1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011,
16 Free Software Foundation, Inc.
18 Permission is granted to copy, distribute and/or modify this document
19 under the terms of the GNU Free Documentation License, Version 1.3 or
20 any later version published by the Free Software Foundation; with the
21 Invariant Sections being ``GNU General Public License'' and ``Funding
22 Free Software'', the Front-Cover texts being (a) (see below), and with
23 the Back-Cover Texts being (b) (see below). A copy of the license is
24 included in the gfdl(7) man page.
26 (a) The FSF's Front-Cover Text is:
30 (b) The FSF's Back-Cover Text is:
32 You have freedom to copy and modify this GNU Manual, like GNU
33 software. Copies published by the Free Software Foundation raise
34 funds for GNU development.
36 @c Set file name and title for the man page.
38 @settitle GNU project C and C++ compiler
40 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
41 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
42 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
43 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
44 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
45 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
46 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
48 Only the most useful options are listed here; see below for the
49 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
52 gpl(7), gfdl(7), fsf-funding(7),
53 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
54 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
55 @file{ld}, @file{binutils} and @file{gdb}.
58 For instructions on reporting bugs, see
62 See the Info entry for @command{gcc}, or
63 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
64 for contributors to GCC@.
69 @chapter GCC Command Options
70 @cindex GCC command options
71 @cindex command options
72 @cindex options, GCC command
74 @c man begin DESCRIPTION
75 When you invoke GCC, it normally does preprocessing, compilation,
76 assembly and linking. The ``overall options'' allow you to stop this
77 process at an intermediate stage. For example, the @option{-c} option
78 says not to run the linker. Then the output consists of object files
79 output by the assembler.
81 Other options are passed on to one stage of processing. Some options
82 control the preprocessor and others the compiler itself. Yet other
83 options control the assembler and linker; most of these are not
84 documented here, since you rarely need to use any of them.
86 @cindex C compilation options
87 Most of the command-line options that you can use with GCC are useful
88 for C programs; when an option is only useful with another language
89 (usually C++), the explanation says so explicitly. If the description
90 for a particular option does not mention a source language, you can use
91 that option with all supported languages.
93 @cindex C++ compilation options
94 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
95 options for compiling C++ programs.
97 @cindex grouping options
98 @cindex options, grouping
99 The @command{gcc} program accepts options and file names as operands. Many
100 options have multi-letter names; therefore multiple single-letter options
101 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
104 @cindex order of options
105 @cindex options, order
106 You can mix options and other arguments. For the most part, the order
107 you use doesn't matter. Order does matter when you use several
108 options of the same kind; for example, if you specify @option{-L} more
109 than once, the directories are searched in the order specified. Also,
110 the placement of the @option{-l} option is significant.
112 Many options have long names starting with @samp{-f} or with
113 @samp{-W}---for example,
114 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
115 these have both positive and negative forms; the negative form of
116 @option{-ffoo} would be @option{-fno-foo}. This manual documents
117 only one of these two forms, whichever one is not the default.
121 @xref{Option Index}, for an index to GCC's options.
124 * Option Summary:: Brief list of all options, without explanations.
125 * Overall Options:: Controlling the kind of output:
126 an executable, object files, assembler files,
127 or preprocessed source.
128 * Invoking G++:: Compiling C++ programs.
129 * C Dialect Options:: Controlling the variant of C language compiled.
130 * C++ Dialect Options:: Variations on C++.
131 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
133 * Language Independent Options:: Controlling how diagnostics should be
135 * Warning Options:: How picky should the compiler be?
136 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
137 * Optimize Options:: How much optimization?
138 * Preprocessor Options:: Controlling header files and macro definitions.
139 Also, getting dependency information for Make.
140 * Assembler Options:: Passing options to the assembler.
141 * Link Options:: Specifying libraries and so on.
142 * Directory Options:: Where to find header files and libraries.
143 Where to find the compiler executable files.
144 * Spec Files:: How to pass switches to sub-processes.
145 * Target Options:: Running a cross-compiler, or an old version of GCC.
146 * Submodel Options:: Specifying minor hardware or convention variations,
147 such as 68010 vs 68020.
148 * Code Gen Options:: Specifying conventions for function calls, data layout
150 * Environment Variables:: Env vars that affect GCC.
151 * Precompiled Headers:: Compiling a header once, and using it many times.
157 @section Option Summary
159 Here is a summary of all the options, grouped by type. Explanations are
160 in the following sections.
163 @item Overall Options
164 @xref{Overall Options,,Options Controlling the Kind of Output}.
165 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
166 -pipe -pass-exit-codes @gol
167 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
168 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
169 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
171 @item C Language Options
172 @xref{C Dialect Options,,Options Controlling C Dialect}.
173 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
174 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
175 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
176 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
177 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
178 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
179 -fsigned-bitfields -fsigned-char @gol
180 -funsigned-bitfields -funsigned-char}
182 @item C++ Language Options
183 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
184 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
185 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
186 -fno-elide-constructors @gol
187 -fno-enforce-eh-specs @gol
188 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
189 -fno-implicit-templates @gol
190 -fno-implicit-inline-templates @gol
191 -fno-implement-inlines -fms-extensions @gol
192 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
193 -fno-optional-diags -fpermissive @gol
194 -fno-pretty-templates @gol
195 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
196 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
197 -fno-default-inline -fvisibility-inlines-hidden @gol
198 -fvisibility-ms-compat @gol
199 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
200 -Wdelete-non-virtual-dtor -Wnarrowing -Wnoexcept @gol
201 -Wnon-virtual-dtor -Wreorder @gol
202 -Weffc++ -Wstrict-null-sentinel @gol
203 -Wno-non-template-friend -Wold-style-cast @gol
204 -Woverloaded-virtual -Wno-pmf-conversions @gol
207 @item Objective-C and Objective-C++ Language Options
208 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
209 Objective-C and Objective-C++ Dialects}.
210 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
211 -fgnu-runtime -fnext-runtime @gol
212 -fno-nil-receivers @gol
213 -fobjc-abi-version=@var{n} @gol
214 -fobjc-call-cxx-cdtors @gol
215 -fobjc-direct-dispatch @gol
216 -fobjc-exceptions @gol
219 -fobjc-std=objc1 @gol
220 -freplace-objc-classes @gol
223 -Wassign-intercept @gol
224 -Wno-protocol -Wselector @gol
225 -Wstrict-selector-match @gol
226 -Wundeclared-selector}
228 @item Language Independent Options
229 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
230 @gccoptlist{-fmessage-length=@var{n} @gol
231 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
232 -fno-diagnostics-show-option}
234 @item Warning Options
235 @xref{Warning Options,,Options to Request or Suppress Warnings}.
236 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
237 -pedantic-errors @gol
238 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
239 -Wno-attributes -Wno-builtin-macro-redefined @gol
240 -Wc++-compat -Wc++11-compat -Wcast-align -Wcast-qual @gol
241 -Wchar-subscripts -Wclobbered -Wcomment @gol
242 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
243 -Wno-deprecated-declarations -Wdisabled-optimization @gol
244 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
245 -Wno-endif-labels -Werror -Werror=* @gol
246 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
247 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
248 -Wformat-security -Wformat-y2k @gol
249 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
250 -Wignored-qualifiers @gol
251 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
252 -Winit-self -Winline -Wmaybe-uninitialized @gol
253 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
254 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
255 -Wlogical-op -Wlong-long @gol
256 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
257 -Wmissing-format-attribute -Wmissing-include-dirs @gol
259 -Wno-multichar -Wnonnull -Wno-overflow @gol
260 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
261 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
262 -Wpointer-arith -Wno-pointer-to-int-cast @gol
263 -Wredundant-decls @gol
264 -Wreturn-type -Wsequence-point -Wshadow @gol
265 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
266 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
267 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
268 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
269 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
270 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
271 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
272 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
273 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
274 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
275 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
276 -Wvariadic-macros -Wvector-operation-performance -Wvla
277 -Wvolatile-register-var -Wwrite-strings -Wzero-as-null-pointer-constant}
279 @item C and Objective-C-only Warning Options
280 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
281 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
282 -Wold-style-declaration -Wold-style-definition @gol
283 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
284 -Wdeclaration-after-statement -Wpointer-sign}
286 @item Debugging Options
287 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
288 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
289 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
290 -fdisable-ipa-@var{pass_name} @gol
291 -fdisable-rtl-@var{pass_name} @gol
292 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
293 -fdisable-tree-@var{pass_name} @gol
294 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
295 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
296 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
297 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
298 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
300 -fdump-statistics @gol
302 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
306 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-nrv -fdump-tree-vect @gol
316 -fdump-tree-sink @gol
317 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
318 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
319 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
320 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
321 -ftree-vectorizer-verbose=@var{n} @gol
322 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
323 -fdump-final-insns=@var{file} @gol
324 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
325 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
326 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
327 -fenable-@var{kind}-@var{pass} @gol
328 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
329 -fdebug-types-section @gol
330 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
331 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
332 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
333 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
334 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
335 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
336 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
337 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
338 -gvms -gxcoff -gxcoff+ @gol
339 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
340 -fdebug-prefix-map=@var{old}=@var{new} @gol
341 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
342 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
343 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
344 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
345 -print-prog-name=@var{program} -print-search-dirs -Q @gol
346 -print-sysroot -print-sysroot-headers-suffix @gol
347 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
349 @item Optimization Options
350 @xref{Optimize Options,,Options that Control Optimization}.
351 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
352 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
353 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
354 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
355 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
356 -fcompare-elim -fcprop-registers -fcrossjumping @gol
357 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
358 -fcx-limited-range @gol
359 -fdata-sections -fdce -fdelayed-branch @gol
360 -fdelete-null-pointer-checks -fdevirtualize -fdse @gol
361 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
362 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
363 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
364 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
365 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
366 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
367 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
368 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
369 -fira-algorithm=@var{algorithm} @gol
370 -fira-region=@var{region} @gol
371 -fira-loop-pressure -fno-ira-share-save-slots @gol
372 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
373 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
374 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
375 -floop-parallelize-all -flto -flto-compression-level @gol
376 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
377 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
378 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
379 -fno-default-inline @gol
380 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
381 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
382 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
383 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
384 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
385 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
386 -fprefetch-loop-arrays @gol
387 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
388 -fprofile-generate=@var{path} @gol
389 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
390 -freciprocal-math -free -fregmove -frename-registers -freorder-blocks @gol
391 -freorder-blocks-and-partition -freorder-functions @gol
392 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
393 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
394 -fsched-spec-load -fsched-spec-load-dangerous @gol
395 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
396 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
397 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
398 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
399 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
400 -fselective-scheduling -fselective-scheduling2 @gol
401 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
402 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
403 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
404 -fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol
405 -fthread-jumps -ftracer -ftree-bit-ccp @gol
406 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
407 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
408 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
409 -ftree-loop-if-convert-stores -ftree-loop-im @gol
410 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
411 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
412 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
413 -ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol
414 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
415 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
416 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
417 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
418 -fwhole-program -fwpa -fuse-linker-plugin @gol
419 --param @var{name}=@var{value}
420 -O -O0 -O1 -O2 -O3 -Os -Ofast}
422 @item Preprocessor Options
423 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
424 @gccoptlist{-A@var{question}=@var{answer} @gol
425 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
426 -C -dD -dI -dM -dN @gol
427 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
428 -idirafter @var{dir} @gol
429 -include @var{file} -imacros @var{file} @gol
430 -iprefix @var{file} -iwithprefix @var{dir} @gol
431 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
432 -imultilib @var{dir} -isysroot @var{dir} @gol
433 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
434 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
435 -remap -trigraphs -undef -U@var{macro} @gol
436 -Wp,@var{option} -Xpreprocessor @var{option}}
438 @item Assembler Option
439 @xref{Assembler Options,,Passing Options to the Assembler}.
440 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
443 @xref{Link Options,,Options for Linking}.
444 @gccoptlist{@var{object-file-name} -l@var{library} @gol
445 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
446 -s -static -static-libgcc -static-libstdc++ -shared @gol
447 -shared-libgcc -symbolic @gol
448 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
451 @item Directory Options
452 @xref{Directory Options,,Options for Directory Search}.
453 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
454 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
457 @item Machine Dependent Options
458 @xref{Submodel Options,,Hardware Models and Configurations}.
459 @c This list is ordered alphanumerically by subsection name.
460 @c Try and put the significant identifier (CPU or system) first,
461 @c so users have a clue at guessing where the ones they want will be.
