input
18009 registers (0--5) are still treated as ``call-saved'' registers and will be
18010 saved on the stack as needed.
18012 With @option{-mno-flat} (the default), the compiler generates save/restore
18013 instructions (except for leaf functions). This is the normal operating mode.
18016 @itemx -mhard-float
18018 @opindex mhard-float
18019 Generate output containing floating-point instructions. This is the
18023 @itemx -msoft-float
18025 @opindex msoft-float
18026 Generate output containing library calls for floating point.
18027 @strong{Warning:} the requisite libraries are not available for all SPARC
18028 targets. Normally the facilities of the machine's usual C compiler are
18029 used, but this cannot be done directly in cross-compilation. You must make
18030 your own arrangements to provide suitable library functions for
18031 cross-compilation. The embedded targets @samp{sparc-*-aout} and
18032 @samp{sparclite-*-*} do provide software floating-point support.
18034 @option{-msoft-float} changes the calling convention in the output file;
18035 therefore, it is only useful if you compile @emph{all} of a program with
18036 this option. In particular, you need to compile @file{libgcc.a}, the
18037 library that comes with GCC, with @option{-msoft-float} in order for
18040 @item -mhard-quad-float
18041 @opindex mhard-quad-float
18042 Generate output containing quad-word (long double) floating-point
18045 @item -msoft-quad-float
18046 @opindex msoft-quad-float
18047 Generate output containing library calls for quad-word (long double)
18048 floating-point instructions. The functions called are those specified
18049 in the SPARC ABI@. This is the default.
18051 As of this writing, there are no SPARC implementations that have hardware
18052 support for the quad-word floating-point instructions. They all invoke
18053 a trap handler for one of these instructions, and then the trap handler
18054 emulates the effect of the instruction. Because of the trap handler overhead,
18055 this is much slower than calling the ABI library routines. Thus the
18056 @option{-msoft-quad-float} option is the default.
18058 @item -mno-unaligned-doubles
18059 @itemx -munaligned-doubles
18060 @opindex mno-unaligned-doubles
18061 @opindex munaligned-doubles
18062 Assume that doubles have 8-byte alignment. This is the default.
18064 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
18065 alignment only if they are contained in another type, or if they have an
18066 absolute address. Otherwise, it assumes they have 4-byte alignment.
18067 Specifying this option avoids some rare compatibility problems with code
18068 generated by other compilers. It is not the default because it results
18069 in a performance loss, especially for floating-point code.
18071 @item -mno-faster-structs
18072 @itemx -mfaster-structs
18073 @opindex mno-faster-structs
18074 @opindex mfaster-structs
18075 With @option{-mfaster-structs}, the compiler assumes that structures
18076 should have 8-byte alignment. This enables the use of pairs of
18077 @code{ldd} and @code{std} instructions for copies in structure
18078 assignment, in place of twice as many @code{ld} and @code{st} pairs.
18079 However, the use of this changed alignment directly violates the SPARC
18080 ABI@. Thus, it's intended only for use on targets where the developer
18081 acknowledges that their resulting code will not be directly in line with
18082 the rules of the ABI@.
18084 @item -mcpu=@var{cpu_type}
18086 Set the instruction set, register set, and instruction scheduling parameters
18087 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
18088 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
18089 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
18090 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
18091 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
18092 and @samp{niagara4}.
18094 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
18095 which selects the best architecture option for the host processor.
18096 @option{-mcpu=native} has no effect if GCC does not recognize
18099 Default instruction scheduling parameters are used for values that select
18100 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
18101 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
18103 Here is a list of each supported architecture and their supported
18111 supersparc, hypersparc, leon
18114 f930, f934, sparclite86x
18120 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
18123 By default (unless configured otherwise), GCC generates code for the V7
18124 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
18125 additionally optimizes it for the Cypress CY7C602 chip, as used in the
18126 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
18127 SPARCStation 1, 2, IPX etc.
18129 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
18130 architecture. The only difference from V7 code is that the compiler emits
18131 the integer multiply and integer divide instructions which exist in SPARC-V8
18132 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
18133 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
18136 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
18137 the SPARC architecture. This adds the integer multiply, integer divide step
18138 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
18139 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
18140 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
18141 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
18142 MB86934 chip, which is the more recent SPARClite with FPU@.
18144 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
18145 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
18146 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
18147 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
18148 optimizes it for the TEMIC SPARClet chip.
18150 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
18151 architecture. This adds 64-bit integer and floating-point move instructions,
18152 3 additional floating-point condition code registers and conditional move
18153 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
18154 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
18155 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
18156 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
18157 @option{-mcpu=niagara}, the compiler additionally optimizes it for
18158 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
18159 additionally optimizes it for Sun UltraSPARC T2 chips. With
18160 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
18161 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
18162 additionally optimizes it for Sun UltraSPARC T4 chips.
18164 @item -mtune=@var{cpu_type}
18166 Set the instruction scheduling parameters for machine type
18167 @var{cpu_type}, but do not set the instruction set or register set that the
18168 option @option{-mcpu=@var{cpu_type}} would.
18170 The same values for @option{-mcpu=@var{cpu_type}} can be used for
18171 @option{-mtune=@var{cpu_type}}, but the only useful values are those
18172 that select a particular CPU implementation. Those are @samp{cypress},
18173 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
18174 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
18175 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
18176 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
18181 @opindex mno-v8plus
18182 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
18183 difference from the V8 ABI is that the global and out registers are
18184 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
18185 mode for all SPARC-V9 processors.
18191 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
18192 Visual Instruction Set extensions. The default is @option{-mno-vis}.
18198 With @option{-mvis2}, GCC generates code that takes advantage of
18199 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
18200 default is @option{-mvis2} when targetting a cpu that supports such
18201 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
18202 also sets @option{-mvis}.
18208 With @option{-mvis3}, GCC generates code that takes advantage of
18209 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
18210 default is @option{-mvis3} when targetting a cpu that supports such
18211 instructions, such as niagara-3 and later. Setting @option{-mvis3}
18212 also sets @option{-mvis2} and @option{-mvis}.
18218 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
18219 population count instruction. The default is @option{-mpopc}
18220 when targetting a cpu that supports such instructions, such as Niagara-2 and
18227 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
18228 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
18229 when targetting a cpu that supports such instructions, such as Niagara-3 and
18233 @opindex mfix-at697f
18234 Enable the documented workaround for the single erratum of the Atmel AT697F
18235 processor (which corresponds to erratum #13 of the AT697E processor).
18238 These @samp{-m} options are supported in addition to the above
18239 on SPARC-V9 processors in 64-bit environments:
18242 @item -mlittle-endian
18243 @opindex mlittle-endian
18244 Generate code for a processor running in little-endian mode. It is only
18245 available for a few configurations and most notably not on Solaris and Linux.
18251 Generate code for a 32-bit or 64-bit environment.
18252 The 32-bit environment sets int, long and pointer to 32 bits.
18253 The 64-bit environment sets int to 32 bits and long and pointer
18256 @item -mcmodel=@var{which}
18258 Set the code model to one of
18262 The Medium/Low code model: 64-bit addresses, programs
18263 must be linked in the low 32 bits of memory. Programs can be statically
18264 or dynamically linked.
