1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.3 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
168 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wdelete-non-virtual-dtor -Wnarrowing -Wnoexcept @gol
200 -Wnon-virtual-dtor -Wreorder @gol
201 -Weffc++ -Wstrict-null-sentinel @gol
202 -Wno-non-template-friend -Wold-style-cast @gol
203 -Woverloaded-virtual -Wno-pmf-conversions @gol
206 @item Objective-C and Objective-C++ Language Options
207 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
208 Objective-C and Objective-C++ Dialects}.
209 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
210 -fgnu-runtime -fnext-runtime @gol
211 -fno-nil-receivers @gol
212 -fobjc-abi-version=@var{n} @gol
213 -fobjc-call-cxx-cdtors @gol
214 -fobjc-direct-dispatch @gol
215 -fobjc-exceptions @gol
218 -fobjc-std=objc1 @gol
219 -freplace-objc-classes @gol
222 -Wassign-intercept @gol
223 -Wno-protocol -Wselector @gol
224 -Wstrict-selector-match @gol
225 -Wundeclared-selector}
227 @item Language Independent Options
228 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
229 @gccoptlist{-fmessage-length=@var{n} @gol
230 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
231 -fno-diagnostics-show-option}
233 @item Warning Options
234 @xref{Warning Options,,Options to Request or Suppress Warnings}.
235 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
236 -pedantic-errors @gol
237 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
238 -Wno-attributes -Wno-builtin-macro-redefined @gol
239 -Wc++-compat -Wc++11-compat -Wcast-align -Wcast-qual @gol
240 -Wchar-subscripts -Wclobbered -Wcomment @gol
241 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
242 -Wno-deprecated-declarations -Wdisabled-optimization @gol
243 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
244 -Wno-endif-labels -Werror -Werror=* @gol
245 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
246 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
247 -Wformat-security -Wformat-y2k @gol
248 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
249 -Wignored-qualifiers @gol
250 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
251 -Winit-self -Winline -Wmaybe-uninitialized @gol
252 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
253 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
254 -Wlogical-op -Wlong-long @gol
255 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
256 -Wmissing-format-attribute -Wmissing-include-dirs @gol
258 -Wno-multichar -Wnonnull -Wno-overflow @gol
259 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
260 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
261 -Wpointer-arith -Wno-pointer-to-int-cast @gol
262 -Wredundant-decls @gol
263 -Wreturn-type -Wsequence-point -Wshadow @gol
264 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
265 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
266 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
267 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
268 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
269 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
270 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
271 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
272 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
273 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
274 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
275 -Wvariadic-macros -Wvector-operation-performance -Wvla
276 -Wvolatile-register-var -Wwrite-strings -Wzero-as-null-pointer-constant}
278 @item C and Objective-C-only Warning Options
279 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
280 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
281 -Wold-style-declaration -Wold-style-definition @gol
282 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
283 -Wdeclaration-after-statement -Wpointer-sign}
285 @item Debugging Options
286 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
287 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
288 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
289 -fdisable-ipa-@var{pass_name} @gol
290 -fdisable-rtl-@var{pass_name} @gol
291 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
292 -fdisable-tree-@var{pass_name} @gol
293 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
294 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
295 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
296 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
297 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
299 -fdump-statistics @gol
301 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
305 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-nrv -fdump-tree-vect @gol
315 -fdump-tree-sink @gol
316 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
318 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
319 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
320 -ftree-vectorizer-verbose=@var{n} @gol
321 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
322 -fdump-final-insns=@var{file} @gol
323 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
324 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
325 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
326 -fenable-@var{kind}-@var{pass} @gol
327 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
328 -fdebug-types-section @gol
329 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
330 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
331 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
332 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
333 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
334 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
335 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
336 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
337 -gvms -gxcoff -gxcoff+ @gol
338 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
339 -fdebug-prefix-map=@var{old}=@var{new} @gol
340 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
341 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
342 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
343 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
344 -print-prog-name=@var{program} -print-search-dirs -Q @gol
345 -print-sysroot -print-sysroot-headers-suffix @gol
346 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
348 @item Optimization Options
349 @xref{Optimize Options,,Options that Control Optimization}.
350 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
351 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
352 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
353 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
354 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
355 -fcompare-elim -fcprop-registers -fcrossjumping @gol
356 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
357 -fcx-limited-range @gol
358 -fdata-sections -fdce -fdce -fdelayed-branch @gol
359 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
360 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
361 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
362 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
363 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
364 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
365 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
366 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
367 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
368 -fira-algorithm=@var{algorithm} @gol
369 -fira-region=@var{region} @gol
370 -fira-loop-pressure -fno-ira-share-save-slots @gol
371 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
372 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
373 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
374 -floop-parallelize-all -flto -flto-compression-level @gol
375 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
376 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
377 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
378 -fno-default-inline @gol
379 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
380 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
381 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
382 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
383 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
384 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
385 -fprefetch-loop-arrays @gol
386 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
387 -fprofile-generate=@var{path} @gol
388 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
389 -freciprocal-math -free -fregmove -frename-registers -freorder-blocks @gol
390 -freorder-blocks-and-partition -freorder-functions @gol
391 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
392 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
393 -fsched-spec-load -fsched-spec-load-dangerous @gol
394 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
395 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
396 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
397 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
398 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
399 -fselective-scheduling -fselective-scheduling2 @gol
400 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
401 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
402 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
403 -fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol
404 -fthread-jumps -ftracer -ftree-bit-ccp @gol
405 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
406 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
407 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
408 -ftree-loop-if-convert-stores -ftree-loop-im @gol
409 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
410 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
411 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
412 -ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol
413 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
414 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
415 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
416 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
417 -fwhole-program -fwpa -fuse-linker-plugin @gol
418 --param @var{name}=@var{value}
419 -O -O0 -O1 -O2 -O3 -Os -Ofast}
421 @item Preprocessor Options
422 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
423 @gccoptlist{-A@var{question}=@var{answer} @gol
424 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
425 -C -dD -dI -dM -dN @gol
426 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
427 -idirafter @var{dir} @gol
428 -include @var{file} -imacros @var{file} @gol
429 -iprefix @var{file} -iwithprefix @var{dir} @gol
430 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
431 -imultilib @var{dir} -isysroot @var{dir} @gol
432 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
433 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
434 -remap -trigraphs -undef -U@var{macro} @gol
435 -Wp,@var{option} -Xpreprocessor @var{option}}
437 @item Assembler Option
438 @xref{Assembler Options,,Passing Options to the Assembler}.
439 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
442 @xref{Link Options,,Options for Linking}.
443 @gccoptlist{@var{object-file-name} -l@var{library} @gol
444 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
445 -s -static -static-libgcc -static-libstdc++ -shared @gol
446 -shared-libgcc -symbolic @gol
447 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
450 @item Directory Options
451 @xref{Directory Options,,Options for Directory Search}.
452 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
453 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
456 @item Machine Dependent Options
457 @xref{Submodel Options,,Hardware Models and Configurations}.
458 @c This list is ordered alphanumerically by subsection name.
459 @c Try and put the significant identifier (CPU or system) first,
460 @c so users have a clue at guessing where the ones they want will be.
462 @emph{Adapteva Epiphany Options}
463 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
464 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
465 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
466 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
467 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
468 -msplit-vecmove-early -m1reg-@var{reg}}
471 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
472 -mabi=@var{name} @gol
473 -mapcs-stack-check -mno-apcs-stack-check @gol
474 -mapcs-float -mno-apcs-float @gol
475 -mapcs-reentrant -mno-apcs-reentrant @gol
476 -msched-prolog -mno-sched-prolog @gol
477 -mlittle-endian -mbig-endian -mwords-little-endian @gol
478 -mfloat-abi=@var{name} -mfpe @gol
479 -mfp16-format=@var{name}
480 -mthumb-interwork -mno-thumb-interwork @gol
481 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
482 -mstructure-size-boundary=@var{n} @gol
483 -mabort-on-noreturn @gol
484 -mlong-calls -mno-long-calls @gol
485 -msingle-pic-base -mno-single-pic-base @gol
486 -mpic-register=@var{reg} @gol
487 -mnop-fun-dllimport @gol
488 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
489 -mpoke-function-name @gol
491 -mtpcs-frame -mtpcs-leaf-frame @gol
492 -mcaller-super-interworking -mcallee-super-interworking @gol
493 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
494 -mword-relocations @gol
495 -mfix-cortex-m3-ldrd}
498 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
499 -mcall-prologues -mtiny-stack -mint8 -mstrict-X}
501 @emph{Blackfin Options}
502 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
503 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
504 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
505 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
506 -mno-id-shared-library -mshared-library-id=@var{n} @gol
507 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
508 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
509 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
513 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
514 -msim -msdata=@var{sdata-type}}
517 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
518 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
519 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
520 -mstack-align -mdata-align -mconst-align @gol
521 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
522 -melf -maout -melinux -mlinux -sim -sim2 @gol
523 -mmul-bug-workaround -mno-mul-bug-workaround}
525 @emph{Darwin Options}
526 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
527 -arch_only -bind_at_load -bundle -bundle_loader @gol
528 -client_name -compatibility_version -current_version @gol
530 -dependency-file -dylib_file -dylinker_install_name @gol
531 -dynamic -dynamiclib -exported_symbols_list @gol
532 -filelist -flat_namespace -force_cpusubtype_ALL @gol
533 -force_flat_namespace -headerpad_max_install_names @gol
535 -image_base -init -install_name -keep_private_externs @gol
536 -multi_module -multiply_defined -multiply_defined_unused @gol
537 -noall_load -no_dead_strip_inits_and_terms @gol
538 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
539 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
540 -private_bundle -read_only_relocs -sectalign @gol
541 -sectobjectsymbols -whyload -seg1addr @gol
542 -sectcreate -sectobjectsymbols -sectorder @gol
543 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
544 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
545 -segprot -segs_read_only_addr -segs_read_write_addr @gol
546 -single_module -static -sub_library -sub_umbrella @gol
547 -twolevel_namespace -umbrella -undefined @gol
548 -unexported_symbols_list -weak_reference_mismatches @gol
549 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
550 -mkernel -mone-byte-bool}
552 @emph{DEC Alpha Options}
553 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
554 -mieee -mieee-with-inexact -mieee-conformant @gol
555 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
556 -mtrap-precision=@var{mode} -mbuild-constants @gol
557 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
558 -mbwx -mmax -mfix -mcix @gol
559 -mfloat-vax -mfloat-ieee @gol
560 -mexplicit-relocs -msmall-data -mlarge-data @gol
561 -msmall-text -mlarge-text @gol
562 -mmemory-latency=@var{time}}
564 @emph{DEC Alpha/VMS Options}
565 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
568 @gccoptlist{-msmall-model -mno-lsim}
571 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
572 -mhard-float -msoft-float @gol
573 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
574 -mdouble -mno-double @gol
575 -mmedia -mno-media -mmuladd -mno-muladd @gol
576 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
577 -mlinked-fp -mlong-calls -malign-labels @gol
578 -mlibrary-pic -macc-4 -macc-8 @gol
579 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
580 -moptimize-membar -mno-optimize-membar @gol
581 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
582 -mvliw-branch -mno-vliw-branch @gol
583 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
584 -mno-nested-cond-exec -mtomcat-stats @gol
588 @emph{GNU/Linux Options}
589 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
590 -tno-android-cc -tno-android-ld}
592 @emph{H8/300 Options}
593 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
596 @gccoptlist{-march=@var{architecture-type} @gol
597 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
598 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
599 -mfixed-range=@var{register-range} @gol
600 -mjump-in-delay -mlinker-opt -mlong-calls @gol
601 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
602 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
603 -mno-jump-in-delay -mno-long-load-store @gol
604 -mno-portable-runtime -mno-soft-float @gol
605 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
606 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
607 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
608 -munix=@var{unix-std} -nolibdld -static -threads}
610 @emph{i386 and x86-64 Options}
611 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
612 -mfpmath=@var{unit} @gol
613 -masm=@var{dialect} -mno-fancy-math-387 @gol
614 -mno-fp-ret-in-387 -msoft-float @gol
615 -mno-wide-multiply -mrtd -malign-double @gol
616 -mpreferred-stack-boundary=@var{num} @gol
617 -mincoming-stack-boundary=@var{num} @gol
618 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
619 -mrecip -mrecip=@var{opt} @gol
621 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
622 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
623 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
624 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
625 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
626 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
627 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
628 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
629 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
630 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
631 -mcmodel=@var{code-model} -mabi=@var{name} @gol
632 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
633 -msse2avx -mfentry -m8bit-idiv @gol
634 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
636 @emph{i386 and x86-64 Windows Options}
637 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
638 -mnop-fun-dllimport -mthread @gol
639 -municode -mwin32 -mwindows -fno-set-stack-executable}
642 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
643 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
644 -mconstant-gp -mauto-pic -mfused-madd @gol
645 -minline-float-divide-min-latency @gol
646 -minline-float-divide-max-throughput @gol
647 -mno-inline-float-divide @gol
648 -minline-int-divide-min-latency @gol
649 -minline-int-divide-max-throughput @gol
650 -mno-inline-int-divide @gol
651 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
652 -mno-inline-sqrt @gol
653 -mdwarf2-asm -mearly-stop-bits @gol
654 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
655 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
656 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
657 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
658 -msched-spec-ldc -msched-spec-control-ldc @gol
659 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
660 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
661 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
662 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
664 @emph{IA-64/VMS Options}
665 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
668 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
669 -msign-extend-enabled -muser-enabled}
671 @emph{M32R/D Options}
672 @gccoptlist{-m32r2 -m32rx -m32r @gol
674 -malign-loops -mno-align-loops @gol
675 -missue-rate=@var{number} @gol
676 -mbranch-cost=@var{number} @gol
677 -mmodel=@var{code-size-model-type} @gol
678 -msdata=@var{sdata-type} @gol
679 -mno-flush-func -mflush-func=@var{name} @gol
680 -mno-flush-trap -mflush-trap=@var{number} @gol
684 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
686 @emph{M680x0 Options}
687 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
688 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
689 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
690 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
691 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
692 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
693 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
694 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
698 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
699 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
700 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
701 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
702 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
705 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
706 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
707 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
708 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
711 @emph{MicroBlaze Options}
712 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
713 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
714 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
715 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
716 -mxl-mode-@var{app-model}}
719 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
720 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
721 -mips64 -mips64r2 @gol
722 -mips16 -mno-mips16 -mflip-mips16 @gol
723 -minterlink-mips16 -mno-interlink-mips16 @gol
724 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
725 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
726 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
727 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
728 -mfpu=@var{fpu-type} @gol
729 -msmartmips -mno-smartmips @gol
730 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
731 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
732 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
733 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
734 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
735 -membedded-data -mno-embedded-data @gol
736 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
737 -mcode-readable=@var{setting} @gol
738 -msplit-addresses -mno-split-addresses @gol
739 -mexplicit-relocs -mno-explicit-relocs @gol
740 -mcheck-zero-division -mno-check-zero-division @gol
741 -mdivide-traps -mdivide-breaks @gol
742 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
743 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
744 -mfix-24k -mno-fix-24k @gol
745 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
746 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
747 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
748 -mflush-func=@var{func} -mno-flush-func @gol
749 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
750 -mfp-exceptions -mno-fp-exceptions @gol
751 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
752 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
755 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
756 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
757 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
758 -mno-base-addresses -msingle-exit -mno-single-exit}
760 @emph{MN10300 Options}
761 @gccoptlist{-mmult-bug -mno-mult-bug @gol
762 -mno-am33 -mam33 -mam33-2 -mam34 @gol
763 -mtune=@var{cpu-type} @gol
764 -mreturn-pointer-on-d0 @gol
765 -mno-crt0 -mrelax -mliw -msetlb}
767 @emph{PDP-11 Options}
768 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
769 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
770 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
771 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
772 -mbranch-expensive -mbranch-cheap @gol
773 -munix-asm -mdec-asm}
775 @emph{picoChip Options}
776 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
777 -msymbol-as-address -mno-inefficient-warnings}
779 @emph{PowerPC Options}
780 See RS/6000 and PowerPC Options.
783 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78}
785 @emph{RS/6000 and PowerPC Options}
786 @gccoptlist{-mcpu=@var{cpu-type} @gol
787 -mtune=@var{cpu-type} @gol
788 -mcmodel=@var{code-model} @gol
789 -mpower -mno-power -mpower2 -mno-power2 @gol
790 -mpowerpc -mpowerpc64 -mno-powerpc @gol
791 -maltivec -mno-altivec @gol
792 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
793 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
794 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
795 -mfprnd -mno-fprnd @gol
796 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
797 -mnew-mnemonics -mold-mnemonics @gol
798 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
799 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
800 -malign-power -malign-natural @gol
801 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
802 -msingle-float -mdouble-float -msimple-fpu @gol
803 -mstring -mno-string -mupdate -mno-update @gol
804 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
805 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
806 -mstrict-align -mno-strict-align -mrelocatable @gol
807 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
808 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
809 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
810 -mprioritize-restricted-insns=@var{priority} @gol
811 -msched-costly-dep=@var{dependence_type} @gol
812 -minsert-sched-nops=@var{scheme} @gol
813 -mcall-sysv -mcall-netbsd @gol
814 -maix-struct-return -msvr4-struct-return @gol
815 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
816 -mblock-move-inline-limit=@var{num} @gol
817 -misel -mno-isel @gol
818 -misel=yes -misel=no @gol
820 -mspe=yes -mspe=no @gol
822 -mgen-cell-microcode -mwarn-cell-microcode @gol
823 -mvrsave -mno-vrsave @gol
824 -mmulhw -mno-mulhw @gol
825 -mdlmzb -mno-dlmzb @gol
826 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
827 -mprototype -mno-prototype @gol
828 -msim -mmvme -mads -myellowknife -memb -msdata @gol
829 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
830 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
831 -mno-recip-precision @gol
832 -mveclibabi=@var{type} -mfriz -mno-friz @gol
833 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
834 -msave-toc-indirect -mno-save-toc-indirect}
837 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
839 -mbig-endian-data -mlittle-endian-data @gol
842 -mas100-syntax -mno-as100-syntax@gol
844 -mmax-constant-size=@gol
847 -msave-acc-in-interrupts}
849 @emph{S/390 and zSeries Options}
850 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
851 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
852 -mlong-double-64 -mlong-double-128 @gol
853 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
854 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
855 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
856 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
857 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
860 @gccoptlist{-meb -mel @gol
864 -mscore5 -mscore5u -mscore7 -mscore7d}
867 @gccoptlist{-m1 -m2 -m2e @gol
868 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
870 -m4-nofpu -m4-single-only -m4-single -m4 @gol
871 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
872 -m5-64media -m5-64media-nofpu @gol
873 -m5-32media -m5-32media-nofpu @gol
874 -m5-compact -m5-compact-nofpu @gol
875 -mb -ml -mdalign -mrelax @gol
876 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
877 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
878 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
879 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
880 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
881 -maccumulate-outgoing-args -minvalid-symbols -msoft-atomic}
883 @emph{Solaris 2 Options}
884 @gccoptlist{-mimpure-text -mno-impure-text @gol
888 @gccoptlist{-mcpu=@var{cpu-type} @gol
889 -mtune=@var{cpu-type} @gol
890 -mcmodel=@var{code-model} @gol
891 -mmemory-model=@var{mem-model} @gol
892 -m32 -m64 -mapp-regs -mno-app-regs @gol
893 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
894 -mfpu -mno-fpu -mhard-float -msoft-float @gol
895 -mhard-quad-float -msoft-quad-float @gol
897 -mstack-bias -mno-stack-bias @gol
898 -munaligned-doubles -mno-unaligned-doubles @gol
899 -mv8plus -mno-v8plus -mvis -mno-vis @gol
900 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
901 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
905 @gccoptlist{-mwarn-reloc -merror-reloc @gol
906 -msafe-dma -munsafe-dma @gol
908 -msmall-mem -mlarge-mem -mstdmain @gol
909 -mfixed-range=@var{register-range} @gol
911 -maddress-space-conversion -mno-address-space-conversion @gol
912 -mcache-size=@var{cache-size} @gol
913 -matomic-updates -mno-atomic-updates}
915 @emph{System V Options}
916 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
919 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
920 -mprolog-function -mno-prolog-function -mspace @gol
921 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
922 -mapp-regs -mno-app-regs @gol
923 -mdisable-callt -mno-disable-callt @gol
926 -mv850e1 -mv850es @gol
931 @gccoptlist{-mg -mgnu -munix}
933 @emph{VxWorks Options}
934 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
935 -Xbind-lazy -Xbind-now}
937 @emph{x86-64 Options}
938 See i386 and x86-64 Options.
940 @emph{Xstormy16 Options}
943 @emph{Xtensa Options}
944 @gccoptlist{-mconst16 -mno-const16 @gol
945 -mfused-madd -mno-fused-madd @gol
947 -mserialize-volatile -mno-serialize-volatile @gol
948 -mtext-section-literals -mno-text-section-literals @gol
949 -mtarget-align -mno-target-align @gol
950 -mlongcalls -mno-longcalls}
952 @emph{zSeries Options}
953 See S/390 and zSeries Options.
955 @item Code Generation Options
956 @xref{Code Gen Options,,Options for Code Generation Conventions}.
957 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
958 -ffixed-@var{reg} -fexceptions @gol
959 -fnon-call-exceptions -funwind-tables @gol
960 -fasynchronous-unwind-tables @gol
961 -finhibit-size-directive -finstrument-functions @gol
962 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
963 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
964 -fno-common -fno-ident @gol
965 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
966 -fno-jump-tables @gol
967 -frecord-gcc-switches @gol
968 -freg-struct-return -fshort-enums @gol
969 -fshort-double -fshort-wchar @gol
970 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
971 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
972 -fno-stack-limit -fsplit-stack @gol
973 -fleading-underscore -ftls-model=@var{model} @gol
974 -ftrapv -fwrapv -fbounds-check @gol
975 -fvisibility -fstrict-volatile-bitfields}
979 * Overall Options:: Controlling the kind of output:
980 an executable, object files, assembler files,
981 or preprocessed source.
982 * C Dialect Options:: Controlling the variant of C language compiled.
983 * C++ Dialect Options:: Variations on C++.
984 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
986 * Language Independent Options:: Controlling how diagnostics should be
988 * Warning Options:: How picky should the compiler be?
989 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
990 * Optimize Options:: How much optimization?
991 * Preprocessor Options:: Controlling header files and macro definitions.
992 Also, getting dependency information for Make.
993 * Assembler Options:: Passing options to the assembler.
994 * Link Options:: Specifying libraries and so on.
995 * Directory Options:: Where to find header files and libraries.
996 Where to find the compiler executable files.
997 * Spec Files:: How to pass switches to sub-processes.
998 * Target Options:: Running a cross-compiler, or an old version of GCC.
1001 @node Overall Options
1002 @section Options Controlling the Kind of Output
1004 Compilation can involve up to four stages: preprocessing, compilation
1005 proper, assembly and linking, always in that order. GCC is capable of
1006 preprocessing and compiling several files either into several
1007 assembler input files, or into one assembler input file; then each
1008 assembler input file produces an object file, and linking combines all
1009 the object files (those newly compiled, and those specified as input)
1010 into an executable file.
1012 @cindex file name suffix
1013 For any given input file, the file name suffix determines what kind of
1014 compilation is done:
1018 C source code which must be preprocessed.
1021 C source code which should not be preprocessed.
1024 C++ source code which should not be preprocessed.
1027 Objective-C source code. Note that you must link with the @file{libobjc}
1028 library to make an Objective-C program work.
1031 Objective-C source code which should not be preprocessed.
1035 Objective-C++ source code. Note that you must link with the @file{libobjc}
1036 library to make an Objective-C++ program work. Note that @samp{.M} refers
1037 to a literal capital M@.
1039 @item @var{file}.mii
1040 Objective-C++ source code which should not be preprocessed.
1043 C, C++, Objective-C or Objective-C++ header file to be turned into a
1044 precompiled header (default), or C, C++ header file to be turned into an
1045 Ada spec (via the @option{-fdump-ada-spec} switch).
1048 @itemx @var{file}.cp
1049 @itemx @var{file}.cxx
1050 @itemx @var{file}.cpp
1051 @itemx @var{file}.CPP
1052 @itemx @var{file}.c++
1054 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1055 the last two letters must both be literally @samp{x}. Likewise,
1056 @samp{.C} refers to a literal capital C@.
1060 Objective-C++ source code which must be preprocessed.
1062 @item @var{file}.mii
1063 Objective-C++ source code which should not be preprocessed.
1067 @itemx @var{file}.hp
1068 @itemx @var{file}.hxx
1069 @itemx @var{file}.hpp
1070 @itemx @var{file}.HPP
1071 @itemx @var{file}.h++
1072 @itemx @var{file}.tcc
1073 C++ header file to be turned into a precompiled header or Ada spec.
1076 @itemx @var{file}.for
1077 @itemx @var{file}.ftn
1078 Fixed form Fortran source code which should not be preprocessed.
1081 @itemx @var{file}.FOR
1082 @itemx @var{file}.fpp
1083 @itemx @var{file}.FPP
1084 @itemx @var{file}.FTN
1085 Fixed form Fortran source code which must be preprocessed (with the traditional
1088 @item @var{file}.f90
1089 @itemx @var{file}.f95
1090 @itemx @var{file}.f03
1091 @itemx @var{file}.f08
1092 Free form Fortran source code which should not be preprocessed.
1094 @item @var{file}.F90
1095 @itemx @var{file}.F95
1096 @itemx @var{file}.F03
1097 @itemx @var{file}.F08
1098 Free form Fortran source code which must be preprocessed (with the
1099 traditional preprocessor).
1104 @c FIXME: Descriptions of Java file types.
1110 @item @var{file}.ads
1111 Ada source code file which contains a library unit declaration (a
1112 declaration of a package, subprogram, or generic, or a generic
1113 instantiation), or a library unit renaming declaration (a package,
1114 generic, or subprogram renaming declaration). Such files are also
1117 @item @var{file}.adb
1118 Ada source code file containing a library unit body (a subprogram or
1119 package body). Such files are also called @dfn{bodies}.
1121 @c GCC also knows about some suffixes for languages not yet included:
1132 @itemx @var{file}.sx
1133 Assembler code which must be preprocessed.
1136 An object file to be fed straight into linking.
1137 Any file name with no recognized suffix is treated this way.
1141 You can specify the input language explicitly with the @option{-x} option:
1144 @item -x @var{language}
1145 Specify explicitly the @var{language} for the following input files
1146 (rather than letting the compiler choose a default based on the file
1147 name suffix). This option applies to all following input files until
1148 the next @option{-x} option. Possible values for @var{language} are:
1150 c c-header cpp-output
1151 c++ c++-header c++-cpp-output
1152 objective-c objective-c-header objective-c-cpp-output
1153 objective-c++ objective-c++-header objective-c++-cpp-output
1154 assembler assembler-with-cpp
1156 f77 f77-cpp-input f95 f95-cpp-input
1162 Turn off any specification of a language, so that subsequent files are
1163 handled according to their file name suffixes (as they are if @option{-x}
1164 has not been used at all).
1166 @item -pass-exit-codes
1167 @opindex pass-exit-codes
1168 Normally the @command{gcc} program will exit with the code of 1 if any
1169 phase of the compiler returns a non-success return code. If you specify
1170 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1171 numerically highest error produced by any phase that returned an error
1172 indication. The C, C++, and Fortran frontends return 4, if an internal
1173 compiler error is encountered.
1176 If you only want some of the stages of compilation, you can use
1177 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1178 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1179 @command{gcc} is to stop. Note that some combinations (for example,
1180 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1185 Compile or assemble the source files, but do not link. The linking
1186 stage simply is not done. The ultimate output is in the form of an
1187 object file for each source file.
1189 By default, the object file name for a source file is made by replacing
1190 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1192 Unrecognized input files, not requiring compilation or assembly, are
1197 Stop after the stage of compilation proper; do not assemble. The output
1198 is in the form of an assembler code file for each non-assembler input
1201 By default, the assembler file name for a source file is made by
1202 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1204 Input files that don't require compilation are ignored.
1208 Stop after the preprocessing stage; do not run the compiler proper. The
1209 output is in the form of preprocessed source code, which is sent to the
1212 Input files which don't require preprocessing are ignored.
1214 @cindex output file option
1217 Place output in file @var{file}. This applies regardless to whatever
1218 sort of output is being produced, whether it be an executable file,
1219 an object file, an assembler file or preprocessed C code.
1221 If @option{-o} is not specified, the default is to put an executable
1222 file in @file{a.out}, the object file for
1223 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1224 assembler file in @file{@var{source}.s}, a precompiled header file in
1225 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1230 Print (on standard error output) the commands executed to run the stages
1231 of compilation. Also print the version number of the compiler driver
1232 program and of the preprocessor and the compiler proper.
1236 Like @option{-v} except the commands are not executed and arguments
1237 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1238 This is useful for shell scripts to capture the driver-generated command lines.
1242 Use pipes rather than temporary files for communication between the
1243 various stages of compilation. This fails to work on some systems where
1244 the assembler is unable to read from a pipe; but the GNU assembler has
1249 Print (on the standard output) a description of the command line options
1250 understood by @command{gcc}. If the @option{-v} option is also specified
1251 then @option{--help} will also be passed on to the various processes
1252 invoked by @command{gcc}, so that they can display the command line options
1253 they accept. If the @option{-Wextra} option has also been specified
1254 (prior to the @option{--help} option), then command line options which
1255 have no documentation associated with them will also be displayed.
1258 @opindex target-help
1259 Print (on the standard output) a description of target-specific command
1260 line options for each tool. For some targets extra target-specific
1261 information may also be printed.
1263 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1264 Print (on the standard output) a description of the command line
1265 options understood by the compiler that fit into all specified classes
1266 and qualifiers. These are the supported classes:
1269 @item @samp{optimizers}
1270 This will display all of the optimization options supported by the
1273 @item @samp{warnings}
1274 This will display all of the options controlling warning messages
1275 produced by the compiler.
1278 This will display target-specific options. Unlike the
1279 @option{--target-help} option however, target-specific options of the
1280 linker and assembler will not be displayed. This is because those
1281 tools do not currently support the extended @option{--help=} syntax.
1284 This will display the values recognized by the @option{--param}
1287 @item @var{language}
1288 This will display the options supported for @var{language}, where
1289 @var{language} is the name of one of the languages supported in this
1293 This will display the options that are common to all languages.
1296 These are the supported qualifiers:
1299 @item @samp{undocumented}
1300 Display only those options which are undocumented.
1303 Display options which take an argument that appears after an equal
1304 sign in the same continuous piece of text, such as:
1305 @samp{--help=target}.
1307 @item @samp{separate}
1308 Display options which take an argument that appears as a separate word
1309 following the original option, such as: @samp{-o output-file}.
1312 Thus for example to display all the undocumented target-specific
1313 switches supported by the compiler the following can be used:
1316 --help=target,undocumented
1319 The sense of a qualifier can be inverted by prefixing it with the
1320 @samp{^} character, so for example to display all binary warning
1321 options (i.e., ones that are either on or off and that do not take an
1322 argument), which have a description the following can be used:
1325 --help=warnings,^joined,^undocumented
1328 The argument to @option{--help=} should not consist solely of inverted
1331 Combining several classes is possible, although this usually
1332 restricts the output by so much that there is nothing to display. One
1333 case where it does work however is when one of the classes is
1334 @var{target}. So for example to display all the target-specific
1335 optimization options the following can be used:
1338 --help=target,optimizers
1341 The @option{--help=} option can be repeated on the command line. Each
1342 successive use will display its requested class of options, skipping
1343 those that have already been displayed.
1345 If the @option{-Q} option appears on the command line before the
1346 @option{--help=} option, then the descriptive text displayed by
1347 @option{--help=} is changed. Instead of describing the displayed
1348 options, an indication is given as to whether the option is enabled,
1349 disabled or set to a specific value (assuming that the compiler
1350 knows this at the point where the @option{--help=} option is used).
1352 Here is a truncated example from the ARM port of @command{gcc}:
1355 % gcc -Q -mabi=2 --help=target -c
1356 The following options are target specific:
1358 -mabort-on-noreturn [disabled]
1362 The output is sensitive to the effects of previous command line
1363 options, so for example it is possible to find out which optimizations
1364 are enabled at @option{-O2} by using:
1367 -Q -O2 --help=optimizers
1370 Alternatively you can discover which binary optimizations are enabled
1371 by @option{-O3} by using:
1374 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1375 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1376 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1379 @item -no-canonical-prefixes
1380 @opindex no-canonical-prefixes
1381 Do not expand any symbolic links, resolve references to @samp{/../}
1382 or @samp{/./}, or make the path absolute when generating a relative
1387 Display the version number and copyrights of the invoked GCC@.
1391 Invoke all subcommands under a wrapper program. The name of the
1392 wrapper program and its parameters are passed as a comma separated
1396 gcc -c t.c -wrapper gdb,--args
1399 This will invoke all subprograms of @command{gcc} under
1400 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1401 @samp{gdb --args cc1 @dots{}}.
1403 @item -fplugin=@var{name}.so
1404 Load the plugin code in file @var{name}.so, assumed to be a
1405 shared object to be dlopen'd by the compiler. The base name of
1406 the shared object file is used to identify the plugin for the
1407 purposes of argument parsing (See
1408 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1409 Each plugin should define the callback functions specified in the
1412 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1413 Define an argument called @var{key} with a value of @var{value}
1414 for the plugin called @var{name}.
1416 @item -fdump-ada-spec@r{[}-slim@r{]}
1417 For C and C++ source and include files, generate corresponding Ada
1418 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1419 GNAT User's Guide}, which provides detailed documentation on this feature.
1421 @item -fdump-go-spec=@var{file}
1422 For input files in any language, generate corresponding Go
1423 declarations in @var{file}. This generates Go @code{const},
1424 @code{type}, @code{var}, and @code{func} declarations which may be a
1425 useful way to start writing a Go interface to code written in some
1428 @include @value{srcdir}/../libiberty/at-file.texi
1432 @section Compiling C++ Programs
1434 @cindex suffixes for C++ source
1435 @cindex C++ source file suffixes
1436 C++ source files conventionally use one of the suffixes @samp{.C},
1437 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1438 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1439 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1440 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1441 files with these names and compiles them as C++ programs even if you
1442 call the compiler the same way as for compiling C programs (usually
1443 with the name @command{gcc}).
1447 However, the use of @command{gcc} does not add the C++ library.
1448 @command{g++} is a program that calls GCC and treats @samp{.c},
1449 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1450 files unless @option{-x} is used, and automatically specifies linking
1451 against the C++ library. This program is also useful when
1452 precompiling a C header file with a @samp{.h} extension for use in C++
1453 compilations. On many systems, @command{g++} is also installed with
1454 the name @command{c++}.
1456 @cindex invoking @command{g++}
1457 When you compile C++ programs, you may specify many of the same
1458 command-line options that you use for compiling programs in any
1459 language; or command-line options meaningful for C and related
1460 languages; or options that are meaningful only for C++ programs.
1461 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1462 explanations of options for languages related to C@.
1463 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1464 explanations of options that are meaningful only for C++ programs.
1466 @node C Dialect Options
1467 @section Options Controlling C Dialect
1468 @cindex dialect options
1469 @cindex language dialect options
1470 @cindex options, dialect
1472 The following options control the dialect of C (or languages derived
1473 from C, such as C++, Objective-C and Objective-C++) that the compiler
1477 @cindex ANSI support
1481 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1482 equivalent to @samp{-std=c++98}.
1484 This turns off certain features of GCC that are incompatible with ISO
1485 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1486 such as the @code{asm} and @code{typeof} keywords, and
1487 predefined macros such as @code{unix} and @code{vax} that identify the
1488 type of system you are using. It also enables the undesirable and
1489 rarely used ISO trigraph feature. For the C compiler,
1490 it disables recognition of C++ style @samp{//} comments as well as
1491 the @code{inline} keyword.
1493 The alternate keywords @code{__asm__}, @code{__extension__},
1494 @code{__inline__} and @code{__typeof__} continue to work despite
1495 @option{-ansi}. You would not want to use them in an ISO C program, of
1496 course, but it is useful to put them in header files that might be included
1497 in compilations done with @option{-ansi}. Alternate predefined macros
1498 such as @code{__unix__} and @code{__vax__} are also available, with or
1499 without @option{-ansi}.
1501 The @option{-ansi} option does not cause non-ISO programs to be
1502 rejected gratuitously. For that, @option{-pedantic} is required in
1503 addition to @option{-ansi}. @xref{Warning Options}.
1505 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1506 option is used. Some header files may notice this macro and refrain
1507 from declaring certain functions or defining certain macros that the
1508 ISO standard doesn't call for; this is to avoid interfering with any
1509 programs that might use these names for other things.
1511 Functions that would normally be built in but do not have semantics
1512 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1513 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1514 built-in functions provided by GCC}, for details of the functions
1519 Determine the language standard. @xref{Standards,,Language Standards
1520 Supported by GCC}, for details of these standard versions. This option
1521 is currently only supported when compiling C or C++.
1523 The compiler can accept several base standards, such as @samp{c90} or
1524 @samp{c++98}, and GNU dialects of those standards, such as
1525 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1526 compiler will accept all programs following that standard and those
1527 using GNU extensions that do not contradict it. For example,
1528 @samp{-std=c90} turns off certain features of GCC that are
1529 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1530 keywords, but not other GNU extensions that do not have a meaning in
1531 ISO C90, such as omitting the middle term of a @code{?:}
1532 expression. On the other hand, by specifying a GNU dialect of a
1533 standard, all features the compiler support are enabled, even when
1534 those features change the meaning of the base standard and some
1535 strict-conforming programs may be rejected. The particular standard
1536 is used by @option{-pedantic} to identify which features are GNU
1537 extensions given that version of the standard. For example
1538 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1539 comments, while @samp{-std=gnu99 -pedantic} would not.
1541 A value for this option must be provided; possible values are
1547 Support all ISO C90 programs (certain GNU extensions that conflict
1548 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1550 @item iso9899:199409
1551 ISO C90 as modified in amendment 1.
1557 ISO C99. Note that this standard is not yet fully supported; see
1558 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1559 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1564 ISO C11, the 2011 revision of the ISO C standard.
1565 Support is incomplete and experimental. The name @samp{c1x} is
1570 GNU dialect of ISO C90 (including some C99 features). This
1571 is the default for C code.
1575 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1576 this will become the default. The name @samp{gnu9x} is deprecated.
1580 GNU dialect of ISO C11. Support is incomplete and experimental. The
1581 name @samp{gnu1x} is deprecated.
1584 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1588 GNU dialect of @option{-std=c++98}. This is the default for
1592 The 2011 ISO C++ standard plus amendments. Support for C++11 is still
1593 experimental, and may change in incompatible ways in future releases.
1596 GNU dialect of @option{-std=c++11}. Support for C++11 is still
1597 experimental, and may change in incompatible ways in future releases.
1600 @item -fgnu89-inline
1601 @opindex fgnu89-inline
1602 The option @option{-fgnu89-inline} tells GCC to use the traditional
1603 GNU semantics for @code{inline} functions when in C99 mode.
1604 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1605 is accepted and ignored by GCC versions 4.1.3 up to but not including
1606 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1607 C99 mode. Using this option is roughly equivalent to adding the
1608 @code{gnu_inline} function attribute to all inline functions
1609 (@pxref{Function Attributes}).
1611 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1612 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1613 specifies the default behavior). This option was first supported in
1614 GCC 4.3. This option is not supported in @option{-std=c90} or
1615 @option{-std=gnu90} mode.
1617 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1618 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1619 in effect for @code{inline} functions. @xref{Common Predefined
1620 Macros,,,cpp,The C Preprocessor}.
1622 @item -aux-info @var{filename}
1624 Output to the given filename prototyped declarations for all functions
1625 declared and/or defined in a translation unit, including those in header
1626 files. This option is silently ignored in any language other than C@.
1628 Besides declarations, the file indicates, in comments, the origin of
1629 each declaration (source file and line), whether the declaration was
1630 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1631 @samp{O} for old, respectively, in the first character after the line
1632 number and the colon), and whether it came from a declaration or a
1633 definition (@samp{C} or @samp{F}, respectively, in the following
1634 character). In the case of function definitions, a K&R-style list of
1635 arguments followed by their declarations is also provided, inside
1636 comments, after the declaration.
1638 @item -fallow-parameterless-variadic-functions
1639 Accept variadic functions without named parameters.
1641 Although it is possible to define such a function, this is not very
1642 useful as it is not possible to read the arguments. This is only
1643 supported for C as this construct is allowed by C++.
1647 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1648 keyword, so that code can use these words as identifiers. You can use
1649 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1650 instead. @option{-ansi} implies @option{-fno-asm}.
1652 In C++, this switch only affects the @code{typeof} keyword, since
1653 @code{asm} and @code{inline} are standard keywords. You may want to
1654 use the @option{-fno-gnu-keywords} flag instead, which has the same
1655 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1656 switch only affects the @code{asm} and @code{typeof} keywords, since
1657 @code{inline} is a standard keyword in ISO C99.
1660 @itemx -fno-builtin-@var{function}
1661 @opindex fno-builtin
1662 @cindex built-in functions
1663 Don't recognize built-in functions that do not begin with
1664 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1665 functions provided by GCC}, for details of the functions affected,
1666 including those which are not built-in functions when @option{-ansi} or
1667 @option{-std} options for strict ISO C conformance are used because they
1668 do not have an ISO standard meaning.
1670 GCC normally generates special code to handle certain built-in functions
1671 more efficiently; for instance, calls to @code{alloca} may become single
1672 instructions that adjust the stack directly, and calls to @code{memcpy}
1673 may become inline copy loops. The resulting code is often both smaller
1674 and faster, but since the function calls no longer appear as such, you
1675 cannot set a breakpoint on those calls, nor can you change the behavior
1676 of the functions by linking with a different library. In addition,
1677 when a function is recognized as a built-in function, GCC may use
1678 information about that function to warn about problems with calls to
1679 that function, or to generate more efficient code, even if the
1680 resulting code still contains calls to that function. For example,
1681 warnings are given with @option{-Wformat} for bad calls to
1682 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1683 known not to modify global memory.
1685 With the @option{-fno-builtin-@var{function}} option
1686 only the built-in function @var{function} is
1687 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1688 function is named that is not built-in in this version of GCC, this
1689 option is ignored. There is no corresponding
1690 @option{-fbuiltin-@var{function}} option; if you wish to enable
1691 built-in functions selectively when using @option{-fno-builtin} or
1692 @option{-ffreestanding}, you may define macros such as:
1695 #define abs(n) __builtin_abs ((n))
1696 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1701 @cindex hosted environment
1703 Assert that compilation takes place in a hosted environment. This implies
1704 @option{-fbuiltin}. A hosted environment is one in which the
1705 entire standard library is available, and in which @code{main} has a return
1706 type of @code{int}. Examples are nearly everything except a kernel.
1707 This is equivalent to @option{-fno-freestanding}.
1709 @item -ffreestanding
1710 @opindex ffreestanding
1711 @cindex hosted environment
1713 Assert that compilation takes place in a freestanding environment. This
1714 implies @option{-fno-builtin}. A freestanding environment
1715 is one in which the standard library may not exist, and program startup may
1716 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1717 This is equivalent to @option{-fno-hosted}.
1719 @xref{Standards,,Language Standards Supported by GCC}, for details of
1720 freestanding and hosted environments.
1724 @cindex OpenMP parallel
1725 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1726 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1727 compiler generates parallel code according to the OpenMP Application
1728 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1729 implies @option{-pthread}, and thus is only supported on targets that
1730 have support for @option{-pthread}.
1734 When the option @option{-fgnu-tm} is specified, the compiler will
1735 generate code for the Linux variant of Intel's current Transactional
1736 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1737 an experimental feature whose interface may change in future versions
1738 of GCC, as the official specification changes. Please note that not
1739 all architectures are supported for this feature.
1741 For more information on GCC's support for transactional memory,
1742 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1743 Transactional Memory Library}.
1745 Note that the transactional memory feature is not supported with
1746 non-call exceptions (@option{-fnon-call-exceptions}).
1748 @item -fms-extensions
1749 @opindex fms-extensions
1750 Accept some non-standard constructs used in Microsoft header files.
1752 In C++ code, this allows member names in structures to be similar
1753 to previous types declarations.
1762 Some cases of unnamed fields in structures and unions are only
1763 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1764 fields within structs/unions}, for details.
1766 @item -fplan9-extensions
1767 Accept some non-standard constructs used in Plan 9 code.
1769 This enables @option{-fms-extensions}, permits passing pointers to
1770 structures with anonymous fields to functions which expect pointers to
1771 elements of the type of the field, and permits referring to anonymous
1772 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1773 struct/union fields within structs/unions}, for details. This is only
1774 supported for C, not C++.
1778 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1779 options for strict ISO C conformance) implies @option{-trigraphs}.
1781 @item -no-integrated-cpp
1782 @opindex no-integrated-cpp
1783 Performs a compilation in two passes: preprocessing and compiling. This
1784 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1785 @option{-B} option. The user supplied compilation step can then add in
1786 an additional preprocessing step after normal preprocessing but before
1787 compiling. The default is to use the integrated cpp (internal cpp)
1789 The semantics of this option will change if "cc1", "cc1plus", and
1790 "cc1obj" are merged.
1792 @cindex traditional C language
1793 @cindex C language, traditional
1795 @itemx -traditional-cpp
1796 @opindex traditional-cpp
1797 @opindex traditional
1798 Formerly, these options caused GCC to attempt to emulate a pre-standard
1799 C compiler. They are now only supported with the @option{-E} switch.
1800 The preprocessor continues to support a pre-standard mode. See the GNU
1801 CPP manual for details.
1803 @item -fcond-mismatch
1804 @opindex fcond-mismatch
1805 Allow conditional expressions with mismatched types in the second and
1806 third arguments. The value of such an expression is void. This option
1807 is not supported for C++.
1809 @item -flax-vector-conversions
1810 @opindex flax-vector-conversions
1811 Allow implicit conversions between vectors with differing numbers of
1812 elements and/or incompatible element types. This option should not be
1815 @item -funsigned-char
1816 @opindex funsigned-char
1817 Let the type @code{char} be unsigned, like @code{unsigned char}.
1819 Each kind of machine has a default for what @code{char} should
1820 be. It is either like @code{unsigned char} by default or like
1821 @code{signed char} by default.
1823 Ideally, a portable program should always use @code{signed char} or
1824 @code{unsigned char} when it depends on the signedness of an object.
1825 But many programs have been written to use plain @code{char} and
1826 expect it to be signed, or expect it to be unsigned, depending on the
1827 machines they were written for. This option, and its inverse, let you
1828 make such a program work with the opposite default.
1830 The type @code{char} is always a distinct type from each of
1831 @code{signed char} or @code{unsigned char}, even though its behavior
1832 is always just like one of those two.
1835 @opindex fsigned-char
1836 Let the type @code{char} be signed, like @code{signed char}.
1838 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1839 the negative form of @option{-funsigned-char}. Likewise, the option
1840 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1842 @item -fsigned-bitfields
1843 @itemx -funsigned-bitfields
1844 @itemx -fno-signed-bitfields
1845 @itemx -fno-unsigned-bitfields
1846 @opindex fsigned-bitfields
1847 @opindex funsigned-bitfields
1848 @opindex fno-signed-bitfields
1849 @opindex fno-unsigned-bitfields
1850 These options control whether a bit-field is signed or unsigned, when the
1851 declaration does not use either @code{signed} or @code{unsigned}. By
1852 default, such a bit-field is signed, because this is consistent: the
1853 basic integer types such as @code{int} are signed types.
1856 @node C++ Dialect Options
1857 @section Options Controlling C++ Dialect
1859 @cindex compiler options, C++
1860 @cindex C++ options, command line
1861 @cindex options, C++
1862 This section describes the command-line options that are only meaningful
1863 for C++ programs; but you can also use most of the GNU compiler options
1864 regardless of what language your program is in. For example, you
1865 might compile a file @code{firstClass.C} like this:
1868 g++ -g -frepo -O -c firstClass.C
1872 In this example, only @option{-frepo} is an option meant
1873 only for C++ programs; you can use the other options with any
1874 language supported by GCC@.
1876 Here is a list of options that are @emph{only} for compiling C++ programs:
1880 @item -fabi-version=@var{n}
1881 @opindex fabi-version
1882 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1883 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1884 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1885 the version that conforms most closely to the C++ ABI specification.
1886 Therefore, the ABI obtained using version 0 will change as ABI bugs
1889 The default is version 2.
1891 Version 3 corrects an error in mangling a constant address as a
1894 Version 4, which first appeared in G++ 4.5, implements a standard
1895 mangling for vector types.
1897 Version 5, which first appeared in G++ 4.6, corrects the mangling of
1898 attribute const/volatile on function pointer types, decltype of a
1899 plain decl, and use of a function parameter in the declaration of
1902 Version 6, which first appeared in G++ 4.7, corrects the promotion
1903 behavior of C++11 scoped enums and the mangling of template argument
1904 packs, const/static_cast, prefix ++ and --, and a class scope function
1905 used as a template argument.
1907 See also @option{-Wabi}.
1909 @item -fno-access-control
1910 @opindex fno-access-control
1911 Turn off all access checking. This switch is mainly useful for working
1912 around bugs in the access control code.
1916 Check that the pointer returned by @code{operator new} is non-null
1917 before attempting to modify the storage allocated. This check is
1918 normally unnecessary because the C++ standard specifies that
1919 @code{operator new} will only return @code{0} if it is declared
1920 @samp{throw()}, in which case the compiler will always check the
1921 return value even without this option. In all other cases, when
1922 @code{operator new} has a non-empty exception specification, memory
1923 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1924 @samp{new (nothrow)}.
1926 @item -fconserve-space
1927 @opindex fconserve-space
1928 Put uninitialized or runtime-initialized global variables into the
1929 common segment, as C does. This saves space in the executable at the
1930 cost of not diagnosing duplicate definitions. If you compile with this
1931 flag and your program mysteriously crashes after @code{main()} has
1932 completed, you may have an object that is being destroyed twice because
1933 two definitions were merged.
1935 This option is no longer useful on most targets, now that support has
1936 been added for putting variables into BSS without making them common.
1938 @item -fconstexpr-depth=@var{n}
1939 @opindex fconstexpr-depth
1940 Set the maximum nested evaluation depth for C++11 constexpr functions
1941 to @var{n}. A limit is needed to detect endless recursion during
1942 constant expression evaluation. The minimum specified by the standard
1945 @item -fdeduce-init-list
1946 @opindex fdeduce-init-list
1947 Enable deduction of a template type parameter as
1948 std::initializer_list from a brace-enclosed initializer list, i.e.
1951 template <class T> auto forward(T t) -> decltype (realfn (t))
1958 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1962 This deduction was implemented as a possible extension to the
1963 originally proposed semantics for the C++11 standard, but was not part
1964 of the final standard, so it is disabled by default. This option is
1965 deprecated, and may be removed in a future version of G++.
1967 @item -ffriend-injection
1968 @opindex ffriend-injection
1969 Inject friend functions into the enclosing namespace, so that they are
1970 visible outside the scope of the class in which they are declared.
1971 Friend functions were documented to work this way in the old Annotated
1972 C++ Reference Manual, and versions of G++ before 4.1 always worked
1973 that way. However, in ISO C++ a friend function which is not declared
1974 in an enclosing scope can only be found using argument dependent
1975 lookup. This option causes friends to be injected as they were in
1978 This option is for compatibility, and may be removed in a future
1981 @item -fno-elide-constructors
1982 @opindex fno-elide-constructors
1983 The C++ standard allows an implementation to omit creating a temporary
1984 which is only used to initialize another object of the same type.
1985 Specifying this option disables that optimization, and forces G++ to
1986 call the copy constructor in all cases.
1988 @item -fno-enforce-eh-specs
1989 @opindex fno-enforce-eh-specs
1990 Don't generate code to check for violation of exception specifications
1991 at runtime. This option violates the C++ standard, but may be useful
1992 for reducing code size in production builds, much like defining
1993 @samp{NDEBUG}. This does not give user code permission to throw
1994 exceptions in violation of the exception specifications; the compiler
1995 will still optimize based on the specifications, so throwing an
1996 unexpected exception will result in undefined behavior.
1999 @itemx -fno-for-scope
2001 @opindex fno-for-scope
2002 If @option{-ffor-scope} is specified, the scope of variables declared in
2003 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2004 as specified by the C++ standard.
2005 If @option{-fno-for-scope} is specified, the scope of variables declared in
2006 a @i{for-init-statement} extends to the end of the enclosing scope,
2007 as was the case in old versions of G++, and other (traditional)
2008 implementations of C++.
2010 The default if neither flag is given to follow the standard,
2011 but to allow and give a warning for old-style code that would
2012 otherwise be invalid, or have different behavior.
2014 @item -fno-gnu-keywords
2015 @opindex fno-gnu-keywords
2016 Do not recognize @code{typeof} as a keyword, so that code can use this
2017 word as an identifier. You can use the keyword @code{__typeof__} instead.
2018 @option{-ansi} implies @option{-fno-gnu-keywords}.
2020 @item -fno-implicit-templates
2021 @opindex fno-implicit-templates
2022 Never emit code for non-inline templates which are instantiated
2023 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2024 @xref{Template Instantiation}, for more information.
2026 @item -fno-implicit-inline-templates
2027 @opindex fno-implicit-inline-templates
2028 Don't emit code for implicit instantiations of inline templates, either.
2029 The default is to handle inlines differently so that compiles with and
2030 without optimization will need the same set of explicit instantiations.
2032 @item -fno-implement-inlines
2033 @opindex fno-implement-inlines
2034 To save space, do not emit out-of-line copies of inline functions
2035 controlled by @samp{#pragma implementation}. This will cause linker
2036 errors if these functions are not inlined everywhere they are called.
2038 @item -fms-extensions
2039 @opindex fms-extensions
2040 Disable pedantic warnings about constructs used in MFC, such as implicit
2041 int and getting a pointer to member function via non-standard syntax.
2043 @item -fno-nonansi-builtins
2044 @opindex fno-nonansi-builtins
2045 Disable built-in declarations of functions that are not mandated by
2046 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2047 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2050 @opindex fnothrow-opt
2051 Treat a @code{throw()} exception specification as though it were a
2052 @code{noexcept} specification to reduce or eliminate the text size
2053 overhead relative to a function with no exception specification. If
2054 the function has local variables of types with non-trivial
2055 destructors, the exception specification will actually make the
2056 function smaller because the EH cleanups for those variables can be
2057 optimized away. The semantic effect is that an exception thrown out of
2058 a function with such an exception specification will result in a call
2059 to @code{terminate} rather than @code{unexpected}.
2061 @item -fno-operator-names
2062 @opindex fno-operator-names
2063 Do not treat the operator name keywords @code{and}, @code{bitand},
2064 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2065 synonyms as keywords.
2067 @item -fno-optional-diags
2068 @opindex fno-optional-diags
2069 Disable diagnostics that the standard says a compiler does not need to
2070 issue. Currently, the only such diagnostic issued by G++ is the one for
2071 a name having multiple meanings within a class.
2074 @opindex fpermissive
2075 Downgrade some diagnostics about nonconformant code from errors to
2076 warnings. Thus, using @option{-fpermissive} will allow some
2077 nonconforming code to compile.
2079 @item -fno-pretty-templates
2080 @opindex fno-pretty-templates
2081 When an error message refers to a specialization of a function
2082 template, the compiler will normally print the signature of the
2083 template followed by the template arguments and any typedefs or
2084 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2085 rather than @code{void f(int)}) so that it's clear which template is
2086 involved. When an error message refers to a specialization of a class
2087 template, the compiler will omit any template arguments which match
2088 the default template arguments for that template. If either of these
2089 behaviors make it harder to understand the error message rather than
2090 easier, using @option{-fno-pretty-templates} will disable them.
2094 Enable automatic template instantiation at link time. This option also
2095 implies @option{-fno-implicit-templates}. @xref{Template
2096 Instantiation}, for more information.
2100 Disable generation of information about every class with virtual
2101 functions for use by the C++ runtime type identification features
2102 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2103 of the language, you can save some space by using this flag. Note that
2104 exception handling uses the same information, but it will generate it as
2105 needed. The @samp{dynamic_cast} operator can still be used for casts that
2106 do not require runtime type information, i.e.@: casts to @code{void *} or to
2107 unambiguous base classes.
2111 Emit statistics about front-end processing at the end of the compilation.
2112 This information is generally only useful to the G++ development team.
2114 @item -fstrict-enums
2115 @opindex fstrict-enums
2116 Allow the compiler to optimize using the assumption that a value of
2117 enumeration type can only be one of the values of the enumeration (as
2118 defined in the C++ standard; basically, a value which can be
2119 represented in the minimum number of bits needed to represent all the
2120 enumerators). This assumption may not be valid if the program uses a
2121 cast to convert an arbitrary integer value to the enumeration type.
2123 @item -ftemplate-depth=@var{n}
2124 @opindex ftemplate-depth
2125 Set the maximum instantiation depth for template classes to @var{n}.
2126 A limit on the template instantiation depth is needed to detect
2127 endless recursions during template class instantiation. ANSI/ISO C++
2128 conforming programs must not rely on a maximum depth greater than 17
2129 (changed to 1024 in C++11). The default value is 900, as the compiler
2130 can run out of stack space before hitting 1024 in some situations.
2132 @item -fno-threadsafe-statics
2133 @opindex fno-threadsafe-statics
2134 Do not emit the extra code to use the routines specified in the C++
2135 ABI for thread-safe initialization of local statics. You can use this
2136 option to reduce code size slightly in code that doesn't need to be
2139 @item -fuse-cxa-atexit
2140 @opindex fuse-cxa-atexit
2141 Register destructors for objects with static storage duration with the
2142 @code{__cxa_atexit} function rather than the @code{atexit} function.
2143 This option is required for fully standards-compliant handling of static
2144 destructors, but will only work if your C library supports
2145 @code{__cxa_atexit}.
2147 @item -fno-use-cxa-get-exception-ptr
2148 @opindex fno-use-cxa-get-exception-ptr
2149 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2150 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2151 if the runtime routine is not available.
2153 @item -fvisibility-inlines-hidden
2154 @opindex fvisibility-inlines-hidden
2155 This switch declares that the user does not attempt to compare
2156 pointers to inline functions or methods where the addresses of the two functions
2157 were taken in different shared objects.
2159 The effect of this is that GCC may, effectively, mark inline methods with
2160 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2161 appear in the export table of a DSO and do not require a PLT indirection
2162 when used within the DSO@. Enabling this option can have a dramatic effect
2163 on load and link times of a DSO as it massively reduces the size of the
2164 dynamic export table when the library makes heavy use of templates.
2166 The behavior of this switch is not quite the same as marking the
2167 methods as hidden directly, because it does not affect static variables
2168 local to the function or cause the compiler to deduce that
2169 the function is defined in only one shared object.
2171 You may mark a method as having a visibility explicitly to negate the
2172 effect of the switch for that method. For example, if you do want to
2173 compare pointers to a particular inline method, you might mark it as
2174 having default visibility. Marking the enclosing class with explicit
2175 visibility will have no effect.
2177 Explicitly instantiated inline methods are unaffected by this option
2178 as their linkage might otherwise cross a shared library boundary.
2179 @xref{Template Instantiation}.
2181 @item -fvisibility-ms-compat
2182 @opindex fvisibility-ms-compat
2183 This flag attempts to use visibility settings to make GCC's C++
2184 linkage model compatible with that of Microsoft Visual Studio.
2186 The flag makes these changes to GCC's linkage model:
2190 It sets the default visibility to @code{hidden}, like
2191 @option{-fvisibility=hidden}.
2194 Types, but not their members, are not hidden by default.
2197 The One Definition Rule is relaxed for types without explicit
2198 visibility specifications which are defined in more than one different
2199 shared object: those declarations are permitted if they would have
2200 been permitted when this option was not used.
2203 In new code it is better to use @option{-fvisibility=hidden} and
2204 export those classes which are intended to be externally visible.
2205 Unfortunately it is possible for code to rely, perhaps accidentally,
2206 on the Visual Studio behavior.
2208 Among the consequences of these changes are that static data members
2209 of the same type with the same name but defined in different shared
2210 objects will be different, so changing one will not change the other;
2211 and that pointers to function members defined in different shared
2212 objects may not compare equal. When this flag is given, it is a
2213 violation of the ODR to define types with the same name differently.
2217 Do not use weak symbol support, even if it is provided by the linker.
2218 By default, G++ will use weak symbols if they are available. This
2219 option exists only for testing, and should not be used by end-users;
2220 it will result in inferior code and has no benefits. This option may
2221 be removed in a future release of G++.
2225 Do not search for header files in the standard directories specific to
2226 C++, but do still search the other standard directories. (This option
2227 is used when building the C++ library.)
2230 In addition, these optimization, warning, and code generation options
2231 have meanings only for C++ programs:
2234 @item -fno-default-inline
2235 @opindex fno-default-inline
2236 Do not assume @samp{inline} for functions defined inside a class scope.
2237 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2238 functions will have linkage like inline functions; they just won't be
2241 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2244 Warn when G++ generates code that is probably not compatible with the
2245 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2246 all such cases, there are probably some cases that are not warned about,
2247 even though G++ is generating incompatible code. There may also be
2248 cases where warnings are emitted even though the code that is generated
2251 You should rewrite your code to avoid these warnings if you are
2252 concerned about the fact that code generated by G++ may not be binary
2253 compatible with code generated by other compilers.
2255 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2260 A template with a non-type template parameter of reference type is
2261 mangled incorrectly:
2264 template <int &> struct S @{@};
2268 This is fixed in @option{-fabi-version=3}.
2271 SIMD vector types declared using @code{__attribute ((vector_size))} are
2272 mangled in a non-standard way that does not allow for overloading of
2273 functions taking vectors of different sizes.
2275 The mangling is changed in @option{-fabi-version=4}.
2278 The known incompatibilities in @option{-fabi-version=1} include:
2283 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2284 pack data into the same byte as a base class. For example:
2287 struct A @{ virtual void f(); int f1 : 1; @};
2288 struct B : public A @{ int f2 : 1; @};
2292 In this case, G++ will place @code{B::f2} into the same byte
2293 as@code{A::f1}; other compilers will not. You can avoid this problem
2294 by explicitly padding @code{A} so that its size is a multiple of the
2295 byte size on your platform; that will cause G++ and other compilers to
2296 layout @code{B} identically.
2299 Incorrect handling of tail-padding for virtual bases. G++ does not use
2300 tail padding when laying out virtual bases. For example:
2303 struct A @{ virtual void f(); char c1; @};
2304 struct B @{ B(); char c2; @};
2305 struct C : public A, public virtual B @{@};
2309 In this case, G++ will not place @code{B} into the tail-padding for
2310 @code{A}; other compilers will. You can avoid this problem by
2311 explicitly padding @code{A} so that its size is a multiple of its
2312 alignment (ignoring virtual base classes); that will cause G++ and other
2313 compilers to layout @code{C} identically.
2316 Incorrect handling of bit-fields with declared widths greater than that
2317 of their underlying types, when the bit-fields appear in a union. For
2321 union U @{ int i : 4096; @};
2325 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2326 union too small by the number of bits in an @code{int}.
2329 Empty classes can be placed at incorrect offsets. For example:
2339 struct C : public B, public A @{@};
2343 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2344 it should be placed at offset zero. G++ mistakenly believes that the
2345 @code{A} data member of @code{B} is already at offset zero.
2348 Names of template functions whose types involve @code{typename} or
2349 template template parameters can be mangled incorrectly.
2352 template <typename Q>
2353 void f(typename Q::X) @{@}
2355 template <template <typename> class Q>
2356 void f(typename Q<int>::X) @{@}
2360 Instantiations of these templates may be mangled incorrectly.
2364 It also warns psABI related changes. The known psABI changes at this
2370 For SYSV/x86-64, when passing union with long double, it is changed to
2371 pass in memory as specified in psABI. For example:
2381 @code{union U} will always be passed in memory.
2385 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2386 @opindex Wctor-dtor-privacy
2387 @opindex Wno-ctor-dtor-privacy
2388 Warn when a class seems unusable because all the constructors or
2389 destructors in that class are private, and it has neither friends nor
2390 public static member functions.
2392 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2393 @opindex Wdelete-non-virtual-dtor
2394 @opindex Wno-delete-non-virtual-dtor
2395 Warn when @samp{delete} is used to destroy an instance of a class which
2396 has virtual functions and non-virtual destructor. It is unsafe to delete
2397 an instance of a derived class through a pointer to a base class if the
2398 base class does not have a virtual destructor. This warning is enabled
2401 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2403 @opindex Wno-narrowing
2404 Warn when a narrowing conversion prohibited by C++11 occurs within
2408 int i = @{ 2.2 @}; // error: narrowing from double to int
2411 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2413 With -std=c++11, @option{-Wno-narrowing} suppresses the diagnostic
2414 required by the standard. Note that this does not affect the meaning
2415 of well-formed code; narrowing conversions are still considered
2416 ill-formed in SFINAE context.
2418 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2420 @opindex Wno-noexcept
2421 Warn when a noexcept-expression evaluates to false because of a call
2422 to a function that does not have a non-throwing exception
2423 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2424 the compiler to never throw an exception.
2426 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2427 @opindex Wnon-virtual-dtor
2428 @opindex Wno-non-virtual-dtor
2429 Warn when a class has virtual functions and accessible non-virtual
2430 destructor, in which case it would be possible but unsafe to delete
2431 an instance of a derived class through a pointer to the base class.
2432 This warning is also enabled if -Weffc++ is specified.
2434 @item -Wreorder @r{(C++ and Objective-C++ only)}
2436 @opindex Wno-reorder
2437 @cindex reordering, warning
2438 @cindex warning for reordering of member initializers
2439 Warn when the order of member initializers given in the code does not
2440 match the order in which they must be executed. For instance:
2446 A(): j (0), i (1) @{ @}
2450 The compiler will rearrange the member initializers for @samp{i}
2451 and @samp{j} to match the declaration order of the members, emitting
2452 a warning to that effect. This warning is enabled by @option{-Wall}.
2455 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2458 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2461 Warn about violations of the following style guidelines from Scott Meyers'
2462 @cite{Effective C++} book:
2466 Item 11: Define a copy constructor and an assignment operator for classes
2467 with dynamically allocated memory.
2470 Item 12: Prefer initialization to assignment in constructors.
2473 Item 14: Make destructors virtual in base classes.
2476 Item 15: Have @code{operator=} return a reference to @code{*this}.
2479 Item 23: Don't try to return a reference when you must return an object.
2483 Also warn about violations of the following style guidelines from
2484 Scott Meyers' @cite{More Effective C++} book:
2488 Item 6: Distinguish between prefix and postfix forms of increment and
2489 decrement operators.
2492 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2496 When selecting this option, be aware that the standard library
2497 headers do not obey all of these guidelines; use @samp{grep -v}
2498 to filter out those warnings.
2500 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2501 @opindex Wstrict-null-sentinel
2502 @opindex Wno-strict-null-sentinel
2503 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2504 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2505 to @code{__null}. Although it is a null pointer constant not a null pointer,
2506 it is guaranteed to be of the same size as a pointer. But this use is
2507 not portable across different compilers.
2509 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2510 @opindex Wno-non-template-friend
2511 @opindex Wnon-template-friend
2512 Disable warnings when non-templatized friend functions are declared
2513 within a template. Since the advent of explicit template specification
2514 support in G++, if the name of the friend is an unqualified-id (i.e.,
2515 @samp{friend foo(int)}), the C++ language specification demands that the
2516 friend declare or define an ordinary, nontemplate function. (Section
2517 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2518 could be interpreted as a particular specialization of a templatized
2519 function. Because this non-conforming behavior is no longer the default
2520 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2521 check existing code for potential trouble spots and is on by default.
2522 This new compiler behavior can be turned off with
2523 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2524 but disables the helpful warning.
2526 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2527 @opindex Wold-style-cast
2528 @opindex Wno-old-style-cast
2529 Warn if an old-style (C-style) cast to a non-void type is used within
2530 a C++ program. The new-style casts (@samp{dynamic_cast},
2531 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2532 less vulnerable to unintended effects and much easier to search for.
2534 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2535 @opindex Woverloaded-virtual
2536 @opindex Wno-overloaded-virtual
2537 @cindex overloaded virtual function, warning
2538 @cindex warning for overloaded virtual function
2539 Warn when a function declaration hides virtual functions from a
2540 base class. For example, in:
2547 struct B: public A @{
2552 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2560 will fail to compile.
2562 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2563 @opindex Wno-pmf-conversions
2564 @opindex Wpmf-conversions
2565 Disable the diagnostic for converting a bound pointer to member function
2568 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2569 @opindex Wsign-promo
2570 @opindex Wno-sign-promo
2571 Warn when overload resolution chooses a promotion from unsigned or
2572 enumerated type to a signed type, over a conversion to an unsigned type of
2573 the same size. Previous versions of G++ would try to preserve
2574 unsignedness, but the standard mandates the current behavior.
2579 A& operator = (int);
2589 In this example, G++ will synthesize a default @samp{A& operator =
2590 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2593 @node Objective-C and Objective-C++ Dialect Options
2594 @section Options Controlling Objective-C and Objective-C++ Dialects
2596 @cindex compiler options, Objective-C and Objective-C++
2597 @cindex Objective-C and Objective-C++ options, command line
2598 @cindex options, Objective-C and Objective-C++
2599 (NOTE: This manual does not describe the Objective-C and Objective-C++
2600 languages themselves. @xref{Standards,,Language Standards
2601 Supported by GCC}, for references.)
2603 This section describes the command-line options that are only meaningful
2604 for Objective-C and Objective-C++ programs, but you can also use most of
2605 the language-independent GNU compiler options.
2606 For example, you might compile a file @code{some_class.m} like this:
2609 gcc -g -fgnu-runtime -O -c some_class.m
2613 In this example, @option{-fgnu-runtime} is an option meant only for
2614 Objective-C and Objective-C++ programs; you can use the other options with
2615 any language supported by GCC@.
2617 Note that since Objective-C is an extension of the C language, Objective-C
2618 compilations may also use options specific to the C front-end (e.g.,
2619 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2620 C++-specific options (e.g., @option{-Wabi}).
2622 Here is a list of options that are @emph{only} for compiling Objective-C
2623 and Objective-C++ programs:
2626 @item -fconstant-string-class=@var{class-name}
2627 @opindex fconstant-string-class
2628 Use @var{class-name} as the name of the class to instantiate for each
2629 literal string specified with the syntax @code{@@"@dots{}"}. The default
2630 class name is @code{NXConstantString} if the GNU runtime is being used, and
2631 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2632 @option{-fconstant-cfstrings} option, if also present, will override the
2633 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2634 to be laid out as constant CoreFoundation strings.
2637 @opindex fgnu-runtime
2638 Generate object code compatible with the standard GNU Objective-C
2639 runtime. This is the default for most types of systems.
2641 @item -fnext-runtime
2642 @opindex fnext-runtime
2643 Generate output compatible with the NeXT runtime. This is the default
2644 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2645 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2648 @item -fno-nil-receivers
2649 @opindex fno-nil-receivers
2650 Assume that all Objective-C message dispatches (@code{[receiver
2651 message:arg]}) in this translation unit ensure that the receiver is
2652 not @code{nil}. This allows for more efficient entry points in the
2653 runtime to be used. This option is only available in conjunction with
2654 the NeXT runtime and ABI version 0 or 1.
2656 @item -fobjc-abi-version=@var{n}
2657 @opindex fobjc-abi-version
2658 Use version @var{n} of the Objective-C ABI for the selected runtime.
2659 This option is currently supported only for the NeXT runtime. In that
2660 case, Version 0 is the traditional (32-bit) ABI without support for
2661 properties and other Objective-C 2.0 additions. Version 1 is the
2662 traditional (32-bit) ABI with support for properties and other
2663 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2664 nothing is specified, the default is Version 0 on 32-bit target
2665 machines, and Version 2 on 64-bit target machines.
2667 @item -fobjc-call-cxx-cdtors
2668 @opindex fobjc-call-cxx-cdtors
2669 For each Objective-C class, check if any of its instance variables is a
2670 C++ object with a non-trivial default constructor. If so, synthesize a
2671 special @code{- (id) .cxx_construct} instance method that will run
2672 non-trivial default constructors on any such instance variables, in order,
2673 and then return @code{self}. Similarly, check if any instance variable
2674 is a C++ object with a non-trivial destructor, and if so, synthesize a
2675 special @code{- (void) .cxx_destruct} method that will run
2676 all such default destructors, in reverse order.
2678 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2679 methods thusly generated will only operate on instance variables
2680 declared in the current Objective-C class, and not those inherited
2681 from superclasses. It is the responsibility of the Objective-C
2682 runtime to invoke all such methods in an object's inheritance
2683 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2684 by the runtime immediately after a new object instance is allocated;
2685 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2686 before the runtime deallocates an object instance.
2688 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2689 support for invoking the @code{- (id) .cxx_construct} and
2690 @code{- (void) .cxx_destruct} methods.
2692 @item -fobjc-direct-dispatch
2693 @opindex fobjc-direct-dispatch
2694 Allow fast jumps to the message dispatcher. On Darwin this is
2695 accomplished via the comm page.
2697 @item -fobjc-exceptions
2698 @opindex fobjc-exceptions
2699 Enable syntactic support for structured exception handling in
2700 Objective-C, similar to what is offered by C++ and Java. This option
2701 is required to use the Objective-C keywords @code{@@try},
2702 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2703 @code{@@synchronized}. This option is available with both the GNU
2704 runtime and the NeXT runtime (but not available in conjunction with
2705 the NeXT runtime on Mac OS X 10.2 and earlier).
2709 Enable garbage collection (GC) in Objective-C and Objective-C++
2710 programs. This option is only available with the NeXT runtime; the
2711 GNU runtime has a different garbage collection implementation that
2712 does not require special compiler flags.
2714 @item -fobjc-nilcheck
2715 @opindex fobjc-nilcheck
2716 For the NeXT runtime with version 2 of the ABI, check for a nil
2717 receiver in method invocations before doing the actual method call.
2718 This is the default and can be disabled using
2719 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2720 checked for nil in this way no matter what this flag is set to.
2721 Currently this flag does nothing when the GNU runtime, or an older
2722 version of the NeXT runtime ABI, is used.
2724 @item -fobjc-std=objc1
2726 Conform to the language syntax of Objective-C 1.0, the language
2727 recognized by GCC 4.0. This only affects the Objective-C additions to
2728 the C/C++ language; it does not affect conformance to C/C++ standards,
2729 which is controlled by the separate C/C++ dialect option flags. When
2730 this option is used with the Objective-C or Objective-C++ compiler,
2731 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2732 This is useful if you need to make sure that your Objective-C code can
2733 be compiled with older versions of GCC.