463 @emph{Adapteva Epiphany Options}
464 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
465 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
466 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
467 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
468 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
469 -msplit-vecmove-early -m1reg-@var{reg}}
472 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
473 -mabi=@var{name} @gol
474 -mapcs-stack-check -mno-apcs-stack-check @gol
475 -mapcs-float -mno-apcs-float @gol
476 -mapcs-reentrant -mno-apcs-reentrant @gol
477 -msched-prolog -mno-sched-prolog @gol
478 -mlittle-endian -mbig-endian -mwords-little-endian @gol
479 -mfloat-abi=@var{name} -mfpe @gol
480 -mfp16-format=@var{name}
481 -mthumb-interwork -mno-thumb-interwork @gol
482 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
483 -mstructure-size-boundary=@var{n} @gol
484 -mabort-on-noreturn @gol
485 -mlong-calls -mno-long-calls @gol
486 -msingle-pic-base -mno-single-pic-base @gol
487 -mpic-register=@var{reg} @gol
488 -mnop-fun-dllimport @gol
489 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
490 -mpoke-function-name @gol
492 -mtpcs-frame -mtpcs-leaf-frame @gol
493 -mcaller-super-interworking -mcallee-super-interworking @gol
494 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
495 -mword-relocations @gol
496 -mfix-cortex-m3-ldrd @gol
500 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
501 -mcall-prologues -mint8 -mno-interrupts -mrelax -mshort-calls @gol
502 -mstrict-X -mtiny-stack}
504 @emph{Blackfin Options}
505 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
506 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
507 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
508 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
509 -mno-id-shared-library -mshared-library-id=@var{n} @gol
510 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
511 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
512 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
516 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
517 -msim -msdata=@var{sdata-type}}
520 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
521 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
522 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
523 -mstack-align -mdata-align -mconst-align @gol
524 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
525 -melf -maout -melinux -mlinux -sim -sim2 @gol
526 -mmul-bug-workaround -mno-mul-bug-workaround}
529 @gccoptlist{-mmac @gol
530 -mcr16cplus -mcr16c @gol
531 -msim -mint32 -mbit-ops
532 -mdata-model=@var{model}}
534 @emph{Darwin Options}
535 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
536 -arch_only -bind_at_load -bundle -bundle_loader @gol
537 -client_name -compatibility_version -current_version @gol
539 -dependency-file -dylib_file -dylinker_install_name @gol
540 -dynamic -dynamiclib -exported_symbols_list @gol
541 -filelist -flat_namespace -force_cpusubtype_ALL @gol
542 -force_flat_namespace -headerpad_max_install_names @gol
544 -image_base -init -install_name -keep_private_externs @gol
545 -multi_module -multiply_defined -multiply_defined_unused @gol
546 -noall_load -no_dead_strip_inits_and_terms @gol
547 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
548 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
549 -private_bundle -read_only_relocs -sectalign @gol
550 -sectobjectsymbols -whyload -seg1addr @gol
551 -sectcreate -sectobjectsymbols -sectorder @gol
552 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
553 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
554 -segprot -segs_read_only_addr -segs_read_write_addr @gol
555 -single_module -static -sub_library -sub_umbrella @gol
556 -twolevel_namespace -umbrella -undefined @gol
557 -unexported_symbols_list -weak_reference_mismatches @gol
558 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
559 -mkernel -mone-byte-bool}
561 @emph{DEC Alpha Options}
562 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
563 -mieee -mieee-with-inexact -mieee-conformant @gol
564 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
565 -mtrap-precision=@var{mode} -mbuild-constants @gol
566 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
567 -mbwx -mmax -mfix -mcix @gol
568 -mfloat-vax -mfloat-ieee @gol
569 -mexplicit-relocs -msmall-data -mlarge-data @gol
570 -msmall-text -mlarge-text @gol
571 -mmemory-latency=@var{time}}
573 @emph{DEC Alpha/VMS Options}
574 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
577 @gccoptlist{-msmall-model -mno-lsim}
580 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
581 -mhard-float -msoft-float @gol
582 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
583 -mdouble -mno-double @gol
584 -mmedia -mno-media -mmuladd -mno-muladd @gol
585 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
586 -mlinked-fp -mlong-calls -malign-labels @gol
587 -mlibrary-pic -macc-4 -macc-8 @gol
588 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
589 -moptimize-membar -mno-optimize-membar @gol
590 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
591 -mvliw-branch -mno-vliw-branch @gol
592 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
593 -mno-nested-cond-exec -mtomcat-stats @gol
597 @emph{GNU/Linux Options}
598 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
599 -tno-android-cc -tno-android-ld}
601 @emph{H8/300 Options}
602 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
605 @gccoptlist{-march=@var{architecture-type} @gol
606 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
607 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
608 -mfixed-range=@var{register-range} @gol
609 -mjump-in-delay -mlinker-opt -mlong-calls @gol
610 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
611 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
612 -mno-jump-in-delay -mno-long-load-store @gol
613 -mno-portable-runtime -mno-soft-float @gol
614 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
615 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
616 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
617 -munix=@var{unix-std} -nolibdld -static -threads}
619 @emph{i386 and x86-64 Options}
620 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
621 -mfpmath=@var{unit} @gol
622 -masm=@var{dialect} -mno-fancy-math-387 @gol
623 -mno-fp-ret-in-387 -msoft-float @gol
624 -mno-wide-multiply -mrtd -malign-double @gol
625 -mpreferred-stack-boundary=@var{num} @gol
626 -mincoming-stack-boundary=@var{num} @gol
627 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
628 -mrecip -mrecip=@var{opt} @gol
629 -mvzeroupper -mprefer-avx128 @gol
630 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
631 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
632 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
633 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
634 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
635 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
636 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
637 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
638 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
639 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
640 -mcmodel=@var{code-model} -mabi=@var{name} @gol
641 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
642 -msse2avx -mfentry -m8bit-idiv @gol
643 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
645 @emph{i386 and x86-64 Windows Options}
646 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
647 -mnop-fun-dllimport -mthread @gol
648 -municode -mwin32 -mwindows -fno-set-stack-executable}
651 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
652 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
653 -mconstant-gp -mauto-pic -mfused-madd @gol
654 -minline-float-divide-min-latency @gol
655 -minline-float-divide-max-throughput @gol
656 -mno-inline-float-divide @gol
657 -minline-int-divide-min-latency @gol
658 -minline-int-divide-max-throughput @gol
659 -mno-inline-int-divide @gol
660 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
661 -mno-inline-sqrt @gol
662 -mdwarf2-asm -mearly-stop-bits @gol
663 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
664 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
665 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
666 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
667 -msched-spec-ldc -msched-spec-control-ldc @gol
668 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
669 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
670 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
671 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
673 @emph{IA-64/VMS Options}
674 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
677 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
678 -msign-extend-enabled -muser-enabled}
680 @emph{M32R/D Options}
681 @gccoptlist{-m32r2 -m32rx -m32r @gol
683 -malign-loops -mno-align-loops @gol
684 -missue-rate=@var{number} @gol
685 -mbranch-cost=@var{number} @gol
686 -mmodel=@var{code-size-model-type} @gol
687 -msdata=@var{sdata-type} @gol
688 -mno-flush-func -mflush-func=@var{name} @gol
689 -mno-flush-trap -mflush-trap=@var{number} @gol
693 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
695 @emph{M680x0 Options}
696 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
697 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
698 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
699 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
700 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
701 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
702 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
703 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
707 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
708 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
709 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
710 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
711 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
714 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
715 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
716 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
717 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
720 @emph{MicroBlaze Options}
721 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
722 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
723 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
724 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
725 -mxl-mode-@var{app-model}}
728 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
729 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
730 -mips64 -mips64r2 @gol
731 -mips16 -mno-mips16 -mflip-mips16 @gol
732 -minterlink-mips16 -mno-interlink-mips16 @gol
733 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
734 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
735 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
736 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
737 -mfpu=@var{fpu-type} @gol
738 -msmartmips -mno-smartmips @gol
739 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
740 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
741 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
742 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
743 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
744 -membedded-data -mno-embedded-data @gol
745 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
746 -mcode-readable=@var{setting} @gol
747 -msplit-addresses -mno-split-addresses @gol
748 -mexplicit-relocs -mno-explicit-relocs @gol
749 -mcheck-zero-division -mno-check-zero-division @gol
750 -mdivide-traps -mdivide-breaks @gol
751 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
752 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
753 -mfix-24k -mno-fix-24k @gol
754 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
755 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
756 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
757 -mflush-func=@var{func} -mno-flush-func @gol
758 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
759 -mfp-exceptions -mno-fp-exceptions @gol
760 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
761 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
764 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
765 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
766 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
767 -mno-base-addresses -msingle-exit -mno-single-exit}
769 @emph{MN10300 Options}
770 @gccoptlist{-mmult-bug -mno-mult-bug @gol
771 -mno-am33 -mam33 -mam33-2 -mam34 @gol
772 -mtune=@var{cpu-type} @gol
773 -mreturn-pointer-on-d0 @gol
774 -mno-crt0 -mrelax -mliw -msetlb}
776 @emph{PDP-11 Options}
777 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
778 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
779 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
780 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
781 -mbranch-expensive -mbranch-cheap @gol
782 -munix-asm -mdec-asm}
784 @emph{picoChip Options}
785 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
786 -msymbol-as-address -mno-inefficient-warnings}
788 @emph{PowerPC Options}
789 See RS/6000 and PowerPC Options.
792 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=rl78}
794 @emph{RS/6000 and PowerPC Options}
795 @gccoptlist{-mcpu=@var{cpu-type} @gol
796 -mtune=@var{cpu-type} @gol
797 -mcmodel=@var{code-model} @gol
798 -mpower -mno-power -mpower2 -mno-power2 @gol
799 -mpowerpc -mpowerpc64 -mno-powerpc @gol
800 -maltivec -mno-altivec @gol
801 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
802 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
803 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
804 -mfprnd -mno-fprnd @gol
805 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
806 -mnew-mnemonics -mold-mnemonics @gol
807 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
808 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
809 -malign-power -malign-natural @gol
810 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
811 -msingle-float -mdouble-float -msimple-fpu @gol
812 -mstring -mno-string -mupdate -mno-update @gol
813 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
814 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
815 -mstrict-align -mno-strict-align -mrelocatable @gol
816 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
817 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
818 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
819 -mprioritize-restricted-insns=@var{priority} @gol
820 -msched-costly-dep=@var{dependence_type} @gol
821 -minsert-sched-nops=@var{scheme} @gol
822 -mcall-sysv -mcall-netbsd @gol
823 -maix-struct-return -msvr4-struct-return @gol
824 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
825 -mblock-move-inline-limit=@var{num} @gol
826 -misel -mno-isel @gol
827 -misel=yes -misel=no @gol
829 -mspe=yes -mspe=no @gol
831 -mgen-cell-microcode -mwarn-cell-microcode @gol
832 -mvrsave -mno-vrsave @gol
833 -mmulhw -mno-mulhw @gol
834 -mdlmzb -mno-dlmzb @gol
835 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
836 -mprototype -mno-prototype @gol
837 -msim -mmvme -mads -myellowknife -memb -msdata @gol
838 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
839 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
840 -mno-recip-precision @gol
841 -mveclibabi=@var{type} -mfriz -mno-friz @gol
842 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
843 -msave-toc-indirect -mno-save-toc-indirect}
846 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
848 -mbig-endian-data -mlittle-endian-data @gol
851 -mas100-syntax -mno-as100-syntax@gol
853 -mmax-constant-size=@gol
856 -msave-acc-in-interrupts}
858 @emph{S/390 and zSeries Options}
859 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
860 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
861 -mlong-double-64 -mlong-double-128 @gol
862 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
863 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
864 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
865 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
866 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
869 @gccoptlist{-meb -mel @gol
873 -mscore5 -mscore5u -mscore7 -mscore7d}
876 @gccoptlist{-m1 -m2 -m2e @gol
877 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
879 -m4-nofpu -m4-single-only -m4-single -m4 @gol
880 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
881 -m5-64media -m5-64media-nofpu @gol
882 -m5-32media -m5-32media-nofpu @gol
883 -m5-compact -m5-compact-nofpu @gol
884 -mb -ml -mdalign -mrelax @gol
885 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
886 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
887 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
888 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
889 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
890 -maccumulate-outgoing-args -minvalid-symbols -msoft-atomic @gol
891 -mbranch-cost=@var{num} -mcbranchdi -mcmpeqdi -mfused-madd -mpretend-cmove}
893 @emph{Solaris 2 Options}
894 @gccoptlist{-mimpure-text -mno-impure-text @gol
898 @gccoptlist{-mcpu=@var{cpu-type} @gol
899 -mtune=@var{cpu-type} @gol
900 -mcmodel=@var{code-model} @gol
901 -mmemory-model=@var{mem-model} @gol
902 -m32 -m64 -mapp-regs -mno-app-regs @gol
903 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
904 -mfpu -mno-fpu -mhard-float -msoft-float @gol
905 -mhard-quad-float -msoft-quad-float @gol
906 -mstack-bias -mno-stack-bias @gol
907 -munaligned-doubles -mno-unaligned-doubles @gol
908 -mv8plus -mno-v8plus -mvis -mno-vis @gol
909 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
910 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
914 @gccoptlist{-mwarn-reloc -merror-reloc @gol
915 -msafe-dma -munsafe-dma @gol
917 -msmall-mem -mlarge-mem -mstdmain @gol
918 -mfixed-range=@var{register-range} @gol
920 -maddress-space-conversion -mno-address-space-conversion @gol
921 -mcache-size=@var{cache-size} @gol
922 -matomic-updates -mno-atomic-updates}
924 @emph{System V Options}
925 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
927 @emph{TILE-Gx Options}
928 @gccoptlist{-mcpu=@var{cpu} -m32 -m64}
930 @emph{TILEPro Options}
931 @gccoptlist{-mcpu=@var{cpu} -m32}
934 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
935 -mprolog-function -mno-prolog-function -mspace @gol
936 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
937 -mapp-regs -mno-app-regs @gol
938 -mdisable-callt -mno-disable-callt @gol
941 -mv850e1 -mv850es @gol
946 @gccoptlist{-mg -mgnu -munix}
948 @emph{VxWorks Options}
949 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
950 -Xbind-lazy -Xbind-now}
952 @emph{x86-64 Options}
953 See i386 and x86-64 Options.
955 @emph{Xstormy16 Options}
958 @emph{Xtensa Options}
959 @gccoptlist{-mconst16 -mno-const16 @gol
960 -mfused-madd -mno-fused-madd @gol
962 -mserialize-volatile -mno-serialize-volatile @gol
963 -mtext-section-literals -mno-text-section-literals @gol
964 -mtarget-align -mno-target-align @gol
965 -mlongcalls -mno-longcalls}
967 @emph{zSeries Options}
968 See S/390 and zSeries Options.
970 @item Code Generation Options
971 @xref{Code Gen Options,,Options for Code Generation Conventions}.
972 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
973 -ffixed-@var{reg} -fexceptions @gol
974 -fnon-call-exceptions -funwind-tables @gol
975 -fasynchronous-unwind-tables @gol
976 -finhibit-size-directive -finstrument-functions @gol
977 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
978 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
979 -fno-common -fno-ident @gol
980 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
981 -fno-jump-tables @gol
982 -frecord-gcc-switches @gol
983 -freg-struct-return -fshort-enums @gol
984 -fshort-double -fshort-wchar @gol
985 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
986 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
987 -fno-stack-limit -fsplit-stack @gol
988 -fleading-underscore -ftls-model=@var{model} @gol
989 -ftrapv -fwrapv -fbounds-check @gol
990 -fvisibility -fstrict-volatile-bitfields}
994 * Overall Options:: Controlling the kind of output:
995 an executable, object files, assembler files,
996 or preprocessed source.
997 * C Dialect Options:: Controlling the variant of C language compiled.
998 * C++ Dialect Options:: Variations on C++.
999 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
1001 * Language Independent Options:: Controlling how diagnostics should be
1003 * Warning Options:: How picky should the compiler be?
1004 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
1005 * Optimize Options:: How much optimization?
1006 * Preprocessor Options:: Controlling header files and macro definitions.
1007 Also, getting dependency information for Make.
1008 * Assembler Options:: Passing options to the assembler.
1009 * Link Options:: Specifying libraries and so on.
1010 * Directory Options:: Where to find header files and libraries.
1011 Where to find the compiler executable files.
1012 * Spec Files:: How to pass switches to sub-processes.
1013 * Target Options:: Running a cross-compiler, or an old version of GCC.
1016 @node Overall Options
1017 @section Options Controlling the Kind of Output
1019 Compilation can involve up to four stages: preprocessing, compilation
1020 proper, assembly and linking, always in that order. GCC is capable of
1021 preprocessing and compiling several files either into several
1022 assembler input files, or into one assembler input file; then each
1023 assembler input file produces an object file, and linking combines all
1024 the object files (those newly compiled, and those specified as input)
1025 into an executable file.
1027 @cindex file name suffix
1028 For any given input file, the file name suffix determines what kind of
1029 compilation is done:
1033 C source code that must be preprocessed.
1036 C source code that should not be preprocessed.
1039 C++ source code that should not be preprocessed.
1042 Objective-C source code. Note that you must link with the @file{libobjc}
1043 library to make an Objective-C program work.
1046 Objective-C source code that should not be preprocessed.
1050 Objective-C++ source code. Note that you must link with the @file{libobjc}
1051 library to make an Objective-C++ program work. Note that @samp{.M} refers
1052 to a literal capital M@.
1054 @item @var{file}.mii
1055 Objective-C++ source code that should not be preprocessed.
1058 C, C++, Objective-C or Objective-C++ header file to be turned into a
1059 precompiled header (default), or C, C++ header file to be turned into an
1060 Ada spec (via the @option{-fdump-ada-spec} switch).
1063 @itemx @var{file}.cp
1064 @itemx @var{file}.cxx
1065 @itemx @var{file}.cpp
1066 @itemx @var{file}.CPP
1067 @itemx @var{file}.c++
1069 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1070 the last two letters must both be literally @samp{x}. Likewise,
1071 @samp{.C} refers to a literal capital C@.
1075 Objective-C++ source code that must be preprocessed.
1077 @item @var{file}.mii
1078 Objective-C++ source code that should not be preprocessed.
1082 @itemx @var{file}.hp
1083 @itemx @var{file}.hxx
1084 @itemx @var{file}.hpp
1085 @itemx @var{file}.HPP
1086 @itemx @var{file}.h++
1087 @itemx @var{file}.tcc
1088 C++ header file to be turned into a precompiled header or Ada spec.
1091 @itemx @var{file}.for
1092 @itemx @var{file}.ftn
1093 Fixed form Fortran source code that should not be preprocessed.
1096 @itemx @var{file}.FOR
1097 @itemx @var{file}.fpp
1098 @itemx @var{file}.FPP
1099 @itemx @var{file}.FTN
1100 Fixed form Fortran source code that must be preprocessed (with the traditional
1103 @item @var{file}.f90
1104 @itemx @var{file}.f95
1105 @itemx @var{file}.f03
1106 @itemx @var{file}.f08
1107 Free form Fortran source code that should not be preprocessed.
1109 @item @var{file}.F90
1110 @itemx @var{file}.F95
1111 @itemx @var{file}.F03
1112 @itemx @var{file}.F08
1113 Free form Fortran source code that must be preprocessed (with the
1114 traditional preprocessor).
1119 @c FIXME: Descriptions of Java file types.
1125 @item @var{file}.ads
1126 Ada source code file that contains a library unit declaration (a
1127 declaration of a package, subprogram, or generic, or a generic
1128 instantiation), or a library unit renaming declaration (a package,
1129 generic, or subprogram renaming declaration). Such files are also
1132 @item @var{file}.adb
1133 Ada source code file containing a library unit body (a subprogram or
1134 package body). Such files are also called @dfn{bodies}.
1136 @c GCC also knows about some suffixes for languages not yet included:
1147 @itemx @var{file}.sx
1148 Assembler code that must be preprocessed.
1151 An object file to be fed straight into linking.
1152 Any file name with no recognized suffix is treated this way.
1156 You can specify the input language explicitly with the @option{-x} option:
1159 @item -x @var{language}
1160 Specify explicitly the @var{language} for the following input files
1161 (rather than letting the compiler choose a default based on the file
1162 name suffix). This option applies to all following input files until
1163 the next @option{-x} option. Possible values for @var{language} are:
1165 c c-header cpp-output
1166 c++ c++-header c++-cpp-output
1167 objective-c objective-c-header objective-c-cpp-output
1168 objective-c++ objective-c++-header objective-c++-cpp-output
1169 assembler assembler-with-cpp
1171 f77 f77-cpp-input f95 f95-cpp-input
1177 Turn off any specification of a language, so that subsequent files are
1178 handled according to their file name suffixes (as they are if @option{-x}
1179 has not been used at all).
1181 @item -pass-exit-codes
1182 @opindex pass-exit-codes
1183 Normally the @command{gcc} program will exit with the code of 1 if any
1184 phase of the compiler returns a non-success return code. If you specify
1185 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1186 numerically highest error produced by any phase that returned an error
1187 indication. The C, C++, and Fortran frontends return 4, if an internal
1188 compiler error is encountered.
1191 If you only want some of the stages of compilation, you can use
1192 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1193 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1194 @command{gcc} is to stop. Note that some combinations (for example,
1195 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1200 Compile or assemble the source files, but do not link. The linking
1201 stage simply is not done. The ultimate output is in the form of an
1202 object file for each source file.
1204 By default, the object file name for a source file is made by replacing
1205 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1207 Unrecognized input files, not requiring compilation or assembly, are
1212 Stop after the stage of compilation proper; do not assemble. The output
1213 is in the form of an assembler code file for each non-assembler input
1216 By default, the assembler file name for a source file is made by
1217 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1219 Input files that don't require compilation are ignored.
1223 Stop after the preprocessing stage; do not run the compiler proper. The
1224 output is in the form of preprocessed source code, which is sent to the
1227 Input files that don't require preprocessing are ignored.
1229 @cindex output file option
1232 Place output in file @var{file}. This applies regardless to whatever
1233 sort of output is being produced, whether it be an executable file,
1234 an object file, an assembler file or preprocessed C code.
1236 If @option{-o} is not specified, the default is to put an executable
1237 file in @file{a.out}, the object file for
1238 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1239 assembler file in @file{@var{source}.s}, a precompiled header file in
1240 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1245 Print (on standard error output) the commands executed to run the stages
1246 of compilation. Also print the version number of the compiler driver
1247 program and of the preprocessor and the compiler proper.
1251 Like @option{-v} except the commands are not executed and arguments
1252 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1253 This is useful for shell scripts to capture the driver-generated command lines.
1257 Use pipes rather than temporary files for communication between the
1258 various stages of compilation. This fails to work on some systems where
1259 the assembler is unable to read from a pipe; but the GNU assembler has
1264 Print (on the standard output) a description of the command-line options
1265 understood by @command{gcc}. If the @option{-v} option is also specified
1266 then @option{--help} will also be passed on to the various processes
1267 invoked by @command{gcc}, so that they can display the command-line options
1268 they accept. If the @option{-Wextra} option has also been specified
1269 (prior to the @option{--help} option), then command-line options that
1270 have no documentation associated with them will also be displayed.
1273 @opindex target-help
1274 Print (on the standard output) a description of target-specific command-line
1275 options for each tool. For some targets extra target-specific
1276 information may also be printed.
1278 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1279 Print (on the standard output) a description of the command-line
1280 options understood by the compiler that fit into all specified classes
1281 and qualifiers. These are the supported classes:
1284 @item @samp{optimizers}
1285 This will display all of the optimization options supported by the
1288 @item @samp{warnings}
1289 This will display all of the options controlling warning messages
1290 produced by the compiler.
1293 This will display target-specific options. Unlike the
1294 @option{--target-help} option however, target-specific options of the
1295 linker and assembler will not be displayed. This is because those
1296 tools do not currently support the extended @option{--help=} syntax.
1299 This will display the values recognized by the @option{--param}
1302 @item @var{language}
1303 This will display the options supported for @var{language}, where
1304 @var{language} is the name of one of the languages supported in this
1308 This will display the options that are common to all languages.
1311 These are the supported qualifiers:
1314 @item @samp{undocumented}
1315 Display only those options that are undocumented.
1318 Display options taking an argument that appears after an equal
1319 sign in the same continuous piece of text, such as:
1320 @samp{--help=target}.
1322 @item @samp{separate}
1323 Display options taking an argument that appears as a separate word
1324 following the original option, such as: @samp{-o output-file}.
1327 Thus for example to display all the undocumented target-specific
1328 switches supported by the compiler the following can be used:
1331 --help=target,undocumented
1334 The sense of a qualifier can be inverted by prefixing it with the
1335 @samp{^} character, so for example to display all binary warning
1336 options (i.e., ones that are either on or off and that do not take an
1337 argument) that have a description, use:
1340 --help=warnings,^joined,^undocumented
1343 The argument to @option{--help=} should not consist solely of inverted
1346 Combining several classes is possible, although this usually
1347 restricts the output by so much that there is nothing to display. One
1348 case where it does work however is when one of the classes is
1349 @var{target}. So for example to display all the target-specific
1350 optimization options the following can be used:
1353 --help=target,optimizers
1356 The @option{--help=} option can be repeated on the command line. Each
1357 successive use will display its requested class of options, skipping
1358 those that have already been displayed.
1360 If the @option{-Q} option appears on the command line before the
1361 @option{--help=} option, then the descriptive text displayed by
1362 @option{--help=} is changed. Instead of describing the displayed
1363 options, an indication is given as to whether the option is enabled,
1364 disabled or set to a specific value (assuming that the compiler
1365 knows this at the point where the @option{--help=} option is used).
1367 Here is a truncated example from the ARM port of @command{gcc}:
1370 % gcc -Q -mabi=2 --help=target -c
1371 The following options are target specific:
1373 -mabort-on-noreturn [disabled]
1377 The output is sensitive to the effects of previous command-line
1378 options, so for example it is possible to find out which optimizations
1379 are enabled at @option{-O2} by using:
1382 -Q -O2 --help=optimizers
1385 Alternatively you can discover which binary optimizations are enabled
1386 by @option{-O3} by using:
1389 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1390 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1391 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1394 @item -no-canonical-prefixes
1395 @opindex no-canonical-prefixes
1396 Do not expand any symbolic links, resolve references to @samp{/../}
1397 or @samp{/./}, or make the path absolute when generating a relative
1402 Display the version number and copyrights of the invoked GCC@.
1406 Invoke all subcommands under a wrapper program. The name of the
1407 wrapper program and its parameters are passed as a comma separated
1411 gcc -c t.c -wrapper gdb,--args
1414 This will invoke all subprograms of @command{gcc} under
1415 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1416 @samp{gdb --args cc1 @dots{}}.