18267 The Medium/Middle code model: 64-bit addresses, programs
18268 must be linked in the low 44 bits of memory, the text and data segments must
18269 be less than 2GB in size and the data segment must be located within 2GB of
18273 The Medium/Anywhere code model: 64-bit addresses, programs
18274 may be linked anywhere in memory, the text and data segments must be less
18275 than 2GB in size and the data segment must be located within 2GB of the
18279 The Medium/Anywhere code model for embedded systems:
18280 64-bit addresses, the text and data segments must be less than 2GB in
18281 size, both starting anywhere in memory (determined at link time). The
18282 global register %g4 points to the base of the data segment. Programs
18283 are statically linked and PIC is not supported.
18286 @item -mmemory-model=@var{mem-model}
18287 @opindex mmemory-model
18288 Set the memory model in force on the processor to one of
18292 The default memory model for the processor and operating system.
18295 Relaxed Memory Order
18298 Partial Store Order
18304 Sequential Consistency
18307 These memory models are formally defined in Appendix D of the Sparc V9
18308 architecture manual, as set in the processor's @code{PSTATE.MM} field.
18311 @itemx -mno-stack-bias
18312 @opindex mstack-bias
18313 @opindex mno-stack-bias
18314 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
18315 frame pointer if present, are offset by @minus{}2047 which must be added back
18316 when making stack frame references. This is the default in 64-bit mode.
18317 Otherwise, assume no such offset is present.
18321 @subsection SPU Options
18322 @cindex SPU options
18324 These @samp{-m} options are supported on the SPU:
18328 @itemx -merror-reloc
18329 @opindex mwarn-reloc
18330 @opindex merror-reloc
18332 The loader for SPU does not handle dynamic relocations. By default, GCC
18333 will give an error when it generates code that requires a dynamic
18334 relocation. @option{-mno-error-reloc} disables the error,
18335 @option{-mwarn-reloc} will generate a warning instead.
18338 @itemx -munsafe-dma
18340 @opindex munsafe-dma
18342 Instructions that initiate or test completion of DMA must not be
18343 reordered with respect to loads and stores of the memory that is being
18344 accessed. Users typically address this problem using the volatile
18345 keyword, but that can lead to inefficient code in places where the
18346 memory is known to not change. Rather than mark the memory as volatile
18347 we treat the DMA instructions as potentially effecting all memory. With
18348 @option{-munsafe-dma} users must use the volatile keyword to protect
18351 @item -mbranch-hints
18352 @opindex mbranch-hints
18354 By default, GCC will generate a branch hint instruction to avoid
18355 pipeline stalls for always taken or probably taken branches. A hint
18356 will not be generated closer than 8 instructions away from its branch.
18357 There is little reason to disable them, except for debugging purposes,
18358 or to make an object a little bit smaller.
18362 @opindex msmall-mem
18363 @opindex mlarge-mem
18365 By default, GCC generates code assuming that addresses are never larger
18366 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
18367 a full 32-bit address.
18372 By default, GCC links against startup code that assumes the SPU-style
18373 main function interface (which has an unconventional parameter list).
18374 With @option{-mstdmain}, GCC will link your program against startup
18375 code that assumes a C99-style interface to @code{main}, including a
18376 local copy of @code{argv} strings.
18378 @item -mfixed-range=@var{register-range}
18379 @opindex mfixed-range
18380 Generate code treating the given register range as fixed registers.
18381 A fixed register is one that the register allocator can not use. This is
18382 useful when compiling kernel code. A register range is specified as
18383 two registers separated by a dash. Multiple register ranges can be
18384 specified separated by a comma.
18390 Compile code assuming that pointers to the PPU address space accessed
18391 via the @code{__ea} named address space qualifier are either 32 or 64
18392 bits wide. The default is 32 bits. As this is an ABI changing option,
18393 all object code in an executable must be compiled with the same setting.
18395 @item -maddress-space-conversion
18396 @itemx -mno-address-space-conversion
18397 @opindex maddress-space-conversion
18398 @opindex mno-address-space-conversion
18399 Allow/disallow treating the @code{__ea} address space as superset
18400 of the generic address space. This enables explicit type casts
18401 between @code{__ea} and generic pointer as well as implicit
18402 conversions of generic pointers to @code{__ea} pointers. The
18403 default is to allow address space pointer conversions.
18405 @item -mcache-size=@var{cache-size}
18406 @opindex mcache-size
18407 This option controls the version of libgcc that the compiler links to an
18408 executable and selects a software-managed cache for accessing variables
18409 in the @code{__ea} address space with a particular cache size. Possible
18410 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
18411 and @samp{128}. The default cache size is 64KB.
18413 @item -matomic-updates
18414 @itemx -mno-atomic-updates
18415 @opindex matomic-updates
18416 @opindex mno-atomic-updates
18417 This option controls the version of libgcc that the compiler links to an
18418 executable and selects whether atomic updates to the software-managed
18419 cache of PPU-side variables are used. If you use atomic updates, changes
18420 to a PPU variable from SPU code using the @code{__ea} named address space
18421 qualifier will not interfere with changes to other PPU variables residing
18422 in the same cache line from PPU code. If you do not use atomic updates,
18423 such interference may occur; however, writing back cache lines will be
18424 more efficient. The default behavior is to use atomic updates.
18427 @itemx -mdual-nops=@var{n}
18428 @opindex mdual-nops
18429 By default, GCC will insert nops to increase dual issue when it expects
18430 it to increase performance. @var{n} can be a value from 0 to 10. A
18431 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
18432 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
18434 @item -mhint-max-nops=@var{n}
18435 @opindex mhint-max-nops
18436 Maximum number of nops to insert for a branch hint. A branch hint must
18437 be at least 8 instructions away from the branch it is effecting. GCC
18438 will insert up to @var{n} nops to enforce this, otherwise it will not
18439 generate the branch hint.
18441 @item -mhint-max-distance=@var{n}
18442 @opindex mhint-max-distance
18443 The encoding of the branch hint instruction limits the hint to be within
18444 256 instructions of the branch it is effecting. By default, GCC makes
18445 sure it is within 125.
18448 @opindex msafe-hints
18449 Work around a hardware bug that causes the SPU to stall indefinitely.
18450 By default, GCC will insert the @code{hbrp} instruction to make sure
18451 this stall won't happen.
18455 @node System V Options
18456 @subsection Options for System V
18458 These additional options are available on System V Release 4 for
18459 compatibility with other compilers on those systems:
18464 Create a shared object.
18465 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
18469 Identify the versions of each tool used by the compiler, in a
18470 @code{.ident} assembler directive in the output.
18474 Refrain from adding @code{.ident} directives to the output file (this is
18477 @item -YP,@var{dirs}
18479 Search the directories @var{dirs}, and no others, for libraries
18480 specified with @option{-l}.
18482 @item -Ym,@var{dir}
18484 Look in the directory @var{dir} to find the M4 preprocessor.
18485 The assembler uses this option.
18486 @c This is supposed to go with a -Yd for predefined M4 macro files, but
18487 @c the generic assembler that comes with Solaris takes just -Ym.