2735 @item -freplace-objc-classes
2736 @opindex freplace-objc-classes
2737 Emit a special marker instructing @command{ld(1)} not to statically link in
2738 the resulting object file, and allow @command{dyld(1)} to load it in at
2739 run time instead. This is used in conjunction with the Fix-and-Continue
2740 debugging mode, where the object file in question may be recompiled and
2741 dynamically reloaded in the course of program execution, without the need
2742 to restart the program itself. Currently, Fix-and-Continue functionality
2743 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2748 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2749 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2750 compile time) with static class references that get initialized at load time,
2751 which improves run-time performance. Specifying the @option{-fzero-link} flag
2752 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2753 to be retained. This is useful in Zero-Link debugging mode, since it allows
2754 for individual class implementations to be modified during program execution.
2755 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2756 regardless of command line options.
2760 Dump interface declarations for all classes seen in the source file to a
2761 file named @file{@var{sourcename}.decl}.
2763 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2764 @opindex Wassign-intercept
2765 @opindex Wno-assign-intercept
2766 Warn whenever an Objective-C assignment is being intercepted by the
2769 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2770 @opindex Wno-protocol
2772 If a class is declared to implement a protocol, a warning is issued for
2773 every method in the protocol that is not implemented by the class. The
2774 default behavior is to issue a warning for every method not explicitly
2775 implemented in the class, even if a method implementation is inherited
2776 from the superclass. If you use the @option{-Wno-protocol} option, then
2777 methods inherited from the superclass are considered to be implemented,
2778 and no warning is issued for them.
2780 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2782 @opindex Wno-selector
2783 Warn if multiple methods of different types for the same selector are
2784 found during compilation. The check is performed on the list of methods
2785 in the final stage of compilation. Additionally, a check is performed
2786 for each selector appearing in a @code{@@selector(@dots{})}
2787 expression, and a corresponding method for that selector has been found
2788 during compilation. Because these checks scan the method table only at
2789 the end of compilation, these warnings are not produced if the final
2790 stage of compilation is not reached, for example because an error is
2791 found during compilation, or because the @option{-fsyntax-only} option is
2794 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2795 @opindex Wstrict-selector-match
2796 @opindex Wno-strict-selector-match
2797 Warn if multiple methods with differing argument and/or return types are
2798 found for a given selector when attempting to send a message using this
2799 selector to a receiver of type @code{id} or @code{Class}. When this flag
2800 is off (which is the default behavior), the compiler will omit such warnings
2801 if any differences found are confined to types which share the same size
2804 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2805 @opindex Wundeclared-selector
2806 @opindex Wno-undeclared-selector
2807 Warn if a @code{@@selector(@dots{})} expression referring to an
2808 undeclared selector is found. A selector is considered undeclared if no
2809 method with that name has been declared before the
2810 @code{@@selector(@dots{})} expression, either explicitly in an
2811 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2812 an @code{@@implementation} section. This option always performs its
2813 checks as soon as a @code{@@selector(@dots{})} expression is found,
2814 while @option{-Wselector} only performs its checks in the final stage of
2815 compilation. This also enforces the coding style convention
2816 that methods and selectors must be declared before being used.
2818 @item -print-objc-runtime-info
2819 @opindex print-objc-runtime-info
2820 Generate C header describing the largest structure that is passed by
2825 @node Language Independent Options
2826 @section Options to Control Diagnostic Messages Formatting
2827 @cindex options to control diagnostics formatting
2828 @cindex diagnostic messages
2829 @cindex message formatting
2831 Traditionally, diagnostic messages have been formatted irrespective of
2832 the output device's aspect (e.g.@: its width, @dots{}). The options described
2833 below can be used to control the diagnostic messages formatting
2834 algorithm, e.g.@: how many characters per line, how often source location
2835 information should be reported. Right now, only the C++ front end can
2836 honor these options. However it is expected, in the near future, that
2837 the remaining front ends would be able to digest them correctly.
2840 @item -fmessage-length=@var{n}
2841 @opindex fmessage-length
2842 Try to format error messages so that they fit on lines of about @var{n}
2843 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2844 the front ends supported by GCC@. If @var{n} is zero, then no
2845 line-wrapping will be done; each error message will appear on a single
2848 @opindex fdiagnostics-show-location
2849 @item -fdiagnostics-show-location=once
2850 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2851 reporter to emit @emph{once} source location information; that is, in
2852 case the message is too long to fit on a single physical line and has to
2853 be wrapped, the source location won't be emitted (as prefix) again,
2854 over and over, in subsequent continuation lines. This is the default
2857 @item -fdiagnostics-show-location=every-line
2858 Only meaningful in line-wrapping mode. Instructs the diagnostic
2859 messages reporter to emit the same source location information (as
2860 prefix) for physical lines that result from the process of breaking
2861 a message which is too long to fit on a single line.
2863 @item -fno-diagnostics-show-option
2864 @opindex fno-diagnostics-show-option
2865 @opindex fdiagnostics-show-option
2866 By default, each diagnostic emitted includes text which indicates the
2867 command line option that directly controls the diagnostic (if such an
2868 option is known to the diagnostic machinery). Specifying the
2869 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2871 @item -Wcoverage-mismatch
2872 @opindex Wcoverage-mismatch
2873 Warn if feedback profiles do not match when using the
2874 @option{-fprofile-use} option.
2875 If a source file was changed between @option{-fprofile-gen} and
2876 @option{-fprofile-use}, the files with the profile feedback can fail
2877 to match the source file and GCC can not use the profile feedback
2878 information. By default, this warning is enabled and is treated as an
2879 error. @option{-Wno-coverage-mismatch} can be used to disable the
2880 warning or @option{-Wno-error=coverage-mismatch} can be used to
2881 disable the error. Disable the error for this warning can result in
2882 poorly optimized code, so disabling the error is useful only in the
2883 case of very minor changes such as bug fixes to an existing code-base.
2884 Completely disabling the warning is not recommended.
2888 @node Warning Options
2889 @section Options to Request or Suppress Warnings
2890 @cindex options to control warnings
2891 @cindex warning messages
2892 @cindex messages, warning
2893 @cindex suppressing warnings
2895 Warnings are diagnostic messages that report constructions which
2896 are not inherently erroneous but which are risky or suggest there
2897 may have been an error.
2899 The following language-independent options do not enable specific
2900 warnings but control the kinds of diagnostics produced by GCC.
2903 @cindex syntax checking
2905 @opindex fsyntax-only
2906 Check the code for syntax errors, but don't do anything beyond that.
2908 @item -fmax-errors=@var{n}
2909 @opindex fmax-errors
2910 Limits the maximum number of error messages to @var{n}, at which point
2911 GCC bails out rather than attempting to continue processing the source
2912 code. If @var{n} is 0 (the default), there is no limit on the number
2913 of error messages produced. If @option{-Wfatal-errors} is also
2914 specified, then @option{-Wfatal-errors} takes precedence over this
2919 Inhibit all warning messages.
2924 Make all warnings into errors.
2929 Make the specified warning into an error. The specifier for a warning
2930 is appended, for example @option{-Werror=switch} turns the warnings
2931 controlled by @option{-Wswitch} into errors. This switch takes a
2932 negative form, to be used to negate @option{-Werror} for specific
2933 warnings, for example @option{-Wno-error=switch} makes
2934 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2937 The warning message for each controllable warning includes the
2938 option which controls the warning. That option can then be used with
2939 @option{-Werror=} and @option{-Wno-error=} as described above.
2940 (Printing of the option in the warning message can be disabled using the
2941 @option{-fno-diagnostics-show-option} flag.)
2943 Note that specifying @option{-Werror=}@var{foo} automatically implies
2944 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2947 @item -Wfatal-errors
2948 @opindex Wfatal-errors
2949 @opindex Wno-fatal-errors
2950 This option causes the compiler to abort compilation on the first error
2951 occurred rather than trying to keep going and printing further error
2956 You can request many specific warnings with options beginning
2957 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2958 implicit declarations. Each of these specific warning options also
2959 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2960 example, @option{-Wno-implicit}. This manual lists only one of the
2961 two forms, whichever is not the default. For further,
2962 language-specific options also refer to @ref{C++ Dialect Options} and
2963 @ref{Objective-C and Objective-C++ Dialect Options}.
2965 When an unrecognized warning option is requested (e.g.,
2966 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2967 that the option is not recognized. However, if the @option{-Wno-} form
2968 is used, the behavior is slightly different: No diagnostic will be
2969 produced for @option{-Wno-unknown-warning} unless other diagnostics
2970 are being produced. This allows the use of new @option{-Wno-} options
2971 with old compilers, but if something goes wrong, the compiler will
2972 warn that an unrecognized option was used.
2977 Issue all the warnings demanded by strict ISO C and ISO C++;
2978 reject all programs that use forbidden extensions, and some other
2979 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2980 version of the ISO C standard specified by any @option{-std} option used.
2982 Valid ISO C and ISO C++ programs should compile properly with or without
2983 this option (though a rare few will require @option{-ansi} or a
2984 @option{-std} option specifying the required version of ISO C)@. However,
2985 without this option, certain GNU extensions and traditional C and C++
2986 features are supported as well. With this option, they are rejected.
2988 @option{-pedantic} does not cause warning messages for use of the
2989 alternate keywords whose names begin and end with @samp{__}. Pedantic
2990 warnings are also disabled in the expression that follows
2991 @code{__extension__}. However, only system header files should use
2992 these escape routes; application programs should avoid them.
2993 @xref{Alternate Keywords}.
2995 Some users try to use @option{-pedantic} to check programs for strict ISO
2996 C conformance. They soon find that it does not do quite what they want:
2997 it finds some non-ISO practices, but not all---only those for which
2998 ISO C @emph{requires} a diagnostic, and some others for which
2999 diagnostics have been added.
3001 A feature to report any failure to conform to ISO C might be useful in
3002 some instances, but would require considerable additional work and would
3003 be quite different from @option{-pedantic}. We don't have plans to
3004 support such a feature in the near future.
3006 Where the standard specified with @option{-std} represents a GNU
3007 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3008 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3009 extended dialect is based. Warnings from @option{-pedantic} are given
3010 where they are required by the base standard. (It would not make sense
3011 for such warnings to be given only for features not in the specified GNU
3012 C dialect, since by definition the GNU dialects of C include all
3013 features the compiler supports with the given option, and there would be
3014 nothing to warn about.)
3016 @item -pedantic-errors
3017 @opindex pedantic-errors
3018 Like @option{-pedantic}, except that errors are produced rather than
3024 This enables all the warnings about constructions that some users
3025 consider questionable, and that are easy to avoid (or modify to
3026 prevent the warning), even in conjunction with macros. This also
3027 enables some language-specific warnings described in @ref{C++ Dialect
3028 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3030 @option{-Wall} turns on the following warning flags:
3032 @gccoptlist{-Waddress @gol
3033 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3035 -Wchar-subscripts @gol
3036 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
3037 -Wimplicit-int @r{(C and Objective-C only)} @gol
3038 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3041 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3042 -Wmaybe-uninitialized @gol
3043 -Wmissing-braces @gol
3049 -Wsequence-point @gol
3050 -Wsign-compare @r{(only in C++)} @gol
3051 -Wstrict-aliasing @gol
3052 -Wstrict-overflow=1 @gol
3055 -Wuninitialized @gol
3056 -Wunknown-pragmas @gol
3057 -Wunused-function @gol
3060 -Wunused-variable @gol
3061 -Wvolatile-register-var @gol
3064 Note that some warning flags are not implied by @option{-Wall}. Some of
3065 them warn about constructions that users generally do not consider
3066 questionable, but which occasionally you might wish to check for;
3067 others warn about constructions that are necessary or hard to avoid in
3068 some cases, and there is no simple way to modify the code to suppress
3069 the warning. Some of them are enabled by @option{-Wextra} but many of
3070 them must be enabled individually.
3076 This enables some extra warning flags that are not enabled by
3077 @option{-Wall}. (This option used to be called @option{-W}. The older
3078 name is still supported, but the newer name is more descriptive.)
3080 @gccoptlist{-Wclobbered @gol
3082 -Wignored-qualifiers @gol
3083 -Wmissing-field-initializers @gol
3084 -Wmissing-parameter-type @r{(C only)} @gol
3085 -Wold-style-declaration @r{(C only)} @gol
3086 -Woverride-init @gol
3089 -Wuninitialized @gol
3090 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3091 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3094 The option @option{-Wextra} also prints warning messages for the
3100 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3101 @samp{>}, or @samp{>=}.
3104 (C++ only) An enumerator and a non-enumerator both appear in a
3105 conditional expression.
3108 (C++ only) Ambiguous virtual bases.
3111 (C++ only) Subscripting an array which has been declared @samp{register}.
3114 (C++ only) Taking the address of a variable which has been declared
3118 (C++ only) A base class is not initialized in a derived class' copy
3123 @item -Wchar-subscripts
3124 @opindex Wchar-subscripts
3125 @opindex Wno-char-subscripts
3126 Warn if an array subscript has type @code{char}. This is a common cause
3127 of error, as programmers often forget that this type is signed on some
3129 This warning is enabled by @option{-Wall}.
3133 @opindex Wno-comment
3134 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3135 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3136 This warning is enabled by @option{-Wall}.
3139 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3141 Suppress warning messages emitted by @code{#warning} directives.
3143 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3144 @opindex Wdouble-promotion
3145 @opindex Wno-double-promotion
3146 Give a warning when a value of type @code{float} is implicitly
3147 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3148 floating-point unit implement @code{float} in hardware, but emulate
3149 @code{double} in software. On such a machine, doing computations
3150 using @code{double} values is much more expensive because of the
3151 overhead required for software emulation.
3153 It is easy to accidentally do computations with @code{double} because
3154 floating-point literals are implicitly of type @code{double}. For
3158 float area(float radius)
3160 return 3.14159 * radius * radius;
3164 the compiler will perform the entire computation with @code{double}
3165 because the floating-point literal is a @code{double}.
3170 @opindex ffreestanding
3171 @opindex fno-builtin
3172 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3173 the arguments supplied have types appropriate to the format string
3174 specified, and that the conversions specified in the format string make
3175 sense. This includes standard functions, and others specified by format
3176 attributes (@pxref{Function Attributes}), in the @code{printf},
3177 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3178 not in the C standard) families (or other target-specific families).
3179 Which functions are checked without format attributes having been
3180 specified depends on the standard version selected, and such checks of
3181 functions without the attribute specified are disabled by
3182 @option{-ffreestanding} or @option{-fno-builtin}.
3184 The formats are checked against the format features supported by GNU
3185 libc version 2.2. These include all ISO C90 and C99 features, as well
3186 as features from the Single Unix Specification and some BSD and GNU
3187 extensions. Other library implementations may not support all these
3188 features; GCC does not support warning about features that go beyond a
3189 particular library's limitations. However, if @option{-pedantic} is used
3190 with @option{-Wformat}, warnings will be given about format features not
3191 in the selected standard version (but not for @code{strfmon} formats,
3192 since those are not in any version of the C standard). @xref{C Dialect
3193 Options,,Options Controlling C Dialect}.
3195 Since @option{-Wformat} also checks for null format arguments for
3196 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3198 @option{-Wformat} is included in @option{-Wall}. For more control over some
3199 aspects of format checking, the options @option{-Wformat-y2k},
3200 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3201 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3202 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3205 @opindex Wformat-y2k
3206 @opindex Wno-format-y2k
3207 If @option{-Wformat} is specified, also warn about @code{strftime}
3208 formats which may yield only a two-digit year.
3210 @item -Wno-format-contains-nul
3211 @opindex Wno-format-contains-nul
3212 @opindex Wformat-contains-nul
3213 If @option{-Wformat} is specified, do not warn about format strings that
3216 @item -Wno-format-extra-args
3217 @opindex Wno-format-extra-args
3218 @opindex Wformat-extra-args
3219 If @option{-Wformat} is specified, do not warn about excess arguments to a
3220 @code{printf} or @code{scanf} format function. The C standard specifies
3221 that such arguments are ignored.
3223 Where the unused arguments lie between used arguments that are
3224 specified with @samp{$} operand number specifications, normally
3225 warnings are still given, since the implementation could not know what
3226 type to pass to @code{va_arg} to skip the unused arguments. However,
3227 in the case of @code{scanf} formats, this option will suppress the
3228 warning if the unused arguments are all pointers, since the Single
3229 Unix Specification says that such unused arguments are allowed.
3231 @item -Wno-format-zero-length
3232 @opindex Wno-format-zero-length
3233 @opindex Wformat-zero-length
3234 If @option{-Wformat} is specified, do not warn about zero-length formats.
3235 The C standard specifies that zero-length formats are allowed.
3237 @item -Wformat-nonliteral
3238 @opindex Wformat-nonliteral
3239 @opindex Wno-format-nonliteral
3240 If @option{-Wformat} is specified, also warn if the format string is not a
3241 string literal and so cannot be checked, unless the format function
3242 takes its format arguments as a @code{va_list}.
3244 @item -Wformat-security
3245 @opindex Wformat-security
3246 @opindex Wno-format-security
3247 If @option{-Wformat} is specified, also warn about uses of format
3248 functions that represent possible security problems. At present, this
3249 warns about calls to @code{printf} and @code{scanf} functions where the
3250 format string is not a string literal and there are no format arguments,
3251 as in @code{printf (foo);}. This may be a security hole if the format
3252 string came from untrusted input and contains @samp{%n}. (This is
3253 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3254 in future warnings may be added to @option{-Wformat-security} that are not
3255 included in @option{-Wformat-nonliteral}.)
3259 @opindex Wno-format=2
3260 Enable @option{-Wformat} plus format checks not included in
3261 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3262 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3266 @opindex Wno-nonnull
3267 Warn about passing a null pointer for arguments marked as
3268 requiring a non-null value by the @code{nonnull} function attribute.
3270 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3271 can be disabled with the @option{-Wno-nonnull} option.
3273 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3275 @opindex Wno-init-self
3276 Warn about uninitialized variables which are initialized with themselves.
3277 Note this option can only be used with the @option{-Wuninitialized} option.
3279 For example, GCC will warn about @code{i} being uninitialized in the
3280 following snippet only when @option{-Winit-self} has been specified:
3291 @item -Wimplicit-int @r{(C and Objective-C only)}
3292 @opindex Wimplicit-int
3293 @opindex Wno-implicit-int
3294 Warn when a declaration does not specify a type.
3295 This warning is enabled by @option{-Wall}.
3297 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3298 @opindex Wimplicit-function-declaration
3299 @opindex Wno-implicit-function-declaration
3300 Give a warning whenever a function is used before being declared. In
3301 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3302 enabled by default and it is made into an error by
3303 @option{-pedantic-errors}. This warning is also enabled by
3306 @item -Wimplicit @r{(C and Objective-C only)}
3308 @opindex Wno-implicit
3309 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3310 This warning is enabled by @option{-Wall}.
3312 @item -Wignored-qualifiers @r{(C and C++ only)}
3313 @opindex Wignored-qualifiers
3314 @opindex Wno-ignored-qualifiers
3315 Warn if the return type of a function has a type qualifier
3316 such as @code{const}. For ISO C such a type qualifier has no effect,
3317 since the value returned by a function is not an lvalue.
3318 For C++, the warning is only emitted for scalar types or @code{void}.
3319 ISO C prohibits qualified @code{void} return types on function
3320 definitions, so such return types always receive a warning
3321 even without this option.
3323 This warning is also enabled by @option{-Wextra}.
3328 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3329 a function with external linkage, returning int, taking either zero
3330 arguments, two, or three arguments of appropriate types. This warning
3331 is enabled by default in C++ and is enabled by either @option{-Wall}
3332 or @option{-pedantic}.
3334 @item -Wmissing-braces
3335 @opindex Wmissing-braces
3336 @opindex Wno-missing-braces
3337 Warn if an aggregate or union initializer is not fully bracketed. In
3338 the following example, the initializer for @samp{a} is not fully
3339 bracketed, but that for @samp{b} is fully bracketed.
3342 int a[2][2] = @{ 0, 1, 2, 3 @};
3343 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3346 This warning is enabled by @option{-Wall}.
3348 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3349 @opindex Wmissing-include-dirs
3350 @opindex Wno-missing-include-dirs
3351 Warn if a user-supplied include directory does not exist.
3354 @opindex Wparentheses
3355 @opindex Wno-parentheses
3356 Warn if parentheses are omitted in certain contexts, such
3357 as when there is an assignment in a context where a truth value
3358 is expected, or when operators are nested whose precedence people
3359 often get confused about.
3361 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3362 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3363 interpretation from that of ordinary mathematical notation.
3365 Also warn about constructions where there may be confusion to which
3366 @code{if} statement an @code{else} branch belongs. Here is an example of
3381 In C/C++, every @code{else} branch belongs to the innermost possible
3382 @code{if} statement, which in this example is @code{if (b)}. This is
3383 often not what the programmer expected, as illustrated in the above
3384 example by indentation the programmer chose. When there is the
3385 potential for this confusion, GCC will issue a warning when this flag
3386 is specified. To eliminate the warning, add explicit braces around
3387 the innermost @code{if} statement so there is no way the @code{else}
3388 could belong to the enclosing @code{if}. The resulting code would
3405 Also warn for dangerous uses of the
3406 ?: with omitted middle operand GNU extension. When the condition
3407 in the ?: operator is a boolean expression the omitted value will
3408 be always 1. Often the user expects it to be a value computed
3409 inside the conditional expression instead.
3411 This warning is enabled by @option{-Wall}.
3413 @item -Wsequence-point
3414 @opindex Wsequence-point
3415 @opindex Wno-sequence-point
3416 Warn about code that may have undefined semantics because of violations
3417 of sequence point rules in the C and C++ standards.
3419 The C and C++ standards defines the order in which expressions in a C/C++
3420 program are evaluated in terms of @dfn{sequence points}, which represent
3421 a partial ordering between the execution of parts of the program: those
3422 executed before the sequence point, and those executed after it. These
3423 occur after the evaluation of a full expression (one which is not part
3424 of a larger expression), after the evaluation of the first operand of a
3425 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3426 function is called (but after the evaluation of its arguments and the
3427 expression denoting the called function), and in certain other places.
3428 Other than as expressed by the sequence point rules, the order of
3429 evaluation of subexpressions of an expression is not specified. All
3430 these rules describe only a partial order rather than a total order,
3431 since, for example, if two functions are called within one expression
3432 with no sequence point between them, the order in which the functions
3433 are called is not specified. However, the standards committee have
3434 ruled that function calls do not overlap.
3436 It is not specified when between sequence points modifications to the
3437 values of objects take effect. Programs whose behavior depends on this
3438 have undefined behavior; the C and C++ standards specify that ``Between
3439 the previous and next sequence point an object shall have its stored
3440 value modified at most once by the evaluation of an expression.
3441 Furthermore, the prior value shall be read only to determine the value
3442 to be stored.''. If a program breaks these rules, the results on any
3443 particular implementation are entirely unpredictable.
3445 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3446 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3447 diagnosed by this option, and it may give an occasional false positive
3448 result, but in general it has been found fairly effective at detecting
3449 this sort of problem in programs.
3451 The standard is worded confusingly, therefore there is some debate
3452 over the precise meaning of the sequence point rules in subtle cases.
3453 Links to discussions of the problem, including proposed formal
3454 definitions, may be found on the GCC readings page, at
3455 @uref{http://gcc.gnu.org/@/readings.html}.
3457 This warning is enabled by @option{-Wall} for C and C++.
3460 @opindex Wreturn-type
3461 @opindex Wno-return-type
3462 Warn whenever a function is defined with a return-type that defaults
3463 to @code{int}. Also warn about any @code{return} statement with no
3464 return-value in a function whose return-type is not @code{void}
3465 (falling off the end of the function body is considered returning
3466 without a value), and about a @code{return} statement with an
3467 expression in a function whose return-type is @code{void}.
3469 For C++, a function without return type always produces a diagnostic
3470 message, even when @option{-Wno-return-type} is specified. The only
3471 exceptions are @samp{main} and functions defined in system headers.
3473 This warning is enabled by @option{-Wall}.
3478 Warn whenever a @code{switch} statement has an index of enumerated type
3479 and lacks a @code{case} for one or more of the named codes of that
3480 enumeration. (The presence of a @code{default} label prevents this
3481 warning.) @code{case} labels outside the enumeration range also
3482 provoke warnings when this option is used (even if there is a
3483 @code{default} label).
3484 This warning is enabled by @option{-Wall}.
3486 @item -Wswitch-default
3487 @opindex Wswitch-default
3488 @opindex Wno-switch-default
3489 Warn whenever a @code{switch} statement does not have a @code{default}
3493 @opindex Wswitch-enum
3494 @opindex Wno-switch-enum
3495 Warn whenever a @code{switch} statement has an index of enumerated type
3496 and lacks a @code{case} for one or more of the named codes of that
3497 enumeration. @code{case} labels outside the enumeration range also
3498 provoke warnings when this option is used. The only difference
3499 between @option{-Wswitch} and this option is that this option gives a
3500 warning about an omitted enumeration code even if there is a
3501 @code{default} label.
3503 @item -Wsync-nand @r{(C and C++ only)}
3505 @opindex Wno-sync-nand
3506 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3507 built-in functions are used. These functions changed semantics in GCC 4.4.
3511 @opindex Wno-trigraphs
3512 Warn if any trigraphs are encountered that might change the meaning of
3513 the program (trigraphs within comments are not warned about).
3514 This warning is enabled by @option{-Wall}.
3516 @item -Wunused-but-set-parameter
3517 @opindex Wunused-but-set-parameter
3518 @opindex Wno-unused-but-set-parameter
3519 Warn whenever a function parameter is assigned to, but otherwise unused
3520 (aside from its declaration).
3522 To suppress this warning use the @samp{unused} attribute
3523 (@pxref{Variable Attributes}).
3525 This warning is also enabled by @option{-Wunused} together with
3528 @item -Wunused-but-set-variable
3529 @opindex Wunused-but-set-variable
3530 @opindex Wno-unused-but-set-variable
3531 Warn whenever a local variable is assigned to, but otherwise unused
3532 (aside from its declaration).
3533 This warning is enabled by @option{-Wall}.
3535 To suppress this warning use the @samp{unused} attribute
3536 (@pxref{Variable Attributes}).
3538 This warning is also enabled by @option{-Wunused}, which is enabled
3541 @item -Wunused-function
3542 @opindex Wunused-function
3543 @opindex Wno-unused-function
3544 Warn whenever a static function is declared but not defined or a
3545 non-inline static function is unused.
3546 This warning is enabled by @option{-Wall}.
3548 @item -Wunused-label
3549 @opindex Wunused-label
3550 @opindex Wno-unused-label
3551 Warn whenever a label is declared but not used.
3552 This warning is enabled by @option{-Wall}.
3554 To suppress this warning use the @samp{unused} attribute
3555 (@pxref{Variable Attributes}).
3557 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3558 @opindex Wunused-local-typedefs
3559 Warn when a typedef locally defined in a function is not used.
3561 @item -Wunused-parameter
3562 @opindex Wunused-parameter
3563 @opindex Wno-unused-parameter
3564 Warn whenever a function parameter is unused aside from its declaration.
3566 To suppress this warning use the @samp{unused} attribute
3567 (@pxref{Variable Attributes}).
3569 @item -Wno-unused-result
3570 @opindex Wunused-result
3571 @opindex Wno-unused-result
3572 Do not warn if a caller of a function marked with attribute
3573 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3574 its return value. The default is @option{-Wunused-result}.
3576 @item -Wunused-variable
3577 @opindex Wunused-variable
3578 @opindex Wno-unused-variable
3579 Warn whenever a local variable or non-constant static variable is unused
3580 aside from its declaration.
3581 This warning is enabled by @option{-Wall}.
3583 To suppress this warning use the @samp{unused} attribute
3584 (@pxref{Variable Attributes}).
3586 @item -Wunused-value
3587 @opindex Wunused-value
3588 @opindex Wno-unused-value
3589 Warn whenever a statement computes a result that is explicitly not
3590 used. To suppress this warning cast the unused expression to
3591 @samp{void}. This includes an expression-statement or the left-hand
3592 side of a comma expression that contains no side effects. For example,
3593 an expression such as @samp{x[i,j]} will cause a warning, while
3594 @samp{x[(void)i,j]} will not.
3596 This warning is enabled by @option{-Wall}.
3601 All the above @option{-Wunused} options combined.
3603 In order to get a warning about an unused function parameter, you must
3604 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3605 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3607 @item -Wuninitialized
3608 @opindex Wuninitialized
3609 @opindex Wno-uninitialized
3610 Warn if an automatic variable is used without first being initialized
3611 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3612 warn if a non-static reference or non-static @samp{const} member
3613 appears in a class without constructors.
3615 If you want to warn about code which uses the uninitialized value of the
3616 variable in its own initializer, use the @option{-Winit-self} option.
3618 These warnings occur for individual uninitialized or clobbered
3619 elements of structure, union or array variables as well as for
3620 variables which are uninitialized or clobbered as a whole. They do
3621 not occur for variables or elements declared @code{volatile}. Because
3622 these warnings depend on optimization, the exact variables or elements
3623 for which there are warnings will depend on the precise optimization
3624 options and version of GCC used.
3626 Note that there may be no warning about a variable that is used only
3627 to compute a value that itself is never used, because such
3628 computations may be deleted by data flow analysis before the warnings
3631 @item -Wmaybe-uninitialized
3632 @opindex Wmaybe-uninitialized
3633 @opindex Wno-maybe-uninitialized
3634 For an automatic variable, if there exists a path from the function
3635 entry to a use of the variable that is initialized, but there exist
3636 some other paths the variable is not initialized, the compiler will
3637 emit a warning if it can not prove the uninitialized paths do not
3638 happen at runtime. These warnings are made optional because GCC is
3639 not smart enough to see all the reasons why the code might be correct
3640 despite appearing to have an error. Here is one example of how
3661 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3662 always initialized, but GCC doesn't know this. To suppress the
3663 warning, the user needs to provide a default case with assert(0) or
3666 @cindex @code{longjmp} warnings
3667 This option also warns when a non-volatile automatic variable might be
3668 changed by a call to @code{longjmp}. These warnings as well are possible
3669 only in optimizing compilation.
3671 The compiler sees only the calls to @code{setjmp}. It cannot know
3672 where @code{longjmp} will be called; in fact, a signal handler could
3673 call it at any point in the code. As a result, you may get a warning
3674 even when there is in fact no problem because @code{longjmp} cannot
3675 in fact be called at the place which would cause a problem.
3677 Some spurious warnings can be avoided if you declare all the functions
3678 you use that never return as @code{noreturn}. @xref{Function
3681 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3683 @item -Wunknown-pragmas
3684 @opindex Wunknown-pragmas
3685 @opindex Wno-unknown-pragmas
3686 @cindex warning for unknown pragmas
3687 @cindex unknown pragmas, warning
3688 @cindex pragmas, warning of unknown
3689 Warn when a #pragma directive is encountered which is not understood by
3690 GCC@. If this command line option is used, warnings will even be issued
3691 for unknown pragmas in system header files. This is not the case if
3692 the warnings were only enabled by the @option{-Wall} command line option.
3695 @opindex Wno-pragmas
3697 Do not warn about misuses of pragmas, such as incorrect parameters,
3698 invalid syntax, or conflicts between pragmas. See also
3699 @samp{-Wunknown-pragmas}.
3701 @item -Wstrict-aliasing
3702 @opindex Wstrict-aliasing
3703 @opindex Wno-strict-aliasing
3704 This option is only active when @option{-fstrict-aliasing} is active.
3705 It warns about code which might break the strict aliasing rules that the
3706 compiler is using for optimization. The warning does not catch all
3707 cases, but does attempt to catch the more common pitfalls. It is
3708 included in @option{-Wall}.
3709 It is equivalent to @option{-Wstrict-aliasing=3}
3711 @item -Wstrict-aliasing=n
3712 @opindex Wstrict-aliasing=n
3713 @opindex Wno-strict-aliasing=n
3714 This option is only active when @option{-fstrict-aliasing} is active.
3715 It warns about code which might break the strict aliasing rules that the
3716 compiler is using for optimization.
3717 Higher levels correspond to higher accuracy (fewer false positives).
3718 Higher levels also correspond to more effort, similar to the way -O works.
3719 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3722 Level 1: Most aggressive, quick, least accurate.
3723 Possibly useful when higher levels
3724 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3725 false negatives. However, it has many false positives.
3726 Warns for all pointer conversions between possibly incompatible types,
3727 even if never dereferenced. Runs in the frontend only.
3729 Level 2: Aggressive, quick, not too precise.
3730 May still have many false positives (not as many as level 1 though),
3731 and few false negatives (but possibly more than level 1).
3732 Unlike level 1, it only warns when an address is taken. Warns about
3733 incomplete types. Runs in the frontend only.
3735 Level 3 (default for @option{-Wstrict-aliasing}):
3736 Should have very few false positives and few false
3737 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3738 Takes care of the common pun+dereference pattern in the frontend:
3739 @code{*(int*)&some_float}.
3740 If optimization is enabled, it also runs in the backend, where it deals
3741 with multiple statement cases using flow-sensitive points-to information.
3742 Only warns when the converted pointer is dereferenced.
3743 Does not warn about incomplete types.
3745 @item -Wstrict-overflow
3746 @itemx -Wstrict-overflow=@var{n}
3747 @opindex Wstrict-overflow
3748 @opindex Wno-strict-overflow
3749 This option is only active when @option{-fstrict-overflow} is active.
3750 It warns about cases where the compiler optimizes based on the
3751 assumption that signed overflow does not occur. Note that it does not
3752 warn about all cases where the code might overflow: it only warns
3753 about cases where the compiler implements some optimization. Thus
3754 this warning depends on the optimization level.
3756 An optimization which assumes that signed overflow does not occur is
3757 perfectly safe if the values of the variables involved are such that
3758 overflow never does, in fact, occur. Therefore this warning can
3759 easily give a false positive: a warning about code which is not
3760 actually a problem. To help focus on important issues, several
3761 warning levels are defined. No warnings are issued for the use of
3762 undefined signed overflow when estimating how many iterations a loop
3763 will require, in particular when determining whether a loop will be
3767 @item -Wstrict-overflow=1
3768 Warn about cases which are both questionable and easy to avoid. For
3769 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3770 compiler will simplify this to @code{1}. This level of
3771 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3772 are not, and must be explicitly requested.
3774 @item -Wstrict-overflow=2
3775 Also warn about other cases where a comparison is simplified to a
3776 constant. For example: @code{abs (x) >= 0}. This can only be
3777 simplified when @option{-fstrict-overflow} is in effect, because
3778 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3779 zero. @option{-Wstrict-overflow} (with no level) is the same as
3780 @option{-Wstrict-overflow=2}.
3782 @item -Wstrict-overflow=3
3783 Also warn about other cases where a comparison is simplified. For
3784 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3786 @item -Wstrict-overflow=4
3787 Also warn about other simplifications not covered by the above cases.
3788 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3790 @item -Wstrict-overflow=5
3791 Also warn about cases where the compiler reduces the magnitude of a
3792 constant involved in a comparison. For example: @code{x + 2 > y} will
3793 be simplified to @code{x + 1 >= y}. This is reported only at the
3794 highest warning level because this simplification applies to many
3795 comparisons, so this warning level will give a very large number of
3799 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3800 @opindex Wsuggest-attribute=
3801 @opindex Wno-suggest-attribute=
3802 Warn for cases where adding an attribute may be beneficial. The
3803 attributes currently supported are listed below.
3806 @item -Wsuggest-attribute=pure
3807 @itemx -Wsuggest-attribute=const
3808 @itemx -Wsuggest-attribute=noreturn
3809 @opindex Wsuggest-attribute=pure
3810 @opindex Wno-suggest-attribute=pure
3811 @opindex Wsuggest-attribute=const
3812 @opindex Wno-suggest-attribute=const
3813 @opindex Wsuggest-attribute=noreturn
3814 @opindex Wno-suggest-attribute=noreturn
3816 Warn about functions which might be candidates for attributes
3817 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3818 functions visible in other compilation units or (in the case of @code{pure} and
3819 @code{const}) if it cannot prove that the function returns normally. A function
3820 returns normally if it doesn't contain an infinite loop nor returns abnormally
3821 by throwing, calling @code{abort()} or trapping. This analysis requires option
3822 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3823 higher. Higher optimization levels improve the accuracy of the analysis.
3826 @item -Warray-bounds
3827 @opindex Wno-array-bounds
3828 @opindex Warray-bounds
3829 This option is only active when @option{-ftree-vrp} is active
3830 (default for @option{-O2} and above). It warns about subscripts to arrays
3831 that are always out of bounds. This warning is enabled by @option{-Wall}.
3833 @item -Wno-div-by-zero
3834 @opindex Wno-div-by-zero
3835 @opindex Wdiv-by-zero
3836 Do not warn about compile-time integer division by zero. Floating point
3837 division by zero is not warned about, as it can be a legitimate way of
3838 obtaining infinities and NaNs.
3840 @item -Wsystem-headers
3841 @opindex Wsystem-headers
3842 @opindex Wno-system-headers
3843 @cindex warnings from system headers
3844 @cindex system headers, warnings from
3845 Print warning messages for constructs found in system header files.
3846 Warnings from system headers are normally suppressed, on the assumption
3847 that they usually do not indicate real problems and would only make the
3848 compiler output harder to read. Using this command line option tells
3849 GCC to emit warnings from system headers as if they occurred in user
3850 code. However, note that using @option{-Wall} in conjunction with this
3851 option will @emph{not} warn about unknown pragmas in system
3852 headers---for that, @option{-Wunknown-pragmas} must also be used.
3855 @opindex Wtrampolines
3856 @opindex Wno-trampolines
3857 Warn about trampolines generated for pointers to nested functions.
3859 A trampoline is a small piece of data or code that is created at run
3860 time on the stack when the address of a nested function is taken, and
3861 is used to call the nested function indirectly. For some targets, it
3862 is made up of data only and thus requires no special treatment. But,
3863 for most targets, it is made up of code and thus requires the stack
3864 to be made executable in order for the program to work properly.
3867 @opindex Wfloat-equal
3868 @opindex Wno-float-equal
3869 Warn if floating point values are used in equality comparisons.
3871 The idea behind this is that sometimes it is convenient (for the
3872 programmer) to consider floating-point values as approximations to
3873 infinitely precise real numbers. If you are doing this, then you need
3874 to compute (by analyzing the code, or in some other way) the maximum or
3875 likely maximum error that the computation introduces, and allow for it
3876 when performing comparisons (and when producing output, but that's a
3877 different problem). In particular, instead of testing for equality, you
3878 would check to see whether the two values have ranges that overlap; and
3879 this is done with the relational operators, so equality comparisons are
3882 @item -Wtraditional @r{(C and Objective-C only)}
3883 @opindex Wtraditional
3884 @opindex Wno-traditional
3885 Warn about certain constructs that behave differently in traditional and
3886 ISO C@. Also warn about ISO C constructs that have no traditional C
3887 equivalent, and/or problematic constructs which should be avoided.
3891 Macro parameters that appear within string literals in the macro body.
3892 In traditional C macro replacement takes place within string literals,
3893 but does not in ISO C@.
3896 In traditional C, some preprocessor directives did not exist.
3897 Traditional preprocessors would only consider a line to be a directive
3898 if the @samp{#} appeared in column 1 on the line. Therefore
3899 @option{-Wtraditional} warns about directives that traditional C
3900 understands but would ignore because the @samp{#} does not appear as the
3901 first character on the line. It also suggests you hide directives like
3902 @samp{#pragma} not understood by traditional C by indenting them. Some
3903 traditional implementations would not recognize @samp{#elif}, so it
3904 suggests avoiding it altogether.
3907 A function-like macro that appears without arguments.
3910 The unary plus operator.
3913 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3914 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3915 constants.) Note, these suffixes appear in macros defined in the system
3916 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3917 Use of these macros in user code might normally lead to spurious
3918 warnings, however GCC's integrated preprocessor has enough context to
3919 avoid warning in these cases.
3922 A function declared external in one block and then used after the end of
3926 A @code{switch} statement has an operand of type @code{long}.
3929 A non-@code{static} function declaration follows a @code{static} one.
3930 This construct is not accepted by some traditional C compilers.
3933 The ISO type of an integer constant has a different width or
3934 signedness from its traditional type. This warning is only issued if
3935 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3936 typically represent bit patterns, are not warned about.
3939 Usage of ISO string concatenation is detected.
3942 Initialization of automatic aggregates.
3945 Identifier conflicts with labels. Traditional C lacks a separate
3946 namespace for labels.
3949 Initialization of unions. If the initializer is zero, the warning is
3950 omitted. This is done under the assumption that the zero initializer in
3951 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3952 initializer warnings and relies on default initialization to zero in the
3956 Conversions by prototypes between fixed/floating point values and vice
3957 versa. The absence of these prototypes when compiling with traditional
3958 C would cause serious problems. This is a subset of the possible
3959 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3962 Use of ISO C style function definitions. This warning intentionally is
3963 @emph{not} issued for prototype declarations or variadic functions
3964 because these ISO C features will appear in your code when using
3965 libiberty's traditional C compatibility macros, @code{PARAMS} and
3966 @code{VPARAMS}. This warning is also bypassed for nested functions
3967 because that feature is already a GCC extension and thus not relevant to
3968 traditional C compatibility.
3971 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3972 @opindex Wtraditional-conversion
3973 @opindex Wno-traditional-conversion
3974 Warn if a prototype causes a type conversion that is different from what
3975 would happen to the same argument in the absence of a prototype. This
3976 includes conversions of fixed point to floating and vice versa, and
3977 conversions changing the width or signedness of a fixed point argument
3978 except when the same as the default promotion.
3980 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3981 @opindex Wdeclaration-after-statement
3982 @opindex Wno-declaration-after-statement
3983 Warn when a declaration is found after a statement in a block. This
3984 construct, known from C++, was introduced with ISO C99 and is by default
3985 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3986 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3991 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3993 @item -Wno-endif-labels
3994 @opindex Wno-endif-labels
3995 @opindex Wendif-labels
3996 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4001 Warn whenever a local variable or type declaration shadows another variable,
4002 parameter, type, or class member (in C++), or whenever a built-in function
4003 is shadowed. Note that in C++, the compiler will not warn if a local variable
4004 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
4006 @item -Wlarger-than=@var{len}
4007 @opindex Wlarger-than=@var{len}
4008 @opindex Wlarger-than-@var{len}
4009 Warn whenever an object of larger than @var{len} bytes is defined.
4011 @item -Wframe-larger-than=@var{len}
4012 @opindex Wframe-larger-than
4013 Warn if the size of a function frame is larger than @var{len} bytes.
4014 The computation done to determine the stack frame size is approximate
4015 and not conservative.
4016 The actual requirements may be somewhat greater than @var{len}
4017 even if you do not get a warning. In addition, any space allocated
4018 via @code{alloca}, variable-length arrays, or related constructs
4019 is not included by the compiler when determining
4020 whether or not to issue a warning.
4022 @item -Wno-free-nonheap-object
4023 @opindex Wno-free-nonheap-object
4024 @opindex Wfree-nonheap-object
4025 Do not warn when attempting to free an object which was not allocated
4028 @item -Wstack-usage=@var{len}
4029 @opindex Wstack-usage
4030 Warn if the stack usage of a function might be larger than @var{len} bytes.
4031 The computation done to determine the stack usage is conservative.
4032 Any space allocated via @code{alloca}, variable-length arrays, or related
4033 constructs is included by the compiler when determining whether or not to
4036 The message is in keeping with the output of @option{-fstack-usage}.
4040 If the stack usage is fully static but exceeds the specified amount, it's:
4043 warning: stack usage is 1120 bytes
4046 If the stack usage is (partly) dynamic but bounded, it's:
4049 warning: stack usage might be 1648 bytes
4052 If the stack usage is (partly) dynamic and not bounded, it's:
4055 warning: stack usage might be unbounded
4059 @item -Wunsafe-loop-optimizations
4060 @opindex Wunsafe-loop-optimizations
4061 @opindex Wno-unsafe-loop-optimizations
4062 Warn if the loop cannot be optimized because the compiler could not
4063 assume anything on the bounds of the loop indices. With
4064 @option{-funsafe-loop-optimizations} warn if the compiler made
4067 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4068 @opindex Wno-pedantic-ms-format
4069 @opindex Wpedantic-ms-format
4070 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4071 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4072 depending on the MS runtime, when you are using the options @option{-Wformat}
4073 and @option{-pedantic} without gnu-extensions.
4075 @item -Wpointer-arith
4076 @opindex Wpointer-arith
4077 @opindex Wno-pointer-arith
4078 Warn about anything that depends on the ``size of'' a function type or
4079 of @code{void}. GNU C assigns these types a size of 1, for
4080 convenience in calculations with @code{void *} pointers and pointers
4081 to functions. In C++, warn also when an arithmetic operation involves
4082 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4085 @opindex Wtype-limits
4086 @opindex Wno-type-limits
4087 Warn if a comparison is always true or always false due to the limited
4088 range of the data type, but do not warn for constant expressions. For
4089 example, warn if an unsigned variable is compared against zero with
4090 @samp{<} or @samp{>=}. This warning is also enabled by
4093 @item -Wbad-function-cast @r{(C and Objective-C only)}
4094 @opindex Wbad-function-cast
4095 @opindex Wno-bad-function-cast
4096 Warn whenever a function call is cast to a non-matching type.
4097 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4099 @item -Wc++-compat @r{(C and Objective-C only)}
4100 Warn about ISO C constructs that are outside of the common subset of
4101 ISO C and ISO C++, e.g.@: request for implicit conversion from
4102 @code{void *} to a pointer to non-@code{void} type.
4104 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4105 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4106 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4107 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4108 enabled by @option{-Wall}.
4112 @opindex Wno-cast-qual
4113 Warn whenever a pointer is cast so as to remove a type qualifier from
4114 the target type. For example, warn if a @code{const char *} is cast
4115 to an ordinary @code{char *}.
4117 Also warn when making a cast which introduces a type qualifier in an
4118 unsafe way. For example, casting @code{char **} to @code{const char **}
4119 is unsafe, as in this example:
4122 /* p is char ** value. */
4123 const char **q = (const char **) p;
4124 /* Assignment of readonly string to const char * is OK. */
4126 /* Now char** pointer points to read-only memory. */
4131 @opindex Wcast-align
4132 @opindex Wno-cast-align
4133 Warn whenever a pointer is cast such that the required alignment of the
4134 target is increased. For example, warn if a @code{char *} is cast to
4135 an @code{int *} on machines where integers can only be accessed at
4136 two- or four-byte boundaries.
4138 @item -Wwrite-strings
4139 @opindex Wwrite-strings
4140 @opindex Wno-write-strings
4141 When compiling C, give string constants the type @code{const
4142 char[@var{length}]} so that copying the address of one into a
4143 non-@code{const} @code{char *} pointer will get a warning. These
4144 warnings will help you find at compile time code that can try to write
4145 into a string constant, but only if you have been very careful about
4146 using @code{const} in declarations and prototypes. Otherwise, it will
4147 just be a nuisance. This is why we did not make @option{-Wall} request
4150 When compiling C++, warn about the deprecated conversion from string
4151 literals to @code{char *}. This warning is enabled by default for C++
4156 @opindex Wno-clobbered
4157 Warn for variables that might be changed by @samp{longjmp} or
4158 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4161 @opindex Wconversion
4162 @opindex Wno-conversion
4163 Warn for implicit conversions that may alter a value. This includes
4164 conversions between real and integer, like @code{abs (x)} when
4165 @code{x} is @code{double}; conversions between signed and unsigned,
4166 like @code{unsigned ui = -1}; and conversions to smaller types, like
4167 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4168 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4169 changed by the conversion like in @code{abs (2.0)}. Warnings about
4170 conversions between signed and unsigned integers can be disabled by
4171 using @option{-Wno-sign-conversion}.
4173 For C++, also warn for confusing overload resolution for user-defined
4174 conversions; and conversions that will never use a type conversion
4175 operator: conversions to @code{void}, the same type, a base class or a
4176 reference to them. Warnings about conversions between signed and
4177 unsigned integers are disabled by default in C++ unless
4178 @option{-Wsign-conversion} is explicitly enabled.
4180 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4181 @opindex Wconversion-null
4182 @opindex Wno-conversion-null
4183 Do not warn for conversions between @code{NULL} and non-pointer
4184 types. @option{-Wconversion-null} is enabled by default.
4186 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4187 @opindex Wzero-as-null-pointer-constant
4188 @opindex Wno-zero-as-null-pointer-constant
4189 Warn when a literal '0' is used as null pointer constant. This can
4190 be useful to facilitate the conversion to @code{nullptr} in C++11.
4193 @opindex Wempty-body
4194 @opindex Wno-empty-body
4195 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4196 while} statement. This warning is also enabled by @option{-Wextra}.
4198 @item -Wenum-compare
4199 @opindex Wenum-compare
4200 @opindex Wno-enum-compare
4201 Warn about a comparison between values of different enum types. In C++
4202 this warning is enabled by default. In C this warning is enabled by
4205 @item -Wjump-misses-init @r{(C, Objective-C only)}
4206 @opindex Wjump-misses-init
4207 @opindex Wno-jump-misses-init
4208 Warn if a @code{goto} statement or a @code{switch} statement jumps
4209 forward across the initialization of a variable, or jumps backward to a
4210 label after the variable has been initialized. This only warns about
4211 variables which are initialized when they are declared. This warning is
4212 only supported for C and Objective C; in C++ this sort of branch is an
4215 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4216 can be disabled with the @option{-Wno-jump-misses-init} option.
4218 @item -Wsign-compare
4219 @opindex Wsign-compare
4220 @opindex Wno-sign-compare
4221 @cindex warning for comparison of signed and unsigned values
4222 @cindex comparison of signed and unsigned values, warning
4223 @cindex signed and unsigned values, comparison warning
4224 Warn when a comparison between signed and unsigned values could produce
4225 an incorrect result when the signed value is converted to unsigned.
4226 This warning is also enabled by @option{-Wextra}; to get the other warnings
4227 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4229 @item -Wsign-conversion
4230 @opindex Wsign-conversion
4231 @opindex Wno-sign-conversion
4232 Warn for implicit conversions that may change the sign of an integer
4233 value, like assigning a signed integer expression to an unsigned
4234 integer variable. An explicit cast silences the warning. In C, this
4235 option is enabled also by @option{-Wconversion}.
4239 @opindex Wno-address
4240 Warn about suspicious uses of memory addresses. These include using
4241 the address of a function in a conditional expression, such as
4242 @code{void func(void); if (func)}, and comparisons against the memory
4243 address of a string literal, such as @code{if (x == "abc")}. Such
4244 uses typically indicate a programmer error: the address of a function
4245 always evaluates to true, so their use in a conditional usually
4246 indicate that the programmer forgot the parentheses in a function
4247 call; and comparisons against string literals result in unspecified
4248 behavior and are not portable in C, so they usually indicate that the
4249 programmer intended to use @code{strcmp}. This warning is enabled by
4253 @opindex Wlogical-op
4254 @opindex Wno-logical-op
4255 Warn about suspicious uses of logical operators in expressions.
4256 This includes using logical operators in contexts where a
4257 bit-wise operator is likely to be expected.
4259 @item -Waggregate-return
4260 @opindex Waggregate-return
4261 @opindex Wno-aggregate-return
4262 Warn if any functions that return structures or unions are defined or
4263 called. (In languages where you can return an array, this also elicits
4266 @item -Wno-attributes
4267 @opindex Wno-attributes
4268 @opindex Wattributes
4269 Do not warn if an unexpected @code{__attribute__} is used, such as
4270 unrecognized attributes, function attributes applied to variables,
4271 etc. This will not stop errors for incorrect use of supported
4274 @item -Wno-builtin-macro-redefined
4275 @opindex Wno-builtin-macro-redefined
4276 @opindex Wbuiltin-macro-redefined
4277 Do not warn if certain built-in macros are redefined. This suppresses
4278 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4279 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4281 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4282 @opindex Wstrict-prototypes
4283 @opindex Wno-strict-prototypes
4284 Warn if a function is declared or defined without specifying the
4285 argument types. (An old-style function definition is permitted without
4286 a warning if preceded by a declaration which specifies the argument
4289 @item -Wold-style-declaration @r{(C and Objective-C only)}
4290 @opindex Wold-style-declaration
4291 @opindex Wno-old-style-declaration
4292 Warn for obsolescent usages, according to the C Standard, in a
4293 declaration. For example, warn if storage-class specifiers like
4294 @code{static} are not the first things in a declaration. This warning
4295 is also enabled by @option{-Wextra}.
4297 @item -Wold-style-definition @r{(C and Objective-C only)}
4298 @opindex Wold-style-definition
4299 @opindex Wno-old-style-definition
4300 Warn if an old-style function definition is used. A warning is given
4301 even if there is a previous prototype.
4303 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4304 @opindex Wmissing-parameter-type
4305 @opindex Wno-missing-parameter-type
4306 A function parameter is declared without a type specifier in K&R-style
4313 This warning is also enabled by @option{-Wextra}.
4315 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4316 @opindex Wmissing-prototypes
4317 @opindex Wno-missing-prototypes
4318 Warn if a global function is defined without a previous prototype
4319 declaration. This warning is issued even if the definition itself
4320 provides a prototype. The aim is to detect global functions that fail
4321 to be declared in header files.
4323 @item -Wmissing-declarations
4324 @opindex Wmissing-declarations
4325 @opindex Wno-missing-declarations
4326 Warn if a global function is defined without a previous declaration.
4327 Do so even if the definition itself provides a prototype.
4328 Use this option to detect global functions that are not declared in
4329 header files. In C++, no warnings are issued for function templates,
4330 or for inline functions, or for functions in anonymous namespaces.
4332 @item -Wmissing-field-initializers
4333 @opindex Wmissing-field-initializers
4334 @opindex Wno-missing-field-initializers
4338 Warn if a structure's initializer has some fields missing. For
4339 example, the following code would cause such a warning, because
4340 @code{x.h} is implicitly zero:
4343 struct s @{ int f, g, h; @};
4344 struct s x = @{ 3, 4 @};
4347 This option does not warn about designated initializers, so the following
4348 modification would not trigger a warning:
4351 struct s @{ int f, g, h; @};
4352 struct s x = @{ .f = 3, .g = 4 @};
4355 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4356 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4358 @item -Wmissing-format-attribute
4359 @opindex Wmissing-format-attribute
4360 @opindex Wno-missing-format-attribute
4363 Warn about function pointers which might be candidates for @code{format}
4364 attributes. Note these are only possible candidates, not absolute ones.
4365 GCC will guess that function pointers with @code{format} attributes that
4366 are used in assignment, initialization, parameter passing or return
4367 statements should have a corresponding @code{format} attribute in the
4368 resulting type. I.e.@: the left-hand side of the assignment or
4369 initialization, the type of the parameter variable, or the return type
4370 of the containing function respectively should also have a @code{format}
4371 attribute to avoid the warning.
4373 GCC will also warn about function definitions which might be
4374 candidates for @code{format} attributes. Again, these are only
4375 possible candidates. GCC will guess that @code{format} attributes
4376 might be appropriate for any function that calls a function like
4377 @code{vprintf} or @code{vscanf}, but this might not always be the
4378 case, and some functions for which @code{format} attributes are
4379 appropriate may not be detected.
4381 @item -Wno-multichar
4382 @opindex Wno-multichar
4384 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4385 Usually they indicate a typo in the user's code, as they have
4386 implementation-defined values, and should not be used in portable code.
4388 @item -Wnormalized=<none|id|nfc|nfkc>
4389 @opindex Wnormalized=
4392 @cindex character set, input normalization
4393 In ISO C and ISO C++, two identifiers are different if they are
4394 different sequences of characters. However, sometimes when characters
4395 outside the basic ASCII character set are used, you can have two
4396 different character sequences that look the same. To avoid confusion,
4397 the ISO 10646 standard sets out some @dfn{normalization rules} which
4398 when applied ensure that two sequences that look the same are turned into
4399 the same sequence. GCC can warn you if you are using identifiers which
4400 have not been normalized; this option controls that warning.
4402 There are four levels of warning that GCC supports. The default is
4403 @option{-Wnormalized=nfc}, which warns about any identifier which is
4404 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4405 recommended form for most uses.
4407 Unfortunately, there are some characters which ISO C and ISO C++ allow
4408 in identifiers that when turned into NFC aren't allowable as
4409 identifiers. That is, there's no way to use these symbols in portable
4410 ISO C or C++ and have all your identifiers in NFC@.
4411 @option{-Wnormalized=id} suppresses the warning for these characters.
4412 It is hoped that future versions of the standards involved will correct
4413 this, which is why this option is not the default.
4415 You can switch the warning off for all characters by writing
4416 @option{-Wnormalized=none}. You would only want to do this if you
4417 were using some other normalization scheme (like ``D''), because
4418 otherwise you can easily create bugs that are literally impossible to see.
4420 Some characters in ISO 10646 have distinct meanings but look identical
4421 in some fonts or display methodologies, especially once formatting has
4422 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4423 LETTER N'', will display just like a regular @code{n} which has been
4424 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4425 normalization scheme to convert all these into a standard form as
4426 well, and GCC will warn if your code is not in NFKC if you use
4427 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4428 about every identifier that contains the letter O because it might be
4429 confused with the digit 0, and so is not the default, but may be
4430 useful as a local coding convention if the programming environment is
4431 unable to be fixed to display these characters distinctly.
4433 @item -Wno-deprecated
4434 @opindex Wno-deprecated
4435 @opindex Wdeprecated
4436 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4438 @item -Wno-deprecated-declarations
4439 @opindex Wno-deprecated-declarations
4440 @opindex Wdeprecated-declarations
4441 Do not warn about uses of functions (@pxref{Function Attributes}),
4442 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4443 Attributes}) marked as deprecated by using the @code{deprecated}
4447 @opindex Wno-overflow
4449 Do not warn about compile-time overflow in constant expressions.
4451 @item -Woverride-init @r{(C and Objective-C only)}
4452 @opindex Woverride-init
4453 @opindex Wno-override-init
4457 Warn if an initialized field without side effects is overridden when
4458 using designated initializers (@pxref{Designated Inits, , Designated
4461 This warning is included in @option{-Wextra}. To get other
4462 @option{-Wextra} warnings without this one, use @samp{-Wextra
4463 -Wno-override-init}.
4468 Warn if a structure is given the packed attribute, but the packed
4469 attribute has no effect on the layout or size of the structure.
4470 Such structures may be mis-aligned for little benefit. For
4471 instance, in this code, the variable @code{f.x} in @code{struct bar}
4472 will be misaligned even though @code{struct bar} does not itself
4473 have the packed attribute:
4480 @} __attribute__((packed));
4488 @item -Wpacked-bitfield-compat
4489 @opindex Wpacked-bitfield-compat
4490 @opindex Wno-packed-bitfield-compat
4491 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4492 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4493 the change can lead to differences in the structure layout. GCC
4494 informs you when the offset of such a field has changed in GCC 4.4.
4495 For example there is no longer a 4-bit padding between field @code{a}
4496 and @code{b} in this structure:
4503 @} __attribute__ ((packed));
4506 This warning is enabled by default. Use
4507 @option{-Wno-packed-bitfield-compat} to disable this warning.
4512 Warn if padding is included in a structure, either to align an element
4513 of the structure or to align the whole structure. Sometimes when this
4514 happens it is possible to rearrange the fields of the structure to
4515 reduce the padding and so make the structure smaller.
4517 @item -Wredundant-decls
4518 @opindex Wredundant-decls
4519 @opindex Wno-redundant-decls
4520 Warn if anything is declared more than once in the same scope, even in
4521 cases where multiple declaration is valid and changes nothing.
4523 @item -Wnested-externs @r{(C and Objective-C only)}
4524 @opindex Wnested-externs
4525 @opindex Wno-nested-externs
4526 Warn if an @code{extern} declaration is encountered within a function.
4531 Warn if a function can not be inlined and it was declared as inline.
4532 Even with this option, the compiler will not warn about failures to
4533 inline functions declared in system headers.
4535 The compiler uses a variety of heuristics to determine whether or not
4536 to inline a function. For example, the compiler takes into account
4537 the size of the function being inlined and the amount of inlining
4538 that has already been done in the current function. Therefore,
4539 seemingly insignificant changes in the source program can cause the
4540 warnings produced by @option{-Winline} to appear or disappear.
4542 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4543 @opindex Wno-invalid-offsetof
4544 @opindex Winvalid-offsetof
4545 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4546 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4547 to a non-POD type is undefined. In existing C++ implementations,
4548 however, @samp{offsetof} typically gives meaningful results even when
4549 applied to certain kinds of non-POD types. (Such as a simple
4550 @samp{struct} that fails to be a POD type only by virtue of having a
4551 constructor.) This flag is for users who are aware that they are
4552 writing nonportable code and who have deliberately chosen to ignore the
4555 The restrictions on @samp{offsetof} may be relaxed in a future version
4556 of the C++ standard.
4558 @item -Wno-int-to-pointer-cast
4559 @opindex Wno-int-to-pointer-cast
4560 @opindex Wint-to-pointer-cast
4561 Suppress warnings from casts to pointer type of an integer of a
4562 different size. In C++, casting to a pointer type of smaller size is
4563 an error. @option{Wint-to-pointer-cast} is enabled by default.
4566 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4567 @opindex Wno-pointer-to-int-cast
4568 @opindex Wpointer-to-int-cast
4569 Suppress warnings from casts from a pointer to an integer type of a
4573 @opindex Winvalid-pch
4574 @opindex Wno-invalid-pch
4575 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4576 the search path but can't be used.
4580 @opindex Wno-long-long
4581 Warn if @samp{long long} type is used. This is enabled by either
4582 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4583 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4585 @item -Wvariadic-macros
4586 @opindex Wvariadic-macros
4587 @opindex Wno-variadic-macros
4588 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4589 alternate syntax when in pedantic ISO C99 mode. This is default.
4590 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4592 @item -Wvector-operation-performance
4593 @opindex Wvector-operation-performance
4594 @opindex Wno-vector-operation-performance
4595 Warn if vector operation is not implemented via SIMD capabilities of the
4596 architecture. Mainly useful for the performance tuning.
4597 Vector operation can be implemented @code{piecewise} which means that the
4598 scalar operation is performed on every vector element;
4599 @code{in parallel} which means that the vector operation is implemented
4600 using scalars of wider type, which normally is more performance efficient;
4601 and @code{as a single scalar} which means that vector fits into a
4607 Warn if variable length array is used in the code.
4608 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4609 the variable length array.
4611 @item -Wvolatile-register-var
4612 @opindex Wvolatile-register-var
4613 @opindex Wno-volatile-register-var
4614 Warn if a register variable is declared volatile. The volatile
4615 modifier does not inhibit all optimizations that may eliminate reads
4616 and/or writes to register variables. This warning is enabled by
4619 @item -Wdisabled-optimization
4620 @opindex Wdisabled-optimization
4621 @opindex Wno-disabled-optimization
4622 Warn if a requested optimization pass is disabled. This warning does
4623 not generally indicate that there is anything wrong with your code; it
4624 merely indicates that GCC's optimizers were unable to handle the code
4625 effectively. Often, the problem is that your code is too big or too
4626 complex; GCC will refuse to optimize programs when the optimization
4627 itself is likely to take inordinate amounts of time.
4629 @item -Wpointer-sign @r{(C and Objective-C only)}
4630 @opindex Wpointer-sign
4631 @opindex Wno-pointer-sign
4632 Warn for pointer argument passing or assignment with different signedness.
4633 This option is only supported for C and Objective-C@. It is implied by
4634 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4635 @option{-Wno-pointer-sign}.
4637 @item -Wstack-protector
4638 @opindex Wstack-protector
4639 @opindex Wno-stack-protector
4640 This option is only active when @option{-fstack-protector} is active. It
4641 warns about functions that will not be protected against stack smashing.
4644 @opindex Wno-mudflap
4645 Suppress warnings about constructs that cannot be instrumented by
4648 @item -Woverlength-strings
4649 @opindex Woverlength-strings
4650 @opindex Wno-overlength-strings
4651 Warn about string constants which are longer than the ``minimum
4652 maximum'' length specified in the C standard. Modern compilers
4653 generally allow string constants which are much longer than the
4654 standard's minimum limit, but very portable programs should avoid
4655 using longer strings.
4657 The limit applies @emph{after} string constant concatenation, and does
4658 not count the trailing NUL@. In C90, the limit was 509 characters; in
4659 C99, it was raised to 4095. C++98 does not specify a normative
4660 minimum maximum, so we do not diagnose overlength strings in C++@.
4662 This option is implied by @option{-pedantic}, and can be disabled with
4663 @option{-Wno-overlength-strings}.
4665 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4666 @opindex Wunsuffixed-float-constants
4668 GCC will issue a warning for any floating constant that does not have
4669 a suffix. When used together with @option{-Wsystem-headers} it will
4670 warn about such constants in system header files. This can be useful
4671 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4672 from the decimal floating-point extension to C99.
4675 @node Debugging Options
4676 @section Options for Debugging Your Program or GCC
4677 @cindex options, debugging
4678 @cindex debugging information options
4680 GCC has various special options that are used for debugging
4681 either your program or GCC:
4686 Produce debugging information in the operating system's native format
4687 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4690 On most systems that use stabs format, @option{-g} enables use of extra
4691 debugging information that only GDB can use; this extra information
4692 makes debugging work better in GDB but will probably make other debuggers
4694 refuse to read the program. If you want to control for certain whether
4695 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4696 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4698 GCC allows you to use @option{-g} with
4699 @option{-O}. The shortcuts taken by optimized code may occasionally
4700 produce surprising results: some variables you declared may not exist
4701 at all; flow of control may briefly move where you did not expect it;
4702 some statements may not be executed because they compute constant
4703 results or their values were already at hand; some statements may
4704 execute in different places because they were moved out of loops.
4706 Nevertheless it proves possible to debug optimized output. This makes
4707 it reasonable to use the optimizer for programs that might have bugs.
4709 The following options are useful when GCC is generated with the
4710 capability for more than one debugging format.
4714 Produce debugging information for use by GDB@. This means to use the
4715 most expressive format available (DWARF 2, stabs, or the native format
4716 if neither of those are supported), including GDB extensions if at all
4721 Produce debugging information in stabs format (if that is supported),
4722 without GDB extensions. This is the format used by DBX on most BSD
4723 systems. On MIPS, Alpha and System V Release 4 systems this option
4724 produces stabs debugging output which is not understood by DBX or SDB@.
4725 On System V Release 4 systems this option requires the GNU assembler.
4727 @item -feliminate-unused-debug-symbols
4728 @opindex feliminate-unused-debug-symbols
4729 Produce debugging information in stabs format (if that is supported),
4730 for only symbols that are actually used.
4732 @item -femit-class-debug-always
4733 Instead of emitting debugging information for a C++ class in only one
4734 object file, emit it in all object files using the class. This option
4735 should be used only with debuggers that are unable to handle the way GCC
4736 normally emits debugging information for classes because using this
4737 option will increase the size of debugging information by as much as a
4740 @item -fno-debug-types-section
4741 @opindex fno-debug-types-section
4742 @opindex fdebug-types-section
4743 By default when using DWARF v4 or higher type DIEs will be put into
4744 their own .debug_types section instead of making them part of the
4745 .debug_info section. It is more efficient to put them in a separate
4746 comdat sections since the linker will then be able to remove duplicates.
4747 But not all DWARF consumers support .debug_types sections yet.
4751 Produce debugging information in stabs format (if that is supported),
4752 using GNU extensions understood only by the GNU debugger (GDB)@. The
4753 use of these extensions is likely to make other debuggers crash or
4754 refuse to read the program.
4758 Produce debugging information in COFF format (if that is supported).
4759 This is the format used by SDB on most System V systems prior to
4764 Produce debugging information in XCOFF format (if that is supported).
4765 This is the format used by the DBX debugger on IBM RS/6000 systems.
4769 Produce debugging information in XCOFF format (if that is supported),
4770 using GNU extensions understood only by the GNU debugger (GDB)@. The
4771 use of these extensions is likely to make other debuggers crash or
4772 refuse to read the program, and may cause assemblers other than the GNU
4773 assembler (GAS) to fail with an error.
4775 @item -gdwarf-@var{version}
4776 @opindex gdwarf-@var{version}
4777 Produce debugging information in DWARF format (if that is
4778 supported). This is the format used by DBX on IRIX 6. The value
4779 of @var{version} may be either 2, 3 or 4; the default version is 2.
4781 Note that with DWARF version 2 some ports require, and will always
4782 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4784 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4785 for maximum benefit.
4787 @item -grecord-gcc-switches
4788 @opindex grecord-gcc-switches
4789 This switch causes the command line options, that were used to invoke the
4790 compiler and may affect code generation, to be appended to the
4791 DW_AT_producer attribute in DWARF debugging information. The options
4792 are concatenated with spaces separating them from each other and from
4793 the compiler version. See also @option{-frecord-gcc-switches} for another
4794 way of storing compiler options into the object file.
4796 @item -gno-record-gcc-switches
4797 @opindex gno-record-gcc-switches
4798 Disallow appending command line options to the DW_AT_producer attribute
4799 in DWARF debugging information. This is the default.
4801 @item -gstrict-dwarf
4802 @opindex gstrict-dwarf
4803 Disallow using extensions of later DWARF standard version than selected
4804 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4805 DWARF extensions from later standard versions is allowed.
4807 @item -gno-strict-dwarf
4808 @opindex gno-strict-dwarf
4809 Allow using extensions of later DWARF standard version than selected with
4810 @option{-gdwarf-@var{version}}.
4814 Produce debugging information in VMS debug format (if that is
4815 supported). This is the format used by DEBUG on VMS systems.
4818 @itemx -ggdb@var{level}
4819 @itemx -gstabs@var{level}
4820 @itemx -gcoff@var{level}
4821 @itemx -gxcoff@var{level}
4822 @itemx -gvms@var{level}
4823 Request debugging information and also use @var{level} to specify how
4824 much information. The default level is 2.
4826 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4829 Level 1 produces minimal information, enough for making backtraces in
4830 parts of the program that you don't plan to debug. This includes
4831 descriptions of functions and external variables, but no information
4832 about local variables and no line numbers.
4834 Level 3 includes extra information, such as all the macro definitions
4835 present in the program. Some debuggers support macro expansion when
4836 you use @option{-g3}.
4838 @option{-gdwarf-2} does not accept a concatenated debug level, because
4839 GCC used to support an option @option{-gdwarf} that meant to generate
4840 debug information in version 1 of the DWARF format (which is very
4841 different from version 2), and it would have been too confusing. That
4842 debug format is long obsolete, but the option cannot be changed now.
4843 Instead use an additional @option{-g@var{level}} option to change the
4844 debug level for DWARF.
4848 Turn off generation of debug info, if leaving out this option would have
4849 generated it, or turn it on at level 2 otherwise. The position of this
4850 argument in the command line does not matter, it takes effect after all
4851 other options are processed, and it does so only once, no matter how
4852 many times it is given. This is mainly intended to be used with
4853 @option{-fcompare-debug}.
4855 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4856 @opindex fdump-final-insns
4857 Dump the final internal representation (RTL) to @var{file}. If the
4858 optional argument is omitted (or if @var{file} is @code{.}), the name
4859 of the dump file will be determined by appending @code{.gkd} to the
4860 compilation output file name.
4862 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4863 @opindex fcompare-debug
4864 @opindex fno-compare-debug
4865 If no error occurs during compilation, run the compiler a second time,
4866 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4867 passed to the second compilation. Dump the final internal
4868 representation in both compilations, and print an error if they differ.
4870 If the equal sign is omitted, the default @option{-gtoggle} is used.
4872 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4873 and nonzero, implicitly enables @option{-fcompare-debug}. If
4874 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4875 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4878 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4879 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4880 of the final representation and the second compilation, preventing even
4881 @env{GCC_COMPARE_DEBUG} from taking effect.
4883 To verify full coverage during @option{-fcompare-debug} testing, set
4884 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4885 which GCC will reject as an invalid option in any actual compilation
4886 (rather than preprocessing, assembly or linking). To get just a
4887 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4888 not overridden} will do.
4890 @item -fcompare-debug-second
4891 @opindex fcompare-debug-second
4892 This option is implicitly passed to the compiler for the second
4893 compilation requested by @option{-fcompare-debug}, along with options to
4894 silence warnings, and omitting other options that would cause
4895 side-effect compiler outputs to files or to the standard output. Dump
4896 files and preserved temporary files are renamed so as to contain the
4897 @code{.gk} additional extension during the second compilation, to avoid
4898 overwriting those generated by the first.
4900 When this option is passed to the compiler driver, it causes the
4901 @emph{first} compilation to be skipped, which makes it useful for little
4902 other than debugging the compiler proper.
4904 @item -feliminate-dwarf2-dups
4905 @opindex feliminate-dwarf2-dups
4906 Compress DWARF2 debugging information by eliminating duplicated
4907 information about each symbol. This option only makes sense when
4908 generating DWARF2 debugging information with @option{-gdwarf-2}.
4910 @item -femit-struct-debug-baseonly
4911 Emit debug information for struct-like types
4912 only when the base name of the compilation source file
4913 matches the base name of file in which the struct was defined.
4915 This option substantially reduces the size of debugging information,
4916 but at significant potential loss in type information to the debugger.
4917 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4918 See @option{-femit-struct-debug-detailed} for more detailed control.
4920 This option works only with DWARF 2.
4922 @item -femit-struct-debug-reduced
4923 Emit debug information for struct-like types
4924 only when the base name of the compilation source file
4925 matches the base name of file in which the type was defined,
4926 unless the struct is a template or defined in a system header.
4928 This option significantly reduces the size of debugging information,
4929 with some potential loss in type information to the debugger.
4930 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4931 See @option{-femit-struct-debug-detailed} for more detailed control.
4933 This option works only with DWARF 2.
4935 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4936 Specify the struct-like types
4937 for which the compiler will generate debug information.
4938 The intent is to reduce duplicate struct debug information
4939 between different object files within the same program.
4941 This option is a detailed version of
4942 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4943 which will serve for most needs.
4945 A specification has the syntax@*
4946 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4948 The optional first word limits the specification to
4949 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4950 A struct type is used directly when it is the type of a variable, member.
4951 Indirect uses arise through pointers to structs.
4952 That is, when use of an incomplete struct would be legal, the use is indirect.
4954 @samp{struct one direct; struct two * indirect;}.
4956 The optional second word limits the specification to
4957 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4958 Generic structs are a bit complicated to explain.
4959 For C++, these are non-explicit specializations of template classes,
4960 or non-template classes within the above.
4961 Other programming languages have generics,
4962 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4964 The third word specifies the source files for those
4965 structs for which the compiler will emit debug information.
4966 The values @samp{none} and @samp{any} have the normal meaning.
4967 The value @samp{base} means that
4968 the base of name of the file in which the type declaration appears
4969 must match the base of the name of the main compilation file.
4970 In practice, this means that
4971 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4972 but types declared in other header will not.
4973 The value @samp{sys} means those types satisfying @samp{base}
4974 or declared in system or compiler headers.