1418 @item -fplugin=@var{name}.so
1419 Load the plugin code in file @var{name}.so, assumed to be a
1420 shared object to be dlopen'd by the compiler. The base name of
1421 the shared object file is used to identify the plugin for the
1422 purposes of argument parsing (See
1423 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1424 Each plugin should define the callback functions specified in the
1427 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1428 Define an argument called @var{key} with a value of @var{value}
1429 for the plugin called @var{name}.
1431 @item -fdump-ada-spec@r{[}-slim@r{]}
1432 @opindex fdump-ada-spec
1433 For C and C++ source and include files, generate corresponding Ada specs.
1434 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1435 GNAT User's Guide}, which provides detailed documentation on this feature.
1437 @item -fdump-go-spec=@var{file}
1438 @opindex fdump-go-spec
1439 For input files in any language, generate corresponding Go
1440 declarations in @var{file}. This generates Go @code{const},
1441 @code{type}, @code{var}, and @code{func} declarations which may be a
1442 useful way to start writing a Go interface to code written in some
1445 @include @value{srcdir}/../libiberty/at-file.texi
1449 @section Compiling C++ Programs
1451 @cindex suffixes for C++ source
1452 @cindex C++ source file suffixes
1453 C++ source files conventionally use one of the suffixes @samp{.C},
1454 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1455 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1456 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1457 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1458 files with these names and compiles them as C++ programs even if you
1459 call the compiler the same way as for compiling C programs (usually
1460 with the name @command{gcc}).
1464 However, the use of @command{gcc} does not add the C++ library.
1465 @command{g++} is a program that calls GCC and treats @samp{.c},
1466 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1467 files unless @option{-x} is used, and automatically specifies linking
1468 against the C++ library. This program is also useful when
1469 precompiling a C header file with a @samp{.h} extension for use in C++
1470 compilations. On many systems, @command{g++} is also installed with
1471 the name @command{c++}.
1473 @cindex invoking @command{g++}
1474 When you compile C++ programs, you may specify many of the same
1475 command-line options that you use for compiling programs in any
1476 language; or command-line options meaningful for C and related
1477 languages; or options that are meaningful only for C++ programs.
1478 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1479 explanations of options for languages related to C@.
1480 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1481 explanations of options that are meaningful only for C++ programs.
1483 @node C Dialect Options
1484 @section Options Controlling C Dialect
1485 @cindex dialect options
1486 @cindex language dialect options
1487 @cindex options, dialect
1489 The following options control the dialect of C (or languages derived
1490 from C, such as C++, Objective-C and Objective-C++) that the compiler
1494 @cindex ANSI support
1498 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1499 equivalent to @samp{-std=c++98}.
1501 This turns off certain features of GCC that are incompatible with ISO
1502 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1503 such as the @code{asm} and @code{typeof} keywords, and
1504 predefined macros such as @code{unix} and @code{vax} that identify the
1505 type of system you are using. It also enables the undesirable and
1506 rarely used ISO trigraph feature. For the C compiler,
1507 it disables recognition of C++ style @samp{//} comments as well as
1508 the @code{inline} keyword.
1510 The alternate keywords @code{__asm__}, @code{__extension__},
1511 @code{__inline__} and @code{__typeof__} continue to work despite
1512 @option{-ansi}. You would not want to use them in an ISO C program, of
1513 course, but it is useful to put them in header files that might be included
1514 in compilations done with @option{-ansi}. Alternate predefined macros
1515 such as @code{__unix__} and @code{__vax__} are also available, with or
1516 without @option{-ansi}.
1518 The @option{-ansi} option does not cause non-ISO programs to be
1519 rejected gratuitously. For that, @option{-pedantic} is required in
1520 addition to @option{-ansi}. @xref{Warning Options}.
1522 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1523 option is used. Some header files may notice this macro and refrain
1524 from declaring certain functions or defining certain macros that the
1525 ISO standard doesn't call for; this is to avoid interfering with any
1526 programs that might use these names for other things.
1528 Functions that would normally be built in but do not have semantics
1529 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1530 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1531 built-in functions provided by GCC}, for details of the functions
1536 Determine the language standard. @xref{Standards,,Language Standards
1537 Supported by GCC}, for details of these standard versions. This option
1538 is currently only supported when compiling C or C++.
1540 The compiler can accept several base standards, such as @samp{c90} or
1541 @samp{c++98}, and GNU dialects of those standards, such as
1542 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1543 compiler will accept all programs following that standard and those
1544 using GNU extensions that do not contradict it. For example,
1545 @samp{-std=c90} turns off certain features of GCC that are
1546 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1547 keywords, but not other GNU extensions that do not have a meaning in
1548 ISO C90, such as omitting the middle term of a @code{?:}
1549 expression. On the other hand, by specifying a GNU dialect of a
1550 standard, all features the compiler support are enabled, even when
1551 those features change the meaning of the base standard and some
1552 strict-conforming programs may be rejected. The particular standard
1553 is used by @option{-pedantic} to identify which features are GNU
1554 extensions given that version of the standard. For example
1555 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1556 comments, while @samp{-std=gnu99 -pedantic} would not.
1558 A value for this option must be provided; possible values are
1564 Support all ISO C90 programs (certain GNU extensions that conflict
1565 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1567 @item iso9899:199409
1568 ISO C90 as modified in amendment 1.
1574 ISO C99. Note that this standard is not yet fully supported; see
1575 @w{@uref{http://gcc.gnu.org/gcc-4.7/c99status.html}} for more information. The
1576 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1581 ISO C11, the 2011 revision of the ISO C standard.
1582 Support is incomplete and experimental. The name @samp{c1x} is
1587 GNU dialect of ISO C90 (including some C99 features). This
1588 is the default for C code.
1592 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1593 this will become the default. The name @samp{gnu9x} is deprecated.
1597 GNU dialect of ISO C11. Support is incomplete and experimental. The
1598 name @samp{gnu1x} is deprecated.
1601 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1605 GNU dialect of @option{-std=c++98}. This is the default for
1609 The 2011 ISO C++ standard plus amendments. Support for C++11 is still
1610 experimental, and may change in incompatible ways in future releases.
1613 GNU dialect of @option{-std=c++11}. Support for C++11 is still
1614 experimental, and may change in incompatible ways in future releases.
1617 @item -fgnu89-inline
1618 @opindex fgnu89-inline
1619 The option @option{-fgnu89-inline} tells GCC to use the traditional
1620 GNU semantics for @code{inline} functions when in C99 mode.
1621 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1622 is accepted and ignored by GCC versions 4.1.3 up to but not including
1623 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1624 C99 mode. Using this option is roughly equivalent to adding the
1625 @code{gnu_inline} function attribute to all inline functions
1626 (@pxref{Function Attributes}).
1628 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1629 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1630 specifies the default behavior). This option was first supported in
1631 GCC 4.3. This option is not supported in @option{-std=c90} or
1632 @option{-std=gnu90} mode.
1634 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1635 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1636 in effect for @code{inline} functions. @xref{Common Predefined
1637 Macros,,,cpp,The C Preprocessor}.
1639 @item -aux-info @var{filename}
1641 Output to the given filename prototyped declarations for all functions
1642 declared and/or defined in a translation unit, including those in header
1643 files. This option is silently ignored in any language other than C@.
1645 Besides declarations, the file indicates, in comments, the origin of
1646 each declaration (source file and line), whether the declaration was
1647 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1648 @samp{O} for old, respectively, in the first character after the line
1649 number and the colon), and whether it came from a declaration or a
1650 definition (@samp{C} or @samp{F}, respectively, in the following
1651 character). In the case of function definitions, a K&R-style list of
1652 arguments followed by their declarations is also provided, inside
1653 comments, after the declaration.
1655 @item -fallow-parameterless-variadic-functions
1656 Accept variadic functions without named parameters.
1658 Although it is possible to define such a function, this is not very
1659 useful as it is not possible to read the arguments. This is only
1660 supported for C as this construct is allowed by C++.
1664 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1665 keyword, so that code can use these words as identifiers. You can use
1666 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1667 instead. @option{-ansi} implies @option{-fno-asm}.
1669 In C++, this switch only affects the @code{typeof} keyword, since
1670 @code{asm} and @code{inline} are standard keywords. You may want to
1671 use the @option{-fno-gnu-keywords} flag instead, which has the same
1672 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1673 switch only affects the @code{asm} and @code{typeof} keywords, since
1674 @code{inline} is a standard keyword in ISO C99.
1677 @itemx -fno-builtin-@var{function}
1678 @opindex fno-builtin
1679 @cindex built-in functions
1680 Don't recognize built-in functions that do not begin with
1681 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1682 functions provided by GCC}, for details of the functions affected,
1683 including those which are not built-in functions when @option{-ansi} or
1684 @option{-std} options for strict ISO C conformance are used because they
1685 do not have an ISO standard meaning.
1687 GCC normally generates special code to handle certain built-in functions
1688 more efficiently; for instance, calls to @code{alloca} may become single
1689 instructions which adjust the stack directly, and calls to @code{memcpy}
1690 may become inline copy loops. The resulting code is often both smaller
1691 and faster, but since the function calls no longer appear as such, you
1692 cannot set a breakpoint on those calls, nor can you change the behavior
1693 of the functions by linking with a different library. In addition,
1694 when a function is recognized as a built-in function, GCC may use
1695 information about that function to warn about problems with calls to
1696 that function, or to generate more efficient code, even if the
1697 resulting code still contains calls to that function. For example,
1698 warnings are given with @option{-Wformat} for bad calls to
1699 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1700 known not to modify global memory.
1702 With the @option{-fno-builtin-@var{function}} option
1703 only the built-in function @var{function} is
1704 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1705 function is named that is not built-in in this version of GCC, this
1706 option is ignored. There is no corresponding
1707 @option{-fbuiltin-@var{function}} option; if you wish to enable
1708 built-in functions selectively when using @option{-fno-builtin} or
1709 @option{-ffreestanding}, you may define macros such as:
1712 #define abs(n) __builtin_abs ((n))
1713 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1718 @cindex hosted environment
1720 Assert that compilation takes place in a hosted environment. This implies
1721 @option{-fbuiltin}. A hosted environment is one in which the
1722 entire standard library is available, and in which @code{main} has a return
1723 type of @code{int}. Examples are nearly everything except a kernel.
1724 This is equivalent to @option{-fno-freestanding}.
1726 @item -ffreestanding
1727 @opindex ffreestanding
1728 @cindex hosted environment
1730 Assert that compilation takes place in a freestanding environment. This
1731 implies @option{-fno-builtin}. A freestanding environment
1732 is one in which the standard library may not exist, and program startup may
1733 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1734 This is equivalent to @option{-fno-hosted}.
1736 @xref{Standards,,Language Standards Supported by GCC}, for details of
1737 freestanding and hosted environments.
1741 @cindex OpenMP parallel
1742 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1743 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1744 compiler generates parallel code according to the OpenMP Application
1745 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1746 implies @option{-pthread}, and thus is only supported on targets that
1747 have support for @option{-pthread}.
1751 When the option @option{-fgnu-tm} is specified, the compiler will
1752 generate code for the Linux variant of Intel's current Transactional
1753 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1754 an experimental feature whose interface may change in future versions
1755 of GCC, as the official specification changes. Please note that not
1756 all architectures are supported for this feature.
1758 For more information on GCC's support for transactional memory,
1759 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1760 Transactional Memory Library}.
1762 Note that the transactional memory feature is not supported with
1763 non-call exceptions (@option{-fnon-call-exceptions}).
1765 @item -fms-extensions
1766 @opindex fms-extensions
1767 Accept some non-standard constructs used in Microsoft header files.
1769 In C++ code, this allows member names in structures to be similar
1770 to previous types declarations.
1779 Some cases of unnamed fields in structures and unions are only
1780 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1781 fields within structs/unions}, for details.
1783 @item -fplan9-extensions
1784 Accept some non-standard constructs used in Plan 9 code.
1786 This enables @option{-fms-extensions}, permits passing pointers to
1787 structures with anonymous fields to functions that expect pointers to
1788 elements of the type of the field, and permits referring to anonymous
1789 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1790 struct/union fields within structs/unions}, for details. This is only
1791 supported for C, not C++.
1795 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1796 options for strict ISO C conformance) implies @option{-trigraphs}.
1798 @item -no-integrated-cpp
1799 @opindex no-integrated-cpp
1800 Performs a compilation in two passes: preprocessing and compiling. This
1801 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1802 @option{-B} option. The user supplied compilation step can then add in
1803 an additional preprocessing step after normal preprocessing but before
1804 compiling. The default is to use the integrated cpp (internal cpp)
1806 The semantics of this option will change if "cc1", "cc1plus", and
1807 "cc1obj" are merged.
1809 @cindex traditional C language
1810 @cindex C language, traditional
1812 @itemx -traditional-cpp
1813 @opindex traditional-cpp
1814 @opindex traditional
1815 Formerly, these options caused GCC to attempt to emulate a pre-standard
1816 C compiler. They are now only supported with the @option{-E} switch.
1817 The preprocessor continues to support a pre-standard mode. See the GNU
1818 CPP manual for details.
1820 @item -fcond-mismatch
1821 @opindex fcond-mismatch
1822 Allow conditional expressions with mismatched types in the second and
1823 third arguments. The value of such an expression is void. This option
1824 is not supported for C++.
1826 @item -flax-vector-conversions
1827 @opindex flax-vector-conversions
1828 Allow implicit conversions between vectors with differing numbers of
1829 elements and/or incompatible element types. This option should not be
1832 @item -funsigned-char
1833 @opindex funsigned-char
1834 Let the type @code{char} be unsigned, like @code{unsigned char}.
1836 Each kind of machine has a default for what @code{char} should
1837 be. It is either like @code{unsigned char} by default or like
1838 @code{signed char} by default.
1840 Ideally, a portable program should always use @code{signed char} or
1841 @code{unsigned char} when it depends on the signedness of an object.
1842 But many programs have been written to use plain @code{char} and
1843 expect it to be signed, or expect it to be unsigned, depending on the
1844 machines they were written for. This option, and its inverse, let you
1845 make such a program work with the opposite default.
1847 The type @code{char} is always a distinct type from each of
1848 @code{signed char} or @code{unsigned char}, even though its behavior
1849 is always just like one of those two.
1852 @opindex fsigned-char
1853 Let the type @code{char} be signed, like @code{signed char}.
1855 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1856 the negative form of @option{-funsigned-char}. Likewise, the option
1857 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1859 @item -fsigned-bitfields
1860 @itemx -funsigned-bitfields
1861 @itemx -fno-signed-bitfields
1862 @itemx -fno-unsigned-bitfields
1863 @opindex fsigned-bitfields
1864 @opindex funsigned-bitfields
1865 @opindex fno-signed-bitfields
1866 @opindex fno-unsigned-bitfields
1867 These options control whether a bit-field is signed or unsigned, when the
1868 declaration does not use either @code{signed} or @code{unsigned}. By
1869 default, such a bit-field is signed, because this is consistent: the
1870 basic integer types such as @code{int} are signed types.
1873 @node C++ Dialect Options
1874 @section Options Controlling C++ Dialect
1876 @cindex compiler options, C++
1877 @cindex C++ options, command-line
1878 @cindex options, C++
1879 This section describes the command-line options that are only meaningful
1880 for C++ programs; but you can also use most of the GNU compiler options
1881 regardless of what language your program is in. For example, you
1882 might compile a file @code{firstClass.C} like this:
1885 g++ -g -frepo -O -c firstClass.C
1889 In this example, only @option{-frepo} is an option meant
1890 only for C++ programs; you can use the other options with any
1891 language supported by GCC@.
1893 Here is a list of options that are @emph{only} for compiling C++ programs:
1897 @item -fabi-version=@var{n}
1898 @opindex fabi-version
1899 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1900 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1901 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1902 the version that conforms most closely to the C++ ABI specification.
1903 Therefore, the ABI obtained using version 0 will change as ABI bugs
1906 The default is version 2.
1908 Version 3 corrects an error in mangling a constant address as a
1911 Version 4, which first appeared in G++ 4.5, implements a standard
1912 mangling for vector types.
1914 Version 5, which first appeared in G++ 4.6, corrects the mangling of
1915 attribute const/volatile on function pointer types, decltype of a
1916 plain decl, and use of a function parameter in the declaration of
1919 Version 6, which first appeared in G++ 4.7, corrects the promotion
1920 behavior of C++11 scoped enums and the mangling of template argument
1921 packs, const/static_cast, prefix ++ and --, and a class scope function
1922 used as a template argument.
1924 See also @option{-Wabi}.
1926 @item -fno-access-control
1927 @opindex fno-access-control
1928 Turn off all access checking. This switch is mainly useful for working
1929 around bugs in the access control code.
1933 Check that the pointer returned by @code{operator new} is non-null
1934 before attempting to modify the storage allocated. This check is
1935 normally unnecessary because the C++ standard specifies that
1936 @code{operator new} will only return @code{0} if it is declared
1937 @samp{throw()}, in which case the compiler will always check the
1938 return value even without this option. In all other cases, when
1939 @code{operator new} has a non-empty exception specification, memory
1940 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1941 @samp{new (nothrow)}.
1943 @item -fconserve-space
1944 @opindex fconserve-space
1945 Put uninitialized or run-time-initialized global variables into the
1946 common segment, as C does. This saves space in the executable at the
1947 cost of not diagnosing duplicate definitions. If you compile with this
1948 flag and your program mysteriously crashes after @code{main()} has
1949 completed, you may have an object that is being destroyed twice because
1950 two definitions were merged.
1952 This option is no longer useful on most targets, now that support has
1953 been added for putting variables into BSS without making them common.
1955 @item -fconstexpr-depth=@var{n}
1956 @opindex fconstexpr-depth
1957 Set the maximum nested evaluation depth for C++11 constexpr functions
1958 to @var{n}. A limit is needed to detect endless recursion during
1959 constant expression evaluation. The minimum specified by the standard
1962 @item -fdeduce-init-list
1963 @opindex fdeduce-init-list
1964 Enable deduction of a template type parameter as
1965 std::initializer_list from a brace-enclosed initializer list, i.e.
1968 template <class T> auto forward(T t) -> decltype (realfn (t))
1975 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1979 This deduction was implemented as a possible extension to the
1980 originally proposed semantics for the C++11 standard, but was not part
1981 of the final standard, so it is disabled by default. This option is
1982 deprecated, and may be removed in a future version of G++.
1984 @item -ffriend-injection
1985 @opindex ffriend-injection
1986 Inject friend functions into the enclosing namespace, so that they are
1987 visible outside the scope of the class in which they are declared.
1988 Friend functions were documented to work this way in the old Annotated
1989 C++ Reference Manual, and versions of G++ before 4.1 always worked
1990 that way. However, in ISO C++ a friend function that is not declared
1991 in an enclosing scope can only be found using argument dependent
1992 lookup. This option causes friends to be injected as they were in
1995 This option is for compatibility, and may be removed in a future
1998 @item -fno-elide-constructors
1999 @opindex fno-elide-constructors
2000 The C++ standard allows an implementation to omit creating a temporary
2001 that is only used to initialize another object of the same type.