18490 @node TILE-Gx Options
18491 @subsection TILE-Gx Options
18492 @cindex TILE-Gx options
18494 These @samp{-m} options are supported on the TILE-Gx:
18497 @item -mcpu=@var{name}
18499 Selects the type of CPU to be targeted. Currently the only supported
18500 type is @samp{tilegx}.
18506 Generate code for a 32-bit or 64-bit environment. The 32-bit
18507 environment sets int, long, and pointer to 32 bits. The 64-bit
18508 environment sets int to 32 bits and long and pointer to 64 bits.
18511 @node TILEPro Options
18512 @subsection TILEPro Options
18513 @cindex TILEPro options
18515 These @samp{-m} options are supported on the TILEPro:
18518 @item -mcpu=@var{name}
18520 Selects the type of CPU to be targeted. Currently the only supported
18521 type is @samp{tilepro}.
18525 Generate code for a 32-bit environment, which sets int, long, and
18526 pointer to 32 bits. This is the only supported behavior so the flag
18527 is essentially ignored.
18531 @subsection V850 Options
18532 @cindex V850 Options
18534 These @samp{-m} options are defined for V850 implementations:
18538 @itemx -mno-long-calls
18539 @opindex mlong-calls
18540 @opindex mno-long-calls
18541 Treat all calls as being far away (near). If calls are assumed to be
18542 far away, the compiler will always load the functions address up into a
18543 register, and call indirect through the pointer.
18549 Do not optimize (do optimize) basic blocks that use the same index
18550 pointer 4 or more times to copy pointer into the @code{ep} register, and
18551 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
18552 option is on by default if you optimize.
18554 @item -mno-prolog-function
18555 @itemx -mprolog-function
18556 @opindex mno-prolog-function
18557 @opindex mprolog-function
18558 Do not use (do use) external functions to save and restore registers
18559 at the prologue and epilogue of a function. The external functions
18560 are slower, but use less code space if more than one function saves
18561 the same number of registers. The @option{-mprolog-function} option
18562 is on by default if you optimize.
18566 Try to make the code as small as possible. At present, this just turns
18567 on the @option{-mep} and @option{-mprolog-function} options.
18569 @item -mtda=@var{n}
18571 Put static or global variables whose size is @var{n} bytes or less into
18572 the tiny data area that register @code{ep} points to. The tiny data
18573 area can hold up to 256 bytes in total (128 bytes for byte references).
18575 @item -msda=@var{n}
18577 Put static or global variables whose size is @var{n} bytes or less into
18578 the small data area that register @code{gp} points to. The small data
18579 area can hold up to 64 kilobytes.
18581 @item -mzda=@var{n}
18583 Put static or global variables whose size is @var{n} bytes or less into
18584 the first 32 kilobytes of memory.
18588 Specify that the target processor is the V850.
18591 @opindex mbig-switch
18592 Generate code suitable for big switch tables. Use this option only if
18593 the assembler/linker complain about out of range branches within a switch
18598 This option will cause r2 and r5 to be used in the code generated by
18599 the compiler. This setting is the default.
18601 @item -mno-app-regs
18602 @opindex mno-app-regs
18603 This option will cause r2 and r5 to be treated as fixed registers.
18607 Specify that the target processor is the V850E2V3. The preprocessor
18608 constants @samp{__v850e2v3__} will be defined if
18609 this option is used.
18613 Specify that the target processor is the V850E2. The preprocessor
18614 constants @samp{__v850e2__} will be defined if this option is used.
18618 Specify that the target processor is the V850E1. The preprocessor
18619 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
18620 this option is used.
18624 Specify that the target processor is the V850ES. This is an alias for
18625 the @option{-mv850e1} option.
18629 Specify that the target processor is the V850E@. The preprocessor
18630 constant @samp{__v850e__} will be defined if this option is used.
18632 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
18633 nor @option{-mv850e2} nor @option{-mv850e2v3}
18634 are defined then a default target processor will be chosen and the
18635 relevant @samp{__v850*__} preprocessor constant will be defined.
18637 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
18638 defined, regardless of which processor variant is the target.
18640 @item -mdisable-callt
18641 @opindex mdisable-callt
18642 This option will suppress generation of the CALLT instruction for the
18643 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
18644 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
18649 @subsection VAX Options
18650 @cindex VAX options
18652 These @samp{-m} options are defined for the VAX:
18657 Do not output certain jump instructions (@code{aobleq} and so on)
18658 that the Unix assembler for the VAX cannot handle across long
18663 Do output those jump instructions, on the assumption that you
18664 will assemble with the GNU assembler.
18668 Output code for G-format floating-point numbers instead of D-format.
18671 @node VxWorks Options
18672 @subsection VxWorks Options
18673 @cindex VxWorks Options
18675 The options in this section are defined for all VxWorks targets.
18676 Options specific to the target hardware are listed with the other
18677 options for that target.
18682 GCC can generate code for both VxWorks kernels and real time processes
18683 (RTPs). This option switches from the former to the latter. It also
18684 defines the preprocessor macro @code{__RTP__}.
18687 @opindex non-static
18688 Link an RTP executable against shared libraries rather than static
18689 libraries. The options @option{-static} and @option{-shared} can
18690 also be used for RTPs (@pxref{Link Options}); @option{-static}
18697 These options are passed down to the linker. They are defined for
18698 compatibility with Diab.
18701 @opindex Xbind-lazy
18702 Enable lazy binding of function calls. This option is equivalent to
18703 @option{-Wl,-z,now} and is defined for compatibility with Diab.
18707 Disable lazy binding of function calls. This option is the default and
18708 is defined for compatibility with Diab.
18711 @node x86-64 Options
18712 @subsection x86-64 Options
18713 @cindex x86-64 options
18715 These are listed under @xref{i386 and x86-64 Options}.
18717 @node Xstormy16 Options
18718 @subsection Xstormy16 Options
18719 @cindex Xstormy16 Options
18721 These options are defined for Xstormy16:
18726 Choose startup files and linker script suitable for the simulator.
18729 @node Xtensa Options
18730 @subsection Xtensa Options
18731 @cindex Xtensa Options
18733 These options are supported for Xtensa targets:
18737 @itemx -mno-const16
18739 @opindex mno-const16
18740 Enable or disable use of @code{CONST16} instructions for loading
18741 constant values. The @code{CONST16} instruction is currently not a
18742 standard option from Tensilica. When enabled, @code{CONST16}
18743 instructions are always used in place of the standard @code{L32R}
18744 instructions. The use of @code{CONST16} is enabled by default only if
18745 the @code{L32R} instruction is not available.