4976 You may need to experiment to determine the best settings for your application.
4978 The default is @samp{-femit-struct-debug-detailed=all}.
4980 This option works only with DWARF 2.
4982 @item -fno-merge-debug-strings
4983 @opindex fmerge-debug-strings
4984 @opindex fno-merge-debug-strings
4985 Direct the linker to not merge together strings in the debugging
4986 information which are identical in different object files. Merging is
4987 not supported by all assemblers or linkers. Merging decreases the size
4988 of the debug information in the output file at the cost of increasing
4989 link processing time. Merging is enabled by default.
4991 @item -fdebug-prefix-map=@var{old}=@var{new}
4992 @opindex fdebug-prefix-map
4993 When compiling files in directory @file{@var{old}}, record debugging
4994 information describing them as in @file{@var{new}} instead.
4996 @item -fno-dwarf2-cfi-asm
4997 @opindex fdwarf2-cfi-asm
4998 @opindex fno-dwarf2-cfi-asm
4999 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5000 instead of using GAS @code{.cfi_*} directives.
5002 @cindex @command{prof}
5005 Generate extra code to write profile information suitable for the
5006 analysis program @command{prof}. You must use this option when compiling
5007 the source files you want data about, and you must also use it when
5010 @cindex @command{gprof}
5013 Generate extra code to write profile information suitable for the
5014 analysis program @command{gprof}. You must use this option when compiling
5015 the source files you want data about, and you must also use it when
5020 Makes the compiler print out each function name as it is compiled, and
5021 print some statistics about each pass when it finishes.
5024 @opindex ftime-report
5025 Makes the compiler print some statistics about the time consumed by each
5026 pass when it finishes.
5029 @opindex fmem-report
5030 Makes the compiler print some statistics about permanent memory
5031 allocation when it finishes.
5033 @item -fpre-ipa-mem-report
5034 @opindex fpre-ipa-mem-report
5035 @item -fpost-ipa-mem-report
5036 @opindex fpost-ipa-mem-report
5037 Makes the compiler print some statistics about permanent memory
5038 allocation before or after interprocedural optimization.
5041 @opindex fstack-usage
5042 Makes the compiler output stack usage information for the program, on a
5043 per-function basis. The filename for the dump is made by appending
5044 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5045 the output file, if explicitly specified and it is not an executable,
5046 otherwise it is the basename of the source file. An entry is made up
5051 The name of the function.
5055 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5058 The qualifier @code{static} means that the function manipulates the stack
5059 statically: a fixed number of bytes are allocated for the frame on function
5060 entry and released on function exit; no stack adjustments are otherwise made
5061 in the function. The second field is this fixed number of bytes.
5063 The qualifier @code{dynamic} means that the function manipulates the stack
5064 dynamically: in addition to the static allocation described above, stack
5065 adjustments are made in the body of the function, for example to push/pop
5066 arguments around function calls. If the qualifier @code{bounded} is also
5067 present, the amount of these adjustments is bounded at compile-time and
5068 the second field is an upper bound of the total amount of stack used by
5069 the function. If it is not present, the amount of these adjustments is
5070 not bounded at compile-time and the second field only represents the
5073 @item -fprofile-arcs
5074 @opindex fprofile-arcs
5075 Add code so that program flow @dfn{arcs} are instrumented. During
5076 execution the program records how many times each branch and call is
5077 executed and how many times it is taken or returns. When the compiled
5078 program exits it saves this data to a file called
5079 @file{@var{auxname}.gcda} for each source file. The data may be used for
5080 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5081 test coverage analysis (@option{-ftest-coverage}). Each object file's
5082 @var{auxname} is generated from the name of the output file, if
5083 explicitly specified and it is not the final executable, otherwise it is
5084 the basename of the source file. In both cases any suffix is removed
5085 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5086 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5087 @xref{Cross-profiling}.
5089 @cindex @command{gcov}
5093 This option is used to compile and link code instrumented for coverage
5094 analysis. The option is a synonym for @option{-fprofile-arcs}
5095 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5096 linking). See the documentation for those options for more details.
5101 Compile the source files with @option{-fprofile-arcs} plus optimization
5102 and code generation options. For test coverage analysis, use the
5103 additional @option{-ftest-coverage} option. You do not need to profile
5104 every source file in a program.
5107 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5108 (the latter implies the former).
5111 Run the program on a representative workload to generate the arc profile
5112 information. This may be repeated any number of times. You can run
5113 concurrent instances of your program, and provided that the file system
5114 supports locking, the data files will be correctly updated. Also
5115 @code{fork} calls are detected and correctly handled (double counting
5119 For profile-directed optimizations, compile the source files again with
5120 the same optimization and code generation options plus
5121 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5122 Control Optimization}).
5125 For test coverage analysis, use @command{gcov} to produce human readable
5126 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5127 @command{gcov} documentation for further information.
5131 With @option{-fprofile-arcs}, for each function of your program GCC
5132 creates a program flow graph, then finds a spanning tree for the graph.
5133 Only arcs that are not on the spanning tree have to be instrumented: the
5134 compiler adds code to count the number of times that these arcs are
5135 executed. When an arc is the only exit or only entrance to a block, the
5136 instrumentation code can be added to the block; otherwise, a new basic
5137 block must be created to hold the instrumentation code.
5140 @item -ftest-coverage
5141 @opindex ftest-coverage
5142 Produce a notes file that the @command{gcov} code-coverage utility
5143 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5144 show program coverage. Each source file's note file is called
5145 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5146 above for a description of @var{auxname} and instructions on how to
5147 generate test coverage data. Coverage data will match the source files
5148 more closely, if you do not optimize.
5150 @item -fdbg-cnt-list
5151 @opindex fdbg-cnt-list
5152 Print the name and the counter upper bound for all debug counters.
5155 @item -fdbg-cnt=@var{counter-value-list}
5157 Set the internal debug counter upper bound. @var{counter-value-list}
5158 is a comma-separated list of @var{name}:@var{value} pairs
5159 which sets the upper bound of each debug counter @var{name} to @var{value}.
5160 All debug counters have the initial upper bound of @var{UINT_MAX},
5161 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5162 e.g. With -fdbg-cnt=dce:10,tail_call:0
5163 dbg_cnt(dce) will return true only for first 10 invocations
5165 @itemx -fenable-@var{kind}-@var{pass}
5166 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5170 This is a set of debugging options that are used to explicitly disable/enable
5171 optimization passes. For compiler users, regular options for enabling/disabling
5172 passes should be used instead.
5176 @item -fdisable-ipa-@var{pass}
5177 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5178 statically invoked in the compiler multiple times, the pass name should be
5179 appended with a sequential number starting from 1.
5181 @item -fdisable-rtl-@var{pass}
5182 @item -fdisable-rtl-@var{pass}=@var{range-list}
5183 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5184 statically invoked in the compiler multiple times, the pass name should be
5185 appended with a sequential number starting from 1. @var{range-list} is a comma
5186 seperated list of function ranges or assembler names. Each range is a number
5187 pair seperated by a colon. The range is inclusive in both ends. If the range
5188 is trivial, the number pair can be simplified as a single number. If the
5189 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5190 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5191 function header of a dump file, and the pass names can be dumped by using
5192 option @option{-fdump-passes}.
5194 @item -fdisable-tree-@var{pass}
5195 @item -fdisable-tree-@var{pass}=@var{range-list}
5196 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5199 @item -fenable-ipa-@var{pass}
5200 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5201 statically invoked in the compiler multiple times, the pass name should be
5202 appended with a sequential number starting from 1.
5204 @item -fenable-rtl-@var{pass}
5205 @item -fenable-rtl-@var{pass}=@var{range-list}
5206 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5207 description and examples.
5209 @item -fenable-tree-@var{pass}
5210 @item -fenable-tree-@var{pass}=@var{range-list}
5211 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5212 of option arguments.
5216 # disable ccp1 for all functions
5218 # disable complete unroll for function whose cgraph node uid is 1
5219 -fenable-tree-cunroll=1
5220 # disable gcse2 for functions at the following ranges [1,1],
5221 # [300,400], and [400,1000]
5222 # disable gcse2 for functions foo and foo2
5223 -fdisable-rtl-gcse2=foo,foo2
5224 # disable early inlining
5225 -fdisable-tree-einline
5226 # disable ipa inlining
5227 -fdisable-ipa-inline
5228 # enable tree full unroll
5229 -fenable-tree-unroll
5235 @item -d@var{letters}
5236 @itemx -fdump-rtl-@var{pass}
5238 Says to make debugging dumps during compilation at times specified by
5239 @var{letters}. This is used for debugging the RTL-based passes of the
5240 compiler. The file names for most of the dumps are made by appending
5241 a pass number and a word to the @var{dumpname}, and the files are
5242 created in the directory of the output file. Note that the pass
5243 number is computed statically as passes get registered into the pass
5244 manager. Thus the numbering is not related to the dynamic order of
5245 execution of passes. In particular, a pass installed by a plugin
5246 could have a number over 200 even if it executed quite early.
5247 @var{dumpname} is generated from the name of the output file, if
5248 explicitly specified and it is not an executable, otherwise it is the
5249 basename of the source file. These switches may have different effects
5250 when @option{-E} is used for preprocessing.
5252 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5253 @option{-d} option @var{letters}. Here are the possible
5254 letters for use in @var{pass} and @var{letters}, and their meanings:
5258 @item -fdump-rtl-alignments
5259 @opindex fdump-rtl-alignments
5260 Dump after branch alignments have been computed.
5262 @item -fdump-rtl-asmcons
5263 @opindex fdump-rtl-asmcons
5264 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5266 @item -fdump-rtl-auto_inc_dec
5267 @opindex fdump-rtl-auto_inc_dec
5268 Dump after auto-inc-dec discovery. This pass is only run on
5269 architectures that have auto inc or auto dec instructions.
5271 @item -fdump-rtl-barriers
5272 @opindex fdump-rtl-barriers
5273 Dump after cleaning up the barrier instructions.
5275 @item -fdump-rtl-bbpart
5276 @opindex fdump-rtl-bbpart
5277 Dump after partitioning hot and cold basic blocks.
5279 @item -fdump-rtl-bbro
5280 @opindex fdump-rtl-bbro
5281 Dump after block reordering.
5283 @item -fdump-rtl-btl1
5284 @itemx -fdump-rtl-btl2
5285 @opindex fdump-rtl-btl2
5286 @opindex fdump-rtl-btl2
5287 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5288 after the two branch
5289 target load optimization passes.
5291 @item -fdump-rtl-bypass
5292 @opindex fdump-rtl-bypass
5293 Dump after jump bypassing and control flow optimizations.
5295 @item -fdump-rtl-combine
5296 @opindex fdump-rtl-combine
5297 Dump after the RTL instruction combination pass.
5299 @item -fdump-rtl-compgotos
5300 @opindex fdump-rtl-compgotos
5301 Dump after duplicating the computed gotos.
5303 @item -fdump-rtl-ce1
5304 @itemx -fdump-rtl-ce2
5305 @itemx -fdump-rtl-ce3
5306 @opindex fdump-rtl-ce1
5307 @opindex fdump-rtl-ce2
5308 @opindex fdump-rtl-ce3
5309 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5310 @option{-fdump-rtl-ce3} enable dumping after the three
5311 if conversion passes.
5313 @itemx -fdump-rtl-cprop_hardreg
5314 @opindex fdump-rtl-cprop_hardreg
5315 Dump after hard register copy propagation.
5317 @itemx -fdump-rtl-csa
5318 @opindex fdump-rtl-csa
5319 Dump after combining stack adjustments.
5321 @item -fdump-rtl-cse1
5322 @itemx -fdump-rtl-cse2
5323 @opindex fdump-rtl-cse1
5324 @opindex fdump-rtl-cse2
5325 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5326 the two common sub-expression elimination passes.
5328 @itemx -fdump-rtl-dce
5329 @opindex fdump-rtl-dce
5330 Dump after the standalone dead code elimination passes.
5332 @itemx -fdump-rtl-dbr
5333 @opindex fdump-rtl-dbr
5334 Dump after delayed branch scheduling.
5336 @item -fdump-rtl-dce1
5337 @itemx -fdump-rtl-dce2
5338 @opindex fdump-rtl-dce1
5339 @opindex fdump-rtl-dce2
5340 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5341 the two dead store elimination passes.
5344 @opindex fdump-rtl-eh
5345 Dump after finalization of EH handling code.
5347 @item -fdump-rtl-eh_ranges
5348 @opindex fdump-rtl-eh_ranges
5349 Dump after conversion of EH handling range regions.
5351 @item -fdump-rtl-expand
5352 @opindex fdump-rtl-expand
5353 Dump after RTL generation.
5355 @item -fdump-rtl-fwprop1
5356 @itemx -fdump-rtl-fwprop2
5357 @opindex fdump-rtl-fwprop1
5358 @opindex fdump-rtl-fwprop2
5359 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5360 dumping after the two forward propagation passes.
5362 @item -fdump-rtl-gcse1
5363 @itemx -fdump-rtl-gcse2
5364 @opindex fdump-rtl-gcse1
5365 @opindex fdump-rtl-gcse2
5366 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5367 after global common subexpression elimination.
5369 @item -fdump-rtl-init-regs
5370 @opindex fdump-rtl-init-regs
5371 Dump after the initialization of the registers.
5373 @item -fdump-rtl-initvals
5374 @opindex fdump-rtl-initvals
5375 Dump after the computation of the initial value sets.
5377 @itemx -fdump-rtl-into_cfglayout
5378 @opindex fdump-rtl-into_cfglayout
5379 Dump after converting to cfglayout mode.
5381 @item -fdump-rtl-ira
5382 @opindex fdump-rtl-ira
5383 Dump after iterated register allocation.
5385 @item -fdump-rtl-jump
5386 @opindex fdump-rtl-jump
5387 Dump after the second jump optimization.
5389 @item -fdump-rtl-loop2
5390 @opindex fdump-rtl-loop2
5391 @option{-fdump-rtl-loop2} enables dumping after the rtl
5392 loop optimization passes.
5394 @item -fdump-rtl-mach
5395 @opindex fdump-rtl-mach
5396 Dump after performing the machine dependent reorganization pass, if that
5399 @item -fdump-rtl-mode_sw
5400 @opindex fdump-rtl-mode_sw
5401 Dump after removing redundant mode switches.
5403 @item -fdump-rtl-rnreg
5404 @opindex fdump-rtl-rnreg
5405 Dump after register renumbering.
5407 @itemx -fdump-rtl-outof_cfglayout
5408 @opindex fdump-rtl-outof_cfglayout
5409 Dump after converting from cfglayout mode.
5411 @item -fdump-rtl-peephole2
5412 @opindex fdump-rtl-peephole2
5413 Dump after the peephole pass.
5415 @item -fdump-rtl-postreload
5416 @opindex fdump-rtl-postreload
5417 Dump after post-reload optimizations.
5419 @itemx -fdump-rtl-pro_and_epilogue
5420 @opindex fdump-rtl-pro_and_epilogue
5421 Dump after generating the function pro and epilogues.
5423 @item -fdump-rtl-regmove
5424 @opindex fdump-rtl-regmove
5425 Dump after the register move pass.
5427 @item -fdump-rtl-sched1
5428 @itemx -fdump-rtl-sched2
5429 @opindex fdump-rtl-sched1
5430 @opindex fdump-rtl-sched2
5431 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5432 after the basic block scheduling passes.
5434 @item -fdump-rtl-see
5435 @opindex fdump-rtl-see
5436 Dump after sign extension elimination.
5438 @item -fdump-rtl-seqabstr
5439 @opindex fdump-rtl-seqabstr
5440 Dump after common sequence discovery.
5442 @item -fdump-rtl-shorten
5443 @opindex fdump-rtl-shorten
5444 Dump after shortening branches.
5446 @item -fdump-rtl-sibling
5447 @opindex fdump-rtl-sibling
5448 Dump after sibling call optimizations.
5450 @item -fdump-rtl-split1
5451 @itemx -fdump-rtl-split2
5452 @itemx -fdump-rtl-split3
5453 @itemx -fdump-rtl-split4
5454 @itemx -fdump-rtl-split5
5455 @opindex fdump-rtl-split1
5456 @opindex fdump-rtl-split2
5457 @opindex fdump-rtl-split3
5458 @opindex fdump-rtl-split4
5459 @opindex fdump-rtl-split5
5460 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5461 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5462 @option{-fdump-rtl-split5} enable dumping after five rounds of
5463 instruction splitting.
5465 @item -fdump-rtl-sms
5466 @opindex fdump-rtl-sms
5467 Dump after modulo scheduling. This pass is only run on some
5470 @item -fdump-rtl-stack
5471 @opindex fdump-rtl-stack
5472 Dump after conversion from GCC's "flat register file" registers to the
5473 x87's stack-like registers. This pass is only run on x86 variants.
5475 @item -fdump-rtl-subreg1
5476 @itemx -fdump-rtl-subreg2
5477 @opindex fdump-rtl-subreg1
5478 @opindex fdump-rtl-subreg2
5479 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5480 the two subreg expansion passes.
5482 @item -fdump-rtl-unshare
5483 @opindex fdump-rtl-unshare
5484 Dump after all rtl has been unshared.
5486 @item -fdump-rtl-vartrack
5487 @opindex fdump-rtl-vartrack
5488 Dump after variable tracking.
5490 @item -fdump-rtl-vregs
5491 @opindex fdump-rtl-vregs
5492 Dump after converting virtual registers to hard registers.
5494 @item -fdump-rtl-web
5495 @opindex fdump-rtl-web
5496 Dump after live range splitting.
5498 @item -fdump-rtl-regclass
5499 @itemx -fdump-rtl-subregs_of_mode_init
5500 @itemx -fdump-rtl-subregs_of_mode_finish
5501 @itemx -fdump-rtl-dfinit
5502 @itemx -fdump-rtl-dfinish
5503 @opindex fdump-rtl-regclass
5504 @opindex fdump-rtl-subregs_of_mode_init
5505 @opindex fdump-rtl-subregs_of_mode_finish
5506 @opindex fdump-rtl-dfinit
5507 @opindex fdump-rtl-dfinish
5508 These dumps are defined but always produce empty files.
5510 @item -fdump-rtl-all
5511 @opindex fdump-rtl-all
5512 Produce all the dumps listed above.
5516 Annotate the assembler output with miscellaneous debugging information.
5520 Dump all macro definitions, at the end of preprocessing, in addition to
5525 Produce a core dump whenever an error occurs.
5529 Print statistics on memory usage, at the end of the run, to
5534 Annotate the assembler output with a comment indicating which
5535 pattern and alternative was used. The length of each instruction is
5540 Dump the RTL in the assembler output as a comment before each instruction.
5541 Also turns on @option{-dp} annotation.
5545 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5546 dump a representation of the control flow graph suitable for viewing with VCG
5547 to @file{@var{file}.@var{pass}.vcg}.
5551 Just generate RTL for a function instead of compiling it. Usually used
5552 with @option{-fdump-rtl-expand}.
5556 @opindex fdump-noaddr
5557 When doing debugging dumps, suppress address output. This makes it more
5558 feasible to use diff on debugging dumps for compiler invocations with
5559 different compiler binaries and/or different
5560 text / bss / data / heap / stack / dso start locations.
5562 @item -fdump-unnumbered
5563 @opindex fdump-unnumbered
5564 When doing debugging dumps, suppress instruction numbers and address output.
5565 This makes it more feasible to use diff on debugging dumps for compiler
5566 invocations with different options, in particular with and without
5569 @item -fdump-unnumbered-links
5570 @opindex fdump-unnumbered-links
5571 When doing debugging dumps (see @option{-d} option above), suppress
5572 instruction numbers for the links to the previous and next instructions
5575 @item -fdump-translation-unit @r{(C++ only)}
5576 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5577 @opindex fdump-translation-unit
5578 Dump a representation of the tree structure for the entire translation
5579 unit to a file. The file name is made by appending @file{.tu} to the
5580 source file name, and the file is created in the same directory as the
5581 output file. If the @samp{-@var{options}} form is used, @var{options}
5582 controls the details of the dump as described for the
5583 @option{-fdump-tree} options.
5585 @item -fdump-class-hierarchy @r{(C++ only)}
5586 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5587 @opindex fdump-class-hierarchy
5588 Dump a representation of each class's hierarchy and virtual function
5589 table layout to a file. The file name is made by appending
5590 @file{.class} to the source file name, and the file is created in the
5591 same directory as the output file. If the @samp{-@var{options}} form
5592 is used, @var{options} controls the details of the dump as described
5593 for the @option{-fdump-tree} options.
5595 @item -fdump-ipa-@var{switch}
5597 Control the dumping at various stages of inter-procedural analysis
5598 language tree to a file. The file name is generated by appending a
5599 switch specific suffix to the source file name, and the file is created
5600 in the same directory as the output file. The following dumps are
5605 Enables all inter-procedural analysis dumps.
5608 Dumps information about call-graph optimization, unused function removal,
5609 and inlining decisions.
5612 Dump after function inlining.
5617 @opindex fdump-passes
5618 Dump the list of optimization passes that are turned on and off by
5619 the current command line options.
5621 @item -fdump-statistics-@var{option}
5622 @opindex fdump-statistics
5623 Enable and control dumping of pass statistics in a separate file. The
5624 file name is generated by appending a suffix ending in
5625 @samp{.statistics} to the source file name, and the file is created in
5626 the same directory as the output file. If the @samp{-@var{option}}
5627 form is used, @samp{-stats} will cause counters to be summed over the
5628 whole compilation unit while @samp{-details} will dump every event as
5629 the passes generate them. The default with no option is to sum
5630 counters for each function compiled.
5632 @item -fdump-tree-@var{switch}
5633 @itemx -fdump-tree-@var{switch}-@var{options}
5635 Control the dumping at various stages of processing the intermediate
5636 language tree to a file. The file name is generated by appending a
5637 switch specific suffix to the source file name, and the file is
5638 created in the same directory as the output file. If the
5639 @samp{-@var{options}} form is used, @var{options} is a list of
5640 @samp{-} separated options that control the details of the dump. Not
5641 all options are applicable to all dumps, those which are not
5642 meaningful will be ignored. The following options are available
5646 Print the address of each node. Usually this is not meaningful as it
5647 changes according to the environment and source file. Its primary use
5648 is for tying up a dump file with a debug environment.
5650 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5651 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5652 use working backward from mangled names in the assembly file.
5654 Inhibit dumping of members of a scope or body of a function merely
5655 because that scope has been reached. Only dump such items when they
5656 are directly reachable by some other path. When dumping pretty-printed
5657 trees, this option inhibits dumping the bodies of control structures.
5659 Print a raw representation of the tree. By default, trees are
5660 pretty-printed into a C-like representation.
5662 Enable more detailed dumps (not honored by every dump option).
5664 Enable dumping various statistics about the pass (not honored by every dump
5667 Enable showing basic block boundaries (disabled in raw dumps).
5669 Enable showing virtual operands for every statement.
5671 Enable showing line numbers for statements.
5673 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5675 Enable showing the tree dump for each statement.
5677 Enable showing the EH region number holding each statement.
5679 Enable showing scalar evolution analysis details.
5681 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5682 and @option{lineno}.
5685 The following tree dumps are possible:
5689 @opindex fdump-tree-original
5690 Dump before any tree based optimization, to @file{@var{file}.original}.
5693 @opindex fdump-tree-optimized
5694 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5697 @opindex fdump-tree-gimple
5698 Dump each function before and after the gimplification pass to a file. The
5699 file name is made by appending @file{.gimple} to the source file name.
5702 @opindex fdump-tree-cfg
5703 Dump the control flow graph of each function to a file. The file name is
5704 made by appending @file{.cfg} to the source file name.
5707 @opindex fdump-tree-vcg
5708 Dump the control flow graph of each function to a file in VCG format. The
5709 file name is made by appending @file{.vcg} to the source file name. Note
5710 that if the file contains more than one function, the generated file cannot
5711 be used directly by VCG@. You will need to cut and paste each function's
5712 graph into its own separate file first.
5715 @opindex fdump-tree-ch
5716 Dump each function after copying loop headers. The file name is made by
5717 appending @file{.ch} to the source file name.
5720 @opindex fdump-tree-ssa
5721 Dump SSA related information to a file. The file name is made by appending
5722 @file{.ssa} to the source file name.
5725 @opindex fdump-tree-alias
5726 Dump aliasing information for each function. The file name is made by
5727 appending @file{.alias} to the source file name.
5730 @opindex fdump-tree-ccp
5731 Dump each function after CCP@. The file name is made by appending
5732 @file{.ccp} to the source file name.
5735 @opindex fdump-tree-storeccp
5736 Dump each function after STORE-CCP@. The file name is made by appending
5737 @file{.storeccp} to the source file name.
5740 @opindex fdump-tree-pre
5741 Dump trees after partial redundancy elimination. The file name is made
5742 by appending @file{.pre} to the source file name.
5745 @opindex fdump-tree-fre
5746 Dump trees after full redundancy elimination. The file name is made
5747 by appending @file{.fre} to the source file name.
5750 @opindex fdump-tree-copyprop
5751 Dump trees after copy propagation. The file name is made
5752 by appending @file{.copyprop} to the source file name.
5754 @item store_copyprop
5755 @opindex fdump-tree-store_copyprop
5756 Dump trees after store copy-propagation. The file name is made
5757 by appending @file{.store_copyprop} to the source file name.
5760 @opindex fdump-tree-dce
5761 Dump each function after dead code elimination. The file name is made by
5762 appending @file{.dce} to the source file name.
5765 @opindex fdump-tree-mudflap
5766 Dump each function after adding mudflap instrumentation. The file name is
5767 made by appending @file{.mudflap} to the source file name.
5770 @opindex fdump-tree-sra
5771 Dump each function after performing scalar replacement of aggregates. The
5772 file name is made by appending @file{.sra} to the source file name.
5775 @opindex fdump-tree-sink
5776 Dump each function after performing code sinking. The file name is made
5777 by appending @file{.sink} to the source file name.
5780 @opindex fdump-tree-dom
5781 Dump each function after applying dominator tree optimizations. The file
5782 name is made by appending @file{.dom} to the source file name.
5785 @opindex fdump-tree-dse
5786 Dump each function after applying dead store elimination. The file
5787 name is made by appending @file{.dse} to the source file name.
5790 @opindex fdump-tree-phiopt
5791 Dump each function after optimizing PHI nodes into straightline code. The file
5792 name is made by appending @file{.phiopt} to the source file name.
5795 @opindex fdump-tree-forwprop
5796 Dump each function after forward propagating single use variables. The file
5797 name is made by appending @file{.forwprop} to the source file name.
5800 @opindex fdump-tree-copyrename
5801 Dump each function after applying the copy rename optimization. The file
5802 name is made by appending @file{.copyrename} to the source file name.
5805 @opindex fdump-tree-nrv
5806 Dump each function after applying the named return value optimization on
5807 generic trees. The file name is made by appending @file{.nrv} to the source
5811 @opindex fdump-tree-vect
5812 Dump each function after applying vectorization of loops. The file name is
5813 made by appending @file{.vect} to the source file name.
5816 @opindex fdump-tree-slp
5817 Dump each function after applying vectorization of basic blocks. The file name
5818 is made by appending @file{.slp} to the source file name.
5821 @opindex fdump-tree-vrp
5822 Dump each function after Value Range Propagation (VRP). The file name
5823 is made by appending @file{.vrp} to the source file name.
5826 @opindex fdump-tree-all
5827 Enable all the available tree dumps with the flags provided in this option.
5830 @item -ftree-vectorizer-verbose=@var{n}
5831 @opindex ftree-vectorizer-verbose
5832 This option controls the amount of debugging output the vectorizer prints.
5833 This information is written to standard error, unless
5834 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5835 in which case it is output to the usual dump listing file, @file{.vect}.
5836 For @var{n}=0 no diagnostic information is reported.
5837 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5838 and the total number of loops that got vectorized.
5839 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5840 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5841 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5842 level that @option{-fdump-tree-vect-stats} uses.
5843 Higher verbosity levels mean either more information dumped for each
5844 reported loop, or same amount of information reported for more loops:
5845 if @var{n}=3, vectorizer cost model information is reported.
5846 If @var{n}=4, alignment related information is added to the reports.
5847 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5848 memory access-patterns) is added to the reports.
5849 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5850 that did not pass the first analysis phase (i.e., may not be countable, or
5851 may have complicated control-flow).
5852 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5853 If @var{n}=8, SLP related information is added to the reports.
5854 For @var{n}=9, all the information the vectorizer generates during its
5855 analysis and transformation is reported. This is the same verbosity level
5856 that @option{-fdump-tree-vect-details} uses.
5858 @item -frandom-seed=@var{string}
5859 @opindex frandom-seed
5860 This option provides a seed that GCC uses when it would otherwise use
5861 random numbers. It is used to generate certain symbol names
5862 that have to be different in every compiled file. It is also used to
5863 place unique stamps in coverage data files and the object files that
5864 produce them. You can use the @option{-frandom-seed} option to produce
5865 reproducibly identical object files.
5867 The @var{string} should be different for every file you compile.
5869 @item -fsched-verbose=@var{n}
5870 @opindex fsched-verbose
5871 On targets that use instruction scheduling, this option controls the
5872 amount of debugging output the scheduler prints. This information is
5873 written to standard error, unless @option{-fdump-rtl-sched1} or
5874 @option{-fdump-rtl-sched2} is specified, in which case it is output
5875 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5876 respectively. However for @var{n} greater than nine, the output is
5877 always printed to standard error.
5879 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5880 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5881 For @var{n} greater than one, it also output basic block probabilities,
5882 detailed ready list information and unit/insn info. For @var{n} greater
5883 than two, it includes RTL at abort point, control-flow and regions info.
5884 And for @var{n} over four, @option{-fsched-verbose} also includes
5888 @itemx -save-temps=cwd
5890 Store the usual ``temporary'' intermediate files permanently; place them
5891 in the current directory and name them based on the source file. Thus,
5892 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5893 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5894 preprocessed @file{foo.i} output file even though the compiler now
5895 normally uses an integrated preprocessor.
5897 When used in combination with the @option{-x} command line option,
5898 @option{-save-temps} is sensible enough to avoid over writing an
5899 input source file with the same extension as an intermediate file.
5900 The corresponding intermediate file may be obtained by renaming the
5901 source file before using @option{-save-temps}.
5903 If you invoke GCC in parallel, compiling several different source
5904 files that share a common base name in different subdirectories or the
5905 same source file compiled for multiple output destinations, it is
5906 likely that the different parallel compilers will interfere with each
5907 other, and overwrite the temporary files. For instance:
5910 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5911 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5914 may result in @file{foo.i} and @file{foo.o} being written to
5915 simultaneously by both compilers.
5917 @item -save-temps=obj
5918 @opindex save-temps=obj
5919 Store the usual ``temporary'' intermediate files permanently. If the
5920 @option{-o} option is used, the temporary files are based on the
5921 object file. If the @option{-o} option is not used, the
5922 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5927 gcc -save-temps=obj -c foo.c
5928 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5929 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5932 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5933 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5934 @file{dir2/yfoobar.o}.
5936 @item -time@r{[}=@var{file}@r{]}
5938 Report the CPU time taken by each subprocess in the compilation
5939 sequence. For C source files, this is the compiler proper and assembler
5940 (plus the linker if linking is done).
5942 Without the specification of an output file, the output looks like this:
5949 The first number on each line is the ``user time'', that is time spent
5950 executing the program itself. The second number is ``system time'',
5951 time spent executing operating system routines on behalf of the program.
5952 Both numbers are in seconds.
5954 With the specification of an output file, the output is appended to the
5955 named file, and it looks like this:
5958 0.12 0.01 cc1 @var{options}
5959 0.00 0.01 as @var{options}
5962 The ``user time'' and the ``system time'' are moved before the program
5963 name, and the options passed to the program are displayed, so that one
5964 can later tell what file was being compiled, and with which options.
5966 @item -fvar-tracking
5967 @opindex fvar-tracking
5968 Run variable tracking pass. It computes where variables are stored at each
5969 position in code. Better debugging information is then generated
5970 (if the debugging information format supports this information).
5972 It is enabled by default when compiling with optimization (@option{-Os},
5973 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5974 the debug info format supports it.
5976 @item -fvar-tracking-assignments
5977 @opindex fvar-tracking-assignments
5978 @opindex fno-var-tracking-assignments
5979 Annotate assignments to user variables early in the compilation and
5980 attempt to carry the annotations over throughout the compilation all the
5981 way to the end, in an attempt to improve debug information while
5982 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5984 It can be enabled even if var-tracking is disabled, in which case
5985 annotations will be created and maintained, but discarded at the end.
5987 @item -fvar-tracking-assignments-toggle
5988 @opindex fvar-tracking-assignments-toggle
5989 @opindex fno-var-tracking-assignments-toggle
5990 Toggle @option{-fvar-tracking-assignments}, in the same way that
5991 @option{-gtoggle} toggles @option{-g}.
5993 @item -print-file-name=@var{library}
5994 @opindex print-file-name
5995 Print the full absolute name of the library file @var{library} that
5996 would be used when linking---and don't do anything else. With this
5997 option, GCC does not compile or link anything; it just prints the
6000 @item -print-multi-directory
6001 @opindex print-multi-directory
6002 Print the directory name corresponding to the multilib selected by any
6003 other switches present in the command line. This directory is supposed
6004 to exist in @env{GCC_EXEC_PREFIX}.
6006 @item -print-multi-lib
6007 @opindex print-multi-lib
6008 Print the mapping from multilib directory names to compiler switches
6009 that enable them. The directory name is separated from the switches by
6010 @samp{;}, and each switch starts with an @samp{@@} instead of the
6011 @samp{-}, without spaces between multiple switches. This is supposed to
6012 ease shell-processing.
6014 @item -print-multi-os-directory
6015 @opindex print-multi-os-directory
6016 Print the path to OS libraries for the selected
6017 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6018 present in the @file{lib} subdirectory and no multilibs are used, this is
6019 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6020 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6021 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6022 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6024 @item -print-prog-name=@var{program}
6025 @opindex print-prog-name
6026 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6028 @item -print-libgcc-file-name
6029 @opindex print-libgcc-file-name
6030 Same as @option{-print-file-name=libgcc.a}.
6032 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6033 but you do want to link with @file{libgcc.a}. You can do
6036 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6039 @item -print-search-dirs
6040 @opindex print-search-dirs
6041 Print the name of the configured installation directory and a list of
6042 program and library directories @command{gcc} will search---and don't do anything else.
6044 This is useful when @command{gcc} prints the error message
6045 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6046 To resolve this you either need to put @file{cpp0} and the other compiler
6047 components where @command{gcc} expects to find them, or you can set the environment
6048 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6049 Don't forget the trailing @samp{/}.
6050 @xref{Environment Variables}.
6052 @item -print-sysroot
6053 @opindex print-sysroot
6054 Print the target sysroot directory that will be used during
6055 compilation. This is the target sysroot specified either at configure
6056 time or using the @option{--sysroot} option, possibly with an extra
6057 suffix that depends on compilation options. If no target sysroot is
6058 specified, the option prints nothing.
6060 @item -print-sysroot-headers-suffix
6061 @opindex print-sysroot-headers-suffix
6062 Print the suffix added to the target sysroot when searching for
6063 headers, or give an error if the compiler is not configured with such
6064 a suffix---and don't do anything else.
6067 @opindex dumpmachine
6068 Print the compiler's target machine (for example,
6069 @samp{i686-pc-linux-gnu})---and don't do anything else.
6072 @opindex dumpversion
6073 Print the compiler version (for example, @samp{3.0})---and don't do
6078 Print the compiler's built-in specs---and don't do anything else. (This
6079 is used when GCC itself is being built.) @xref{Spec Files}.
6081 @item -feliminate-unused-debug-types
6082 @opindex feliminate-unused-debug-types
6083 Normally, when producing DWARF2 output, GCC will emit debugging
6084 information for all types declared in a compilation
6085 unit, regardless of whether or not they are actually used
6086 in that compilation unit. Sometimes this is useful, such as
6087 if, in the debugger, you want to cast a value to a type that is
6088 not actually used in your program (but is declared). More often,
6089 however, this results in a significant amount of wasted space.
6090 With this option, GCC will avoid producing debug symbol output
6091 for types that are nowhere used in the source file being compiled.
6094 @node Optimize Options
6095 @section Options That Control Optimization
6096 @cindex optimize options
6097 @cindex options, optimization
6099 These options control various sorts of optimizations.
6101 Without any optimization option, the compiler's goal is to reduce the
6102 cost of compilation and to make debugging produce the expected
6103 results. Statements are independent: if you stop the program with a
6104 breakpoint between statements, you can then assign a new value to any
6105 variable or change the program counter to any other statement in the
6106 function and get exactly the results you would expect from the source
6109 Turning on optimization flags makes the compiler attempt to improve
6110 the performance and/or code size at the expense of compilation time
6111 and possibly the ability to debug the program.
6113 The compiler performs optimization based on the knowledge it has of the
6114 program. Compiling multiple files at once to a single output file mode allows
6115 the compiler to use information gained from all of the files when compiling
6118 Not all optimizations are controlled directly by a flag. Only
6119 optimizations that have a flag are listed in this section.
6121 Most optimizations are only enabled if an @option{-O} level is set on
6122 the command line. Otherwise they are disabled, even if individual
6123 optimization flags are specified.
6125 Depending on the target and how GCC was configured, a slightly different
6126 set of optimizations may be enabled at each @option{-O} level than
6127 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6128 to find out the exact set of optimizations that are enabled at each level.
6129 @xref{Overall Options}, for examples.
6136 Optimize. Optimizing compilation takes somewhat more time, and a lot
6137 more memory for a large function.
6139 With @option{-O}, the compiler tries to reduce code size and execution
6140 time, without performing any optimizations that take a great deal of
6143 @option{-O} turns on the following optimization flags:
6147 -fcprop-registers @gol
6150 -fdelayed-branch @gol
6152 -fguess-branch-probability @gol
6153 -fif-conversion2 @gol
6154 -fif-conversion @gol
6155 -fipa-pure-const @gol
6157 -fipa-reference @gol
6159 -fsplit-wide-types @gol
6161 -ftree-builtin-call-dce @gol
6164 -ftree-copyrename @gol
6166 -ftree-dominator-opts @gol
6168 -ftree-forwprop @gol
6176 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6177 where doing so does not interfere with debugging.
6181 Optimize even more. GCC performs nearly all supported optimizations
6182 that do not involve a space-speed tradeoff.
6183 As compared to @option{-O}, this option increases both compilation time
6184 and the performance of the generated code.
6186 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6187 also turns on the following optimization flags:
6188 @gccoptlist{-fthread-jumps @gol
6189 -falign-functions -falign-jumps @gol
6190 -falign-loops -falign-labels @gol
6193 -fcse-follow-jumps -fcse-skip-blocks @gol
6194 -fdelete-null-pointer-checks @gol
6196 -fexpensive-optimizations @gol
6197 -fgcse -fgcse-lm @gol
6198 -finline-small-functions @gol
6199 -findirect-inlining @gol
6201 -foptimize-sibling-calls @gol
6202 -fpartial-inlining @gol
6205 -freorder-blocks -freorder-functions @gol
6206 -frerun-cse-after-loop @gol
6207 -fsched-interblock -fsched-spec @gol
6208 -fschedule-insns -fschedule-insns2 @gol
6209 -fstrict-aliasing -fstrict-overflow @gol
6210 -ftree-switch-conversion -ftree-tail-merge @gol
6214 Please note the warning under @option{-fgcse} about
6215 invoking @option{-O2} on programs that use computed gotos.
6219 Optimize yet more. @option{-O3} turns on all optimizations specified
6220 by @option{-O2} and also turns on the @option{-finline-functions},
6221 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6222 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6223 @option{-fipa-cp-clone} options.
6227 Reduce compilation time and make debugging produce the expected
6228 results. This is the default.
6232 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6233 do not typically increase code size. It also performs further
6234 optimizations designed to reduce code size.
6236 @option{-Os} disables the following optimization flags:
6237 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6238 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6239 -fprefetch-loop-arrays -ftree-vect-loop-version}
6243 Disregard strict standards compliance. @option{-Ofast} enables all
6244 @option{-O3} optimizations. It also enables optimizations that are not
6245 valid for all standard compliant programs.
6246 It turns on @option{-ffast-math} and the Fortran-specific
6247 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6249 If you use multiple @option{-O} options, with or without level numbers,
6250 the last such option is the one that is effective.
6253 Options of the form @option{-f@var{flag}} specify machine-independent
6254 flags. Most flags have both positive and negative forms; the negative
6255 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6256 below, only one of the forms is listed---the one you typically will
6257 use. You can figure out the other form by either removing @samp{no-}
6260 The following options control specific optimizations. They are either
6261 activated by @option{-O} options or are related to ones that are. You
6262 can use the following flags in the rare cases when ``fine-tuning'' of
6263 optimizations to be performed is desired.
6266 @item -fno-default-inline
6267 @opindex fno-default-inline
6268 Do not make member functions inline by default merely because they are
6269 defined inside the class scope (C++ only). Otherwise, when you specify
6270 @w{@option{-O}}, member functions defined inside class scope are compiled
6271 inline by default; i.e., you don't need to add @samp{inline} in front of
6272 the member function name.
6274 @item -fno-defer-pop
6275 @opindex fno-defer-pop
6276 Always pop the arguments to each function call as soon as that function
6277 returns. For machines which must pop arguments after a function call,
6278 the compiler normally lets arguments accumulate on the stack for several
6279 function calls and pops them all at once.
6281 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6283 @item -fforward-propagate
6284 @opindex fforward-propagate
6285 Perform a forward propagation pass on RTL@. The pass tries to combine two
6286 instructions and checks if the result can be simplified. If loop unrolling
6287 is active, two passes are performed and the second is scheduled after
6290 This option is enabled by default at optimization levels @option{-O},
6291 @option{-O2}, @option{-O3}, @option{-Os}.
6293 @item -ffp-contract=@var{style}
6294 @opindex ffp-contract
6295 @option{-ffp-contract=off} disables floating-point expression contraction.
6296 @option{-ffp-contract=fast} enables floating-point expression contraction
6297 such as forming of fused multiply-add operations if the target has
6298 native support for them.
6299 @option{-ffp-contract=on} enables floating-point expression contraction
6300 if allowed by the language standard. This is currently not implemented
6301 and treated equal to @option{-ffp-contract=off}.
6303 The default is @option{-ffp-contract=fast}.
6305 @item -fomit-frame-pointer
6306 @opindex fomit-frame-pointer
6307 Don't keep the frame pointer in a register for functions that
6308 don't need one. This avoids the instructions to save, set up and
6309 restore frame pointers; it also makes an extra register available
6310 in many functions. @strong{It also makes debugging impossible on
6313 On some machines, such as the VAX, this flag has no effect, because
6314 the standard calling sequence automatically handles the frame pointer
6315 and nothing is saved by pretending it doesn't exist. The
6316 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6317 whether a target machine supports this flag. @xref{Registers,,Register
6318 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6320 Starting with GCC version 4.6, the default setting (when not optimizing for
6321 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6322 @option{-fomit-frame-pointer}. The default can be reverted to
6323 @option{-fno-omit-frame-pointer} by configuring GCC with the
6324 @option{--enable-frame-pointer} configure option.
6326 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6328 @item -foptimize-sibling-calls
6329 @opindex foptimize-sibling-calls
6330 Optimize sibling and tail recursive calls.
6332 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6336 Don't pay attention to the @code{inline} keyword. Normally this option
6337 is used to keep the compiler from expanding any functions inline.
6338 Note that if you are not optimizing, no functions can be expanded inline.
6340 @item -finline-small-functions
6341 @opindex finline-small-functions
6342 Integrate functions into their callers when their body is smaller than expected
6343 function call code (so overall size of program gets smaller). The compiler
6344 heuristically decides which functions are simple enough to be worth integrating
6347 Enabled at level @option{-O2}.
6349 @item -findirect-inlining
6350 @opindex findirect-inlining
6351 Inline also indirect calls that are discovered to be known at compile
6352 time thanks to previous inlining. This option has any effect only
6353 when inlining itself is turned on by the @option{-finline-functions}
6354 or @option{-finline-small-functions} options.
6356 Enabled at level @option{-O2}.
6358 @item -finline-functions
6359 @opindex finline-functions
6360 Integrate all simple functions into their callers. The compiler
6361 heuristically decides which functions are simple enough to be worth
6362 integrating in this way.
6364 If all calls to a given function are integrated, and the function is
6365 declared @code{static}, then the function is normally not output as
6366 assembler code in its own right.
6368 Enabled at level @option{-O3}.
6370 @item -finline-functions-called-once
6371 @opindex finline-functions-called-once
6372 Consider all @code{static} functions called once for inlining into their
6373 caller even if they are not marked @code{inline}. If a call to a given
6374 function is integrated, then the function is not output as assembler code
6377 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6379 @item -fearly-inlining
6380 @opindex fearly-inlining
6381 Inline functions marked by @code{always_inline} and functions whose body seems
6382 smaller than the function call overhead early before doing
6383 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6384 makes profiling significantly cheaper and usually inlining faster on programs
6385 having large chains of nested wrapper functions.
6391 Perform interprocedural scalar replacement of aggregates, removal of
6392 unused parameters and replacement of parameters passed by reference
6393 by parameters passed by value.
6395 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6397 @item -finline-limit=@var{n}
6398 @opindex finline-limit
6399 By default, GCC limits the size of functions that can be inlined. This flag
6400 allows coarse control of this limit. @var{n} is the size of functions that
6401 can be inlined in number of pseudo instructions.
6403 Inlining is actually controlled by a number of parameters, which may be
6404 specified individually by using @option{--param @var{name}=@var{value}}.
6405 The @option{-finline-limit=@var{n}} option sets some of these parameters
6409 @item max-inline-insns-single
6410 is set to @var{n}/2.
6411 @item max-inline-insns-auto
6412 is set to @var{n}/2.
6415 See below for a documentation of the individual
6416 parameters controlling inlining and for the defaults of these parameters.
6418 @emph{Note:} there may be no value to @option{-finline-limit} that results
6419 in default behavior.
6421 @emph{Note:} pseudo instruction represents, in this particular context, an
6422 abstract measurement of function's size. In no way does it represent a count
6423 of assembly instructions and as such its exact meaning might change from one
6424 release to an another.
6426 @item -fno-keep-inline-dllexport
6427 @opindex -fno-keep-inline-dllexport
6428 This is a more fine-grained version of @option{-fkeep-inline-functions},
6429 which applies only to functions that are declared using the @code{dllexport}
6430 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6433 @item -fkeep-inline-functions
6434 @opindex fkeep-inline-functions
6435 In C, emit @code{static} functions that are declared @code{inline}
6436 into the object file, even if the function has been inlined into all
6437 of its callers. This switch does not affect functions using the
6438 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6439 inline functions into the object file.
6441 @item -fkeep-static-consts
6442 @opindex fkeep-static-consts
6443 Emit variables declared @code{static const} when optimization isn't turned
6444 on, even if the variables aren't referenced.
6446 GCC enables this option by default. If you want to force the compiler to
6447 check if the variable was referenced, regardless of whether or not
6448 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6450 @item -fmerge-constants
6451 @opindex fmerge-constants
6452 Attempt to merge identical constants (string constants and floating point
6453 constants) across compilation units.
6455 This option is the default for optimized compilation if the assembler and
6456 linker support it. Use @option{-fno-merge-constants} to inhibit this
6459 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6461 @item -fmerge-all-constants
6462 @opindex fmerge-all-constants
6463 Attempt to merge identical constants and identical variables.
6465 This option implies @option{-fmerge-constants}. In addition to
6466 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6467 arrays or initialized constant variables with integral or floating point
6468 types. Languages like C or C++ require each variable, including multiple
6469 instances of the same variable in recursive calls, to have distinct locations,
6470 so using this option will result in non-conforming
6473 @item -fmodulo-sched
6474 @opindex fmodulo-sched
6475 Perform swing modulo scheduling immediately before the first scheduling
6476 pass. This pass looks at innermost loops and reorders their
6477 instructions by overlapping different iterations.
6479 @item -fmodulo-sched-allow-regmoves
6480 @opindex fmodulo-sched-allow-regmoves
6481 Perform more aggressive SMS based modulo scheduling with register moves
6482 allowed. By setting this flag certain anti-dependences edges will be
6483 deleted which will trigger the generation of reg-moves based on the
6484 life-range analysis. This option is effective only with
6485 @option{-fmodulo-sched} enabled.
6487 @item -fno-branch-count-reg
6488 @opindex fno-branch-count-reg
6489 Do not use ``decrement and branch'' instructions on a count register,
6490 but instead generate a sequence of instructions that decrement a
6491 register, compare it against zero, then branch based upon the result.
6492 This option is only meaningful on architectures that support such
6493 instructions, which include x86, PowerPC, IA-64 and S/390.
6495 The default is @option{-fbranch-count-reg}.
6497 @item -fno-function-cse
6498 @opindex fno-function-cse
6499 Do not put function addresses in registers; make each instruction that
6500 calls a constant function contain the function's address explicitly.
6502 This option results in less efficient code, but some strange hacks
6503 that alter the assembler output may be confused by the optimizations
6504 performed when this option is not used.
6506 The default is @option{-ffunction-cse}
6508 @item -fno-zero-initialized-in-bss
6509 @opindex fno-zero-initialized-in-bss
6510 If the target supports a BSS section, GCC by default puts variables that
6511 are initialized to zero into BSS@. This can save space in the resulting
6514 This option turns off this behavior because some programs explicitly
6515 rely on variables going to the data section. E.g., so that the
6516 resulting executable can find the beginning of that section and/or make
6517 assumptions based on that.
6519 The default is @option{-fzero-initialized-in-bss}.
6521 @item -fmudflap -fmudflapth -fmudflapir
6525 @cindex bounds checking
6527 For front-ends that support it (C and C++), instrument all risky
6528 pointer/array dereferencing operations, some standard library
6529 string/heap functions, and some other associated constructs with
6530 range/validity tests. Modules so instrumented should be immune to
6531 buffer overflows, invalid heap use, and some other classes of C/C++
6532 programming errors. The instrumentation relies on a separate runtime
6533 library (@file{libmudflap}), which will be linked into a program if
6534 @option{-fmudflap} is given at link time. Run-time behavior of the
6535 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6536 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6539 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6540 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6541 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6542 instrumentation should ignore pointer reads. This produces less
6543 instrumentation (and therefore faster execution) and still provides
6544 some protection against outright memory corrupting writes, but allows
6545 erroneously read data to propagate within a program.
6547 @item -fthread-jumps
6548 @opindex fthread-jumps
6549 Perform optimizations where we check to see if a jump branches to a
6550 location where another comparison subsumed by the first is found. If
6551 so, the first branch is redirected to either the destination of the
6552 second branch or a point immediately following it, depending on whether
6553 the condition is known to be true or false.
6555 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6557 @item -fsplit-wide-types
6558 @opindex fsplit-wide-types
6559 When using a type that occupies multiple registers, such as @code{long
6560 long} on a 32-bit system, split the registers apart and allocate them
6561 independently. This normally generates better code for those types,
6562 but may make debugging more difficult.
6564 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6567 @item -fcse-follow-jumps
6568 @opindex fcse-follow-jumps
6569 In common subexpression elimination (CSE), scan through jump instructions
6570 when the target of the jump is not reached by any other path. For
6571 example, when CSE encounters an @code{if} statement with an
6572 @code{else} clause, CSE will follow the jump when the condition
6575 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6577 @item -fcse-skip-blocks
6578 @opindex fcse-skip-blocks
6579 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6580 follow jumps which conditionally skip over blocks. When CSE
6581 encounters a simple @code{if} statement with no else clause,
6582 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6583 body of the @code{if}.
6585 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6587 @item -frerun-cse-after-loop
6588 @opindex frerun-cse-after-loop
6589 Re-run common subexpression elimination after loop optimizations has been
6592 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6596 Perform a global common subexpression elimination pass.
6597 This pass also performs global constant and copy propagation.
6599 @emph{Note:} When compiling a program using computed gotos, a GCC
6600 extension, you may get better runtime performance if you disable
6601 the global common subexpression elimination pass by adding
6602 @option{-fno-gcse} to the command line.
6604 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6608 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6609 attempt to move loads which are only killed by stores into themselves. This
6610 allows a loop containing a load/store sequence to be changed to a load outside
6611 the loop, and a copy/store within the loop.
6613 Enabled by default when gcse is enabled.
6617 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6618 global common subexpression elimination. This pass will attempt to move
6619 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6620 loops containing a load/store sequence can be changed to a load before
6621 the loop and a store after the loop.
6623 Not enabled at any optimization level.
6627 When @option{-fgcse-las} is enabled, the global common subexpression
6628 elimination pass eliminates redundant loads that come after stores to the
6629 same memory location (both partial and full redundancies).
6631 Not enabled at any optimization level.
6633 @item -fgcse-after-reload
6634 @opindex fgcse-after-reload
6635 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6636 pass is performed after reload. The purpose of this pass is to cleanup
6639 @item -funsafe-loop-optimizations
6640 @opindex funsafe-loop-optimizations
6641 If given, the loop optimizer will assume that loop indices do not
6642 overflow, and that the loops with nontrivial exit condition are not
6643 infinite. This enables a wider range of loop optimizations even if
6644 the loop optimizer itself cannot prove that these assumptions are valid.
6645 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6646 if it finds this kind of loop.
6648 @item -fcrossjumping
6649 @opindex fcrossjumping
6650 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6651 resulting code may or may not perform better than without cross-jumping.
6653 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6655 @item -fauto-inc-dec
6656 @opindex fauto-inc-dec
6657 Combine increments or decrements of addresses with memory accesses.
6658 This pass is always skipped on architectures that do not have
6659 instructions to support this. Enabled by default at @option{-O} and
6660 higher on architectures that support this.
6664 Perform dead code elimination (DCE) on RTL@.
6665 Enabled by default at @option{-O} and higher.
6669 Perform dead store elimination (DSE) on RTL@.
6670 Enabled by default at @option{-O} and higher.
6672 @item -fif-conversion
6673 @opindex fif-conversion
6674 Attempt to transform conditional jumps into branch-less equivalents. This
6675 include use of conditional moves, min, max, set flags and abs instructions, and
6676 some tricks doable by standard arithmetics. The use of conditional execution
6677 on chips where it is available is controlled by @code{if-conversion2}.
6679 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6681 @item -fif-conversion2
6682 @opindex fif-conversion2
6683 Use conditional execution (where available) to transform conditional jumps into
6684 branch-less equivalents.
6686 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6688 @item -fdelete-null-pointer-checks
6689 @opindex fdelete-null-pointer-checks
6690 Assume that programs cannot safely dereference null pointers, and that
6691 no code or data element resides there. This enables simple constant
6692 folding optimizations at all optimization levels. In addition, other
6693 optimization passes in GCC use this flag to control global dataflow
6694 analyses that eliminate useless checks for null pointers; these assume
6695 that if a pointer is checked after it has already been dereferenced,
6698 Note however that in some environments this assumption is not true.
6699 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6700 for programs which depend on that behavior.
6702 Some targets, especially embedded ones, disable this option at all levels.
6703 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6704 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6705 are enabled independently at different optimization levels.
6707 @item -fdevirtualize
6708 @opindex fdevirtualize
6709 Attempt to convert calls to virtual functions to direct calls. This
6710 is done both within a procedure and interprocedurally as part of
6711 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6712 propagation (@option{-fipa-cp}).
6713 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6715 @item -fexpensive-optimizations
6716 @opindex fexpensive-optimizations
6717 Perform a number of minor optimizations that are relatively expensive.
6719 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6723 Attempt to remove redundant extension instructions. This is especially
6724 helpful for the x86-64 architecture which implicitly zero-extends in 64-bit
6725 registers after writing to their lower 32-bit half.
6727 Enabled for x86 at levels @option{-O2}, @option{-O3}.
6729 @item -foptimize-register-move
6731 @opindex foptimize-register-move
6733 Attempt to reassign register numbers in move instructions and as
6734 operands of other simple instructions in order to maximize the amount of
6735 register tying. This is especially helpful on machines with two-operand
6738 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6741 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6743 @item -fira-algorithm=@var{algorithm}
6744 Use specified coloring algorithm for the integrated register
6745 allocator. The @var{algorithm} argument should be @code{priority} or
6746 @code{CB}. The first algorithm specifies Chow's priority coloring,
6747 the second one specifies Chaitin-Briggs coloring. The second
6748 algorithm can be unimplemented for some architectures. If it is
6749 implemented, it is the default because Chaitin-Briggs coloring as a
6750 rule generates a better code.
6752 @item -fira-region=@var{region}
6753 Use specified regions for the integrated register allocator. The
6754 @var{region} argument should be one of @code{all}, @code{mixed}, or
6755 @code{one}. The first value means using all loops as register
6756 allocation regions, the second value which is enabled by default when
6757 compiling with optimization for speed (@option{-O}, @option{-O2},
6758 @dots{}) means using all loops except for loops with small register
6759 pressure as the regions, and third one which is enabled by default for
6760 @option{-Os} or @option{-O0} means using all function as a single
6761 region. The first value can give best result for machines with small
6762 size and irregular register set, the third one results in faster and
6763 generates decent code and the smallest size code, and the second value
6764 usually give the best results in most cases and for most
6767 @item -fira-loop-pressure
6768 @opindex fira-loop-pressure
6769 Use IRA to evaluate register pressure in loops for decision to move
6770 loop invariants. Usage of this option usually results in generation
6771 of faster and smaller code on machines with big register files (>= 32
6772 registers) but it can slow compiler down.
6774 This option is enabled at level @option{-O3} for some targets.
6776 @item -fno-ira-share-save-slots
6777 @opindex fno-ira-share-save-slots
6778 Switch off sharing stack slots used for saving call used hard
6779 registers living through a call. Each hard register will get a
6780 separate stack slot and as a result function stack frame will be
6783 @item -fno-ira-share-spill-slots
6784 @opindex fno-ira-share-spill-slots
6785 Switch off sharing stack slots allocated for pseudo-registers. Each
6786 pseudo-register which did not get a hard register will get a separate
6787 stack slot and as a result function stack frame will be bigger.
6789 @item -fira-verbose=@var{n}
6790 @opindex fira-verbose
6791 Set up how verbose dump file for the integrated register allocator
6792 will be. Default value is 5. If the value is greater or equal to 10,
6793 the dump file will be stderr as if the value were @var{n} minus 10.
6795 @item -fdelayed-branch
6796 @opindex fdelayed-branch
6797 If supported for the target machine, attempt to reorder instructions
6798 to exploit instruction slots available after delayed branch
6801 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6803 @item -fschedule-insns
6804 @opindex fschedule-insns
6805 If supported for the target machine, attempt to reorder instructions to
6806 eliminate execution stalls due to required data being unavailable. This
6807 helps machines that have slow floating point or memory load instructions
6808 by allowing other instructions to be issued until the result of the load
6809 or floating point instruction is required.
6811 Enabled at levels @option{-O2}, @option{-O3}.
6813 @item -fschedule-insns2
6814 @opindex fschedule-insns2
6815 Similar to @option{-fschedule-insns}, but requests an additional pass of
6816 instruction scheduling after register allocation has been done. This is
6817 especially useful on machines with a relatively small number of
6818 registers and where memory load instructions take more than one cycle.
6820 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6822 @item -fno-sched-interblock
6823 @opindex fno-sched-interblock
6824 Don't schedule instructions across basic blocks. This is normally
6825 enabled by default when scheduling before register allocation, i.e.@:
6826 with @option{-fschedule-insns} or at @option{-O2} or higher.
6828 @item -fno-sched-spec
6829 @opindex fno-sched-spec
6830 Don't allow speculative motion of non-load instructions. This is normally
6831 enabled by default when scheduling before register allocation, i.e.@:
6832 with @option{-fschedule-insns} or at @option{-O2} or higher.
6834 @item -fsched-pressure
6835 @opindex fsched-pressure
6836 Enable register pressure sensitive insn scheduling before the register
6837 allocation. This only makes sense when scheduling before register
6838 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6839 @option{-O2} or higher. Usage of this option can improve the
6840 generated code and decrease its size by preventing register pressure
6841 increase above the number of available hard registers and as a
6842 consequence register spills in the register allocation.
6844 @item -fsched-spec-load
6845 @opindex fsched-spec-load
6846 Allow speculative motion of some load instructions. This only makes
6847 sense when scheduling before register allocation, i.e.@: with
6848 @option{-fschedule-insns} or at @option{-O2} or higher.
6850 @item -fsched-spec-load-dangerous
6851 @opindex fsched-spec-load-dangerous
6852 Allow speculative motion of more load instructions. This only makes
6853 sense when scheduling before register allocation, i.e.@: with
6854 @option{-fschedule-insns} or at @option{-O2} or higher.
6856 @item -fsched-stalled-insns
6857 @itemx -fsched-stalled-insns=@var{n}
6858 @opindex fsched-stalled-insns
6859 Define how many insns (if any) can be moved prematurely from the queue
6860 of stalled insns into the ready list, during the second scheduling pass.
6861 @option{-fno-sched-stalled-insns} means that no insns will be moved
6862 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6863 on how many queued insns can be moved prematurely.
6864 @option{-fsched-stalled-insns} without a value is equivalent to
6865 @option{-fsched-stalled-insns=1}.
6867 @item -fsched-stalled-insns-dep
6868 @itemx -fsched-stalled-insns-dep=@var{n}
6869 @opindex fsched-stalled-insns-dep
6870 Define how many insn groups (cycles) will be examined for a dependency
6871 on a stalled insn that is candidate for premature removal from the queue
6872 of stalled insns. This has an effect only during the second scheduling pass,
6873 and only if @option{-fsched-stalled-insns} is used.
6874 @option{-fno-sched-stalled-insns-dep} is equivalent to
6875 @option{-fsched-stalled-insns-dep=0}.
6876 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6877 @option{-fsched-stalled-insns-dep=1}.
6879 @item -fsched2-use-superblocks
6880 @opindex fsched2-use-superblocks
6881 When scheduling after register allocation, do use superblock scheduling
6882 algorithm. Superblock scheduling allows motion across basic block boundaries
6883 resulting on faster schedules. This option is experimental, as not all machine
6884 descriptions used by GCC model the CPU closely enough to avoid unreliable
6885 results from the algorithm.
6887 This only makes sense when scheduling after register allocation, i.e.@: with
6888 @option{-fschedule-insns2} or at @option{-O2} or higher.
6890 @item -fsched-group-heuristic
6891 @opindex fsched-group-heuristic
6892 Enable the group heuristic in the scheduler. This heuristic favors
6893 the instruction that belongs to a schedule group. This is enabled
6894 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6895 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6897 @item -fsched-critical-path-heuristic
6898 @opindex fsched-critical-path-heuristic
6899 Enable the critical-path heuristic in the scheduler. This heuristic favors
6900 instructions on the critical path. This is enabled by default when
6901 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6902 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6904 @item -fsched-spec-insn-heuristic
6905 @opindex fsched-spec-insn-heuristic
6906 Enable the speculative instruction heuristic in the scheduler. This
6907 heuristic favors speculative instructions with greater dependency weakness.
6908 This is enabled by default when scheduling is enabled, i.e.@:
6909 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6910 or at @option{-O2} or higher.
6912 @item -fsched-rank-heuristic
6913 @opindex fsched-rank-heuristic
6914 Enable the rank heuristic in the scheduler. This heuristic favors
6915 the instruction belonging to a basic block with greater size or frequency.
6916 This is enabled by default when scheduling is enabled, i.e.@:
6917 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6918 at @option{-O2} or higher.
6920 @item -fsched-last-insn-heuristic
6921 @opindex fsched-last-insn-heuristic
6922 Enable the last-instruction heuristic in the scheduler. This heuristic
6923 favors the instruction that is less dependent on the last instruction
6924 scheduled. This is enabled by default when scheduling is enabled,
6925 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6926 at @option{-O2} or higher.
6928 @item -fsched-dep-count-heuristic
6929 @opindex fsched-dep-count-heuristic
6930 Enable the dependent-count heuristic in the scheduler. This heuristic
6931 favors the instruction that has more instructions depending on it.
6932 This is enabled by default when scheduling is enabled, i.e.@:
6933 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6934 at @option{-O2} or higher.
6936 @item -freschedule-modulo-scheduled-loops
6937 @opindex freschedule-modulo-scheduled-loops
6938 The modulo scheduling comes before the traditional scheduling, if a loop
6939 was modulo scheduled we may want to prevent the later scheduling passes
6940 from changing its schedule, we use this option to control that.
6942 @item -fselective-scheduling
6943 @opindex fselective-scheduling
6944 Schedule instructions using selective scheduling algorithm. Selective
6945 scheduling runs instead of the first scheduler pass.
6947 @item -fselective-scheduling2
6948 @opindex fselective-scheduling2
6949 Schedule instructions using selective scheduling algorithm. Selective
6950 scheduling runs instead of the second scheduler pass.
6952 @item -fsel-sched-pipelining
6953 @opindex fsel-sched-pipelining
6954 Enable software pipelining of innermost loops during selective scheduling.
6955 This option has no effect until one of @option{-fselective-scheduling} or
6956 @option{-fselective-scheduling2} is turned on.
6958 @item -fsel-sched-pipelining-outer-loops
6959 @opindex fsel-sched-pipelining-outer-loops
6960 When pipelining loops during selective scheduling, also pipeline outer loops.
6961 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6964 @opindex fshrink-wrap
6965 Emit function prologues only before parts of the function that need it,
6966 rather than at the top of the function. This flag is enabled by default at
6967 @option{-O} and higher.
6969 @item -fcaller-saves
6970 @opindex fcaller-saves
6971 Enable values to be allocated in registers that will be clobbered by
6972 function calls, by emitting extra instructions to save and restore the
6973 registers around such calls. Such allocation is done only when it
6974 seems to result in better code than would otherwise be produced.
6976 This option is always enabled by default on certain machines, usually
6977 those which have no call-preserved registers to use instead.
6979 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6981 @item -fcombine-stack-adjustments
6982 @opindex fcombine-stack-adjustments
6983 Tracks stack adjustments (pushes and pops) and stack memory references
6984 and then tries to find ways to combine them.
6986 Enabled by default at @option{-O1} and higher.
6988 @item -fconserve-stack
6989 @opindex fconserve-stack
6990 Attempt to minimize stack usage. The compiler will attempt to use less
6991 stack space, even if that makes the program slower. This option
6992 implies setting the @option{large-stack-frame} parameter to 100
6993 and the @option{large-stack-frame-growth} parameter to 400.
6995 @item -ftree-reassoc
6996 @opindex ftree-reassoc
6997 Perform reassociation on trees. This flag is enabled by default
6998 at @option{-O} and higher.
7002 Perform partial redundancy elimination (PRE) on trees. This flag is
7003 enabled by default at @option{-O2} and @option{-O3}.
7005 @item -ftree-forwprop
7006 @opindex ftree-forwprop
7007 Perform forward propagation on trees. This flag is enabled by default
7008 at @option{-O} and higher.
7012 Perform full redundancy elimination (FRE) on trees. The difference
7013 between FRE and PRE is that FRE only considers expressions
7014 that are computed on all paths leading to the redundant computation.
7015 This analysis is faster than PRE, though it exposes fewer redundancies.
7016 This flag is enabled by default at @option{-O} and higher.
7018 @item -ftree-phiprop
7019 @opindex ftree-phiprop
7020 Perform hoisting of loads from conditional pointers on trees. This
7021 pass is enabled by default at @option{-O} and higher.
7023 @item -ftree-copy-prop
7024 @opindex ftree-copy-prop
7025 Perform copy propagation on trees. This pass eliminates unnecessary
7026 copy operations. This flag is enabled by default at @option{-O} and
7029 @item -fipa-pure-const
7030 @opindex fipa-pure-const
7031 Discover which functions are pure or constant.
7032 Enabled by default at @option{-O} and higher.
7034 @item -fipa-reference
7035 @opindex fipa-reference
7036 Discover which static variables do not escape cannot escape the
7038 Enabled by default at @option{-O} and higher.
7042 Perform interprocedural pointer analysis and interprocedural modification
7043 and reference analysis. This option can cause excessive memory and
7044 compile-time usage on large compilation units. It is not enabled by
7045 default at any optimization level.
7048 @opindex fipa-profile
7049 Perform interprocedural profile propagation. The functions called only from
7050 cold functions are marked as cold. Also functions executed once (such as
7051 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7052 functions and loop less parts of functions executed once are then optimized for
7054 Enabled by default at @option{-O} and higher.
7058 Perform interprocedural constant propagation.
7059 This optimization analyzes the program to determine when values passed
7060 to functions are constants and then optimizes accordingly.
7061 This optimization can substantially increase performance
7062 if the application has constants passed to functions.
7063 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7065 @item -fipa-cp-clone
7066 @opindex fipa-cp-clone
7067 Perform function cloning to make interprocedural constant propagation stronger.
7068 When enabled, interprocedural constant propagation will perform function cloning
7069 when externally visible function can be called with constant arguments.
7070 Because this optimization can create multiple copies of functions,
7071 it may significantly increase code size
7072 (see @option{--param ipcp-unit-growth=@var{value}}).
7073 This flag is enabled by default at @option{-O3}.
7075 @item -fipa-matrix-reorg
7076 @opindex fipa-matrix-reorg
7077 Perform matrix flattening and transposing.
7078 Matrix flattening tries to replace an @math{m}-dimensional matrix
7079 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7080 This reduces the level of indirection needed for accessing the elements
7081 of the matrix. The second optimization is matrix transposing that
7082 attempts to change the order of the matrix's dimensions in order to
7083 improve cache locality.
7084 Both optimizations need the @option{-fwhole-program} flag.
7085 Transposing is enabled only if profiling information is available.
7089 Perform forward store motion on trees. This flag is
7090 enabled by default at @option{-O} and higher.
7092 @item -ftree-bit-ccp
7093 @opindex ftree-bit-ccp
7094 Perform sparse conditional bit constant propagation on trees and propagate
7095 pointer alignment information.
7096 This pass only operates on local scalar variables and is enabled by default
7097 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7101 Perform sparse conditional constant propagation (CCP) on trees. This
7102 pass only operates on local scalar variables and is enabled by default
7103 at @option{-O} and higher.
7105 @item -ftree-switch-conversion
7106 Perform conversion of simple initializations in a switch to
7107 initializations from a scalar array. This flag is enabled by default
7108 at @option{-O2} and higher.
7110 @item -ftree-tail-merge
7111 Look for identical code sequences. When found, replace one with a jump to the
7112 other. This optimization is known as tail merging or cross jumping. This flag
7113 is enabled by default at @option{-O2} and higher. The run time of this pass can
7114 be limited using @option{max-tail-merge-comparisons} parameter and
7115 @option{max-tail-merge-iterations} parameter.
7119 Perform dead code elimination (DCE) on trees. This flag is enabled by
7120 default at @option{-O} and higher.
7122 @item -ftree-builtin-call-dce
7123 @opindex ftree-builtin-call-dce
7124 Perform conditional dead code elimination (DCE) for calls to builtin functions
7125 that may set @code{errno} but are otherwise side-effect free. This flag is
7126 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7129 @item -ftree-dominator-opts
7130 @opindex ftree-dominator-opts
7131 Perform a variety of simple scalar cleanups (constant/copy
7132 propagation, redundancy elimination, range propagation and expression
7133 simplification) based on a dominator tree traversal. This also
7134 performs jump threading (to reduce jumps to jumps). This flag is
7135 enabled by default at @option{-O} and higher.
7139 Perform dead store elimination (DSE) on trees. A dead store is a store into
7140 a memory location which will later be overwritten by another store without
7141 any intervening loads. In this case the earlier store can be deleted. This
7142 flag is enabled by default at @option{-O} and higher.
7146 Perform loop header copying on trees. This is beneficial since it increases
7147 effectiveness of code motion optimizations. It also saves one jump. This flag
7148 is enabled by default at @option{-O} and higher. It is not enabled
7149 for @option{-Os}, since it usually increases code size.
7151 @item -ftree-loop-optimize
7152 @opindex ftree-loop-optimize
7153 Perform loop optimizations on trees. This flag is enabled by default
7154 at @option{-O} and higher.
7156 @item -ftree-loop-linear
7157 @opindex ftree-loop-linear
7158 Perform loop interchange transformations on tree. Same as
7159 @option{-floop-interchange}. To use this code transformation, GCC has
7160 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7161 enable the Graphite loop transformation infrastructure.
7163 @item -floop-interchange
7164 @opindex floop-interchange
7165 Perform loop interchange transformations on loops. Interchanging two
7166 nested loops switches the inner and outer loops. For example, given a
7171 A(J, I) = A(J, I) * C
7175 loop interchange will transform the loop as if the user had written:
7179 A(J, I) = A(J, I) * C
7183 which can be beneficial when @code{N} is larger than the caches,
7184 because in Fortran, the elements of an array are stored in memory
7185 contiguously by column, and the original loop iterates over rows,
7186 potentially creating at each access a cache miss. This optimization
7187 applies to all the languages supported by GCC and is not limited to
7188 Fortran. To use this code transformation, GCC has to be configured
7189 with @option{--with-ppl} and @option{--with-cloog} to enable the
7190 Graphite loop transformation infrastructure.
7192 @item -floop-strip-mine
7193 @opindex floop-strip-mine
7194 Perform loop strip mining transformations on loops. Strip mining
7195 splits a loop into two nested loops. The outer loop has strides
7196 equal to the strip size and the inner loop has strides of the
7197 original loop within a strip. The strip length can be changed
7198 using the @option{loop-block-tile-size} parameter. For example,
7205 loop strip mining will transform the loop as if the user had written:
7208 DO I = II, min (II + 50, N)
7213 This optimization applies to all the languages supported by GCC and is
7214 not limited to Fortran. To use this code transformation, GCC has to
7215 be configured with @option{--with-ppl} and @option{--with-cloog} to
7216 enable the Graphite loop transformation infrastructure.
7219 @opindex floop-block
7220 Perform loop blocking transformations on loops. Blocking strip mines
7221 each loop in the loop nest such that the memory accesses of the
7222 element loops fit inside caches. The strip length can be changed
7223 using the @option{loop-block-tile-size} parameter. For example, given
7228 A(J, I) = B(I) + C(J)
7232 loop blocking will transform the loop as if the user had written:
7236 DO I = II, min (II + 50, N)
7237 DO J = JJ, min (JJ + 50, M)
7238 A(J, I) = B(I) + C(J)
7244 which can be beneficial when @code{M} is larger than the caches,
7245 because the innermost loop will iterate over a smaller amount of data
7246 that can be kept in the caches. This optimization applies to all the
7247 languages supported by GCC and is not limited to Fortran. To use this
7248 code transformation, GCC has to be configured with @option{--with-ppl}
7249 and @option{--with-cloog} to enable the Graphite loop transformation
7252 @item -fgraphite-identity
7253 @opindex fgraphite-identity
7254 Enable the identity transformation for graphite. For every SCoP we generate
7255 the polyhedral representation and transform it back to gimple. Using
7256 @option{-fgraphite-identity} we can check the costs or benefits of the
7257 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7258 are also performed by the code generator CLooG, like index splitting and
7259 dead code elimination in loops.