2002 Specifying this option disables that optimization, and forces G++ to
2003 call the copy constructor in all cases.
2005 @item -fno-enforce-eh-specs
2006 @opindex fno-enforce-eh-specs
2007 Don't generate code to check for violation of exception specifications
2008 at run time. This option violates the C++ standard, but may be useful
2009 for reducing code size in production builds, much like defining
2010 @samp{NDEBUG}. This does not give user code permission to throw
2011 exceptions in violation of the exception specifications; the compiler
2012 will still optimize based on the specifications, so throwing an
2013 unexpected exception will result in undefined behavior.
2016 @itemx -fno-for-scope
2018 @opindex fno-for-scope
2019 If @option{-ffor-scope} is specified, the scope of variables declared in
2020 a @i{for-init-statement} is limited to the @samp{for} loop itself,
2021 as specified by the C++ standard.
2022 If @option{-fno-for-scope} is specified, the scope of variables declared in
2023 a @i{for-init-statement} extends to the end of the enclosing scope,
2024 as was the case in old versions of G++, and other (traditional)
2025 implementations of C++.
2027 The default if neither flag is given to follow the standard,
2028 but to allow and give a warning for old-style code that would
2029 otherwise be invalid, or have different behavior.
2031 @item -fno-gnu-keywords
2032 @opindex fno-gnu-keywords
2033 Do not recognize @code{typeof} as a keyword, so that code can use this
2034 word as an identifier. You can use the keyword @code{__typeof__} instead.
2035 @option{-ansi} implies @option{-fno-gnu-keywords}.
2037 @item -fno-implicit-templates
2038 @opindex fno-implicit-templates
2039 Never emit code for non-inline templates that are instantiated
2040 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2041 @xref{Template Instantiation}, for more information.
2043 @item -fno-implicit-inline-templates
2044 @opindex fno-implicit-inline-templates
2045 Don't emit code for implicit instantiations of inline templates, either.
2046 The default is to handle inlines differently so that compiles with and
2047 without optimization will need the same set of explicit instantiations.
2049 @item -fno-implement-inlines
2050 @opindex fno-implement-inlines
2051 To save space, do not emit out-of-line copies of inline functions
2052 controlled by @samp{#pragma implementation}. This will cause linker
2053 errors if these functions are not inlined everywhere they are called.
2055 @item -fms-extensions
2056 @opindex fms-extensions
2057 Disable pedantic warnings about constructs used in MFC, such as implicit
2058 int and getting a pointer to member function via non-standard syntax.
2060 @item -fno-nonansi-builtins
2061 @opindex fno-nonansi-builtins
2062 Disable built-in declarations of functions that are not mandated by
2063 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2064 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2067 @opindex fnothrow-opt
2068 Treat a @code{throw()} exception specification as though it were a
2069 @code{noexcept} specification to reduce or eliminate the text size
2070 overhead relative to a function with no exception specification. If
2071 the function has local variables of types with non-trivial
2072 destructors, the exception specification will actually make the
2073 function smaller because the EH cleanups for those variables can be
2074 optimized away. The semantic effect is that an exception thrown out of
2075 a function with such an exception specification will result in a call
2076 to @code{terminate} rather than @code{unexpected}.
2078 @item -fno-operator-names
2079 @opindex fno-operator-names
2080 Do not treat the operator name keywords @code{and}, @code{bitand},
2081 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2082 synonyms as keywords.
2084 @item -fno-optional-diags
2085 @opindex fno-optional-diags
2086 Disable diagnostics that the standard says a compiler does not need to
2087 issue. Currently, the only such diagnostic issued by G++ is the one for
2088 a name having multiple meanings within a class.
2091 @opindex fpermissive
2092 Downgrade some diagnostics about nonconformant code from errors to
2093 warnings. Thus, using @option{-fpermissive} will allow some
2094 nonconforming code to compile.
2096 @item -fno-pretty-templates
2097 @opindex fno-pretty-templates
2098 When an error message refers to a specialization of a function
2099 template, the compiler will normally print the signature of the
2100 template followed by the template arguments and any typedefs or
2101 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2102 rather than @code{void f(int)}) so that it's clear which template is
2103 involved. When an error message refers to a specialization of a class
2104 template, the compiler will omit any template arguments that match
2105 the default template arguments for that template. If either of these
2106 behaviors make it harder to understand the error message rather than
2107 easier, using @option{-fno-pretty-templates} will disable them.
2111 Enable automatic template instantiation at link time. This option also
2112 implies @option{-fno-implicit-templates}. @xref{Template
2113 Instantiation}, for more information.
2117 Disable generation of information about every class with virtual
2118 functions for use by the C++ run-time type identification features
2119 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2120 of the language, you can save some space by using this flag. Note that
2121 exception handling uses the same information, but it will generate it as
2122 needed. The @samp{dynamic_cast} operator can still be used for casts that
2123 do not require run-time type information, i.e.@: casts to @code{void *} or to
2124 unambiguous base classes.
2128 Emit statistics about front-end processing at the end of the compilation.
2129 This information is generally only useful to the G++ development team.
2131 @item -fstrict-enums
2132 @opindex fstrict-enums
2133 Allow the compiler to optimize using the assumption that a value of
2134 enumerated type can only be one of the values of the enumeration (as
2135 defined in the C++ standard; basically, a value that can be
2136 represented in the minimum number of bits needed to represent all the
2137 enumerators). This assumption may not be valid if the program uses a
2138 cast to convert an arbitrary integer value to the enumerated type.
2140 @item -ftemplate-depth=@var{n}
2141 @opindex ftemplate-depth
2142 Set the maximum instantiation depth for template classes to @var{n}.
2143 A limit on the template instantiation depth is needed to detect
2144 endless recursions during template class instantiation. ANSI/ISO C++
2145 conforming programs must not rely on a maximum depth greater than 17
2146 (changed to 1024 in C++11). The default value is 900, as the compiler
2147 can run out of stack space before hitting 1024 in some situations.
2149 @item -fno-threadsafe-statics
2150 @opindex fno-threadsafe-statics
2151 Do not emit the extra code to use the routines specified in the C++
2152 ABI for thread-safe initialization of local statics. You can use this
2153 option to reduce code size slightly in code that doesn't need to be
2156 @item -fuse-cxa-atexit
2157 @opindex fuse-cxa-atexit
2158 Register destructors for objects with static storage duration with the
2159 @code{__cxa_atexit} function rather than the @code{atexit} function.
2160 This option is required for fully standards-compliant handling of static
2161 destructors, but will only work if your C library supports
2162 @code{__cxa_atexit}.
2164 @item -fno-use-cxa-get-exception-ptr
2165 @opindex fno-use-cxa-get-exception-ptr
2166 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2167 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2168 if the runtime routine is not available.
2170 @item -fvisibility-inlines-hidden
2171 @opindex fvisibility-inlines-hidden
2172 This switch declares that the user does not attempt to compare
2173 pointers to inline functions or methods where the addresses of the two functions
2174 were taken in different shared objects.
2176 The effect of this is that GCC may, effectively, mark inline methods with
2177 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2178 appear in the export table of a DSO and do not require a PLT indirection
2179 when used within the DSO@. Enabling this option can have a dramatic effect
2180 on load and link times of a DSO as it massively reduces the size of the
2181 dynamic export table when the library makes heavy use of templates.
2183 The behavior of this switch is not quite the same as marking the
2184 methods as hidden directly, because it does not affect static variables
2185 local to the function or cause the compiler to deduce that
2186 the function is defined in only one shared object.
2188 You may mark a method as having a visibility explicitly to negate the
2189 effect of the switch for that method. For example, if you do want to
2190 compare pointers to a particular inline method, you might mark it as
2191 having default visibility. Marking the enclosing class with explicit
2192 visibility will have no effect.
2194 Explicitly instantiated inline methods are unaffected by this option
2195 as their linkage might otherwise cross a shared library boundary.
2196 @xref{Template Instantiation}.
2198 @item -fvisibility-ms-compat
2199 @opindex fvisibility-ms-compat
2200 This flag attempts to use visibility settings to make GCC's C++
2201 linkage model compatible with that of Microsoft Visual Studio.
2203 The flag makes these changes to GCC's linkage model:
2207 It sets the default visibility to @code{hidden}, like
2208 @option{-fvisibility=hidden}.
2211 Types, but not their members, are not hidden by default.
2214 The One Definition Rule is relaxed for types without explicit
2215 visibility specifications that are defined in more than one different
2216 shared object: those declarations are permitted if they would have
2217 been permitted when this option was not used.
2220 In new code it is better to use @option{-fvisibility=hidden} and
2221 export those classes that are intended to be externally visible.
2222 Unfortunately it is possible for code to rely, perhaps accidentally,
2223 on the Visual Studio behavior.
2225 Among the consequences of these changes are that static data members
2226 of the same type with the same name but defined in different shared
2227 objects will be different, so changing one will not change the other;
2228 and that pointers to function members defined in different shared
2229 objects may not compare equal. When this flag is given, it is a
2230 violation of the ODR to define types with the same name differently.
2234 Do not use weak symbol support, even if it is provided by the linker.
2235 By default, G++ will use weak symbols if they are available. This
2236 option exists only for testing, and should not be used by end-users;
2237 it will result in inferior code and has no benefits. This option may
2238 be removed in a future release of G++.
2242 Do not search for header files in the standard directories specific to
2243 C++, but do still search the other standard directories. (This option
2244 is used when building the C++ library.)
2247 In addition, these optimization, warning, and code generation options
2248 have meanings only for C++ programs:
2251 @item -fno-default-inline
2252 @opindex fno-default-inline
2253 Do not assume @samp{inline} for functions defined inside a class scope.
2254 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2255 functions will have linkage like inline functions; they just won't be
2258 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2261 Warn when G++ generates code that is probably not compatible with the
2262 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2263 all such cases, there are probably some cases that are not warned about,
2264 even though G++ is generating incompatible code. There may also be
2265 cases where warnings are emitted even though the code that is generated
2268 You should rewrite your code to avoid these warnings if you are
2269 concerned about the fact that code generated by G++ may not be binary
2270 compatible with code generated by other compilers.
2272 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2277 A template with a non-type template parameter of reference type is
2278 mangled incorrectly:
2281 template <int &> struct S @{@};
2285 This is fixed in @option{-fabi-version=3}.
2288 SIMD vector types declared using @code{__attribute ((vector_size))} are
2289 mangled in a non-standard way that does not allow for overloading of
2290 functions taking vectors of different sizes.
2292 The mangling is changed in @option{-fabi-version=4}.
2295 The known incompatibilities in @option{-fabi-version=1} include:
2300 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2301 pack data into the same byte as a base class. For example:
2304 struct A @{ virtual void f(); int f1 : 1; @};
2305 struct B : public A @{ int f2 : 1; @};
2309 In this case, G++ will place @code{B::f2} into the same byte
2310 as@code{A::f1}; other compilers will not. You can avoid this problem
2311 by explicitly padding @code{A} so that its size is a multiple of the
2312 byte size on your platform; that will cause G++ and other compilers to
2313 layout @code{B} identically.
2316 Incorrect handling of tail-padding for virtual bases. G++ does not use
2317 tail padding when laying out virtual bases. For example:
2320 struct A @{ virtual void f(); char c1; @};
2321 struct B @{ B(); char c2; @};
2322 struct C : public A, public virtual B @{@};
2326 In this case, G++ will not place @code{B} into the tail-padding for
2327 @code{A}; other compilers will. You can avoid this problem by
2328 explicitly padding @code{A} so that its size is a multiple of its
2329 alignment (ignoring virtual base classes); that will cause G++ and other
2330 compilers to layout @code{C} identically.
2333 Incorrect handling of bit-fields with declared widths greater than that
2334 of their underlying types, when the bit-fields appear in a union. For
2338 union U @{ int i : 4096; @};
2342 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2343 union too small by the number of bits in an @code{int}.
2346 Empty classes can be placed at incorrect offsets. For example:
2356 struct C : public B, public A @{@};
2360 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2361 it should be placed at offset zero. G++ mistakenly believes that the
2362 @code{A} data member of @code{B} is already at offset zero.
2365 Names of template functions whose types involve @code{typename} or
2366 template template parameters can be mangled incorrectly.
2369 template <typename Q>
2370 void f(typename Q::X) @{@}
2372 template <template <typename> class Q>
2373 void f(typename Q<int>::X) @{@}
2377 Instantiations of these templates may be mangled incorrectly.
2381 It also warns psABI related changes. The known psABI changes at this
2387 For SYSV/x86-64, when passing union with long double, it is changed to
2388 pass in memory as specified in psABI. For example:
2398 @code{union U} will always be passed in memory.
2402 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2403 @opindex Wctor-dtor-privacy
2404 @opindex Wno-ctor-dtor-privacy
2405 Warn when a class seems unusable because all the constructors or
2406 destructors in that class are private, and it has neither friends nor
2407 public static member functions.
2409 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2410 @opindex Wdelete-non-virtual-dtor
2411 @opindex Wno-delete-non-virtual-dtor
2412 Warn when @samp{delete} is used to destroy an instance of a class that
2413 has virtual functions and non-virtual destructor. It is unsafe to delete
2414 an instance of a derived class through a pointer to a base class if the
2415 base class does not have a virtual destructor. This warning is enabled
2418 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2420 @opindex Wno-narrowing
2421 Warn when a narrowing conversion prohibited by C++11 occurs within
2425 int i = @{ 2.2 @}; // error: narrowing from double to int
2428 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2430 With -std=c++11, @option{-Wno-narrowing} suppresses the diagnostic
2431 required by the standard. Note that this does not affect the meaning
2432 of well-formed code; narrowing conversions are still considered
2433 ill-formed in SFINAE context.
2435 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2437 @opindex Wno-noexcept
2438 Warn when a noexcept-expression evaluates to false because of a call
2439 to a function that does not have a non-throwing exception
2440 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2441 the compiler to never throw an exception.
2443 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2444 @opindex Wnon-virtual-dtor
2445 @opindex Wno-non-virtual-dtor
2446 Warn when a class has virtual functions and accessible non-virtual
2447 destructor, in which case it would be possible but unsafe to delete
2448 an instance of a derived class through a pointer to the base class.
2449 This warning is also enabled if @option{-Weffc++} is specified.
2451 @item -Wreorder @r{(C++ and Objective-C++ only)}
2453 @opindex Wno-reorder
2454 @cindex reordering, warning
2455 @cindex warning for reordering of member initializers
2456 Warn when the order of member initializers given in the code does not
2457 match the order in which they must be executed. For instance:
2463 A(): j (0), i (1) @{ @}
2467 The compiler will rearrange the member initializers for @samp{i}
2468 and @samp{j} to match the declaration order of the members, emitting
2469 a warning to that effect. This warning is enabled by @option{-Wall}.
2472 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2475 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2478 Warn about violations of the following style guidelines from Scott Meyers'
2479 @cite{Effective C++, Second Edition} book:
2483 Item 11: Define a copy constructor and an assignment operator for classes
2484 with dynamically allocated memory.
2487 Item 12: Prefer initialization to assignment in constructors.
2490 Item 14: Make destructors virtual in base classes.
2493 Item 15: Have @code{operator=} return a reference to @code{*this}.
2496 Item 23: Don't try to return a reference when you must return an object.
2500 Also warn about violations of the following style guidelines from
2501 Scott Meyers' @cite{More Effective C++} book:
2505 Item 6: Distinguish between prefix and postfix forms of increment and
2506 decrement operators.
2509 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2513 When selecting this option, be aware that the standard library
2514 headers do not obey all of these guidelines; use @samp{grep -v}
2515 to filter out those warnings.
2517 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2518 @opindex Wstrict-null-sentinel
2519 @opindex Wno-strict-null-sentinel
2520 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2521 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2522 to @code{__null}. Although it is a null pointer constant not a null pointer,
2523 it is guaranteed to be of the same size as a pointer. But this use is
2524 not portable across different compilers.
2526 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2527 @opindex Wno-non-template-friend
2528 @opindex Wnon-template-friend
2529 Disable warnings when non-templatized friend functions are declared
2530 within a template. Since the advent of explicit template specification
2531 support in G++, if the name of the friend is an unqualified-id (i.e.,
2532 @samp{friend foo(int)}), the C++ language specification demands that the
2533 friend declare or define an ordinary, nontemplate function. (Section
2534 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2535 could be interpreted as a particular specialization of a templatized
2536 function. Because this non-conforming behavior is no longer the default
2537 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2538 check existing code for potential trouble spots and is on by default.
2539 This new compiler behavior can be turned off with
2540 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2541 but disables the helpful warning.
2543 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2544 @opindex Wold-style-cast
2545 @opindex Wno-old-style-cast
2546 Warn if an old-style (C-style) cast to a non-void type is used within
2547 a C++ program. The new-style casts (@samp{dynamic_cast},
2548 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2549 less vulnerable to unintended effects and much easier to search for.
2551 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2552 @opindex Woverloaded-virtual
2553 @opindex Wno-overloaded-virtual
2554 @cindex overloaded virtual function, warning
2555 @cindex warning for overloaded virtual function
2556 Warn when a function declaration hides virtual functions from a
2557 base class. For example, in:
2564 struct B: public A @{
2569 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2577 will fail to compile.
2579 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2580 @opindex Wno-pmf-conversions
2581 @opindex Wpmf-conversions
2582 Disable the diagnostic for converting a bound pointer to member function
2585 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2586 @opindex Wsign-promo
2587 @opindex Wno-sign-promo
2588 Warn when overload resolution chooses a promotion from unsigned or
2589 enumerated type to a signed type, over a conversion to an unsigned type of
2590 the same size. Previous versions of G++ would try to preserve
2591 unsignedness, but the standard mandates the current behavior.
2596 A& operator = (int);
2606 In this example, G++ will synthesize a default @samp{A& operator =
2607 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2610 @node Objective-C and Objective-C++ Dialect Options
2611 @section Options Controlling Objective-C and Objective-C++ Dialects
2613 @cindex compiler options, Objective-C and Objective-C++
2614 @cindex Objective-C and Objective-C++ options, command-line
2615 @cindex options, Objective-C and Objective-C++
2616 (NOTE: This manual does not describe the Objective-C and Objective-C++
2617 languages themselves. @xref{Standards,,Language Standards
2618 Supported by GCC}, for references.)
2620 This section describes the command-line options that are only meaningful
2621 for Objective-C and Objective-C++ programs, but you can also use most of
2622 the language-independent GNU compiler options.
2623 For example, you might compile a file @code{some_class.m} like this:
2626 gcc -g -fgnu-runtime -O -c some_class.m
2630 In this example, @option{-fgnu-runtime} is an option meant only for
2631 Objective-C and Objective-C++ programs; you can use the other options with
2632 any language supported by GCC@.
2634 Note that since Objective-C is an extension of the C language, Objective-C
2635 compilations may also use options specific to the C front-end (e.g.,
2636 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2637 C++-specific options (e.g., @option{-Wabi}).
2639 Here is a list of options that are @emph{only} for compiling Objective-C
2640 and Objective-C++ programs:
2643 @item -fconstant-string-class=@var{class-name}
2644 @opindex fconstant-string-class
2645 Use @var{class-name} as the name of the class to instantiate for each
2646 literal string specified with the syntax @code{@@"@dots{}"}. The default
2647 class name is @code{NXConstantString} if the GNU runtime is being used, and
2648 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2649 @option{-fconstant-cfstrings} option, if also present, will override the
2650 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2651 to be laid out as constant CoreFoundation strings.