18748 @itemx -mno-fused-madd
18749 @opindex mfused-madd
18750 @opindex mno-fused-madd
18751 Enable or disable use of fused multiply/add and multiply/subtract
18752 instructions in the floating-point option. This has no effect if the
18753 floating-point option is not also enabled. Disabling fused multiply/add
18754 and multiply/subtract instructions forces the compiler to use separate
18755 instructions for the multiply and add/subtract operations. This may be
18756 desirable in some cases where strict IEEE 754-compliant results are
18757 required: the fused multiply add/subtract instructions do not round the
18758 intermediate result, thereby producing results with @emph{more} bits of
18759 precision than specified by the IEEE standard. Disabling fused multiply
18760 add/subtract instructions also ensures that the program output is not
18761 sensitive to the compiler's ability to combine multiply and add/subtract
18764 @item -mserialize-volatile
18765 @itemx -mno-serialize-volatile
18766 @opindex mserialize-volatile
18767 @opindex mno-serialize-volatile
18768 When this option is enabled, GCC inserts @code{MEMW} instructions before
18769 @code{volatile} memory references to guarantee sequential consistency.
18770 The default is @option{-mserialize-volatile}. Use
18771 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
18773 @item -mforce-no-pic
18774 @opindex mforce-no-pic
18775 For targets, like GNU/Linux, where all user-mode Xtensa code must be
18776 position-independent code (PIC), this option disables PIC for compiling
18779 @item -mtext-section-literals
18780 @itemx -mno-text-section-literals
18781 @opindex mtext-section-literals
18782 @opindex mno-text-section-literals
18783 Control the treatment of literal pools. The default is
18784 @option{-mno-text-section-literals}, which places literals in a separate
18785 section in the output file. This allows the literal pool to be placed
18786 in a data RAM/ROM, and it also allows the linker to combine literal
18787 pools from separate object files to remove redundant literals and
18788 improve code size. With @option{-mtext-section-literals}, the literals
18789 are interspersed in the text section in order to keep them as close as
18790 possible to their references. This may be necessary for large assembly
18793 @item -mtarget-align
18794 @itemx -mno-target-align
18795 @opindex mtarget-align
18796 @opindex mno-target-align
18797 When this option is enabled, GCC instructs the assembler to
18798 automatically align instructions to reduce branch penalties at the
18799 expense of some code density. The assembler attempts to widen density
18800 instructions to align branch targets and the instructions following call
18801 instructions. If there are not enough preceding safe density
18802 instructions to align a target, no widening will be performed. The
18803 default is @option{-mtarget-align}. These options do not affect the
18804 treatment of auto-aligned instructions like @code{LOOP}, which the
18805 assembler will always align, either by widening density instructions or
18806 by inserting no-op instructions.
18809 @itemx -mno-longcalls
18810 @opindex mlongcalls
18811 @opindex mno-longcalls
18812 When this option is enabled, GCC instructs the assembler to translate
18813 direct calls to indirect calls unless it can determine that the target
18814 of a direct call is in the range allowed by the call instruction. This
18815 translation typically occurs for calls to functions in other source
18816 files. Specifically, the assembler translates a direct @code{CALL}
18817 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
18818 The default is @option{-mno-longcalls}. This option should be used in
18819 programs where the call target can potentially be out of range. This
18820 option is implemented in the assembler, not the compiler, so the
18821 assembly code generated by GCC will still show direct call
18822 instructions---look at the disassembled object code to see the actual
18823 instructions. Note that the assembler will use an indirect call for
18824 every cross-file call, not just those that really will be out of range.
18827 @node zSeries Options
18828 @subsection zSeries Options
18829 @cindex zSeries options
18831 These are listed under @xref{S/390 and zSeries Options}.
18833 @node Code Gen Options
18834 @section Options for Code Generation Conventions
18835 @cindex code generation conventions
18836 @cindex options, code generation
18837 @cindex run-time options
18839 These machine-independent options control the interface conventions
18840 used in code generation.
18842 Most of them have both positive and negative forms; the negative form
18843 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
18844 one of the forms is listed---the one that is not the default. You
18845 can figure out the other form by either removing @samp{no-} or adding
18849 @item -fbounds-check
18850 @opindex fbounds-check
18851 For front-ends that support it, generate additional code to check that
18852 indices used to access arrays are within the declared range. This is
18853 currently only supported by the Java and Fortran front-ends, where
18854 this option defaults to true and false respectively.
18858 This option generates traps for signed overflow on addition, subtraction,
18859 multiplication operations.
18863 This option instructs the compiler to assume that signed arithmetic
18864 overflow of addition, subtraction and multiplication wraps around
18865 using twos-complement representation. This flag enables some optimizations
18866 and disables others. This option is enabled by default for the Java
18867 front-end, as required by the Java language specification.
18870 @opindex fexceptions
18871 Enable exception handling. Generates extra code needed to propagate
18872 exceptions. For some targets, this implies GCC will generate frame
18873 unwind information for all functions, which can produce significant data
18874 size overhead, although it does not affect execution. If you do not
18875 specify this option, GCC will enable it by default for languages like
18876 C++ that normally require exception handling, and disable it for
18877 languages like C that do not normally require it. However, you may need
18878 to enable this option when compiling C code that needs to interoperate
18879 properly with exception handlers written in C++. You may also wish to
18880 disable this option if you are compiling older C++ programs that don't
18881 use exception handling.
18883 @item -fnon-call-exceptions
18884 @opindex fnon-call-exceptions
18885 Generate code that allows trapping instructions to throw exceptions.
18886 Note that this requires platform-specific runtime support that does
18887 not exist everywhere. Moreover, it only allows @emph{trapping}
18888 instructions to throw exceptions, i.e.@: memory references or floating-point
18889 instructions. It does not allow exceptions to be thrown from
18890 arbitrary signal handlers such as @code{SIGALRM}.
18892 @item -funwind-tables
18893 @opindex funwind-tables
18894 Similar to @option{-fexceptions}, except that it will just generate any needed
18895 static data, but will not affect the generated code in any other way.
18896 You will normally not enable this option; instead, a language processor
18897 that needs this handling would enable it on your behalf.
18899 @item -fasynchronous-unwind-tables
18900 @opindex fasynchronous-unwind-tables
18901 Generate unwind table in dwarf2 format, if supported by target machine. The
18902 table is exact at each instruction boundary, so it can be used for stack
18903 unwinding from asynchronous events (such as debugger or garbage collector).
18905 @item -fpcc-struct-return
18906 @opindex fpcc-struct-return
18907 Return ``short'' @code{struct} and @code{union} values in memory like
18908 longer ones, rather than in registers. This convention is less
18909 efficient, but it has the advantage of allowing intercallability between
18910 GCC-compiled files and files compiled with other compilers, particularly
18911 the Portable C Compiler (pcc).
18913 The precise convention for returning structures in memory depends
18914 on the target configuration macros.
18916 Short structures and unions are those whose size and alignment match
18917 that of some integer type.
18919 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
18920 switch is not binary compatible with code compiled with the
18921 @option{-freg-struct-return} switch.
18922 Use it to conform to a non-default application binary interface.
18924 @item -freg-struct-return
18925 @opindex freg-struct-return
18926 Return @code{struct} and @code{union} values in registers when possible.
18927 This is more efficient for small structures than
18928 @option{-fpcc-struct-return}.