7261 @item -floop-flatten
7262 @opindex floop-flatten
7263 Removes the loop nesting structure: transforms the loop nest into a
7264 single loop. This transformation can be useful to vectorize all the
7265 levels of the loop nest.
7267 @item -floop-parallelize-all
7268 @opindex floop-parallelize-all
7269 Use the Graphite data dependence analysis to identify loops that can
7270 be parallelized. Parallelize all the loops that can be analyzed to
7271 not contain loop carried dependences without checking that it is
7272 profitable to parallelize the loops.
7274 @item -fcheck-data-deps
7275 @opindex fcheck-data-deps
7276 Compare the results of several data dependence analyzers. This option
7277 is used for debugging the data dependence analyzers.
7279 @item -ftree-loop-if-convert
7280 Attempt to transform conditional jumps in the innermost loops to
7281 branch-less equivalents. The intent is to remove control-flow from
7282 the innermost loops in order to improve the ability of the
7283 vectorization pass to handle these loops. This is enabled by default
7284 if vectorization is enabled.
7286 @item -ftree-loop-if-convert-stores
7287 Attempt to also if-convert conditional jumps containing memory writes.
7288 This transformation can be unsafe for multi-threaded programs as it
7289 transforms conditional memory writes into unconditional memory writes.
7292 for (i = 0; i < N; i++)
7296 would be transformed to
7298 for (i = 0; i < N; i++)
7299 A[i] = cond ? expr : A[i];
7301 potentially producing data races.
7303 @item -ftree-loop-distribution
7304 Perform loop distribution. This flag can improve cache performance on
7305 big loop bodies and allow further loop optimizations, like
7306 parallelization or vectorization, to take place. For example, the loop
7323 @item -ftree-loop-distribute-patterns
7324 Perform loop distribution of patterns that can be code generated with
7325 calls to a library. This flag is enabled by default at @option{-O3}.
7327 This pass distributes the initialization loops and generates a call to
7328 memset zero. For example, the loop
7344 and the initialization loop is transformed into a call to memset zero.
7346 @item -ftree-loop-im
7347 @opindex ftree-loop-im
7348 Perform loop invariant motion on trees. This pass moves only invariants that
7349 would be hard to handle at RTL level (function calls, operations that expand to
7350 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7351 operands of conditions that are invariant out of the loop, so that we can use
7352 just trivial invariantness analysis in loop unswitching. The pass also includes
7355 @item -ftree-loop-ivcanon
7356 @opindex ftree-loop-ivcanon
7357 Create a canonical counter for number of iterations in the loop for that
7358 determining number of iterations requires complicated analysis. Later
7359 optimizations then may determine the number easily. Useful especially
7360 in connection with unrolling.
7364 Perform induction variable optimizations (strength reduction, induction
7365 variable merging and induction variable elimination) on trees.
7367 @item -ftree-parallelize-loops=n
7368 @opindex ftree-parallelize-loops
7369 Parallelize loops, i.e., split their iteration space to run in n threads.
7370 This is only possible for loops whose iterations are independent
7371 and can be arbitrarily reordered. The optimization is only
7372 profitable on multiprocessor machines, for loops that are CPU-intensive,
7373 rather than constrained e.g.@: by memory bandwidth. This option
7374 implies @option{-pthread}, and thus is only supported on targets
7375 that have support for @option{-pthread}.
7379 Perform function-local points-to analysis on trees. This flag is
7380 enabled by default at @option{-O} and higher.
7384 Perform scalar replacement of aggregates. This pass replaces structure
7385 references with scalars to prevent committing structures to memory too
7386 early. This flag is enabled by default at @option{-O} and higher.
7388 @item -ftree-copyrename
7389 @opindex ftree-copyrename
7390 Perform copy renaming on trees. This pass attempts to rename compiler
7391 temporaries to other variables at copy locations, usually resulting in
7392 variable names which more closely resemble the original variables. This flag
7393 is enabled by default at @option{-O} and higher.
7397 Perform temporary expression replacement during the SSA->normal phase. Single
7398 use/single def temporaries are replaced at their use location with their
7399 defining expression. This results in non-GIMPLE code, but gives the expanders
7400 much more complex trees to work on resulting in better RTL generation. This is
7401 enabled by default at @option{-O} and higher.
7403 @item -ftree-vectorize
7404 @opindex ftree-vectorize
7405 Perform loop vectorization on trees. This flag is enabled by default at
7408 @item -ftree-slp-vectorize
7409 @opindex ftree-slp-vectorize
7410 Perform basic block vectorization on trees. This flag is enabled by default at
7411 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7413 @item -ftree-vect-loop-version
7414 @opindex ftree-vect-loop-version
7415 Perform loop versioning when doing loop vectorization on trees. When a loop
7416 appears to be vectorizable except that data alignment or data dependence cannot
7417 be determined at compile time then vectorized and non-vectorized versions of
7418 the loop are generated along with runtime checks for alignment or dependence
7419 to control which version is executed. This option is enabled by default
7420 except at level @option{-Os} where it is disabled.
7422 @item -fvect-cost-model
7423 @opindex fvect-cost-model
7424 Enable cost model for vectorization.
7428 Perform Value Range Propagation on trees. This is similar to the
7429 constant propagation pass, but instead of values, ranges of values are
7430 propagated. This allows the optimizers to remove unnecessary range
7431 checks like array bound checks and null pointer checks. This is
7432 enabled by default at @option{-O2} and higher. Null pointer check
7433 elimination is only done if @option{-fdelete-null-pointer-checks} is
7438 Perform tail duplication to enlarge superblock size. This transformation
7439 simplifies the control flow of the function allowing other optimizations to do
7442 @item -funroll-loops
7443 @opindex funroll-loops
7444 Unroll loops whose number of iterations can be determined at compile
7445 time or upon entry to the loop. @option{-funroll-loops} implies
7446 @option{-frerun-cse-after-loop}. This option makes code larger,
7447 and may or may not make it run faster.
7449 @item -funroll-all-loops
7450 @opindex funroll-all-loops
7451 Unroll all loops, even if their number of iterations is uncertain when
7452 the loop is entered. This usually makes programs run more slowly.
7453 @option{-funroll-all-loops} implies the same options as
7454 @option{-funroll-loops},
7456 @item -fsplit-ivs-in-unroller
7457 @opindex fsplit-ivs-in-unroller
7458 Enables expressing of values of induction variables in later iterations
7459 of the unrolled loop using the value in the first iteration. This breaks
7460 long dependency chains, thus improving efficiency of the scheduling passes.
7462 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7463 same effect. However in cases the loop body is more complicated than
7464 a single basic block, this is not reliable. It also does not work at all
7465 on some of the architectures due to restrictions in the CSE pass.
7467 This optimization is enabled by default.
7469 @item -fvariable-expansion-in-unroller
7470 @opindex fvariable-expansion-in-unroller
7471 With this option, the compiler will create multiple copies of some
7472 local variables when unrolling a loop which can result in superior code.
7474 @item -fpartial-inlining
7475 @opindex fpartial-inlining
7476 Inline parts of functions. This option has any effect only
7477 when inlining itself is turned on by the @option{-finline-functions}
7478 or @option{-finline-small-functions} options.
7480 Enabled at level @option{-O2}.
7482 @item -fpredictive-commoning
7483 @opindex fpredictive-commoning
7484 Perform predictive commoning optimization, i.e., reusing computations
7485 (especially memory loads and stores) performed in previous
7486 iterations of loops.
7488 This option is enabled at level @option{-O3}.
7490 @item -fprefetch-loop-arrays
7491 @opindex fprefetch-loop-arrays
7492 If supported by the target machine, generate instructions to prefetch
7493 memory to improve the performance of loops that access large arrays.
7495 This option may generate better or worse code; results are highly
7496 dependent on the structure of loops within the source code.
7498 Disabled at level @option{-Os}.
7501 @itemx -fno-peephole2
7502 @opindex fno-peephole
7503 @opindex fno-peephole2
7504 Disable any machine-specific peephole optimizations. The difference
7505 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7506 are implemented in the compiler; some targets use one, some use the
7507 other, a few use both.
7509 @option{-fpeephole} is enabled by default.
7510 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7512 @item -fno-guess-branch-probability
7513 @opindex fno-guess-branch-probability
7514 Do not guess branch probabilities using heuristics.
7516 GCC will use heuristics to guess branch probabilities if they are
7517 not provided by profiling feedback (@option{-fprofile-arcs}). These
7518 heuristics are based on the control flow graph. If some branch probabilities
7519 are specified by @samp{__builtin_expect}, then the heuristics will be
7520 used to guess branch probabilities for the rest of the control flow graph,
7521 taking the @samp{__builtin_expect} info into account. The interactions
7522 between the heuristics and @samp{__builtin_expect} can be complex, and in
7523 some cases, it may be useful to disable the heuristics so that the effects
7524 of @samp{__builtin_expect} are easier to understand.
7526 The default is @option{-fguess-branch-probability} at levels
7527 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7529 @item -freorder-blocks
7530 @opindex freorder-blocks
7531 Reorder basic blocks in the compiled function in order to reduce number of
7532 taken branches and improve code locality.
7534 Enabled at levels @option{-O2}, @option{-O3}.
7536 @item -freorder-blocks-and-partition
7537 @opindex freorder-blocks-and-partition
7538 In addition to reordering basic blocks in the compiled function, in order
7539 to reduce number of taken branches, partitions hot and cold basic blocks
7540 into separate sections of the assembly and .o files, to improve
7541 paging and cache locality performance.
7543 This optimization is automatically turned off in the presence of
7544 exception handling, for linkonce sections, for functions with a user-defined
7545 section attribute and on any architecture that does not support named
7548 @item -freorder-functions
7549 @opindex freorder-functions
7550 Reorder functions in the object file in order to
7551 improve code locality. This is implemented by using special
7552 subsections @code{.text.hot} for most frequently executed functions and
7553 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7554 the linker so object file format must support named sections and linker must
7555 place them in a reasonable way.
7557 Also profile feedback must be available in to make this option effective. See
7558 @option{-fprofile-arcs} for details.
7560 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7562 @item -fstrict-aliasing
7563 @opindex fstrict-aliasing
7564 Allow the compiler to assume the strictest aliasing rules applicable to
7565 the language being compiled. For C (and C++), this activates
7566 optimizations based on the type of expressions. In particular, an
7567 object of one type is assumed never to reside at the same address as an
7568 object of a different type, unless the types are almost the same. For
7569 example, an @code{unsigned int} can alias an @code{int}, but not a
7570 @code{void*} or a @code{double}. A character type may alias any other
7573 @anchor{Type-punning}Pay special attention to code like this:
7586 The practice of reading from a different union member than the one most
7587 recently written to (called ``type-punning'') is common. Even with
7588 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7589 is accessed through the union type. So, the code above will work as
7590 expected. @xref{Structures unions enumerations and bit-fields
7591 implementation}. However, this code might not:
7602 Similarly, access by taking the address, casting the resulting pointer
7603 and dereferencing the result has undefined behavior, even if the cast
7604 uses a union type, e.g.:
7608 return ((union a_union *) &d)->i;
7612 The @option{-fstrict-aliasing} option is enabled at levels
7613 @option{-O2}, @option{-O3}, @option{-Os}.
7615 @item -fstrict-overflow
7616 @opindex fstrict-overflow
7617 Allow the compiler to assume strict signed overflow rules, depending
7618 on the language being compiled. For C (and C++) this means that
7619 overflow when doing arithmetic with signed numbers is undefined, which
7620 means that the compiler may assume that it will not happen. This
7621 permits various optimizations. For example, the compiler will assume
7622 that an expression like @code{i + 10 > i} will always be true for
7623 signed @code{i}. This assumption is only valid if signed overflow is
7624 undefined, as the expression is false if @code{i + 10} overflows when
7625 using twos complement arithmetic. When this option is in effect any
7626 attempt to determine whether an operation on signed numbers will
7627 overflow must be written carefully to not actually involve overflow.
7629 This option also allows the compiler to assume strict pointer
7630 semantics: given a pointer to an object, if adding an offset to that
7631 pointer does not produce a pointer to the same object, the addition is
7632 undefined. This permits the compiler to conclude that @code{p + u >
7633 p} is always true for a pointer @code{p} and unsigned integer
7634 @code{u}. This assumption is only valid because pointer wraparound is
7635 undefined, as the expression is false if @code{p + u} overflows using
7636 twos complement arithmetic.
7638 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7639 that integer signed overflow is fully defined: it wraps. When
7640 @option{-fwrapv} is used, there is no difference between
7641 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7642 integers. With @option{-fwrapv} certain types of overflow are
7643 permitted. For example, if the compiler gets an overflow when doing
7644 arithmetic on constants, the overflowed value can still be used with
7645 @option{-fwrapv}, but not otherwise.
7647 The @option{-fstrict-overflow} option is enabled at levels
7648 @option{-O2}, @option{-O3}, @option{-Os}.
7650 @item -falign-functions
7651 @itemx -falign-functions=@var{n}
7652 @opindex falign-functions
7653 Align the start of functions to the next power-of-two greater than
7654 @var{n}, skipping up to @var{n} bytes. For instance,
7655 @option{-falign-functions=32} aligns functions to the next 32-byte
7656 boundary, but @option{-falign-functions=24} would align to the next
7657 32-byte boundary only if this can be done by skipping 23 bytes or less.
7659 @option{-fno-align-functions} and @option{-falign-functions=1} are
7660 equivalent and mean that functions will not be aligned.
7662 Some assemblers only support this flag when @var{n} is a power of two;
7663 in that case, it is rounded up.
7665 If @var{n} is not specified or is zero, use a machine-dependent default.
7667 Enabled at levels @option{-O2}, @option{-O3}.
7669 @item -falign-labels
7670 @itemx -falign-labels=@var{n}
7671 @opindex falign-labels
7672 Align all branch targets to a power-of-two boundary, skipping up to
7673 @var{n} bytes like @option{-falign-functions}. This option can easily
7674 make code slower, because it must insert dummy operations for when the
7675 branch target is reached in the usual flow of the code.
7677 @option{-fno-align-labels} and @option{-falign-labels=1} are
7678 equivalent and mean that labels will not be aligned.
7680 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7681 are greater than this value, then their values are used instead.
7683 If @var{n} is not specified or is zero, use a machine-dependent default
7684 which is very likely to be @samp{1}, meaning no alignment.
7686 Enabled at levels @option{-O2}, @option{-O3}.
7689 @itemx -falign-loops=@var{n}
7690 @opindex falign-loops
7691 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7692 like @option{-falign-functions}. The hope is that the loop will be
7693 executed many times, which will make up for any execution of the dummy
7696 @option{-fno-align-loops} and @option{-falign-loops=1} are
7697 equivalent and mean that loops will not be aligned.
7699 If @var{n} is not specified or is zero, use a machine-dependent default.
7701 Enabled at levels @option{-O2}, @option{-O3}.
7704 @itemx -falign-jumps=@var{n}
7705 @opindex falign-jumps
7706 Align branch targets to a power-of-two boundary, for branch targets
7707 where the targets can only be reached by jumping, skipping up to @var{n}
7708 bytes like @option{-falign-functions}. In this case, no dummy operations
7711 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7712 equivalent and mean that loops will not be aligned.
7714 If @var{n} is not specified or is zero, use a machine-dependent default.
7716 Enabled at levels @option{-O2}, @option{-O3}.
7718 @item -funit-at-a-time
7719 @opindex funit-at-a-time
7720 This option is left for compatibility reasons. @option{-funit-at-a-time}
7721 has no effect, while @option{-fno-unit-at-a-time} implies
7722 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7726 @item -fno-toplevel-reorder
7727 @opindex fno-toplevel-reorder
7728 Do not reorder top-level functions, variables, and @code{asm}
7729 statements. Output them in the same order that they appear in the
7730 input file. When this option is used, unreferenced static variables
7731 will not be removed. This option is intended to support existing code
7732 which relies on a particular ordering. For new code, it is better to
7735 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7736 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7741 Constructs webs as commonly used for register allocation purposes and assign
7742 each web individual pseudo register. This allows the register allocation pass
7743 to operate on pseudos directly, but also strengthens several other optimization
7744 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7745 however, make debugging impossible, since variables will no longer stay in a
7748 Enabled by default with @option{-funroll-loops}.
7750 @item -fwhole-program
7751 @opindex fwhole-program
7752 Assume that the current compilation unit represents the whole program being
7753 compiled. All public functions and variables with the exception of @code{main}
7754 and those merged by attribute @code{externally_visible} become static functions
7755 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.
7756 While this option is equivalent to proper use of the @code{static} keyword for
7757 programs consisting of a single file, in combination with option
7758 @option{-flto} this flag can be used to
7759 compile many smaller scale programs since the functions and variables become
7760 local for the whole combined compilation unit, not for the single source file
7763 This option implies @option{-fwhole-file} for Fortran programs.
7765 @item -flto[=@var{n}]
7767 This option runs the standard link-time optimizer. When invoked
7768 with source code, it generates GIMPLE (one of GCC's internal
7769 representations) and writes it to special ELF sections in the object
7770 file. When the object files are linked together, all the function
7771 bodies are read from these ELF sections and instantiated as if they
7772 had been part of the same translation unit.
7774 To use the link-time optimizer, @option{-flto} needs to be specified at
7775 compile time and during the final link. For example:
7778 gcc -c -O2 -flto foo.c
7779 gcc -c -O2 -flto bar.c
7780 gcc -o myprog -flto -O2 foo.o bar.o
7783 The first two invocations to GCC save a bytecode representation
7784 of GIMPLE into special ELF sections inside @file{foo.o} and
7785 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
7786 @file{foo.o} and @file{bar.o}, merges the two files into a single
7787 internal image, and compiles the result as usual. Since both
7788 @file{foo.o} and @file{bar.o} are merged into a single image, this
7789 causes all the interprocedural analyses and optimizations in GCC to
7790 work across the two files as if they were a single one. This means,
7791 for example, that the inliner is able to inline functions in
7792 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7794 Another (simpler) way to enable link-time optimization is:
7797 gcc -o myprog -flto -O2 foo.c bar.c
7800 The above generates bytecode for @file{foo.c} and @file{bar.c},
7801 merges them together into a single GIMPLE representation and optimizes
7802 them as usual to produce @file{myprog}.
7804 The only important thing to keep in mind is that to enable link-time
7805 optimizations the @option{-flto} flag needs to be passed to both the
7806 compile and the link commands.
7808 To make whole program optimization effective, it is necessary to make
7809 certain whole program assumptions. The compiler needs to know
7810 what functions and variables can be accessed by libraries and runtime
7811 outside of the link-time optimized unit. When supported by the linker,
7812 the linker plugin (see @option{-fuse-linker-plugin}) passes information
7813 to the compiler about used and externally visible symbols. When
7814 the linker plugin is not available, @option{-fwhole-program} should be
7815 used to allow the compiler to make these assumptions, which leads
7816 to more aggressive optimization decisions.
7818 Note that when a file is compiled with @option{-flto}, the generated
7819 object file is larger than a regular object file because it
7820 contains GIMPLE bytecodes and the usual final code. This means that
7821 object files with LTO information can be linked as normal object
7822 files; if @option{-flto} is not passed to the linker, no
7823 interprocedural optimizations are applied.
7825 Additionally, the optimization flags used to compile individual files
7826 are not necessarily related to those used at link time. For instance,
7829 gcc -c -O0 -flto foo.c
7830 gcc -c -O0 -flto bar.c
7831 gcc -o myprog -flto -O3 foo.o bar.o
7834 This produces individual object files with unoptimized assembler
7835 code, but the resulting binary @file{myprog} is optimized at
7836 @option{-O3}. If, instead, the final binary is generated without
7837 @option{-flto}, then @file{myprog} is not optimized.
7839 When producing the final binary with @option{-flto}, GCC only
7840 applies link-time optimizations to those files that contain bytecode.
7841 Therefore, you can mix and match object files and libraries with
7842 GIMPLE bytecodes and final object code. GCC automatically selects
7843 which files to optimize in LTO mode and which files to link without
7846 There are some code generation flags that GCC preserves when
7847 generating bytecodes, as they need to be used during the final link
7848 stage. Currently, the following options are saved into the GIMPLE
7849 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7850 @option{-m} target flags.
7852 At link time, these options are read in and reapplied. Note that the
7853 current implementation makes no attempt to recognize conflicting
7854 values for these options. If different files have conflicting option
7855 values (e.g., one file is compiled with @option{-fPIC} and another
7856 isn't), the compiler simply uses the last value read from the
7857 bytecode files. It is recommended, then, that you compile all the files
7858 participating in the same link with the same options.
7860 If LTO encounters objects with C linkage declared with incompatible
7861 types in separate translation units to be linked together (undefined
7862 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7863 issued. The behavior is still undefined at runtime.
7865 Another feature of LTO is that it is possible to apply interprocedural
7866 optimizations on files written in different languages. This requires
7867 support in the language front end. Currently, the C, C++ and
7868 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7869 something like this should work:
7874 gfortran -c -flto baz.f90
7875 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7878 Notice that the final link is done with @command{g++} to get the C++
7879 runtime libraries and @option{-lgfortran} is added to get the Fortran
7880 runtime libraries. In general, when mixing languages in LTO mode, you
7881 should use the same link command options as when mixing languages in a
7882 regular (non-LTO) compilation; all you need to add is @option{-flto} to
7883 all the compile and link commands.
7885 If object files containing GIMPLE bytecode are stored in a library archive, say
7886 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7887 are using a linker with plugin support. To enable this feature, use
7888 the flag @option{-fuse-linker-plugin} at link time:
7891 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7894 With the linker plugin enabled, the linker extracts the needed
7895 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
7896 to make them part of the aggregated GIMPLE image to be optimized.
7898 If you are not using a linker with plugin support and/or do not
7899 enable the linker plugin, then the objects inside @file{libfoo.a}
7900 are extracted and linked as usual, but they do not participate
7901 in the LTO optimization process.
7903 Link-time optimizations do not require the presence of the whole program to
7904 operate. If the program does not require any symbols to be exported, it is
7905 possible to combine @option{-flto} and @option{-fwhole-program} to allow
7906 the interprocedural optimizers to use more aggressive assumptions which may
7907 lead to improved optimization opportunities.
7908 Use of @option{-fwhole-program} is not needed when linker plugin is
7909 active (see @option{-fuse-linker-plugin}).
7911 The current implementation of LTO makes no
7912 attempt to generate bytecode that is portable between different
7913 types of hosts. The bytecode files are versioned and there is a
7914 strict version check, so bytecode files generated in one version of
7915 GCC will not work with an older/newer version of GCC.
7917 Link-time optimization does not work well with generation of debugging
7918 information. Combining @option{-flto} with
7919 @option{-g} is currently experimental and expected to produce wrong
7922 If you specify the optional @var{n}, the optimization and code
7923 generation done at link time is executed in parallel using @var{n}
7924 parallel jobs by utilizing an installed @command{make} program. The
7925 environment variable @env{MAKE} may be used to override the program
7926 used. The default value for @var{n} is 1.
7928 You can also specify @option{-flto=jobserver} to use GNU make's
7929 job server mode to determine the number of parallel jobs. This
7930 is useful when the Makefile calling GCC is already executing in parallel.
7931 You must prepend a @samp{+} to the command recipe in the parent Makefile
7932 for this to work. This option likely only works if @env{MAKE} is
7935 This option is disabled by default
7937 @item -flto-partition=@var{alg}
7938 @opindex flto-partition
7939 Specify the partitioning algorithm used by the link-time optimizer.
7940 The value is either @code{1to1} to specify a partitioning mirroring
7941 the original source files or @code{balanced} to specify partitioning
7942 into equally sized chunks (whenever possible). Specifying @code{none}
7943 as an algorithm disables partitioning and streaming completely. The
7944 default value is @code{balanced}.
7946 @item -flto-compression-level=@var{n}
7947 This option specifies the level of compression used for intermediate
7948 language written to LTO object files, and is only meaningful in
7949 conjunction with LTO mode (@option{-flto}). Valid
7950 values are 0 (no compression) to 9 (maximum compression). Values
7951 outside this range are clamped to either 0 or 9. If the option is not
7952 given, a default balanced compression setting is used.
7955 Prints a report with internal details on the workings of the link-time
7956 optimizer. The contents of this report vary from version to version.
7957 It is meant to be useful to GCC developers when processing object
7958 files in LTO mode (via @option{-flto}).
7960 Disabled by default.
7962 @item -fuse-linker-plugin
7963 Enables the use of a linker plugin during link-time optimization. This
7964 option relies on plugin support in the linker, which is available in gold
7965 or in GNU ld 2.21 or newer.
7967 This option enables the extraction of object files with GIMPLE bytecode out
7968 of library archives. This improves the quality of optimization by exposing
7969 more code to the link-time optimizer. This information specifies what
7970 symbols can be accessed externally (by non-LTO object or during dynamic
7971 linking). Resulting code quality improvements on binaries (and shared
7972 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7973 See @option{-flto} for a description of the effect of this flag and how to
7976 This option is enabled by default when LTO support in GCC is enabled
7977 and GCC was configured for use with
7978 a linker supporting plugins (GNU ld 2.21 or newer or gold).
7980 @item -ffat-lto-objects
7981 @opindex ffat-lto-objects
7982 Fat LTO objects are object files that contain both the intermediate language
7983 and the object code. This makes them usable for both LTO linking and normal
7984 linking. This option is effective only when compiling with @option{-flto}
7985 and is ignored at link time.
7987 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
7988 requires the complete toolchain to be aware of LTO. It requires a linker with
7989 linker plugin support for basic functionality. Additionally, nm, ar and ranlib
7990 need to support linker plugins to allow a full-featured build environment
7991 (capable of building static libraries etc).
7993 The default is @option{-ffat-lto-objects} but this default is intended to
7994 change in future releases when linker plugin enabled environments become more
7997 @item -fcompare-elim
7998 @opindex fcompare-elim
7999 After register allocation and post-register allocation instruction splitting,
8000 identify arithmetic instructions that compute processor flags similar to a
8001 comparison operation based on that arithmetic. If possible, eliminate the
8002 explicit comparison operation.
8004 This pass only applies to certain targets that cannot explicitly represent
8005 the comparison operation before register allocation is complete.
8007 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8009 @item -fcprop-registers
8010 @opindex fcprop-registers
8011 After register allocation and post-register allocation instruction splitting,
8012 we perform a copy-propagation pass to try to reduce scheduling dependencies
8013 and occasionally eliminate the copy.
8015 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8017 @item -fprofile-correction
8018 @opindex fprofile-correction
8019 Profiles collected using an instrumented binary for multi-threaded programs may
8020 be inconsistent due to missed counter updates. When this option is specified,
8021 GCC will use heuristics to correct or smooth out such inconsistencies. By
8022 default, GCC will emit an error message when an inconsistent profile is detected.
8024 @item -fprofile-dir=@var{path}
8025 @opindex fprofile-dir
8027 Set the directory to search for the profile data files in to @var{path}.
8028 This option affects only the profile data generated by
8029 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8030 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8031 and its related options. Both absolute and relative paths can be used.
8032 By default, GCC will use the current directory as @var{path}, thus the
8033 profile data file will appear in the same directory as the object file.
8035 @item -fprofile-generate
8036 @itemx -fprofile-generate=@var{path}
8037 @opindex fprofile-generate
8039 Enable options usually used for instrumenting application to produce
8040 profile useful for later recompilation with profile feedback based
8041 optimization. You must use @option{-fprofile-generate} both when
8042 compiling and when linking your program.
8044 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
8046 If @var{path} is specified, GCC will look at the @var{path} to find
8047 the profile feedback data files. See @option{-fprofile-dir}.
8050 @itemx -fprofile-use=@var{path}
8051 @opindex fprofile-use
8052 Enable profile feedback directed optimizations, and optimizations
8053 generally profitable only with profile feedback available.
8055 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8056 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
8058 By default, GCC emits an error message if the feedback profiles do not
8059 match the source code. This error can be turned into a warning by using
8060 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8063 If @var{path} is specified, GCC will look at the @var{path} to find
8064 the profile feedback data files. See @option{-fprofile-dir}.
8067 The following options control compiler behavior regarding floating
8068 point arithmetic. These options trade off between speed and
8069 correctness. All must be specifically enabled.
8073 @opindex ffloat-store
8074 Do not store floating point variables in registers, and inhibit other
8075 options that might change whether a floating point value is taken from a
8078 @cindex floating point precision
8079 This option prevents undesirable excess precision on machines such as
8080 the 68000 where the floating registers (of the 68881) keep more
8081 precision than a @code{double} is supposed to have. Similarly for the
8082 x86 architecture. For most programs, the excess precision does only
8083 good, but a few programs rely on the precise definition of IEEE floating
8084 point. Use @option{-ffloat-store} for such programs, after modifying
8085 them to store all pertinent intermediate computations into variables.
8087 @item -fexcess-precision=@var{style}
8088 @opindex fexcess-precision
8089 This option allows further control over excess precision on machines
8090 where floating-point registers have more precision than the IEEE
8091 @code{float} and @code{double} types and the processor does not
8092 support operations rounding to those types. By default,
8093 @option{-fexcess-precision=fast} is in effect; this means that
8094 operations are carried out in the precision of the registers and that
8095 it is unpredictable when rounding to the types specified in the source
8096 code takes place. When compiling C, if
8097 @option{-fexcess-precision=standard} is specified then excess
8098 precision will follow the rules specified in ISO C99; in particular,
8099 both casts and assignments cause values to be rounded to their
8100 semantic types (whereas @option{-ffloat-store} only affects
8101 assignments). This option is enabled by default for C if a strict
8102 conformance option such as @option{-std=c99} is used.
8105 @option{-fexcess-precision=standard} is not implemented for languages
8106 other than C, and has no effect if
8107 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8108 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8109 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8110 semantics apply without excess precision, and in the latter, rounding
8115 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8116 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8117 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8119 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8121 This option is not turned on by any @option{-O} option besides
8122 @option{-Ofast} since it can result in incorrect output for programs
8123 which depend on an exact implementation of IEEE or ISO rules/specifications
8124 for math functions. It may, however, yield faster code for programs
8125 that do not require the guarantees of these specifications.
8127 @item -fno-math-errno
8128 @opindex fno-math-errno
8129 Do not set ERRNO after calling math functions that are executed
8130 with a single instruction, e.g., sqrt. A program that relies on
8131 IEEE exceptions for math error handling may want to use this flag
8132 for speed while maintaining IEEE arithmetic compatibility.
8134 This option is not turned on by any @option{-O} option since
8135 it can result in incorrect output for programs which depend on
8136 an exact implementation of IEEE or ISO rules/specifications for
8137 math functions. It may, however, yield faster code for programs
8138 that do not require the guarantees of these specifications.
8140 The default is @option{-fmath-errno}.
8142 On Darwin systems, the math library never sets @code{errno}. There is
8143 therefore no reason for the compiler to consider the possibility that
8144 it might, and @option{-fno-math-errno} is the default.
8146 @item -funsafe-math-optimizations
8147 @opindex funsafe-math-optimizations
8149 Allow optimizations for floating-point arithmetic that (a) assume
8150 that arguments and results are valid and (b) may violate IEEE or
8151 ANSI standards. When used at link-time, it may include libraries
8152 or startup files that change the default FPU control word or other
8153 similar optimizations.
8155 This option is not turned on by any @option{-O} option since
8156 it can result in incorrect output for programs which depend on
8157 an exact implementation of IEEE or ISO rules/specifications for
8158 math functions. It may, however, yield faster code for programs
8159 that do not require the guarantees of these specifications.
8160 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8161 @option{-fassociative-math} and @option{-freciprocal-math}.
8163 The default is @option{-fno-unsafe-math-optimizations}.
8165 @item -fassociative-math
8166 @opindex fassociative-math
8168 Allow re-association of operands in series of floating-point operations.
8169 This violates the ISO C and C++ language standard by possibly changing
8170 computation result. NOTE: re-ordering may change the sign of zero as
8171 well as ignore NaNs and inhibit or create underflow or overflow (and
8172 thus cannot be used on a code which relies on rounding behavior like
8173 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8174 and thus may not be used when ordered comparisons are required.
8175 This option requires that both @option{-fno-signed-zeros} and
8176 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8177 much sense with @option{-frounding-math}. For Fortran the option
8178 is automatically enabled when both @option{-fno-signed-zeros} and
8179 @option{-fno-trapping-math} are in effect.
8181 The default is @option{-fno-associative-math}.
8183 @item -freciprocal-math
8184 @opindex freciprocal-math
8186 Allow the reciprocal of a value to be used instead of dividing by
8187 the value if this enables optimizations. For example @code{x / y}
8188 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8189 is subject to common subexpression elimination. Note that this loses
8190 precision and increases the number of flops operating on the value.
8192 The default is @option{-fno-reciprocal-math}.
8194 @item -ffinite-math-only
8195 @opindex ffinite-math-only
8196 Allow optimizations for floating-point arithmetic that assume
8197 that arguments and results are not NaNs or +-Infs.
8199 This option is not turned on by any @option{-O} option since
8200 it can result in incorrect output for programs which depend on
8201 an exact implementation of IEEE or ISO rules/specifications for
8202 math functions. It may, however, yield faster code for programs
8203 that do not require the guarantees of these specifications.
8205 The default is @option{-fno-finite-math-only}.
8207 @item -fno-signed-zeros
8208 @opindex fno-signed-zeros
8209 Allow optimizations for floating point arithmetic that ignore the
8210 signedness of zero. IEEE arithmetic specifies the behavior of
8211 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8212 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8213 This option implies that the sign of a zero result isn't significant.
8215 The default is @option{-fsigned-zeros}.
8217 @item -fno-trapping-math
8218 @opindex fno-trapping-math
8219 Compile code assuming that floating-point operations cannot generate
8220 user-visible traps. These traps include division by zero, overflow,
8221 underflow, inexact result and invalid operation. This option requires
8222 that @option{-fno-signaling-nans} be in effect. Setting this option may
8223 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8225 This option should never be turned on by any @option{-O} option since
8226 it can result in incorrect output for programs which depend on
8227 an exact implementation of IEEE or ISO rules/specifications for
8230 The default is @option{-ftrapping-math}.
8232 @item -frounding-math
8233 @opindex frounding-math
8234 Disable transformations and optimizations that assume default floating
8235 point rounding behavior. This is round-to-zero for all floating point
8236 to integer conversions, and round-to-nearest for all other arithmetic
8237 truncations. This option should be specified for programs that change
8238 the FP rounding mode dynamically, or that may be executed with a
8239 non-default rounding mode. This option disables constant folding of
8240 floating point expressions at compile-time (which may be affected by
8241 rounding mode) and arithmetic transformations that are unsafe in the
8242 presence of sign-dependent rounding modes.
8244 The default is @option{-fno-rounding-math}.
8246 This option is experimental and does not currently guarantee to
8247 disable all GCC optimizations that are affected by rounding mode.
8248 Future versions of GCC may provide finer control of this setting
8249 using C99's @code{FENV_ACCESS} pragma. This command line option
8250 will be used to specify the default state for @code{FENV_ACCESS}.
8252 @item -fsignaling-nans
8253 @opindex fsignaling-nans
8254 Compile code assuming that IEEE signaling NaNs may generate user-visible
8255 traps during floating-point operations. Setting this option disables
8256 optimizations that may change the number of exceptions visible with
8257 signaling NaNs. This option implies @option{-ftrapping-math}.
8259 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8262 The default is @option{-fno-signaling-nans}.
8264 This option is experimental and does not currently guarantee to
8265 disable all GCC optimizations that affect signaling NaN behavior.
8267 @item -fsingle-precision-constant
8268 @opindex fsingle-precision-constant
8269 Treat floating point constant as single precision constant instead of
8270 implicitly converting it to double precision constant.
8272 @item -fcx-limited-range
8273 @opindex fcx-limited-range
8274 When enabled, this option states that a range reduction step is not
8275 needed when performing complex division. Also, there is no checking
8276 whether the result of a complex multiplication or division is @code{NaN
8277 + I*NaN}, with an attempt to rescue the situation in that case. The
8278 default is @option{-fno-cx-limited-range}, but is enabled by
8279 @option{-ffast-math}.
8281 This option controls the default setting of the ISO C99
8282 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8285 @item -fcx-fortran-rules
8286 @opindex fcx-fortran-rules
8287 Complex multiplication and division follow Fortran rules. Range
8288 reduction is done as part of complex division, but there is no checking
8289 whether the result of a complex multiplication or division is @code{NaN
8290 + I*NaN}, with an attempt to rescue the situation in that case.
8292 The default is @option{-fno-cx-fortran-rules}.
8296 The following options control optimizations that may improve
8297 performance, but are not enabled by any @option{-O} options. This
8298 section includes experimental options that may produce broken code.
8301 @item -fbranch-probabilities
8302 @opindex fbranch-probabilities
8303 After running a program compiled with @option{-fprofile-arcs}
8304 (@pxref{Debugging Options,, Options for Debugging Your Program or
8305 @command{gcc}}), you can compile it a second time using
8306 @option{-fbranch-probabilities}, to improve optimizations based on
8307 the number of times each branch was taken. When the program
8308 compiled with @option{-fprofile-arcs} exits it saves arc execution
8309 counts to a file called @file{@var{sourcename}.gcda} for each source
8310 file. The information in this data file is very dependent on the
8311 structure of the generated code, so you must use the same source code
8312 and the same optimization options for both compilations.
8314 With @option{-fbranch-probabilities}, GCC puts a
8315 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8316 These can be used to improve optimization. Currently, they are only
8317 used in one place: in @file{reorg.c}, instead of guessing which path a
8318 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8319 exactly determine which path is taken more often.
8321 @item -fprofile-values
8322 @opindex fprofile-values
8323 If combined with @option{-fprofile-arcs}, it adds code so that some
8324 data about values of expressions in the program is gathered.
8326 With @option{-fbranch-probabilities}, it reads back the data gathered
8327 from profiling values of expressions for usage in optimizations.
8329 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8333 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8334 a code to gather information about values of expressions.
8336 With @option{-fbranch-probabilities}, it reads back the data gathered
8337 and actually performs the optimizations based on them.
8338 Currently the optimizations include specialization of division operation
8339 using the knowledge about the value of the denominator.
8341 @item -frename-registers
8342 @opindex frename-registers
8343 Attempt to avoid false dependencies in scheduled code by making use
8344 of registers left over after register allocation. This optimization
8345 will most benefit processors with lots of registers. Depending on the
8346 debug information format adopted by the target, however, it can
8347 make debugging impossible, since variables will no longer stay in
8348 a ``home register''.
8350 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8354 Perform tail duplication to enlarge superblock size. This transformation
8355 simplifies the control flow of the function allowing other optimizations to do
8358 Enabled with @option{-fprofile-use}.
8360 @item -funroll-loops
8361 @opindex funroll-loops
8362 Unroll loops whose number of iterations can be determined at compile time or
8363 upon entry to the loop. @option{-funroll-loops} implies
8364 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8365 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8366 small constant number of iterations). This option makes code larger, and may
8367 or may not make it run faster.
8369 Enabled with @option{-fprofile-use}.
8371 @item -funroll-all-loops
8372 @opindex funroll-all-loops
8373 Unroll all loops, even if their number of iterations is uncertain when
8374 the loop is entered. This usually makes programs run more slowly.
8375 @option{-funroll-all-loops} implies the same options as
8376 @option{-funroll-loops}.
8379 @opindex fpeel-loops
8380 Peels the loops for that there is enough information that they do not
8381 roll much (from profile feedback). It also turns on complete loop peeling
8382 (i.e.@: complete removal of loops with small constant number of iterations).
8384 Enabled with @option{-fprofile-use}.
8386 @item -fmove-loop-invariants
8387 @opindex fmove-loop-invariants
8388 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8389 at level @option{-O1}
8391 @item -funswitch-loops
8392 @opindex funswitch-loops
8393 Move branches with loop invariant conditions out of the loop, with duplicates
8394 of the loop on both branches (modified according to result of the condition).
8396 @item -ffunction-sections
8397 @itemx -fdata-sections
8398 @opindex ffunction-sections
8399 @opindex fdata-sections
8400 Place each function or data item into its own section in the output
8401 file if the target supports arbitrary sections. The name of the
8402 function or the name of the data item determines the section's name
8405 Use these options on systems where the linker can perform optimizations
8406 to improve locality of reference in the instruction space. Most systems
8407 using the ELF object format and SPARC processors running Solaris 2 have
8408 linkers with such optimizations. AIX may have these optimizations in
8411 Only use these options when there are significant benefits from doing
8412 so. When you specify these options, the assembler and linker will
8413 create larger object and executable files and will also be slower.
8414 You will not be able to use @code{gprof} on all systems if you
8415 specify this option and you may have problems with debugging if
8416 you specify both this option and @option{-g}.
8418 @item -fbranch-target-load-optimize
8419 @opindex fbranch-target-load-optimize
8420 Perform branch target register load optimization before prologue / epilogue
8422 The use of target registers can typically be exposed only during reload,
8423 thus hoisting loads out of loops and doing inter-block scheduling needs
8424 a separate optimization pass.
8426 @item -fbranch-target-load-optimize2
8427 @opindex fbranch-target-load-optimize2
8428 Perform branch target register load optimization after prologue / epilogue
8431 @item -fbtr-bb-exclusive
8432 @opindex fbtr-bb-exclusive
8433 When performing branch target register load optimization, don't reuse
8434 branch target registers in within any basic block.
8436 @item -fstack-protector
8437 @opindex fstack-protector
8438 Emit extra code to check for buffer overflows, such as stack smashing
8439 attacks. This is done by adding a guard variable to functions with
8440 vulnerable objects. This includes functions that call alloca, and
8441 functions with buffers larger than 8 bytes. The guards are initialized
8442 when a function is entered and then checked when the function exits.
8443 If a guard check fails, an error message is printed and the program exits.
8445 @item -fstack-protector-all
8446 @opindex fstack-protector-all
8447 Like @option{-fstack-protector} except that all functions are protected.
8449 @item -fsection-anchors
8450 @opindex fsection-anchors
8451 Try to reduce the number of symbolic address calculations by using
8452 shared ``anchor'' symbols to address nearby objects. This transformation
8453 can help to reduce the number of GOT entries and GOT accesses on some
8456 For example, the implementation of the following function @code{foo}:
8460 int foo (void) @{ return a + b + c; @}
8463 would usually calculate the addresses of all three variables, but if you
8464 compile it with @option{-fsection-anchors}, it will access the variables
8465 from a common anchor point instead. The effect is similar to the
8466 following pseudocode (which isn't valid C):
8471 register int *xr = &x;
8472 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8476 Not all targets support this option.
8478 @item --param @var{name}=@var{value}
8480 In some places, GCC uses various constants to control the amount of
8481 optimization that is done. For example, GCC will not inline functions
8482 that contain more that a certain number of instructions. You can
8483 control some of these constants on the command-line using the
8484 @option{--param} option.
8486 The names of specific parameters, and the meaning of the values, are
8487 tied to the internals of the compiler, and are subject to change
8488 without notice in future releases.
8490 In each case, the @var{value} is an integer. The allowable choices for
8491 @var{name} are given in the following table:
8494 @item predictable-branch-outcome
8495 When branch is predicted to be taken with probability lower than this threshold
8496 (in percent), then it is considered well predictable. The default is 10.
8498 @item max-crossjump-edges
8499 The maximum number of incoming edges to consider for crossjumping.
8500 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8501 the number of edges incoming to each block. Increasing values mean
8502 more aggressive optimization, making the compile time increase with
8503 probably small improvement in executable size.
8505 @item min-crossjump-insns
8506 The minimum number of instructions which must be matched at the end
8507 of two blocks before crossjumping will be performed on them. This
8508 value is ignored in the case where all instructions in the block being
8509 crossjumped from are matched. The default value is 5.
8511 @item max-grow-copy-bb-insns
8512 The maximum code size expansion factor when copying basic blocks
8513 instead of jumping. The expansion is relative to a jump instruction.
8514 The default value is 8.
8516 @item max-goto-duplication-insns
8517 The maximum number of instructions to duplicate to a block that jumps
8518 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8519 passes, GCC factors computed gotos early in the compilation process,
8520 and unfactors them as late as possible. Only computed jumps at the
8521 end of a basic blocks with no more than max-goto-duplication-insns are
8522 unfactored. The default value is 8.
8524 @item max-delay-slot-insn-search
8525 The maximum number of instructions to consider when looking for an
8526 instruction to fill a delay slot. If more than this arbitrary number of
8527 instructions is searched, the time savings from filling the delay slot
8528 will be minimal so stop searching. Increasing values mean more
8529 aggressive optimization, making the compile time increase with probably
8530 small improvement in executable run time.
8532 @item max-delay-slot-live-search
8533 When trying to fill delay slots, the maximum number of instructions to
8534 consider when searching for a block with valid live register
8535 information. Increasing this arbitrarily chosen value means more
8536 aggressive optimization, increasing the compile time. This parameter
8537 should be removed when the delay slot code is rewritten to maintain the
8540 @item max-gcse-memory
8541 The approximate maximum amount of memory that will be allocated in
8542 order to perform the global common subexpression elimination
8543 optimization. If more memory than specified is required, the
8544 optimization will not be done.
8546 @item max-gcse-insertion-ratio
8547 If the ratio of expression insertions to deletions is larger than this value
8548 for any expression, then RTL PRE will insert or remove the expression and thus
8549 leave partially redundant computations in the instruction stream. The default value is 20.
8551 @item max-pending-list-length
8552 The maximum number of pending dependencies scheduling will allow
8553 before flushing the current state and starting over. Large functions
8554 with few branches or calls can create excessively large lists which
8555 needlessly consume memory and resources.
8557 @item max-modulo-backtrack-attempts
8558 The maximum number of backtrack attempts the scheduler should make
8559 when modulo scheduling a loop. Larger values can exponentially increase
8562 @item max-inline-insns-single
8563 Several parameters control the tree inliner used in gcc.
8564 This number sets the maximum number of instructions (counted in GCC's
8565 internal representation) in a single function that the tree inliner
8566 will consider for inlining. This only affects functions declared
8567 inline and methods implemented in a class declaration (C++).
8568 The default value is 400.
8570 @item max-inline-insns-auto
8571 When you use @option{-finline-functions} (included in @option{-O3}),
8572 a lot of functions that would otherwise not be considered for inlining
8573 by the compiler will be investigated. To those functions, a different
8574 (more restrictive) limit compared to functions declared inline can
8576 The default value is 40.
8578 @item large-function-insns
8579 The limit specifying really large functions. For functions larger than this
8580 limit after inlining, inlining is constrained by
8581 @option{--param large-function-growth}. This parameter is useful primarily
8582 to avoid extreme compilation time caused by non-linear algorithms used by the
8584 The default value is 2700.
8586 @item large-function-growth
8587 Specifies maximal growth of large function caused by inlining in percents.
8588 The default value is 100 which limits large function growth to 2.0 times
8591 @item large-unit-insns
8592 The limit specifying large translation unit. Growth caused by inlining of
8593 units larger than this limit is limited by @option{--param inline-unit-growth}.
8594 For small units this might be too tight (consider unit consisting of function A
8595 that is inline and B that just calls A three time. If B is small relative to
8596 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8597 large units consisting of small inlineable functions however the overall unit
8598 growth limit is needed to avoid exponential explosion of code size. Thus for
8599 smaller units, the size is increased to @option{--param large-unit-insns}
8600 before applying @option{--param inline-unit-growth}. The default is 10000
8602 @item inline-unit-growth
8603 Specifies maximal overall growth of the compilation unit caused by inlining.
8604 The default value is 30 which limits unit growth to 1.3 times the original
8607 @item ipcp-unit-growth
8608 Specifies maximal overall growth of the compilation unit caused by
8609 interprocedural constant propagation. The default value is 10 which limits
8610 unit growth to 1.1 times the original size.
8612 @item large-stack-frame
8613 The limit specifying large stack frames. While inlining the algorithm is trying
8614 to not grow past this limit too much. Default value is 256 bytes.
8616 @item large-stack-frame-growth
8617 Specifies maximal growth of large stack frames caused by inlining in percents.
8618 The default value is 1000 which limits large stack frame growth to 11 times
8621 @item max-inline-insns-recursive
8622 @itemx max-inline-insns-recursive-auto
8623 Specifies maximum number of instructions out-of-line copy of self recursive inline
8624 function can grow into by performing recursive inlining.
8626 For functions declared inline @option{--param max-inline-insns-recursive} is
8627 taken into account. For function not declared inline, recursive inlining
8628 happens only when @option{-finline-functions} (included in @option{-O3}) is
8629 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8630 default value is 450.
8632 @item max-inline-recursive-depth
8633 @itemx max-inline-recursive-depth-auto
8634 Specifies maximum recursion depth used by the recursive inlining.
8636 For functions declared inline @option{--param max-inline-recursive-depth} is
8637 taken into account. For function not declared inline, recursive inlining
8638 happens only when @option{-finline-functions} (included in @option{-O3}) is
8639 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8642 @item min-inline-recursive-probability
8643 Recursive inlining is profitable only for function having deep recursion
8644 in average and can hurt for function having little recursion depth by
8645 increasing the prologue size or complexity of function body to other
8648 When profile feedback is available (see @option{-fprofile-generate}) the actual
8649 recursion depth can be guessed from probability that function will recurse via
8650 given call expression. This parameter limits inlining only to call expression
8651 whose probability exceeds given threshold (in percents). The default value is
8654 @item early-inlining-insns
8655 Specify growth that early inliner can make. In effect it increases amount of
8656 inlining for code having large abstraction penalty. The default value is 10.
8658 @item max-early-inliner-iterations
8659 @itemx max-early-inliner-iterations
8660 Limit of iterations of early inliner. This basically bounds number of nested
8661 indirect calls early inliner can resolve. Deeper chains are still handled by
8664 @item comdat-sharing-probability
8665 @itemx comdat-sharing-probability
8666 Probability (in percent) that C++ inline function with comdat visibility
8667 will be shared across multiple compilation units. The default value is 20.
8669 @item min-vect-loop-bound
8670 The minimum number of iterations under which a loop will not get vectorized
8671 when @option{-ftree-vectorize} is used. The number of iterations after
8672 vectorization needs to be greater than the value specified by this option
8673 to allow vectorization. The default value is 0.
8675 @item gcse-cost-distance-ratio
8676 Scaling factor in calculation of maximum distance an expression
8677 can be moved by GCSE optimizations. This is currently supported only in the
8678 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8679 will be with simple expressions, i.e., the expressions which have cost
8680 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8681 hoisting of simple expressions. The default value is 10.
8683 @item gcse-unrestricted-cost
8684 Cost, roughly measured as the cost of a single typical machine
8685 instruction, at which GCSE optimizations will not constrain
8686 the distance an expression can travel. This is currently
8687 supported only in the code hoisting pass. The lesser the cost,
8688 the more aggressive code hoisting will be. Specifying 0 will
8689 allow all expressions to travel unrestricted distances.
8690 The default value is 3.
8692 @item max-hoist-depth
8693 The depth of search in the dominator tree for expressions to hoist.
8694 This is used to avoid quadratic behavior in hoisting algorithm.
8695 The value of 0 will avoid limiting the search, but may slow down compilation
8696 of huge functions. The default value is 30.
8698 @item max-tail-merge-comparisons
8699 The maximum amount of similar bbs to compare a bb with. This is used to
8700 avoid quadratic behaviour in tree tail merging. The default value is 10.
8702 @item max-tail-merge-iterations
8703 The maximum amount of iterations of the pass over the function. This is used to
8704 limit run time in tree tail merging. The default value is 2.
8706 @item max-unrolled-insns
8707 The maximum number of instructions that a loop should have if that loop
8708 is unrolled, and if the loop is unrolled, it determines how many times
8709 the loop code is unrolled.
8711 @item max-average-unrolled-insns
8712 The maximum number of instructions biased by probabilities of their execution
8713 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8714 it determines how many times the loop code is unrolled.
8716 @item max-unroll-times
8717 The maximum number of unrollings of a single loop.
8719 @item max-peeled-insns
8720 The maximum number of instructions that a loop should have if that loop
8721 is peeled, and if the loop is peeled, it determines how many times
8722 the loop code is peeled.
8724 @item max-peel-times
8725 The maximum number of peelings of a single loop.
8727 @item max-completely-peeled-insns
8728 The maximum number of insns of a completely peeled loop.
8730 @item max-completely-peel-times
8731 The maximum number of iterations of a loop to be suitable for complete peeling.
8733 @item max-completely-peel-loop-nest-depth
8734 The maximum depth of a loop nest suitable for complete peeling.
8736 @item max-unswitch-insns
8737 The maximum number of insns of an unswitched loop.
8739 @item max-unswitch-level
8740 The maximum number of branches unswitched in a single loop.
8743 The minimum cost of an expensive expression in the loop invariant motion.
8745 @item iv-consider-all-candidates-bound
8746 Bound on number of candidates for induction variables below that
8747 all candidates are considered for each use in induction variable
8748 optimizations. Only the most relevant candidates are considered
8749 if there are more candidates, to avoid quadratic time complexity.
8751 @item iv-max-considered-uses
8752 The induction variable optimizations give up on loops that contain more
8753 induction variable uses.
8755 @item iv-always-prune-cand-set-bound
8756 If number of candidates in the set is smaller than this value,
8757 we always try to remove unnecessary ivs from the set during its
8758 optimization when a new iv is added to the set.
8760 @item scev-max-expr-size
8761 Bound on size of expressions used in the scalar evolutions analyzer.
8762 Large expressions slow the analyzer.
8764 @item scev-max-expr-complexity
8765 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8766 Complex expressions slow the analyzer.
8768 @item omega-max-vars
8769 The maximum number of variables in an Omega constraint system.
8770 The default value is 128.
8772 @item omega-max-geqs
8773 The maximum number of inequalities in an Omega constraint system.
8774 The default value is 256.
8777 The maximum number of equalities in an Omega constraint system.
8778 The default value is 128.
8780 @item omega-max-wild-cards
8781 The maximum number of wildcard variables that the Omega solver will
8782 be able to insert. The default value is 18.
8784 @item omega-hash-table-size
8785 The size of the hash table in the Omega solver. The default value is
8788 @item omega-max-keys
8789 The maximal number of keys used by the Omega solver. The default
8792 @item omega-eliminate-redundant-constraints
8793 When set to 1, use expensive methods to eliminate all redundant
8794 constraints. The default value is 0.
8796 @item vect-max-version-for-alignment-checks
8797 The maximum number of runtime checks that can be performed when
8798 doing loop versioning for alignment in the vectorizer. See option
8799 ftree-vect-loop-version for more information.
8801 @item vect-max-version-for-alias-checks
8802 The maximum number of runtime checks that can be performed when
8803 doing loop versioning for alias in the vectorizer. See option
8804 ftree-vect-loop-version for more information.
8806 @item max-iterations-to-track
8808 The maximum number of iterations of a loop the brute force algorithm
8809 for analysis of # of iterations of the loop tries to evaluate.
8811 @item hot-bb-count-fraction
8812 Select fraction of the maximal count of repetitions of basic block in program
8813 given basic block needs to have to be considered hot.
8815 @item hot-bb-frequency-fraction
8816 Select fraction of the entry block frequency of executions of basic block in
8817 function given basic block needs to have to be considered hot.
8819 @item max-predicted-iterations
8820 The maximum number of loop iterations we predict statically. This is useful
8821 in cases where function contain single loop with known bound and other loop
8822 with unknown. We predict the known number of iterations correctly, while
8823 the unknown number of iterations average to roughly 10. This means that the
8824 loop without bounds would appear artificially cold relative to the other one.
8826 @item align-threshold
8828 Select fraction of the maximal frequency of executions of basic block in
8829 function given basic block will get aligned.
8831 @item align-loop-iterations
8833 A loop expected to iterate at lest the selected number of iterations will get
8836 @item tracer-dynamic-coverage
8837 @itemx tracer-dynamic-coverage-feedback
8839 This value is used to limit superblock formation once the given percentage of
8840 executed instructions is covered. This limits unnecessary code size
8843 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8844 feedback is available. The real profiles (as opposed to statically estimated
8845 ones) are much less balanced allowing the threshold to be larger value.
8847 @item tracer-max-code-growth
8848 Stop tail duplication once code growth has reached given percentage. This is
8849 rather hokey argument, as most of the duplicates will be eliminated later in
8850 cross jumping, so it may be set to much higher values than is the desired code
8853 @item tracer-min-branch-ratio
8855 Stop reverse growth when the reverse probability of best edge is less than this
8856 threshold (in percent).
8858 @item tracer-min-branch-ratio
8859 @itemx tracer-min-branch-ratio-feedback
8861 Stop forward growth if the best edge do have probability lower than this
8864 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8865 compilation for profile feedback and one for compilation without. The value
8866 for compilation with profile feedback needs to be more conservative (higher) in
8867 order to make tracer effective.
8869 @item max-cse-path-length
8871 Maximum number of basic blocks on path that cse considers. The default is 10.
8874 The maximum instructions CSE process before flushing. The default is 1000.
8876 @item ggc-min-expand
8878 GCC uses a garbage collector to manage its own memory allocation. This
8879 parameter specifies the minimum percentage by which the garbage
8880 collector's heap should be allowed to expand between collections.
8881 Tuning this may improve compilation speed; it has no effect on code
8884 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8885 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8886 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8887 GCC is not able to calculate RAM on a particular platform, the lower
8888 bound of 30% is used. Setting this parameter and
8889 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8890 every opportunity. This is extremely slow, but can be useful for
8893 @item ggc-min-heapsize
8895 Minimum size of the garbage collector's heap before it begins bothering
8896 to collect garbage. The first collection occurs after the heap expands
8897 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8898 tuning this may improve compilation speed, and has no effect on code
8901 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8902 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8903 with a lower bound of 4096 (four megabytes) and an upper bound of
8904 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8905 particular platform, the lower bound is used. Setting this parameter
8906 very large effectively disables garbage collection. Setting this
8907 parameter and @option{ggc-min-expand} to zero causes a full collection
8908 to occur at every opportunity.
8910 @item max-reload-search-insns
8911 The maximum number of instruction reload should look backward for equivalent
8912 register. Increasing values mean more aggressive optimization, making the
8913 compile time increase with probably slightly better performance. The default
8916 @item max-cselib-memory-locations
8917 The maximum number of memory locations cselib should take into account.
8918 Increasing values mean more aggressive optimization, making the compile time
8919 increase with probably slightly better performance. The default value is 500.
8921 @item reorder-blocks-duplicate
8922 @itemx reorder-blocks-duplicate-feedback
8924 Used by basic block reordering pass to decide whether to use unconditional
8925 branch or duplicate the code on its destination. Code is duplicated when its
8926 estimated size is smaller than this value multiplied by the estimated size of
8927 unconditional jump in the hot spots of the program.
8929 The @option{reorder-block-duplicate-feedback} is used only when profile
8930 feedback is available and may be set to higher values than
8931 @option{reorder-block-duplicate} since information about the hot spots is more
8934 @item max-sched-ready-insns
8935 The maximum number of instructions ready to be issued the scheduler should
8936 consider at any given time during the first scheduling pass. Increasing
8937 values mean more thorough searches, making the compilation time increase
8938 with probably little benefit. The default value is 100.
8940 @item max-sched-region-blocks
8941 The maximum number of blocks in a region to be considered for
8942 interblock scheduling. The default value is 10.
8944 @item max-pipeline-region-blocks
8945 The maximum number of blocks in a region to be considered for
8946 pipelining in the selective scheduler. The default value is 15.
8948 @item max-sched-region-insns
8949 The maximum number of insns in a region to be considered for
8950 interblock scheduling. The default value is 100.
8952 @item max-pipeline-region-insns
8953 The maximum number of insns in a region to be considered for
8954 pipelining in the selective scheduler. The default value is 200.
8957 The minimum probability (in percents) of reaching a source block
8958 for interblock speculative scheduling. The default value is 40.
8960 @item max-sched-extend-regions-iters
8961 The maximum number of iterations through CFG to extend regions.
8962 0 - disable region extension,
8963 N - do at most N iterations.
8964 The default value is 0.
8966 @item max-sched-insn-conflict-delay
8967 The maximum conflict delay for an insn to be considered for speculative motion.
8968 The default value is 3.
8970 @item sched-spec-prob-cutoff
8971 The minimal probability of speculation success (in percents), so that
8972 speculative insn will be scheduled.
8973 The default value is 40.
8975 @item sched-mem-true-dep-cost
8976 Minimal distance (in CPU cycles) between store and load targeting same
8977 memory locations. The default value is 1.
8979 @item selsched-max-lookahead
8980 The maximum size of the lookahead window of selective scheduling. It is a
8981 depth of search for available instructions.
8982 The default value is 50.
8984 @item selsched-max-sched-times
8985 The maximum number of times that an instruction will be scheduled during
8986 selective scheduling. This is the limit on the number of iterations
8987 through which the instruction may be pipelined. The default value is 2.
8989 @item selsched-max-insns-to-rename
8990 The maximum number of best instructions in the ready list that are considered
8991 for renaming in the selective scheduler. The default value is 2.
8994 The minimum value of stage count that swing modulo scheduler will
8995 generate. The default value is 2.
8997 @item max-last-value-rtl
8998 The maximum size measured as number of RTLs that can be recorded in an expression
8999 in combiner for a pseudo register as last known value of that register. The default
9002 @item integer-share-limit
9003 Small integer constants can use a shared data structure, reducing the
9004 compiler's memory usage and increasing its speed. This sets the maximum
9005 value of a shared integer constant. The default value is 256.
9007 @item min-virtual-mappings
9008 Specifies the minimum number of virtual mappings in the incremental
9009 SSA updater that should be registered to trigger the virtual mappings
9010 heuristic defined by virtual-mappings-ratio. The default value is
9013 @item virtual-mappings-ratio
9014 If the number of virtual mappings is virtual-mappings-ratio bigger
9015 than the number of virtual symbols to be updated, then the incremental
9016 SSA updater switches to a full update for those symbols. The default
9019 @item ssp-buffer-size
9020 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
9021 protection when @option{-fstack-protection} is used.
9023 @item max-jump-thread-duplication-stmts
9024 Maximum number of statements allowed in a block that needs to be
9025 duplicated when threading jumps.
9027 @item max-fields-for-field-sensitive
9028 Maximum number of fields in a structure we will treat in
9029 a field sensitive manner during pointer analysis. The default is zero
9030 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
9032 @item prefetch-latency
9033 Estimate on average number of instructions that are executed before
9034 prefetch finishes. The distance we prefetch ahead is proportional
9035 to this constant. Increasing this number may also lead to less
9036 streams being prefetched (see @option{simultaneous-prefetches}).
9038 @item simultaneous-prefetches
9039 Maximum number of prefetches that can run at the same time.
9041 @item l1-cache-line-size
9042 The size of cache line in L1 cache, in bytes.
9045 The size of L1 cache, in kilobytes.
9048 The size of L2 cache, in kilobytes.
9050 @item min-insn-to-prefetch-ratio
9051 The minimum ratio between the number of instructions and the
9052 number of prefetches to enable prefetching in a loop.
9054 @item prefetch-min-insn-to-mem-ratio
9055 The minimum ratio between the number of instructions and the
9056 number of memory references to enable prefetching in a loop.
9058 @item use-canonical-types
9059 Whether the compiler should use the ``canonical'' type system. By
9060 default, this should always be 1, which uses a more efficient internal
9061 mechanism for comparing types in C++ and Objective-C++. However, if
9062 bugs in the canonical type system are causing compilation failures,
9063 set this value to 0 to disable canonical types.
9065 @item switch-conversion-max-branch-ratio
9066 Switch initialization conversion will refuse to create arrays that are
9067 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9068 branches in the switch.
9070 @item max-partial-antic-length
9071 Maximum length of the partial antic set computed during the tree
9072 partial redundancy elimination optimization (@option{-ftree-pre}) when
9073 optimizing at @option{-O3} and above. For some sorts of source code
9074 the enhanced partial redundancy elimination optimization can run away,
9075 consuming all of the memory available on the host machine. This
9076 parameter sets a limit on the length of the sets that are computed,
9077 which prevents the runaway behavior. Setting a value of 0 for
9078 this parameter will allow an unlimited set length.
9080 @item sccvn-max-scc-size
9081 Maximum size of a strongly connected component (SCC) during SCCVN
9082 processing. If this limit is hit, SCCVN processing for the whole
9083 function will not be done and optimizations depending on it will
9084 be disabled. The default maximum SCC size is 10000.
9086 @item ira-max-loops-num
9087 IRA uses a regional register allocation by default. If a function
9088 contains loops more than number given by the parameter, only at most
9089 given number of the most frequently executed loops will form regions
9090 for the regional register allocation. The default value of the
9093 @item ira-max-conflict-table-size
9094 Although IRA uses a sophisticated algorithm of compression conflict
9095 table, the table can be still big for huge functions. If the conflict
9096 table for a function could be more than size in MB given by the
9097 parameter, the conflict table is not built and faster, simpler, and
9098 lower quality register allocation algorithm will be used. The
9099 algorithm do not use pseudo-register conflicts. The default value of
9100 the parameter is 2000.
9102 @item ira-loop-reserved-regs
9103 IRA can be used to evaluate more accurate register pressure in loops
9104 for decision to move loop invariants (see @option{-O3}). The number
9105 of available registers reserved for some other purposes is described
9106 by this parameter. The default value of the parameter is 2 which is
9107 minimal number of registers needed for execution of typical
9108 instruction. This value is the best found from numerous experiments.
9110 @item loop-invariant-max-bbs-in-loop
9111 Loop invariant motion can be very expensive, both in compile time and
9112 in amount of needed compile time memory, with very large loops. Loops
9113 with more basic blocks than this parameter won't have loop invariant
9114 motion optimization performed on them. The default value of the
9115 parameter is 1000 for -O1 and 10000 for -O2 and above.
9117 @item max-vartrack-size
9118 Sets a maximum number of hash table slots to use during variable
9119 tracking dataflow analysis of any function. If this limit is exceeded
9120 with variable tracking at assignments enabled, analysis for that
9121 function is retried without it, after removing all debug insns from
9122 the function. If the limit is exceeded even without debug insns, var
9123 tracking analysis is completely disabled for the function. Setting
9124 the parameter to zero makes it unlimited.
9126 @item max-vartrack-expr-depth
9127 Sets a maximum number of recursion levels when attempting to map
9128 variable names or debug temporaries to value expressions. This trades
9129 compile time for more complete debug information. If this is set too
9130 low, value expressions that are available and could be represented in
9131 debug information may end up not being used; setting this higher may
9132 enable the compiler to find more complex debug expressions, but compile
9133 time and memory use may grow. The default is 12.
9135 @item min-nondebug-insn-uid
9136 Use uids starting at this parameter for nondebug insns. The range below
9137 the parameter is reserved exclusively for debug insns created by
9138 @option{-fvar-tracking-assignments}, but debug insns may get
9139 (non-overlapping) uids above it if the reserved range is exhausted.
9141 @item ipa-sra-ptr-growth-factor
9142 IPA-SRA will replace a pointer to an aggregate with one or more new
9143 parameters only when their cumulative size is less or equal to
9144 @option{ipa-sra-ptr-growth-factor} times the size of the original
9147 @item tm-max-aggregate-size
9148 When making copies of thread-local variables in a transaction, this
9149 parameter specifies the size in bytes after which variables will be
9150 saved with the logging functions as opposed to save/restore code
9151 sequence pairs. This option only applies when using
9154 @item graphite-max-nb-scop-params
9155 To avoid exponential effects in the Graphite loop transforms, the
9156 number of parameters in a Static Control Part (SCoP) is bounded. The
9157 default value is 10 parameters. A variable whose value is unknown at
9158 compile time and defined outside a SCoP is a parameter of the SCoP.
9160 @item graphite-max-bbs-per-function
9161 To avoid exponential effects in the detection of SCoPs, the size of
9162 the functions analyzed by Graphite is bounded. The default value is
9165 @item loop-block-tile-size
9166 Loop blocking or strip mining transforms, enabled with
9167 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9168 loop in the loop nest by a given number of iterations. The strip
9169 length can be changed using the @option{loop-block-tile-size}
9170 parameter. The default value is 51 iterations.
9172 @item ipa-cp-value-list-size
9173 IPA-CP attempts to track all possible values and types passed to a function's
9174 parameter in order to propagate them and perform devirtualization.
9175 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9176 stores per one formal parameter of a function.
9178 @item lto-partitions
9179 Specify desired number of partitions produced during WHOPR compilation.
9180 The number of partitions should exceed the number of CPUs used for compilation.
9181 The default value is 32.
9183 @item lto-minpartition
9184 Size of minimal partition for WHOPR (in estimated instructions).
9185 This prevents expenses of splitting very small programs into too many
9188 @item cxx-max-namespaces-for-diagnostic-help
9189 The maximum number of namespaces to consult for suggestions when C++
9190 name lookup fails for an identifier. The default is 1000.
9192 @item sink-frequency-threshold
9193 The maximum relative execution frequency (in percents) of the target block
9194 relative to a statement's original block to allow statement sinking of a
9195 statement. Larger numbers result in more aggressive statement sinking.
9196 The default value is 75. A small positive adjustment is applied for
9197 statements with memory operands as those are even more profitable so sink.
9199 @item max-stores-to-sink
9200 The maximum number of conditional stores paires that can be sunk. Set to 0
9201 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9202 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9204 @item allow-load-data-races
9205 Allow optimizers to introduce new data races on loads.
9206 Set to 1 to allow, otherwise to 0. This option is enabled by default
9207 unless implicitly set by the @option{-fmemory-model=} option.
9209 @item allow-store-data-races
9210 Allow optimizers to introduce new data races on stores.
9211 Set to 1 to allow, otherwise to 0. This option is enabled by default
9212 unless implicitly set by the @option{-fmemory-model=} option.
9214 @item allow-packed-load-data-races
9215 Allow optimizers to introduce new data races on packed data loads.
9216 Set to 1 to allow, otherwise to 0. This option is enabled by default
9217 unless implicitly set by the @option{-fmemory-model=} option.
9219 @item allow-packed-store-data-races
9220 Allow optimizers to introduce new data races on packed data stores.
9221 Set to 1 to allow, otherwise to 0. This option is enabled by default
9222 unless implicitly set by the @option{-fmemory-model=} option.
9224 @item case-values-threshold
9225 The smallest number of different values for which it is best to use a
9226 jump-table instead of a tree of conditional branches. If the value is
9227 0, use the default for the machine. The default is 0.
9229 @item tree-reassoc-width
9230 Set the maximum number of instructions executed in parallel in
9231 reassociated tree. This parameter overrides target dependent
9232 heuristics used by default if has non zero value.
9237 @node Preprocessor Options
9238 @section Options Controlling the Preprocessor
9239 @cindex preprocessor options
9240 @cindex options, preprocessor
9242 These options control the C preprocessor, which is run on each C source
9243 file before actual compilation.
9245 If you use the @option{-E} option, nothing is done except preprocessing.
9246 Some of these options make sense only together with @option{-E} because
9247 they cause the preprocessor output to be unsuitable for actual
9251 @item -Wp,@var{option}
9253 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9254 and pass @var{option} directly through to the preprocessor. If
9255 @var{option} contains commas, it is split into multiple options at the
9256 commas. However, many options are modified, translated or interpreted
9257 by the compiler driver before being passed to the preprocessor, and
9258 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9259 interface is undocumented and subject to change, so whenever possible
9260 you should avoid using @option{-Wp} and let the driver handle the
9263 @item -Xpreprocessor @var{option}
9264 @opindex Xpreprocessor
9265 Pass @var{option} as an option to the preprocessor. You can use this to
9266 supply system-specific preprocessor options which GCC does not know how to
9269 If you want to pass an option that takes an argument, you must use
9270 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9273 @include cppopts.texi
9275 @node Assembler Options
9276 @section Passing Options to the Assembler
9278 @c prevent bad page break with this line
9279 You can pass options to the assembler.
9282 @item -Wa,@var{option}
9284 Pass @var{option} as an option to the assembler. If @var{option}
9285 contains commas, it is split into multiple options at the commas.
9287 @item -Xassembler @var{option}
9289 Pass @var{option} as an option to the assembler. You can use this to
9290 supply system-specific assembler options which GCC does not know how to
9293 If you want to pass an option that takes an argument, you must use
9294 @option{-Xassembler} twice, once for the option and once for the argument.
9299 @section Options for Linking
9300 @cindex link options
9301 @cindex options, linking
9303 These options come into play when the compiler links object files into
9304 an executable output file. They are meaningless if the compiler is
9305 not doing a link step.
9309 @item @var{object-file-name}
9310 A file name that does not end in a special recognized suffix is
9311 considered to name an object file or library. (Object files are
9312 distinguished from libraries by the linker according to the file
9313 contents.) If linking is done, these object files are used as input
9322 If any of these options is used, then the linker is not run, and
9323 object file names should not be used as arguments. @xref{Overall
9327 @item -l@var{library}
9328 @itemx -l @var{library}
9330 Search the library named @var{library} when linking. (The second
9331 alternative with the library as a separate argument is only for
9332 POSIX compliance and is not recommended.)
9334 It makes a difference where in the command you write this option; the
9335 linker searches and processes libraries and object files in the order they
9336 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9337 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9338 to functions in @samp{z}, those functions may not be loaded.
9340 The linker searches a standard list of directories for the library,
9341 which is actually a file named @file{lib@var{library}.a}. The linker
9342 then uses this file as if it had been specified precisely by name.
9344 The directories searched include several standard system directories
9345 plus any that you specify with @option{-L}.
9347 Normally the files found this way are library files---archive files
9348 whose members are object files. The linker handles an archive file by
9349 scanning through it for members which define symbols that have so far
9350 been referenced but not defined. But if the file that is found is an
9351 ordinary object file, it is linked in the usual fashion. The only
9352 difference between using an @option{-l} option and specifying a file name
9353 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9354 and searches several directories.
9358 You need this special case of the @option{-l} option in order to
9359 link an Objective-C or Objective-C++ program.
9362 @opindex nostartfiles
9363 Do not use the standard system startup files when linking.
9364 The standard system libraries are used normally, unless @option{-nostdlib}
9365 or @option{-nodefaultlibs} is used.
9367 @item -nodefaultlibs
9368 @opindex nodefaultlibs
9369 Do not use the standard system libraries when linking.
9370 Only the libraries you specify will be passed to the linker, options
9371 specifying linkage of the system libraries, such as @code{-static-libgcc}
9372 or @code{-shared-libgcc}, will be ignored.
9373 The standard startup files are used normally, unless @option{-nostartfiles}
9374 is used. The compiler may generate calls to @code{memcmp},
9375 @code{memset}, @code{memcpy} and @code{memmove}.
9376 These entries are usually resolved by entries in
9377 libc. These entry points should be supplied through some other
9378 mechanism when this option is specified.