2654 @opindex fgnu-runtime
2655 Generate object code compatible with the standard GNU Objective-C
2656 runtime. This is the default for most types of systems.
2658 @item -fnext-runtime
2659 @opindex fnext-runtime
2660 Generate output compatible with the NeXT runtime. This is the default
2661 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2662 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2665 @item -fno-nil-receivers
2666 @opindex fno-nil-receivers
2667 Assume that all Objective-C message dispatches (@code{[receiver
2668 message:arg]}) in this translation unit ensure that the receiver is
2669 not @code{nil}. This allows for more efficient entry points in the
2670 runtime to be used. This option is only available in conjunction with
2671 the NeXT runtime and ABI version 0 or 1.
2673 @item -fobjc-abi-version=@var{n}
2674 @opindex fobjc-abi-version
2675 Use version @var{n} of the Objective-C ABI for the selected runtime.
2676 This option is currently supported only for the NeXT runtime. In that
2677 case, Version 0 is the traditional (32-bit) ABI without support for
2678 properties and other Objective-C 2.0 additions. Version 1 is the
2679 traditional (32-bit) ABI with support for properties and other
2680 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2681 nothing is specified, the default is Version 0 on 32-bit target
2682 machines, and Version 2 on 64-bit target machines.
2684 @item -fobjc-call-cxx-cdtors
2685 @opindex fobjc-call-cxx-cdtors
2686 For each Objective-C class, check if any of its instance variables is a
2687 C++ object with a non-trivial default constructor. If so, synthesize a
2688 special @code{- (id) .cxx_construct} instance method which will run
2689 non-trivial default constructors on any such instance variables, in order,
2690 and then return @code{self}. Similarly, check if any instance variable
2691 is a C++ object with a non-trivial destructor, and if so, synthesize a
2692 special @code{- (void) .cxx_destruct} method which will run
2693 all such default destructors, in reverse order.
2695 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2696 methods thusly generated will only operate on instance variables
2697 declared in the current Objective-C class, and not those inherited
2698 from superclasses. It is the responsibility of the Objective-C
2699 runtime to invoke all such methods in an object's inheritance
2700 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2701 by the runtime immediately after a new object instance is allocated;
2702 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2703 before the runtime deallocates an object instance.
2705 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2706 support for invoking the @code{- (id) .cxx_construct} and
2707 @code{- (void) .cxx_destruct} methods.
2709 @item -fobjc-direct-dispatch
2710 @opindex fobjc-direct-dispatch
2711 Allow fast jumps to the message dispatcher. On Darwin this is
2712 accomplished via the comm page.
2714 @item -fobjc-exceptions
2715 @opindex fobjc-exceptions
2716 Enable syntactic support for structured exception handling in
2717 Objective-C, similar to what is offered by C++ and Java. This option
2718 is required to use the Objective-C keywords @code{@@try},
2719 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2720 @code{@@synchronized}. This option is available with both the GNU
2721 runtime and the NeXT runtime (but not available in conjunction with
2722 the NeXT runtime on Mac OS X 10.2 and earlier).
2726 Enable garbage collection (GC) in Objective-C and Objective-C++
2727 programs. This option is only available with the NeXT runtime; the
2728 GNU runtime has a different garbage collection implementation that
2729 does not require special compiler flags.
2731 @item -fobjc-nilcheck
2732 @opindex fobjc-nilcheck
2733 For the NeXT runtime with version 2 of the ABI, check for a nil
2734 receiver in method invocations before doing the actual method call.
2735 This is the default and can be disabled using
2736 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2737 checked for nil in this way no matter what this flag is set to.
2738 Currently this flag does nothing when the GNU runtime, or an older
2739 version of the NeXT runtime ABI, is used.
2741 @item -fobjc-std=objc1
2743 Conform to the language syntax of Objective-C 1.0, the language
2744 recognized by GCC 4.0. This only affects the Objective-C additions to
2745 the C/C++ language; it does not affect conformance to C/C++ standards,
2746 which is controlled by the separate C/C++ dialect option flags. When
2747 this option is used with the Objective-C or Objective-C++ compiler,
2748 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2749 This is useful if you need to make sure that your Objective-C code can
2750 be compiled with older versions of GCC.
2752 @item -freplace-objc-classes
2753 @opindex freplace-objc-classes
2754 Emit a special marker instructing @command{ld(1)} not to statically link in
2755 the resulting object file, and allow @command{dyld(1)} to load it in at
2756 run time instead. This is used in conjunction with the Fix-and-Continue
2757 debugging mode, where the object file in question may be recompiled and
2758 dynamically reloaded in the course of program execution, without the need
2759 to restart the program itself. Currently, Fix-and-Continue functionality
2760 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2765 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2766 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2767 compile time) with static class references that get initialized at load time,
2768 which improves run-time performance. Specifying the @option{-fzero-link} flag
2769 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2770 to be retained. This is useful in Zero-Link debugging mode, since it allows
2771 for individual class implementations to be modified during program execution.
2772 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2773 regardless of command-line options.
2777 Dump interface declarations for all classes seen in the source file to a
2778 file named @file{@var{sourcename}.decl}.
2780 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2781 @opindex Wassign-intercept
2782 @opindex Wno-assign-intercept
2783 Warn whenever an Objective-C assignment is being intercepted by the
2786 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2787 @opindex Wno-protocol
2789 If a class is declared to implement a protocol, a warning is issued for
2790 every method in the protocol that is not implemented by the class. The
2791 default behavior is to issue a warning for every method not explicitly
2792 implemented in the class, even if a method implementation is inherited
2793 from the superclass. If you use the @option{-Wno-protocol} option, then
2794 methods inherited from the superclass are considered to be implemented,
2795 and no warning is issued for them.
2797 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2799 @opindex Wno-selector
2800 Warn if multiple methods of different types for the same selector are
2801 found during compilation. The check is performed on the list of methods
2802 in the final stage of compilation. Additionally, a check is performed
2803 for each selector appearing in a @code{@@selector(@dots{})}
2804 expression, and a corresponding method for that selector has been found
2805 during compilation. Because these checks scan the method table only at
2806 the end of compilation, these warnings are not produced if the final
2807 stage of compilation is not reached, for example because an error is
2808 found during compilation, or because the @option{-fsyntax-only} option is
2811 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2812 @opindex Wstrict-selector-match
2813 @opindex Wno-strict-selector-match
2814 Warn if multiple methods with differing argument and/or return types are
2815 found for a given selector when attempting to send a message using this
2816 selector to a receiver of type @code{id} or @code{Class}. When this flag
2817 is off (which is the default behavior), the compiler will omit such warnings
2818 if any differences found are confined to types that share the same size
2821 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2822 @opindex Wundeclared-selector
2823 @opindex Wno-undeclared-selector
2824 Warn if a @code{@@selector(@dots{})} expression referring to an
2825 undeclared selector is found. A selector is considered undeclared if no
2826 method with that name has been declared before the
2827 @code{@@selector(@dots{})} expression, either explicitly in an
2828 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2829 an @code{@@implementation} section. This option always performs its
2830 checks as soon as a @code{@@selector(@dots{})} expression is found,
2831 while @option{-Wselector} only performs its checks in the final stage of
2832 compilation. This also enforces the coding style convention
2833 that methods and selectors must be declared before being used.
2835 @item -print-objc-runtime-info
2836 @opindex print-objc-runtime-info
2837 Generate C header describing the largest structure that is passed by
2842 @node Language Independent Options
2843 @section Options to Control Diagnostic Messages Formatting
2844 @cindex options to control diagnostics formatting
2845 @cindex diagnostic messages
2846 @cindex message formatting
2848 Traditionally, diagnostic messages have been formatted irrespective of
2849 the output device's aspect (e.g.@: its width, @dots{}). The options described
2850 below can be used to control the diagnostic messages formatting
2851 algorithm, e.g.@: how many characters per line, how often source location
2852 information should be reported. Right now, only the C++ front end can
2853 honor these options. However it is expected, in the near future, that
2854 the remaining front ends would be able to digest them correctly.
2857 @item -fmessage-length=@var{n}
2858 @opindex fmessage-length
2859 Try to format error messages so that they fit on lines of about @var{n}
2860 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2861 the front ends supported by GCC@. If @var{n} is zero, then no
2862 line-wrapping will be done; each error message will appear on a single
2865 @opindex fdiagnostics-show-location
2866 @item -fdiagnostics-show-location=once
2867 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2868 reporter to emit @emph{once} source location information; that is, in
2869 case the message is too long to fit on a single physical line and has to
2870 be wrapped, the source location won't be emitted (as prefix) again,
2871 over and over, in subsequent continuation lines. This is the default
2874 @item -fdiagnostics-show-location=every-line
2875 Only meaningful in line-wrapping mode. Instructs the diagnostic
2876 messages reporter to emit the same source location information (as
2877 prefix) for physical lines that result from the process of breaking
2878 a message which is too long to fit on a single line.
2880 @item -fno-diagnostics-show-option
2881 @opindex fno-diagnostics-show-option
2882 @opindex fdiagnostics-show-option
2883 By default, each diagnostic emitted includes text indicating the
2884 command-line option that directly controls the diagnostic (if such an
2885 option is known to the diagnostic machinery). Specifying the
2886 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2890 @node Warning Options
2891 @section Options to Request or Suppress Warnings
2892 @cindex options to control warnings
2893 @cindex warning messages
2894 @cindex messages, warning
2895 @cindex suppressing warnings
2897 Warnings are diagnostic messages that report constructions that
2898 are not inherently erroneous but that are risky or suggest there
2899 may have been an error.
2901 The following language-independent options do not enable specific
2902 warnings but control the kinds of diagnostics produced by GCC.
2905 @cindex syntax checking
2907 @opindex fsyntax-only
2908 Check the code for syntax errors, but don't do anything beyond that.
2910 @item -fmax-errors=@var{n}
2911 @opindex fmax-errors
2912 Limits the maximum number of error messages to @var{n}, at which point
2913 GCC bails out rather than attempting to continue processing the source
2914 code. If @var{n} is 0 (the default), there is no limit on the number
2915 of error messages produced. If @option{-Wfatal-errors} is also
2916 specified, then @option{-Wfatal-errors} takes precedence over this
2921 Inhibit all warning messages.
2926 Make all warnings into errors.
2931 Make the specified warning into an error. The specifier for a warning
2932 is appended, for example @option{-Werror=switch} turns the warnings
2933 controlled by @option{-Wswitch} into errors. This switch takes a
2934 negative form, to be used to negate @option{-Werror} for specific
2935 warnings, for example @option{-Wno-error=switch} makes
2936 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2939 The warning message for each controllable warning includes the
2940 option that controls the warning. That option can then be used with
2941 @option{-Werror=} and @option{-Wno-error=} as described above.
2942 (Printing of the option in the warning message can be disabled using the
2943 @option{-fno-diagnostics-show-option} flag.)
2945 Note that specifying @option{-Werror=}@var{foo} automatically implies
2946 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2949 @item -Wfatal-errors
2950 @opindex Wfatal-errors
2951 @opindex Wno-fatal-errors
2952 This option causes the compiler to abort compilation on the first error
2953 occurred rather than trying to keep going and printing further error
2958 You can request many specific warnings with options beginning
2959 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2960 implicit declarations. Each of these specific warning options also
2961 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2962 example, @option{-Wno-implicit}. This manual lists only one of the
2963 two forms, whichever is not the default. For further,
2964 language-specific options also refer to @ref{C++ Dialect Options} and
2965 @ref{Objective-C and Objective-C++ Dialect Options}.
2967 When an unrecognized warning option is requested (e.g.,
2968 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2969 that the option is not recognized. However, if the @option{-Wno-} form
2970 is used, the behavior is slightly different: No diagnostic will be
2971 produced for @option{-Wno-unknown-warning} unless other diagnostics
2972 are being produced. This allows the use of new @option{-Wno-} options
2973 with old compilers, but if something goes wrong, the compiler will
2974 warn that an unrecognized option was used.
2979 Issue all the warnings demanded by strict ISO C and ISO C++;
2980 reject all programs that use forbidden extensions, and some other
2981 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2982 version of the ISO C standard specified by any @option{-std} option used.
2984 Valid ISO C and ISO C++ programs should compile properly with or without
2985 this option (though a rare few will require @option{-ansi} or a
2986 @option{-std} option specifying the required version of ISO C)@. However,
2987 without this option, certain GNU extensions and traditional C and C++
2988 features are supported as well. With this option, they are rejected.
2990 @option{-pedantic} does not cause warning messages for use of the
2991 alternate keywords whose names begin and end with @samp{__}. Pedantic
2992 warnings are also disabled in the expression that follows
2993 @code{__extension__}. However, only system header files should use
2994 these escape routes; application programs should avoid them.
2995 @xref{Alternate Keywords}.
2997 Some users try to use @option{-pedantic} to check programs for strict ISO
2998 C conformance. They soon find that it does not do quite what they want:
2999 it finds some non-ISO practices, but not all---only those for which
3000 ISO C @emph{requires} a diagnostic, and some others for which
3001 diagnostics have been added.
3003 A feature to report any failure to conform to ISO C might be useful in
3004 some instances, but would require considerable additional work and would
3005 be quite different from @option{-pedantic}. We don't have plans to
3006 support such a feature in the near future.
3008 Where the standard specified with @option{-std} represents a GNU
3009 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3010 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3011 extended dialect is based. Warnings from @option{-pedantic} are given
3012 where they are required by the base standard. (It would not make sense
3013 for such warnings to be given only for features not in the specified GNU
3014 C dialect, since by definition the GNU dialects of C include all
3015 features the compiler supports with the given option, and there would be
3016 nothing to warn about.)
3018 @item -pedantic-errors
3019 @opindex pedantic-errors
3020 Like @option{-pedantic}, except that errors are produced rather than
3026 This enables all the warnings about constructions that some users
3027 consider questionable, and that are easy to avoid (or modify to
3028 prevent the warning), even in conjunction with macros. This also
3029 enables some language-specific warnings described in @ref{C++ Dialect
3030 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3032 @option{-Wall} turns on the following warning flags:
3034 @gccoptlist{-Waddress @gol
3035 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
3037 -Wchar-subscripts @gol
3038 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
3039 -Wimplicit-int @r{(C and Objective-C only)} @gol
3040 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3043 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3044 -Wmaybe-uninitialized @gol
3045 -Wmissing-braces @gol
3051 -Wsequence-point @gol
3052 -Wsign-compare @r{(only in C++)} @gol
3053 -Wstrict-aliasing @gol
3054 -Wstrict-overflow=1 @gol
3057 -Wuninitialized @gol
3058 -Wunknown-pragmas @gol
3059 -Wunused-function @gol
3062 -Wunused-variable @gol
3063 -Wvolatile-register-var @gol
3066 Note that some warning flags are not implied by @option{-Wall}. Some of
3067 them warn about constructions that users generally do not consider
3068 questionable, but which occasionally you might wish to check for;
3069 others warn about constructions that are necessary or hard to avoid in
3070 some cases, and there is no simple way to modify the code to suppress
3071 the warning. Some of them are enabled by @option{-Wextra} but many of
3072 them must be enabled individually.
3078 This enables some extra warning flags that are not enabled by
3079 @option{-Wall}. (This option used to be called @option{-W}. The older
3080 name is still supported, but the newer name is more descriptive.)
3082 @gccoptlist{-Wclobbered @gol
3084 -Wignored-qualifiers @gol
3085 -Wmissing-field-initializers @gol
3086 -Wmissing-parameter-type @r{(C only)} @gol
3087 -Wold-style-declaration @r{(C only)} @gol
3088 -Woverride-init @gol
3091 -Wuninitialized @gol
3092 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3093 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3096 The option @option{-Wextra} also prints warning messages for the
3102 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3103 @samp{>}, or @samp{>=}.
3106 (C++ only) An enumerator and a non-enumerator both appear in a
3107 conditional expression.
3110 (C++ only) Ambiguous virtual bases.
3113 (C++ only) Subscripting an array that has been declared @samp{register}.
3116 (C++ only) Taking the address of a variable that has been declared
3120 (C++ only) A base class is not initialized in a derived class' copy
3125 @item -Wchar-subscripts
3126 @opindex Wchar-subscripts
3127 @opindex Wno-char-subscripts
3128 Warn if an array subscript has type @code{char}. This is a common cause
3129 of error, as programmers often forget that this type is signed on some
3131 This warning is enabled by @option{-Wall}.
3135 @opindex Wno-comment
3136 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3137 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3138 This warning is enabled by @option{-Wall}.
3140 @item -Wno-coverage-mismatch
3141 @opindex Wno-coverage-mismatch
3142 Warn if feedback profiles do not match when using the
3143 @option{-fprofile-use} option.
3144 If a source file was changed between @option{-fprofile-gen} and
3145 @option{-fprofile-use}, the files with the profile feedback can fail
3146 to match the source file and GCC cannot use the profile feedback
3147 information. By default, this warning is enabled and is treated as an
3148 error. @option{-Wno-coverage-mismatch} can be used to disable the
3149 warning or @option{-Wno-error=coverage-mismatch} can be used to
3150 disable the error. Disabling the error for this warning can result in
3151 poorly optimized code and is useful only in the
3152 case of very minor changes such as bug fixes to an existing code-base.
3153 Completely disabling the warning is not recommended.
3156 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3158 Suppress warning messages emitted by @code{#warning} directives.
3160 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3161 @opindex Wdouble-promotion
3162 @opindex Wno-double-promotion
3163 Give a warning when a value of type @code{float} is implicitly
3164 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3165 floating-point unit implement @code{float} in hardware, but emulate
3166 @code{double} in software. On such a machine, doing computations
3167 using @code{double} values is much more expensive because of the
3168 overhead required for software emulation.
3170 It is easy to accidentally do computations with @code{double} because
3171 floating-point literals are implicitly of type @code{double}. For
3175 float area(float radius)
3177 return 3.14159 * radius * radius;
3181 the compiler will perform the entire computation with @code{double}
3182 because the floating-point literal is a @code{double}.
3187 @opindex ffreestanding
3188 @opindex fno-builtin
3189 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3190 the arguments supplied have types appropriate to the format string
3191 specified, and that the conversions specified in the format string make
3192 sense. This includes standard functions, and others specified by format
3193 attributes (@pxref{Function Attributes}), in the @code{printf},
3194 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3195 not in the C standard) families (or other target-specific families).
3196 Which functions are checked without format attributes having been
3197 specified depends on the standard version selected, and such checks of
3198 functions without the attribute specified are disabled by
3199 @option{-ffreestanding} or @option{-fno-builtin}.
3201 The formats are checked against the format features supported by GNU
3202 libc version 2.2. These include all ISO C90 and C99 features, as well
3203 as features from the Single Unix Specification and some BSD and GNU
3204 extensions. Other library implementations may not support all these
3205 features; GCC does not support warning about features that go beyond a
3206 particular library's limitations. However, if @option{-pedantic} is used
3207 with @option{-Wformat}, warnings will be given about format features not
3208 in the selected standard version (but not for @code{strfmon} formats,
3209 since those are not in any version of the C standard). @xref{C Dialect
3210 Options,,Options Controlling C Dialect}.
3212 Since @option{-Wformat} also checks for null format arguments for
3213 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3215 @option{-Wformat} is included in @option{-Wall}. For more control over some
3216 aspects of format checking, the options @option{-Wformat-y2k},
3217 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3218 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3219 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3222 @opindex Wformat-y2k
3223 @opindex Wno-format-y2k
3224 If @option{-Wformat} is specified, also warn about @code{strftime}
3225 formats that may yield only a two-digit year.