18930 If you specify neither @option{-fpcc-struct-return} nor
18931 @option{-freg-struct-return}, GCC defaults to whichever convention is
18932 standard for the target. If there is no standard convention, GCC
18933 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18934 the principal compiler. In those cases, we can choose the standard, and
18935 we chose the more efficient register return alternative.
18937 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18938 switch is not binary compatible with code compiled with the
18939 @option{-fpcc-struct-return} switch.
18940 Use it to conform to a non-default application binary interface.
18942 @item -fshort-enums
18943 @opindex fshort-enums
18944 Allocate to an @code{enum} type only as many bytes as it needs for the
18945 declared range of possible values. Specifically, the @code{enum} type
18946 will be equivalent to the smallest integer type that has enough room.
18948 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18949 code that is not binary compatible with code generated without that switch.
18950 Use it to conform to a non-default application binary interface.
18952 @item -fshort-double
18953 @opindex fshort-double
18954 Use the same size for @code{double} as for @code{float}.
18956 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18957 code that is not binary compatible with code generated without that switch.
18958 Use it to conform to a non-default application binary interface.
18960 @item -fshort-wchar
18961 @opindex fshort-wchar
18962 Override the underlying type for @samp{wchar_t} to be @samp{short
18963 unsigned int} instead of the default for the target. This option is
18964 useful for building programs to run under WINE@.
18966 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18967 code that is not binary compatible with code generated without that switch.
18968 Use it to conform to a non-default application binary interface.
18971 @opindex fno-common
18972 In C code, controls the placement of uninitialized global variables.
18973 Unix C compilers have traditionally permitted multiple definitions of
18974 such variables in different compilation units by placing the variables
18976 This is the behavior specified by @option{-fcommon}, and is the default
18977 for GCC on most targets.
18978 On the other hand, this behavior is not required by ISO C, and on some
18979 targets may carry a speed or code size penalty on variable references.
18980 The @option{-fno-common} option specifies that the compiler should place
18981 uninitialized global variables in the data section of the object file,
18982 rather than generating them as common blocks.
18983 This has the effect that if the same variable is declared
18984 (without @code{extern}) in two different compilations,
18985 you will get a multiple-definition error when you link them.
18986 In this case, you must compile with @option{-fcommon} instead.
18987 Compiling with @option{-fno-common} is useful on targets for which
18988 it provides better performance, or if you wish to verify that the
18989 program will work on other systems that always treat uninitialized
18990 variable declarations this way.
18994 Ignore the @samp{#ident} directive.
18996 @item -finhibit-size-directive
18997 @opindex finhibit-size-directive
18998 Don't output a @code{.size} assembler directive, or anything else that
18999 would cause trouble if the function is split in the middle, and the
19000 two halves are placed at locations far apart in memory. This option is
19001 used when compiling @file{crtstuff.c}; you should not need to use it
19004 @item -fverbose-asm
19005 @opindex fverbose-asm
19006 Put extra commentary information in the generated assembly code to
19007 make it more readable. This option is generally only of use to those
19008 who actually need to read the generated assembly code (perhaps while
19009 debugging the compiler itself).
19011 @option{-fno-verbose-asm}, the default, causes the
19012 extra information to be omitted and is useful when comparing two assembler
19015 @item -frecord-gcc-switches
19016 @opindex frecord-gcc-switches
19017 This switch causes the command line that was used to invoke the
19018 compiler to be recorded into the object file that is being created.
19019 This switch is only implemented on some targets and the exact format
19020 of the recording is target and binary file format dependent, but it
19021 usually takes the form of a section containing ASCII text. This
19022 switch is related to the @option{-fverbose-asm} switch, but that
19023 switch only records information in the assembler output file as
19024 comments, so it never reaches the object file.
19025 See also @option{-grecord-gcc-switches} for another
19026 way of storing compiler options into the object file.
19030 @cindex global offset table
19032 Generate position-independent code (PIC) suitable for use in a shared
19033 library, if supported for the target machine. Such code accesses all
19034 constant addresses through a global offset table (GOT)@. The dynamic
19035 loader resolves the GOT entries when the program starts (the dynamic
19036 loader is not part of GCC; it is part of the operating system). If
19037 the GOT size for the linked executable exceeds a machine-specific
19038 maximum size, you get an error message from the linker indicating that
19039 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
19040 instead. (These maximums are 8k on the SPARC and 32k
19041 on the m68k and RS/6000. The 386 has no such limit.)
19043 Position-independent code requires special support, and therefore works
19044 only on certain machines. For the 386, GCC supports PIC for System V
19045 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
19046 position-independent.
19048 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19053 If supported for the target machine, emit position-independent code,
19054 suitable for dynamic linking and avoiding any limit on the size of the
19055 global offset table. This option makes a difference on the m68k,
19056 PowerPC and SPARC@.
19058 Position-independent code requires special support, and therefore works
19059 only on certain machines.
19061 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19068 These options are similar to @option{-fpic} and @option{-fPIC}, but
19069 generated position independent code can be only linked into executables.
19070 Usually these options are used when @option{-pie} GCC option will be
19071 used during linking.
19073 @option{-fpie} and @option{-fPIE} both define the macros
19074 @code{__pie__} and @code{__PIE__}. The macros have the value 1
19075 for @option{-fpie} and 2 for @option{-fPIE}.
19077 @item -fno-jump-tables
19078 @opindex fno-jump-tables
19079 Do not use jump tables for switch statements even where it would be
19080 more efficient than other code generation strategies. This option is
19081 of use in conjunction with @option{-fpic} or @option{-fPIC} for
19082 building code that forms part of a dynamic linker and cannot
19083 reference the address of a jump table. On some targets, jump tables
19084 do not require a GOT and this option is not needed.
19086 @item -ffixed-@var{reg}
19088 Treat the register named @var{reg} as a fixed register; generated code
19089 should never refer to it (except perhaps as a stack pointer, frame
19090 pointer or in some other fixed role).
19092 @var{reg} must be the name of a register. The register names accepted
19093 are machine-specific and are defined in the @code{REGISTER_NAMES}
19094 macro in the machine description macro file.
19096 This flag does not have a negative form, because it specifies a
19099 @item -fcall-used-@var{reg}
19100 @opindex fcall-used
19101 Treat the register named @var{reg} as an allocable register that is
19102 clobbered by function calls. It may be allocated for temporaries or
19103 variables that do not live across a call. Functions compiled this way
19104 will not save and restore the register @var{reg}.
19106 It is an error to used this flag with the frame pointer or stack pointer.
19107 Use of this flag for other registers that have fixed pervasive roles in
19108 the machine's execution model will produce disastrous results.
19110 This flag does not have a negative form, because it specifies a
19113 @item -fcall-saved-@var{reg}
19114 @opindex fcall-saved
19115 Treat the register named @var{reg} as an allocable register saved by
19116 functions. It may be allocated even for temporaries or variables that
19117 live across a call. Functions compiled this way will save and restore
19118 the register @var{reg} if they use it.
19120 It is an error to used this flag with the frame pointer or stack pointer.
19121 Use of this flag for other registers that have fixed pervasive roles in
19122 the machine's execution model will produce disastrous results.
19124 A different sort of disaster will result from the use of this flag for
19125 a register in which function values may be returned.