9382 Do not use the standard system startup files or libraries when linking.
9383 No startup files and only the libraries you specify will be passed to
9384 the linker, options specifying linkage of the system libraries, such as
9385 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9386 The compiler may generate calls to @code{memcmp}, @code{memset},
9387 @code{memcpy} and @code{memmove}.
9388 These entries are usually resolved by entries in
9389 libc. These entry points should be supplied through some other
9390 mechanism when this option is specified.
9392 @cindex @option{-lgcc}, use with @option{-nostdlib}
9393 @cindex @option{-nostdlib} and unresolved references
9394 @cindex unresolved references and @option{-nostdlib}
9395 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9396 @cindex @option{-nodefaultlibs} and unresolved references
9397 @cindex unresolved references and @option{-nodefaultlibs}
9398 One of the standard libraries bypassed by @option{-nostdlib} and
9399 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9400 that GCC uses to overcome shortcomings of particular machines, or special
9401 needs for some languages.
9402 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9403 Collection (GCC) Internals},
9404 for more discussion of @file{libgcc.a}.)
9405 In most cases, you need @file{libgcc.a} even when you want to avoid
9406 other standard libraries. In other words, when you specify @option{-nostdlib}
9407 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9408 This ensures that you have no unresolved references to internal GCC
9409 library subroutines. (For example, @samp{__main}, used to ensure C++
9410 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9411 GNU Compiler Collection (GCC) Internals}.)
9415 Produce a position independent executable on targets which support it.
9416 For predictable results, you must also specify the same set of options
9417 that were used to generate code (@option{-fpie}, @option{-fPIE},
9418 or model suboptions) when you specify this option.
9422 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9423 that support it. This instructs the linker to add all symbols, not
9424 only used ones, to the dynamic symbol table. This option is needed
9425 for some uses of @code{dlopen} or to allow obtaining backtraces
9426 from within a program.
9430 Remove all symbol table and relocation information from the executable.
9434 On systems that support dynamic linking, this prevents linking with the shared
9435 libraries. On other systems, this option has no effect.
9439 Produce a shared object which can then be linked with other objects to
9440 form an executable. Not all systems support this option. For predictable
9441 results, you must also specify the same set of options that were used to
9442 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9443 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9444 needs to build supplementary stub code for constructors to work. On
9445 multi-libbed systems, @samp{gcc -shared} must select the correct support
9446 libraries to link against. Failing to supply the correct flags may lead
9447 to subtle defects. Supplying them in cases where they are not necessary
9450 @item -shared-libgcc
9451 @itemx -static-libgcc
9452 @opindex shared-libgcc
9453 @opindex static-libgcc
9454 On systems that provide @file{libgcc} as a shared library, these options
9455 force the use of either the shared or static version respectively.
9456 If no shared version of @file{libgcc} was built when the compiler was
9457 configured, these options have no effect.
9459 There are several situations in which an application should use the
9460 shared @file{libgcc} instead of the static version. The most common
9461 of these is when the application wishes to throw and catch exceptions
9462 across different shared libraries. In that case, each of the libraries
9463 as well as the application itself should use the shared @file{libgcc}.
9465 Therefore, the G++ and GCJ drivers automatically add
9466 @option{-shared-libgcc} whenever you build a shared library or a main
9467 executable, because C++ and Java programs typically use exceptions, so
9468 this is the right thing to do.
9470 If, instead, you use the GCC driver to create shared libraries, you may
9471 find that they will not always be linked with the shared @file{libgcc}.
9472 If GCC finds, at its configuration time, that you have a non-GNU linker
9473 or a GNU linker that does not support option @option{--eh-frame-hdr},
9474 it will link the shared version of @file{libgcc} into shared libraries
9475 by default. Otherwise, it will take advantage of the linker and optimize
9476 away the linking with the shared version of @file{libgcc}, linking with
9477 the static version of libgcc by default. This allows exceptions to
9478 propagate through such shared libraries, without incurring relocation
9479 costs at library load time.
9481 However, if a library or main executable is supposed to throw or catch
9482 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9483 for the languages used in the program, or using the option
9484 @option{-shared-libgcc}, such that it is linked with the shared
9487 @item -static-libstdc++
9488 When the @command{g++} program is used to link a C++ program, it will
9489 normally automatically link against @option{libstdc++}. If
9490 @file{libstdc++} is available as a shared library, and the
9491 @option{-static} option is not used, then this will link against the
9492 shared version of @file{libstdc++}. That is normally fine. However, it
9493 is sometimes useful to freeze the version of @file{libstdc++} used by
9494 the program without going all the way to a fully static link. The
9495 @option{-static-libstdc++} option directs the @command{g++} driver to
9496 link @file{libstdc++} statically, without necessarily linking other
9497 libraries statically.
9501 Bind references to global symbols when building a shared object. Warn
9502 about any unresolved references (unless overridden by the link editor
9503 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9506 @item -T @var{script}
9508 @cindex linker script
9509 Use @var{script} as the linker script. This option is supported by most
9510 systems using the GNU linker. On some targets, such as bare-board
9511 targets without an operating system, the @option{-T} option may be required
9512 when linking to avoid references to undefined symbols.
9514 @item -Xlinker @var{option}
9516 Pass @var{option} as an option to the linker. You can use this to
9517 supply system-specific linker options which GCC does not know how to
9520 If you want to pass an option that takes a separate argument, you must use
9521 @option{-Xlinker} twice, once for the option and once for the argument.
9522 For example, to pass @option{-assert definitions}, you must write
9523 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9524 @option{-Xlinker "-assert definitions"}, because this passes the entire
9525 string as a single argument, which is not what the linker expects.
9527 When using the GNU linker, it is usually more convenient to pass
9528 arguments to linker options using the @option{@var{option}=@var{value}}
9529 syntax than as separate arguments. For example, you can specify
9530 @samp{-Xlinker -Map=output.map} rather than
9531 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9532 this syntax for command-line options.
9534 @item -Wl,@var{option}
9536 Pass @var{option} as an option to the linker. If @var{option} contains
9537 commas, it is split into multiple options at the commas. You can use this
9538 syntax to pass an argument to the option.
9539 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9540 linker. When using the GNU linker, you can also get the same effect with
9541 @samp{-Wl,-Map=output.map}.
9543 @item -u @var{symbol}
9545 Pretend the symbol @var{symbol} is undefined, to force linking of
9546 library modules to define it. You can use @option{-u} multiple times with
9547 different symbols to force loading of additional library modules.
9550 @node Directory Options
9551 @section Options for Directory Search
9552 @cindex directory options
9553 @cindex options, directory search
9556 These options specify directories to search for header files, for
9557 libraries and for parts of the compiler:
9562 Add the directory @var{dir} to the head of the list of directories to be
9563 searched for header files. This can be used to override a system header
9564 file, substituting your own version, since these directories are
9565 searched before the system header file directories. However, you should
9566 not use this option to add directories that contain vendor-supplied
9567 system header files (use @option{-isystem} for that). If you use more than
9568 one @option{-I} option, the directories are scanned in left-to-right
9569 order; the standard system directories come after.
9571 If a standard system include directory, or a directory specified with
9572 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9573 option will be ignored. The directory will still be searched but as a
9574 system directory at its normal position in the system include chain.
9575 This is to ensure that GCC's procedure to fix buggy system headers and
9576 the ordering for the include_next directive are not inadvertently changed.
9577 If you really need to change the search order for system directories,
9578 use the @option{-nostdinc} and/or @option{-isystem} options.
9580 @item -iplugindir=@var{dir}
9581 Set the directory to search for plugins which are passed
9582 by @option{-fplugin=@var{name}} instead of
9583 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9584 to be used by the user, but only passed by the driver.
9586 @item -iquote@var{dir}
9588 Add the directory @var{dir} to the head of the list of directories to
9589 be searched for header files only for the case of @samp{#include
9590 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9591 otherwise just like @option{-I}.
9595 Add directory @var{dir} to the list of directories to be searched
9598 @item -B@var{prefix}
9600 This option specifies where to find the executables, libraries,
9601 include files, and data files of the compiler itself.
9603 The compiler driver program runs one or more of the subprograms
9604 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9605 @var{prefix} as a prefix for each program it tries to run, both with and
9606 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9608 For each subprogram to be run, the compiler driver first tries the
9609 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9610 was not specified, the driver tries two standard prefixes, which are
9611 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9612 those results in a file name that is found, the unmodified program
9613 name is searched for using the directories specified in your
9614 @env{PATH} environment variable.
9616 The compiler will check to see if the path provided by the @option{-B}
9617 refers to a directory, and if necessary it will add a directory
9618 separator character at the end of the path.
9620 @option{-B} prefixes that effectively specify directory names also apply
9621 to libraries in the linker, because the compiler translates these
9622 options into @option{-L} options for the linker. They also apply to
9623 includes files in the preprocessor, because the compiler translates these
9624 options into @option{-isystem} options for the preprocessor. In this case,
9625 the compiler appends @samp{include} to the prefix.
9627 The run-time support file @file{libgcc.a} can also be searched for using
9628 the @option{-B} prefix, if needed. If it is not found there, the two
9629 standard prefixes above are tried, and that is all. The file is left
9630 out of the link if it is not found by those means.
9632 Another way to specify a prefix much like the @option{-B} prefix is to use
9633 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9636 As a special kludge, if the path provided by @option{-B} is
9637 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9638 9, then it will be replaced by @file{[dir/]include}. This is to help
9639 with boot-strapping the compiler.
9641 @item -specs=@var{file}
9643 Process @var{file} after the compiler reads in the standard @file{specs}
9644 file, in order to override the defaults that the @file{gcc} driver
9645 program uses when determining what switches to pass to @file{cc1},
9646 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9647 @option{-specs=@var{file}} can be specified on the command line, and they
9648 are processed in order, from left to right.
9650 @item --sysroot=@var{dir}
9652 Use @var{dir} as the logical root directory for headers and libraries.
9653 For example, if the compiler would normally search for headers in
9654 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9655 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9657 If you use both this option and the @option{-isysroot} option, then
9658 the @option{--sysroot} option will apply to libraries, but the
9659 @option{-isysroot} option will apply to header files.
9661 The GNU linker (beginning with version 2.16) has the necessary support
9662 for this option. If your linker does not support this option, the
9663 header file aspect of @option{--sysroot} will still work, but the
9664 library aspect will not.
9668 This option has been deprecated. Please use @option{-iquote} instead for
9669 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9670 Any directories you specify with @option{-I} options before the @option{-I-}
9671 option are searched only for the case of @samp{#include "@var{file}"};
9672 they are not searched for @samp{#include <@var{file}>}.
9674 If additional directories are specified with @option{-I} options after
9675 the @option{-I-}, these directories are searched for all @samp{#include}
9676 directives. (Ordinarily @emph{all} @option{-I} directories are used
9679 In addition, the @option{-I-} option inhibits the use of the current
9680 directory (where the current input file came from) as the first search
9681 directory for @samp{#include "@var{file}"}. There is no way to
9682 override this effect of @option{-I-}. With @option{-I.} you can specify
9683 searching the directory which was current when the compiler was
9684 invoked. That is not exactly the same as what the preprocessor does
9685 by default, but it is often satisfactory.
9687 @option{-I-} does not inhibit the use of the standard system directories
9688 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9695 @section Specifying subprocesses and the switches to pass to them
9698 @command{gcc} is a driver program. It performs its job by invoking a
9699 sequence of other programs to do the work of compiling, assembling and
9700 linking. GCC interprets its command-line parameters and uses these to
9701 deduce which programs it should invoke, and which command-line options
9702 it ought to place on their command lines. This behavior is controlled
9703 by @dfn{spec strings}. In most cases there is one spec string for each
9704 program that GCC can invoke, but a few programs have multiple spec
9705 strings to control their behavior. The spec strings built into GCC can
9706 be overridden by using the @option{-specs=} command-line switch to specify
9709 @dfn{Spec files} are plaintext files that are used to construct spec
9710 strings. They consist of a sequence of directives separated by blank
9711 lines. The type of directive is determined by the first non-whitespace
9712 character on the line and it can be one of the following:
9715 @item %@var{command}
9716 Issues a @var{command} to the spec file processor. The commands that can
9720 @item %include <@var{file}>
9721 @cindex @code{%include}
9722 Search for @var{file} and insert its text at the current point in the
9725 @item %include_noerr <@var{file}>
9726 @cindex @code{%include_noerr}
9727 Just like @samp{%include}, but do not generate an error message if the include
9728 file cannot be found.
9730 @item %rename @var{old_name} @var{new_name}
9731 @cindex @code{%rename}
9732 Rename the spec string @var{old_name} to @var{new_name}.
9736 @item *[@var{spec_name}]:
9737 This tells the compiler to create, override or delete the named spec
9738 string. All lines after this directive up to the next directive or
9739 blank line are considered to be the text for the spec string. If this
9740 results in an empty string then the spec will be deleted. (Or, if the
9741 spec did not exist, then nothing will happen.) Otherwise, if the spec
9742 does not currently exist a new spec will be created. If the spec does
9743 exist then its contents will be overridden by the text of this
9744 directive, unless the first character of that text is the @samp{+}
9745 character, in which case the text will be appended to the spec.
9747 @item [@var{suffix}]:
9748 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9749 and up to the next directive or blank line are considered to make up the
9750 spec string for the indicated suffix. When the compiler encounters an
9751 input file with the named suffix, it will processes the spec string in
9752 order to work out how to compile that file. For example:
9759 This says that any input file whose name ends in @samp{.ZZ} should be
9760 passed to the program @samp{z-compile}, which should be invoked with the
9761 command-line switch @option{-input} and with the result of performing the
9762 @samp{%i} substitution. (See below.)
9764 As an alternative to providing a spec string, the text that follows a
9765 suffix directive can be one of the following:
9768 @item @@@var{language}
9769 This says that the suffix is an alias for a known @var{language}. This is
9770 similar to using the @option{-x} command-line switch to GCC to specify a
9771 language explicitly. For example:
9778 Says that .ZZ files are, in fact, C++ source files.
9781 This causes an error messages saying:
9784 @var{name} compiler not installed on this system.
9788 GCC already has an extensive list of suffixes built into it.
9789 This directive will add an entry to the end of the list of suffixes, but
9790 since the list is searched from the end backwards, it is effectively
9791 possible to override earlier entries using this technique.
9795 GCC has the following spec strings built into it. Spec files can
9796 override these strings or create their own. Note that individual
9797 targets can also add their own spec strings to this list.
9800 asm Options to pass to the assembler
9801 asm_final Options to pass to the assembler post-processor
9802 cpp Options to pass to the C preprocessor
9803 cc1 Options to pass to the C compiler
9804 cc1plus Options to pass to the C++ compiler
9805 endfile Object files to include at the end of the link
9806 link Options to pass to the linker
9807 lib Libraries to include on the command line to the linker
9808 libgcc Decides which GCC support library to pass to the linker
9809 linker Sets the name of the linker
9810 predefines Defines to be passed to the C preprocessor
9811 signed_char Defines to pass to CPP to say whether @code{char} is signed
9813 startfile Object files to include at the start of the link
9816 Here is a small example of a spec file:
9822 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9825 This example renames the spec called @samp{lib} to @samp{old_lib} and
9826 then overrides the previous definition of @samp{lib} with a new one.
9827 The new definition adds in some extra command-line options before
9828 including the text of the old definition.
9830 @dfn{Spec strings} are a list of command-line options to be passed to their
9831 corresponding program. In addition, the spec strings can contain
9832 @samp{%}-prefixed sequences to substitute variable text or to
9833 conditionally insert text into the command line. Using these constructs
9834 it is possible to generate quite complex command lines.
9836 Here is a table of all defined @samp{%}-sequences for spec
9837 strings. Note that spaces are not generated automatically around the
9838 results of expanding these sequences. Therefore you can concatenate them
9839 together or combine them with constant text in a single argument.
9843 Substitute one @samp{%} into the program name or argument.
9846 Substitute the name of the input file being processed.
9849 Substitute the basename of the input file being processed.
9850 This is the substring up to (and not including) the last period
9851 and not including the directory.
9854 This is the same as @samp{%b}, but include the file suffix (text after
9858 Marks the argument containing or following the @samp{%d} as a
9859 temporary file name, so that that file will be deleted if GCC exits
9860 successfully. Unlike @samp{%g}, this contributes no text to the
9863 @item %g@var{suffix}
9864 Substitute a file name that has suffix @var{suffix} and is chosen
9865 once per compilation, and mark the argument in the same way as
9866 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9867 name is now chosen in a way that is hard to predict even when previously
9868 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9869 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9870 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9871 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9872 was simply substituted with a file name chosen once per compilation,
9873 without regard to any appended suffix (which was therefore treated
9874 just like ordinary text), making such attacks more likely to succeed.
9876 @item %u@var{suffix}
9877 Like @samp{%g}, but generates a new temporary file name even if
9878 @samp{%u@var{suffix}} was already seen.
9880 @item %U@var{suffix}
9881 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9882 new one if there is no such last file name. In the absence of any
9883 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9884 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9885 would involve the generation of two distinct file names, one
9886 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9887 simply substituted with a file name chosen for the previous @samp{%u},
9888 without regard to any appended suffix.
9890 @item %j@var{suffix}
9891 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9892 writable, and if save-temps is off; otherwise, substitute the name
9893 of a temporary file, just like @samp{%u}. This temporary file is not
9894 meant for communication between processes, but rather as a junk
9897 @item %|@var{suffix}
9898 @itemx %m@var{suffix}
9899 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9900 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9901 all. These are the two most common ways to instruct a program that it
9902 should read from standard input or write to standard output. If you
9903 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9904 construct: see for example @file{f/lang-specs.h}.
9906 @item %.@var{SUFFIX}
9907 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9908 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9909 terminated by the next space or %.
9912 Marks the argument containing or following the @samp{%w} as the
9913 designated output file of this compilation. This puts the argument
9914 into the sequence of arguments that @samp{%o} will substitute later.
9917 Substitutes the names of all the output files, with spaces
9918 automatically placed around them. You should write spaces
9919 around the @samp{%o} as well or the results are undefined.
9920 @samp{%o} is for use in the specs for running the linker.
9921 Input files whose names have no recognized suffix are not compiled
9922 at all, but they are included among the output files, so they will
9926 Substitutes the suffix for object files. Note that this is
9927 handled specially when it immediately follows @samp{%g, %u, or %U},
9928 because of the need for those to form complete file names. The
9929 handling is such that @samp{%O} is treated exactly as if it had already
9930 been substituted, except that @samp{%g, %u, and %U} do not currently
9931 support additional @var{suffix} characters following @samp{%O} as they would
9932 following, for example, @samp{.o}.
9935 Substitutes the standard macro predefinitions for the
9936 current target machine. Use this when running @code{cpp}.
9939 Like @samp{%p}, but puts @samp{__} before and after the name of each
9940 predefined macro, except for macros that start with @samp{__} or with
9941 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9945 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9946 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9947 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9948 and @option{-imultilib} as necessary.
9951 Current argument is the name of a library or startup file of some sort.
9952 Search for that file in a standard list of directories and substitute
9953 the full name found. The current working directory is included in the
9954 list of directories scanned.
9957 Current argument is the name of a linker script. Search for that file
9958 in the current list of directories to scan for libraries. If the file
9959 is located insert a @option{--script} option into the command line
9960 followed by the full path name found. If the file is not found then
9961 generate an error message. Note: the current working directory is not
9965 Print @var{str} as an error message. @var{str} is terminated by a newline.
9966 Use this when inconsistent options are detected.
9969 Substitute the contents of spec string @var{name} at this point.
9971 @item %x@{@var{option}@}
9972 Accumulate an option for @samp{%X}.
9975 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9979 Output the accumulated assembler options specified by @option{-Wa}.
9982 Output the accumulated preprocessor options specified by @option{-Wp}.
9985 Process the @code{asm} spec. This is used to compute the
9986 switches to be passed to the assembler.
9989 Process the @code{asm_final} spec. This is a spec string for
9990 passing switches to an assembler post-processor, if such a program is
9994 Process the @code{link} spec. This is the spec for computing the
9995 command line passed to the linker. Typically it will make use of the
9996 @samp{%L %G %S %D and %E} sequences.
9999 Dump out a @option{-L} option for each directory that GCC believes might
10000 contain startup files. If the target supports multilibs then the
10001 current multilib directory will be prepended to each of these paths.
10004 Process the @code{lib} spec. This is a spec string for deciding which
10005 libraries should be included on the command line to the linker.
10008 Process the @code{libgcc} spec. This is a spec string for deciding
10009 which GCC support library should be included on the command line to the linker.
10012 Process the @code{startfile} spec. This is a spec for deciding which
10013 object files should be the first ones passed to the linker. Typically
10014 this might be a file named @file{crt0.o}.
10017 Process the @code{endfile} spec. This is a spec string that specifies
10018 the last object files that will be passed to the linker.
10021 Process the @code{cpp} spec. This is used to construct the arguments
10022 to be passed to the C preprocessor.
10025 Process the @code{cc1} spec. This is used to construct the options to be
10026 passed to the actual C compiler (@samp{cc1}).
10029 Process the @code{cc1plus} spec. This is used to construct the options to be
10030 passed to the actual C++ compiler (@samp{cc1plus}).
10033 Substitute the variable part of a matched option. See below.
10034 Note that each comma in the substituted string is replaced by
10038 Remove all occurrences of @code{-S} from the command line. Note---this
10039 command is position dependent. @samp{%} commands in the spec string
10040 before this one will see @code{-S}, @samp{%} commands in the spec string
10041 after this one will not.
10043 @item %:@var{function}(@var{args})
10044 Call the named function @var{function}, passing it @var{args}.
10045 @var{args} is first processed as a nested spec string, then split
10046 into an argument vector in the usual fashion. The function returns
10047 a string which is processed as if it had appeared literally as part
10048 of the current spec.
10050 The following built-in spec functions are provided:
10053 @item @code{getenv}
10054 The @code{getenv} spec function takes two arguments: an environment
10055 variable name and a string. If the environment variable is not
10056 defined, a fatal error is issued. Otherwise, the return value is the
10057 value of the environment variable concatenated with the string. For
10058 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
10061 %:getenv(TOPDIR /include)
10064 expands to @file{/path/to/top/include}.
10066 @item @code{if-exists}
10067 The @code{if-exists} spec function takes one argument, an absolute
10068 pathname to a file. If the file exists, @code{if-exists} returns the
10069 pathname. Here is a small example of its usage:
10073 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
10076 @item @code{if-exists-else}
10077 The @code{if-exists-else} spec function is similar to the @code{if-exists}
10078 spec function, except that it takes two arguments. The first argument is
10079 an absolute pathname to a file. If the file exists, @code{if-exists-else}
10080 returns the pathname. If it does not exist, it returns the second argument.
10081 This way, @code{if-exists-else} can be used to select one file or another,
10082 based on the existence of the first. Here is a small example of its usage:
10086 crt0%O%s %:if-exists(crti%O%s) \
10087 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
10090 @item @code{replace-outfile}
10091 The @code{replace-outfile} spec function takes two arguments. It looks for the
10092 first argument in the outfiles array and replaces it with the second argument. Here
10093 is a small example of its usage:
10096 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10099 @item @code{remove-outfile}
10100 The @code{remove-outfile} spec function takes one argument. It looks for the
10101 first argument in the outfiles array and removes it. Here is a small example
10105 %:remove-outfile(-lm)
10108 @item @code{pass-through-libs}
10109 The @code{pass-through-libs} spec function takes any number of arguments. It
10110 finds any @option{-l} options and any non-options ending in ".a" (which it
10111 assumes are the names of linker input library archive files) and returns a
10112 result containing all the found arguments each prepended by
10113 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10114 intended to be passed to the LTO linker plugin.
10117 %:pass-through-libs(%G %L %G)
10120 @item @code{print-asm-header}
10121 The @code{print-asm-header} function takes no arguments and simply
10122 prints a banner like:
10128 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10131 It is used to separate compiler options from assembler options
10132 in the @option{--target-help} output.
10135 @item %@{@code{S}@}
10136 Substitutes the @code{-S} switch, if that switch was given to GCC@.
10137 If that switch was not specified, this substitutes nothing. Note that
10138 the leading dash is omitted when specifying this option, and it is
10139 automatically inserted if the substitution is performed. Thus the spec
10140 string @samp{%@{foo@}} would match the command-line option @option{-foo}
10141 and would output the command line option @option{-foo}.
10143 @item %W@{@code{S}@}
10144 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10145 deleted on failure.
10147 @item %@{@code{S}*@}
10148 Substitutes all the switches specified to GCC whose names start
10149 with @code{-S}, but which also take an argument. This is used for
10150 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10151 GCC considers @option{-o foo} as being
10152 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10153 text, including the space. Thus two arguments would be generated.
10155 @item %@{@code{S}*&@code{T}*@}
10156 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10157 (the order of @code{S} and @code{T} in the spec is not significant).
10158 There can be any number of ampersand-separated variables; for each the
10159 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10161 @item %@{@code{S}:@code{X}@}
10162 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10164 @item %@{!@code{S}:@code{X}@}
10165 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10167 @item %@{@code{S}*:@code{X}@}
10168 Substitutes @code{X} if one or more switches whose names start with
10169 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10170 once, no matter how many such switches appeared. However, if @code{%*}
10171 appears somewhere in @code{X}, then @code{X} will be substituted once
10172 for each matching switch, with the @code{%*} replaced by the part of
10173 that switch that matched the @code{*}.
10175 @item %@{.@code{S}:@code{X}@}
10176 Substitutes @code{X}, if processing a file with suffix @code{S}.
10178 @item %@{!.@code{S}:@code{X}@}
10179 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10181 @item %@{,@code{S}:@code{X}@}
10182 Substitutes @code{X}, if processing a file for language @code{S}.
10184 @item %@{!,@code{S}:@code{X}@}
10185 Substitutes @code{X}, if not processing a file for language @code{S}.
10187 @item %@{@code{S}|@code{P}:@code{X}@}
10188 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10189 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10190 @code{*} sequences as well, although they have a stronger binding than
10191 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10192 alternatives must be starred, and only the first matching alternative
10195 For example, a spec string like this:
10198 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10201 will output the following command-line options from the following input
10202 command-line options:
10207 -d fred.c -foo -baz -boggle
10208 -d jim.d -bar -baz -boggle
10211 @item %@{S:X; T:Y; :D@}
10213 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10214 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10215 be as many clauses as you need. This may be combined with @code{.},
10216 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10221 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10222 construct may contain other nested @samp{%} constructs or spaces, or
10223 even newlines. They are processed as usual, as described above.
10224 Trailing white space in @code{X} is ignored. White space may also
10225 appear anywhere on the left side of the colon in these constructs,
10226 except between @code{.} or @code{*} and the corresponding word.
10228 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10229 handled specifically in these constructs. If another value of
10230 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10231 @option{-W} switch is found later in the command line, the earlier
10232 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10233 just one letter, which passes all matching options.
10235 The character @samp{|} at the beginning of the predicate text is used to
10236 indicate that a command should be piped to the following command, but
10237 only if @option{-pipe} is specified.
10239 It is built into GCC which switches take arguments and which do not.
10240 (You might think it would be useful to generalize this to allow each
10241 compiler's spec to say which switches take arguments. But this cannot
10242 be done in a consistent fashion. GCC cannot even decide which input
10243 files have been specified without knowing which switches take arguments,
10244 and it must know which input files to compile in order to tell which
10247 GCC also knows implicitly that arguments starting in @option{-l} are to be
10248 treated as compiler output files, and passed to the linker in their
10249 proper position among the other output files.
10251 @c man begin OPTIONS
10253 @node Target Options
10254 @section Specifying Target Machine and Compiler Version
10255 @cindex target options
10256 @cindex cross compiling
10257 @cindex specifying machine version
10258 @cindex specifying compiler version and target machine
10259 @cindex compiler version, specifying
10260 @cindex target machine, specifying
10262 The usual way to run GCC is to run the executable called @command{gcc}, or
10263 @command{@var{machine}-gcc} when cross-compiling, or
10264 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10265 one that was installed last.
10267 @node Submodel Options
10268 @section Hardware Models and Configurations
10269 @cindex submodel options
10270 @cindex specifying hardware config
10271 @cindex hardware models and configurations, specifying
10272 @cindex machine dependent options
10274 Each target machine types can have its own
10275 special options, starting with @samp{-m}, to choose among various
10276 hardware models or configurations---for example, 68010 vs 68020,
10277 floating coprocessor or none. A single installed version of the
10278 compiler can compile for any model or configuration, according to the
10281 Some configurations of the compiler also support additional special
10282 options, usually for compatibility with other compilers on the same
10285 @c This list is ordered alphanumerically by subsection name.
10286 @c It should be the same order and spelling as these options are listed
10287 @c in Machine Dependent Options
10290 * Adapteva Epiphany Options::
10293 * Blackfin Options::
10297 * DEC Alpha Options::
10298 * DEC Alpha/VMS Options::
10301 * GNU/Linux Options::
10304 * i386 and x86-64 Options::
10305 * i386 and x86-64 Windows Options::
10307 * IA-64/VMS Options::
10314 * MicroBlaze Options::
10317 * MN10300 Options::
10319 * picoChip Options::
10320 * PowerPC Options::
10322 * RS/6000 and PowerPC Options::
10324 * S/390 and zSeries Options::
10327 * Solaris 2 Options::
10330 * System V Options::
10333 * VxWorks Options::
10335 * Xstormy16 Options::
10337 * zSeries Options::
10340 @node Adapteva Epiphany Options
10341 @subsection Adapteva Epiphany Options
10343 These @samp{-m} options are defined for Adapteva Epiphany:
10346 @item -mhalf-reg-file
10347 @opindex mhalf-reg-file
10348 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
10349 That allows code to run on hardware variants that lack these registers.
10351 @item -mprefer-short-insn-regs
10352 @opindex mprefer-short-insn-regs
10353 Preferrentially allocate registers that allow short instruction generation.
10354 This can result in increasesd instruction count, so if this reduces or
10355 increases code size might vary from case to case.
10357 @item -mbranch-cost=@var{num}
10358 @opindex mbranch-cost
10359 Set the cost of branches to roughly @var{num} ``simple'' instructions.
10360 This cost is only a heuristic and is not guaranteed to produce
10361 consistent results across releases.
10365 Enable the generation of conditional moves.
10367 @item -mnops=@var{num}
10369 Emit @var{num} nops before every other generated instruction.
10371 @item -mno-soft-cmpsf
10372 @opindex mno-soft-cmpsf
10373 For single-precision floating point comparisons, emit an fsub instruction
10374 and test the flags. This is faster than a software comparison, but can
10375 get incorrect results in the presence of NaNs, or when two different small
10376 numbers are compared such that their difference is calculated as zero.
10377 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
10378 software comparisons.
10380 @item -mstack-offset=@var{num}
10381 @opindex mstack-offset
10382 Set the offset between the top of the stack and the stack pointer.
10383 E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7
10384 can be used by leaf functions without stack allocation.
10385 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
10386 Note also that this option changes the ABI, compiling a program with a
10387 different stack offset than the libraries have been compiled with
10388 will generally not work.
10389 This option can be useful if you want to evaluate if a different stack
10390 offset would give you better code, but to actually use a different stack
10391 offset to build working programs, it is recommended to configure the
10392 toolchain with the appropriate @samp{--with-stack-offset=@var{num}} option.
10394 @item -mno-round-nearest
10395 @opindex mno-round-nearest
10396 Make the scheduler assume that the rounding mode has been set to
10397 truncating. The default is @option{-mround-nearest}.
10400 @opindex mlong-calls
10401 If not otherwise specified by an attribute, assume all calls might be beyond
10402 the offset range of the b / bl instructions, and therefore load the
10403 function address into a register before performing a (otherwise direct) call.
10404 This is the default.
10406 @item -mshort-calls
10407 @opindex short-calls
10408 If not otherwise specified by an attribute, assume all direct calls are
10409 in the range of the b / bl instructions, so use these instructions
10410 for direct calls. The default is @option{-mlong-calls}.
10414 Assume addresses can be loaded as 16 bit unsigned values. This does not
10415 apply to function addresses for which @option{-mlong-calls} semantics
10418 @item -mfp-mode=@var{mode}
10420 Set the prevailing mode of the floating point unit.
10421 This determines the floating point mode that is provided and expected
10422 at function call and return time. Making this mode match the mode you
10423 predominantly need at function start can make your programs smaller and
10424 faster by avoiding unnecessary mode switches.
10426 @var{mode} can be set to one the following values:
10430 Any mode at function entry is valid, and retained or restored when
10431 the function returns, and when it calls other functions.
10432 This mode is useful for compiling libraries or other compilation units
10433 you might want to incorporate into different programs with different
10434 prevailing FPU modes, and the convenience of being able to use a single
10435 object file outweighs the size and speed overhead for any extra
10436 mode switching that might be needed, compared with what would be needed
10437 with a more specific choice of prevailing FPU mode.
10440 This is the mode used for floating point calculations with
10441 truncating (i.e.@: round towards zero) rounding mode. That includes
10442 conversion from floating point to integer.
10444 @item round-nearest
10445 This is the mode used for floating point calculations with
10446 round-to-nearest-or-even rounding mode.
10449 This is the mode used to perform integer calculations in the FPU, e.g.@:
10450 integer multiply, or integer multiply-and-accumulate.
10453 The default is @option{-mfp-mode=caller}
10455 @item -mnosplit-lohi
10456 @opindex mnosplit-lohi
10458 @opindex mno-postinc
10459 @item -mno-postmodify
10460 @opindex mno-postmodify
10461 Code generation tweaks that disable, respectively, splitting of 32
10462 bit loads, generation of post-increment addresses, and generation of
10463 post-modify addresses. The defaults are @option{msplit-lohi},
10464 @option{-mpost-inc}, and @option{-mpost-modify}.
10466 @item -mnovect-double
10467 @opindex mno-vect-double
10468 Change the preferred SIMD mode to SImode. The default is
10469 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
10471 @item -max-vect-align=@var{num}
10472 @opindex max-vect-align
10473 The maximum alignment for SIMD vector mode types.
10474 @var{num} may be 4 or 8. The default is 8.
10475 Note that this is an ABI change, even though many library function
10476 interfaces will be unaffected, if they don't use SIMD vector modes
10477 in places where they affect size and/or alignment of relevant types.
10479 @item -msplit-vecmove-early
10480 @opindex msplit-vecmove-early
10481 Split vector moves into single word moves before reload. In theory this
10482 could give better register allocation, but so far the reverse seems to be
10483 generally the case.
10485 @item -m1reg-@var{reg}
10487 Specify a register to hold the constant @minus{}1, which makes loading small negative
10488 constants and certain bitmasks faster.
10489 Allowable values for reg are r43 and r63, which specify to use that register
10490 as a fixed register, and none, which means that no register is used for this
10491 purpose. The default is @option{-m1reg-none}.
10496 @subsection ARM Options
10497 @cindex ARM options
10499 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10503 @item -mabi=@var{name}
10505 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10506 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10509 @opindex mapcs-frame
10510 Generate a stack frame that is compliant with the ARM Procedure Call
10511 Standard for all functions, even if this is not strictly necessary for
10512 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10513 with this option will cause the stack frames not to be generated for
10514 leaf functions. The default is @option{-mno-apcs-frame}.
10518 This is a synonym for @option{-mapcs-frame}.
10521 @c not currently implemented
10522 @item -mapcs-stack-check
10523 @opindex mapcs-stack-check
10524 Generate code to check the amount of stack space available upon entry to
10525 every function (that actually uses some stack space). If there is
10526 insufficient space available then either the function
10527 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10528 called, depending upon the amount of stack space required. The run time
10529 system is required to provide these functions. The default is
10530 @option{-mno-apcs-stack-check}, since this produces smaller code.
10532 @c not currently implemented
10534 @opindex mapcs-float
10535 Pass floating point arguments using the float point registers. This is
10536 one of the variants of the APCS@. This option is recommended if the
10537 target hardware has a floating point unit or if a lot of floating point
10538 arithmetic is going to be performed by the code. The default is
10539 @option{-mno-apcs-float}, since integer only code is slightly increased in
10540 size if @option{-mapcs-float} is used.
10542 @c not currently implemented
10543 @item -mapcs-reentrant
10544 @opindex mapcs-reentrant
10545 Generate reentrant, position independent code. The default is
10546 @option{-mno-apcs-reentrant}.
10549 @item -mthumb-interwork
10550 @opindex mthumb-interwork
10551 Generate code which supports calling between the ARM and Thumb
10552 instruction sets. Without this option, on pre-v5 architectures, the
10553 two instruction sets cannot be reliably used inside one program. The
10554 default is @option{-mno-thumb-interwork}, since slightly larger code
10555 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10556 configurations this option is meaningless.
10558 @item -mno-sched-prolog
10559 @opindex mno-sched-prolog
10560 Prevent the reordering of instructions in the function prolog, or the
10561 merging of those instruction with the instructions in the function's
10562 body. This means that all functions will start with a recognizable set
10563 of instructions (or in fact one of a choice from a small set of
10564 different function prologues), and this information can be used to
10565 locate the start if functions inside an executable piece of code. The
10566 default is @option{-msched-prolog}.
10568 @item -mfloat-abi=@var{name}
10569 @opindex mfloat-abi
10570 Specifies which floating-point ABI to use. Permissible values
10571 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10573 Specifying @samp{soft} causes GCC to generate output containing
10574 library calls for floating-point operations.
10575 @samp{softfp} allows the generation of code using hardware floating-point
10576 instructions, but still uses the soft-float calling conventions.
10577 @samp{hard} allows generation of floating-point instructions
10578 and uses FPU-specific calling conventions.
10580 The default depends on the specific target configuration. Note that
10581 the hard-float and soft-float ABIs are not link-compatible; you must
10582 compile your entire program with the same ABI, and link with a
10583 compatible set of libraries.
10585 @item -mlittle-endian
10586 @opindex mlittle-endian
10587 Generate code for a processor running in little-endian mode. This is
10588 the default for all standard configurations.
10591 @opindex mbig-endian
10592 Generate code for a processor running in big-endian mode; the default is
10593 to compile code for a little-endian processor.
10595 @item -mwords-little-endian
10596 @opindex mwords-little-endian
10597 This option only applies when generating code for big-endian processors.
10598 Generate code for a little-endian word order but a big-endian byte
10599 order. That is, a byte order of the form @samp{32107654}. Note: this
10600 option should only be used if you require compatibility with code for
10601 big-endian ARM processors generated by versions of the compiler prior to
10602 2.8. This option is now deprecated.
10604 @item -mcpu=@var{name}
10606 This specifies the name of the target ARM processor. GCC uses this name
10607 to determine what kind of instructions it can emit when generating
10608 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10609 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10610 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10611 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10612 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10614 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10615 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10616 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10617 @samp{strongarm1110},
10618 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10619 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10620 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10621 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10622 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10623 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10624 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10625 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
10626 @samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10627 @samp{cortex-m4}, @samp{cortex-m3},
10630 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10631 @samp{fa526}, @samp{fa626},
10632 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10635 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10636 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10637 See @option{-mtune} for more information.
10639 @option{-mcpu=native} causes the compiler to auto-detect the CPU
10640 of the build computer. At present, this feature is only supported on
10641 Linux, and not all architectures are recognised. If the auto-detect is
10642 unsuccessful the option has no effect.
10644 @item -mtune=@var{name}
10646 This option is very similar to the @option{-mcpu=} option, except that
10647 instead of specifying the actual target processor type, and hence
10648 restricting which instructions can be used, it specifies that GCC should
10649 tune the performance of the code as if the target were of the type
10650 specified in this option, but still choosing the instructions that it
10651 will generate based on the CPU specified by a @option{-mcpu=} option.
10652 For some ARM implementations better performance can be obtained by using
10655 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10656 performance for a blend of processors within architecture @var{arch}.
10657 The aim is to generate code that run well on the current most popular
10658 processors, balancing between optimizations that benefit some CPUs in the
10659 range, and avoiding performance pitfalls of other CPUs. The effects of
10660 this option may change in future GCC versions as CPU models come and go.
10662 @option{-mtune=native} causes the compiler to auto-detect the CPU
10663 of the build computer. At present, this feature is only supported on
10664 Linux, and not all architectures are recognised. If the auto-detect is
10665 unsuccessful the option has no effect.
10667 @item -march=@var{name}
10669 This specifies the name of the target ARM architecture. GCC uses this
10670 name to determine what kind of instructions it can emit when generating
10671 assembly code. This option can be used in conjunction with or instead
10672 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10673 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10674 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10675 @samp{armv6}, @samp{armv6j},
10676 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10677 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10678 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10680 @option{-march=native} causes the compiler to auto-detect the architecture
10681 of the build computer. At present, this feature is only supported on
10682 Linux, and not all architectures are recognised. If the auto-detect is
10683 unsuccessful the option has no effect.
10685 @item -mfpu=@var{name}
10686 @itemx -mfpe=@var{number}
10687 @itemx -mfp=@var{number}
10691 This specifies what floating point hardware (or hardware emulation) is
10692 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10693 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10694 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10695 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10696 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10697 @option{-mfp} and @option{-mfpe} are synonyms for
10698 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10701 If @option{-msoft-float} is specified this specifies the format of
10702 floating point values.
10704 If the selected floating-point hardware includes the NEON extension
10705 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10706 operations will not be used by GCC's auto-vectorization pass unless
10707 @option{-funsafe-math-optimizations} is also specified. This is
10708 because NEON hardware does not fully implement the IEEE 754 standard for
10709 floating-point arithmetic (in particular denormal values are treated as
10710 zero), so the use of NEON instructions may lead to a loss of precision.
10712 @item -mfp16-format=@var{name}
10713 @opindex mfp16-format
10714 Specify the format of the @code{__fp16} half-precision floating-point type.
10715 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10716 the default is @samp{none}, in which case the @code{__fp16} type is not
10717 defined. @xref{Half-Precision}, for more information.
10719 @item -mstructure-size-boundary=@var{n}
10720 @opindex mstructure-size-boundary
10721 The size of all structures and unions will be rounded up to a multiple
10722 of the number of bits set by this option. Permissible values are 8, 32
10723 and 64. The default value varies for different toolchains. For the COFF
10724 targeted toolchain the default value is 8. A value of 64 is only allowed
10725 if the underlying ABI supports it.
10727 Specifying the larger number can produce faster, more efficient code, but
10728 can also increase the size of the program. Different values are potentially
10729 incompatible. Code compiled with one value cannot necessarily expect to
10730 work with code or libraries compiled with another value, if they exchange
10731 information using structures or unions.
10733 @item -mabort-on-noreturn
10734 @opindex mabort-on-noreturn
10735 Generate a call to the function @code{abort} at the end of a
10736 @code{noreturn} function. It will be executed if the function tries to
10740 @itemx -mno-long-calls
10741 @opindex mlong-calls
10742 @opindex mno-long-calls
10743 Tells the compiler to perform function calls by first loading the
10744 address of the function into a register and then performing a subroutine
10745 call on this register. This switch is needed if the target function
10746 will lie outside of the 64 megabyte addressing range of the offset based
10747 version of subroutine call instruction.
10749 Even if this switch is enabled, not all function calls will be turned
10750 into long calls. The heuristic is that static functions, functions
10751 which have the @samp{short-call} attribute, functions that are inside
10752 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10753 definitions have already been compiled within the current compilation
10754 unit, will not be turned into long calls. The exception to this rule is
10755 that weak function definitions, functions with the @samp{long-call}
10756 attribute or the @samp{section} attribute, and functions that are within
10757 the scope of a @samp{#pragma long_calls} directive, will always be
10758 turned into long calls.
10760 This feature is not enabled by default. Specifying
10761 @option{-mno-long-calls} will restore the default behavior, as will
10762 placing the function calls within the scope of a @samp{#pragma
10763 long_calls_off} directive. Note these switches have no effect on how
10764 the compiler generates code to handle function calls via function
10767 @item -msingle-pic-base
10768 @opindex msingle-pic-base
10769 Treat the register used for PIC addressing as read-only, rather than
10770 loading it in the prologue for each function. The run-time system is
10771 responsible for initializing this register with an appropriate value
10772 before execution begins.
10774 @item -mpic-register=@var{reg}
10775 @opindex mpic-register
10776 Specify the register to be used for PIC addressing. The default is R10
10777 unless stack-checking is enabled, when R9 is used.
10779 @item -mcirrus-fix-invalid-insns
10780 @opindex mcirrus-fix-invalid-insns
10781 @opindex mno-cirrus-fix-invalid-insns
10782 Insert NOPs into the instruction stream to in order to work around
10783 problems with invalid Maverick instruction combinations. This option
10784 is only valid if the @option{-mcpu=ep9312} option has been used to
10785 enable generation of instructions for the Cirrus Maverick floating
10786 point co-processor. This option is not enabled by default, since the
10787 problem is only present in older Maverick implementations. The default
10788 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10791 @item -mpoke-function-name
10792 @opindex mpoke-function-name
10793 Write the name of each function into the text section, directly
10794 preceding the function prologue. The generated code is similar to this:
10798 .ascii "arm_poke_function_name", 0
10801 .word 0xff000000 + (t1 - t0)
10802 arm_poke_function_name
10804 stmfd sp!, @{fp, ip, lr, pc@}
10808 When performing a stack backtrace, code can inspect the value of
10809 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10810 location @code{pc - 12} and the top 8 bits are set, then we know that
10811 there is a function name embedded immediately preceding this location
10812 and has length @code{((pc[-3]) & 0xff000000)}.
10819 Select between generating code that executes in ARM and Thumb
10820 states. The default for most configurations is to generate code
10821 that executes in ARM state, but the default can be changed by
10822 configuring GCC with the @option{--with-mode=}@var{state}
10826 @opindex mtpcs-frame
10827 Generate a stack frame that is compliant with the Thumb Procedure Call
10828 Standard for all non-leaf functions. (A leaf function is one that does
10829 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10831 @item -mtpcs-leaf-frame
10832 @opindex mtpcs-leaf-frame
10833 Generate a stack frame that is compliant with the Thumb Procedure Call
10834 Standard for all leaf functions. (A leaf function is one that does
10835 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10837 @item -mcallee-super-interworking
10838 @opindex mcallee-super-interworking
10839 Gives all externally visible functions in the file being compiled an ARM
10840 instruction set header which switches to Thumb mode before executing the
10841 rest of the function. This allows these functions to be called from
10842 non-interworking code. This option is not valid in AAPCS configurations
10843 because interworking is enabled by default.
10845 @item -mcaller-super-interworking
10846 @opindex mcaller-super-interworking
10847 Allows calls via function pointers (including virtual functions) to
10848 execute correctly regardless of whether the target code has been
10849 compiled for interworking or not. There is a small overhead in the cost
10850 of executing a function pointer if this option is enabled. This option
10851 is not valid in AAPCS configurations because interworking is enabled
10854 @item -mtp=@var{name}
10856 Specify the access model for the thread local storage pointer. The valid
10857 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10858 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10859 (supported in the arm6k architecture), and @option{auto}, which uses the
10860 best available method for the selected processor. The default setting is
10863 @item -mtls-dialect=@var{dialect}
10864 @opindex mtls-dialect
10865 Specify the dialect to use for accessing thread local storage. Two
10866 dialects are supported --- @option{gnu} and @option{gnu2}. The
10867 @option{gnu} dialect selects the original GNU scheme for supporting
10868 local and global dynamic TLS models. The @option{gnu2} dialect
10869 selects the GNU descriptor scheme, which provides better performance
10870 for shared libraries. The GNU descriptor scheme is compatible with
10871 the original scheme, but does require new assembler, linker and
10872 library support. Initial and local exec TLS models are unaffected by
10873 this option and always use the original scheme.
10875 @item -mword-relocations
10876 @opindex mword-relocations
10877 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10878 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10879 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10882 @item -mfix-cortex-m3-ldrd
10883 @opindex mfix-cortex-m3-ldrd
10884 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10885 with overlapping destination and base registers are used. This option avoids
10886 generating these instructions. This option is enabled by default when
10887 @option{-mcpu=cortex-m3} is specified.
10892 @subsection AVR Options
10893 @cindex AVR Options
10895 These options are defined for AVR implementations:
10898 @item -mmcu=@var{mcu}
10900 Specify ATMEL AVR instruction set or MCU type.
10902 Instruction set avr1 is for the minimal AVR core, not supported by the C
10903 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10904 attiny11, attiny12, attiny15, attiny28).
10906 Instruction set avr2 (default) is for the classic AVR core with up to
10907 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10908 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10909 at90c8534, at90s8535).
10911 Instruction set avr3 is for the classic AVR core with up to 128K program
10912 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10914 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10915 memory space (MCU types: atmega8, atmega83, atmega85).
10917 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10918 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10919 atmega64, atmega128, at43usb355, at94k).
10921 @item -mno-interrupts
10922 @opindex mno-interrupts
10923 Generated code is not compatible with hardware interrupts.
10924 Code size will be smaller.
10926 @item -mcall-prologues
10927 @opindex mcall-prologues
10928 Functions prologues/epilogues expanded as call to appropriate
10929 subroutines. Code size will be smaller.
10932 @opindex mtiny-stack
10933 Change only the low 8 bits of the stack pointer.
10937 Assume int to be 8 bit integer. This affects the sizes of all types: A
10938 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10939 and long long will be 4 bytes. Please note that this option does not
10940 comply to the C standards, but it will provide you with smaller code
10945 Use register @code{X} in a way proposed by the hardware. This means
10946 that @code{X} will only be used in indirect, post-increment or
10947 pre-decrement addressing.
10949 Without this option, the @code{X} register may be used in the same way
10950 as @code{Y} or @code{Z} which then is emulated by additional
10952 For example, loading a value with @code{X+const} addressing with a
10953 small @code{const <= 63} to a register @var{Rn} will be printed as
10961 @subsubsection @code{EIND} and Devices with more than 128k Bytes of Flash
10963 Pointers in the implementation are 16 bits wide.
10964 The address of a function or label is represented as word address so
10965 that indirect jumps and calls can address any code address in the
10966 range of 64k words.
10968 In order to faciliate indirect jump on devices with more than 128k
10969 bytes of program memory space, there is a special function register called
10970 @code{EIND} that serves as most significant part of the target address
10971 when @code{EICALL} or @code{EIJMP} instructions are used.
10973 Indirect jumps and calls on these devices are handled as follows and
10974 are subject to some limitations:
10979 The compiler never sets @code{EIND}.
10982 The startup code from libgcc never sets @code{EIND}.
10983 Notice that startup code is a blend of code from libgcc and avr-libc.
10984 For the impact of avr-libc on @code{EIND}, see the
10985 @w{@uref{http://nongnu.org/avr-libc/user-manual,avr-libc user manual}}.
10988 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
10989 instructions or might read @code{EIND} directly.
10992 The compiler assumes that @code{EIND} never changes during the startup
10993 code or run of the application. In particular, @code{EIND} is not
10994 saved/restored in function or interrupt service routine
10998 It is legitimate for user-specific startup code to set up @code{EIND}
10999 early, for example by means of initialization code located in
11000 section @code{.init3}, and thus prior to general startup code that
11001 initializes RAM and calls constructors.
11004 For indirect calls to functions and computed goto, the linker will
11005 generate @emph{stubs}. Stubs are jump pads sometimes also called
11006 @emph{trampolines}. Thus, the indirect call/jump will jump to such a stub.
11007 The stub contains a direct jump to the desired address.
11010 Stubs will be generated automatically by the linker if
11011 the following two conditions are met:
11014 @item The address of a label is taken by means of the @code{gs} modifier
11015 (short for @emph{generate stubs}) like so:
11017 LDI r24, lo8(gs(@var{func}))
11018 LDI r25, hi8(gs(@var{func}))
11020 @item The final location of that label is in a code segment
11021 @emph{outside} the segment where the stubs are located.
11025 The compiler will emit such @code{gs} modifiers for code labels in the
11026 following situations:
11028 @item Taking address of a function or code label.
11029 @item Computed goto.
11030 @item If prologue-save function is used, see @option{-mcall-prologues}
11031 command line option.
11032 @item Switch/case dispatch tables. If you do not want such dispatch
11033 tables you can specify the @option{-fno-jump-tables} command line option.
11034 @item C and C++ constructors/destructors called during startup/shutdown.
11035 @item If the tools hit a @code{gs()} modifier explained above.
11039 The default linker script is arranged for code with @code{EIND = 0}.
11040 If code is supposed to work for a setup with @code{EIND != 0}, a custom
11041 linker script has to be used in order to place the sections whose
11042 name start with @code{.trampolines} into the segment where @code{EIND}
11046 Jumping to non-symbolic addresses like so is @emph{not} supported:
11051 /* Call function at word address 0x2 */
11052 return ((int(*)(void)) 0x2)();
11056 Instead, a stub has to be set up:
11061 extern int func_4 (void);
11063 /* Call function at byte address 0x4 */
11068 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
11069 Alternatively, @code{func_4} can be defined in the linker script.
11072 @node Blackfin Options
11073 @subsection Blackfin Options
11074 @cindex Blackfin Options
11077 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
11079 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
11080 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
11081 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
11082 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
11083 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
11084 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
11085 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
11086 @samp{bf561}, @samp{bf592}.
11087 The optional @var{sirevision} specifies the silicon revision of the target
11088 Blackfin processor. Any workarounds available for the targeted silicon revision
11089 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
11090 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
11091 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
11092 hexadecimal digits representing the major and minor numbers in the silicon
11093 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
11094 is not defined. If @var{sirevision} is @samp{any}, the
11095 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
11096 If this optional @var{sirevision} is not used, GCC assumes the latest known
11097 silicon revision of the targeted Blackfin processor.
11099 Support for @samp{bf561} is incomplete. For @samp{bf561},
11100 Only the processor macro is defined.
11101 Without this option, @samp{bf532} is used as the processor by default.
11102 The corresponding predefined processor macros for @var{cpu} is to
11103 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
11104 provided by libgloss to be linked in if @option{-msim} is not given.
11108 Specifies that the program will be run on the simulator. This causes
11109 the simulator BSP provided by libgloss to be linked in. This option
11110 has effect only for @samp{bfin-elf} toolchain.
11111 Certain other options, such as @option{-mid-shared-library} and
11112 @option{-mfdpic}, imply @option{-msim}.
11114 @item -momit-leaf-frame-pointer
11115 @opindex momit-leaf-frame-pointer
11116 Don't keep the frame pointer in a register for leaf functions. This
11117 avoids the instructions to save, set up and restore frame pointers and
11118 makes an extra register available in leaf functions. The option
11119 @option{-fomit-frame-pointer} removes the frame pointer for all functions
11120 which might make debugging harder.
11122 @item -mspecld-anomaly
11123 @opindex mspecld-anomaly
11124 When enabled, the compiler will ensure that the generated code does not
11125 contain speculative loads after jump instructions. If this option is used,
11126 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
11128 @item -mno-specld-anomaly
11129 @opindex mno-specld-anomaly
11130 Don't generate extra code to prevent speculative loads from occurring.
11132 @item -mcsync-anomaly
11133 @opindex mcsync-anomaly
11134 When enabled, the compiler will ensure that the generated code does not
11135 contain CSYNC or SSYNC instructions too soon after conditional branches.
11136 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
11138 @item -mno-csync-anomaly
11139 @opindex mno-csync-anomaly
11140 Don't generate extra code to prevent CSYNC or SSYNC instructions from
11141 occurring too soon after a conditional branch.
11145 When enabled, the compiler is free to take advantage of the knowledge that
11146 the entire program fits into the low 64k of memory.
11149 @opindex mno-low-64k
11150 Assume that the program is arbitrarily large. This is the default.
11152 @item -mstack-check-l1
11153 @opindex mstack-check-l1
11154 Do stack checking using information placed into L1 scratchpad memory by the
11157 @item -mid-shared-library
11158 @opindex mid-shared-library
11159 Generate code that supports shared libraries via the library ID method.
11160 This allows for execute in place and shared libraries in an environment
11161 without virtual memory management. This option implies @option{-fPIC}.
11162 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11164 @item -mno-id-shared-library
11165 @opindex mno-id-shared-library
11166 Generate code that doesn't assume ID based shared libraries are being used.
11167 This is the default.
11169 @item -mleaf-id-shared-library
11170 @opindex mleaf-id-shared-library
11171 Generate code that supports shared libraries via the library ID method,
11172 but assumes that this library or executable won't link against any other
11173 ID shared libraries. That allows the compiler to use faster code for jumps
11176 @item -mno-leaf-id-shared-library
11177 @opindex mno-leaf-id-shared-library
11178 Do not assume that the code being compiled won't link against any ID shared
11179 libraries. Slower code will be generated for jump and call insns.
11181 @item -mshared-library-id=n
11182 @opindex mshared-library-id
11183 Specified the identification number of the ID based shared library being
11184 compiled. Specifying a value of 0 will generate more compact code, specifying
11185 other values will force the allocation of that number to the current
11186 library but is no more space or time efficient than omitting this option.
11190 Generate code that allows the data segment to be located in a different
11191 area of memory from the text segment. This allows for execute in place in
11192 an environment without virtual memory management by eliminating relocations
11193 against the text section.
11195 @item -mno-sep-data
11196 @opindex mno-sep-data
11197 Generate code that assumes that the data segment follows the text segment.
11198 This is the default.
11201 @itemx -mno-long-calls
11202 @opindex mlong-calls
11203 @opindex mno-long-calls
11204 Tells the compiler to perform function calls by first loading the
11205 address of the function into a register and then performing a subroutine
11206 call on this register. This switch is needed if the target function
11207 will lie outside of the 24 bit addressing range of the offset based
11208 version of subroutine call instruction.
11210 This feature is not enabled by default. Specifying
11211 @option{-mno-long-calls} will restore the default behavior. Note these
11212 switches have no effect on how the compiler generates code to handle
11213 function calls via function pointers.
11217 Link with the fast floating-point library. This library relaxes some of
11218 the IEEE floating-point standard's rules for checking inputs against
11219 Not-a-Number (NAN), in the interest of performance.
11222 @opindex minline-plt
11223 Enable inlining of PLT entries in function calls to functions that are
11224 not known to bind locally. It has no effect without @option{-mfdpic}.
11227 @opindex mmulticore
11228 Build standalone application for multicore Blackfin processor. Proper
11229 start files and link scripts will be used to support multicore.
11230 This option defines @code{__BFIN_MULTICORE}. It can only be used with
11231 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
11232 @option{-mcorea} or @option{-mcoreb}. If it's used without
11233 @option{-mcorea} or @option{-mcoreb}, single application/dual core
11234 programming model is used. In this model, the main function of Core B
11235 should be named as coreb_main. If it's used with @option{-mcorea} or
11236 @option{-mcoreb}, one application per core programming model is used.
11237 If this option is not used, single core application programming
11242 Build standalone application for Core A of BF561 when using
11243 one application per core programming model. Proper start files
11244 and link scripts will be used to support Core A. This option
11245 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
11249 Build standalone application for Core B of BF561 when using
11250 one application per core programming model. Proper start files
11251 and link scripts will be used to support Core B. This option
11252 defines @code{__BFIN_COREB}. When this option is used, coreb_main
11253 should be used instead of main. It must be used with
11254 @option{-mmulticore}.
11258 Build standalone application for SDRAM. Proper start files and
11259 link scripts will be used to put the application into SDRAM.
11260 Loader should initialize SDRAM before loading the application
11261 into SDRAM. This option defines @code{__BFIN_SDRAM}.
11265 Assume that ICPLBs are enabled at runtime. This has an effect on certain
11266 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
11267 are enabled; for standalone applications the default is off.
11271 @subsection C6X Options
11272 @cindex C6X Options
11275 @item -march=@var{name}
11277 This specifies the name of the target architecture. GCC uses this
11278 name to determine what kind of instructions it can emit when generating
11279 assembly code. Permissible names are: @samp{c62x},
11280 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
11283 @opindex mbig-endian
11284 Generate code for a big endian target.
11286 @item -mlittle-endian
11287 @opindex mlittle-endian
11288 Generate code for a little endian target. This is the default.
11292 Choose startup files and linker script suitable for the simulator.
11294 @item -msdata=default
11295 @opindex msdata=default
11296 Put small global and static data in the @samp{.neardata} section,
11297 which is pointed to by register @code{B14}. Put small uninitialized
11298 global and static data in the @samp{.bss} section, which is adjacent
11299 to the @samp{.neardata} section. Put small read-only data into the
11300 @samp{.rodata} section. The corresponding sections used for large
11301 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
11304 @opindex msdata=all
11305 Put all data, not just small objets, into the sections reserved for
11306 small data, and use addressing relative to the @code{B14} register to
11310 @opindex msdata=none
11311 Make no use of the sections reserved for small data, and use absolute
11312 addresses to access all data. Put all initialized global and static
11313 data in the @samp{.fardata} section, and all uninitialized data in the
11314 @samp{.far} section. Put all constant data into the @samp{.const}
11319 @subsection CRIS Options
11320 @cindex CRIS Options
11322 These options are defined specifically for the CRIS ports.
11325 @item -march=@var{architecture-type}
11326 @itemx -mcpu=@var{architecture-type}
11329 Generate code for the specified architecture. The choices for
11330 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
11331 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
11332 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
11335 @item -mtune=@var{architecture-type}
11337 Tune to @var{architecture-type} everything applicable about the generated
11338 code, except for the ABI and the set of available instructions. The
11339 choices for @var{architecture-type} are the same as for
11340 @option{-march=@var{architecture-type}}.
11342 @item -mmax-stack-frame=@var{n}
11343 @opindex mmax-stack-frame
11344 Warn when the stack frame of a function exceeds @var{n} bytes.
11350 The options @option{-metrax4} and @option{-metrax100} are synonyms for
11351 @option{-march=v3} and @option{-march=v8} respectively.
11353 @item -mmul-bug-workaround
11354 @itemx -mno-mul-bug-workaround
11355 @opindex mmul-bug-workaround
11356 @opindex mno-mul-bug-workaround
11357 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
11358 models where it applies. This option is active by default.
11362 Enable CRIS-specific verbose debug-related information in the assembly
11363 code. This option also has the effect to turn off the @samp{#NO_APP}
11364 formatted-code indicator to the assembler at the beginning of the
11369 Do not use condition-code results from previous instruction; always emit
11370 compare and test instructions before use of condition codes.
11372 @item -mno-side-effects
11373 @opindex mno-side-effects
11374 Do not emit instructions with side-effects in addressing modes other than
11377 @item -mstack-align
11378 @itemx -mno-stack-align
11379 @itemx -mdata-align
11380 @itemx -mno-data-align
11381 @itemx -mconst-align
11382 @itemx -mno-const-align
11383 @opindex mstack-align
11384 @opindex mno-stack-align
11385 @opindex mdata-align
11386 @opindex mno-data-align
11387 @opindex mconst-align
11388 @opindex mno-const-align
11389 These options (no-options) arranges (eliminate arrangements) for the
11390 stack-frame, individual data and constants to be aligned for the maximum
11391 single data access size for the chosen CPU model. The default is to
11392 arrange for 32-bit alignment. ABI details such as structure layout are
11393 not affected by these options.
11401 Similar to the stack- data- and const-align options above, these options
11402 arrange for stack-frame, writable data and constants to all be 32-bit,
11403 16-bit or 8-bit aligned. The default is 32-bit alignment.
11405 @item -mno-prologue-epilogue
11406 @itemx -mprologue-epilogue
11407 @opindex mno-prologue-epilogue
11408 @opindex mprologue-epilogue
11409 With @option{-mno-prologue-epilogue}, the normal function prologue and
11410 epilogue that sets up the stack-frame are omitted and no return
11411 instructions or return sequences are generated in the code. Use this
11412 option only together with visual inspection of the compiled code: no
11413 warnings or errors are generated when call-saved registers must be saved,
11414 or storage for local variable needs to be allocated.
11418 @opindex mno-gotplt
11420 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11421 instruction sequences that load addresses for functions from the PLT part
11422 of the GOT rather than (traditional on other architectures) calls to the
11423 PLT@. The default is @option{-mgotplt}.
11427 Legacy no-op option only recognized with the cris-axis-elf and
11428 cris-axis-linux-gnu targets.
11432 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11436 This option, recognized for the cris-axis-elf arranges
11437 to link with input-output functions from a simulator library. Code,
11438 initialized data and zero-initialized data are allocated consecutively.
11442 Like @option{-sim}, but pass linker options to locate initialized data at
11443 0x40000000 and zero-initialized data at 0x80000000.
11446 @node Darwin Options
11447 @subsection Darwin Options
11448 @cindex Darwin options
11450 These options are defined for all architectures running the Darwin operating
11453 FSF GCC on Darwin does not create ``fat'' object files; it will create
11454 an object file for the single architecture that it was built to
11455 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11456 @option{-arch} options are used; it does so by running the compiler or
11457 linker multiple times and joining the results together with
11460 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11461 @samp{i686}) is determined by the flags that specify the ISA
11462 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11463 @option{-force_cpusubtype_ALL} option can be used to override this.
11465 The Darwin tools vary in their behavior when presented with an ISA
11466 mismatch. The assembler, @file{as}, will only permit instructions to
11467 be used that are valid for the subtype of the file it is generating,
11468 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11469 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11470 and print an error if asked to create a shared library with a less
11471 restrictive subtype than its input files (for instance, trying to put
11472 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11473 for executables, @file{ld}, will quietly give the executable the most
11474 restrictive subtype of any of its input files.
11479 Add the framework directory @var{dir} to the head of the list of
11480 directories to be searched for header files. These directories are
11481 interleaved with those specified by @option{-I} options and are
11482 scanned in a left-to-right order.
11484 A framework directory is a directory with frameworks in it. A
11485 framework is a directory with a @samp{"Headers"} and/or
11486 @samp{"PrivateHeaders"} directory contained directly in it that ends
11487 in @samp{".framework"}. The name of a framework is the name of this
11488 directory excluding the @samp{".framework"}. Headers associated with
11489 the framework are found in one of those two directories, with
11490 @samp{"Headers"} being searched first. A subframework is a framework
11491 directory that is in a framework's @samp{"Frameworks"} directory.
11492 Includes of subframework headers can only appear in a header of a
11493 framework that contains the subframework, or in a sibling subframework
11494 header. Two subframeworks are siblings if they occur in the same
11495 framework. A subframework should not have the same name as a
11496 framework, a warning will be issued if this is violated. Currently a
11497 subframework cannot have subframeworks, in the future, the mechanism
11498 may be extended to support this. The standard frameworks can be found
11499 in @samp{"/System/Library/Frameworks"} and
11500 @samp{"/Library/Frameworks"}. An example include looks like
11501 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11502 the name of the framework and header.h is found in the
11503 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11505 @item -iframework@var{dir}
11506 @opindex iframework
11507 Like @option{-F} except the directory is a treated as a system
11508 directory. The main difference between this @option{-iframework} and
11509 @option{-F} is that with @option{-iframework} the compiler does not
11510 warn about constructs contained within header files found via
11511 @var{dir}. This option is valid only for the C family of languages.
11515 Emit debugging information for symbols that are used. For STABS
11516 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11517 This is by default ON@.
11521 Emit debugging information for all symbols and types.
11523 @item -mmacosx-version-min=@var{version}
11524 The earliest version of MacOS X that this executable will run on
11525 is @var{version}. Typical values of @var{version} include @code{10.1},
11526 @code{10.2}, and @code{10.3.9}.
11528 If the compiler was built to use the system's headers by default,
11529 then the default for this option is the system version on which the
11530 compiler is running, otherwise the default is to make choices which
11531 are compatible with as many systems and code bases as possible.
11535 Enable kernel development mode. The @option{-mkernel} option sets
11536 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11537 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11538 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11539 applicable. This mode also sets @option{-mno-altivec},
11540 @option{-msoft-float}, @option{-fno-builtin} and
11541 @option{-mlong-branch} for PowerPC targets.
11543 @item -mone-byte-bool
11544 @opindex mone-byte-bool
11545 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11546 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11547 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11548 option has no effect on x86.
11550 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11551 to generate code that is not binary compatible with code generated
11552 without that switch. Using this switch may require recompiling all
11553 other modules in a program, including system libraries. Use this
11554 switch to conform to a non-default data model.
11556 @item -mfix-and-continue
11557 @itemx -ffix-and-continue
11558 @itemx -findirect-data
11559 @opindex mfix-and-continue
11560 @opindex ffix-and-continue
11561 @opindex findirect-data
11562 Generate code suitable for fast turn around development. Needed to
11563 enable gdb to dynamically load @code{.o} files into already running
11564 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11565 are provided for backwards compatibility.
11569 Loads all members of static archive libraries.
11570 See man ld(1) for more information.
11572 @item -arch_errors_fatal
11573 @opindex arch_errors_fatal
11574 Cause the errors having to do with files that have the wrong architecture
11577 @item -bind_at_load
11578 @opindex bind_at_load
11579 Causes the output file to be marked such that the dynamic linker will
11580 bind all undefined references when the file is loaded or launched.
11584 Produce a Mach-o bundle format file.
11585 See man ld(1) for more information.
11587 @item -bundle_loader @var{executable}
11588 @opindex bundle_loader
11589 This option specifies the @var{executable} that will be loading the build
11590 output file being linked. See man ld(1) for more information.
11593 @opindex dynamiclib
11594 When passed this option, GCC will produce a dynamic library instead of
11595 an executable when linking, using the Darwin @file{libtool} command.
11597 @item -force_cpusubtype_ALL
11598 @opindex force_cpusubtype_ALL
11599 This causes GCC's output file to have the @var{ALL} subtype, instead of
11600 one controlled by the @option{-mcpu} or @option{-march} option.
11602 @item -allowable_client @var{client_name}
11603 @itemx -client_name
11604 @itemx -compatibility_version
11605 @itemx -current_version
11607 @itemx -dependency-file
11609 @itemx -dylinker_install_name
11611 @itemx -exported_symbols_list
11614 @itemx -flat_namespace
11615 @itemx -force_flat_namespace
11616 @itemx -headerpad_max_install_names
11619 @itemx -install_name
11620 @itemx -keep_private_externs
11621 @itemx -multi_module
11622 @itemx -multiply_defined
11623 @itemx -multiply_defined_unused
11626 @itemx -no_dead_strip_inits_and_terms
11627 @itemx -nofixprebinding
11628 @itemx -nomultidefs
11630 @itemx -noseglinkedit
11631 @itemx -pagezero_size
11633 @itemx -prebind_all_twolevel_modules
11634 @itemx -private_bundle
11636 @itemx -read_only_relocs
11638 @itemx -sectobjectsymbols
11642 @itemx -sectobjectsymbols
11645 @itemx -segs_read_only_addr
11647 @itemx -segs_read_write_addr
11648 @itemx -seg_addr_table
11649 @itemx -seg_addr_table_filename
11650 @itemx -seglinkedit
11652 @itemx -segs_read_only_addr
11653 @itemx -segs_read_write_addr
11654 @itemx -single_module
11656 @itemx -sub_library
11658 @itemx -sub_umbrella
11659 @itemx -twolevel_namespace
11662 @itemx -unexported_symbols_list
11663 @itemx -weak_reference_mismatches
11664 @itemx -whatsloaded
11665 @opindex allowable_client
11666 @opindex client_name
11667 @opindex compatibility_version
11668 @opindex current_version
11669 @opindex dead_strip
11670 @opindex dependency-file
11671 @opindex dylib_file
11672 @opindex dylinker_install_name
11674 @opindex exported_symbols_list
11676 @opindex flat_namespace
11677 @opindex force_flat_namespace
11678 @opindex headerpad_max_install_names
11679 @opindex image_base
11681 @opindex install_name
11682 @opindex keep_private_externs
11683 @opindex multi_module
11684 @opindex multiply_defined
11685 @opindex multiply_defined_unused
11686 @opindex noall_load
11687 @opindex no_dead_strip_inits_and_terms
11688 @opindex nofixprebinding
11689 @opindex nomultidefs
11691 @opindex noseglinkedit
11692 @opindex pagezero_size
11694 @opindex prebind_all_twolevel_modules
11695 @opindex private_bundle
11696 @opindex read_only_relocs
11698 @opindex sectobjectsymbols
11701 @opindex sectcreate
11702 @opindex sectobjectsymbols
11705 @opindex segs_read_only_addr
11706 @opindex segs_read_write_addr
11707 @opindex seg_addr_table
11708 @opindex seg_addr_table_filename
11709 @opindex seglinkedit
11711 @opindex segs_read_only_addr
11712 @opindex segs_read_write_addr
11713 @opindex single_module
11715 @opindex sub_library
11716 @opindex sub_umbrella
11717 @opindex twolevel_namespace
11720 @opindex unexported_symbols_list
11721 @opindex weak_reference_mismatches
11722 @opindex whatsloaded
11723 These options are passed to the Darwin linker. The Darwin linker man page
11724 describes them in detail.
11727 @node DEC Alpha Options
11728 @subsection DEC Alpha Options
11730 These @samp{-m} options are defined for the DEC Alpha implementations:
11733 @item -mno-soft-float
11734 @itemx -msoft-float
11735 @opindex mno-soft-float
11736 @opindex msoft-float
11737 Use (do not use) the hardware floating-point instructions for
11738 floating-point operations. When @option{-msoft-float} is specified,
11739 functions in @file{libgcc.a} will be used to perform floating-point
11740 operations. Unless they are replaced by routines that emulate the
11741 floating-point operations, or compiled in such a way as to call such
11742 emulations routines, these routines will issue floating-point
11743 operations. If you are compiling for an Alpha without floating-point
11744 operations, you must ensure that the library is built so as not to call
11747 Note that Alpha implementations without floating-point operations are
11748 required to have floating-point registers.
11751 @itemx -mno-fp-regs
11753 @opindex mno-fp-regs
11754 Generate code that uses (does not use) the floating-point register set.
11755 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11756 register set is not used, floating point operands are passed in integer
11757 registers as if they were integers and floating-point results are passed
11758 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11759 so any function with a floating-point argument or return value called by code
11760 compiled with @option{-mno-fp-regs} must also be compiled with that
11763 A typical use of this option is building a kernel that does not use,
11764 and hence need not save and restore, any floating-point registers.
11768 The Alpha architecture implements floating-point hardware optimized for
11769 maximum performance. It is mostly compliant with the IEEE floating
11770 point standard. However, for full compliance, software assistance is
11771 required. This option generates code fully IEEE compliant code
11772 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11773 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11774 defined during compilation. The resulting code is less efficient but is
11775 able to correctly support denormalized numbers and exceptional IEEE
11776 values such as not-a-number and plus/minus infinity. Other Alpha
11777 compilers call this option @option{-ieee_with_no_inexact}.
11779 @item -mieee-with-inexact
11780 @opindex mieee-with-inexact
11781 This is like @option{-mieee} except the generated code also maintains
11782 the IEEE @var{inexact-flag}. Turning on this option causes the
11783 generated code to implement fully-compliant IEEE math. In addition to
11784 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11785 macro. On some Alpha implementations the resulting code may execute
11786 significantly slower than the code generated by default. Since there is
11787 very little code that depends on the @var{inexact-flag}, you should
11788 normally not specify this option. Other Alpha compilers call this
11789 option @option{-ieee_with_inexact}.