3227 @item -Wno-format-contains-nul
3228 @opindex Wno-format-contains-nul
3229 @opindex Wformat-contains-nul
3230 If @option{-Wformat} is specified, do not warn about format strings that
3233 @item -Wno-format-extra-args
3234 @opindex Wno-format-extra-args
3235 @opindex Wformat-extra-args
3236 If @option{-Wformat} is specified, do not warn about excess arguments to a
3237 @code{printf} or @code{scanf} format function. The C standard specifies
3238 that such arguments are ignored.
3240 Where the unused arguments lie between used arguments that are
3241 specified with @samp{$} operand number specifications, normally
3242 warnings are still given, since the implementation could not know what
3243 type to pass to @code{va_arg} to skip the unused arguments. However,
3244 in the case of @code{scanf} formats, this option will suppress the
3245 warning if the unused arguments are all pointers, since the Single
3246 Unix Specification says that such unused arguments are allowed.
3248 @item -Wno-format-zero-length
3249 @opindex Wno-format-zero-length
3250 @opindex Wformat-zero-length
3251 If @option{-Wformat} is specified, do not warn about zero-length formats.
3252 The C standard specifies that zero-length formats are allowed.
3254 @item -Wformat-nonliteral
3255 @opindex Wformat-nonliteral
3256 @opindex Wno-format-nonliteral
3257 If @option{-Wformat} is specified, also warn if the format string is not a
3258 string literal and so cannot be checked, unless the format function
3259 takes its format arguments as a @code{va_list}.
3261 @item -Wformat-security
3262 @opindex Wformat-security
3263 @opindex Wno-format-security
3264 If @option{-Wformat} is specified, also warn about uses of format
3265 functions that represent possible security problems. At present, this
3266 warns about calls to @code{printf} and @code{scanf} functions where the
3267 format string is not a string literal and there are no format arguments,
3268 as in @code{printf (foo);}. This may be a security hole if the format
3269 string came from untrusted input and contains @samp{%n}. (This is
3270 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3271 in future warnings may be added to @option{-Wformat-security} that are not
3272 included in @option{-Wformat-nonliteral}.)
3276 @opindex Wno-format=2
3277 Enable @option{-Wformat} plus format checks not included in
3278 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3279 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3283 @opindex Wno-nonnull
3284 Warn about passing a null pointer for arguments marked as
3285 requiring a non-null value by the @code{nonnull} function attribute.
3287 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3288 can be disabled with the @option{-Wno-nonnull} option.
3290 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3292 @opindex Wno-init-self
3293 Warn about uninitialized variables that are initialized with themselves.
3294 Note this option can only be used with the @option{-Wuninitialized} option.
3296 For example, GCC will warn about @code{i} being uninitialized in the
3297 following snippet only when @option{-Winit-self} has been specified:
3308 @item -Wimplicit-int @r{(C and Objective-C only)}
3309 @opindex Wimplicit-int
3310 @opindex Wno-implicit-int
3311 Warn when a declaration does not specify a type.
3312 This warning is enabled by @option{-Wall}.
3314 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3315 @opindex Wimplicit-function-declaration
3316 @opindex Wno-implicit-function-declaration
3317 Give a warning whenever a function is used before being declared. In
3318 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3319 enabled by default and it is made into an error by
3320 @option{-pedantic-errors}. This warning is also enabled by
3323 @item -Wimplicit @r{(C and Objective-C only)}
3325 @opindex Wno-implicit
3326 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3327 This warning is enabled by @option{-Wall}.
3329 @item -Wignored-qualifiers @r{(C and C++ only)}
3330 @opindex Wignored-qualifiers
3331 @opindex Wno-ignored-qualifiers
3332 Warn if the return type of a function has a type qualifier
3333 such as @code{const}. For ISO C such a type qualifier has no effect,
3334 since the value returned by a function is not an lvalue.
3335 For C++, the warning is only emitted for scalar types or @code{void}.
3336 ISO C prohibits qualified @code{void} return types on function
3337 definitions, so such return types always receive a warning
3338 even without this option.
3340 This warning is also enabled by @option{-Wextra}.
3345 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3346 a function with external linkage, returning int, taking either zero
3347 arguments, two, or three arguments of appropriate types. This warning
3348 is enabled by default in C++ and is enabled by either @option{-Wall}
3349 or @option{-pedantic}.
3351 @item -Wmissing-braces
3352 @opindex Wmissing-braces
3353 @opindex Wno-missing-braces
3354 Warn if an aggregate or union initializer is not fully bracketed. In
3355 the following example, the initializer for @samp{a} is not fully
3356 bracketed, but that for @samp{b} is fully bracketed.
3359 int a[2][2] = @{ 0, 1, 2, 3 @};
3360 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3363 This warning is enabled by @option{-Wall}.
3365 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3366 @opindex Wmissing-include-dirs
3367 @opindex Wno-missing-include-dirs
3368 Warn if a user-supplied include directory does not exist.
3371 @opindex Wparentheses
3372 @opindex Wno-parentheses
3373 Warn if parentheses are omitted in certain contexts, such
3374 as when there is an assignment in a context where a truth value
3375 is expected, or when operators are nested whose precedence people
3376 often get confused about.
3378 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3379 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3380 interpretation from that of ordinary mathematical notation.
3382 Also warn about constructions where there may be confusion to which
3383 @code{if} statement an @code{else} branch belongs. Here is an example of
3398 In C/C++, every @code{else} branch belongs to the innermost possible
3399 @code{if} statement, which in this example is @code{if (b)}. This is
3400 often not what the programmer expected, as illustrated in the above
3401 example by indentation the programmer chose. When there is the
3402 potential for this confusion, GCC will issue a warning when this flag
3403 is specified. To eliminate the warning, add explicit braces around
3404 the innermost @code{if} statement so there is no way the @code{else}
3405 could belong to the enclosing @code{if}. The resulting code would
3422 Also warn for dangerous uses of the
3423 ?: with omitted middle operand GNU extension. When the condition
3424 in the ?: operator is a boolean expression the omitted value will
3425 be always 1. Often the user expects it to be a value computed
3426 inside the conditional expression instead.
3428 This warning is enabled by @option{-Wall}.
3430 @item -Wsequence-point
3431 @opindex Wsequence-point
3432 @opindex Wno-sequence-point
3433 Warn about code that may have undefined semantics because of violations
3434 of sequence point rules in the C and C++ standards.
3436 The C and C++ standards defines the order in which expressions in a C/C++
3437 program are evaluated in terms of @dfn{sequence points}, which represent
3438 a partial ordering between the execution of parts of the program: those
3439 executed before the sequence point, and those executed after it. These
3440 occur after the evaluation of a full expression (one which is not part
3441 of a larger expression), after the evaluation of the first operand of a
3442 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3443 function is called (but after the evaluation of its arguments and the
3444 expression denoting the called function), and in certain other places.
3445 Other than as expressed by the sequence point rules, the order of
3446 evaluation of subexpressions of an expression is not specified. All
3447 these rules describe only a partial order rather than a total order,
3448 since, for example, if two functions are called within one expression
3449 with no sequence point between them, the order in which the functions
3450 are called is not specified. However, the standards committee have
3451 ruled that function calls do not overlap.
3453 It is not specified when between sequence points modifications to the
3454 values of objects take effect. Programs whose behavior depends on this
3455 have undefined behavior; the C and C++ standards specify that ``Between
3456 the previous and next sequence point an object shall have its stored
3457 value modified at most once by the evaluation of an expression.
3458 Furthermore, the prior value shall be read only to determine the value
3459 to be stored.''. If a program breaks these rules, the results on any
3460 particular implementation are entirely unpredictable.
3462 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3463 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3464 diagnosed by this option, and it may give an occasional false positive
3465 result, but in general it has been found fairly effective at detecting
3466 this sort of problem in programs.
3468 The standard is worded confusingly, therefore there is some debate
3469 over the precise meaning of the sequence point rules in subtle cases.
3470 Links to discussions of the problem, including proposed formal
3471 definitions, may be found on the GCC readings page, at
3472 @uref{http://gcc.gnu.org/@/readings.html}.
3474 This warning is enabled by @option{-Wall} for C and C++.
3477 @opindex Wreturn-type
3478 @opindex Wno-return-type
3479 Warn whenever a function is defined with a return-type that defaults
3480 to @code{int}. Also warn about any @code{return} statement with no
3481 return-value in a function whose return-type is not @code{void}
3482 (falling off the end of the function body is considered returning
3483 without a value), and about a @code{return} statement with an
3484 expression in a function whose return-type is @code{void}.
3486 For C++, a function without return type always produces a diagnostic
3487 message, even when @option{-Wno-return-type} is specified. The only
3488 exceptions are @samp{main} and functions defined in system headers.
3490 This warning is enabled by @option{-Wall}.
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. (The presence of a @code{default} label prevents this
3498 warning.) @code{case} labels outside the enumeration range also
3499 provoke warnings when this option is used (even if there is a
3500 @code{default} label).
3501 This warning is enabled by @option{-Wall}.
3503 @item -Wswitch-default
3504 @opindex Wswitch-default
3505 @opindex Wno-switch-default
3506 Warn whenever a @code{switch} statement does not have a @code{default}
3510 @opindex Wswitch-enum
3511 @opindex Wno-switch-enum
3512 Warn whenever a @code{switch} statement has an index of enumerated type
3513 and lacks a @code{case} for one or more of the named codes of that
3514 enumeration. @code{case} labels outside the enumeration range also
3515 provoke warnings when this option is used. The only difference
3516 between @option{-Wswitch} and this option is that this option gives a
3517 warning about an omitted enumeration code even if there is a
3518 @code{default} label.
3520 @item -Wsync-nand @r{(C and C++ only)}
3522 @opindex Wno-sync-nand
3523 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3524 built-in functions are used. These functions changed semantics in GCC 4.4.
3528 @opindex Wno-trigraphs
3529 Warn if any trigraphs are encountered that might change the meaning of
3530 the program (trigraphs within comments are not warned about).
3531 This warning is enabled by @option{-Wall}.
3533 @item -Wunused-but-set-parameter
3534 @opindex Wunused-but-set-parameter
3535 @opindex Wno-unused-but-set-parameter
3536 Warn whenever a function parameter is assigned to, but otherwise unused
3537 (aside from its declaration).
3539 To suppress this warning use the @samp{unused} attribute
3540 (@pxref{Variable Attributes}).
3542 This warning is also enabled by @option{-Wunused} together with
3545 @item -Wunused-but-set-variable
3546 @opindex Wunused-but-set-variable
3547 @opindex Wno-unused-but-set-variable
3548 Warn whenever a local variable is assigned to, but otherwise unused
3549 (aside from its declaration).
3550 This warning is enabled by @option{-Wall}.
3552 To suppress this warning use the @samp{unused} attribute
3553 (@pxref{Variable Attributes}).
3555 This warning is also enabled by @option{-Wunused}, which is enabled
3558 @item -Wunused-function
3559 @opindex Wunused-function
3560 @opindex Wno-unused-function
3561 Warn whenever a static function is declared but not defined or a
3562 non-inline static function is unused.
3563 This warning is enabled by @option{-Wall}.
3565 @item -Wunused-label
3566 @opindex Wunused-label
3567 @opindex Wno-unused-label
3568 Warn whenever a label is declared but not used.
3569 This warning is enabled by @option{-Wall}.
3571 To suppress this warning use the @samp{unused} attribute
3572 (@pxref{Variable Attributes}).
3574 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3575 @opindex Wunused-local-typedefs
3576 Warn when a typedef locally defined in a function is not used.
3578 @item -Wunused-parameter
3579 @opindex Wunused-parameter
3580 @opindex Wno-unused-parameter
3581 Warn whenever a function parameter is unused aside from its declaration.
3583 To suppress this warning use the @samp{unused} attribute
3584 (@pxref{Variable Attributes}).
3586 @item -Wno-unused-result
3587 @opindex Wunused-result
3588 @opindex Wno-unused-result
3589 Do not warn if a caller of a function marked with attribute
3590 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
3591 its return value. The default is @option{-Wunused-result}.
3593 @item -Wunused-variable
3594 @opindex Wunused-variable
3595 @opindex Wno-unused-variable
3596 Warn whenever a local variable or non-constant static variable is unused
3597 aside from its declaration.
3598 This warning is enabled by @option{-Wall}.
3600 To suppress this warning use the @samp{unused} attribute
3601 (@pxref{Variable Attributes}).
3603 @item -Wunused-value
3604 @opindex Wunused-value
3605 @opindex Wno-unused-value
3606 Warn whenever a statement computes a result that is explicitly not
3607 used. To suppress this warning cast the unused expression to
3608 @samp{void}. This includes an expression-statement or the left-hand
3609 side of a comma expression that contains no side effects. For example,
3610 an expression such as @samp{x[i,j]} will cause a warning, while
3611 @samp{x[(void)i,j]} will not.
3613 This warning is enabled by @option{-Wall}.
3618 All the above @option{-Wunused} options combined.
3620 In order to get a warning about an unused function parameter, you must
3621 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3622 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3624 @item -Wuninitialized
3625 @opindex Wuninitialized
3626 @opindex Wno-uninitialized
3627 Warn if an automatic variable is used without first being initialized
3628 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3629 warn if a non-static reference or non-static @samp{const} member
3630 appears in a class without constructors.
3632 If you want to warn about code that uses the uninitialized value of the
3633 variable in its own initializer, use the @option{-Winit-self} option.
3635 These warnings occur for individual uninitialized or clobbered
3636 elements of structure, union or array variables as well as for
3637 variables that are uninitialized or clobbered as a whole. They do
3638 not occur for variables or elements declared @code{volatile}. Because
3639 these warnings depend on optimization, the exact variables or elements
3640 for which there are warnings will depend on the precise optimization
3641 options and version of GCC used.
3643 Note that there may be no warning about a variable that is used only
3644 to compute a value that itself is never used, because such
3645 computations may be deleted by data flow analysis before the warnings
3648 @item -Wmaybe-uninitialized
3649 @opindex Wmaybe-uninitialized
3650 @opindex Wno-maybe-uninitialized
3651 For an automatic variable, if there exists a path from the function
3652 entry to a use of the variable that is initialized, but there exist
3653 some other paths the variable is not initialized, the compiler will
3654 emit a warning if it can not prove the uninitialized paths do not
3655 happen at run time. These warnings are made optional because GCC is
3656 not smart enough to see all the reasons why the code might be correct
3657 despite appearing to have an error. Here is one example of how
3678 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3679 always initialized, but GCC doesn't know this. To suppress the
3680 warning, the user needs to provide a default case with assert(0) or
3683 @cindex @code{longjmp} warnings
3684 This option also warns when a non-volatile automatic variable might be
3685 changed by a call to @code{longjmp}. These warnings as well are possible
3686 only in optimizing compilation.
3688 The compiler sees only the calls to @code{setjmp}. It cannot know
3689 where @code{longjmp} will be called; in fact, a signal handler could
3690 call it at any point in the code. As a result, you may get a warning
3691 even when there is in fact no problem because @code{longjmp} cannot
3692 in fact be called at the place that would cause a problem.
3694 Some spurious warnings can be avoided if you declare all the functions
3695 you use that never return as @code{noreturn}. @xref{Function
3698 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3700 @item -Wunknown-pragmas
3701 @opindex Wunknown-pragmas
3702 @opindex Wno-unknown-pragmas
3703 @cindex warning for unknown pragmas
3704 @cindex unknown pragmas, warning
3705 @cindex pragmas, warning of unknown
3706 Warn when a @code{#pragma} directive is encountered that is not understood by
3707 GCC@. If this command-line option is used, warnings will even be issued
3708 for unknown pragmas in system header files. This is not the case if
3709 the warnings were only enabled by the @option{-Wall} command-line option.
3712 @opindex Wno-pragmas
3714 Do not warn about misuses of pragmas, such as incorrect parameters,
3715 invalid syntax, or conflicts between pragmas. See also
3716 @samp{-Wunknown-pragmas}.
3718 @item -Wstrict-aliasing
3719 @opindex Wstrict-aliasing
3720 @opindex Wno-strict-aliasing
3721 This option is only active when @option{-fstrict-aliasing} is active.
3722 It warns about code that might break the strict aliasing rules that the
3723 compiler is using for optimization. The warning does not catch all
3724 cases, but does attempt to catch the more common pitfalls. It is
3725 included in @option{-Wall}.
3726 It is equivalent to @option{-Wstrict-aliasing=3}
3728 @item -Wstrict-aliasing=n
3729 @opindex Wstrict-aliasing=n
3730 @opindex Wno-strict-aliasing=n
3731 This option is only active when @option{-fstrict-aliasing} is active.
3732 It warns about code that might break the strict aliasing rules that the
3733 compiler is using for optimization.
3734 Higher levels correspond to higher accuracy (fewer false positives).
3735 Higher levels also correspond to more effort, similar to the way -O works.
3736 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3739 Level 1: Most aggressive, quick, least accurate.
3740 Possibly useful when higher levels
3741 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3742 false negatives. However, it has many false positives.
3743 Warns for all pointer conversions between possibly incompatible types,
3744 even if never dereferenced. Runs in the front end only.
3746 Level 2: Aggressive, quick, not too precise.
3747 May still have many false positives (not as many as level 1 though),
3748 and few false negatives (but possibly more than level 1).
3749 Unlike level 1, it only warns when an address is taken. Warns about
3750 incomplete types. Runs in the front end only.
3752 Level 3 (default for @option{-Wstrict-aliasing}):
3753 Should have very few false positives and few false
3754 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3755 Takes care of the common pun+dereference pattern in the front end:
3756 @code{*(int*)&some_float}.
3757 If optimization is enabled, it also runs in the back end, where it deals
3758 with multiple statement cases using flow-sensitive points-to information.
3759 Only warns when the converted pointer is dereferenced.
3760 Does not warn about incomplete types.
3762 @item -Wstrict-overflow
3763 @itemx -Wstrict-overflow=@var{n}
3764 @opindex Wstrict-overflow
3765 @opindex Wno-strict-overflow
3766 This option is only active when @option{-fstrict-overflow} is active.
3767 It warns about cases where the compiler optimizes based on the
3768 assumption that signed overflow does not occur. Note that it does not
3769 warn about all cases where the code might overflow: it only warns
3770 about cases where the compiler implements some optimization. Thus
3771 this warning depends on the optimization level.
3773 An optimization that assumes that signed overflow does not occur is
3774 perfectly safe if the values of the variables involved are such that
3775 overflow never does, in fact, occur. Therefore this warning can
3776 easily give a false positive: a warning about code that is not
3777 actually a problem. To help focus on important issues, several
3778 warning levels are defined. No warnings are issued for the use of
3779 undefined signed overflow when estimating how many iterations a loop
3780 will require, in particular when determining whether a loop will be
3784 @item -Wstrict-overflow=1
3785 Warn about cases that are both questionable and easy to avoid. For
3786 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3787 compiler will simplify this to @code{1}. This level of
3788 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3789 are not, and must be explicitly requested.
3791 @item -Wstrict-overflow=2
3792 Also warn about other cases where a comparison is simplified to a
3793 constant. For example: @code{abs (x) >= 0}. This can only be
3794 simplified when @option{-fstrict-overflow} is in effect, because
3795 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3796 zero. @option{-Wstrict-overflow} (with no level) is the same as
3797 @option{-Wstrict-overflow=2}.