19127 This flag does not have a negative form, because it specifies a
19130 @item -fpack-struct[=@var{n}]
19131 @opindex fpack-struct
19132 Without a value specified, pack all structure members together without
19133 holes. When a value is specified (which must be a small power of two), pack
19134 structure members according to this value, representing the maximum
19135 alignment (that is, objects with default alignment requirements larger than
19136 this will be output potentially unaligned at the next fitting location.
19138 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
19139 code that is not binary compatible with code generated without that switch.
19140 Additionally, it makes the code suboptimal.
19141 Use it to conform to a non-default application binary interface.
19143 @item -finstrument-functions
19144 @opindex finstrument-functions
19145 Generate instrumentation calls for entry and exit to functions. Just
19146 after function entry and just before function exit, the following
19147 profiling functions will be called with the address of the current
19148 function and its call site. (On some platforms,
19149 @code{__builtin_return_address} does not work beyond the current
19150 function, so the call site information may not be available to the
19151 profiling functions otherwise.)
19154 void __cyg_profile_func_enter (void *this_fn,
19156 void __cyg_profile_func_exit (void *this_fn,
19160 The first argument is the address of the start of the current function,
19161 which may be looked up exactly in the symbol table.
19163 This instrumentation is also done for functions expanded inline in other
19164 functions. The profiling calls will indicate where, conceptually, the
19165 inline function is entered and exited. This means that addressable
19166 versions of such functions must be available. If all your uses of a
19167 function are expanded inline, this may mean an additional expansion of
19168 code size. If you use @samp{extern inline} in your C code, an
19169 addressable version of such functions must be provided. (This is
19170 normally the case anyways, but if you get lucky and the optimizer always
19171 expands the functions inline, you might have gotten away without
19172 providing static copies.)
19174 A function may be given the attribute @code{no_instrument_function}, in
19175 which case this instrumentation will not be done. This can be used, for
19176 example, for the profiling functions listed above, high-priority
19177 interrupt routines, and any functions from which the profiling functions
19178 cannot safely be called (perhaps signal handlers, if the profiling
19179 routines generate output or allocate memory).
19181 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
19182 @opindex finstrument-functions-exclude-file-list
19184 Set the list of functions that are excluded from instrumentation (see
19185 the description of @code{-finstrument-functions}). If the file that
19186 contains a function definition matches with one of @var{file}, then
19187 that function is not instrumented. The match is done on substrings:
19188 if the @var{file} parameter is a substring of the file name, it is
19189 considered to be a match.
19194 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
19198 will exclude any inline function defined in files whose pathnames
19199 contain @code{/bits/stl} or @code{include/sys}.
19201 If, for some reason, you want to include letter @code{','} in one of
19202 @var{sym}, write @code{'\,'}. For example,
19203 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
19204 (note the single quote surrounding the option).
19206 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
19207 @opindex finstrument-functions-exclude-function-list
19209 This is similar to @code{-finstrument-functions-exclude-file-list},
19210 but this option sets the list of function names to be excluded from
19211 instrumentation. The function name to be matched is its user-visible
19212 name, such as @code{vector<int> blah(const vector<int> &)}, not the
19213 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
19214 match is done on substrings: if the @var{sym} parameter is a substring
19215 of the function name, it is considered to be a match. For C99 and C++
19216 extended identifiers, the function name must be given in UTF-8, not
19217 using universal character names.
19219 @item -fstack-check
19220 @opindex fstack-check
19221 Generate code to verify that you do not go beyond the boundary of the
19222 stack. You should specify this flag if you are running in an
19223 environment with multiple threads, but only rarely need to specify it in
19224 a single-threaded environment since stack overflow is automatically
19225 detected on nearly all systems if there is only one stack.
19227 Note that this switch does not actually cause checking to be done; the
19228 operating system or the language runtime must do that. The switch causes
19229 generation of code to ensure that they see the stack being extended.
19231 You can additionally specify a string parameter: @code{no} means no
19232 checking, @code{generic} means force the use of old-style checking,
19233 @code{specific} means use the best checking method and is equivalent
19234 to bare @option{-fstack-check}.
19236 Old-style checking is a generic mechanism that requires no specific
19237 target support in the compiler but comes with the following drawbacks:
19241 Modified allocation strategy for large objects: they will always be
19242 allocated dynamically if their size exceeds a fixed threshold.
19245 Fixed limit on the size of the static frame of functions: when it is
19246 topped by a particular function, stack checking is not reliable and
19247 a warning is issued by the compiler.
19250 Inefficiency: because of both the modified allocation strategy and the
19251 generic implementation, the performances of the code are hampered.
19254 Note that old-style stack checking is also the fallback method for
19255 @code{specific} if no target support has been added in the compiler.
19257 @item -fstack-limit-register=@var{reg}
19258 @itemx -fstack-limit-symbol=@var{sym}
19259 @itemx -fno-stack-limit
19260 @opindex fstack-limit-register
19261 @opindex fstack-limit-symbol
19262 @opindex fno-stack-limit
19263 Generate code to ensure that the stack does not grow beyond a certain value,
19264 either the value of a register or the address of a symbol. If the stack
19265 would grow beyond the value, a signal is raised. For most targets,
19266 the signal is raised before the stack overruns the boundary, so
19267 it is possible to catch the signal without taking special precautions.
19269 For instance, if the stack starts at absolute address @samp{0x80000000}
19270 and grows downwards, you can use the flags
19271 @option{-fstack-limit-symbol=__stack_limit} and
19272 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
19273 of 128KB@. Note that this may only work with the GNU linker.
19275 @item -fsplit-stack
19276 @opindex fsplit-stack
19277 Generate code to automatically split the stack before it overflows.
19278 The resulting program has a discontiguous stack which can only
19279 overflow if the program is unable to allocate any more memory. This
19280 is most useful when running threaded programs, as it is no longer
19281 necessary to calculate a good stack size to use for each thread. This
19282 is currently only implemented for the i386 and x86_64 backends running
19285 When code compiled with @option{-fsplit-stack} calls code compiled
19286 without @option{-fsplit-stack}, there may not be much stack space
19287 available for the latter code to run. If compiling all code,
19288 including library code, with @option{-fsplit-stack} is not an option,
19289 then the linker can fix up these calls so that the code compiled
19290 without @option{-fsplit-stack} always has a large stack. Support for
19291 this is implemented in the gold linker in GNU binutils release 2.21
19294 @item -fleading-underscore
19295 @opindex fleading-underscore
19296 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
19297 change the way C symbols are represented in the object file. One use
19298 is to help link with legacy assembly code.
19300 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
19301 generate code that is not binary compatible with code generated without that
19302 switch. Use it to conform to a non-default application binary interface.
19303 Not all targets provide complete support for this switch.
19305 @item -ftls-model=@var{model}
19306 @opindex ftls-model
19307 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
19308 The @var{model} argument should be one of @code{global-dynamic},
19309 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
19311 The default without @option{-fpic} is @code{initial-exec}; with
19312 @option{-fpic} the default is @code{global-dynamic}.