11791 @item -mfp-trap-mode=@var{trap-mode}
11792 @opindex mfp-trap-mode
11793 This option controls what floating-point related traps are enabled.
11794 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11795 The trap mode can be set to one of four values:
11799 This is the default (normal) setting. The only traps that are enabled
11800 are the ones that cannot be disabled in software (e.g., division by zero
11804 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11808 Like @samp{u}, but the instructions are marked to be safe for software
11809 completion (see Alpha architecture manual for details).
11812 Like @samp{su}, but inexact traps are enabled as well.
11815 @item -mfp-rounding-mode=@var{rounding-mode}
11816 @opindex mfp-rounding-mode
11817 Selects the IEEE rounding mode. Other Alpha compilers call this option
11818 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11823 Normal IEEE rounding mode. Floating point numbers are rounded towards
11824 the nearest machine number or towards the even machine number in case
11828 Round towards minus infinity.
11831 Chopped rounding mode. Floating point numbers are rounded towards zero.
11834 Dynamic rounding mode. A field in the floating point control register
11835 (@var{fpcr}, see Alpha architecture reference manual) controls the
11836 rounding mode in effect. The C library initializes this register for
11837 rounding towards plus infinity. Thus, unless your program modifies the
11838 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11841 @item -mtrap-precision=@var{trap-precision}
11842 @opindex mtrap-precision
11843 In the Alpha architecture, floating point traps are imprecise. This
11844 means without software assistance it is impossible to recover from a
11845 floating trap and program execution normally needs to be terminated.
11846 GCC can generate code that can assist operating system trap handlers
11847 in determining the exact location that caused a floating point trap.
11848 Depending on the requirements of an application, different levels of
11849 precisions can be selected:
11853 Program precision. This option is the default and means a trap handler
11854 can only identify which program caused a floating point exception.
11857 Function precision. The trap handler can determine the function that
11858 caused a floating point exception.
11861 Instruction precision. The trap handler can determine the exact
11862 instruction that caused a floating point exception.
11865 Other Alpha compilers provide the equivalent options called
11866 @option{-scope_safe} and @option{-resumption_safe}.
11868 @item -mieee-conformant
11869 @opindex mieee-conformant
11870 This option marks the generated code as IEEE conformant. You must not
11871 use this option unless you also specify @option{-mtrap-precision=i} and either
11872 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11873 is to emit the line @samp{.eflag 48} in the function prologue of the
11874 generated assembly file. Under DEC Unix, this has the effect that
11875 IEEE-conformant math library routines will be linked in.
11877 @item -mbuild-constants
11878 @opindex mbuild-constants
11879 Normally GCC examines a 32- or 64-bit integer constant to
11880 see if it can construct it from smaller constants in two or three
11881 instructions. If it cannot, it will output the constant as a literal and
11882 generate code to load it from the data segment at runtime.
11884 Use this option to require GCC to construct @emph{all} integer constants
11885 using code, even if it takes more instructions (the maximum is six).
11887 You would typically use this option to build a shared library dynamic
11888 loader. Itself a shared library, it must relocate itself in memory
11889 before it can find the variables and constants in its own data segment.
11895 Select whether to generate code to be assembled by the vendor-supplied
11896 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11914 Indicate whether GCC should generate code to use the optional BWX,
11915 CIX, FIX and MAX instruction sets. The default is to use the instruction
11916 sets supported by the CPU type specified via @option{-mcpu=} option or that
11917 of the CPU on which GCC was built if none was specified.
11920 @itemx -mfloat-ieee
11921 @opindex mfloat-vax
11922 @opindex mfloat-ieee
11923 Generate code that uses (does not use) VAX F and G floating point
11924 arithmetic instead of IEEE single and double precision.
11926 @item -mexplicit-relocs
11927 @itemx -mno-explicit-relocs
11928 @opindex mexplicit-relocs
11929 @opindex mno-explicit-relocs
11930 Older Alpha assemblers provided no way to generate symbol relocations
11931 except via assembler macros. Use of these macros does not allow
11932 optimal instruction scheduling. GNU binutils as of version 2.12
11933 supports a new syntax that allows the compiler to explicitly mark
11934 which relocations should apply to which instructions. This option
11935 is mostly useful for debugging, as GCC detects the capabilities of
11936 the assembler when it is built and sets the default accordingly.
11939 @itemx -mlarge-data
11940 @opindex msmall-data
11941 @opindex mlarge-data
11942 When @option{-mexplicit-relocs} is in effect, static data is
11943 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11944 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11945 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11946 16-bit relocations off of the @code{$gp} register. This limits the
11947 size of the small data area to 64KB, but allows the variables to be
11948 directly accessed via a single instruction.
11950 The default is @option{-mlarge-data}. With this option the data area
11951 is limited to just below 2GB@. Programs that require more than 2GB of
11952 data must use @code{malloc} or @code{mmap} to allocate the data in the
11953 heap instead of in the program's data segment.
11955 When generating code for shared libraries, @option{-fpic} implies
11956 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11959 @itemx -mlarge-text
11960 @opindex msmall-text
11961 @opindex mlarge-text
11962 When @option{-msmall-text} is used, the compiler assumes that the
11963 code of the entire program (or shared library) fits in 4MB, and is
11964 thus reachable with a branch instruction. When @option{-msmall-data}
11965 is used, the compiler can assume that all local symbols share the
11966 same @code{$gp} value, and thus reduce the number of instructions
11967 required for a function call from 4 to 1.
11969 The default is @option{-mlarge-text}.
11971 @item -mcpu=@var{cpu_type}
11973 Set the instruction set and instruction scheduling parameters for
11974 machine type @var{cpu_type}. You can specify either the @samp{EV}
11975 style name or the corresponding chip number. GCC supports scheduling
11976 parameters for the EV4, EV5 and EV6 family of processors and will
11977 choose the default values for the instruction set from the processor
11978 you specify. If you do not specify a processor type, GCC will default
11979 to the processor on which the compiler was built.
11981 Supported values for @var{cpu_type} are
11987 Schedules as an EV4 and has no instruction set extensions.
11991 Schedules as an EV5 and has no instruction set extensions.
11995 Schedules as an EV5 and supports the BWX extension.
12000 Schedules as an EV5 and supports the BWX and MAX extensions.
12004 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
12008 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
12011 Native toolchains also support the value @samp{native},
12012 which selects the best architecture option for the host processor.
12013 @option{-mcpu=native} has no effect if GCC does not recognize
12016 @item -mtune=@var{cpu_type}
12018 Set only the instruction scheduling parameters for machine type
12019 @var{cpu_type}. The instruction set is not changed.
12021 Native toolchains also support the value @samp{native},
12022 which selects the best architecture option for the host processor.
12023 @option{-mtune=native} has no effect if GCC does not recognize
12026 @item -mmemory-latency=@var{time}
12027 @opindex mmemory-latency
12028 Sets the latency the scheduler should assume for typical memory
12029 references as seen by the application. This number is highly
12030 dependent on the memory access patterns used by the application
12031 and the size of the external cache on the machine.
12033 Valid options for @var{time} are
12037 A decimal number representing clock cycles.
12043 The compiler contains estimates of the number of clock cycles for
12044 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
12045 (also called Dcache, Scache, and Bcache), as well as to main memory.
12046 Note that L3 is only valid for EV5.
12051 @node DEC Alpha/VMS Options
12052 @subsection DEC Alpha/VMS Options
12054 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
12057 @item -mvms-return-codes
12058 @opindex mvms-return-codes
12059 Return VMS condition codes from main. The default is to return POSIX
12060 style condition (e.g.@: error) codes.
12062 @item -mdebug-main=@var{prefix}
12063 @opindex mdebug-main=@var{prefix}
12064 Flag the first routine whose name starts with @var{prefix} as the main
12065 routine for the debugger.
12069 Default to 64bit memory allocation routines.
12073 @subsection FR30 Options
12074 @cindex FR30 Options
12076 These options are defined specifically for the FR30 port.
12080 @item -msmall-model
12081 @opindex msmall-model
12082 Use the small address space model. This can produce smaller code, but
12083 it does assume that all symbolic values and addresses will fit into a
12088 Assume that run-time support has been provided and so there is no need
12089 to include the simulator library (@file{libsim.a}) on the linker
12095 @subsection FRV Options
12096 @cindex FRV Options
12102 Only use the first 32 general purpose registers.
12107 Use all 64 general purpose registers.
12112 Use only the first 32 floating point registers.
12117 Use all 64 floating point registers
12120 @opindex mhard-float
12122 Use hardware instructions for floating point operations.
12125 @opindex msoft-float
12127 Use library routines for floating point operations.
12132 Dynamically allocate condition code registers.
12137 Do not try to dynamically allocate condition code registers, only
12138 use @code{icc0} and @code{fcc0}.
12143 Change ABI to use double word insns.
12148 Do not use double word instructions.
12153 Use floating point double instructions.
12156 @opindex mno-double
12158 Do not use floating point double instructions.
12163 Use media instructions.
12168 Do not use media instructions.
12173 Use multiply and add/subtract instructions.
12176 @opindex mno-muladd
12178 Do not use multiply and add/subtract instructions.
12183 Select the FDPIC ABI, that uses function descriptors to represent
12184 pointers to functions. Without any PIC/PIE-related options, it
12185 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
12186 assumes GOT entries and small data are within a 12-bit range from the
12187 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
12188 are computed with 32 bits.
12189 With a @samp{bfin-elf} target, this option implies @option{-msim}.
12192 @opindex minline-plt
12194 Enable inlining of PLT entries in function calls to functions that are
12195 not known to bind locally. It has no effect without @option{-mfdpic}.
12196 It's enabled by default if optimizing for speed and compiling for
12197 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
12198 optimization option such as @option{-O3} or above is present in the
12204 Assume a large TLS segment when generating thread-local code.
12209 Do not assume a large TLS segment when generating thread-local code.
12214 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
12215 that is known to be in read-only sections. It's enabled by default,
12216 except for @option{-fpic} or @option{-fpie}: even though it may help
12217 make the global offset table smaller, it trades 1 instruction for 4.
12218 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
12219 one of which may be shared by multiple symbols, and it avoids the need
12220 for a GOT entry for the referenced symbol, so it's more likely to be a
12221 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
12223 @item -multilib-library-pic
12224 @opindex multilib-library-pic
12226 Link with the (library, not FD) pic libraries. It's implied by
12227 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
12228 @option{-fpic} without @option{-mfdpic}. You should never have to use
12232 @opindex mlinked-fp
12234 Follow the EABI requirement of always creating a frame pointer whenever
12235 a stack frame is allocated. This option is enabled by default and can
12236 be disabled with @option{-mno-linked-fp}.
12239 @opindex mlong-calls
12241 Use indirect addressing to call functions outside the current
12242 compilation unit. This allows the functions to be placed anywhere
12243 within the 32-bit address space.
12245 @item -malign-labels
12246 @opindex malign-labels
12248 Try to align labels to an 8-byte boundary by inserting nops into the
12249 previous packet. This option only has an effect when VLIW packing
12250 is enabled. It doesn't create new packets; it merely adds nops to
12253 @item -mlibrary-pic
12254 @opindex mlibrary-pic
12256 Generate position-independent EABI code.
12261 Use only the first four media accumulator registers.
12266 Use all eight media accumulator registers.
12271 Pack VLIW instructions.
12276 Do not pack VLIW instructions.
12279 @opindex mno-eflags
12281 Do not mark ABI switches in e_flags.
12284 @opindex mcond-move
12286 Enable the use of conditional-move instructions (default).
12288 This switch is mainly for debugging the compiler and will likely be removed
12289 in a future version.
12291 @item -mno-cond-move
12292 @opindex mno-cond-move
12294 Disable the use of conditional-move instructions.
12296 This switch is mainly for debugging the compiler and will likely be removed
12297 in a future version.
12302 Enable the use of conditional set instructions (default).
12304 This switch is mainly for debugging the compiler and will likely be removed
12305 in a future version.
12310 Disable the use of conditional set instructions.
12312 This switch is mainly for debugging the compiler and will likely be removed
12313 in a future version.
12316 @opindex mcond-exec
12318 Enable the use of conditional execution (default).
12320 This switch is mainly for debugging the compiler and will likely be removed
12321 in a future version.
12323 @item -mno-cond-exec
12324 @opindex mno-cond-exec
12326 Disable the use of conditional execution.
12328 This switch is mainly for debugging the compiler and will likely be removed
12329 in a future version.
12331 @item -mvliw-branch
12332 @opindex mvliw-branch
12334 Run a pass to pack branches into VLIW instructions (default).
12336 This switch is mainly for debugging the compiler and will likely be removed
12337 in a future version.
12339 @item -mno-vliw-branch
12340 @opindex mno-vliw-branch
12342 Do not run a pass to pack branches into VLIW instructions.
12344 This switch is mainly for debugging the compiler and will likely be removed
12345 in a future version.
12347 @item -mmulti-cond-exec
12348 @opindex mmulti-cond-exec
12350 Enable optimization of @code{&&} and @code{||} in conditional execution
12353 This switch is mainly for debugging the compiler and will likely be removed
12354 in a future version.
12356 @item -mno-multi-cond-exec
12357 @opindex mno-multi-cond-exec
12359 Disable optimization of @code{&&} and @code{||} in conditional execution.
12361 This switch is mainly for debugging the compiler and will likely be removed
12362 in a future version.
12364 @item -mnested-cond-exec
12365 @opindex mnested-cond-exec
12367 Enable nested conditional execution optimizations (default).
12369 This switch is mainly for debugging the compiler and will likely be removed
12370 in a future version.
12372 @item -mno-nested-cond-exec
12373 @opindex mno-nested-cond-exec
12375 Disable nested conditional execution optimizations.
12377 This switch is mainly for debugging the compiler and will likely be removed
12378 in a future version.
12380 @item -moptimize-membar
12381 @opindex moptimize-membar
12383 This switch removes redundant @code{membar} instructions from the
12384 compiler generated code. It is enabled by default.
12386 @item -mno-optimize-membar
12387 @opindex mno-optimize-membar
12389 This switch disables the automatic removal of redundant @code{membar}
12390 instructions from the generated code.
12392 @item -mtomcat-stats
12393 @opindex mtomcat-stats
12395 Cause gas to print out tomcat statistics.
12397 @item -mcpu=@var{cpu}
12400 Select the processor type for which to generate code. Possible values are
12401 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12402 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12406 @node GNU/Linux Options
12407 @subsection GNU/Linux Options
12409 These @samp{-m} options are defined for GNU/Linux targets:
12414 Use the GNU C library. This is the default except
12415 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12419 Use uClibc C library. This is the default on
12420 @samp{*-*-linux-*uclibc*} targets.
12424 Use Bionic C library. This is the default on
12425 @samp{*-*-linux-*android*} targets.
12429 Compile code compatible with Android platform. This is the default on
12430 @samp{*-*-linux-*android*} targets.
12432 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12433 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12434 this option makes the GCC driver pass Android-specific options to the linker.
12435 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12438 @item -tno-android-cc
12439 @opindex tno-android-cc
12440 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12441 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12442 @option{-fno-rtti} by default.
12444 @item -tno-android-ld
12445 @opindex tno-android-ld
12446 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12447 linking options to the linker.
12451 @node H8/300 Options
12452 @subsection H8/300 Options
12454 These @samp{-m} options are defined for the H8/300 implementations:
12459 Shorten some address references at link time, when possible; uses the
12460 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12461 ld, Using ld}, for a fuller description.
12465 Generate code for the H8/300H@.
12469 Generate code for the H8S@.
12473 Generate code for the H8S and H8/300H in the normal mode. This switch
12474 must be used either with @option{-mh} or @option{-ms}.
12478 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12482 Make @code{int} data 32 bits by default.
12485 @opindex malign-300
12486 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12487 The default for the H8/300H and H8S is to align longs and floats on 4
12489 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12490 This option has no effect on the H8/300.
12494 @subsection HPPA Options
12495 @cindex HPPA Options
12497 These @samp{-m} options are defined for the HPPA family of computers:
12500 @item -march=@var{architecture-type}
12502 Generate code for the specified architecture. The choices for
12503 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12504 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12505 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12506 architecture option for your machine. Code compiled for lower numbered
12507 architectures will run on higher numbered architectures, but not the
12510 @item -mpa-risc-1-0
12511 @itemx -mpa-risc-1-1
12512 @itemx -mpa-risc-2-0
12513 @opindex mpa-risc-1-0
12514 @opindex mpa-risc-1-1
12515 @opindex mpa-risc-2-0
12516 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12519 @opindex mbig-switch
12520 Generate code suitable for big switch tables. Use this option only if
12521 the assembler/linker complain about out of range branches within a switch
12524 @item -mjump-in-delay
12525 @opindex mjump-in-delay
12526 Fill delay slots of function calls with unconditional jump instructions
12527 by modifying the return pointer for the function call to be the target
12528 of the conditional jump.
12530 @item -mdisable-fpregs
12531 @opindex mdisable-fpregs
12532 Prevent floating point registers from being used in any manner. This is
12533 necessary for compiling kernels which perform lazy context switching of
12534 floating point registers. If you use this option and attempt to perform
12535 floating point operations, the compiler will abort.
12537 @item -mdisable-indexing
12538 @opindex mdisable-indexing
12539 Prevent the compiler from using indexing address modes. This avoids some
12540 rather obscure problems when compiling MIG generated code under MACH@.
12542 @item -mno-space-regs
12543 @opindex mno-space-regs
12544 Generate code that assumes the target has no space registers. This allows
12545 GCC to generate faster indirect calls and use unscaled index address modes.
12547 Such code is suitable for level 0 PA systems and kernels.
12549 @item -mfast-indirect-calls
12550 @opindex mfast-indirect-calls
12551 Generate code that assumes calls never cross space boundaries. This
12552 allows GCC to emit code which performs faster indirect calls.
12554 This option will not work in the presence of shared libraries or nested
12557 @item -mfixed-range=@var{register-range}
12558 @opindex mfixed-range
12559 Generate code treating the given register range as fixed registers.
12560 A fixed register is one that the register allocator can not use. This is
12561 useful when compiling kernel code. A register range is specified as
12562 two registers separated by a dash. Multiple register ranges can be
12563 specified separated by a comma.
12565 @item -mlong-load-store
12566 @opindex mlong-load-store
12567 Generate 3-instruction load and store sequences as sometimes required by
12568 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12571 @item -mportable-runtime
12572 @opindex mportable-runtime
12573 Use the portable calling conventions proposed by HP for ELF systems.
12577 Enable the use of assembler directives only GAS understands.
12579 @item -mschedule=@var{cpu-type}
12581 Schedule code according to the constraints for the machine type
12582 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12583 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12584 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12585 proper scheduling option for your machine. The default scheduling is
12589 @opindex mlinker-opt
12590 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12591 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12592 linkers in which they give bogus error messages when linking some programs.
12595 @opindex msoft-float
12596 Generate output containing library calls for floating point.
12597 @strong{Warning:} the requisite libraries are not available for all HPPA
12598 targets. Normally the facilities of the machine's usual C compiler are
12599 used, but this cannot be done directly in cross-compilation. You must make
12600 your own arrangements to provide suitable library functions for
12603 @option{-msoft-float} changes the calling convention in the output file;
12604 therefore, it is only useful if you compile @emph{all} of a program with
12605 this option. In particular, you need to compile @file{libgcc.a}, the
12606 library that comes with GCC, with @option{-msoft-float} in order for
12611 Generate the predefine, @code{_SIO}, for server IO@. The default is
12612 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12613 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12614 options are available under HP-UX and HI-UX@.
12618 Use GNU ld specific options. This passes @option{-shared} to ld when
12619 building a shared library. It is the default when GCC is configured,
12620 explicitly or implicitly, with the GNU linker. This option does not
12621 have any affect on which ld is called, it only changes what parameters
12622 are passed to that ld. The ld that is called is determined by the
12623 @option{--with-ld} configure option, GCC's program search path, and
12624 finally by the user's @env{PATH}. The linker used by GCC can be printed
12625 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12626 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12630 Use HP ld specific options. This passes @option{-b} to ld when building
12631 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12632 links. It is the default when GCC is configured, explicitly or
12633 implicitly, with the HP linker. This option does not have any affect on
12634 which ld is called, it only changes what parameters are passed to that
12635 ld. The ld that is called is determined by the @option{--with-ld}
12636 configure option, GCC's program search path, and finally by the user's
12637 @env{PATH}. The linker used by GCC can be printed using @samp{which
12638 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12639 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12642 @opindex mno-long-calls
12643 Generate code that uses long call sequences. This ensures that a call
12644 is always able to reach linker generated stubs. The default is to generate
12645 long calls only when the distance from the call site to the beginning
12646 of the function or translation unit, as the case may be, exceeds a
12647 predefined limit set by the branch type being used. The limits for
12648 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12649 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12652 Distances are measured from the beginning of functions when using the
12653 @option{-ffunction-sections} option, or when using the @option{-mgas}
12654 and @option{-mno-portable-runtime} options together under HP-UX with
12657 It is normally not desirable to use this option as it will degrade
12658 performance. However, it may be useful in large applications,
12659 particularly when partial linking is used to build the application.
12661 The types of long calls used depends on the capabilities of the
12662 assembler and linker, and the type of code being generated. The
12663 impact on systems that support long absolute calls, and long pic
12664 symbol-difference or pc-relative calls should be relatively small.
12665 However, an indirect call is used on 32-bit ELF systems in pic code
12666 and it is quite long.
12668 @item -munix=@var{unix-std}
12670 Generate compiler predefines and select a startfile for the specified
12671 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12672 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12673 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12674 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12675 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12678 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12679 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12680 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12681 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12682 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12683 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12685 It is @emph{important} to note that this option changes the interfaces
12686 for various library routines. It also affects the operational behavior
12687 of the C library. Thus, @emph{extreme} care is needed in using this
12690 Library code that is intended to operate with more than one UNIX
12691 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12692 as appropriate. Most GNU software doesn't provide this capability.
12696 Suppress the generation of link options to search libdld.sl when the
12697 @option{-static} option is specified on HP-UX 10 and later.
12701 The HP-UX implementation of setlocale in libc has a dependency on
12702 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12703 when the @option{-static} option is specified, special link options
12704 are needed to resolve this dependency.
12706 On HP-UX 10 and later, the GCC driver adds the necessary options to
12707 link with libdld.sl when the @option{-static} option is specified.
12708 This causes the resulting binary to be dynamic. On the 64-bit port,
12709 the linkers generate dynamic binaries by default in any case. The
12710 @option{-nolibdld} option can be used to prevent the GCC driver from
12711 adding these link options.
12715 Add support for multithreading with the @dfn{dce thread} library
12716 under HP-UX@. This option sets flags for both the preprocessor and
12720 @node i386 and x86-64 Options
12721 @subsection Intel 386 and AMD x86-64 Options
12722 @cindex i386 Options
12723 @cindex x86-64 Options
12724 @cindex Intel 386 Options
12725 @cindex AMD x86-64 Options
12727 These @samp{-m} options are defined for the i386 and x86-64 family of
12731 @item -mtune=@var{cpu-type}
12733 Tune to @var{cpu-type} everything applicable about the generated code, except
12734 for the ABI and the set of available instructions. The choices for
12735 @var{cpu-type} are:
12738 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12739 If you know the CPU on which your code will run, then you should use
12740 the corresponding @option{-mtune} option instead of
12741 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12742 of your application will have, then you should use this option.
12744 As new processors are deployed in the marketplace, the behavior of this
12745 option will change. Therefore, if you upgrade to a newer version of
12746 GCC, the code generated option will change to reflect the processors
12747 that were most common when that version of GCC was released.
12749 There is no @option{-march=generic} option because @option{-march}
12750 indicates the instruction set the compiler can use, and there is no
12751 generic instruction set applicable to all processors. In contrast,
12752 @option{-mtune} indicates the processor (or, in this case, collection of
12753 processors) for which the code is optimized.
12755 This selects the CPU to tune for at compilation time by determining
12756 the processor type of the compiling machine. Using @option{-mtune=native}
12757 will produce code optimized for the local machine under the constraints
12758 of the selected instruction set. Using @option{-march=native} will
12759 enable all instruction subsets supported by the local machine (hence
12760 the result might not run on different machines).
12762 Original Intel's i386 CPU@.
12764 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12765 @item i586, pentium
12766 Intel Pentium CPU with no MMX support.
12768 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12770 Intel PentiumPro CPU@.
12772 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12773 instruction set will be used, so the code will run on all i686 family chips.
12775 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12776 @item pentium3, pentium3m
12777 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12780 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12781 support. Used by Centrino notebooks.
12782 @item pentium4, pentium4m
12783 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12785 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12788 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12789 SSE2 and SSE3 instruction set support.
12791 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12792 instruction set support.
12794 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12795 and SSE4.2 instruction set support.
12797 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12798 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12800 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12801 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
12804 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12805 instruction set support.
12807 AMD K6 CPU with MMX instruction set support.
12809 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12810 @item athlon, athlon-tbird
12811 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12813 @item athlon-4, athlon-xp, athlon-mp
12814 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12815 instruction set support.
12816 @item k8, opteron, athlon64, athlon-fx
12817 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12818 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12819 @item k8-sse3, opteron-sse3, athlon64-sse3
12820 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12821 @item amdfam10, barcelona
12822 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12823 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12824 instruction set extensions.)
12826 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
12827 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
12828 SSSE3, SSE4.1, SSE4.2, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12829 instruction set extensions.)
12831 AMD Family 14h core based CPUs with x86-64 instruction set support. (This
12832 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
12833 instruction set extensions.)
12835 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12838 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12839 instruction set support.
12841 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12842 implemented for this chip.)
12844 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12845 implemented for this chip.)
12847 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12850 While picking a specific @var{cpu-type} will schedule things appropriately
12851 for that particular chip, the compiler will not generate any code that
12852 does not run on the default machine type without the @option{-march=@var{cpu-type}}
12853 option being used. For example, if GCC is configured for i686-pc-linux-gnu
12854 then @option{-mtune=pentium4} will generate code that is tuned for Pentium4
12855 but will still run on i686 machines.
12857 @item -march=@var{cpu-type}
12859 Generate instructions for the machine type @var{cpu-type}. The choices
12860 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12861 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12863 @item -mcpu=@var{cpu-type}
12865 A deprecated synonym for @option{-mtune}.
12867 @item -mfpmath=@var{unit}
12869 Generate floating point arithmetics for selected unit @var{unit}. The choices
12870 for @var{unit} are:
12874 Use the standard 387 floating point coprocessor present majority of chips and
12875 emulated otherwise. Code compiled with this option will run almost everywhere.
12876 The temporary results are computed in 80bit precision instead of precision
12877 specified by the type resulting in slightly different results compared to most
12878 of other chips. See @option{-ffloat-store} for more detailed description.
12880 This is the default choice for i386 compiler.
12883 Use scalar floating point instructions present in the SSE instruction set.
12884 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12885 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12886 instruction set supports only single precision arithmetics, thus the double and
12887 extended precision arithmetics is still done using 387. Later version, present
12888 only in Pentium4 and the future AMD x86-64 chips supports double precision
12891 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12892 or @option{-msse2} switches to enable SSE extensions and make this option
12893 effective. For the x86-64 compiler, these extensions are enabled by default.
12895 The resulting code should be considerably faster in the majority of cases and avoid
12896 the numerical instability problems of 387 code, but may break some existing
12897 code that expects temporaries to be 80bit.
12899 This is the default choice for the x86-64 compiler.
12904 Attempt to utilize both instruction sets at once. This effectively double the
12905 amount of available registers and on chips with separate execution units for
12906 387 and SSE the execution resources too. Use this option with care, as it is
12907 still experimental, because the GCC register allocator does not model separate
12908 functional units well resulting in instable performance.
12911 @item -masm=@var{dialect}
12912 @opindex masm=@var{dialect}
12913 Output asm instructions using selected @var{dialect}. Supported
12914 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12915 not support @samp{intel}.
12918 @itemx -mno-ieee-fp
12920 @opindex mno-ieee-fp
12921 Control whether or not the compiler uses IEEE floating point
12922 comparisons. These handle correctly the case where the result of a
12923 comparison is unordered.
12926 @opindex msoft-float
12927 Generate output containing library calls for floating point.
12928 @strong{Warning:} the requisite libraries are not part of GCC@.
12929 Normally the facilities of the machine's usual C compiler are used, but
12930 this can't be done directly in cross-compilation. You must make your
12931 own arrangements to provide suitable library functions for
12934 On machines where a function returns floating point results in the 80387
12935 register stack, some floating point opcodes may be emitted even if
12936 @option{-msoft-float} is used.
12938 @item -mno-fp-ret-in-387
12939 @opindex mno-fp-ret-in-387
12940 Do not use the FPU registers for return values of functions.
12942 The usual calling convention has functions return values of types
12943 @code{float} and @code{double} in an FPU register, even if there
12944 is no FPU@. The idea is that the operating system should emulate
12947 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12948 in ordinary CPU registers instead.
12950 @item -mno-fancy-math-387
12951 @opindex mno-fancy-math-387
12952 Some 387 emulators do not support the @code{sin}, @code{cos} and
12953 @code{sqrt} instructions for the 387. Specify this option to avoid
12954 generating those instructions. This option is the default on FreeBSD,
12955 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12956 indicates that the target CPU will always have an FPU and so the
12957 instruction will not need emulation. As of revision 2.6.1, these
12958 instructions are not generated unless you also use the
12959 @option{-funsafe-math-optimizations} switch.
12961 @item -malign-double
12962 @itemx -mno-align-double
12963 @opindex malign-double
12964 @opindex mno-align-double
12965 Control whether GCC aligns @code{double}, @code{long double}, and
12966 @code{long long} variables on a two word boundary or a one word
12967 boundary. Aligning @code{double} variables on a two word boundary will
12968 produce code that runs somewhat faster on a @samp{Pentium} at the
12969 expense of more memory.
12971 On x86-64, @option{-malign-double} is enabled by default.
12973 @strong{Warning:} if you use the @option{-malign-double} switch,
12974 structures containing the above types will be aligned differently than
12975 the published application binary interface specifications for the 386
12976 and will not be binary compatible with structures in code compiled
12977 without that switch.
12979 @item -m96bit-long-double
12980 @itemx -m128bit-long-double
12981 @opindex m96bit-long-double
12982 @opindex m128bit-long-double
12983 These switches control the size of @code{long double} type. The i386
12984 application binary interface specifies the size to be 96 bits,
12985 so @option{-m96bit-long-double} is the default in 32 bit mode.
12987 Modern architectures (Pentium and newer) would prefer @code{long double}
12988 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12989 conforming to the ABI, this would not be possible. So specifying a
12990 @option{-m128bit-long-double} will align @code{long double}
12991 to a 16 byte boundary by padding the @code{long double} with an additional
12994 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12995 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12997 Notice that neither of these options enable any extra precision over the x87
12998 standard of 80 bits for a @code{long double}.
13000 @strong{Warning:} if you override the default value for your target ABI, the
13001 structures and arrays containing @code{long double} variables will change
13002 their size as well as function calling convention for function taking
13003 @code{long double} will be modified. Hence they will not be binary
13004 compatible with arrays or structures in code compiled without that switch.
13006 @item -mlarge-data-threshold=@var{number}
13007 @opindex mlarge-data-threshold=@var{number}
13008 When @option{-mcmodel=medium} is specified, the data greater than
13009 @var{threshold} are placed in large data section. This value must be the
13010 same across all object linked into the binary and defaults to 65535.
13014 Use a different function-calling convention, in which functions that
13015 take a fixed number of arguments return with the @code{ret} @var{num}
13016 instruction, which pops their arguments while returning. This saves one
13017 instruction in the caller since there is no need to pop the arguments
13020 You can specify that an individual function is called with this calling
13021 sequence with the function attribute @samp{stdcall}. You can also
13022 override the @option{-mrtd} option by using the function attribute
13023 @samp{cdecl}. @xref{Function Attributes}.
13025 @strong{Warning:} this calling convention is incompatible with the one
13026 normally used on Unix, so you cannot use it if you need to call
13027 libraries compiled with the Unix compiler.
13029 Also, you must provide function prototypes for all functions that
13030 take variable numbers of arguments (including @code{printf});
13031 otherwise incorrect code will be generated for calls to those
13034 In addition, seriously incorrect code will result if you call a
13035 function with too many arguments. (Normally, extra arguments are
13036 harmlessly ignored.)
13038 @item -mregparm=@var{num}
13040 Control how many registers are used to pass integer arguments. By
13041 default, no registers are used to pass arguments, and at most 3
13042 registers can be used. You can control this behavior for a specific
13043 function by using the function attribute @samp{regparm}.
13044 @xref{Function Attributes}.
13046 @strong{Warning:} if you use this switch, and
13047 @var{num} is nonzero, then you must build all modules with the same
13048 value, including any libraries. This includes the system libraries and
13052 @opindex msseregparm
13053 Use SSE register passing conventions for float and double arguments
13054 and return values. You can control this behavior for a specific
13055 function by using the function attribute @samp{sseregparm}.
13056 @xref{Function Attributes}.
13058 @strong{Warning:} if you use this switch then you must build all
13059 modules with the same value, including any libraries. This includes
13060 the system libraries and startup modules.
13062 @item -mvect8-ret-in-mem
13063 @opindex mvect8-ret-in-mem
13064 Return 8-byte vectors in memory instead of MMX registers. This is the
13065 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
13066 Studio compilers until version 12. Later compiler versions (starting
13067 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
13068 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
13069 you need to remain compatible with existing code produced by those
13070 previous compiler versions or older versions of GCC.
13079 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
13080 is specified, the significands of results of floating-point operations are
13081 rounded to 24 bits (single precision); @option{-mpc64} rounds the
13082 significands of results of floating-point operations to 53 bits (double
13083 precision) and @option{-mpc80} rounds the significands of results of
13084 floating-point operations to 64 bits (extended double precision), which is
13085 the default. When this option is used, floating-point operations in higher
13086 precisions are not available to the programmer without setting the FPU
13087 control word explicitly.
13089 Setting the rounding of floating-point operations to less than the default
13090 80 bits can speed some programs by 2% or more. Note that some mathematical
13091 libraries assume that extended precision (80 bit) floating-point operations
13092 are enabled by default; routines in such libraries could suffer significant
13093 loss of accuracy, typically through so-called "catastrophic cancellation",
13094 when this option is used to set the precision to less than extended precision.
13096 @item -mstackrealign
13097 @opindex mstackrealign
13098 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
13099 option will generate an alternate prologue and epilogue that realigns the
13100 runtime stack if necessary. This supports mixing legacy codes that keep
13101 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
13102 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
13103 applicable to individual functions.
13105 @item -mpreferred-stack-boundary=@var{num}
13106 @opindex mpreferred-stack-boundary
13107 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
13108 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
13109 the default is 4 (16 bytes or 128 bits).
13111 @item -mincoming-stack-boundary=@var{num}
13112 @opindex mincoming-stack-boundary
13113 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
13114 boundary. If @option{-mincoming-stack-boundary} is not specified,
13115 the one specified by @option{-mpreferred-stack-boundary} will be used.
13117 On Pentium and PentiumPro, @code{double} and @code{long double} values
13118 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
13119 suffer significant run time performance penalties. On Pentium III, the
13120 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
13121 properly if it is not 16 byte aligned.
13123 To ensure proper alignment of this values on the stack, the stack boundary
13124 must be as aligned as that required by any value stored on the stack.
13125 Further, every function must be generated such that it keeps the stack
13126 aligned. Thus calling a function compiled with a higher preferred
13127 stack boundary from a function compiled with a lower preferred stack
13128 boundary will most likely misalign the stack. It is recommended that
13129 libraries that use callbacks always use the default setting.
13131 This extra alignment does consume extra stack space, and generally
13132 increases code size. Code that is sensitive to stack space usage, such
13133 as embedded systems and operating system kernels, may want to reduce the
13134 preferred alignment to @option{-mpreferred-stack-boundary=2}.
13163 @itemx -mno-fsgsbase
13199 These switches enable or disable the use of instructions in the MMX, SSE,
13200 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
13201 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
13202 @: extended instruction sets.
13203 These extensions are also available as built-in functions: see
13204 @ref{X86 Built-in Functions}, for details of the functions enabled and
13205 disabled by these switches.
13207 To have SSE/SSE2 instructions generated automatically from floating-point
13208 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
13210 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
13211 generates new AVX instructions or AVX equivalence for all SSEx instructions
13214 These options will enable GCC to use these extended instructions in
13215 generated code, even without @option{-mfpmath=sse}. Applications which
13216 perform runtime CPU detection must compile separate files for each
13217 supported architecture, using the appropriate flags. In particular,
13218 the file containing the CPU detection code should be compiled without
13223 This option instructs GCC to emit a @code{cld} instruction in the prologue
13224 of functions that use string instructions. String instructions depend on
13225 the DF flag to select between autoincrement or autodecrement mode. While the
13226 ABI specifies the DF flag to be cleared on function entry, some operating
13227 systems violate this specification by not clearing the DF flag in their
13228 exception dispatchers. The exception handler can be invoked with the DF flag
13229 set which leads to wrong direction mode, when string instructions are used.
13230 This option can be enabled by default on 32-bit x86 targets by configuring
13231 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
13232 instructions can be suppressed with the @option{-mno-cld} compiler option
13236 @opindex mvzeroupper
13237 This option instructs GCC to emit a @code{vzeroupper} instruction
13238 before a transfer of control flow out of the function to minimize
13239 AVX to SSE transition penalty as well as remove unnecessary zeroupper
13244 This option will enable GCC to use CMPXCHG16B instruction in generated code.
13245 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
13246 data types. This is useful for high resolution counters that could be updated
13247 by multiple processors (or cores). This instruction is generated as part of
13248 atomic built-in functions: see @ref{__sync Builtins} or
13249 @ref{__atomic Builtins} for details.
13253 This option will enable GCC to use SAHF instruction in generated 64-bit code.
13254 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
13255 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
13256 SAHF are load and store instructions, respectively, for certain status flags.
13257 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
13258 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
13262 This option will enable GCC to use movbe instruction to implement
13263 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
13267 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
13268 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
13269 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
13273 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
13274 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
13275 to increase precision instead of DIVSS and SQRTSS (and their vectorized
13276 variants) for single precision floating point arguments. These instructions
13277 are generated only when @option{-funsafe-math-optimizations} is enabled
13278 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
13279 Note that while the throughput of the sequence is higher than the throughput
13280 of the non-reciprocal instruction, the precision of the sequence can be
13281 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
13283 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS
13284 (or RSQRTPS) already with @option{-ffast-math} (or the above option
13285 combination), and doesn't need @option{-mrecip}.
13287 Also note that GCC emits the above sequence with additional Newton-Raphson step
13288 for vectorized single float division and vectorized @code{sqrtf(@var{x})}
13289 already with @option{-ffast-math} (or the above option combination), and
13290 doesn't need @option{-mrecip}.
13292 @item -mrecip=@var{opt}
13293 @opindex mrecip=opt
13294 This option allows to control which reciprocal estimate instructions
13295 may be used. @var{opt} is a comma separated list of options, that may
13296 be preceded by a @code{!} to invert the option:
13297 @code{all}: enable all estimate instructions,
13298 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
13299 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip},
13300 @code{div}: enable the approximation for scalar division,
13301 @code{vec-div}: enable the approximation for vectorized division,
13302 @code{sqrt}: enable the approximation for scalar square root,
13303 @code{vec-sqrt}: enable the approximation for vectorized square root.
13305 So for example, @option{-mrecip=all,!sqrt} would enable
13306 all of the reciprocal approximations, except for square root.
13308 @item -mveclibabi=@var{type}
13309 @opindex mveclibabi
13310 Specifies the ABI type to use for vectorizing intrinsics using an
13311 external library. Supported types are @code{svml} for the Intel short
13312 vector math library and @code{acml} for the AMD math core library style
13313 of interfacing. GCC will currently emit calls to @code{vmldExp2},
13314 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
13315 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
13316 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
13317 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
13318 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
13319 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
13320 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
13321 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
13322 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
13323 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
13324 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
13325 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
13326 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
13327 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
13328 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
13329 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
13330 compatible library will have to be specified at link time.
13332 @item -mabi=@var{name}
13334 Generate code for the specified calling convention. Permissible values
13335 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
13336 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
13337 ABI when targeting Windows. On all other systems, the default is the
13338 SYSV ABI. You can control this behavior for a specific function by
13339 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
13340 @xref{Function Attributes}.
13342 @item -mtls-dialect=@var{type}
13343 @opindex mtls-dialect
13344 Generate code to access thread-local storage using the @samp{gnu} or
13345 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
13346 @samp{gnu2} is more efficient, but it may add compile- and run-time
13347 requirements that cannot be satisfied on all systems.
13350 @itemx -mno-push-args
13351 @opindex mpush-args
13352 @opindex mno-push-args
13353 Use PUSH operations to store outgoing parameters. This method is shorter
13354 and usually equally fast as method using SUB/MOV operations and is enabled
13355 by default. In some cases disabling it may improve performance because of
13356 improved scheduling and reduced dependencies.
13358 @item -maccumulate-outgoing-args
13359 @opindex maccumulate-outgoing-args
13360 If enabled, the maximum amount of space required for outgoing arguments will be
13361 computed in the function prologue. This is faster on most modern CPUs
13362 because of reduced dependencies, improved scheduling and reduced stack usage
13363 when preferred stack boundary is not equal to 2. The drawback is a notable
13364 increase in code size. This switch implies @option{-mno-push-args}.
13368 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
13369 on thread-safe exception handling must compile and link all code with the
13370 @option{-mthreads} option. When compiling, @option{-mthreads} defines
13371 @option{-D_MT}; when linking, it links in a special thread helper library
13372 @option{-lmingwthrd} which cleans up per thread exception handling data.
13374 @item -mno-align-stringops
13375 @opindex mno-align-stringops
13376 Do not align destination of inlined string operations. This switch reduces
13377 code size and improves performance in case the destination is already aligned,
13378 but GCC doesn't know about it.
13380 @item -minline-all-stringops
13381 @opindex minline-all-stringops
13382 By default GCC inlines string operations only when destination is known to be
13383 aligned at least to 4 byte boundary. This enables more inlining, increase code
13384 size, but may improve performance of code that depends on fast memcpy, strlen
13385 and memset for short lengths.
13387 @item -minline-stringops-dynamically
13388 @opindex minline-stringops-dynamically
13389 For string operation of unknown size, inline runtime checks so for small
13390 blocks inline code is used, while for large blocks library call is used.
13392 @item -mstringop-strategy=@var{alg}
13393 @opindex mstringop-strategy=@var{alg}
13394 Overwrite internal decision heuristic about particular algorithm to inline
13395 string operation with. The allowed values are @code{rep_byte},
13396 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
13397 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
13398 expanding inline loop, @code{libcall} for always expanding library call.
13400 @item -momit-leaf-frame-pointer
13401 @opindex momit-leaf-frame-pointer
13402 Don't keep the frame pointer in a register for leaf functions. This
13403 avoids the instructions to save, set up and restore frame pointers and
13404 makes an extra register available in leaf functions. The option
13405 @option{-fomit-frame-pointer} removes the frame pointer for all functions
13406 which might make debugging harder.
13408 @item -mtls-direct-seg-refs
13409 @itemx -mno-tls-direct-seg-refs
13410 @opindex mtls-direct-seg-refs
13411 Controls whether TLS variables may be accessed with offsets from the
13412 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
13413 or whether the thread base pointer must be added. Whether or not this
13414 is legal depends on the operating system, and whether it maps the
13415 segment to cover the entire TLS area.
13417 For systems that use GNU libc, the default is on.
13420 @itemx -mno-sse2avx
13422 Specify that the assembler should encode SSE instructions with VEX
13423 prefix. The option @option{-mavx} turns this on by default.
13428 If profiling is active @option{-pg} put the profiling
13429 counter call before prologue.
13430 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13431 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13434 @itemx -mno-8bit-idiv
13436 On some processors, like Intel Atom, 8bit unsigned integer divide is
13437 much faster than 32bit/64bit integer divide. This option will generate a
13438 runt-time check. If both dividend and divisor are within range of 0
13439 to 255, 8bit unsigned integer divide will be used instead of
13440 32bit/64bit integer divide.
13442 @item -mavx256-split-unaligned-load
13443 @item -mavx256-split-unaligned-store
13444 @opindex avx256-split-unaligned-load
13445 @opindex avx256-split-unaligned-store
13446 Split 32-byte AVX unaligned load and store.
13450 These @samp{-m} switches are supported in addition to the above
13451 on AMD x86-64 processors in 64-bit environments.
13460 Generate code for a 32-bit or 64-bit environment.
13461 The @option{-m32} option sets int, long and pointer to 32 bits and
13462 generates code that runs on any i386 system.
13463 The @option{-m64} option sets int to 32 bits and long and pointer
13464 to 64 bits and generates code for AMD's x86-64 architecture.
13465 The @option{-mx32} option sets int, long and pointer to 32 bits and
13466 generates code for AMD's x86-64 architecture.
13467 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
13468 and @option{-mdynamic-no-pic} options.
13470 @item -mno-red-zone
13471 @opindex mno-red-zone
13472 Do not use a so called red zone for x86-64 code. The red zone is mandated
13473 by the x86-64 ABI, it is a 128-byte area beyond the location of the
13474 stack pointer that will not be modified by signal or interrupt handlers
13475 and therefore can be used for temporary data without adjusting the stack
13476 pointer. The flag @option{-mno-red-zone} disables this red zone.
13478 @item -mcmodel=small
13479 @opindex mcmodel=small
13480 Generate code for the small code model: the program and its symbols must
13481 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13482 Programs can be statically or dynamically linked. This is the default
13485 @item -mcmodel=kernel
13486 @opindex mcmodel=kernel
13487 Generate code for the kernel code model. The kernel runs in the
13488 negative 2 GB of the address space.
13489 This model has to be used for Linux kernel code.
13491 @item -mcmodel=medium
13492 @opindex mcmodel=medium
13493 Generate code for the medium model: The program is linked in the lower 2
13494 GB of the address space. Small symbols are also placed there. Symbols
13495 with sizes larger than @option{-mlarge-data-threshold} are put into
13496 large data or bss sections and can be located above 2GB. Programs can
13497 be statically or dynamically linked.
13499 @item -mcmodel=large
13500 @opindex mcmodel=large
13501 Generate code for the large model: This model makes no assumptions
13502 about addresses and sizes of sections.
13505 @node i386 and x86-64 Windows Options
13506 @subsection i386 and x86-64 Windows Options
13507 @cindex i386 and x86-64 Windows Options
13509 These additional options are available for Windows targets:
13514 This option is available for Cygwin and MinGW targets. It
13515 specifies that a console application is to be generated, by
13516 instructing the linker to set the PE header subsystem type
13517 required for console applications.
13518 This is the default behavior for Cygwin and MinGW targets.
13522 This option is available for Cygwin and MinGW targets. It
13523 specifies that a DLL - a dynamic link library - is to be
13524 generated, enabling the selection of the required runtime
13525 startup object and entry point.
13527 @item -mnop-fun-dllimport
13528 @opindex mnop-fun-dllimport
13529 This option is available for Cygwin and MinGW targets. It
13530 specifies that the dllimport attribute should be ignored.
13534 This option is available for MinGW targets. It specifies
13535 that MinGW-specific thread support is to be used.
13539 This option is available for mingw-w64 targets. It specifies
13540 that the UNICODE macro is getting pre-defined and that the
13541 unicode capable runtime startup code is chosen.
13545 This option is available for Cygwin and MinGW targets. It
13546 specifies that the typical Windows pre-defined macros are to
13547 be set in the pre-processor, but does not influence the choice
13548 of runtime library/startup code.
13552 This option is available for Cygwin and MinGW targets. It
13553 specifies that a GUI application is to be generated by
13554 instructing the linker to set the PE header subsystem type
13557 @item -fno-set-stack-executable
13558 @opindex fno-set-stack-executable
13559 This option is available for MinGW targets. It specifies that
13560 the executable flag for stack used by nested functions isn't
13561 set. This is necessary for binaries running in kernel mode of
13562 Windows, as there the user32 API, which is used to set executable
13563 privileges, isn't available.
13565 @item -mpe-aligned-commons
13566 @opindex mpe-aligned-commons
13567 This option is available for Cygwin and MinGW targets. It
13568 specifies that the GNU extension to the PE file format that
13569 permits the correct alignment of COMMON variables should be
13570 used when generating code. It will be enabled by default if
13571 GCC detects that the target assembler found during configuration
13572 supports the feature.
13575 See also under @ref{i386 and x86-64 Options} for standard options.
13577 @node IA-64 Options
13578 @subsection IA-64 Options
13579 @cindex IA-64 Options
13581 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13585 @opindex mbig-endian
13586 Generate code for a big endian target. This is the default for HP-UX@.
13588 @item -mlittle-endian
13589 @opindex mlittle-endian
13590 Generate code for a little endian target. This is the default for AIX5
13596 @opindex mno-gnu-as
13597 Generate (or don't) code for the GNU assembler. This is the default.
13598 @c Also, this is the default if the configure option @option{--with-gnu-as}
13604 @opindex mno-gnu-ld
13605 Generate (or don't) code for the GNU linker. This is the default.
13606 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13611 Generate code that does not use a global pointer register. The result
13612 is not position independent code, and violates the IA-64 ABI@.
13614 @item -mvolatile-asm-stop
13615 @itemx -mno-volatile-asm-stop
13616 @opindex mvolatile-asm-stop
13617 @opindex mno-volatile-asm-stop
13618 Generate (or don't) a stop bit immediately before and after volatile asm
13621 @item -mregister-names
13622 @itemx -mno-register-names
13623 @opindex mregister-names
13624 @opindex mno-register-names
13625 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13626 the stacked registers. This may make assembler output more readable.
13632 Disable (or enable) optimizations that use the small data section. This may
13633 be useful for working around optimizer bugs.
13635 @item -mconstant-gp
13636 @opindex mconstant-gp
13637 Generate code that uses a single constant global pointer value. This is
13638 useful when compiling kernel code.
13642 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13643 This is useful when compiling firmware code.
13645 @item -minline-float-divide-min-latency
13646 @opindex minline-float-divide-min-latency
13647 Generate code for inline divides of floating point values
13648 using the minimum latency algorithm.
13650 @item -minline-float-divide-max-throughput
13651 @opindex minline-float-divide-max-throughput
13652 Generate code for inline divides of floating point values
13653 using the maximum throughput algorithm.
13655 @item -mno-inline-float-divide
13656 @opindex mno-inline-float-divide
13657 Do not generate inline code for divides of floating point values.
13659 @item -minline-int-divide-min-latency
13660 @opindex minline-int-divide-min-latency
13661 Generate code for inline divides of integer values
13662 using the minimum latency algorithm.
13664 @item -minline-int-divide-max-throughput
13665 @opindex minline-int-divide-max-throughput
13666 Generate code for inline divides of integer values
13667 using the maximum throughput algorithm.
13669 @item -mno-inline-int-divide
13670 @opindex mno-inline-int-divide
13671 Do not generate inline code for divides of integer values.
13673 @item -minline-sqrt-min-latency
13674 @opindex minline-sqrt-min-latency
13675 Generate code for inline square roots
13676 using the minimum latency algorithm.
13678 @item -minline-sqrt-max-throughput
13679 @opindex minline-sqrt-max-throughput
13680 Generate code for inline square roots
13681 using the maximum throughput algorithm.
13683 @item -mno-inline-sqrt
13684 @opindex mno-inline-sqrt
13685 Do not generate inline code for sqrt.
13688 @itemx -mno-fused-madd
13689 @opindex mfused-madd
13690 @opindex mno-fused-madd
13691 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13692 instructions. The default is to use these instructions.
13694 @item -mno-dwarf2-asm
13695 @itemx -mdwarf2-asm
13696 @opindex mno-dwarf2-asm
13697 @opindex mdwarf2-asm
13698 Don't (or do) generate assembler code for the DWARF2 line number debugging
13699 info. This may be useful when not using the GNU assembler.
13701 @item -mearly-stop-bits
13702 @itemx -mno-early-stop-bits
13703 @opindex mearly-stop-bits
13704 @opindex mno-early-stop-bits
13705 Allow stop bits to be placed earlier than immediately preceding the
13706 instruction that triggered the stop bit. This can improve instruction
13707 scheduling, but does not always do so.
13709 @item -mfixed-range=@var{register-range}
13710 @opindex mfixed-range
13711 Generate code treating the given register range as fixed registers.
13712 A fixed register is one that the register allocator can not use. This is
13713 useful when compiling kernel code. A register range is specified as
13714 two registers separated by a dash. Multiple register ranges can be
13715 specified separated by a comma.
13717 @item -mtls-size=@var{tls-size}
13719 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13722 @item -mtune=@var{cpu-type}
13724 Tune the instruction scheduling for a particular CPU, Valid values are
13725 itanium, itanium1, merced, itanium2, and mckinley.
13731 Generate code for a 32-bit or 64-bit environment.
13732 The 32-bit environment sets int, long and pointer to 32 bits.
13733 The 64-bit environment sets int to 32 bits and long and pointer
13734 to 64 bits. These are HP-UX specific flags.
13736 @item -mno-sched-br-data-spec
13737 @itemx -msched-br-data-spec
13738 @opindex mno-sched-br-data-spec
13739 @opindex msched-br-data-spec
13740 (Dis/En)able data speculative scheduling before reload.
13741 This will result in generation of the ld.a instructions and
13742 the corresponding check instructions (ld.c / chk.a).
13743 The default is 'disable'.
13745 @item -msched-ar-data-spec
13746 @itemx -mno-sched-ar-data-spec
13747 @opindex msched-ar-data-spec
13748 @opindex mno-sched-ar-data-spec
13749 (En/Dis)able data speculative scheduling after reload.
13750 This will result in generation of the ld.a instructions and
13751 the corresponding check instructions (ld.c / chk.a).
13752 The default is 'enable'.
13754 @item -mno-sched-control-spec
13755 @itemx -msched-control-spec
13756 @opindex mno-sched-control-spec
13757 @opindex msched-control-spec
13758 (Dis/En)able control speculative scheduling. This feature is
13759 available only during region scheduling (i.e.@: before reload).
13760 This will result in generation of the ld.s instructions and
13761 the corresponding check instructions chk.s .
13762 The default is 'disable'.
13764 @item -msched-br-in-data-spec
13765 @itemx -mno-sched-br-in-data-spec
13766 @opindex msched-br-in-data-spec
13767 @opindex mno-sched-br-in-data-spec
13768 (En/Dis)able speculative scheduling of the instructions that
13769 are dependent on the data speculative loads before reload.
13770 This is effective only with @option{-msched-br-data-spec} enabled.
13771 The default is 'enable'.
13773 @item -msched-ar-in-data-spec
13774 @itemx -mno-sched-ar-in-data-spec
13775 @opindex msched-ar-in-data-spec
13776 @opindex mno-sched-ar-in-data-spec
13777 (En/Dis)able speculative scheduling of the instructions that
13778 are dependent on the data speculative loads after reload.
13779 This is effective only with @option{-msched-ar-data-spec} enabled.
13780 The default is 'enable'.
13782 @item -msched-in-control-spec
13783 @itemx -mno-sched-in-control-spec
13784 @opindex msched-in-control-spec
13785 @opindex mno-sched-in-control-spec
13786 (En/Dis)able speculative scheduling of the instructions that
13787 are dependent on the control speculative loads.
13788 This is effective only with @option{-msched-control-spec} enabled.
13789 The default is 'enable'.
13791 @item -mno-sched-prefer-non-data-spec-insns
13792 @itemx -msched-prefer-non-data-spec-insns
13793 @opindex mno-sched-prefer-non-data-spec-insns
13794 @opindex msched-prefer-non-data-spec-insns
13795 If enabled, data speculative instructions will be chosen for schedule
13796 only if there are no other choices at the moment. This will make
13797 the use of the data speculation much more conservative.
13798 The default is 'disable'.
13800 @item -mno-sched-prefer-non-control-spec-insns
13801 @itemx -msched-prefer-non-control-spec-insns
13802 @opindex mno-sched-prefer-non-control-spec-insns
13803 @opindex msched-prefer-non-control-spec-insns
13804 If enabled, control speculative instructions will be chosen for schedule
13805 only if there are no other choices at the moment. This will make
13806 the use of the control speculation much more conservative.
13807 The default is 'disable'.
13809 @item -mno-sched-count-spec-in-critical-path
13810 @itemx -msched-count-spec-in-critical-path
13811 @opindex mno-sched-count-spec-in-critical-path
13812 @opindex msched-count-spec-in-critical-path
13813 If enabled, speculative dependencies will be considered during
13814 computation of the instructions priorities. This will make the use of the
13815 speculation a bit more conservative.
13816 The default is 'disable'.
13818 @item -msched-spec-ldc
13819 @opindex msched-spec-ldc
13820 Use a simple data speculation check. This option is on by default.
13822 @item -msched-control-spec-ldc
13823 @opindex msched-spec-ldc
13824 Use a simple check for control speculation. This option is on by default.
13826 @item -msched-stop-bits-after-every-cycle
13827 @opindex msched-stop-bits-after-every-cycle
13828 Place a stop bit after every cycle when scheduling. This option is on
13831 @item -msched-fp-mem-deps-zero-cost
13832 @opindex msched-fp-mem-deps-zero-cost
13833 Assume that floating-point stores and loads are not likely to cause a conflict
13834 when placed into the same instruction group. This option is disabled by
13837 @item -msel-sched-dont-check-control-spec
13838 @opindex msel-sched-dont-check-control-spec
13839 Generate checks for control speculation in selective scheduling.
13840 This flag is disabled by default.
13842 @item -msched-max-memory-insns=@var{max-insns}
13843 @opindex msched-max-memory-insns
13844 Limit on the number of memory insns per instruction group, giving lower
13845 priority to subsequent memory insns attempting to schedule in the same
13846 instruction group. Frequently useful to prevent cache bank conflicts.
13847 The default value is 1.
13849 @item -msched-max-memory-insns-hard-limit
13850 @opindex msched-max-memory-insns-hard-limit
13851 Disallow more than `msched-max-memory-insns' in instruction group.
13852 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13853 when limit is reached but may still schedule memory operations.
13857 @node IA-64/VMS Options
13858 @subsection IA-64/VMS Options
13860 These @samp{-m} options are defined for the IA-64/VMS implementations:
13863 @item -mvms-return-codes
13864 @opindex mvms-return-codes
13865 Return VMS condition codes from main. The default is to return POSIX
13866 style condition (e.g.@ error) codes.
13868 @item -mdebug-main=@var{prefix}
13869 @opindex mdebug-main=@var{prefix}
13870 Flag the first routine whose name starts with @var{prefix} as the main
13871 routine for the debugger.
13875 Default to 64bit memory allocation routines.
13879 @subsection LM32 Options
13880 @cindex LM32 options
13882 These @option{-m} options are defined for the Lattice Mico32 architecture:
13885 @item -mbarrel-shift-enabled
13886 @opindex mbarrel-shift-enabled
13887 Enable barrel-shift instructions.
13889 @item -mdivide-enabled
13890 @opindex mdivide-enabled
13891 Enable divide and modulus instructions.
13893 @item -mmultiply-enabled
13894 @opindex multiply-enabled
13895 Enable multiply instructions.
13897 @item -msign-extend-enabled
13898 @opindex msign-extend-enabled
13899 Enable sign extend instructions.
13901 @item -muser-enabled
13902 @opindex muser-enabled
13903 Enable user-defined instructions.
13908 @subsection M32C Options
13909 @cindex M32C options
13912 @item -mcpu=@var{name}
13914 Select the CPU for which code is generated. @var{name} may be one of
13915 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13916 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13917 the M32C/80 series.
13921 Specifies that the program will be run on the simulator. This causes
13922 an alternate runtime library to be linked in which supports, for
13923 example, file I/O@. You must not use this option when generating
13924 programs that will run on real hardware; you must provide your own
13925 runtime library for whatever I/O functions are needed.
13927 @item -memregs=@var{number}
13929 Specifies the number of memory-based pseudo-registers GCC will use
13930 during code generation. These pseudo-registers will be used like real
13931 registers, so there is a tradeoff between GCC's ability to fit the
13932 code into available registers, and the performance penalty of using
13933 memory instead of registers. Note that all modules in a program must
13934 be compiled with the same value for this option. Because of that, you
13935 must not use this option with the default runtime libraries gcc
13940 @node M32R/D Options
13941 @subsection M32R/D Options
13942 @cindex M32R/D options
13944 These @option{-m} options are defined for Renesas M32R/D architectures:
13949 Generate code for the M32R/2@.
13953 Generate code for the M32R/X@.
13957 Generate code for the M32R@. This is the default.
13959 @item -mmodel=small
13960 @opindex mmodel=small
13961 Assume all objects live in the lower 16MB of memory (so that their addresses
13962 can be loaded with the @code{ld24} instruction), and assume all subroutines
13963 are reachable with the @code{bl} instruction.
13964 This is the default.
13966 The addressability of a particular object can be set with the
13967 @code{model} attribute.
13969 @item -mmodel=medium
13970 @opindex mmodel=medium
13971 Assume objects may be anywhere in the 32-bit address space (the compiler
13972 will generate @code{seth/add3} instructions to load their addresses), and
13973 assume all subroutines are reachable with the @code{bl} instruction.
13975 @item -mmodel=large
13976 @opindex mmodel=large
13977 Assume objects may be anywhere in the 32-bit address space (the compiler
13978 will generate @code{seth/add3} instructions to load their addresses), and
13979 assume subroutines may not be reachable with the @code{bl} instruction
13980 (the compiler will generate the much slower @code{seth/add3/jl}
13981 instruction sequence).
13984 @opindex msdata=none
13985 Disable use of the small data area. Variables will be put into
13986 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13987 @code{section} attribute has been specified).
13988 This is the default.
13990 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13991 Objects may be explicitly put in the small data area with the
13992 @code{section} attribute using one of these sections.
13994 @item -msdata=sdata
13995 @opindex msdata=sdata
13996 Put small global and static data in the small data area, but do not
13997 generate special code to reference them.
14000 @opindex msdata=use
14001 Put small global and static data in the small data area, and generate
14002 special instructions to reference them.
14006 @cindex smaller data references
14007 Put global and static objects less than or equal to @var{num} bytes
14008 into the small data or bss sections instead of the normal data or bss
14009 sections. The default value of @var{num} is 8.
14010 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
14011 for this option to have any effect.
14013 All modules should be compiled with the same @option{-G @var{num}} value.
14014 Compiling with different values of @var{num} may or may not work; if it
14015 doesn't the linker will give an error message---incorrect code will not be
14020 Makes the M32R specific code in the compiler display some statistics
14021 that might help in debugging programs.
14023 @item -malign-loops
14024 @opindex malign-loops
14025 Align all loops to a 32-byte boundary.
14027 @item -mno-align-loops
14028 @opindex mno-align-loops
14029 Do not enforce a 32-byte alignment for loops. This is the default.
14031 @item -missue-rate=@var{number}
14032 @opindex missue-rate=@var{number}
14033 Issue @var{number} instructions per cycle. @var{number} can only be 1
14036 @item -mbranch-cost=@var{number}
14037 @opindex mbranch-cost=@var{number}
14038 @var{number} can only be 1 or 2. If it is 1 then branches will be
14039 preferred over conditional code, if it is 2, then the opposite will
14042 @item -mflush-trap=@var{number}
14043 @opindex mflush-trap=@var{number}
14044 Specifies the trap number to use to flush the cache. The default is
14045 12. Valid numbers are between 0 and 15 inclusive.
14047 @item -mno-flush-trap
14048 @opindex mno-flush-trap
14049 Specifies that the cache cannot be flushed by using a trap.
14051 @item -mflush-func=@var{name}
14052 @opindex mflush-func=@var{name}
14053 Specifies the name of the operating system function to call to flush
14054 the cache. The default is @emph{_flush_cache}, but a function call
14055 will only be used if a trap is not available.
14057 @item -mno-flush-func
14058 @opindex mno-flush-func
14059 Indicates that there is no OS function for flushing the cache.
14063 @node M680x0 Options
14064 @subsection M680x0 Options
14065 @cindex M680x0 options
14067 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
14068 The default settings depend on which architecture was selected when
14069 the compiler was configured; the defaults for the most common choices
14073 @item -march=@var{arch}
14075 Generate code for a specific M680x0 or ColdFire instruction set
14076 architecture. Permissible values of @var{arch} for M680x0
14077 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
14078 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
14079 architectures are selected according to Freescale's ISA classification
14080 and the permissible values are: @samp{isaa}, @samp{isaaplus},
14081 @samp{isab} and @samp{isac}.
14083 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
14084 code for a ColdFire target. The @var{arch} in this macro is one of the
14085 @option{-march} arguments given above.
14087 When used together, @option{-march} and @option{-mtune} select code
14088 that runs on a family of similar processors but that is optimized
14089 for a particular microarchitecture.
14091 @item -mcpu=@var{cpu}
14093 Generate code for a specific M680x0 or ColdFire processor.
14094 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
14095 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
14096 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
14097 below, which also classifies the CPUs into families:
14099 @multitable @columnfractions 0.20 0.80
14100 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
14101 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
14102 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
14103 @item @samp{5206e} @tab @samp{5206e}
14104 @item @samp{5208} @tab @samp{5207} @samp{5208}
14105 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
14106 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
14107 @item @samp{5216} @tab @samp{5214} @samp{5216}
14108 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
14109 @item @samp{5225} @tab @samp{5224} @samp{5225}
14110 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
14111 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
14112 @item @samp{5249} @tab @samp{5249}
14113 @item @samp{5250} @tab @samp{5250}
14114 @item @samp{5271} @tab @samp{5270} @samp{5271}
14115 @item @samp{5272} @tab @samp{5272}
14116 @item @samp{5275} @tab @samp{5274} @samp{5275}
14117 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
14118 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
14119 @item @samp{5307} @tab @samp{5307}
14120 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
14121 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
14122 @item @samp{5407} @tab @samp{5407}
14123 @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}
14126 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
14127 @var{arch} is compatible with @var{cpu}. Other combinations of
14128 @option{-mcpu} and @option{-march} are rejected.
14130 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
14131 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
14132 where the value of @var{family} is given by the table above.
14134 @item -mtune=@var{tune}
14136 Tune the code for a particular microarchitecture, within the
14137 constraints set by @option{-march} and @option{-mcpu}.
14138 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
14139 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
14140 and @samp{cpu32}. The ColdFire microarchitectures
14141 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
14143 You can also use @option{-mtune=68020-40} for code that needs
14144 to run relatively well on 68020, 68030 and 68040 targets.
14145 @option{-mtune=68020-60} is similar but includes 68060 targets
14146 as well. These two options select the same tuning decisions as
14147 @option{-m68020-40} and @option{-m68020-60} respectively.
14149 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
14150 when tuning for 680x0 architecture @var{arch}. It also defines
14151 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
14152 option is used. If gcc is tuning for a range of architectures,
14153 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
14154 it defines the macros for every architecture in the range.
14156 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
14157 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
14158 of the arguments given above.
14164 Generate output for a 68000. This is the default
14165 when the compiler is configured for 68000-based systems.
14166 It is equivalent to @option{-march=68000}.
14168 Use this option for microcontrollers with a 68000 or EC000 core,
14169 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
14173 Generate output for a 68010. This is the default
14174 when the compiler is configured for 68010-based systems.
14175 It is equivalent to @option{-march=68010}.
14181 Generate output for a 68020. This is the default
14182 when the compiler is configured for 68020-based systems.
14183 It is equivalent to @option{-march=68020}.
14187 Generate output for a 68030. This is the default when the compiler is
14188 configured for 68030-based systems. It is equivalent to
14189 @option{-march=68030}.
14193 Generate output for a 68040. This is the default when the compiler is
14194 configured for 68040-based systems. It is equivalent to
14195 @option{-march=68040}.
14197 This option inhibits the use of 68881/68882 instructions that have to be
14198 emulated by software on the 68040. Use this option if your 68040 does not
14199 have code to emulate those instructions.
14203 Generate output for a 68060. This is the default when the compiler is
14204 configured for 68060-based systems. It is equivalent to
14205 @option{-march=68060}.
14207 This option inhibits the use of 68020 and 68881/68882 instructions that
14208 have to be emulated by software on the 68060. Use this option if your 68060
14209 does not have code to emulate those instructions.
14213 Generate output for a CPU32. This is the default
14214 when the compiler is configured for CPU32-based systems.
14215 It is equivalent to @option{-march=cpu32}.
14217 Use this option for microcontrollers with a
14218 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
14219 68336, 68340, 68341, 68349 and 68360.
14223 Generate output for a 520X ColdFire CPU@. This is the default
14224 when the compiler is configured for 520X-based systems.
14225 It is equivalent to @option{-mcpu=5206}, and is now deprecated
14226 in favor of that option.
14228 Use this option for microcontroller with a 5200 core, including
14229 the MCF5202, MCF5203, MCF5204 and MCF5206.
14233 Generate output for a 5206e ColdFire CPU@. The option is now
14234 deprecated in favor of the equivalent @option{-mcpu=5206e}.
14238 Generate output for a member of the ColdFire 528X family.
14239 The option is now deprecated in favor of the equivalent
14240 @option{-mcpu=528x}.
14244 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
14245 in favor of the equivalent @option{-mcpu=5307}.
14249 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
14250 in favor of the equivalent @option{-mcpu=5407}.
14254 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
14255 This includes use of hardware floating point instructions.
14256 The option is equivalent to @option{-mcpu=547x}, and is now
14257 deprecated in favor of that option.
14261 Generate output for a 68040, without using any of the new instructions.
14262 This results in code which can run relatively efficiently on either a
14263 68020/68881 or a 68030 or a 68040. The generated code does use the
14264 68881 instructions that are emulated on the 68040.
14266 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
14270 Generate output for a 68060, without using any of the new instructions.
14271 This results in code which can run relatively efficiently on either a
14272 68020/68881 or a 68030 or a 68040. The generated code does use the
14273 68881 instructions that are emulated on the 68060.
14275 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
14279 @opindex mhard-float
14281 Generate floating-point instructions. This is the default for 68020
14282 and above, and for ColdFire devices that have an FPU@. It defines the
14283 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
14284 on ColdFire targets.
14287 @opindex msoft-float
14288 Do not generate floating-point instructions; use library calls instead.
14289 This is the default for 68000, 68010, and 68832 targets. It is also
14290 the default for ColdFire devices that have no FPU.
14296 Generate (do not generate) ColdFire hardware divide and remainder
14297 instructions. If @option{-march} is used without @option{-mcpu},
14298 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
14299 architectures. Otherwise, the default is taken from the target CPU
14300 (either the default CPU, or the one specified by @option{-mcpu}). For
14301 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
14302 @option{-mcpu=5206e}.
14304 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
14308 Consider type @code{int} to be 16 bits wide, like @code{short int}.
14309 Additionally, parameters passed on the stack are also aligned to a
14310 16-bit boundary even on targets whose API mandates promotion to 32-bit.
14314 Do not consider type @code{int} to be 16 bits wide. This is the default.
14317 @itemx -mno-bitfield
14318 @opindex mnobitfield
14319 @opindex mno-bitfield
14320 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
14321 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
14325 Do use the bit-field instructions. The @option{-m68020} option implies
14326 @option{-mbitfield}. This is the default if you use a configuration
14327 designed for a 68020.
14331 Use a different function-calling convention, in which functions
14332 that take a fixed number of arguments return with the @code{rtd}
14333 instruction, which pops their arguments while returning. This
14334 saves one instruction in the caller since there is no need to pop
14335 the arguments there.
14337 This calling convention is incompatible with the one normally
14338 used on Unix, so you cannot use it if you need to call libraries
14339 compiled with the Unix compiler.
14341 Also, you must provide function prototypes for all functions that
14342 take variable numbers of arguments (including @code{printf});
14343 otherwise incorrect code will be generated for calls to those
14346 In addition, seriously incorrect code will result if you call a
14347 function with too many arguments. (Normally, extra arguments are
14348 harmlessly ignored.)
14350 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
14351 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
14355 Do not use the calling conventions selected by @option{-mrtd}.
14356 This is the default.
14359 @itemx -mno-align-int
14360 @opindex malign-int
14361 @opindex mno-align-int
14362 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
14363 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
14364 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
14365 Aligning variables on 32-bit boundaries produces code that runs somewhat
14366 faster on processors with 32-bit busses at the expense of more memory.
14368 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
14369 align structures containing the above types differently than
14370 most published application binary interface specifications for the m68k.
14374 Use the pc-relative addressing mode of the 68000 directly, instead of
14375 using a global offset table. At present, this option implies @option{-fpic},
14376 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
14377 not presently supported with @option{-mpcrel}, though this could be supported for
14378 68020 and higher processors.
14380 @item -mno-strict-align
14381 @itemx -mstrict-align
14382 @opindex mno-strict-align
14383 @opindex mstrict-align
14384 Do not (do) assume that unaligned memory references will be handled by
14388 Generate code that allows the data segment to be located in a different
14389 area of memory from the text segment. This allows for execute in place in
14390 an environment without virtual memory management. This option implies
14393 @item -mno-sep-data
14394 Generate code that assumes that the data segment follows the text segment.
14395 This is the default.
14397 @item -mid-shared-library
14398 Generate code that supports shared libraries via the library ID method.
14399 This allows for execute in place and shared libraries in an environment
14400 without virtual memory management. This option implies @option{-fPIC}.
14402 @item -mno-id-shared-library
14403 Generate code that doesn't assume ID based shared libraries are being used.
14404 This is the default.
14406 @item -mshared-library-id=n
14407 Specified the identification number of the ID based shared library being
14408 compiled. Specifying a value of 0 will generate more compact code, specifying
14409 other values will force the allocation of that number to the current
14410 library but is no more space or time efficient than omitting this option.
14416 When generating position-independent code for ColdFire, generate code
14417 that works if the GOT has more than 8192 entries. This code is
14418 larger and slower than code generated without this option. On M680x0
14419 processors, this option is not needed; @option{-fPIC} suffices.
14421 GCC normally uses a single instruction to load values from the GOT@.
14422 While this is relatively efficient, it only works if the GOT
14423 is smaller than about 64k. Anything larger causes the linker
14424 to report an error such as:
14426 @cindex relocation truncated to fit (ColdFire)
14428 relocation truncated to fit: R_68K_GOT16O foobar
14431 If this happens, you should recompile your code with @option{-mxgot}.
14432 It should then work with very large GOTs. However, code generated with
14433 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14434 the value of a global symbol.
14436 Note that some linkers, including newer versions of the GNU linker,
14437 can create multiple GOTs and sort GOT entries. If you have such a linker,
14438 you should only need to use @option{-mxgot} when compiling a single
14439 object file that accesses more than 8192 GOT entries. Very few do.
14441 These options have no effect unless GCC is generating
14442 position-independent code.
14446 @node MCore Options
14447 @subsection MCore Options
14448 @cindex MCore options
14450 These are the @samp{-m} options defined for the Motorola M*Core
14456 @itemx -mno-hardlit
14458 @opindex mno-hardlit
14459 Inline constants into the code stream if it can be done in two
14460 instructions or less.