3799 @item -Wstrict-overflow=3
3800 Also warn about other cases where a comparison is simplified. For
3801 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3803 @item -Wstrict-overflow=4
3804 Also warn about other simplifications not covered by the above cases.
3805 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3807 @item -Wstrict-overflow=5
3808 Also warn about cases where the compiler reduces the magnitude of a
3809 constant involved in a comparison. For example: @code{x + 2 > y} will
3810 be simplified to @code{x + 1 >= y}. This is reported only at the
3811 highest warning level because this simplification applies to many
3812 comparisons, so this warning level will give a very large number of
3816 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3817 @opindex Wsuggest-attribute=
3818 @opindex Wno-suggest-attribute=
3819 Warn for cases where adding an attribute may be beneficial. The
3820 attributes currently supported are listed below.
3823 @item -Wsuggest-attribute=pure
3824 @itemx -Wsuggest-attribute=const
3825 @itemx -Wsuggest-attribute=noreturn
3826 @opindex Wsuggest-attribute=pure
3827 @opindex Wno-suggest-attribute=pure
3828 @opindex Wsuggest-attribute=const
3829 @opindex Wno-suggest-attribute=const
3830 @opindex Wsuggest-attribute=noreturn
3831 @opindex Wno-suggest-attribute=noreturn
3833 Warn about functions that might be candidates for attributes
3834 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3835 functions visible in other compilation units or (in the case of @code{pure} and
3836 @code{const}) if it cannot prove that the function returns normally. A function
3837 returns normally if it doesn't contain an infinite loop nor returns abnormally
3838 by throwing, calling @code{abort()} or trapping. This analysis requires option
3839 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3840 higher. Higher optimization levels improve the accuracy of the analysis.
3843 @item -Warray-bounds
3844 @opindex Wno-array-bounds
3845 @opindex Warray-bounds
3846 This option is only active when @option{-ftree-vrp} is active
3847 (default for @option{-O2} and above). It warns about subscripts to arrays
3848 that are always out of bounds. This warning is enabled by @option{-Wall}.
3850 @item -Wno-div-by-zero
3851 @opindex Wno-div-by-zero
3852 @opindex Wdiv-by-zero
3853 Do not warn about compile-time integer division by zero. Floating-point
3854 division by zero is not warned about, as it can be a legitimate way of
3855 obtaining infinities and NaNs.
3857 @item -Wsystem-headers
3858 @opindex Wsystem-headers
3859 @opindex Wno-system-headers
3860 @cindex warnings from system headers
3861 @cindex system headers, warnings from
3862 Print warning messages for constructs found in system header files.
3863 Warnings from system headers are normally suppressed, on the assumption
3864 that they usually do not indicate real problems and would only make the
3865 compiler output harder to read. Using this command-line option tells
3866 GCC to emit warnings from system headers as if they occurred in user
3867 code. However, note that using @option{-Wall} in conjunction with this
3868 option will @emph{not} warn about unknown pragmas in system
3869 headers---for that, @option{-Wunknown-pragmas} must also be used.
3872 @opindex Wtrampolines
3873 @opindex Wno-trampolines
3874 Warn about trampolines generated for pointers to nested functions.
3876 A trampoline is a small piece of data or code that is created at run
3877 time on the stack when the address of a nested function is taken, and
3878 is used to call the nested function indirectly. For some targets, it
3879 is made up of data only and thus requires no special treatment. But,
3880 for most targets, it is made up of code and thus requires the stack
3881 to be made executable in order for the program to work properly.
3884 @opindex Wfloat-equal
3885 @opindex Wno-float-equal
3886 Warn if floating-point values are used in equality comparisons.
3888 The idea behind this is that sometimes it is convenient (for the
3889 programmer) to consider floating-point values as approximations to
3890 infinitely precise real numbers. If you are doing this, then you need
3891 to compute (by analyzing the code, or in some other way) the maximum or
3892 likely maximum error that the computation introduces, and allow for it
3893 when performing comparisons (and when producing output, but that's a
3894 different problem). In particular, instead of testing for equality, you
3895 would check to see whether the two values have ranges that overlap; and
3896 this is done with the relational operators, so equality comparisons are
3899 @item -Wtraditional @r{(C and Objective-C only)}
3900 @opindex Wtraditional
3901 @opindex Wno-traditional
3902 Warn about certain constructs that behave differently in traditional and
3903 ISO C@. Also warn about ISO C constructs that have no traditional C
3904 equivalent, and/or problematic constructs that should be avoided.
3908 Macro parameters that appear within string literals in the macro body.
3909 In traditional C macro replacement takes place within string literals,
3910 but does not in ISO C@.
3913 In traditional C, some preprocessor directives did not exist.
3914 Traditional preprocessors would only consider a line to be a directive
3915 if the @samp{#} appeared in column 1 on the line. Therefore
3916 @option{-Wtraditional} warns about directives that traditional C
3917 understands but would ignore because the @samp{#} does not appear as the
3918 first character on the line. It also suggests you hide directives like
3919 @samp{#pragma} not understood by traditional C by indenting them. Some
3920 traditional implementations would not recognize @samp{#elif}, so it
3921 suggests avoiding it altogether.
3924 A function-like macro that appears without arguments.
3927 The unary plus operator.
3930 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
3931 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3932 constants.) Note, these suffixes appear in macros defined in the system
3933 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3934 Use of these macros in user code might normally lead to spurious
3935 warnings, however GCC's integrated preprocessor has enough context to
3936 avoid warning in these cases.
3939 A function declared external in one block and then used after the end of
3943 A @code{switch} statement has an operand of type @code{long}.
3946 A non-@code{static} function declaration follows a @code{static} one.
3947 This construct is not accepted by some traditional C compilers.
3950 The ISO type of an integer constant has a different width or
3951 signedness from its traditional type. This warning is only issued if
3952 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3953 typically represent bit patterns, are not warned about.
3956 Usage of ISO string concatenation is detected.
3959 Initialization of automatic aggregates.
3962 Identifier conflicts with labels. Traditional C lacks a separate
3963 namespace for labels.
3966 Initialization of unions. If the initializer is zero, the warning is
3967 omitted. This is done under the assumption that the zero initializer in
3968 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3969 initializer warnings and relies on default initialization to zero in the
3973 Conversions by prototypes between fixed/floating-point values and vice
3974 versa. The absence of these prototypes when compiling with traditional
3975 C would cause serious problems. This is a subset of the possible
3976 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3979 Use of ISO C style function definitions. This warning intentionally is
3980 @emph{not} issued for prototype declarations or variadic functions
3981 because these ISO C features will appear in your code when using
3982 libiberty's traditional C compatibility macros, @code{PARAMS} and
3983 @code{VPARAMS}. This warning is also bypassed for nested functions
3984 because that feature is already a GCC extension and thus not relevant to
3985 traditional C compatibility.
3988 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3989 @opindex Wtraditional-conversion
3990 @opindex Wno-traditional-conversion
3991 Warn if a prototype causes a type conversion that is different from what
3992 would happen to the same argument in the absence of a prototype. This
3993 includes conversions of fixed point to floating and vice versa, and
3994 conversions changing the width or signedness of a fixed-point argument
3995 except when the same as the default promotion.
3997 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3998 @opindex Wdeclaration-after-statement
3999 @opindex Wno-declaration-after-statement
4000 Warn when a declaration is found after a statement in a block. This
4001 construct, known from C++, was introduced with ISO C99 and is by default
4002 allowed in GCC@. It is not supported by ISO C90 and was not supported by
4003 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
4008 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
4010 @item -Wno-endif-labels
4011 @opindex Wno-endif-labels
4012 @opindex Wendif-labels
4013 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
4018 Warn whenever a local variable or type declaration shadows another variable,
4019 parameter, type, or class member (in C++), or whenever a built-in function
4020 is shadowed. Note that in C++, the compiler will not warn if a local variable
4021 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
4023 @item -Wlarger-than=@var{len}
4024 @opindex Wlarger-than=@var{len}
4025 @opindex Wlarger-than-@var{len}
4026 Warn whenever an object of larger than @var{len} bytes is defined.
4028 @item -Wframe-larger-than=@var{len}
4029 @opindex Wframe-larger-than
4030 Warn if the size of a function frame is larger than @var{len} bytes.
4031 The computation done to determine the stack frame size is approximate
4032 and not conservative.
4033 The actual requirements may be somewhat greater than @var{len}
4034 even if you do not get a warning. In addition, any space allocated
4035 via @code{alloca}, variable-length arrays, or related constructs
4036 is not included by the compiler when determining
4037 whether or not to issue a warning.
4039 @item -Wno-free-nonheap-object
4040 @opindex Wno-free-nonheap-object
4041 @opindex Wfree-nonheap-object
4042 Do not warn when attempting to free an object that was not allocated
4045 @item -Wstack-usage=@var{len}
4046 @opindex Wstack-usage
4047 Warn if the stack usage of a function might be larger than @var{len} bytes.
4048 The computation done to determine the stack usage is conservative.
4049 Any space allocated via @code{alloca}, variable-length arrays, or related
4050 constructs is included by the compiler when determining whether or not to
4053 The message is in keeping with the output of @option{-fstack-usage}.
4057 If the stack usage is fully static but exceeds the specified amount, it's:
4060 warning: stack usage is 1120 bytes
4063 If the stack usage is (partly) dynamic but bounded, it's:
4066 warning: stack usage might be 1648 bytes
4069 If the stack usage is (partly) dynamic and not bounded, it's:
4072 warning: stack usage might be unbounded
4076 @item -Wunsafe-loop-optimizations
4077 @opindex Wunsafe-loop-optimizations
4078 @opindex Wno-unsafe-loop-optimizations
4079 Warn if the loop cannot be optimized because the compiler could not
4080 assume anything on the bounds of the loop indices. With
4081 @option{-funsafe-loop-optimizations} warn if the compiler made
4084 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4085 @opindex Wno-pedantic-ms-format
4086 @opindex Wpedantic-ms-format
4087 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4088 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4089 depending on the MS runtime, when you are using the options @option{-Wformat}
4090 and @option{-pedantic} without gnu-extensions.
4092 @item -Wpointer-arith
4093 @opindex Wpointer-arith
4094 @opindex Wno-pointer-arith
4095 Warn about anything that depends on the ``size of'' a function type or
4096 of @code{void}. GNU C assigns these types a size of 1, for
4097 convenience in calculations with @code{void *} pointers and pointers
4098 to functions. In C++, warn also when an arithmetic operation involves
4099 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4102 @opindex Wtype-limits
4103 @opindex Wno-type-limits
4104 Warn if a comparison is always true or always false due to the limited
4105 range of the data type, but do not warn for constant expressions. For
4106 example, warn if an unsigned variable is compared against zero with
4107 @samp{<} or @samp{>=}. This warning is also enabled by
4110 @item -Wbad-function-cast @r{(C and Objective-C only)}
4111 @opindex Wbad-function-cast
4112 @opindex Wno-bad-function-cast
4113 Warn whenever a function call is cast to a non-matching type.
4114 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4116 @item -Wc++-compat @r{(C and Objective-C only)}
4117 Warn about ISO C constructs that are outside of the common subset of
4118 ISO C and ISO C++, e.g.@: request for implicit conversion from
4119 @code{void *} to a pointer to non-@code{void} type.
4121 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4122 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4123 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4124 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4125 enabled by @option{-Wall}.
4129 @opindex Wno-cast-qual
4130 Warn whenever a pointer is cast so as to remove a type qualifier from
4131 the target type. For example, warn if a @code{const char *} is cast
4132 to an ordinary @code{char *}.
4134 Also warn when making a cast that introduces a type qualifier in an
4135 unsafe way. For example, casting @code{char **} to @code{const char **}
4136 is unsafe, as in this example:
4139 /* p is char ** value. */
4140 const char **q = (const char **) p;
4141 /* Assignment of readonly string to const char * is OK. */
4143 /* Now char** pointer points to read-only memory. */
4148 @opindex Wcast-align
4149 @opindex Wno-cast-align
4150 Warn whenever a pointer is cast such that the required alignment of the
4151 target is increased. For example, warn if a @code{char *} is cast to
4152 an @code{int *} on machines where integers can only be accessed at
4153 two- or four-byte boundaries.
4155 @item -Wwrite-strings
4156 @opindex Wwrite-strings
4157 @opindex Wno-write-strings
4158 When compiling C, give string constants the type @code{const
4159 char[@var{length}]} so that copying the address of one into a
4160 non-@code{const} @code{char *} pointer will get a warning. These
4161 warnings will help you find at compile time code that can try to write
4162 into a string constant, but only if you have been very careful about
4163 using @code{const} in declarations and prototypes. Otherwise, it will
4164 just be a nuisance. This is why we did not make @option{-Wall} request
4167 When compiling C++, warn about the deprecated conversion from string
4168 literals to @code{char *}. This warning is enabled by default for C++
4173 @opindex Wno-clobbered
4174 Warn for variables that might be changed by @samp{longjmp} or
4175 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4178 @opindex Wconversion
4179 @opindex Wno-conversion
4180 Warn for implicit conversions that may alter a value. This includes
4181 conversions between real and integer, like @code{abs (x)} when
4182 @code{x} is @code{double}; conversions between signed and unsigned,
4183 like @code{unsigned ui = -1}; and conversions to smaller types, like
4184 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4185 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4186 changed by the conversion like in @code{abs (2.0)}. Warnings about
4187 conversions between signed and unsigned integers can be disabled by
4188 using @option{-Wno-sign-conversion}.
4190 For C++, also warn for confusing overload resolution for user-defined
4191 conversions; and conversions that will never use a type conversion
4192 operator: conversions to @code{void}, the same type, a base class or a
4193 reference to them. Warnings about conversions between signed and
4194 unsigned integers are disabled by default in C++ unless
4195 @option{-Wsign-conversion} is explicitly enabled.
4197 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4198 @opindex Wconversion-null
4199 @opindex Wno-conversion-null
4200 Do not warn for conversions between @code{NULL} and non-pointer
4201 types. @option{-Wconversion-null} is enabled by default.
4203 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4204 @opindex Wzero-as-null-pointer-constant
4205 @opindex Wno-zero-as-null-pointer-constant
4206 Warn when a literal '0' is used as null pointer constant. This can
4207 be useful to facilitate the conversion to @code{nullptr} in C++11.
4210 @opindex Wempty-body
4211 @opindex Wno-empty-body
4212 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4213 while} statement. This warning is also enabled by @option{-Wextra}.
4215 @item -Wenum-compare
4216 @opindex Wenum-compare
4217 @opindex Wno-enum-compare
4218 Warn about a comparison between values of different enumerated types.
4219 In C++ enumeral mismatches in conditional expressions are also
4220 diagnosed and the warning is enabled by default. In C this warning is
4221 enabled by @option{-Wall}.
4223 @item -Wjump-misses-init @r{(C, Objective-C only)}
4224 @opindex Wjump-misses-init
4225 @opindex Wno-jump-misses-init
4226 Warn if a @code{goto} statement or a @code{switch} statement jumps
4227 forward across the initialization of a variable, or jumps backward to a
4228 label after the variable has been initialized. This only warns about
4229 variables that are initialized when they are declared. This warning is
4230 only supported for C and Objective-C; in C++ this sort of branch is an
4233 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4234 can be disabled with the @option{-Wno-jump-misses-init} option.
4236 @item -Wsign-compare
4237 @opindex Wsign-compare
4238 @opindex Wno-sign-compare
4239 @cindex warning for comparison of signed and unsigned values
4240 @cindex comparison of signed and unsigned values, warning
4241 @cindex signed and unsigned values, comparison warning
4242 Warn when a comparison between signed and unsigned values could produce
4243 an incorrect result when the signed value is converted to unsigned.
4244 This warning is also enabled by @option{-Wextra}; to get the other warnings
4245 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4247 @item -Wsign-conversion
4248 @opindex Wsign-conversion
4249 @opindex Wno-sign-conversion
4250 Warn for implicit conversions that may change the sign of an integer
4251 value, like assigning a signed integer expression to an unsigned
4252 integer variable. An explicit cast silences the warning. In C, this
4253 option is enabled also by @option{-Wconversion}.
4257 @opindex Wno-address
4258 Warn about suspicious uses of memory addresses. These include using
4259 the address of a function in a conditional expression, such as
4260 @code{void func(void); if (func)}, and comparisons against the memory
4261 address of a string literal, such as @code{if (x == "abc")}. Such
4262 uses typically indicate a programmer error: the address of a function
4263 always evaluates to true, so their use in a conditional usually
4264 indicate that the programmer forgot the parentheses in a function
4265 call; and comparisons against string literals result in unspecified
4266 behavior and are not portable in C, so they usually indicate that the
4267 programmer intended to use @code{strcmp}. This warning is enabled by
4271 @opindex Wlogical-op
4272 @opindex Wno-logical-op
4273 Warn about suspicious uses of logical operators in expressions.
4274 This includes using logical operators in contexts where a
4275 bit-wise operator is likely to be expected.
4277 @item -Waggregate-return
4278 @opindex Waggregate-return
4279 @opindex Wno-aggregate-return
4280 Warn if any functions that return structures or unions are defined or
4281 called. (In languages where you can return an array, this also elicits
4284 @item -Wno-attributes
4285 @opindex Wno-attributes
4286 @opindex Wattributes
4287 Do not warn if an unexpected @code{__attribute__} is used, such as
4288 unrecognized attributes, function attributes applied to variables,
4289 etc. This will not stop errors for incorrect use of supported
4292 @item -Wno-builtin-macro-redefined
4293 @opindex Wno-builtin-macro-redefined
4294 @opindex Wbuiltin-macro-redefined
4295 Do not warn if certain built-in macros are redefined. This suppresses
4296 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4297 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4299 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4300 @opindex Wstrict-prototypes
4301 @opindex Wno-strict-prototypes
4302 Warn if a function is declared or defined without specifying the
4303 argument types. (An old-style function definition is permitted without
4304 a warning if preceded by a declaration that specifies the argument
4307 @item -Wold-style-declaration @r{(C and Objective-C only)}
4308 @opindex Wold-style-declaration
4309 @opindex Wno-old-style-declaration
4310 Warn for obsolescent usages, according to the C Standard, in a
4311 declaration. For example, warn if storage-class specifiers like
4312 @code{static} are not the first things in a declaration. This warning
4313 is also enabled by @option{-Wextra}.
4315 @item -Wold-style-definition @r{(C and Objective-C only)}
4316 @opindex Wold-style-definition
4317 @opindex Wno-old-style-definition
4318 Warn if an old-style function definition is used. A warning is given
4319 even if there is a previous prototype.
4321 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4322 @opindex Wmissing-parameter-type
4323 @opindex Wno-missing-parameter-type
4324 A function parameter is declared without a type specifier in K&R-style
4331 This warning is also enabled by @option{-Wextra}.
4333 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4334 @opindex Wmissing-prototypes
4335 @opindex Wno-missing-prototypes
4336 Warn if a global function is defined without a previous prototype
4337 declaration. This warning is issued even if the definition itself
4338 provides a prototype. The aim is to detect global functions that
4339 are not declared in header files.
4341 @item -Wmissing-declarations
4342 @opindex Wmissing-declarations
4343 @opindex Wno-missing-declarations
4344 Warn if a global function is defined without a previous declaration.
4345 Do so even if the definition itself provides a prototype.
4346 Use this option to detect global functions that are not declared in
4347 header files. In C++, no warnings are issued for function templates,
4348 or for inline functions, or for functions in anonymous namespaces.