19314 @item -fvisibility=@var{default|internal|hidden|protected}
19315 @opindex fvisibility
19316 Set the default ELF image symbol visibility to the specified option---all
19317 symbols will be marked with this unless overridden within the code.
19318 Using this feature can very substantially improve linking and
19319 load times of shared object libraries, produce more optimized
19320 code, provide near-perfect API export and prevent symbol clashes.
19321 It is @strong{strongly} recommended that you use this in any shared objects
19324 Despite the nomenclature, @code{default} always means public; i.e.,
19325 available to be linked against from outside the shared object.
19326 @code{protected} and @code{internal} are pretty useless in real-world
19327 usage so the only other commonly used option will be @code{hidden}.
19328 The default if @option{-fvisibility} isn't specified is
19329 @code{default}, i.e., make every
19330 symbol public---this causes the same behavior as previous versions of
19333 A good explanation of the benefits offered by ensuring ELF
19334 symbols have the correct visibility is given by ``How To Write
19335 Shared Libraries'' by Ulrich Drepper (which can be found at
19336 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
19337 solution made possible by this option to marking things hidden when
19338 the default is public is to make the default hidden and mark things
19339 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
19340 and @code{__attribute__ ((visibility("default")))} instead of
19341 @code{__declspec(dllexport)} you get almost identical semantics with
19342 identical syntax. This is a great boon to those working with
19343 cross-platform projects.
19345 For those adding visibility support to existing code, you may find
19346 @samp{#pragma GCC visibility} of use. This works by you enclosing
19347 the declarations you wish to set visibility for with (for example)
19348 @samp{#pragma GCC visibility push(hidden)} and
19349 @samp{#pragma GCC visibility pop}.
19350 Bear in mind that symbol visibility should be viewed @strong{as
19351 part of the API interface contract} and thus all new code should
19352 always specify visibility when it is not the default; i.e., declarations
19353 only for use within the local DSO should @strong{always} be marked explicitly
19354 as hidden as so to avoid PLT indirection overheads---making this
19355 abundantly clear also aids readability and self-documentation of the code.
19356 Note that due to ISO C++ specification requirements, operator new and
19357 operator delete must always be of default visibility.
19359 Be aware that headers from outside your project, in particular system
19360 headers and headers from any other library you use, may not be
19361 expecting to be compiled with visibility other than the default. You
19362 may need to explicitly say @samp{#pragma GCC visibility push(default)}
19363 before including any such headers.
19365 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
19366 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
19367 no modifications. However, this means that calls to @samp{extern}
19368 functions with no explicit visibility will use the PLT, so it is more
19369 effective to use @samp{__attribute ((visibility))} and/or
19370 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
19371 declarations should be treated as hidden.
19373 Note that @samp{-fvisibility} does affect C++ vague linkage
19374 entities. This means that, for instance, an exception class that will
19375 be thrown between DSOs must be explicitly marked with default
19376 visibility so that the @samp{type_info} nodes will be unified between
19379 An overview of these techniques, their benefits and how to use them
19380 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
19382 @item -fstrict-volatile-bitfields
19383 @opindex fstrict-volatile-bitfields
19384 This option should be used if accesses to volatile bitfields (or other
19385 structure fields, although the compiler usually honors those types
19386 anyway) should use a single access of the width of the
19387 field's type, aligned to a natural alignment if possible. For
19388 example, targets with memory-mapped peripheral registers might require
19389 all such accesses to be 16 bits wide; with this flag the user could
19390 declare all peripheral bitfields as ``unsigned short'' (assuming short
19391 is 16 bits on these targets) to force GCC to use 16-bit accesses
19392 instead of, perhaps, a more efficient 32-bit access.
19394 If this option is disabled, the compiler will use the most efficient
19395 instruction. In the previous example, that might be a 32-bit load
19396 instruction, even though that will access bytes that do not contain
19397 any portion of the bitfield, or memory-mapped registers unrelated to
19398 the one being updated.
19400 If the target requires strict alignment, and honoring the field
19401 type would require violating this alignment, a warning is issued.
19402 If the field has @code{packed} attribute, the access is done without
19403 honoring the field type. If the field doesn't have @code{packed}
19404 attribute, the access is done honoring the field type. In both cases,
19405 GCC assumes that the user knows something about the target hardware
19406 that it is unaware of.
19408 The default value of this option is determined by the application binary
19409 interface for the target processor.
19415 @node Environment Variables
19416 @section Environment Variables Affecting GCC
19417 @cindex environment variables
19419 @c man begin ENVIRONMENT
19420 This section describes several environment variables that affect how GCC
19421 operates. Some of them work by specifying directories or prefixes to use
19422 when searching for various kinds of files. Some are used to specify other
19423 aspects of the compilation environment.
19425 Note that you can also specify places to search using options such as
19426 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
19427 take precedence over places specified using environment variables, which
19428 in turn take precedence over those specified by the configuration of GCC@.
19429 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
19430 GNU Compiler Collection (GCC) Internals}.
19435 @c @itemx LC_COLLATE
19437 @c @itemx LC_MONETARY
19438 @c @itemx LC_NUMERIC
19443 @c @findex LC_COLLATE
19444 @findex LC_MESSAGES
19445 @c @findex LC_MONETARY
19446 @c @findex LC_NUMERIC
19450 These environment variables control the way that GCC uses
19451 localization information which allows GCC to work with different
19452 national conventions. GCC inspects the locale categories
19453 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
19454 so. These locale categories can be set to any value supported by your
19455 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
19456 Kingdom encoded in UTF-8.
19458 The @env{LC_CTYPE} environment variable specifies character
19459 classification. GCC uses it to determine the character boundaries in
19460 a string; this is needed for some multibyte encodings that contain quote
19461 and escape characters that would otherwise be interpreted as a string
19464 The @env{LC_MESSAGES} environment variable specifies the language to
19465 use in diagnostic messages.
19467 If the @env{LC_ALL} environment variable is set, it overrides the value
19468 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
19469 and @env{LC_MESSAGES} default to the value of the @env{LANG}
19470 environment variable. If none of these variables are set, GCC
19471 defaults to traditional C English behavior.
19475 If @env{TMPDIR} is set, it specifies the directory to use for temporary
19476 files. GCC uses temporary files to hold the output of one stage of
19477 compilation which is to be used as input to the next stage: for example,
19478 the output of the preprocessor, which is the input to the compiler
19481 @item GCC_COMPARE_DEBUG
19482 @findex GCC_COMPARE_DEBUG
19483 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
19484 @option{-fcompare-debug} to the compiler driver. See the documentation
19485 of this option for more details.
19487 @item GCC_EXEC_PREFIX
19488 @findex GCC_EXEC_PREFIX
19489 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
19490 names of the subprograms executed by the compiler. No slash is added
19491 when this prefix is combined with the name of a subprogram, but you can
19492 specify a prefix that ends with a slash if you wish.
19494 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
19495 an appropriate prefix to use based on the pathname it was invoked with.
19497 If GCC cannot find the subprogram using the specified prefix, it
19498 tries looking in the usual places for the subprogram.