14466 Use the divide instruction. (Enabled by default).
14468 @item -mrelax-immediate
14469 @itemx -mno-relax-immediate
14470 @opindex mrelax-immediate
14471 @opindex mno-relax-immediate
14472 Allow arbitrary sized immediates in bit operations.
14474 @item -mwide-bitfields
14475 @itemx -mno-wide-bitfields
14476 @opindex mwide-bitfields
14477 @opindex mno-wide-bitfields
14478 Always treat bit-fields as int-sized.
14480 @item -m4byte-functions
14481 @itemx -mno-4byte-functions
14482 @opindex m4byte-functions
14483 @opindex mno-4byte-functions
14484 Force all functions to be aligned to a four byte boundary.
14486 @item -mcallgraph-data
14487 @itemx -mno-callgraph-data
14488 @opindex mcallgraph-data
14489 @opindex mno-callgraph-data
14490 Emit callgraph information.
14493 @itemx -mno-slow-bytes
14494 @opindex mslow-bytes
14495 @opindex mno-slow-bytes
14496 Prefer word access when reading byte quantities.
14498 @item -mlittle-endian
14499 @itemx -mbig-endian
14500 @opindex mlittle-endian
14501 @opindex mbig-endian
14502 Generate code for a little endian target.
14508 Generate code for the 210 processor.
14512 Assume that run-time support has been provided and so omit the
14513 simulator library (@file{libsim.a)} from the linker command line.
14515 @item -mstack-increment=@var{size}
14516 @opindex mstack-increment
14517 Set the maximum amount for a single stack increment operation. Large
14518 values can increase the speed of programs which contain functions
14519 that need a large amount of stack space, but they can also trigger a
14520 segmentation fault if the stack is extended too much. The default
14526 @subsection MeP Options
14527 @cindex MeP options
14533 Enables the @code{abs} instruction, which is the absolute difference
14534 between two registers.
14538 Enables all the optional instructions - average, multiply, divide, bit
14539 operations, leading zero, absolute difference, min/max, clip, and
14545 Enables the @code{ave} instruction, which computes the average of two
14548 @item -mbased=@var{n}
14550 Variables of size @var{n} bytes or smaller will be placed in the
14551 @code{.based} section by default. Based variables use the @code{$tp}
14552 register as a base register, and there is a 128 byte limit to the
14553 @code{.based} section.
14557 Enables the bit operation instructions - bit test (@code{btstm}), set
14558 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14559 test-and-set (@code{tas}).
14561 @item -mc=@var{name}
14563 Selects which section constant data will be placed in. @var{name} may
14564 be @code{tiny}, @code{near}, or @code{far}.
14568 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14569 useful unless you also provide @code{-mminmax}.
14571 @item -mconfig=@var{name}
14573 Selects one of the build-in core configurations. Each MeP chip has
14574 one or more modules in it; each module has a core CPU and a variety of
14575 coprocessors, optional instructions, and peripherals. The
14576 @code{MeP-Integrator} tool, not part of GCC, provides these
14577 configurations through this option; using this option is the same as
14578 using all the corresponding command line options. The default
14579 configuration is @code{default}.
14583 Enables the coprocessor instructions. By default, this is a 32-bit
14584 coprocessor. Note that the coprocessor is normally enabled via the
14585 @code{-mconfig=} option.
14589 Enables the 32-bit coprocessor's instructions.
14593 Enables the 64-bit coprocessor's instructions.
14597 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14601 Causes constant variables to be placed in the @code{.near} section.
14605 Enables the @code{div} and @code{divu} instructions.
14609 Generate big-endian code.
14613 Generate little-endian code.
14615 @item -mio-volatile
14616 @opindex mio-volatile
14617 Tells the compiler that any variable marked with the @code{io}
14618 attribute is to be considered volatile.
14622 Causes variables to be assigned to the @code{.far} section by default.
14626 Enables the @code{leadz} (leading zero) instruction.
14630 Causes variables to be assigned to the @code{.near} section by default.
14634 Enables the @code{min} and @code{max} instructions.
14638 Enables the multiplication and multiply-accumulate instructions.
14642 Disables all the optional instructions enabled by @code{-mall-opts}.
14646 Enables the @code{repeat} and @code{erepeat} instructions, used for
14647 low-overhead looping.
14651 Causes all variables to default to the @code{.tiny} section. Note
14652 that there is a 65536 byte limit to this section. Accesses to these
14653 variables use the @code{%gp} base register.
14657 Enables the saturation instructions. Note that the compiler does not
14658 currently generate these itself, but this option is included for
14659 compatibility with other tools, like @code{as}.
14663 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14667 Link the simulator runtime libraries.
14671 Link the simulator runtime libraries, excluding built-in support
14672 for reset and exception vectors and tables.
14676 Causes all functions to default to the @code{.far} section. Without
14677 this option, functions default to the @code{.near} section.
14679 @item -mtiny=@var{n}
14681 Variables that are @var{n} bytes or smaller will be allocated to the
14682 @code{.tiny} section. These variables use the @code{$gp} base
14683 register. The default for this option is 4, but note that there's a
14684 65536 byte limit to the @code{.tiny} section.
14688 @node MicroBlaze Options
14689 @subsection MicroBlaze Options
14690 @cindex MicroBlaze Options
14695 @opindex msoft-float
14696 Use software emulation for floating point (default).
14699 @opindex mhard-float
14700 Use hardware floating point instructions.
14704 Do not optimize block moves, use @code{memcpy}.
14706 @item -mno-clearbss
14707 @opindex mno-clearbss
14708 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14710 @item -mcpu=@var{cpu-type}
14712 Use features of and schedule code for given CPU.
14713 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14714 where @var{X} is a major version, @var{YY} is the minor version, and
14715 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14716 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14718 @item -mxl-soft-mul
14719 @opindex mxl-soft-mul
14720 Use software multiply emulation (default).
14722 @item -mxl-soft-div
14723 @opindex mxl-soft-div
14724 Use software emulation for divides (default).
14726 @item -mxl-barrel-shift
14727 @opindex mxl-barrel-shift
14728 Use the hardware barrel shifter.
14730 @item -mxl-pattern-compare
14731 @opindex mxl-pattern-compare
14732 Use pattern compare instructions.
14734 @item -msmall-divides
14735 @opindex msmall-divides
14736 Use table lookup optimization for small signed integer divisions.
14738 @item -mxl-stack-check
14739 @opindex mxl-stack-check
14740 This option is deprecated. Use -fstack-check instead.
14743 @opindex mxl-gp-opt
14744 Use GP relative sdata/sbss sections.
14746 @item -mxl-multiply-high
14747 @opindex mxl-multiply-high
14748 Use multiply high instructions for high part of 32x32 multiply.
14750 @item -mxl-float-convert
14751 @opindex mxl-float-convert
14752 Use hardware floating point conversion instructions.
14754 @item -mxl-float-sqrt
14755 @opindex mxl-float-sqrt
14756 Use hardware floating point square root instruction.
14758 @item -mxl-mode-@var{app-model}
14759 Select application model @var{app-model}. Valid models are
14762 normal executable (default), uses startup code @file{crt0.o}.
14765 for use with Xilinx Microprocessor Debugger (XMD) based
14766 software intrusive debug agent called xmdstub. This uses startup file
14767 @file{crt1.o} and sets the start address of the program to be 0x800.
14770 for applications that are loaded using a bootloader.
14771 This model uses startup file @file{crt2.o} which does not contain a processor
14772 reset vector handler. This is suitable for transferring control on a
14773 processor reset to the bootloader rather than the application.
14776 for applications that do not require any of the
14777 MicroBlaze vectors. This option may be useful for applications running
14778 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14781 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14782 @option{-mxl-mode-@var{app-model}}.
14787 @subsection MIPS Options
14788 @cindex MIPS options
14794 Generate big-endian code.
14798 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14801 @item -march=@var{arch}
14803 Generate code that will run on @var{arch}, which can be the name of a
14804 generic MIPS ISA, or the name of a particular processor.
14806 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14807 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14808 The processor names are:
14809 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14810 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14811 @samp{5kc}, @samp{5kf},
14813 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14814 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14815 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14816 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14817 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14818 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14820 @samp{octeon}, @samp{octeon+}, @samp{octeon2},
14822 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14823 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14824 @samp{rm7000}, @samp{rm9000},
14825 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14828 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14829 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14831 The special value @samp{from-abi} selects the
14832 most compatible architecture for the selected ABI (that is,
14833 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14835 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
14836 which selects the best architecture option for the host processor.
14837 @option{-march=native} has no effect if GCC does not recognize
14840 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14841 (for example, @samp{-march=r2k}). Prefixes are optional, and
14842 @samp{vr} may be written @samp{r}.
14844 Names of the form @samp{@var{n}f2_1} refer to processors with
14845 FPUs clocked at half the rate of the core, names of the form
14846 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14847 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14848 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14849 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14850 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14851 accepted as synonyms for @samp{@var{n}f1_1}.
14853 GCC defines two macros based on the value of this option. The first
14854 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14855 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14856 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14857 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14858 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14860 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14861 above. In other words, it will have the full prefix and will not
14862 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14863 the macro names the resolved architecture (either @samp{"mips1"} or
14864 @samp{"mips3"}). It names the default architecture when no
14865 @option{-march} option is given.
14867 @item -mtune=@var{arch}
14869 Optimize for @var{arch}. Among other things, this option controls
14870 the way instructions are scheduled, and the perceived cost of arithmetic
14871 operations. The list of @var{arch} values is the same as for
14874 When this option is not used, GCC will optimize for the processor
14875 specified by @option{-march}. By using @option{-march} and
14876 @option{-mtune} together, it is possible to generate code that will
14877 run on a family of processors, but optimize the code for one
14878 particular member of that family.
14880 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14881 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14882 @samp{-march} ones described above.
14886 Equivalent to @samp{-march=mips1}.
14890 Equivalent to @samp{-march=mips2}.
14894 Equivalent to @samp{-march=mips3}.
14898 Equivalent to @samp{-march=mips4}.
14902 Equivalent to @samp{-march=mips32}.
14906 Equivalent to @samp{-march=mips32r2}.
14910 Equivalent to @samp{-march=mips64}.
14914 Equivalent to @samp{-march=mips64r2}.
14919 @opindex mno-mips16
14920 Generate (do not generate) MIPS16 code. If GCC is targetting a
14921 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14923 MIPS16 code generation can also be controlled on a per-function basis
14924 by means of @code{mips16} and @code{nomips16} attributes.
14925 @xref{Function Attributes}, for more information.
14927 @item -mflip-mips16
14928 @opindex mflip-mips16
14929 Generate MIPS16 code on alternating functions. This option is provided
14930 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14931 not intended for ordinary use in compiling user code.
14933 @item -minterlink-mips16
14934 @itemx -mno-interlink-mips16
14935 @opindex minterlink-mips16
14936 @opindex mno-interlink-mips16
14937 Require (do not require) that non-MIPS16 code be link-compatible with
14940 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14941 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14942 therefore disables direct jumps unless GCC knows that the target of the
14943 jump is not MIPS16.
14955 Generate code for the given ABI@.
14957 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14958 generates 64-bit code when you select a 64-bit architecture, but you
14959 can use @option{-mgp32} to get 32-bit code instead.
14961 For information about the O64 ABI, see
14962 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14964 GCC supports a variant of the o32 ABI in which floating-point registers
14965 are 64 rather than 32 bits wide. You can select this combination with
14966 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14967 and @samp{mfhc1} instructions and is therefore only supported for
14968 MIPS32R2 processors.
14970 The register assignments for arguments and return values remain the
14971 same, but each scalar value is passed in a single 64-bit register
14972 rather than a pair of 32-bit registers. For example, scalar
14973 floating-point values are returned in @samp{$f0} only, not a
14974 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14975 remains the same, but all 64 bits are saved.
14978 @itemx -mno-abicalls
14980 @opindex mno-abicalls
14981 Generate (do not generate) code that is suitable for SVR4-style
14982 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14987 Generate (do not generate) code that is fully position-independent,
14988 and that can therefore be linked into shared libraries. This option
14989 only affects @option{-mabicalls}.
14991 All @option{-mabicalls} code has traditionally been position-independent,
14992 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14993 as an extension, the GNU toolchain allows executables to use absolute
14994 accesses for locally-binding symbols. It can also use shorter GP
14995 initialization sequences and generate direct calls to locally-defined
14996 functions. This mode is selected by @option{-mno-shared}.
14998 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14999 objects that can only be linked by the GNU linker. However, the option
15000 does not affect the ABI of the final executable; it only affects the ABI
15001 of relocatable objects. Using @option{-mno-shared} will generally make
15002 executables both smaller and quicker.
15004 @option{-mshared} is the default.
15010 Assume (do not assume) that the static and dynamic linkers
15011 support PLTs and copy relocations. This option only affects
15012 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
15013 has no effect without @samp{-msym32}.
15015 You can make @option{-mplt} the default by configuring
15016 GCC with @option{--with-mips-plt}. The default is
15017 @option{-mno-plt} otherwise.
15023 Lift (do not lift) the usual restrictions on the size of the global
15026 GCC normally uses a single instruction to load values from the GOT@.
15027 While this is relatively efficient, it will only work if the GOT
15028 is smaller than about 64k. Anything larger will cause the linker
15029 to report an error such as:
15031 @cindex relocation truncated to fit (MIPS)
15033 relocation truncated to fit: R_MIPS_GOT16 foobar
15036 If this happens, you should recompile your code with @option{-mxgot}.
15037 It should then work with very large GOTs, although it will also be
15038 less efficient, since it will take three instructions to fetch the
15039 value of a global symbol.
15041 Note that some linkers can create multiple GOTs. If you have such a
15042 linker, you should only need to use @option{-mxgot} when a single object
15043 file accesses more than 64k's worth of GOT entries. Very few do.
15045 These options have no effect unless GCC is generating position
15050 Assume that general-purpose registers are 32 bits wide.
15054 Assume that general-purpose registers are 64 bits wide.
15058 Assume that floating-point registers are 32 bits wide.
15062 Assume that floating-point registers are 64 bits wide.
15065 @opindex mhard-float
15066 Use floating-point coprocessor instructions.
15069 @opindex msoft-float
15070 Do not use floating-point coprocessor instructions. Implement
15071 floating-point calculations using library calls instead.
15073 @item -msingle-float
15074 @opindex msingle-float
15075 Assume that the floating-point coprocessor only supports single-precision
15078 @item -mdouble-float
15079 @opindex mdouble-float
15080 Assume that the floating-point coprocessor supports double-precision
15081 operations. This is the default.
15087 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
15088 implement atomic memory built-in functions. When neither option is
15089 specified, GCC will use the instructions if the target architecture
15092 @option{-mllsc} is useful if the runtime environment can emulate the
15093 instructions and @option{-mno-llsc} can be useful when compiling for
15094 nonstandard ISAs. You can make either option the default by
15095 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
15096 respectively. @option{--with-llsc} is the default for some
15097 configurations; see the installation documentation for details.
15103 Use (do not use) revision 1 of the MIPS DSP ASE@.
15104 @xref{MIPS DSP Built-in Functions}. This option defines the
15105 preprocessor macro @samp{__mips_dsp}. It also defines
15106 @samp{__mips_dsp_rev} to 1.
15112 Use (do not use) revision 2 of the MIPS DSP ASE@.
15113 @xref{MIPS DSP Built-in Functions}. This option defines the
15114 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
15115 It also defines @samp{__mips_dsp_rev} to 2.
15118 @itemx -mno-smartmips
15119 @opindex msmartmips
15120 @opindex mno-smartmips
15121 Use (do not use) the MIPS SmartMIPS ASE.
15123 @item -mpaired-single
15124 @itemx -mno-paired-single
15125 @opindex mpaired-single
15126 @opindex mno-paired-single
15127 Use (do not use) paired-single floating-point instructions.
15128 @xref{MIPS Paired-Single Support}. This option requires
15129 hardware floating-point support to be enabled.
15135 Use (do not use) MIPS Digital Media Extension instructions.
15136 This option can only be used when generating 64-bit code and requires
15137 hardware floating-point support to be enabled.
15142 @opindex mno-mips3d
15143 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
15144 The option @option{-mips3d} implies @option{-mpaired-single}.
15150 Use (do not use) MT Multithreading instructions.
15154 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
15155 an explanation of the default and the way that the pointer size is
15160 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
15162 The default size of @code{int}s, @code{long}s and pointers depends on
15163 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
15164 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
15165 32-bit @code{long}s. Pointers are the same size as @code{long}s,
15166 or the same size as integer registers, whichever is smaller.
15172 Assume (do not assume) that all symbols have 32-bit values, regardless
15173 of the selected ABI@. This option is useful in combination with
15174 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
15175 to generate shorter and faster references to symbolic addresses.
15179 Put definitions of externally-visible data in a small data section
15180 if that data is no bigger than @var{num} bytes. GCC can then access
15181 the data more efficiently; see @option{-mgpopt} for details.
15183 The default @option{-G} option depends on the configuration.
15185 @item -mlocal-sdata
15186 @itemx -mno-local-sdata
15187 @opindex mlocal-sdata
15188 @opindex mno-local-sdata
15189 Extend (do not extend) the @option{-G} behavior to local data too,
15190 such as to static variables in C@. @option{-mlocal-sdata} is the
15191 default for all configurations.
15193 If the linker complains that an application is using too much small data,
15194 you might want to try rebuilding the less performance-critical parts with
15195 @option{-mno-local-sdata}. You might also want to build large
15196 libraries with @option{-mno-local-sdata}, so that the libraries leave
15197 more room for the main program.
15199 @item -mextern-sdata
15200 @itemx -mno-extern-sdata
15201 @opindex mextern-sdata
15202 @opindex mno-extern-sdata
15203 Assume (do not assume) that externally-defined data will be in
15204 a small data section if that data is within the @option{-G} limit.
15205 @option{-mextern-sdata} is the default for all configurations.
15207 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
15208 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
15209 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
15210 is placed in a small data section. If @var{Var} is defined by another
15211 module, you must either compile that module with a high-enough
15212 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
15213 definition. If @var{Var} is common, you must link the application
15214 with a high-enough @option{-G} setting.
15216 The easiest way of satisfying these restrictions is to compile
15217 and link every module with the same @option{-G} option. However,
15218 you may wish to build a library that supports several different
15219 small data limits. You can do this by compiling the library with
15220 the highest supported @option{-G} setting and additionally using
15221 @option{-mno-extern-sdata} to stop the library from making assumptions
15222 about externally-defined data.
15228 Use (do not use) GP-relative accesses for symbols that are known to be
15229 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
15230 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
15233 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
15234 might not hold the value of @code{_gp}. For example, if the code is
15235 part of a library that might be used in a boot monitor, programs that
15236 call boot monitor routines will pass an unknown value in @code{$gp}.
15237 (In such situations, the boot monitor itself would usually be compiled
15238 with @option{-G0}.)
15240 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
15241 @option{-mno-extern-sdata}.
15243 @item -membedded-data
15244 @itemx -mno-embedded-data
15245 @opindex membedded-data
15246 @opindex mno-embedded-data
15247 Allocate variables to the read-only data section first if possible, then
15248 next in the small data section if possible, otherwise in data. This gives
15249 slightly slower code than the default, but reduces the amount of RAM required
15250 when executing, and thus may be preferred for some embedded systems.
15252 @item -muninit-const-in-rodata
15253 @itemx -mno-uninit-const-in-rodata
15254 @opindex muninit-const-in-rodata
15255 @opindex mno-uninit-const-in-rodata
15256 Put uninitialized @code{const} variables in the read-only data section.
15257 This option is only meaningful in conjunction with @option{-membedded-data}.
15259 @item -mcode-readable=@var{setting}
15260 @opindex mcode-readable
15261 Specify whether GCC may generate code that reads from executable sections.
15262 There are three possible settings:
15265 @item -mcode-readable=yes
15266 Instructions may freely access executable sections. This is the
15269 @item -mcode-readable=pcrel
15270 MIPS16 PC-relative load instructions can access executable sections,
15271 but other instructions must not do so. This option is useful on 4KSc
15272 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
15273 It is also useful on processors that can be configured to have a dual
15274 instruction/data SRAM interface and that, like the M4K, automatically
15275 redirect PC-relative loads to the instruction RAM.
15277 @item -mcode-readable=no
15278 Instructions must not access executable sections. This option can be
15279 useful on targets that are configured to have a dual instruction/data
15280 SRAM interface but that (unlike the M4K) do not automatically redirect
15281 PC-relative loads to the instruction RAM.
15284 @item -msplit-addresses
15285 @itemx -mno-split-addresses
15286 @opindex msplit-addresses
15287 @opindex mno-split-addresses
15288 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
15289 relocation operators. This option has been superseded by
15290 @option{-mexplicit-relocs} but is retained for backwards compatibility.
15292 @item -mexplicit-relocs
15293 @itemx -mno-explicit-relocs
15294 @opindex mexplicit-relocs
15295 @opindex mno-explicit-relocs
15296 Use (do not use) assembler relocation operators when dealing with symbolic
15297 addresses. The alternative, selected by @option{-mno-explicit-relocs},
15298 is to use assembler macros instead.
15300 @option{-mexplicit-relocs} is the default if GCC was configured
15301 to use an assembler that supports relocation operators.
15303 @item -mcheck-zero-division
15304 @itemx -mno-check-zero-division
15305 @opindex mcheck-zero-division
15306 @opindex mno-check-zero-division
15307 Trap (do not trap) on integer division by zero.
15309 The default is @option{-mcheck-zero-division}.
15311 @item -mdivide-traps
15312 @itemx -mdivide-breaks
15313 @opindex mdivide-traps
15314 @opindex mdivide-breaks
15315 MIPS systems check for division by zero by generating either a
15316 conditional trap or a break instruction. Using traps results in
15317 smaller code, but is only supported on MIPS II and later. Also, some
15318 versions of the Linux kernel have a bug that prevents trap from
15319 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
15320 allow conditional traps on architectures that support them and
15321 @option{-mdivide-breaks} to force the use of breaks.
15323 The default is usually @option{-mdivide-traps}, but this can be
15324 overridden at configure time using @option{--with-divide=breaks}.
15325 Divide-by-zero checks can be completely disabled using
15326 @option{-mno-check-zero-division}.
15331 @opindex mno-memcpy
15332 Force (do not force) the use of @code{memcpy()} for non-trivial block
15333 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
15334 most constant-sized copies.
15337 @itemx -mno-long-calls
15338 @opindex mlong-calls
15339 @opindex mno-long-calls
15340 Disable (do not disable) use of the @code{jal} instruction. Calling
15341 functions using @code{jal} is more efficient but requires the caller
15342 and callee to be in the same 256 megabyte segment.
15344 This option has no effect on abicalls code. The default is
15345 @option{-mno-long-calls}.
15351 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
15352 instructions, as provided by the R4650 ISA@.
15355 @itemx -mno-fused-madd
15356 @opindex mfused-madd
15357 @opindex mno-fused-madd
15358 Enable (disable) use of the floating point multiply-accumulate
15359 instructions, when they are available. The default is
15360 @option{-mfused-madd}.
15362 When multiply-accumulate instructions are used, the intermediate
15363 product is calculated to infinite precision and is not subject to
15364 the FCSR Flush to Zero bit. This may be undesirable in some
15369 Tell the MIPS assembler to not run its preprocessor over user
15370 assembler files (with a @samp{.s} suffix) when assembling them.
15375 @opindex mno-fix-24k
15376 Work around the 24K E48 (lost data on stores during refill) errata.
15377 The workarounds are implemented by the assembler rather than by GCC.
15380 @itemx -mno-fix-r4000
15381 @opindex mfix-r4000
15382 @opindex mno-fix-r4000
15383 Work around certain R4000 CPU errata:
15386 A double-word or a variable shift may give an incorrect result if executed
15387 immediately after starting an integer division.
15389 A double-word or a variable shift may give an incorrect result if executed
15390 while an integer multiplication is in progress.
15392 An integer division may give an incorrect result if started in a delay slot
15393 of a taken branch or a jump.
15397 @itemx -mno-fix-r4400
15398 @opindex mfix-r4400
15399 @opindex mno-fix-r4400
15400 Work around certain R4400 CPU errata:
15403 A double-word or a variable shift may give an incorrect result if executed
15404 immediately after starting an integer division.
15408 @itemx -mno-fix-r10000
15409 @opindex mfix-r10000
15410 @opindex mno-fix-r10000
15411 Work around certain R10000 errata:
15414 @code{ll}/@code{sc} sequences may not behave atomically on revisions
15415 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15418 This option can only be used if the target architecture supports
15419 branch-likely instructions. @option{-mfix-r10000} is the default when
15420 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15424 @itemx -mno-fix-vr4120
15425 @opindex mfix-vr4120
15426 Work around certain VR4120 errata:
15429 @code{dmultu} does not always produce the correct result.
15431 @code{div} and @code{ddiv} do not always produce the correct result if one
15432 of the operands is negative.
15434 The workarounds for the division errata rely on special functions in
15435 @file{libgcc.a}. At present, these functions are only provided by
15436 the @code{mips64vr*-elf} configurations.
15438 Other VR4120 errata require a nop to be inserted between certain pairs of
15439 instructions. These errata are handled by the assembler, not by GCC itself.
15442 @opindex mfix-vr4130
15443 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15444 workarounds are implemented by the assembler rather than by GCC,
15445 although GCC will avoid using @code{mflo} and @code{mfhi} if the
15446 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15447 instructions are available instead.
15450 @itemx -mno-fix-sb1
15452 Work around certain SB-1 CPU core errata.
15453 (This flag currently works around the SB-1 revision 2
15454 ``F1'' and ``F2'' floating point errata.)
15456 @item -mr10k-cache-barrier=@var{setting}
15457 @opindex mr10k-cache-barrier
15458 Specify whether GCC should insert cache barriers to avoid the
15459 side-effects of speculation on R10K processors.
15461 In common with many processors, the R10K tries to predict the outcome
15462 of a conditional branch and speculatively executes instructions from
15463 the ``taken'' branch. It later aborts these instructions if the
15464 predicted outcome was wrong. However, on the R10K, even aborted
15465 instructions can have side effects.
15467 This problem only affects kernel stores and, depending on the system,
15468 kernel loads. As an example, a speculatively-executed store may load
15469 the target memory into cache and mark the cache line as dirty, even if
15470 the store itself is later aborted. If a DMA operation writes to the
15471 same area of memory before the ``dirty'' line is flushed, the cached
15472 data will overwrite the DMA-ed data. See the R10K processor manual
15473 for a full description, including other potential problems.
15475 One workaround is to insert cache barrier instructions before every memory
15476 access that might be speculatively executed and that might have side
15477 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15478 controls GCC's implementation of this workaround. It assumes that
15479 aborted accesses to any byte in the following regions will not have
15484 the memory occupied by the current function's stack frame;
15487 the memory occupied by an incoming stack argument;
15490 the memory occupied by an object with a link-time-constant address.
15493 It is the kernel's responsibility to ensure that speculative
15494 accesses to these regions are indeed safe.
15496 If the input program contains a function declaration such as:
15502 then the implementation of @code{foo} must allow @code{j foo} and
15503 @code{jal foo} to be executed speculatively. GCC honors this
15504 restriction for functions it compiles itself. It expects non-GCC
15505 functions (such as hand-written assembly code) to do the same.
15507 The option has three forms:
15510 @item -mr10k-cache-barrier=load-store
15511 Insert a cache barrier before a load or store that might be
15512 speculatively executed and that might have side effects even
15515 @item -mr10k-cache-barrier=store
15516 Insert a cache barrier before a store that might be speculatively
15517 executed and that might have side effects even if aborted.
15519 @item -mr10k-cache-barrier=none
15520 Disable the insertion of cache barriers. This is the default setting.
15523 @item -mflush-func=@var{func}
15524 @itemx -mno-flush-func
15525 @opindex mflush-func
15526 Specifies the function to call to flush the I and D caches, or to not
15527 call any such function. If called, the function must take the same
15528 arguments as the common @code{_flush_func()}, that is, the address of the
15529 memory range for which the cache is being flushed, the size of the
15530 memory range, and the number 3 (to flush both caches). The default
15531 depends on the target GCC was configured for, but commonly is either
15532 @samp{_flush_func} or @samp{__cpu_flush}.
15534 @item mbranch-cost=@var{num}
15535 @opindex mbranch-cost
15536 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15537 This cost is only a heuristic and is not guaranteed to produce
15538 consistent results across releases. A zero cost redundantly selects
15539 the default, which is based on the @option{-mtune} setting.
15541 @item -mbranch-likely
15542 @itemx -mno-branch-likely
15543 @opindex mbranch-likely
15544 @opindex mno-branch-likely
15545 Enable or disable use of Branch Likely instructions, regardless of the
15546 default for the selected architecture. By default, Branch Likely
15547 instructions may be generated if they are supported by the selected
15548 architecture. An exception is for the MIPS32 and MIPS64 architectures
15549 and processors which implement those architectures; for those, Branch
15550 Likely instructions will not be generated by default because the MIPS32
15551 and MIPS64 architectures specifically deprecate their use.
15553 @item -mfp-exceptions
15554 @itemx -mno-fp-exceptions
15555 @opindex mfp-exceptions
15556 Specifies whether FP exceptions are enabled. This affects how we schedule
15557 FP instructions for some processors. The default is that FP exceptions are
15560 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15561 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15564 @item -mvr4130-align
15565 @itemx -mno-vr4130-align
15566 @opindex mvr4130-align
15567 The VR4130 pipeline is two-way superscalar, but can only issue two
15568 instructions together if the first one is 8-byte aligned. When this
15569 option is enabled, GCC will align pairs of instructions that it
15570 thinks should execute in parallel.
15572 This option only has an effect when optimizing for the VR4130.
15573 It normally makes code faster, but at the expense of making it bigger.
15574 It is enabled by default at optimization level @option{-O3}.
15579 Enable (disable) generation of @code{synci} instructions on
15580 architectures that support it. The @code{synci} instructions (if
15581 enabled) will be generated when @code{__builtin___clear_cache()} is
15584 This option defaults to @code{-mno-synci}, but the default can be
15585 overridden by configuring with @code{--with-synci}.
15587 When compiling code for single processor systems, it is generally safe
15588 to use @code{synci}. However, on many multi-core (SMP) systems, it
15589 will not invalidate the instruction caches on all cores and may lead
15590 to undefined behavior.
15592 @item -mrelax-pic-calls
15593 @itemx -mno-relax-pic-calls
15594 @opindex mrelax-pic-calls
15595 Try to turn PIC calls that are normally dispatched via register
15596 @code{$25} into direct calls. This is only possible if the linker can
15597 resolve the destination at link-time and if the destination is within
15598 range for a direct call.
15600 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15601 an assembler and a linker that supports the @code{.reloc} assembly
15602 directive and @code{-mexplicit-relocs} is in effect. With
15603 @code{-mno-explicit-relocs}, this optimization can be performed by the
15604 assembler and the linker alone without help from the compiler.
15606 @item -mmcount-ra-address
15607 @itemx -mno-mcount-ra-address
15608 @opindex mmcount-ra-address
15609 @opindex mno-mcount-ra-address
15610 Emit (do not emit) code that allows @code{_mcount} to modify the
15611 calling function's return address. When enabled, this option extends
15612 the usual @code{_mcount} interface with a new @var{ra-address}
15613 parameter, which has type @code{intptr_t *} and is passed in register
15614 @code{$12}. @code{_mcount} can then modify the return address by
15615 doing both of the following:
15618 Returning the new address in register @code{$31}.
15620 Storing the new address in @code{*@var{ra-address}},
15621 if @var{ra-address} is nonnull.
15624 The default is @option{-mno-mcount-ra-address}.
15629 @subsection MMIX Options
15630 @cindex MMIX Options
15632 These options are defined for the MMIX:
15636 @itemx -mno-libfuncs
15638 @opindex mno-libfuncs
15639 Specify that intrinsic library functions are being compiled, passing all
15640 values in registers, no matter the size.
15643 @itemx -mno-epsilon
15645 @opindex mno-epsilon
15646 Generate floating-point comparison instructions that compare with respect
15647 to the @code{rE} epsilon register.
15649 @item -mabi=mmixware
15651 @opindex mabi=mmixware
15653 Generate code that passes function parameters and return values that (in
15654 the called function) are seen as registers @code{$0} and up, as opposed to
15655 the GNU ABI which uses global registers @code{$231} and up.
15657 @item -mzero-extend
15658 @itemx -mno-zero-extend
15659 @opindex mzero-extend
15660 @opindex mno-zero-extend
15661 When reading data from memory in sizes shorter than 64 bits, use (do not
15662 use) zero-extending load instructions by default, rather than
15663 sign-extending ones.
15666 @itemx -mno-knuthdiv
15668 @opindex mno-knuthdiv
15669 Make the result of a division yielding a remainder have the same sign as
15670 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15671 remainder follows the sign of the dividend. Both methods are
15672 arithmetically valid, the latter being almost exclusively used.
15674 @item -mtoplevel-symbols
15675 @itemx -mno-toplevel-symbols
15676 @opindex mtoplevel-symbols
15677 @opindex mno-toplevel-symbols
15678 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15679 code can be used with the @code{PREFIX} assembly directive.
15683 Generate an executable in the ELF format, rather than the default
15684 @samp{mmo} format used by the @command{mmix} simulator.
15686 @item -mbranch-predict
15687 @itemx -mno-branch-predict
15688 @opindex mbranch-predict
15689 @opindex mno-branch-predict
15690 Use (do not use) the probable-branch instructions, when static branch
15691 prediction indicates a probable branch.
15693 @item -mbase-addresses
15694 @itemx -mno-base-addresses
15695 @opindex mbase-addresses
15696 @opindex mno-base-addresses
15697 Generate (do not generate) code that uses @emph{base addresses}. Using a
15698 base address automatically generates a request (handled by the assembler
15699 and the linker) for a constant to be set up in a global register. The
15700 register is used for one or more base address requests within the range 0
15701 to 255 from the value held in the register. The generally leads to short
15702 and fast code, but the number of different data items that can be
15703 addressed is limited. This means that a program that uses lots of static
15704 data may require @option{-mno-base-addresses}.
15706 @item -msingle-exit
15707 @itemx -mno-single-exit
15708 @opindex msingle-exit
15709 @opindex mno-single-exit
15710 Force (do not force) generated code to have a single exit point in each
15714 @node MN10300 Options
15715 @subsection MN10300 Options
15716 @cindex MN10300 options
15718 These @option{-m} options are defined for Matsushita MN10300 architectures:
15723 Generate code to avoid bugs in the multiply instructions for the MN10300
15724 processors. This is the default.
15726 @item -mno-mult-bug
15727 @opindex mno-mult-bug
15728 Do not generate code to avoid bugs in the multiply instructions for the
15729 MN10300 processors.
15733 Generate code which uses features specific to the AM33 processor.
15737 Do not generate code which uses features specific to the AM33 processor. This
15742 Generate code which uses features specific to the AM33/2.0 processor.
15746 Generate code which uses features specific to the AM34 processor.
15748 @item -mtune=@var{cpu-type}
15750 Use the timing characteristics of the indicated CPU type when
15751 scheduling instructions. This does not change the targeted processor
15752 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15753 @samp{am33-2} or @samp{am34}.
15755 @item -mreturn-pointer-on-d0
15756 @opindex mreturn-pointer-on-d0
15757 When generating a function which returns a pointer, return the pointer
15758 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15759 only in a0, and attempts to call such functions without a prototype
15760 would result in errors. Note that this option is on by default; use
15761 @option{-mno-return-pointer-on-d0} to disable it.
15765 Do not link in the C run-time initialization object file.
15769 Indicate to the linker that it should perform a relaxation optimization pass
15770 to shorten branches, calls and absolute memory addresses. This option only
15771 has an effect when used on the command line for the final link step.
15773 This option makes symbolic debugging impossible.
15777 Allow the compiler to generate @emph{Long Instruction Word}
15778 instructions if the target is the @samp{AM33} or later. This is the
15779 default. This option defines the preprocessor macro @samp{__LIW__}.
15783 Do not allow the compiler to generate @emph{Long Instruction Word}
15784 instructions. This option defines the preprocessor macro
15789 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15790 instructions if the target is the @samp{AM33} or later. This is the
15791 default. This option defines the preprocessor macro @samp{__SETLB__}.
15795 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15796 instructions. This option defines the preprocessor macro
15797 @samp{__NO_SETLB__}.
15801 @node PDP-11 Options
15802 @subsection PDP-11 Options
15803 @cindex PDP-11 Options
15805 These options are defined for the PDP-11:
15810 Use hardware FPP floating point. This is the default. (FIS floating
15811 point on the PDP-11/40 is not supported.)
15814 @opindex msoft-float
15815 Do not use hardware floating point.
15819 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15823 Return floating-point results in memory. This is the default.
15827 Generate code for a PDP-11/40.
15831 Generate code for a PDP-11/45. This is the default.
15835 Generate code for a PDP-11/10.
15837 @item -mbcopy-builtin
15838 @opindex mbcopy-builtin
15839 Use inline @code{movmemhi} patterns for copying memory. This is the
15844 Do not use inline @code{movmemhi} patterns for copying memory.
15850 Use 16-bit @code{int}. This is the default.
15856 Use 32-bit @code{int}.
15859 @itemx -mno-float32
15861 @opindex mno-float32
15862 Use 64-bit @code{float}. This is the default.
15865 @itemx -mno-float64
15867 @opindex mno-float64
15868 Use 32-bit @code{float}.
15872 Use @code{abshi2} pattern. This is the default.
15876 Do not use @code{abshi2} pattern.
15878 @item -mbranch-expensive
15879 @opindex mbranch-expensive
15880 Pretend that branches are expensive. This is for experimenting with
15881 code generation only.
15883 @item -mbranch-cheap
15884 @opindex mbranch-cheap
15885 Do not pretend that branches are expensive. This is the default.
15889 Use Unix assembler syntax. This is the default when configured for
15890 @samp{pdp11-*-bsd}.
15894 Use DEC assembler syntax. This is the default when configured for any
15895 PDP-11 target other than @samp{pdp11-*-bsd}.
15898 @node picoChip Options
15899 @subsection picoChip Options
15900 @cindex picoChip options
15902 These @samp{-m} options are defined for picoChip implementations:
15906 @item -mae=@var{ae_type}
15908 Set the instruction set, register set, and instruction scheduling
15909 parameters for array element type @var{ae_type}. Supported values
15910 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15912 @option{-mae=ANY} selects a completely generic AE type. Code
15913 generated with this option will run on any of the other AE types. The
15914 code will not be as efficient as it would be if compiled for a specific
15915 AE type, and some types of operation (e.g., multiplication) will not
15916 work properly on all types of AE.
15918 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15919 for compiled code, and is the default.
15921 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15922 option may suffer from poor performance of byte (char) manipulation,
15923 since the DSP AE does not provide hardware support for byte load/stores.
15925 @item -msymbol-as-address
15926 Enable the compiler to directly use a symbol name as an address in a
15927 load/store instruction, without first loading it into a
15928 register. Typically, the use of this option will generate larger
15929 programs, which run faster than when the option isn't used. However, the
15930 results vary from program to program, so it is left as a user option,
15931 rather than being permanently enabled.
15933 @item -mno-inefficient-warnings
15934 Disables warnings about the generation of inefficient code. These
15935 warnings can be generated, for example, when compiling code which
15936 performs byte-level memory operations on the MAC AE type. The MAC AE has
15937 no hardware support for byte-level memory operations, so all byte
15938 load/stores must be synthesized from word load/store operations. This is
15939 inefficient and a warning will be generated indicating to the programmer
15940 that they should rewrite the code to avoid byte operations, or to target
15941 an AE type which has the necessary hardware support. This option enables
15942 the warning to be turned off.
15946 @node PowerPC Options
15947 @subsection PowerPC Options
15948 @cindex PowerPC options
15950 These are listed under @xref{RS/6000 and PowerPC Options}.
15953 @subsection RL78 Options
15954 @cindex RL78 Options
15960 Links in additional target libraries to support operation within a
15967 Specifies the type of hardware multiplication support to be used. The
15968 default is @code{none}, which uses software multiplication functions.
15969 The @code{g13} option is for the hardware multiply/divide peripheral
15970 only on the RL78/G13 targets. The @code{rl78} option is for the
15971 standard hardware multiplication defined in the RL78 software manual.
15975 @node RS/6000 and PowerPC Options
15976 @subsection IBM RS/6000 and PowerPC Options
15977 @cindex RS/6000 and PowerPC Options
15978 @cindex IBM RS/6000 and PowerPC Options
15980 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15987 @itemx -mno-powerpc
15988 @itemx -mpowerpc-gpopt
15989 @itemx -mno-powerpc-gpopt
15990 @itemx -mpowerpc-gfxopt
15991 @itemx -mno-powerpc-gfxopt
15994 @itemx -mno-powerpc64
15998 @itemx -mno-popcntb
16000 @itemx -mno-popcntd
16009 @itemx -mno-hard-dfp
16013 @opindex mno-power2
16015 @opindex mno-powerpc
16016 @opindex mpowerpc-gpopt
16017 @opindex mno-powerpc-gpopt
16018 @opindex mpowerpc-gfxopt
16019 @opindex mno-powerpc-gfxopt
16020 @opindex mpowerpc64
16021 @opindex mno-powerpc64
16025 @opindex mno-popcntb
16027 @opindex mno-popcntd
16033 @opindex mno-mfpgpr
16035 @opindex mno-hard-dfp
16036 GCC supports two related instruction set architectures for the
16037 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
16038 instructions supported by the @samp{rios} chip set used in the original
16039 RS/6000 systems and the @dfn{PowerPC} instruction set is the
16040 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
16041 the IBM 4xx, 6xx, and follow-on microprocessors.
16043 Neither architecture is a subset of the other. However there is a
16044 large common subset of instructions supported by both. An MQ
16045 register is included in processors supporting the POWER architecture.
16047 You use these options to specify which instructions are available on the
16048 processor you are using. The default value of these options is
16049 determined when configuring GCC@. Specifying the
16050 @option{-mcpu=@var{cpu_type}} overrides the specification of these
16051 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
16052 rather than the options listed above.
16054 The @option{-mpower} option allows GCC to generate instructions that
16055 are found only in the POWER architecture and to use the MQ register.
16056 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
16057 to generate instructions that are present in the POWER2 architecture but
16058 not the original POWER architecture.
16060 The @option{-mpowerpc} option allows GCC to generate instructions that
16061 are found only in the 32-bit subset of the PowerPC architecture.
16062 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
16063 GCC to use the optional PowerPC architecture instructions in the
16064 General Purpose group, including floating-point square root. Specifying
16065 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
16066 use the optional PowerPC architecture instructions in the Graphics
16067 group, including floating-point select.
16069 The @option{-mmfcrf} option allows GCC to generate the move from
16070 condition register field instruction implemented on the POWER4
16071 processor and other processors that support the PowerPC V2.01
16073 The @option{-mpopcntb} option allows GCC to generate the popcount and
16074 double precision FP reciprocal estimate instruction implemented on the
16075 POWER5 processor and other processors that support the PowerPC V2.02
16077 The @option{-mpopcntd} option allows GCC to generate the popcount
16078 instruction implemented on the POWER7 processor and other processors
16079 that support the PowerPC V2.06 architecture.
16080 The @option{-mfprnd} option allows GCC to generate the FP round to
16081 integer instructions implemented on the POWER5+ processor and other
16082 processors that support the PowerPC V2.03 architecture.
16083 The @option{-mcmpb} option allows GCC to generate the compare bytes
16084 instruction implemented on the POWER6 processor and other processors
16085 that support the PowerPC V2.05 architecture.
16086 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
16087 general purpose register instructions implemented on the POWER6X
16088 processor and other processors that support the extended PowerPC V2.05
16090 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
16091 point instructions implemented on some POWER processors.
16093 The @option{-mpowerpc64} option allows GCC to generate the additional
16094 64-bit instructions that are found in the full PowerPC64 architecture
16095 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
16096 @option{-mno-powerpc64}.
16098 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
16099 will use only the instructions in the common subset of both
16100 architectures plus some special AIX common-mode calls, and will not use
16101 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
16102 permits GCC to use any instruction from either architecture and to
16103 allow use of the MQ register; specify this for the Motorola MPC601.
16105 @item -mnew-mnemonics
16106 @itemx -mold-mnemonics
16107 @opindex mnew-mnemonics
16108 @opindex mold-mnemonics
16109 Select which mnemonics to use in the generated assembler code. With
16110 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
16111 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
16112 assembler mnemonics defined for the POWER architecture. Instructions
16113 defined in only one architecture have only one mnemonic; GCC uses that
16114 mnemonic irrespective of which of these options is specified.
16116 GCC defaults to the mnemonics appropriate for the architecture in
16117 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
16118 value of these option. Unless you are building a cross-compiler, you
16119 should normally not specify either @option{-mnew-mnemonics} or
16120 @option{-mold-mnemonics}, but should instead accept the default.
16122 @item -mcpu=@var{cpu_type}
16124 Set architecture type, register usage, choice of mnemonics, and
16125 instruction scheduling parameters for machine type @var{cpu_type}.
16126 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
16127 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
16128 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
16129 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
16130 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
16131 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
16132 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
16133 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
16134 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
16135 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
16136 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
16138 @option{-mcpu=common} selects a completely generic processor. Code
16139 generated under this option will run on any POWER or PowerPC processor.
16140 GCC will use only the instructions in the common subset of both
16141 architectures, and will not use the MQ register. GCC assumes a generic
16142 processor model for scheduling purposes.
16144 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
16145 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
16146 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
16147 types, with an appropriate, generic processor model assumed for
16148 scheduling purposes.
16150 The other options specify a specific processor. Code generated under
16151 those options will run best on that processor, and may not run at all on
16154 The @option{-mcpu} options automatically enable or disable the
16157 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
16158 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
16159 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
16160 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
16162 The particular options set for any particular CPU will vary between
16163 compiler versions, depending on what setting seems to produce optimal
16164 code for that CPU; it doesn't necessarily reflect the actual hardware's
16165 capabilities. If you wish to set an individual option to a particular
16166 value, you may specify it after the @option{-mcpu} option, like
16167 @samp{-mcpu=970 -mno-altivec}.
16169 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
16170 not enabled or disabled by the @option{-mcpu} option at present because
16171 AIX does not have full support for these options. You may still
16172 enable or disable them individually if you're sure it'll work in your
16175 @item -mtune=@var{cpu_type}
16177 Set the instruction scheduling parameters for machine type
16178 @var{cpu_type}, but do not set the architecture type, register usage, or
16179 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
16180 values for @var{cpu_type} are used for @option{-mtune} as for
16181 @option{-mcpu}. If both are specified, the code generated will use the
16182 architecture, registers, and mnemonics set by @option{-mcpu}, but the
16183 scheduling parameters set by @option{-mtune}.
16185 @item -mcmodel=small
16186 @opindex mcmodel=small
16187 Generate PowerPC64 code for the small model: The TOC is limited to
16190 @item -mcmodel=medium
16191 @opindex mcmodel=medium
16192 Generate PowerPC64 code for the medium model: The TOC and other static
16193 data may be up to a total of 4G in size.
16195 @item -mcmodel=large
16196 @opindex mcmodel=large
16197 Generate PowerPC64 code for the large model: The TOC may be up to 4G
16198 in size. Other data and code is only limited by the 64-bit address
16202 @itemx -mno-altivec
16204 @opindex mno-altivec
16205 Generate code that uses (does not use) AltiVec instructions, and also
16206 enable the use of built-in functions that allow more direct access to
16207 the AltiVec instruction set. You may also need to set
16208 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
16214 @opindex mno-vrsave
16215 Generate VRSAVE instructions when generating AltiVec code.
16217 @item -mgen-cell-microcode
16218 @opindex mgen-cell-microcode
16219 Generate Cell microcode instructions
16221 @item -mwarn-cell-microcode
16222 @opindex mwarn-cell-microcode
16223 Warning when a Cell microcode instruction is going to emitted. An example
16224 of a Cell microcode instruction is a variable shift.
16227 @opindex msecure-plt
16228 Generate code that allows ld and ld.so to build executables and shared
16229 libraries with non-exec .plt and .got sections. This is a PowerPC
16230 32-bit SYSV ABI option.
16234 Generate code that uses a BSS .plt section that ld.so fills in, and
16235 requires .plt and .got sections that are both writable and executable.
16236 This is a PowerPC 32-bit SYSV ABI option.
16242 This switch enables or disables the generation of ISEL instructions.
16244 @item -misel=@var{yes/no}
16245 This switch has been deprecated. Use @option{-misel} and
16246 @option{-mno-isel} instead.
16252 This switch enables or disables the generation of SPE simd
16258 @opindex mno-paired
16259 This switch enables or disables the generation of PAIRED simd
16262 @item -mspe=@var{yes/no}
16263 This option has been deprecated. Use @option{-mspe} and
16264 @option{-mno-spe} instead.
16270 Generate code that uses (does not use) vector/scalar (VSX)
16271 instructions, and also enable the use of built-in functions that allow
16272 more direct access to the VSX instruction set.
16274 @item -mfloat-gprs=@var{yes/single/double/no}
16275 @itemx -mfloat-gprs
16276 @opindex mfloat-gprs
16277 This switch enables or disables the generation of floating point
16278 operations on the general purpose registers for architectures that
16281 The argument @var{yes} or @var{single} enables the use of
16282 single-precision floating point operations.
16284 The argument @var{double} enables the use of single and
16285 double-precision floating point operations.
16287 The argument @var{no} disables floating point operations on the
16288 general purpose registers.
16290 This option is currently only available on the MPC854x.
16296 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
16297 targets (including GNU/Linux). The 32-bit environment sets int, long
16298 and pointer to 32 bits and generates code that runs on any PowerPC
16299 variant. The 64-bit environment sets int to 32 bits and long and
16300 pointer to 64 bits, and generates code for PowerPC64, as for
16301 @option{-mpowerpc64}.
16304 @itemx -mno-fp-in-toc
16305 @itemx -mno-sum-in-toc
16306 @itemx -mminimal-toc
16308 @opindex mno-fp-in-toc
16309 @opindex mno-sum-in-toc
16310 @opindex mminimal-toc
16311 Modify generation of the TOC (Table Of Contents), which is created for
16312 every executable file. The @option{-mfull-toc} option is selected by
16313 default. In that case, GCC will allocate at least one TOC entry for
16314 each unique non-automatic variable reference in your program. GCC
16315 will also place floating-point constants in the TOC@. However, only
16316 16,384 entries are available in the TOC@.
16318 If you receive a linker error message that saying you have overflowed
16319 the available TOC space, you can reduce the amount of TOC space used
16320 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
16321 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
16322 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
16323 generate code to calculate the sum of an address and a constant at
16324 run-time instead of putting that sum into the TOC@. You may specify one
16325 or both of these options. Each causes GCC to produce very slightly
16326 slower and larger code at the expense of conserving TOC space.
16328 If you still run out of space in the TOC even when you specify both of
16329 these options, specify @option{-mminimal-toc} instead. This option causes
16330 GCC to make only one TOC entry for every file. When you specify this
16331 option, GCC will produce code that is slower and larger but which
16332 uses extremely little TOC space. You may wish to use this option
16333 only on files that contain less frequently executed code.
16339 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
16340 @code{long} type, and the infrastructure needed to support them.
16341 Specifying @option{-maix64} implies @option{-mpowerpc64} and
16342 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
16343 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
16346 @itemx -mno-xl-compat
16347 @opindex mxl-compat
16348 @opindex mno-xl-compat
16349 Produce code that conforms more closely to IBM XL compiler semantics
16350 when using AIX-compatible ABI@. Pass floating-point arguments to
16351 prototyped functions beyond the register save area (RSA) on the stack
16352 in addition to argument FPRs. Do not assume that most significant
16353 double in 128-bit long double value is properly rounded when comparing
16354 values and converting to double. Use XL symbol names for long double
16357 The AIX calling convention was extended but not initially documented to
16358 handle an obscure K&R C case of calling a function that takes the
16359 address of its arguments with fewer arguments than declared. IBM XL
16360 compilers access floating point arguments which do not fit in the
16361 RSA from the stack when a subroutine is compiled without
16362 optimization. Because always storing floating-point arguments on the
16363 stack is inefficient and rarely needed, this option is not enabled by
16364 default and only is necessary when calling subroutines compiled by IBM
16365 XL compilers without optimization.
16369 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
16370 application written to use message passing with special startup code to
16371 enable the application to run. The system must have PE installed in the
16372 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
16373 must be overridden with the @option{-specs=} option to specify the
16374 appropriate directory location. The Parallel Environment does not
16375 support threads, so the @option{-mpe} option and the @option{-pthread}
16376 option are incompatible.
16378 @item -malign-natural
16379 @itemx -malign-power
16380 @opindex malign-natural
16381 @opindex malign-power
16382 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
16383 @option{-malign-natural} overrides the ABI-defined alignment of larger
16384 types, such as floating-point doubles, on their natural size-based boundary.
16385 The option @option{-malign-power} instructs GCC to follow the ABI-specified
16386 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
16388 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
16392 @itemx -mhard-float
16393 @opindex msoft-float
16394 @opindex mhard-float
16395 Generate code that does not use (uses) the floating-point register set.
16396 Software floating point emulation is provided if you use the
16397 @option{-msoft-float} option, and pass the option to GCC when linking.
16399 @item -msingle-float
16400 @itemx -mdouble-float
16401 @opindex msingle-float
16402 @opindex mdouble-float
16403 Generate code for single or double-precision floating point operations.
16404 @option{-mdouble-float} implies @option{-msingle-float}.
16407 @opindex msimple-fpu
16408 Do not generate sqrt and div instructions for hardware floating point unit.
16412 Specify type of floating point unit. Valid values are @var{sp_lite}
16413 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
16414 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
16415 and @var{dp_full} (equivalent to -mdouble-float).
16418 @opindex mxilinx-fpu
16419 Perform optimizations for floating point unit on Xilinx PPC 405/440.
16422 @itemx -mno-multiple
16424 @opindex mno-multiple
16425 Generate code that uses (does not use) the load multiple word
16426 instructions and the store multiple word instructions. These
16427 instructions are generated by default on POWER systems, and not
16428 generated on PowerPC systems. Do not use @option{-mmultiple} on little
16429 endian PowerPC systems, since those instructions do not work when the
16430 processor is in little endian mode. The exceptions are PPC740 and
16431 PPC750 which permit the instructions usage in little endian mode.
16436 @opindex mno-string
16437 Generate code that uses (does not use) the load string instructions
16438 and the store string word instructions to save multiple registers and
16439 do small block moves. These instructions are generated by default on
16440 POWER systems, and not generated on PowerPC systems. Do not use
16441 @option{-mstring} on little endian PowerPC systems, since those
16442 instructions do not work when the processor is in little endian mode.
16443 The exceptions are PPC740 and PPC750 which permit the instructions
16444 usage in little endian mode.
16449 @opindex mno-update
16450 Generate code that uses (does not use) the load or store instructions
16451 that update the base register to the address of the calculated memory
16452 location. These instructions are generated by default. If you use
16453 @option{-mno-update}, there is a small window between the time that the
16454 stack pointer is updated and the address of the previous frame is
16455 stored, which means code that walks the stack frame across interrupts or
16456 signals may get corrupted data.
16458 @item -mavoid-indexed-addresses
16459 @itemx -mno-avoid-indexed-addresses
16460 @opindex mavoid-indexed-addresses
16461 @opindex mno-avoid-indexed-addresses
16462 Generate code that tries to avoid (not avoid) the use of indexed load
16463 or store instructions. These instructions can incur a performance
16464 penalty on Power6 processors in certain situations, such as when
16465 stepping through large arrays that cross a 16M boundary. This option
16466 is enabled by default when targetting Power6 and disabled otherwise.
16469 @itemx -mno-fused-madd
16470 @opindex mfused-madd
16471 @opindex mno-fused-madd
16472 Generate code that uses (does not use) the floating point multiply and
16473 accumulate instructions. These instructions are generated by default
16474 if hardware floating point is used. The machine dependent
16475 @option{-mfused-madd} option is now mapped to the machine independent
16476 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16477 mapped to @option{-ffp-contract=off}.
16483 Generate code that uses (does not use) the half-word multiply and
16484 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
16485 These instructions are generated by default when targetting those
16492 Generate code that uses (does not use) the string-search @samp{dlmzb}
16493 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
16494 generated by default when targetting those processors.
16496 @item -mno-bit-align
16498 @opindex mno-bit-align
16499 @opindex mbit-align
16500 On System V.4 and embedded PowerPC systems do not (do) force structures
16501 and unions that contain bit-fields to be aligned to the base type of the
16504 For example, by default a structure containing nothing but 8
16505 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
16506 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
16507 the structure would be aligned to a 1 byte boundary and be one byte in
16510 @item -mno-strict-align
16511 @itemx -mstrict-align
16512 @opindex mno-strict-align
16513 @opindex mstrict-align
16514 On System V.4 and embedded PowerPC systems do not (do) assume that
16515 unaligned memory references will be handled by the system.
16517 @item -mrelocatable
16518 @itemx -mno-relocatable
16519 @opindex mrelocatable
16520 @opindex mno-relocatable
16521 Generate code that allows (does not allow) a static executable to be
16522 relocated to a different address at runtime. A simple embedded
16523 PowerPC system loader should relocate the entire contents of
16524 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16525 a table of 32-bit addresses generated by this option. For this to
16526 work, all objects linked together must be compiled with
16527 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16528 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
16530 @item -mrelocatable-lib
16531 @itemx -mno-relocatable-lib
16532 @opindex mrelocatable-lib
16533 @opindex mno-relocatable-lib
16534 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16535 @code{.fixup} section to allow static executables to be relocated at
16536 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
16537 alignment of @option{-mrelocatable}. Objects compiled with
16538 @option{-mrelocatable-lib} may be linked with objects compiled with
16539 any combination of the @option{-mrelocatable} options.
16545 On System V.4 and embedded PowerPC systems do not (do) assume that
16546 register 2 contains a pointer to a global area pointing to the addresses
16547 used in the program.
16550 @itemx -mlittle-endian
16552 @opindex mlittle-endian
16553 On System V.4 and embedded PowerPC systems compile code for the
16554 processor in little endian mode. The @option{-mlittle-endian} option is
16555 the same as @option{-mlittle}.
16558 @itemx -mbig-endian
16560 @opindex mbig-endian
16561 On System V.4 and embedded PowerPC systems compile code for the
16562 processor in big endian mode. The @option{-mbig-endian} option is
16563 the same as @option{-mbig}.
16565 @item -mdynamic-no-pic
16566 @opindex mdynamic-no-pic
16567 On Darwin and Mac OS X systems, compile code so that it is not
16568 relocatable, but that its external references are relocatable. The
16569 resulting code is suitable for applications, but not shared
16572 @item -msingle-pic-base
16573 @opindex msingle-pic-base
16574 Treat the register used for PIC addressing as read-only, rather than
16575 loading it in the prologue for each function. The run-time system is
16576 responsible for initializing this register with an appropriate value
16577 before execution begins.
16579 @item -mprioritize-restricted-insns=@var{priority}
16580 @opindex mprioritize-restricted-insns
16581 This option controls the priority that is assigned to
16582 dispatch-slot restricted instructions during the second scheduling
16583 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16584 @var{no/highest/second-highest} priority to dispatch slot restricted
16587 @item -msched-costly-dep=@var{dependence_type}
16588 @opindex msched-costly-dep
16589 This option controls which dependences are considered costly
16590 by the target during instruction scheduling. The argument
16591 @var{dependence_type} takes one of the following values:
16592 @var{no}: no dependence is costly,
16593 @var{all}: all dependences are costly,
16594 @var{true_store_to_load}: a true dependence from store to load is costly,
16595 @var{store_to_load}: any dependence from store to load is costly,
16596 @var{number}: any dependence which latency >= @var{number} is costly.
16598 @item -minsert-sched-nops=@var{scheme}
16599 @opindex minsert-sched-nops
16600 This option controls which nop insertion scheme will be used during
16601 the second scheduling pass. The argument @var{scheme} takes one of the
16603 @var{no}: Don't insert nops.
16604 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16605 according to the scheduler's grouping.
16606 @var{regroup_exact}: Insert nops to force costly dependent insns into
16607 separate groups. Insert exactly as many nops as needed to force an insn
16608 to a new group, according to the estimated processor grouping.
16609 @var{number}: Insert nops to force costly dependent insns into
16610 separate groups. Insert @var{number} nops to force an insn to a new group.
16613 @opindex mcall-sysv
16614 On System V.4 and embedded PowerPC systems compile code using calling
16615 conventions that adheres to the March 1995 draft of the System V
16616 Application Binary Interface, PowerPC processor supplement. This is the
16617 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16619 @item -mcall-sysv-eabi
16621 @opindex mcall-sysv-eabi
16622 @opindex mcall-eabi
16623 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16625 @item -mcall-sysv-noeabi
16626 @opindex mcall-sysv-noeabi
16627 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16629 @item -mcall-aixdesc
16631 On System V.4 and embedded PowerPC systems compile code for the AIX
16635 @opindex mcall-linux
16636 On System V.4 and embedded PowerPC systems compile code for the
16637 Linux-based GNU system.
16639 @item -mcall-freebsd
16640 @opindex mcall-freebsd
16641 On System V.4 and embedded PowerPC systems compile code for the
16642 FreeBSD operating system.
16644 @item -mcall-netbsd
16645 @opindex mcall-netbsd
16646 On System V.4 and embedded PowerPC systems compile code for the
16647 NetBSD operating system.
16649 @item -mcall-openbsd
16650 @opindex mcall-netbsd
16651 On System V.4 and embedded PowerPC systems compile code for the
16652 OpenBSD operating system.
16654 @item -maix-struct-return
16655 @opindex maix-struct-return
16656 Return all structures in memory (as specified by the AIX ABI)@.
16658 @item -msvr4-struct-return
16659 @opindex msvr4-struct-return
16660 Return structures smaller than 8 bytes in registers (as specified by the
16663 @item -mabi=@var{abi-type}
16665 Extend the current ABI with a particular extension, or remove such extension.
16666 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16667 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16671 Extend the current ABI with SPE ABI extensions. This does not change
16672 the default ABI, instead it adds the SPE ABI extensions to the current
16676 @opindex mabi=no-spe
16677 Disable Booke SPE ABI extensions for the current ABI@.
16679 @item -mabi=ibmlongdouble
16680 @opindex mabi=ibmlongdouble
16681 Change the current ABI to use IBM extended precision long double.
16682 This is a PowerPC 32-bit SYSV ABI option.
16684 @item -mabi=ieeelongdouble
16685 @opindex mabi=ieeelongdouble
16686 Change the current ABI to use IEEE extended precision long double.
16687 This is a PowerPC 32-bit Linux ABI option.
16690 @itemx -mno-prototype
16691 @opindex mprototype
16692 @opindex mno-prototype
16693 On System V.4 and embedded PowerPC systems assume that all calls to
16694 variable argument functions are properly prototyped. Otherwise, the
16695 compiler must insert an instruction before every non prototyped call to
16696 set or clear bit 6 of the condition code register (@var{CR}) to
16697 indicate whether floating point values were passed in the floating point
16698 registers in case the function takes a variable arguments. With
16699 @option{-mprototype}, only calls to prototyped variable argument functions
16700 will set or clear the bit.
16704 On embedded PowerPC systems, assume that the startup module is called
16705 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16706 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16711 On embedded PowerPC systems, assume that the startup module is called
16712 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16717 On embedded PowerPC systems, assume that the startup module is called
16718 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16721 @item -myellowknife
16722 @opindex myellowknife
16723 On embedded PowerPC systems, assume that the startup module is called
16724 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16729 On System V.4 and embedded PowerPC systems, specify that you are
16730 compiling for a VxWorks system.
16734 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16735 header to indicate that @samp{eabi} extended relocations are used.
16741 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16742 Embedded Applications Binary Interface (eabi) which is a set of
16743 modifications to the System V.4 specifications. Selecting @option{-meabi}
16744 means that the stack is aligned to an 8 byte boundary, a function
16745 @code{__eabi} is called to from @code{main} to set up the eabi
16746 environment, and the @option{-msdata} option can use both @code{r2} and
16747 @code{r13} to point to two separate small data areas. Selecting
16748 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16749 do not call an initialization function from @code{main}, and the
16750 @option{-msdata} option will only use @code{r13} to point to a single
16751 small data area. The @option{-meabi} option is on by default if you
16752 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16755 @opindex msdata=eabi
16756 On System V.4 and embedded PowerPC systems, put small initialized
16757 @code{const} global and static data in the @samp{.sdata2} section, which
16758 is pointed to by register @code{r2}. Put small initialized
16759 non-@code{const} global and static data in the @samp{.sdata} section,
16760 which is pointed to by register @code{r13}. Put small uninitialized
16761 global and static data in the @samp{.sbss} section, which is adjacent to
16762 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16763 incompatible with the @option{-mrelocatable} option. The
16764 @option{-msdata=eabi} option also sets the @option{-memb} option.
16767 @opindex msdata=sysv
16768 On System V.4 and embedded PowerPC systems, put small global and static
16769 data in the @samp{.sdata} section, which is pointed to by register
16770 @code{r13}. Put small uninitialized global and static data in the
16771 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16772 The @option{-msdata=sysv} option is incompatible with the
16773 @option{-mrelocatable} option.
16775 @item -msdata=default
16777 @opindex msdata=default
16779 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16780 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16781 same as @option{-msdata=sysv}.
16784 @opindex msdata=data
16785 On System V.4 and embedded PowerPC systems, put small global
16786 data in the @samp{.sdata} section. Put small uninitialized global
16787 data in the @samp{.sbss} section. Do not use register @code{r13}
16788 to address small data however. This is the default behavior unless
16789 other @option{-msdata} options are used.
16793 @opindex msdata=none
16795 On embedded PowerPC systems, put all initialized global and static data
16796 in the @samp{.data} section, and all uninitialized data in the
16797 @samp{.bss} section.
16799 @item -mblock-move-inline-limit=@var{num}
16800 @opindex mblock-move-inline-limit
16801 Inline all block moves (such as calls to @code{memcpy} or structure
16802 copies) less than or equal to @var{num} bytes. The minimum value for
16803 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16804 targets. The default value is target-specific.
16808 @cindex smaller data references (PowerPC)
16809 @cindex .sdata/.sdata2 references (PowerPC)
16810 On embedded PowerPC systems, put global and static items less than or
16811 equal to @var{num} bytes into the small data or bss sections instead of
16812 the normal data or bss section. By default, @var{num} is 8. The
16813 @option{-G @var{num}} switch is also passed to the linker.
16814 All modules should be compiled with the same @option{-G @var{num}} value.
16817 @itemx -mno-regnames
16819 @opindex mno-regnames
16820 On System V.4 and embedded PowerPC systems do (do not) emit register
16821 names in the assembly language output using symbolic forms.
16824 @itemx -mno-longcall
16826 @opindex mno-longcall
16827 By default assume that all calls are far away so that a longer more
16828 expensive calling sequence is required. This is required for calls
16829 further than 32 megabytes (33,554,432 bytes) from the current location.
16830 A short call will be generated if the compiler knows
16831 the call cannot be that far away. This setting can be overridden by
16832 the @code{shortcall} function attribute, or by @code{#pragma
16835 Some linkers are capable of detecting out-of-range calls and generating
16836 glue code on the fly. On these systems, long calls are unnecessary and
16837 generate slower code. As of this writing, the AIX linker can do this,
16838 as can the GNU linker for PowerPC/64. It is planned to add this feature
16839 to the GNU linker for 32-bit PowerPC systems as well.
16841 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16842 callee, L42'', plus a ``branch island'' (glue code). The two target
16843 addresses represent the callee and the ``branch island''. The
16844 Darwin/PPC linker will prefer the first address and generate a ``bl
16845 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16846 otherwise, the linker will generate ``bl L42'' to call the ``branch
16847 island''. The ``branch island'' is appended to the body of the
16848 calling function; it computes the full 32-bit address of the callee
16851 On Mach-O (Darwin) systems, this option directs the compiler emit to
16852 the glue for every direct call, and the Darwin linker decides whether
16853 to use or discard it.
16855 In the future, we may cause GCC to ignore all longcall specifications
16856 when the linker is known to generate glue.
16858 @item -mtls-markers
16859 @itemx -mno-tls-markers
16860 @opindex mtls-markers
16861 @opindex mno-tls-markers
16862 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16863 specifying the function argument. The relocation allows ld to
16864 reliably associate function call with argument setup instructions for
16865 TLS optimization, which in turn allows gcc to better schedule the
16870 Adds support for multithreading with the @dfn{pthreads} library.
16871 This option sets flags for both the preprocessor and linker.
16876 This option will enable GCC to use the reciprocal estimate and
16877 reciprocal square root estimate instructions with additional
16878 Newton-Raphson steps to increase precision instead of doing a divide or
16879 square root and divide for floating point arguments. You should use
16880 the @option{-ffast-math} option when using @option{-mrecip} (or at
16881 least @option{-funsafe-math-optimizations},
16882 @option{-finite-math-only}, @option{-freciprocal-math} and
16883 @option{-fno-trapping-math}). Note that while the throughput of the
16884 sequence is generally higher than the throughput of the non-reciprocal
16885 instruction, the precision of the sequence can be decreased by up to 2
16886 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16889 @item -mrecip=@var{opt}
16890 @opindex mrecip=opt
16891 This option allows to control which reciprocal estimate instructions
16892 may be used. @var{opt} is a comma separated list of options, that may
16893 be preceded by a @code{!} to invert the option:
16894 @code{all}: enable all estimate instructions,
16895 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16896 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16897 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16898 @code{divf}: enable the single precision reciprocal approximation instructions;
16899 @code{divd}: enable the double precision reciprocal approximation instructions;
16900 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16901 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16902 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16904 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16905 all of the reciprocal estimate instructions, except for the
16906 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16907 which handle the double precision reciprocal square root calculations.
16909 @item -mrecip-precision
16910 @itemx -mno-recip-precision
16911 @opindex mrecip-precision
16912 Assume (do not assume) that the reciprocal estimate instructions
16913 provide higher precision estimates than is mandated by the powerpc
16914 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16915 automatically selects @option{-mrecip-precision}. The double
16916 precision square root estimate instructions are not generated by
16917 default on low precision machines, since they do not provide an
16918 estimate that converges after three steps.
16920 @item -mveclibabi=@var{type}
16921 @opindex mveclibabi
16922 Specifies the ABI type to use for vectorizing intrinsics using an
16923 external library. The only type supported at present is @code{mass},
16924 which specifies to use IBM's Mathematical Acceleration Subsystem
16925 (MASS) libraries for vectorizing intrinsics using external libraries.
16926 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16927 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16928 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16929 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16930 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16931 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16932 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16933 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16934 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16935 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16936 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16937 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16938 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16939 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16940 for power7. Both @option{-ftree-vectorize} and
16941 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16942 libraries will have to be specified at link time.
16947 Generate (do not generate) the @code{friz} instruction when the
16948 @option{-funsafe-math-optimizations} option is used to optimize
16949 rounding a floating point value to 64-bit integer and back to floating
16950 point. The @code{friz} instruction does not return the same value if
16951 the floating point number is too large to fit in an integer.
16953 @item -mpointers-to-nested-functions
16954 @itemx -mno-pointers-to-nested-functions
16955 @opindex mpointers-to-nested-functions
16956 Generate (do not generate) code to load up the static chain register
16957 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16958 systems where a function pointer points to a 3 word descriptor giving
16959 the function address, TOC value to be loaded in register @var{r2}, and
16960 static chain value to be loaded in register @var{r11}. The
16961 @option{-mpointers-to-nested-functions} is on by default. You will
16962 not be able to call through pointers to nested functions or pointers
16963 to functions compiled in other languages that use the static chain if
16964 you use the @option{-mno-pointers-to-nested-functions}.
16966 @item -msave-toc-indirect
16967 @itemx -mno-save-toc-indirect
16968 @opindex msave-toc-indirect
16969 Generate (do not generate) code to save the TOC value in the reserved
16970 stack location in the function prologue if the function calls through
16971 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
16972 saved in the prologue, it is saved just before the call through the
16973 pointer. The @option{-mno-save-toc-indirect} option is the default.
16977 @subsection RX Options
16980 These command line options are defined for RX targets:
16983 @item -m64bit-doubles
16984 @itemx -m32bit-doubles
16985 @opindex m64bit-doubles
16986 @opindex m32bit-doubles
16987 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16988 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16989 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16990 works on 32-bit values, which is why the default is
16991 @option{-m32bit-doubles}.
16997 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16998 floating point hardware. The default is enabled for the @var{RX600}
16999 series and disabled for the @var{RX200} series.
17001 Floating point instructions will only be generated for 32-bit floating
17002 point values however, so if the @option{-m64bit-doubles} option is in
17003 use then the FPU hardware will not be used for doubles.
17005 @emph{Note} If the @option{-fpu} option is enabled then
17006 @option{-funsafe-math-optimizations} is also enabled automatically.
17007 This is because the RX FPU instructions are themselves unsafe.
17009 @item -mcpu=@var{name}
17011 Selects the type of RX CPU to be targeted. Currently three types are
17012 supported, the generic @var{RX600} and @var{RX200} series hardware and
17013 the specific @var{RX610} CPU. The default is @var{RX600}.
17015 The only difference between @var{RX600} and @var{RX610} is that the
17016 @var{RX610} does not support the @code{MVTIPL} instruction.
17018 The @var{RX200} series does not have a hardware floating point unit
17019 and so @option{-nofpu} is enabled by default when this type is
17022 @item -mbig-endian-data
17023 @itemx -mlittle-endian-data
17024 @opindex mbig-endian-data
17025 @opindex mlittle-endian-data
17026 Store data (but not code) in the big-endian format. The default is
17027 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
17030 @item -msmall-data-limit=@var{N}
17031 @opindex msmall-data-limit
17032 Specifies the maximum size in bytes of global and static variables
17033 which can be placed into the small data area. Using the small data
17034 area can lead to smaller and faster code, but the size of area is
17035 limited and it is up to the programmer to ensure that the area does
17036 not overflow. Also when the small data area is used one of the RX's
17037 registers (usually @code{r13}) is reserved for use pointing to this
17038 area, so it is no longer available for use by the compiler. This
17039 could result in slower and/or larger code if variables which once
17040 could have been held in the reserved register are now pushed onto the
17043 Note, common variables (variables which have not been initialised) and
17044 constants are not placed into the small data area as they are assigned
17045 to other sections in the output executable.
17047 The default value is zero, which disables this feature. Note, this
17048 feature is not enabled by default with higher optimization levels
17049 (@option{-O2} etc) because of the potentially detrimental effects of
17050 reserving a register. It is up to the programmer to experiment and
17051 discover whether this feature is of benefit to their program. See the
17052 description of the @option{-mpid} option for a description of how the
17053 actual register to hold the small data area pointer is chosen.