19500 The default value of @env{GCC_EXEC_PREFIX} is
19501 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
19502 the installed compiler. In many cases @var{prefix} is the value
19503 of @code{prefix} when you ran the @file{configure} script.
19505 Other prefixes specified with @option{-B} take precedence over this prefix.
19507 This prefix is also used for finding files such as @file{crt0.o} that are
19510 In addition, the prefix is used in an unusual way in finding the
19511 directories to search for header files. For each of the standard
19512 directories whose name normally begins with @samp{/usr/local/lib/gcc}
19513 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
19514 replacing that beginning with the specified prefix to produce an
19515 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
19516 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
19517 These alternate directories are searched first; the standard directories
19518 come next. If a standard directory begins with the configured
19519 @var{prefix} then the value of @var{prefix} is replaced by
19520 @env{GCC_EXEC_PREFIX} when looking for header files.
19522 @item COMPILER_PATH
19523 @findex COMPILER_PATH
19524 The value of @env{COMPILER_PATH} is a colon-separated list of
19525 directories, much like @env{PATH}. GCC tries the directories thus
19526 specified when searching for subprograms, if it can't find the
19527 subprograms using @env{GCC_EXEC_PREFIX}.
19530 @findex LIBRARY_PATH
19531 The value of @env{LIBRARY_PATH} is a colon-separated list of
19532 directories, much like @env{PATH}. When configured as a native compiler,
19533 GCC tries the directories thus specified when searching for special
19534 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
19535 using GCC also uses these directories when searching for ordinary
19536 libraries for the @option{-l} option (but directories specified with
19537 @option{-L} come first).
19541 @cindex locale definition
19542 This variable is used to pass locale information to the compiler. One way in
19543 which this information is used is to determine the character set to be used
19544 when character literals, string literals and comments are parsed in C and C++.
19545 When the compiler is configured to allow multibyte characters,
19546 the following values for @env{LANG} are recognized:
19550 Recognize JIS characters.
19552 Recognize SJIS characters.
19554 Recognize EUCJP characters.
19557 If @env{LANG} is not defined, or if it has some other value, then the
19558 compiler will use mblen and mbtowc as defined by the default locale to
19559 recognize and translate multibyte characters.
19563 Some additional environments variables affect the behavior of the
19566 @include cppenv.texi
19570 @node Precompiled Headers
19571 @section Using Precompiled Headers
19572 @cindex precompiled headers
19573 @cindex speed of compilation
19575 Often large projects have many header files that are included in every
19576 source file. The time the compiler takes to process these header files
19577 over and over again can account for nearly all of the time required to
19578 build the project. To make builds faster, GCC allows users to
19579 `precompile' a header file; then, if builds can use the precompiled
19580 header file they will be much faster.
19582 To create a precompiled header file, simply compile it as you would any
19583 other file, if necessary using the @option{-x} option to make the driver
19584 treat it as a C or C++ header file. You will probably want to use a
19585 tool like @command{make} to keep the precompiled header up-to-date when
19586 the headers it contains change.
19588 A precompiled header file will be searched for when @code{#include} is
19589 seen in the compilation. As it searches for the included file
19590 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
19591 compiler looks for a precompiled header in each directory just before it
19592 looks for the include file in that directory. The name searched for is
19593 the name specified in the @code{#include} with @samp{.gch} appended. If
19594 the precompiled header file can't be used, it is ignored.
19596 For instance, if you have @code{#include "all.h"}, and you have
19597 @file{all.h.gch} in the same directory as @file{all.h}, then the
19598 precompiled header file will be used if possible, and the original
19599 header will be used otherwise.
19601 Alternatively, you might decide to put the precompiled header file in a
19602 directory and use @option{-I} to ensure that directory is searched
19603 before (or instead of) the directory containing the original header.
19604 Then, if you want to check that the precompiled header file is always
19605 used, you can put a file of the same name as the original header in this
19606 directory containing an @code{#error} command.
19608 This also works with @option{-include}. So yet another way to use
19609 precompiled headers, good for projects not designed with precompiled
19610 header files in mind, is to simply take most of the header files used by
19611 a project, include them from another header file, precompile that header
19612 file, and @option{-include} the precompiled header. If the header files
19613 have guards against multiple inclusion, they will be skipped because
19614 they've already been included (in the precompiled header).
19616 If you need to precompile the same header file for different
19617 languages, targets, or compiler options, you can instead make a
19618 @emph{directory} named like @file{all.h.gch}, and put each precompiled
19619 header in the directory, perhaps using @option{-o}. It doesn't matter
19620 what you call the files in the directory, every precompiled header in
19621 the directory will be considered. The first precompiled header
19622 encountered in the directory that is valid for this compilation will
19623 be used; they're searched in no particular order.
19625 There are many other possibilities, limited only by your imagination,
19626 good sense, and the constraints of your build system.
19628 A precompiled header file can be used only when these conditions apply:
19632 Only one precompiled header can be used in a particular compilation.
19635 A precompiled header can't be used once the first C token is seen. You
19636 can have preprocessor directives before a precompiled header; you can
19637 even include a precompiled header from inside another header, so long as
19638 there are no C tokens before the @code{#include}.
19641 The precompiled header file must be produced for the same language as
19642 the current compilation. You can't use a C precompiled header for a C++
19646 The precompiled header file must have been produced by the same compiler
19647 binary as the current compilation is using.
19650 Any macros defined before the precompiled header is included must
19651 either be defined in the same way as when the precompiled header was
19652 generated, or must not affect the precompiled header, which usually
19653 means that they don't appear in the precompiled header at all.
19655 The @option{-D} option is one way to define a macro before a
19656 precompiled header is included; using a @code{#define} can also do it.
19657 There are also some options that define macros implicitly, like
19658 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
19661 @item If debugging information is output when using the precompiled
19662 header, using @option{-g} or similar, the same kind of debugging information
19663 must have been output when building the precompiled header. However,
19664 a precompiled header built using @option{-g} can be used in a compilation
19665 when no debugging information is being output.
19667 @item The same @option{-m} options must generally be used when building
19668 and using the precompiled header. @xref{Submodel Options},
19669 for any cases where this rule is relaxed.
19671 @item Each of the following options must be the same when building and using
19672 the precompiled header:
19674 @gccoptlist{-fexceptions}
19677 Some other command-line options starting with @option{-f},
19678 @option{-p}, or @option{-O} must be defined in the same way as when
19679 the precompiled header was generated. At present, it's not clear
19680 which options are safe to change and which are not; the safest choice
19681 is to use exactly the same options when generating and using the
19682 precompiled header. The following are known to be safe:
19684 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
19685 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
19686 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
19691 For all of these except the last, the compiler will automatically
19692 ignore the precompiled header if the conditions aren't met. If you
19693 find an option combination that doesn't work and doesn't cause the
19694 precompiled header to be ignored, please consider filing a bug report,
19697 If you do use differing options when generating and using the
19698 precompiled header, the actual behavior will be a mixture of the
19699 behavior for the options. For instance, if you use @option{-g} to
19700 generate the precompiled header but not when using it, you may or may
19701 not get debugging information for routines in the precompiled header.