17059 Use the simulator runtime. The default is to use the libgloss board
17062 @item -mas100-syntax
17063 @itemx -mno-as100-syntax
17064 @opindex mas100-syntax
17065 @opindex mno-as100-syntax
17066 When generating assembler output use a syntax that is compatible with
17067 Renesas's AS100 assembler. This syntax can also be handled by the GAS
17068 assembler but it has some restrictions so generating it is not the
17071 @item -mmax-constant-size=@var{N}
17072 @opindex mmax-constant-size
17073 Specifies the maximum size, in bytes, of a constant that can be used as
17074 an operand in a RX instruction. Although the RX instruction set does
17075 allow constants of up to 4 bytes in length to be used in instructions,
17076 a longer value equates to a longer instruction. Thus in some
17077 circumstances it can be beneficial to restrict the size of constants
17078 that are used in instructions. Constants that are too big are instead
17079 placed into a constant pool and referenced via register indirection.
17081 The value @var{N} can be between 0 and 4. A value of 0 (the default)
17082 or 4 means that constants of any size are allowed.
17086 Enable linker relaxation. Linker relaxation is a process whereby the
17087 linker will attempt to reduce the size of a program by finding shorter
17088 versions of various instructions. Disabled by default.
17090 @item -mint-register=@var{N}
17091 @opindex mint-register
17092 Specify the number of registers to reserve for fast interrupt handler
17093 functions. The value @var{N} can be between 0 and 4. A value of 1
17094 means that register @code{r13} will be reserved for the exclusive use
17095 of fast interrupt handlers. A value of 2 reserves @code{r13} and
17096 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
17097 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
17098 A value of 0, the default, does not reserve any registers.
17100 @item -msave-acc-in-interrupts
17101 @opindex msave-acc-in-interrupts
17102 Specifies that interrupt handler functions should preserve the
17103 accumulator register. This is only necessary if normal code might use
17104 the accumulator register, for example because it performs 64-bit
17105 multiplications. The default is to ignore the accumulator as this
17106 makes the interrupt handlers faster.
17112 Enables the generation of position independent data. When enabled any
17113 access to constant data will done via an offset from a base address
17114 held in a register. This allows the location of constant data to be
17115 determined at run-time without requiring the executable to be
17116 relocated, which is a benefit to embedded applications with tight
17117 memory constraints. Data that can be modified is not affected by this
17120 Note, using this feature reserves a register, usually @code{r13}, for
17121 the constant data base address. This can result in slower and/or
17122 larger code, especially in complicated functions.
17124 The actual register chosen to hold the constant data base address
17125 depends upon whether the @option{-msmall-data-limit} and/or the
17126 @option{-mint-register} command line options are enabled. Starting
17127 with register @code{r13} and proceeding downwards, registers are
17128 allocated first to satisfy the requirements of @option{-mint-register},
17129 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
17130 is possible for the small data area register to be @code{r8} if both
17131 @option{-mint-register=4} and @option{-mpid} are specified on the
17134 By default this feature is not enabled. The default can be restored
17135 via the @option{-mno-pid} command line option.
17139 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
17140 has special significance to the RX port when used with the
17141 @code{interrupt} function attribute. This attribute indicates a
17142 function intended to process fast interrupts. GCC will will ensure
17143 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
17144 and/or @code{r13} and only provided that the normal use of the
17145 corresponding registers have been restricted via the
17146 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
17149 @node S/390 and zSeries Options
17150 @subsection S/390 and zSeries Options
17151 @cindex S/390 and zSeries Options
17153 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
17157 @itemx -msoft-float
17158 @opindex mhard-float
17159 @opindex msoft-float
17160 Use (do not use) the hardware floating-point instructions and registers
17161 for floating-point operations. When @option{-msoft-float} is specified,
17162 functions in @file{libgcc.a} will be used to perform floating-point
17163 operations. When @option{-mhard-float} is specified, the compiler
17164 generates IEEE floating-point instructions. This is the default.
17167 @itemx -mno-hard-dfp
17169 @opindex mno-hard-dfp
17170 Use (do not use) the hardware decimal-floating-point instructions for
17171 decimal-floating-point operations. When @option{-mno-hard-dfp} is
17172 specified, functions in @file{libgcc.a} will be used to perform
17173 decimal-floating-point operations. When @option{-mhard-dfp} is
17174 specified, the compiler generates decimal-floating-point hardware
17175 instructions. This is the default for @option{-march=z9-ec} or higher.
17177 @item -mlong-double-64
17178 @itemx -mlong-double-128
17179 @opindex mlong-double-64
17180 @opindex mlong-double-128
17181 These switches control the size of @code{long double} type. A size
17182 of 64bit makes the @code{long double} type equivalent to the @code{double}
17183 type. This is the default.
17186 @itemx -mno-backchain
17187 @opindex mbackchain
17188 @opindex mno-backchain
17189 Store (do not store) the address of the caller's frame as backchain pointer
17190 into the callee's stack frame.
17191 A backchain may be needed to allow debugging using tools that do not understand
17192 DWARF-2 call frame information.
17193 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
17194 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
17195 the backchain is placed into the topmost word of the 96/160 byte register
17198 In general, code compiled with @option{-mbackchain} is call-compatible with
17199 code compiled with @option{-mmo-backchain}; however, use of the backchain
17200 for debugging purposes usually requires that the whole binary is built with
17201 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
17202 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17203 to build a linux kernel use @option{-msoft-float}.
17205 The default is to not maintain the backchain.
17207 @item -mpacked-stack
17208 @itemx -mno-packed-stack
17209 @opindex mpacked-stack
17210 @opindex mno-packed-stack
17211 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
17212 specified, the compiler uses the all fields of the 96/160 byte register save
17213 area only for their default purpose; unused fields still take up stack space.
17214 When @option{-mpacked-stack} is specified, register save slots are densely
17215 packed at the top of the register save area; unused space is reused for other
17216 purposes, allowing for more efficient use of the available stack space.
17217 However, when @option{-mbackchain} is also in effect, the topmost word of
17218 the save area is always used to store the backchain, and the return address
17219 register is always saved two words below the backchain.
17221 As long as the stack frame backchain is not used, code generated with
17222 @option{-mpacked-stack} is call-compatible with code generated with
17223 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
17224 S/390 or zSeries generated code that uses the stack frame backchain at run
17225 time, not just for debugging purposes. Such code is not call-compatible
17226 with code compiled with @option{-mpacked-stack}. Also, note that the
17227 combination of @option{-mbackchain},
17228 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17229 to build a linux kernel use @option{-msoft-float}.
17231 The default is to not use the packed stack layout.
17234 @itemx -mno-small-exec
17235 @opindex msmall-exec
17236 @opindex mno-small-exec
17237 Generate (or do not generate) code using the @code{bras} instruction
17238 to do subroutine calls.
17239 This only works reliably if the total executable size does not
17240 exceed 64k. The default is to use the @code{basr} instruction instead,
17241 which does not have this limitation.
17247 When @option{-m31} is specified, generate code compliant to the
17248 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
17249 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
17250 particular to generate 64-bit instructions. For the @samp{s390}
17251 targets, the default is @option{-m31}, while the @samp{s390x}
17252 targets default to @option{-m64}.
17258 When @option{-mzarch} is specified, generate code using the
17259 instructions available on z/Architecture.
17260 When @option{-mesa} is specified, generate code using the
17261 instructions available on ESA/390. Note that @option{-mesa} is
17262 not possible with @option{-m64}.
17263 When generating code compliant to the GNU/Linux for S/390 ABI,
17264 the default is @option{-mesa}. When generating code compliant
17265 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
17271 Generate (or do not generate) code using the @code{mvcle} instruction
17272 to perform block moves. When @option{-mno-mvcle} is specified,
17273 use a @code{mvc} loop instead. This is the default unless optimizing for
17280 Print (or do not print) additional debug information when compiling.
17281 The default is to not print debug information.
17283 @item -march=@var{cpu-type}
17285 Generate code that will run on @var{cpu-type}, which is the name of a system
17286 representing a certain processor type. Possible values for
17287 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
17288 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
17289 When generating code using the instructions available on z/Architecture,
17290 the default is @option{-march=z900}. Otherwise, the default is
17291 @option{-march=g5}.
17293 @item -mtune=@var{cpu-type}
17295 Tune to @var{cpu-type} everything applicable about the generated code,
17296 except for the ABI and the set of available instructions.
17297 The list of @var{cpu-type} values is the same as for @option{-march}.
17298 The default is the value used for @option{-march}.
17301 @itemx -mno-tpf-trace
17302 @opindex mtpf-trace
17303 @opindex mno-tpf-trace
17304 Generate code that adds (does not add) in TPF OS specific branches to trace
17305 routines in the operating system. This option is off by default, even
17306 when compiling for the TPF OS@.
17309 @itemx -mno-fused-madd
17310 @opindex mfused-madd
17311 @opindex mno-fused-madd
17312 Generate code that uses (does not use) the floating point multiply and
17313 accumulate instructions. These instructions are generated by default if
17314 hardware floating point is used.
17316 @item -mwarn-framesize=@var{framesize}
17317 @opindex mwarn-framesize
17318 Emit a warning if the current function exceeds the given frame size. Because
17319 this is a compile time check it doesn't need to be a real problem when the program
17320 runs. It is intended to identify functions which most probably cause
17321 a stack overflow. It is useful to be used in an environment with limited stack
17322 size e.g.@: the linux kernel.
17324 @item -mwarn-dynamicstack
17325 @opindex mwarn-dynamicstack
17326 Emit a warning if the function calls alloca or uses dynamically
17327 sized arrays. This is generally a bad idea with a limited stack size.
17329 @item -mstack-guard=@var{stack-guard}
17330 @itemx -mstack-size=@var{stack-size}
17331 @opindex mstack-guard
17332 @opindex mstack-size
17333 If these options are provided the s390 back end emits additional instructions in
17334 the function prologue which trigger a trap if the stack size is @var{stack-guard}
17335 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
17336 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
17337 the frame size of the compiled function is chosen.
17338 These options are intended to be used to help debugging stack overflow problems.
17339 The additionally emitted code causes only little overhead and hence can also be
17340 used in production like systems without greater performance degradation. The given
17341 values have to be exact powers of 2 and @var{stack-size} has to be greater than
17342 @var{stack-guard} without exceeding 64k.
17343 In order to be efficient the extra code makes the assumption that the stack starts
17344 at an address aligned to the value given by @var{stack-size}.
17345 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
17348 @node Score Options
17349 @subsection Score Options
17350 @cindex Score Options
17352 These options are defined for Score implementations:
17357 Compile code for big endian mode. This is the default.
17361 Compile code for little endian mode.
17365 Disable generate bcnz instruction.
17369 Enable generate unaligned load and store instruction.
17373 Enable the use of multiply-accumulate instructions. Disabled by default.
17377 Specify the SCORE5 as the target architecture.
17381 Specify the SCORE5U of the target architecture.
17385 Specify the SCORE7 as the target architecture. This is the default.
17389 Specify the SCORE7D as the target architecture.
17393 @subsection SH Options
17395 These @samp{-m} options are defined for the SH implementations:
17400 Generate code for the SH1.
17404 Generate code for the SH2.
17407 Generate code for the SH2e.
17411 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
17412 that the floating-point unit is not used.
17414 @item -m2a-single-only
17415 @opindex m2a-single-only
17416 Generate code for the SH2a-FPU, in such a way that no double-precision
17417 floating point operations are used.
17420 @opindex m2a-single
17421 Generate code for the SH2a-FPU assuming the floating-point unit is in
17422 single-precision mode by default.
17426 Generate code for the SH2a-FPU assuming the floating-point unit is in
17427 double-precision mode by default.
17431 Generate code for the SH3.
17435 Generate code for the SH3e.
17439 Generate code for the SH4 without a floating-point unit.
17441 @item -m4-single-only
17442 @opindex m4-single-only
17443 Generate code for the SH4 with a floating-point unit that only
17444 supports single-precision arithmetic.
17448 Generate code for the SH4 assuming the floating-point unit is in
17449 single-precision mode by default.
17453 Generate code for the SH4.
17457 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
17458 floating-point unit is not used.
17460 @item -m4a-single-only
17461 @opindex m4a-single-only
17462 Generate code for the SH4a, in such a way that no double-precision
17463 floating point operations are used.
17466 @opindex m4a-single
17467 Generate code for the SH4a assuming the floating-point unit is in
17468 single-precision mode by default.
17472 Generate code for the SH4a.
17476 Same as @option{-m4a-nofpu}, except that it implicitly passes
17477 @option{-dsp} to the assembler. GCC doesn't generate any DSP
17478 instructions at the moment.
17482 Compile code for the processor in big endian mode.
17486 Compile code for the processor in little endian mode.
17490 Align doubles at 64-bit boundaries. Note that this changes the calling
17491 conventions, and thus some functions from the standard C library will
17492 not work unless you recompile it first with @option{-mdalign}.
17496 Shorten some address references at link time, when possible; uses the
17497 linker option @option{-relax}.
17501 Use 32-bit offsets in @code{switch} tables. The default is to use
17506 Enable the use of bit manipulation instructions on SH2A.
17510 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
17511 alignment constraints.
17515 Comply with the calling conventions defined by Renesas.
17519 Comply with the calling conventions defined by Renesas.
17523 Comply with the calling conventions defined for GCC before the Renesas
17524 conventions were available. This option is the default for all
17525 targets of the SH toolchain.
17528 @opindex mnomacsave
17529 Mark the @code{MAC} register as call-clobbered, even if
17530 @option{-mhitachi} is given.
17534 Increase IEEE-compliance of floating-point code.
17535 At the moment, this is equivalent to @option{-fno-finite-math-only}.
17536 When generating 16 bit SH opcodes, getting IEEE-conforming results for
17537 comparisons of NANs / infinities incurs extra overhead in every
17538 floating point comparison, therefore the default is set to
17539 @option{-ffinite-math-only}.
17541 @item -minline-ic_invalidate
17542 @opindex minline-ic_invalidate
17543 Inline code to invalidate instruction cache entries after setting up
17544 nested function trampolines.
17545 This option has no effect if -musermode is in effect and the selected
17546 code generation option (e.g. -m4) does not allow the use of the icbi
17548 If the selected code generation option does not allow the use of the icbi
17549 instruction, and -musermode is not in effect, the inlined code will
17550 manipulate the instruction cache address array directly with an associative
17551 write. This not only requires privileged mode, but it will also
17552 fail if the cache line had been mapped via the TLB and has become unmapped.
17556 Dump instruction size and location in the assembly code.
17559 @opindex mpadstruct
17560 This option is deprecated. It pads structures to multiple of 4 bytes,
17561 which is incompatible with the SH ABI@.
17563 @item -msoft-atomic
17564 @opindex msoft-atomic
17565 Generate software atomic sequences for the atomic operations.
17566 This is the default when the target is @code{sh-*-linux*}.
17570 Optimize for space instead of speed. Implied by @option{-Os}.
17573 @opindex mprefergot
17574 When generating position-independent code, emit function calls using
17575 the Global Offset Table instead of the Procedure Linkage Table.
17579 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17580 if the inlined code would not work in user mode.
17581 This is the default when the target is @code{sh-*-linux*}.
17583 @item -multcost=@var{number}
17584 @opindex multcost=@var{number}
17585 Set the cost to assume for a multiply insn.
17587 @item -mdiv=@var{strategy}
17588 @opindex mdiv=@var{strategy}
17589 Set the division strategy to use for SHmedia code. @var{strategy} must be
17590 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17591 inv:call2, inv:fp .
17592 "fp" performs the operation in floating point. This has a very high latency,
17593 but needs only a few instructions, so it might be a good choice if
17594 your code has enough easily exploitable ILP to allow the compiler to
17595 schedule the floating point instructions together with other instructions.
17596 Division by zero causes a floating point exception.
17597 "inv" uses integer operations to calculate the inverse of the divisor,
17598 and then multiplies the dividend with the inverse. This strategy allows
17599 cse and hoisting of the inverse calculation. Division by zero calculates
17600 an unspecified result, but does not trap.
17601 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17602 have been found, or if the entire operation has been hoisted to the same
17603 place, the last stages of the inverse calculation are intertwined with the
17604 final multiply to reduce the overall latency, at the expense of using a few
17605 more instructions, and thus offering fewer scheduling opportunities with
17607 "call" calls a library function that usually implements the inv:minlat
17609 This gives high code density for m5-*media-nofpu compilations.
17610 "call2" uses a different entry point of the same library function, where it
17611 assumes that a pointer to a lookup table has already been set up, which
17612 exposes the pointer load to cse / code hoisting optimizations.
17613 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17614 code generation, but if the code stays unoptimized, revert to the "call",
17615 "call2", or "fp" strategies, respectively. Note that the
17616 potentially-trapping side effect of division by zero is carried by a
17617 separate instruction, so it is possible that all the integer instructions
17618 are hoisted out, but the marker for the side effect stays where it is.
17619 A recombination to fp operations or a call is not possible in that case.
17620 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17621 that the inverse calculation was nor separated from the multiply, they speed
17622 up division where the dividend fits into 20 bits (plus sign where applicable),
17623 by inserting a test to skip a number of operations in this case; this test
17624 slows down the case of larger dividends. inv20u assumes the case of a such
17625 a small dividend to be unlikely, and inv20l assumes it to be likely.
17627 @item -maccumulate-outgoing-args
17628 @opindex maccumulate-outgoing-args
17629 Reserve space once for outgoing arguments in the function prologue rather
17630 than around each call. Generally beneficial for performance and size. Also
17631 needed for unwinding to avoid changing the stack frame around conditional code.
17633 @item -mdivsi3_libfunc=@var{name}
17634 @opindex mdivsi3_libfunc=@var{name}
17635 Set the name of the library function used for 32 bit signed division to
17636 @var{name}. This only affect the name used in the call and inv:call
17637 division strategies, and the compiler will still expect the same
17638 sets of input/output/clobbered registers as if this option was not present.
17640 @item -mfixed-range=@var{register-range}
17641 @opindex mfixed-range
17642 Generate code treating the given register range as fixed registers.
17643 A fixed register is one that the register allocator can not use. This is
17644 useful when compiling kernel code. A register range is specified as
17645 two registers separated by a dash. Multiple register ranges can be
17646 specified separated by a comma.
17648 @item -madjust-unroll
17649 @opindex madjust-unroll
17650 Throttle unrolling to avoid thrashing target registers.
17651 This option only has an effect if the gcc code base supports the
17652 TARGET_ADJUST_UNROLL_MAX target hook.
17654 @item -mindexed-addressing
17655 @opindex mindexed-addressing
17656 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17657 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17658 semantics for the indexed addressing mode. The architecture allows the
17659 implementation of processors with 64 bit MMU, which the OS could use to
17660 get 32 bit addressing, but since no current hardware implementation supports
17661 this or any other way to make the indexed addressing mode safe to use in
17662 the 32 bit ABI, the default is -mno-indexed-addressing.
17664 @item -mgettrcost=@var{number}
17665 @opindex mgettrcost=@var{number}
17666 Set the cost assumed for the gettr instruction to @var{number}.
17667 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17671 Assume pt* instructions won't trap. This will generally generate better
17672 scheduled code, but is unsafe on current hardware. The current architecture
17673 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17674 This has the unintentional effect of making it unsafe to schedule ptabs /
17675 ptrel before a branch, or hoist it out of a loop. For example,
17676 __do_global_ctors, a part of libgcc that runs constructors at program
17677 startup, calls functions in a list which is delimited by @minus{}1. With the
17678 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17679 That means that all the constructors will be run a bit quicker, but when
17680 the loop comes to the end of the list, the program crashes because ptabs
17681 loads @minus{}1 into a target register. Since this option is unsafe for any
17682 hardware implementing the current architecture specification, the default
17683 is -mno-pt-fixed. Unless the user specifies a specific cost with
17684 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17685 this deters register allocation using target registers for storing
17688 @item -minvalid-symbols
17689 @opindex minvalid-symbols
17690 Assume symbols might be invalid. Ordinary function symbols generated by
17691 the compiler will always be valid to load with movi/shori/ptabs or
17692 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17693 to generate symbols that will cause ptabs / ptrel to trap.
17694 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17695 It will then prevent cross-basic-block cse, hoisting and most scheduling
17696 of symbol loads. The default is @option{-mno-invalid-symbols}.
17699 @node Solaris 2 Options
17700 @subsection Solaris 2 Options
17701 @cindex Solaris 2 options
17703 These @samp{-m} options are supported on Solaris 2:
17706 @item -mimpure-text
17707 @opindex mimpure-text
17708 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17709 the compiler to not pass @option{-z text} to the linker when linking a
17710 shared object. Using this option, you can link position-dependent
17711 code into a shared object.
17713 @option{-mimpure-text} suppresses the ``relocations remain against
17714 allocatable but non-writable sections'' linker error message.
17715 However, the necessary relocations will trigger copy-on-write, and the
17716 shared object is not actually shared across processes. Instead of
17717 using @option{-mimpure-text}, you should compile all source code with
17718 @option{-fpic} or @option{-fPIC}.
17722 These switches are supported in addition to the above on Solaris 2:
17727 Add support for multithreading using the POSIX threads library. This
17728 option sets flags for both the preprocessor and linker. This option does
17729 not affect the thread safety of object code produced by the compiler or
17730 that of libraries supplied with it.
17734 This is a synonym for @option{-pthreads}.
17737 @node SPARC Options
17738 @subsection SPARC Options
17739 @cindex SPARC options
17741 These @samp{-m} options are supported on the SPARC:
17744 @item -mno-app-regs
17746 @opindex mno-app-regs
17748 Specify @option{-mapp-regs} to generate output using the global registers
17749 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17752 To be fully SVR4 ABI compliant at the cost of some performance loss,
17753 specify @option{-mno-app-regs}. You should compile libraries and system
17754 software with this option.
17760 With @option{-mflat}, the compiler does not generate save/restore instructions
17761 and uses a ``flat'' or single register window model. This model is compatible
17762 with the regular register window model. The local registers and the input
17763 registers (0--5) are still treated as ``call-saved'' registers and will be
17764 saved on the stack as needed.
17766 With @option{-mno-flat} (the default), the compiler generates save/restore
17767 instructions (except for leaf functions). This is the normal operating mode.
17770 @itemx -mhard-float
17772 @opindex mhard-float
17773 Generate output containing floating point instructions. This is the
17777 @itemx -msoft-float
17779 @opindex msoft-float
17780 Generate output containing library calls for floating point.
17781 @strong{Warning:} the requisite libraries are not available for all SPARC
17782 targets. Normally the facilities of the machine's usual C compiler are
17783 used, but this cannot be done directly in cross-compilation. You must make
17784 your own arrangements to provide suitable library functions for
17785 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17786 @samp{sparclite-*-*} do provide software floating point support.
17788 @option{-msoft-float} changes the calling convention in the output file;
17789 therefore, it is only useful if you compile @emph{all} of a program with
17790 this option. In particular, you need to compile @file{libgcc.a}, the
17791 library that comes with GCC, with @option{-msoft-float} in order for
17794 @item -mhard-quad-float
17795 @opindex mhard-quad-float
17796 Generate output containing quad-word (long double) floating point
17799 @item -msoft-quad-float
17800 @opindex msoft-quad-float
17801 Generate output containing library calls for quad-word (long double)
17802 floating point instructions. The functions called are those specified
17803 in the SPARC ABI@. This is the default.
17805 As of this writing, there are no SPARC implementations that have hardware
17806 support for the quad-word floating point instructions. They all invoke
17807 a trap handler for one of these instructions, and then the trap handler
17808 emulates the effect of the instruction. Because of the trap handler overhead,
17809 this is much slower than calling the ABI library routines. Thus the
17810 @option{-msoft-quad-float} option is the default.
17812 @item -mno-unaligned-doubles
17813 @itemx -munaligned-doubles
17814 @opindex mno-unaligned-doubles
17815 @opindex munaligned-doubles
17816 Assume that doubles have 8 byte alignment. This is the default.
17818 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17819 alignment only if they are contained in another type, or if they have an
17820 absolute address. Otherwise, it assumes they have 4 byte alignment.
17821 Specifying this option avoids some rare compatibility problems with code
17822 generated by other compilers. It is not the default because it results
17823 in a performance loss, especially for floating point code.
17825 @item -mno-faster-structs
17826 @itemx -mfaster-structs
17827 @opindex mno-faster-structs
17828 @opindex mfaster-structs
17829 With @option{-mfaster-structs}, the compiler assumes that structures
17830 should have 8 byte alignment. This enables the use of pairs of
17831 @code{ldd} and @code{std} instructions for copies in structure
17832 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17833 However, the use of this changed alignment directly violates the SPARC
17834 ABI@. Thus, it's intended only for use on targets where the developer
17835 acknowledges that their resulting code will not be directly in line with
17836 the rules of the ABI@.
17838 @item -mcpu=@var{cpu_type}
17840 Set the instruction set, register set, and instruction scheduling parameters
17841 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17842 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17843 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17844 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17845 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
17846 and @samp{niagara4}.
17848 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
17849 which selects the best architecture option for the host processor.
17850 @option{-mcpu=native} has no effect if GCC does not recognize
17853 Default instruction scheduling parameters are used for values that select
17854 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17855 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17857 Here is a list of each supported architecture and their supported
17865 supersparc, hypersparc, leon
17868 f930, f934, sparclite86x
17874 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
17877 By default (unless configured otherwise), GCC generates code for the V7
17878 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17879 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17880 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17881 SPARCStation 1, 2, IPX etc.
17883 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17884 architecture. The only difference from V7 code is that the compiler emits
17885 the integer multiply and integer divide instructions which exist in SPARC-V8
17886 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17887 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17890 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17891 the SPARC architecture. This adds the integer multiply, integer divide step
17892 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17893 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17894 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17895 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17896 MB86934 chip, which is the more recent SPARClite with FPU@.
17898 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17899 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17900 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17901 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17902 optimizes it for the TEMIC SPARClet chip.
17904 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17905 architecture. This adds 64-bit integer and floating-point move instructions,
17906 3 additional floating-point condition code registers and conditional move
17907 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17908 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17909 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17910 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17911 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17912 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17913 additionally optimizes it for Sun UltraSPARC T2 chips. With
17914 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
17915 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
17916 additionally optimizes it for Sun UltraSPARC T4 chips.
17918 @item -mtune=@var{cpu_type}
17920 Set the instruction scheduling parameters for machine type
17921 @var{cpu_type}, but do not set the instruction set or register set that the
17922 option @option{-mcpu=@var{cpu_type}} would.
17924 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17925 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17926 that select a particular CPU implementation. Those are @samp{cypress},
17927 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17928 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17929 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
17930 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
17935 @opindex mno-v8plus
17936 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17937 difference from the V8 ABI is that the global and out registers are
17938 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17939 mode for all SPARC-V9 processors.
17945 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17946 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17952 With @option{-mvis2}, GCC generates code that takes advantage of
17953 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
17954 default is @option{-mvis2} when targetting a cpu that supports such
17955 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
17956 also sets @option{-mvis}.
17962 With @option{-mvis3}, GCC generates code that takes advantage of
17963 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
17964 default is @option{-mvis3} when targetting a cpu that supports such
17965 instructions, such as niagara-3 and later. Setting @option{-mvis3}
17966 also sets @option{-mvis2} and @option{-mvis}.
17972 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
17973 population count instruction. The default is @option{-mpopc}
17974 when targetting a cpu that supports such instructions, such as Niagara-2 and
17981 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
17982 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
17983 when targetting a cpu that supports such instructions, such as Niagara-3 and
17987 @opindex mfix-at697f
17988 Enable the documented workaround for the single erratum of the Atmel AT697F
17989 processor (which corresponds to erratum #13 of the AT697E processor).
17992 These @samp{-m} options are supported in addition to the above
17993 on SPARC-V9 processors in 64-bit environments:
17996 @item -mlittle-endian
17997 @opindex mlittle-endian
17998 Generate code for a processor running in little-endian mode. It is only
17999 available for a few configurations and most notably not on Solaris and Linux.
18005 Generate code for a 32-bit or 64-bit environment.
18006 The 32-bit environment sets int, long and pointer to 32 bits.
18007 The 64-bit environment sets int to 32 bits and long and pointer
18010 @item -mcmodel=@var{which}
18012 Set the code model to one of
18016 The Medium/Low code model: 64-bit addresses, programs
18017 must be linked in the low 32 bits of memory. Programs can be statically
18018 or dynamically linked.
18021 The Medium/Middle code model: 64-bit addresses, programs
18022 must be linked in the low 44 bits of memory, the text and data segments must
18023 be less than 2GB in size and the data segment must be located within 2GB of
18027 The Medium/Anywhere code model: 64-bit addresses, programs
18028 may be linked anywhere in memory, the text and data segments must be less
18029 than 2GB in size and the data segment must be located within 2GB of the
18033 The Medium/Anywhere code model for embedded systems:
18034 64-bit addresses, the text and data segments must be less than 2GB in
18035 size, both starting anywhere in memory (determined at link time). The
18036 global register %g4 points to the base of the data segment. Programs
18037 are statically linked and PIC is not supported.
18040 @item -mmemory-model=@var{mem-model}
18041 @opindex mmemory-model
18042 Set the memory model in force on the processor to one of
18046 The default memory model for the processor and operating system.
18049 Relaxed Memory Order
18052 Partial Store Order
18058 Sequential Consistency
18061 These memory models are formally defined in Appendix D of the Sparc V9
18062 architecture manual, as set in the processor's @code{PSTATE.MM} field.
18065 @itemx -mno-stack-bias
18066 @opindex mstack-bias
18067 @opindex mno-stack-bias
18068 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
18069 frame pointer if present, are offset by @minus{}2047 which must be added back
18070 when making stack frame references. This is the default in 64-bit mode.
18071 Otherwise, assume no such offset is present.
18075 @subsection SPU Options
18076 @cindex SPU options
18078 These @samp{-m} options are supported on the SPU:
18082 @itemx -merror-reloc
18083 @opindex mwarn-reloc
18084 @opindex merror-reloc
18086 The loader for SPU does not handle dynamic relocations. By default, GCC
18087 will give an error when it generates code that requires a dynamic
18088 relocation. @option{-mno-error-reloc} disables the error,
18089 @option{-mwarn-reloc} will generate a warning instead.
18092 @itemx -munsafe-dma
18094 @opindex munsafe-dma
18096 Instructions which initiate or test completion of DMA must not be
18097 reordered with respect to loads and stores of the memory which is being
18098 accessed. Users typically address this problem using the volatile
18099 keyword, but that can lead to inefficient code in places where the
18100 memory is known to not change. Rather than mark the memory as volatile
18101 we treat the DMA instructions as potentially effecting all memory. With
18102 @option{-munsafe-dma} users must use the volatile keyword to protect
18105 @item -mbranch-hints
18106 @opindex mbranch-hints
18108 By default, GCC will generate a branch hint instruction to avoid
18109 pipeline stalls for always taken or probably taken branches. A hint
18110 will not be generated closer than 8 instructions away from its branch.
18111 There is little reason to disable them, except for debugging purposes,
18112 or to make an object a little bit smaller.
18116 @opindex msmall-mem
18117 @opindex mlarge-mem
18119 By default, GCC generates code assuming that addresses are never larger
18120 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
18121 a full 32 bit address.
18126 By default, GCC links against startup code that assumes the SPU-style
18127 main function interface (which has an unconventional parameter list).
18128 With @option{-mstdmain}, GCC will link your program against startup
18129 code that assumes a C99-style interface to @code{main}, including a
18130 local copy of @code{argv} strings.
18132 @item -mfixed-range=@var{register-range}
18133 @opindex mfixed-range
18134 Generate code treating the given register range as fixed registers.
18135 A fixed register is one that the register allocator can not use. This is
18136 useful when compiling kernel code. A register range is specified as
18137 two registers separated by a dash. Multiple register ranges can be
18138 specified separated by a comma.
18144 Compile code assuming that pointers to the PPU address space accessed
18145 via the @code{__ea} named address space qualifier are either 32 or 64
18146 bits wide. The default is 32 bits. As this is an ABI changing option,
18147 all object code in an executable must be compiled with the same setting.
18149 @item -maddress-space-conversion
18150 @itemx -mno-address-space-conversion
18151 @opindex maddress-space-conversion
18152 @opindex mno-address-space-conversion
18153 Allow/disallow treating the @code{__ea} address space as superset
18154 of the generic address space. This enables explicit type casts
18155 between @code{__ea} and generic pointer as well as implicit
18156 conversions of generic pointers to @code{__ea} pointers. The
18157 default is to allow address space pointer conversions.
18159 @item -mcache-size=@var{cache-size}
18160 @opindex mcache-size
18161 This option controls the version of libgcc that the compiler links to an
18162 executable and selects a software-managed cache for accessing variables
18163 in the @code{__ea} address space with a particular cache size. Possible
18164 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
18165 and @samp{128}. The default cache size is 64KB.
18167 @item -matomic-updates
18168 @itemx -mno-atomic-updates
18169 @opindex matomic-updates
18170 @opindex mno-atomic-updates
18171 This option controls the version of libgcc that the compiler links to an
18172 executable and selects whether atomic updates to the software-managed
18173 cache of PPU-side variables are used. If you use atomic updates, changes
18174 to a PPU variable from SPU code using the @code{__ea} named address space
18175 qualifier will not interfere with changes to other PPU variables residing
18176 in the same cache line from PPU code. If you do not use atomic updates,
18177 such interference may occur; however, writing back cache lines will be
18178 more efficient. The default behavior is to use atomic updates.
18181 @itemx -mdual-nops=@var{n}
18182 @opindex mdual-nops
18183 By default, GCC will insert nops to increase dual issue when it expects
18184 it to increase performance. @var{n} can be a value from 0 to 10. A
18185 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
18186 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
18188 @item -mhint-max-nops=@var{n}
18189 @opindex mhint-max-nops
18190 Maximum number of nops to insert for a branch hint. A branch hint must
18191 be at least 8 instructions away from the branch it is effecting. GCC
18192 will insert up to @var{n} nops to enforce this, otherwise it will not
18193 generate the branch hint.
18195 @item -mhint-max-distance=@var{n}
18196 @opindex mhint-max-distance
18197 The encoding of the branch hint instruction limits the hint to be within
18198 256 instructions of the branch it is effecting. By default, GCC makes
18199 sure it is within 125.
18202 @opindex msafe-hints
18203 Work around a hardware bug which causes the SPU to stall indefinitely.
18204 By default, GCC will insert the @code{hbrp} instruction to make sure
18205 this stall won't happen.
18209 @node System V Options
18210 @subsection Options for System V
18212 These additional options are available on System V Release 4 for
18213 compatibility with other compilers on those systems:
18218 Create a shared object.
18219 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
18223 Identify the versions of each tool used by the compiler, in a
18224 @code{.ident} assembler directive in the output.
18228 Refrain from adding @code{.ident} directives to the output file (this is
18231 @item -YP,@var{dirs}
18233 Search the directories @var{dirs}, and no others, for libraries
18234 specified with @option{-l}.
18236 @item -Ym,@var{dir}
18238 Look in the directory @var{dir} to find the M4 preprocessor.
18239 The assembler uses this option.
18240 @c This is supposed to go with a -Yd for predefined M4 macro files, but
18241 @c the generic assembler that comes with Solaris takes just -Ym.
18245 @subsection V850 Options
18246 @cindex V850 Options
18248 These @samp{-m} options are defined for V850 implementations:
18252 @itemx -mno-long-calls
18253 @opindex mlong-calls
18254 @opindex mno-long-calls
18255 Treat all calls as being far away (near). If calls are assumed to be
18256 far away, the compiler will always load the functions address up into a
18257 register, and call indirect through the pointer.
18263 Do not optimize (do optimize) basic blocks that use the same index
18264 pointer 4 or more times to copy pointer into the @code{ep} register, and
18265 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
18266 option is on by default if you optimize.
18268 @item -mno-prolog-function
18269 @itemx -mprolog-function
18270 @opindex mno-prolog-function
18271 @opindex mprolog-function
18272 Do not use (do use) external functions to save and restore registers
18273 at the prologue and epilogue of a function. The external functions
18274 are slower, but use less code space if more than one function saves
18275 the same number of registers. The @option{-mprolog-function} option
18276 is on by default if you optimize.
18280 Try to make the code as small as possible. At present, this just turns
18281 on the @option{-mep} and @option{-mprolog-function} options.
18283 @item -mtda=@var{n}
18285 Put static or global variables whose size is @var{n} bytes or less into
18286 the tiny data area that register @code{ep} points to. The tiny data
18287 area can hold up to 256 bytes in total (128 bytes for byte references).
18289 @item -msda=@var{n}
18291 Put static or global variables whose size is @var{n} bytes or less into
18292 the small data area that register @code{gp} points to. The small data
18293 area can hold up to 64 kilobytes.
18295 @item -mzda=@var{n}
18297 Put static or global variables whose size is @var{n} bytes or less into
18298 the first 32 kilobytes of memory.
18302 Specify that the target processor is the V850.
18305 @opindex mbig-switch
18306 Generate code suitable for big switch tables. Use this option only if
18307 the assembler/linker complain about out of range branches within a switch
18312 This option will cause r2 and r5 to be used in the code generated by
18313 the compiler. This setting is the default.
18315 @item -mno-app-regs
18316 @opindex mno-app-regs
18317 This option will cause r2 and r5 to be treated as fixed registers.
18321 Specify that the target processor is the V850E2V3. The preprocessor
18322 constants @samp{__v850e2v3__} will be defined if
18323 this option is used.
18327 Specify that the target processor is the V850E2. The preprocessor
18328 constants @samp{__v850e2__} will be defined if this option is used.
18332 Specify that the target processor is the V850E1. The preprocessor
18333 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
18334 this option is used.
18338 Specify that the target processor is the V850ES. This is an alias for
18339 the @option{-mv850e1} option.
18343 Specify that the target processor is the V850E@. The preprocessor
18344 constant @samp{__v850e__} will be defined if this option is used.
18346 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
18347 nor @option{-mv850e2} nor @option{-mv850e2v3}
18348 are defined then a default target processor will be chosen and the
18349 relevant @samp{__v850*__} preprocessor constant will be defined.
18351 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
18352 defined, regardless of which processor variant is the target.
18354 @item -mdisable-callt
18355 @opindex mdisable-callt
18356 This option will suppress generation of the CALLT instruction for the
18357 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
18358 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
18363 @subsection VAX Options
18364 @cindex VAX options
18366 These @samp{-m} options are defined for the VAX:
18371 Do not output certain jump instructions (@code{aobleq} and so on)
18372 that the Unix assembler for the VAX cannot handle across long
18377 Do output those jump instructions, on the assumption that you
18378 will assemble with the GNU assembler.
18382 Output code for g-format floating point numbers instead of d-format.
18385 @node VxWorks Options
18386 @subsection VxWorks Options
18387 @cindex VxWorks Options
18389 The options in this section are defined for all VxWorks targets.
18390 Options specific to the target hardware are listed with the other
18391 options for that target.
18396 GCC can generate code for both VxWorks kernels and real time processes
18397 (RTPs). This option switches from the former to the latter. It also
18398 defines the preprocessor macro @code{__RTP__}.
18401 @opindex non-static
18402 Link an RTP executable against shared libraries rather than static
18403 libraries. The options @option{-static} and @option{-shared} can
18404 also be used for RTPs (@pxref{Link Options}); @option{-static}
18411 These options are passed down to the linker. They are defined for
18412 compatibility with Diab.
18415 @opindex Xbind-lazy
18416 Enable lazy binding of function calls. This option is equivalent to
18417 @option{-Wl,-z,now} and is defined for compatibility with Diab.
18421 Disable lazy binding of function calls. This option is the default and
18422 is defined for compatibility with Diab.
18425 @node x86-64 Options
18426 @subsection x86-64 Options
18427 @cindex x86-64 options
18429 These are listed under @xref{i386 and x86-64 Options}.
18431 @node Xstormy16 Options
18432 @subsection Xstormy16 Options
18433 @cindex Xstormy16 Options
18435 These options are defined for Xstormy16:
18440 Choose startup files and linker script suitable for the simulator.
18443 @node Xtensa Options
18444 @subsection Xtensa Options
18445 @cindex Xtensa Options
18447 These options are supported for Xtensa targets:
18451 @itemx -mno-const16
18453 @opindex mno-const16
18454 Enable or disable use of @code{CONST16} instructions for loading
18455 constant values. The @code{CONST16} instruction is currently not a
18456 standard option from Tensilica. When enabled, @code{CONST16}
18457 instructions are always used in place of the standard @code{L32R}
18458 instructions. The use of @code{CONST16} is enabled by default only if
18459 the @code{L32R} instruction is not available.
18462 @itemx -mno-fused-madd
18463 @opindex mfused-madd
18464 @opindex mno-fused-madd
18465 Enable or disable use of fused multiply/add and multiply/subtract
18466 instructions in the floating-point option. This has no effect if the
18467 floating-point option is not also enabled. Disabling fused multiply/add
18468 and multiply/subtract instructions forces the compiler to use separate
18469 instructions for the multiply and add/subtract operations. This may be
18470 desirable in some cases where strict IEEE 754-compliant results are
18471 required: the fused multiply add/subtract instructions do not round the
18472 intermediate result, thereby producing results with @emph{more} bits of
18473 precision than specified by the IEEE standard. Disabling fused multiply
18474 add/subtract instructions also ensures that the program output is not
18475 sensitive to the compiler's ability to combine multiply and add/subtract
18478 @item -mserialize-volatile
18479 @itemx -mno-serialize-volatile
18480 @opindex mserialize-volatile
18481 @opindex mno-serialize-volatile
18482 When this option is enabled, GCC inserts @code{MEMW} instructions before
18483 @code{volatile} memory references to guarantee sequential consistency.
18484 The default is @option{-mserialize-volatile}. Use
18485 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
18487 @item -mforce-no-pic
18488 @opindex mforce-no-pic
18489 For targets, like GNU/Linux, where all user-mode Xtensa code must be
18490 position-independent code (PIC), this option disables PIC for compiling
18493 @item -mtext-section-literals
18494 @itemx -mno-text-section-literals
18495 @opindex mtext-section-literals
18496 @opindex mno-text-section-literals
18497 Control the treatment of literal pools. The default is
18498 @option{-mno-text-section-literals}, which places literals in a separate
18499 section in the output file. This allows the literal pool to be placed
18500 in a data RAM/ROM, and it also allows the linker to combine literal
18501 pools from separate object files to remove redundant literals and
18502 improve code size. With @option{-mtext-section-literals}, the literals
18503 are interspersed in the text section in order to keep them as close as
18504 possible to their references. This may be necessary for large assembly
18507 @item -mtarget-align
18508 @itemx -mno-target-align
18509 @opindex mtarget-align
18510 @opindex mno-target-align
18511 When this option is enabled, GCC instructs the assembler to
18512 automatically align instructions to reduce branch penalties at the
18513 expense of some code density. The assembler attempts to widen density
18514 instructions to align branch targets and the instructions following call
18515 instructions. If there are not enough preceding safe density
18516 instructions to align a target, no widening will be performed. The
18517 default is @option{-mtarget-align}. These options do not affect the
18518 treatment of auto-aligned instructions like @code{LOOP}, which the
18519 assembler will always align, either by widening density instructions or
18520 by inserting no-op instructions.
18523 @itemx -mno-longcalls
18524 @opindex mlongcalls
18525 @opindex mno-longcalls
18526 When this option is enabled, GCC instructs the assembler to translate
18527 direct calls to indirect calls unless it can determine that the target
18528 of a direct call is in the range allowed by the call instruction. This
18529 translation typically occurs for calls to functions in other source
18530 files. Specifically, the assembler translates a direct @code{CALL}
18531 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
18532 The default is @option{-mno-longcalls}. This option should be used in
18533 programs where the call target can potentially be out of range. This
18534 option is implemented in the assembler, not the compiler, so the
18535 assembly code generated by GCC will still show direct call
18536 instructions---look at the disassembled object code to see the actual
18537 instructions. Note that the assembler will use an indirect call for
18538 every cross-file call, not just those that really will be out of range.
18541 @node zSeries Options
18542 @subsection zSeries Options
18543 @cindex zSeries options
18545 These are listed under @xref{S/390 and zSeries Options}.
18547 @node Code Gen Options
18548 @section Options for Code Generation Conventions
18549 @cindex code generation conventions
18550 @cindex options, code generation
18551 @cindex run-time options
18553 These machine-independent options control the interface conventions
18554 used in code generation.
18556 Most of them have both positive and negative forms; the negative form
18557 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
18558 one of the forms is listed---the one which is not the default. You
18559 can figure out the other form by either removing @samp{no-} or adding
18563 @item -fbounds-check
18564 @opindex fbounds-check
18565 For front-ends that support it, generate additional code to check that
18566 indices used to access arrays are within the declared range. This is
18567 currently only supported by the Java and Fortran front-ends, where
18568 this option defaults to true and false respectively.
18572 This option generates traps for signed overflow on addition, subtraction,
18573 multiplication operations.
18577 This option instructs the compiler to assume that signed arithmetic
18578 overflow of addition, subtraction and multiplication wraps around
18579 using twos-complement representation. This flag enables some optimizations
18580 and disables others. This option is enabled by default for the Java
18581 front-end, as required by the Java language specification.
18584 @opindex fexceptions
18585 Enable exception handling. Generates extra code needed to propagate
18586 exceptions. For some targets, this implies GCC will generate frame
18587 unwind information for all functions, which can produce significant data
18588 size overhead, although it does not affect execution. If you do not
18589 specify this option, GCC will enable it by default for languages like
18590 C++ which normally require exception handling, and disable it for
18591 languages like C that do not normally require it. However, you may need
18592 to enable this option when compiling C code that needs to interoperate
18593 properly with exception handlers written in C++. You may also wish to
18594 disable this option if you are compiling older C++ programs that don't
18595 use exception handling.
18597 @item -fnon-call-exceptions
18598 @opindex fnon-call-exceptions
18599 Generate code that allows trapping instructions to throw exceptions.
18600 Note that this requires platform-specific runtime support that does
18601 not exist everywhere. Moreover, it only allows @emph{trapping}
18602 instructions to throw exceptions, i.e.@: memory references or floating
18603 point instructions. It does not allow exceptions to be thrown from
18604 arbitrary signal handlers such as @code{SIGALRM}.
18606 @item -funwind-tables
18607 @opindex funwind-tables
18608 Similar to @option{-fexceptions}, except that it will just generate any needed
18609 static data, but will not affect the generated code in any other way.
18610 You will normally not enable this option; instead, a language processor
18611 that needs this handling would enable it on your behalf.
18613 @item -fasynchronous-unwind-tables
18614 @opindex fasynchronous-unwind-tables
18615 Generate unwind table in dwarf2 format, if supported by target machine. The
18616 table is exact at each instruction boundary, so it can be used for stack
18617 unwinding from asynchronous events (such as debugger or garbage collector).
18619 @item -fpcc-struct-return
18620 @opindex fpcc-struct-return
18621 Return ``short'' @code{struct} and @code{union} values in memory like
18622 longer ones, rather than in registers. This convention is less
18623 efficient, but it has the advantage of allowing intercallability between
18624 GCC-compiled files and files compiled with other compilers, particularly
18625 the Portable C Compiler (pcc).
18627 The precise convention for returning structures in memory depends
18628 on the target configuration macros.
18630 Short structures and unions are those whose size and alignment match
18631 that of some integer type.
18633 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
18634 switch is not binary compatible with code compiled with the
18635 @option{-freg-struct-return} switch.
18636 Use it to conform to a non-default application binary interface.
18638 @item -freg-struct-return
18639 @opindex freg-struct-return
18640 Return @code{struct} and @code{union} values in registers when possible.
18641 This is more efficient for small structures than
18642 @option{-fpcc-struct-return}.
18644 If you specify neither @option{-fpcc-struct-return} nor
18645 @option{-freg-struct-return}, GCC defaults to whichever convention is
18646 standard for the target. If there is no standard convention, GCC
18647 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18648 the principal compiler. In those cases, we can choose the standard, and
18649 we chose the more efficient register return alternative.
18651 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18652 switch is not binary compatible with code compiled with the
18653 @option{-fpcc-struct-return} switch.
18654 Use it to conform to a non-default application binary interface.
18656 @item -fshort-enums
18657 @opindex fshort-enums
18658 Allocate to an @code{enum} type only as many bytes as it needs for the
18659 declared range of possible values. Specifically, the @code{enum} type
18660 will be equivalent to the smallest integer type which has enough room.
18662 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18663 code that is not binary compatible with code generated without that switch.
18664 Use it to conform to a non-default application binary interface.
18666 @item -fshort-double
18667 @opindex fshort-double
18668 Use the same size for @code{double} as for @code{float}.
18670 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18671 code that is not binary compatible with code generated without that switch.
18672 Use it to conform to a non-default application binary interface.
18674 @item -fshort-wchar
18675 @opindex fshort-wchar
18676 Override the underlying type for @samp{wchar_t} to be @samp{short
18677 unsigned int} instead of the default for the target. This option is
18678 useful for building programs to run under WINE@.
18680 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18681 code that is not binary compatible with code generated without that switch.
18682 Use it to conform to a non-default application binary interface.
18685 @opindex fno-common
18686 In C code, controls the placement of uninitialized global variables.
18687 Unix C compilers have traditionally permitted multiple definitions of
18688 such variables in different compilation units by placing the variables
18690 This is the behavior specified by @option{-fcommon}, and is the default
18691 for GCC on most targets.
18692 On the other hand, this behavior is not required by ISO C, and on some
18693 targets may carry a speed or code size penalty on variable references.
18694 The @option{-fno-common} option specifies that the compiler should place
18695 uninitialized global variables in the data section of the object file,
18696 rather than generating them as common blocks.
18697 This has the effect that if the same variable is declared
18698 (without @code{extern}) in two different compilations,
18699 you will get a multiple-definition error when you link them.
18700 In this case, you must compile with @option{-fcommon} instead.
18701 Compiling with @option{-fno-common} is useful on targets for which
18702 it provides better performance, or if you wish to verify that the
18703 program will work on other systems which always treat uninitialized
18704 variable declarations this way.
18708 Ignore the @samp{#ident} directive.
18710 @item -finhibit-size-directive
18711 @opindex finhibit-size-directive
18712 Don't output a @code{.size} assembler directive, or anything else that
18713 would cause trouble if the function is split in the middle, and the
18714 two halves are placed at locations far apart in memory. This option is
18715 used when compiling @file{crtstuff.c}; you should not need to use it
18718 @item -fverbose-asm
18719 @opindex fverbose-asm
18720 Put extra commentary information in the generated assembly code to
18721 make it more readable. This option is generally only of use to those
18722 who actually need to read the generated assembly code (perhaps while
18723 debugging the compiler itself).
18725 @option{-fno-verbose-asm}, the default, causes the
18726 extra information to be omitted and is useful when comparing two assembler
18729 @item -frecord-gcc-switches
18730 @opindex frecord-gcc-switches
18731 This switch causes the command line that was used to invoke the
18732 compiler to be recorded into the object file that is being created.
18733 This switch is only implemented on some targets and the exact format
18734 of the recording is target and binary file format dependent, but it
18735 usually takes the form of a section containing ASCII text. This
18736 switch is related to the @option{-fverbose-asm} switch, but that
18737 switch only records information in the assembler output file as
18738 comments, so it never reaches the object file.
18739 See also @option{-grecord-gcc-switches} for another
18740 way of storing compiler options into the object file.
18744 @cindex global offset table
18746 Generate position-independent code (PIC) suitable for use in a shared
18747 library, if supported for the target machine. Such code accesses all
18748 constant addresses through a global offset table (GOT)@. The dynamic
18749 loader resolves the GOT entries when the program starts (the dynamic
18750 loader is not part of GCC; it is part of the operating system). If
18751 the GOT size for the linked executable exceeds a machine-specific
18752 maximum size, you get an error message from the linker indicating that
18753 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18754 instead. (These maximums are 8k on the SPARC and 32k
18755 on the m68k and RS/6000. The 386 has no such limit.)
18757 Position-independent code requires special support, and therefore works
18758 only on certain machines. For the 386, GCC supports PIC for System V
18759 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18760 position-independent.
18762 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18767 If supported for the target machine, emit position-independent code,
18768 suitable for dynamic linking and avoiding any limit on the size of the
18769 global offset table. This option makes a difference on the m68k,
18770 PowerPC and SPARC@.
18772 Position-independent code requires special support, and therefore works
18773 only on certain machines.
18775 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18782 These options are similar to @option{-fpic} and @option{-fPIC}, but
18783 generated position independent code can be only linked into executables.
18784 Usually these options are used when @option{-pie} GCC option will be
18785 used during linking.
18787 @option{-fpie} and @option{-fPIE} both define the macros
18788 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18789 for @option{-fpie} and 2 for @option{-fPIE}.
18791 @item -fno-jump-tables
18792 @opindex fno-jump-tables
18793 Do not use jump tables for switch statements even where it would be
18794 more efficient than other code generation strategies. This option is
18795 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18796 building code which forms part of a dynamic linker and cannot
18797 reference the address of a jump table. On some targets, jump tables
18798 do not require a GOT and this option is not needed.
18800 @item -ffixed-@var{reg}
18802 Treat the register named @var{reg} as a fixed register; generated code
18803 should never refer to it (except perhaps as a stack pointer, frame
18804 pointer or in some other fixed role).
18806 @var{reg} must be the name of a register. The register names accepted
18807 are machine-specific and are defined in the @code{REGISTER_NAMES}
18808 macro in the machine description macro file.
18810 This flag does not have a negative form, because it specifies a
18813 @item -fcall-used-@var{reg}
18814 @opindex fcall-used
18815 Treat the register named @var{reg} as an allocable register that is
18816 clobbered by function calls. It may be allocated for temporaries or
18817 variables that do not live across a call. Functions compiled this way
18818 will not save and restore the register @var{reg}.
18820 It is an error to used this flag with the frame pointer or stack pointer.
18821 Use of this flag for other registers that have fixed pervasive roles in
18822 the machine's execution model will produce disastrous results.
18824 This flag does not have a negative form, because it specifies a
18827 @item -fcall-saved-@var{reg}
18828 @opindex fcall-saved
18829 Treat the register named @var{reg} as an allocable register saved by
18830 functions. It may be allocated even for temporaries or variables that
18831 live across a call. Functions compiled this way will save and restore
18832 the register @var{reg} if they use it.
18834 It is an error to used this flag with the frame pointer or stack pointer.
18835 Use of this flag for other registers that have fixed pervasive roles in
18836 the machine's execution model will produce disastrous results.
18838 A different sort of disaster will result from the use of this flag for
18839 a register in which function values may be returned.
18841 This flag does not have a negative form, because it specifies a
18844 @item -fpack-struct[=@var{n}]
18845 @opindex fpack-struct
18846 Without a value specified, pack all structure members together without
18847 holes. When a value is specified (which must be a small power of two), pack
18848 structure members according to this value, representing the maximum
18849 alignment (that is, objects with default alignment requirements larger than
18850 this will be output potentially unaligned at the next fitting location.
18852 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18853 code that is not binary compatible with code generated without that switch.
18854 Additionally, it makes the code suboptimal.
18855 Use it to conform to a non-default application binary interface.
18857 @item -finstrument-functions
18858 @opindex finstrument-functions
18859 Generate instrumentation calls for entry and exit to functions. Just
18860 after function entry and just before function exit, the following
18861 profiling functions will be called with the address of the current
18862 function and its call site. (On some platforms,
18863 @code{__builtin_return_address} does not work beyond the current
18864 function, so the call site information may not be available to the
18865 profiling functions otherwise.)
18868 void __cyg_profile_func_enter (void *this_fn,
18870 void __cyg_profile_func_exit (void *this_fn,
18874 The first argument is the address of the start of the current function,
18875 which may be looked up exactly in the symbol table.
18877 This instrumentation is also done for functions expanded inline in other
18878 functions. The profiling calls will indicate where, conceptually, the
18879 inline function is entered and exited. This means that addressable
18880 versions of such functions must be available. If all your uses of a
18881 function are expanded inline, this may mean an additional expansion of
18882 code size. If you use @samp{extern inline} in your C code, an
18883 addressable version of such functions must be provided. (This is
18884 normally the case anyways, but if you get lucky and the optimizer always
18885 expands the functions inline, you might have gotten away without
18886 providing static copies.)
18888 A function may be given the attribute @code{no_instrument_function}, in
18889 which case this instrumentation will not be done. This can be used, for
18890 example, for the profiling functions listed above, high-priority
18891 interrupt routines, and any functions from which the profiling functions
18892 cannot safely be called (perhaps signal handlers, if the profiling
18893 routines generate output or allocate memory).
18895 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18896 @opindex finstrument-functions-exclude-file-list
18898 Set the list of functions that are excluded from instrumentation (see
18899 the description of @code{-finstrument-functions}). If the file that
18900 contains a function definition matches with one of @var{file}, then
18901 that function is not instrumented. The match is done on substrings:
18902 if the @var{file} parameter is a substring of the file name, it is
18903 considered to be a match.
18908 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18912 will exclude any inline function defined in files whose pathnames
18913 contain @code{/bits/stl} or @code{include/sys}.
18915 If, for some reason, you want to include letter @code{','} in one of
18916 @var{sym}, write @code{'\,'}. For example,
18917 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18918 (note the single quote surrounding the option).
18920 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18921 @opindex finstrument-functions-exclude-function-list
18923 This is similar to @code{-finstrument-functions-exclude-file-list},
18924 but this option sets the list of function names to be excluded from
18925 instrumentation. The function name to be matched is its user-visible
18926 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18927 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18928 match is done on substrings: if the @var{sym} parameter is a substring
18929 of the function name, it is considered to be a match. For C99 and C++
18930 extended identifiers, the function name must be given in UTF-8, not
18931 using universal character names.
18933 @item -fstack-check
18934 @opindex fstack-check
18935 Generate code to verify that you do not go beyond the boundary of the
18936 stack. You should specify this flag if you are running in an
18937 environment with multiple threads, but only rarely need to specify it in
18938 a single-threaded environment since stack overflow is automatically
18939 detected on nearly all systems if there is only one stack.
18941 Note that this switch does not actually cause checking to be done; the
18942 operating system or the language runtime must do that. The switch causes
18943 generation of code to ensure that they see the stack being extended.
18945 You can additionally specify a string parameter: @code{no} means no
18946 checking, @code{generic} means force the use of old-style checking,
18947 @code{specific} means use the best checking method and is equivalent
18948 to bare @option{-fstack-check}.
18950 Old-style checking is a generic mechanism that requires no specific
18951 target support in the compiler but comes with the following drawbacks:
18955 Modified allocation strategy for large objects: they will always be
18956 allocated dynamically if their size exceeds a fixed threshold.
18959 Fixed limit on the size of the static frame of functions: when it is
18960 topped by a particular function, stack checking is not reliable and
18961 a warning is issued by the compiler.
18964 Inefficiency: because of both the modified allocation strategy and the
18965 generic implementation, the performances of the code are hampered.
18968 Note that old-style stack checking is also the fallback method for
18969 @code{specific} if no target support has been added in the compiler.
18971 @item -fstack-limit-register=@var{reg}
18972 @itemx -fstack-limit-symbol=@var{sym}
18973 @itemx -fno-stack-limit
18974 @opindex fstack-limit-register
18975 @opindex fstack-limit-symbol
18976 @opindex fno-stack-limit
18977 Generate code to ensure that the stack does not grow beyond a certain value,
18978 either the value of a register or the address of a symbol. If the stack
18979 would grow beyond the value, a signal is raised. For most targets,
18980 the signal is raised before the stack overruns the boundary, so
18981 it is possible to catch the signal without taking special precautions.
18983 For instance, if the stack starts at absolute address @samp{0x80000000}
18984 and grows downwards, you can use the flags
18985 @option{-fstack-limit-symbol=__stack_limit} and
18986 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18987 of 128KB@. Note that this may only work with the GNU linker.
18989 @item -fsplit-stack
18990 @opindex fsplit-stack
18991 Generate code to automatically split the stack before it overflows.
18992 The resulting program has a discontiguous stack which can only
18993 overflow if the program is unable to allocate any more memory. This
18994 is most useful when running threaded programs, as it is no longer
18995 necessary to calculate a good stack size to use for each thread. This
18996 is currently only implemented for the i386 and x86_64 backends running
18999 When code compiled with @option{-fsplit-stack} calls code compiled
19000 without @option{-fsplit-stack}, there may not be much stack space
19001 available for the latter code to run. If compiling all code,
19002 including library code, with @option{-fsplit-stack} is not an option,
19003 then the linker can fix up these calls so that the code compiled
19004 without @option{-fsplit-stack} always has a large stack. Support for
19005 this is implemented in the gold linker in GNU binutils release 2.21
19008 @item -fleading-underscore
19009 @opindex fleading-underscore
19010 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
19011 change the way C symbols are represented in the object file. One use
19012 is to help link with legacy assembly code.
19014 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
19015 generate code that is not binary compatible with code generated without that
19016 switch. Use it to conform to a non-default application binary interface.
19017 Not all targets provide complete support for this switch.
19019 @item -ftls-model=@var{model}
19020 @opindex ftls-model
19021 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
19022 The @var{model} argument should be one of @code{global-dynamic},
19023 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
19025 The default without @option{-fpic} is @code{initial-exec}; with
19026 @option{-fpic} the default is @code{global-dynamic}.
19028 @item -fvisibility=@var{default|internal|hidden|protected}
19029 @opindex fvisibility
19030 Set the default ELF image symbol visibility to the specified option---all
19031 symbols will be marked with this unless overridden within the code.
19032 Using this feature can very substantially improve linking and
19033 load times of shared object libraries, produce more optimized
19034 code, provide near-perfect API export and prevent symbol clashes.
19035 It is @strong{strongly} recommended that you use this in any shared objects
19038 Despite the nomenclature, @code{default} always means public; i.e.,
19039 available to be linked against from outside the shared object.
19040 @code{protected} and @code{internal} are pretty useless in real-world
19041 usage so the only other commonly used option will be @code{hidden}.
19042 The default if @option{-fvisibility} isn't specified is
19043 @code{default}, i.e., make every
19044 symbol public---this causes the same behavior as previous versions of
19047 A good explanation of the benefits offered by ensuring ELF
19048 symbols have the correct visibility is given by ``How To Write
19049 Shared Libraries'' by Ulrich Drepper (which can be found at
19050 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
19051 solution made possible by this option to marking things hidden when
19052 the default is public is to make the default hidden and mark things
19053 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
19054 and @code{__attribute__ ((visibility("default")))} instead of
19055 @code{__declspec(dllexport)} you get almost identical semantics with
19056 identical syntax. This is a great boon to those working with
19057 cross-platform projects.
19059 For those adding visibility support to existing code, you may find
19060 @samp{#pragma GCC visibility} of use. This works by you enclosing
19061 the declarations you wish to set visibility for with (for example)
19062 @samp{#pragma GCC visibility push(hidden)} and
19063 @samp{#pragma GCC visibility pop}.
19064 Bear in mind that symbol visibility should be viewed @strong{as
19065 part of the API interface contract} and thus all new code should
19066 always specify visibility when it is not the default; i.e., declarations
19067 only for use within the local DSO should @strong{always} be marked explicitly
19068 as hidden as so to avoid PLT indirection overheads---making this
19069 abundantly clear also aids readability and self-documentation of the code.
19070 Note that due to ISO C++ specification requirements, operator new and
19071 operator delete must always be of default visibility.
19073 Be aware that headers from outside your project, in particular system
19074 headers and headers from any other library you use, may not be
19075 expecting to be compiled with visibility other than the default. You
19076 may need to explicitly say @samp{#pragma GCC visibility push(default)}
19077 before including any such headers.
19079 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
19080 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
19081 no modifications. However, this means that calls to @samp{extern}
19082 functions with no explicit visibility will use the PLT, so it is more
19083 effective to use @samp{__attribute ((visibility))} and/or
19084 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
19085 declarations should be treated as hidden.
19087 Note that @samp{-fvisibility} does affect C++ vague linkage
19088 entities. This means that, for instance, an exception class that will
19089 be thrown between DSOs must be explicitly marked with default
19090 visibility so that the @samp{type_info} nodes will be unified between
19093 An overview of these techniques, their benefits and how to use them
19094 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
19096 @item -fstrict-volatile-bitfields
19097 @opindex fstrict-volatile-bitfields
19098 This option should be used if accesses to volatile bitfields (or other
19099 structure fields, although the compiler usually honors those types
19100 anyway) should use a single access of the width of the
19101 field's type, aligned to a natural alignment if possible. For
19102 example, targets with memory-mapped peripheral registers might require
19103 all such accesses to be 16 bits wide; with this flag the user could
19104 declare all peripheral bitfields as ``unsigned short'' (assuming short
19105 is 16 bits on these targets) to force GCC to use 16 bit accesses
19106 instead of, perhaps, a more efficient 32 bit access.
19108 If this option is disabled, the compiler will use the most efficient
19109 instruction. In the previous example, that might be a 32-bit load
19110 instruction, even though that will access bytes that do not contain
19111 any portion of the bitfield, or memory-mapped registers unrelated to
19112 the one being updated.
19114 If the target requires strict alignment, and honoring the field
19115 type would require violating this alignment, a warning is issued.
19116 If the field has @code{packed} attribute, the access is done without
19117 honoring the field type. If the field doesn't have @code{packed}
19118 attribute, the access is done honoring the field type. In both cases,
19119 GCC assumes that the user knows something about the target hardware
19120 that it is unaware of.
19122 The default value of this option is determined by the application binary
19123 interface for the target processor.
19129 @node Environment Variables
19130 @section Environment Variables Affecting GCC
19131 @cindex environment variables
19133 @c man begin ENVIRONMENT
19134 This section describes several environment variables that affect how GCC
19135 operates. Some of them work by specifying directories or prefixes to use
19136 when searching for various kinds of files. Some are used to specify other
19137 aspects of the compilation environment.
19139 Note that you can also specify places to search using options such as
19140 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
19141 take precedence over places specified using environment variables, which
19142 in turn take precedence over those specified by the configuration of GCC@.
19143 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
19144 GNU Compiler Collection (GCC) Internals}.
19149 @c @itemx LC_COLLATE
19151 @c @itemx LC_MONETARY
19152 @c @itemx LC_NUMERIC
19157 @c @findex LC_COLLATE
19158 @findex LC_MESSAGES
19159 @c @findex LC_MONETARY
19160 @c @findex LC_NUMERIC
19164 These environment variables control the way that GCC uses
19165 localization information that allow GCC to work with different
19166 national conventions. GCC inspects the locale categories
19167 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
19168 so. These locale categories can be set to any value supported by your
19169 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
19170 Kingdom encoded in UTF-8.
19172 The @env{LC_CTYPE} environment variable specifies character
19173 classification. GCC uses it to determine the character boundaries in
19174 a string; this is needed for some multibyte encodings that contain quote
19175 and escape characters that would otherwise be interpreted as a string
19178 The @env{LC_MESSAGES} environment variable specifies the language to
19179 use in diagnostic messages.
19181 If the @env{LC_ALL} environment variable is set, it overrides the value
19182 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
19183 and @env{LC_MESSAGES} default to the value of the @env{LANG}
19184 environment variable. If none of these variables are set, GCC
19185 defaults to traditional C English behavior.
19189 If @env{TMPDIR} is set, it specifies the directory to use for temporary
19190 files. GCC uses temporary files to hold the output of one stage of
19191 compilation which is to be used as input to the next stage: for example,
19192 the output of the preprocessor, which is the input to the compiler
19195 @item GCC_COMPARE_DEBUG
19196 @findex GCC_COMPARE_DEBUG
19197 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
19198 @option{-fcompare-debug} to the compiler driver. See the documentation
19199 of this option for more details.
19201 @item GCC_EXEC_PREFIX
19202 @findex GCC_EXEC_PREFIX
19203 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
19204 names of the subprograms executed by the compiler. No slash is added
19205 when this prefix is combined with the name of a subprogram, but you can
19206 specify a prefix that ends with a slash if you wish.
19208 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
19209 an appropriate prefix to use based on the pathname it was invoked with.
19211 If GCC cannot find the subprogram using the specified prefix, it
19212 tries looking in the usual places for the subprogram.
19214 The default value of @env{GCC_EXEC_PREFIX} is
19215 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
19216 the installed compiler. In many cases @var{prefix} is the value
19217 of @code{prefix} when you ran the @file{configure} script.
19219 Other prefixes specified with @option{-B} take precedence over this prefix.
19221 This prefix is also used for finding files such as @file{crt0.o} that are
19224 In addition, the prefix is used in an unusual way in finding the
19225 directories to search for header files. For each of the standard
19226 directories whose name normally begins with @samp{/usr/local/lib/gcc}
19227 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
19228 replacing that beginning with the specified prefix to produce an
19229 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
19230 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
19231 These alternate directories are searched first; the standard directories
19232 come next. If a standard directory begins with the configured
19233 @var{prefix} then the value of @var{prefix} is replaced by
19234 @env{GCC_EXEC_PREFIX} when looking for header files.
19236 @item COMPILER_PATH
19237 @findex COMPILER_PATH
19238 The value of @env{COMPILER_PATH} is a colon-separated list of
19239 directories, much like @env{PATH}. GCC tries the directories thus
19240 specified when searching for subprograms, if it can't find the
19241 subprograms using @env{GCC_EXEC_PREFIX}.
19244 @findex LIBRARY_PATH
19245 The value of @env{LIBRARY_PATH} is a colon-separated list of
19246 directories, much like @env{PATH}. When configured as a native compiler,
19247 GCC tries the directories thus specified when searching for special
19248 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
19249 using GCC also uses these directories when searching for ordinary
19250 libraries for the @option{-l} option (but directories specified with
19251 @option{-L} come first).
19255 @cindex locale definition
19256 This variable is used to pass locale information to the compiler. One way in
19257 which this information is used is to determine the character set to be used
19258 when character literals, string literals and comments are parsed in C and C++.
19259 When the compiler is configured to allow multibyte characters,
19260 the following values for @env{LANG} are recognized:
19264 Recognize JIS characters.
19266 Recognize SJIS characters.
19268 Recognize EUCJP characters.
19271 If @env{LANG} is not defined, or if it has some other value, then the
19272 compiler will use mblen and mbtowc as defined by the default locale to
19273 recognize and translate multibyte characters.
19277 Some additional environments variables affect the behavior of the
19280 @include cppenv.texi
19284 @node Precompiled Headers
19285 @section Using Precompiled Headers
19286 @cindex precompiled headers
19287 @cindex speed of compilation
19289 Often large projects have many header files that are included in every
19290 source file. The time the compiler takes to process these header files
19291 over and over again can account for nearly all of the time required to
19292 build the project. To make builds faster, GCC allows users to
19293 `precompile' a header file; then, if builds can use the precompiled
19294 header file they will be much faster.
19296 To create a precompiled header file, simply compile it as you would any
19297 other file, if necessary using the @option{-x} option to make the driver
19298 treat it as a C or C++ header file. You will probably want to use a
19299 tool like @command{make} to keep the precompiled header up-to-date when
19300 the headers it contains change.
19302 A precompiled header file will be searched for when @code{#include} is
19303 seen in the compilation. As it searches for the included file
19304 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
19305 compiler looks for a precompiled header in each directory just before it
19306 looks for the include file in that directory. The name searched for is
19307 the name specified in the @code{#include} with @samp{.gch} appended. If
19308 the precompiled header file can't be used, it is ignored.
19310 For instance, if you have @code{#include "all.h"}, and you have
19311 @file{all.h.gch} in the same directory as @file{all.h}, then the
19312 precompiled header file will be used if possible, and the original
19313 header will be used otherwise.
19315 Alternatively, you might decide to put the precompiled header file in a
19316 directory and use @option{-I} to ensure that directory is searched
19317 before (or instead of) the directory containing the original header.
19318 Then, if you want to check that the precompiled header file is always
19319 used, you can put a file of the same name as the original header in this
19320 directory containing an @code{#error} command.
19322 This also works with @option{-include}. So yet another way to use
19323 precompiled headers, good for projects not designed with precompiled
19324 header files in mind, is to simply take most of the header files used by
19325 a project, include them from another header file, precompile that header
19326 file, and @option{-include} the precompiled header. If the header files
19327 have guards against multiple inclusion, they will be skipped because
19328 they've already been included (in the precompiled header).
19330 If you need to precompile the same header file for different
19331 languages, targets, or compiler options, you can instead make a
19332 @emph{directory} named like @file{all.h.gch}, and put each precompiled
19333 header in the directory, perhaps using @option{-o}. It doesn't matter
19334 what you call the files in the directory, every precompiled header in
19335 the directory will be considered. The first precompiled header
19336 encountered in the directory that is valid for this compilation will
19337 be used; they're searched in no particular order.
19339 There are many other possibilities, limited only by your imagination,
19340 good sense, and the constraints of your build system.
19342 A precompiled header file can be used only when these conditions apply:
19346 Only one precompiled header can be used in a particular compilation.
19349 A precompiled header can't be used once the first C token is seen. You
19350 can have preprocessor directives before a precompiled header; you can
19351 even include a precompiled header from inside another header, so long as
19352 there are no C tokens before the @code{#include}.
19355 The precompiled header file must be produced for the same language as
19356 the current compilation. You can't use a C precompiled header for a C++
19360 The precompiled header file must have been produced by the same compiler
19361 binary as the current compilation is using.
19364 Any macros defined before the precompiled header is included must
19365 either be defined in the same way as when the precompiled header was
19366 generated, or must not affect the precompiled header, which usually
19367 means that they don't appear in the precompiled header at all.
19369 The @option{-D} option is one way to define a macro before a
19370 precompiled header is included; using a @code{#define} can also do it.
19371 There are also some options that define macros implicitly, like
19372 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
19375 @item If debugging information is output when using the precompiled
19376 header, using @option{-g} or similar, the same kind of debugging information
19377 must have been output when building the precompiled header. However,
19378 a precompiled header built using @option{-g} can be used in a compilation
19379 when no debugging information is being output.
19381 @item The same @option{-m} options must generally be used when building
19382 and using the precompiled header. @xref{Submodel Options},
19383 for any cases where this rule is relaxed.
19385 @item Each of the following options must be the same when building and using
19386 the precompiled header:
19388 @gccoptlist{-fexceptions}
19391 Some other command-line options starting with @option{-f},
19392 @option{-p}, or @option{-O} must be defined in the same way as when
19393 the precompiled header was generated. At present, it's not clear
19394 which options are safe to change and which are not; the safest choice
19395 is to use exactly the same options when generating and using the
19396 precompiled header. The following are known to be safe:
19398 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
19399 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
19400 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
19405 For all of these except the last, the compiler will automatically
19406 ignore the precompiled header if the conditions aren't met. If you
19407 find an option combination that doesn't work and doesn't cause the
19408 precompiled header to be ignored, please consider filing a bug report,
19411 If you do use differing options when generating and using the
19412 precompiled header, the actual behavior will be a mixture of the
19413 behavior for the options. For instance, if you use @option{-g} to
19414 generate the precompiled header but not when using it, you may or may
19415 not get debugging information for routines in the precompiled header.