4350 @item -Wmissing-field-initializers
4351 @opindex Wmissing-field-initializers
4352 @opindex Wno-missing-field-initializers
4356 Warn if a structure's initializer has some fields missing. For
4357 example, the following code would cause such a warning, because
4358 @code{x.h} is implicitly zero:
4361 struct s @{ int f, g, h; @};
4362 struct s x = @{ 3, 4 @};
4365 This option does not warn about designated initializers, so the following
4366 modification would not trigger a warning:
4369 struct s @{ int f, g, h; @};
4370 struct s x = @{ .f = 3, .g = 4 @};
4373 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4374 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4376 @item -Wmissing-format-attribute
4377 @opindex Wmissing-format-attribute
4378 @opindex Wno-missing-format-attribute
4381 Warn about function pointers that might be candidates for @code{format}
4382 attributes. Note these are only possible candidates, not absolute ones.
4383 GCC will guess that function pointers with @code{format} attributes that
4384 are used in assignment, initialization, parameter passing or return
4385 statements should have a corresponding @code{format} attribute in the
4386 resulting type. I.e.@: the left-hand side of the assignment or
4387 initialization, the type of the parameter variable, or the return type
4388 of the containing function respectively should also have a @code{format}
4389 attribute to avoid the warning.
4391 GCC will also warn about function definitions that might be
4392 candidates for @code{format} attributes. Again, these are only
4393 possible candidates. GCC will guess that @code{format} attributes
4394 might be appropriate for any function that calls a function like
4395 @code{vprintf} or @code{vscanf}, but this might not always be the
4396 case, and some functions for which @code{format} attributes are
4397 appropriate may not be detected.
4399 @item -Wno-multichar
4400 @opindex Wno-multichar
4402 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4403 Usually they indicate a typo in the user's code, as they have
4404 implementation-defined values, and should not be used in portable code.
4406 @item -Wnormalized=<none|id|nfc|nfkc>
4407 @opindex Wnormalized=
4410 @cindex character set, input normalization
4411 In ISO C and ISO C++, two identifiers are different if they are
4412 different sequences of characters. However, sometimes when characters
4413 outside the basic ASCII character set are used, you can have two
4414 different character sequences that look the same. To avoid confusion,
4415 the ISO 10646 standard sets out some @dfn{normalization rules} which
4416 when applied ensure that two sequences that look the same are turned into
4417 the same sequence. GCC can warn you if you are using identifiers that
4418 have not been normalized; this option controls that warning.
4420 There are four levels of warning supported by GCC. The default is
4421 @option{-Wnormalized=nfc}, which warns about any identifier that is
4422 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4423 recommended form for most uses.
4425 Unfortunately, there are some characters allowed in identifiers by
4426 ISO C and ISO C++ that, when turned into NFC, are not allowed in
4427 identifiers. That is, there's no way to use these symbols in portable
4428 ISO C or C++ and have all your identifiers in NFC@.
4429 @option{-Wnormalized=id} suppresses the warning for these characters.
4430 It is hoped that future versions of the standards involved will correct
4431 this, which is why this option is not the default.
4433 You can switch the warning off for all characters by writing
4434 @option{-Wnormalized=none}. You would only want to do this if you
4435 were using some other normalization scheme (like ``D''), because
4436 otherwise you can easily create bugs that are literally impossible to see.
4438 Some characters in ISO 10646 have distinct meanings but look identical
4439 in some fonts or display methodologies, especially once formatting has
4440 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4441 LETTER N'', will display just like a regular @code{n} that has been
4442 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4443 normalization scheme to convert all these into a standard form as
4444 well, and GCC will warn if your code is not in NFKC if you use
4445 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4446 about every identifier that contains the letter O because it might be
4447 confused with the digit 0, and so is not the default, but may be
4448 useful as a local coding convention if the programming environment is
4449 unable to be fixed to display these characters distinctly.
4451 @item -Wno-deprecated
4452 @opindex Wno-deprecated
4453 @opindex Wdeprecated
4454 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4456 @item -Wno-deprecated-declarations
4457 @opindex Wno-deprecated-declarations
4458 @opindex Wdeprecated-declarations
4459 Do not warn about uses of functions (@pxref{Function Attributes}),
4460 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4461 Attributes}) marked as deprecated by using the @code{deprecated}
4465 @opindex Wno-overflow
4467 Do not warn about compile-time overflow in constant expressions.
4469 @item -Woverride-init @r{(C and Objective-C only)}
4470 @opindex Woverride-init
4471 @opindex Wno-override-init
4475 Warn if an initialized field without side effects is overridden when
4476 using designated initializers (@pxref{Designated Inits, , Designated
4479 This warning is included in @option{-Wextra}. To get other
4480 @option{-Wextra} warnings without this one, use @samp{-Wextra
4481 -Wno-override-init}.
4486 Warn if a structure is given the packed attribute, but the packed
4487 attribute has no effect on the layout or size of the structure.
4488 Such structures may be mis-aligned for little benefit. For
4489 instance, in this code, the variable @code{f.x} in @code{struct bar}
4490 will be misaligned even though @code{struct bar} does not itself
4491 have the packed attribute:
4498 @} __attribute__((packed));
4506 @item -Wpacked-bitfield-compat
4507 @opindex Wpacked-bitfield-compat
4508 @opindex Wno-packed-bitfield-compat
4509 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4510 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4511 the change can lead to differences in the structure layout. GCC
4512 informs you when the offset of such a field has changed in GCC 4.4.
4513 For example there is no longer a 4-bit padding between field @code{a}
4514 and @code{b} in this structure:
4521 @} __attribute__ ((packed));
4524 This warning is enabled by default. Use
4525 @option{-Wno-packed-bitfield-compat} to disable this warning.
4530 Warn if padding is included in a structure, either to align an element
4531 of the structure or to align the whole structure. Sometimes when this
4532 happens it is possible to rearrange the fields of the structure to
4533 reduce the padding and so make the structure smaller.
4535 @item -Wredundant-decls
4536 @opindex Wredundant-decls
4537 @opindex Wno-redundant-decls
4538 Warn if anything is declared more than once in the same scope, even in
4539 cases where multiple declaration is valid and changes nothing.
4541 @item -Wnested-externs @r{(C and Objective-C only)}
4542 @opindex Wnested-externs
4543 @opindex Wno-nested-externs
4544 Warn if an @code{extern} declaration is encountered within a function.
4549 Warn if a function can not be inlined and it was declared as inline.
4550 Even with this option, the compiler will not warn about failures to
4551 inline functions declared in system headers.
4553 The compiler uses a variety of heuristics to determine whether or not
4554 to inline a function. For example, the compiler takes into account
4555 the size of the function being inlined and the amount of inlining
4556 that has already been done in the current function. Therefore,
4557 seemingly insignificant changes in the source program can cause the
4558 warnings produced by @option{-Winline} to appear or disappear.
4560 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4561 @opindex Wno-invalid-offsetof
4562 @opindex Winvalid-offsetof
4563 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4564 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4565 to a non-POD type is undefined. In existing C++ implementations,
4566 however, @samp{offsetof} typically gives meaningful results even when
4567 applied to certain kinds of non-POD types. (Such as a simple
4568 @samp{struct} that fails to be a POD type only by virtue of having a
4569 constructor.) This flag is for users who are aware that they are
4570 writing nonportable code and who have deliberately chosen to ignore the
4573 The restrictions on @samp{offsetof} may be relaxed in a future version
4574 of the C++ standard.
4576 @item -Wno-int-to-pointer-cast
4577 @opindex Wno-int-to-pointer-cast
4578 @opindex Wint-to-pointer-cast
4579 Suppress warnings from casts to pointer type of an integer of a
4580 different size. In C++, casting to a pointer type of smaller size is
4581 an error. @option{Wint-to-pointer-cast} is enabled by default.
4584 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4585 @opindex Wno-pointer-to-int-cast
4586 @opindex Wpointer-to-int-cast
4587 Suppress warnings from casts from a pointer to an integer type of a
4591 @opindex Winvalid-pch
4592 @opindex Wno-invalid-pch
4593 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4594 the search path but can't be used.
4598 @opindex Wno-long-long
4599 Warn if @samp{long long} type is used. This is enabled by either
4600 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4601 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4603 @item -Wvariadic-macros
4604 @opindex Wvariadic-macros
4605 @opindex Wno-variadic-macros
4606 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4607 alternate syntax when in pedantic ISO C99 mode. This is default.
4608 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4610 @item -Wvector-operation-performance
4611 @opindex Wvector-operation-performance
4612 @opindex Wno-vector-operation-performance
4613 Warn if vector operation is not implemented via SIMD capabilities of the
4614 architecture. Mainly useful for the performance tuning.
4615 Vector operation can be implemented @code{piecewise}, which means that the
4616 scalar operation is performed on every vector element;
4617 @code{in parallel}, which means that the vector operation is implemented
4618 using scalars of wider type, which normally is more performance efficient;
4619 and @code{as a single scalar}, which means that vector fits into a
4625 Warn if variable length array is used in the code.
4626 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4627 the variable length array.
4629 @item -Wvolatile-register-var
4630 @opindex Wvolatile-register-var
4631 @opindex Wno-volatile-register-var
4632 Warn if a register variable is declared volatile. The volatile
4633 modifier does not inhibit all optimizations that may eliminate reads
4634 and/or writes to register variables. This warning is enabled by
4637 @item -Wdisabled-optimization
4638 @opindex Wdisabled-optimization
4639 @opindex Wno-disabled-optimization
4640 Warn if a requested optimization pass is disabled. This warning does
4641 not generally indicate that there is anything wrong with your code; it
4642 merely indicates that GCC's optimizers were unable to handle the code
4643 effectively. Often, the problem is that your code is too big or too
4644 complex; GCC will refuse to optimize programs when the optimization
4645 itself is likely to take inordinate amounts of time.
4647 @item -Wpointer-sign @r{(C and Objective-C only)}
4648 @opindex Wpointer-sign
4649 @opindex Wno-pointer-sign
4650 Warn for pointer argument passing or assignment with different signedness.
4651 This option is only supported for C and Objective-C@. It is implied by
4652 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4653 @option{-Wno-pointer-sign}.
4655 @item -Wstack-protector
4656 @opindex Wstack-protector
4657 @opindex Wno-stack-protector
4658 This option is only active when @option{-fstack-protector} is active. It
4659 warns about functions that will not be protected against stack smashing.
4662 @opindex Wno-mudflap
4663 Suppress warnings about constructs that cannot be instrumented by
4666 @item -Woverlength-strings
4667 @opindex Woverlength-strings
4668 @opindex Wno-overlength-strings
4669 Warn about string constants that are longer than the ``minimum
4670 maximum'' length specified in the C standard. Modern compilers
4671 generally allow string constants that are much longer than the
4672 standard's minimum limit, but very portable programs should avoid
4673 using longer strings.
4675 The limit applies @emph{after} string constant concatenation, and does
4676 not count the trailing NUL@. In C90, the limit was 509 characters; in
4677 C99, it was raised to 4095. C++98 does not specify a normative
4678 minimum maximum, so we do not diagnose overlength strings in C++@.
4680 This option is implied by @option{-pedantic}, and can be disabled with
4681 @option{-Wno-overlength-strings}.
4683 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4684 @opindex Wunsuffixed-float-constants
4686 GCC will issue a warning for any floating constant that does not have
4687 a suffix. When used together with @option{-Wsystem-headers} it will
4688 warn about such constants in system header files. This can be useful
4689 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4690 from the decimal floating-point extension to C99.
4693 @node Debugging Options
4694 @section Options for Debugging Your Program or GCC
4695 @cindex options, debugging
4696 @cindex debugging information options
4698 GCC has various special options that are used for debugging
4699 either your program or GCC:
4704 Produce debugging information in the operating system's native format
4705 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4708 On most systems that use stabs format, @option{-g} enables use of extra
4709 debugging information that only GDB can use; this extra information
4710 makes debugging work better in GDB but will probably make other debuggers
4712 refuse to read the program. If you want to control for certain whether
4713 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4714 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4716 GCC allows you to use @option{-g} with
4717 @option{-O}. The shortcuts taken by optimized code may occasionally
4718 produce surprising results: some variables you declared may not exist
4719 at all; flow of control may briefly move where you did not expect it;
4720 some statements may not be executed because they compute constant
4721 results or their values were already at hand; some statements may
4722 execute in different places because they were moved out of loops.
4724 Nevertheless it proves possible to debug optimized output. This makes
4725 it reasonable to use the optimizer for programs that might have bugs.
4727 The following options are useful when GCC is generated with the
4728 capability for more than one debugging format.
4732 Produce debugging information for use by GDB@. This means to use the
4733 most expressive format available (DWARF 2, stabs, or the native format
4734 if neither of those are supported), including GDB extensions if at all
4739 Produce debugging information in stabs format (if that is supported),
4740 without GDB extensions. This is the format used by DBX on most BSD
4741 systems. On MIPS, Alpha and System V Release 4 systems this option
4742 produces stabs debugging output that is not understood by DBX or SDB@.
4743 On System V Release 4 systems this option requires the GNU assembler.
4745 @item -feliminate-unused-debug-symbols
4746 @opindex feliminate-unused-debug-symbols
4747 Produce debugging information in stabs format (if that is supported),
4748 for only symbols that are actually used.
4750 @item -femit-class-debug-always
4751 Instead of emitting debugging information for a C++ class in only one
4752 object file, emit it in all object files using the class. This option
4753 should be used only with debuggers that are unable to handle the way GCC
4754 normally emits debugging information for classes because using this
4755 option will increase the size of debugging information by as much as a
4758 @item -fno-debug-types-section
4759 @opindex fno-debug-types-section
4760 @opindex fdebug-types-section
4761 By default when using DWARF v4 or higher type DIEs will be put into
4762 their own .debug_types section instead of making them part of the
4763 .debug_info section. It is more efficient to put them in a separate
4764 comdat sections since the linker will then be able to remove duplicates.
4765 But not all DWARF consumers support .debug_types sections yet.
4769 Produce debugging information in stabs 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.
4776 Produce debugging information in COFF format (if that is supported).
4777 This is the format used by SDB on most System V systems prior to
4782 Produce debugging information in XCOFF format (if that is supported).
4783 This is the format used by the DBX debugger on IBM RS/6000 systems.
4787 Produce debugging information in XCOFF format (if that is supported),
4788 using GNU extensions understood only by the GNU debugger (GDB)@. The
4789 use of these extensions is likely to make other debuggers crash or
4790 refuse to read the program, and may cause assemblers other than the GNU
4791 assembler (GAS) to fail with an error.
4793 @item -gdwarf-@var{version}
4794 @opindex gdwarf-@var{version}
4795 Produce debugging information in DWARF format (if that is
4796 supported). This is the format used by DBX on IRIX 6. The value
4797 of @var{version} may be either 2, 3 or 4; the default version is 2.
4799 Note that with DWARF version 2 some ports require, and will always
4800 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4802 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4803 for maximum benefit.
4805 @item -grecord-gcc-switches
4806 @opindex grecord-gcc-switches
4807 This switch causes the command-line options used to invoke the
4808 compiler that may affect code generation to be appended to the
4809 DW_AT_producer attribute in DWARF debugging information. The options
4810 are concatenated with spaces separating them from each other and from
4811 the compiler version. See also @option{-frecord-gcc-switches} for another
4812 way of storing compiler options into the object file.
4814 @item -gno-record-gcc-switches
4815 @opindex gno-record-gcc-switches
4816 Disallow appending command-line options to the DW_AT_producer attribute
4817 in DWARF debugging information. This is the default.
4819 @item -gstrict-dwarf
4820 @opindex gstrict-dwarf
4821 Disallow using extensions of later DWARF standard version than selected
4822 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4823 DWARF extensions from later standard versions is allowed.
4825 @item -gno-strict-dwarf
4826 @opindex gno-strict-dwarf
4827 Allow using extensions of later DWARF standard version than selected with
4828 @option{-gdwarf-@var{version}}.
4832 Produce debugging information in VMS debug format (if that is
4833 supported). This is the format used by DEBUG on VMS systems.
4836 @itemx -ggdb@var{level}
4837 @itemx -gstabs@var{level}
4838 @itemx -gcoff@var{level}
4839 @itemx -gxcoff@var{level}
4840 @itemx -gvms@var{level}
4841 Request debugging information and also use @var{level} to specify how
4842 much information. The default level is 2.
4844 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4847 Level 1 produces minimal information, enough for making backtraces in
4848 parts of the program that you don't plan to debug. This includes
4849 descriptions of functions and external variables, but no information
4850 about local variables and no line numbers.
4852 Level 3 includes extra information, such as all the macro definitions
4853 present in the program. Some debuggers support macro expansion when
4854 you use @option{-g3}.
4856 @option{-gdwarf-2} does not accept a concatenated debug level, because
4857 GCC used to support an option @option{-gdwarf} that meant to generate
4858 debug information in version 1 of the DWARF format (which is very
4859 different from version 2), and it would have been too confusing. That
4860 debug format is long obsolete, but the option cannot be changed now.
4861 Instead use an additional @option{-g@var{level}} option to change the
4862 debug level for DWARF.
4866 Turn off generation of debug info, if leaving out this option would have
4867 generated it, or turn it on at level 2 otherwise. The position of this
4868 argument in the command line does not matter, it takes effect after all
4869 other options are processed, and it does so only once, no matter how
4870 many times it is given. This is mainly intended to be used with
4871 @option{-fcompare-debug}.
4873 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4874 @opindex fdump-final-insns
4875 Dump the final internal representation (RTL) to @var{file}. If the
4876 optional argument is omitted (or if @var{file} is @code{.}), the name
4877 of the dump file will be determined by appending @code{.gkd} to the
4878 compilation output file name.
4880 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4881 @opindex fcompare-debug
4882 @opindex fno-compare-debug
4883 If no error occurs during compilation, run the compiler a second time,
4884 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4885 passed to the second compilation. Dump the final internal
4886 representation in both compilations, and print an error if they differ.
4888 If the equal sign is omitted, the default @option{-gtoggle} is used.
4890 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4891 and nonzero, implicitly enables @option{-fcompare-debug}. If
4892 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4893 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4896 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4897 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4898 of the final representation and the second compilation, preventing even
4899 @env{GCC_COMPARE_DEBUG} from taking effect.
4901 To verify full coverage during @option{-fcompare-debug} testing, set
4902 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4903 which GCC will reject as an invalid option in any actual compilation
4904 (rather than preprocessing, assembly or linking). To get just a
4905 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4906 not overridden} will do.
4908 @item -fcompare-debug-second
4909 @opindex fcompare-debug-second
4910 This option is implicitly passed to the compiler for the second
4911 compilation requested by @option{-fcompare-debug}, along with options to
4912 silence warnings, and omitting other options that would cause
4913 side-effect compiler outputs to files or to the standard output. Dump
4914 files and preserved temporary files are renamed so as to contain the
4915 @code{.gk} additional extension during the second compilation, to avoid
4916 overwriting those generated by the first.
4918 When this option is passed to the compiler driver, it causes the
4919 @emph{first} compilation to be skipped, which makes it useful for little
4920 other than debugging the compiler proper.
4922 @item -feliminate-dwarf2-dups
4923 @opindex feliminate-dwarf2-dups
4924 Compress DWARF2 debugging information by eliminating duplicated
4925 information about each symbol. This option only makes sense when
4926 generating DWARF2 debugging information with @option{-gdwarf-2}.
4928 @item -femit-struct-debug-baseonly
4929 Emit debug information for struct-like types
4930 only when the base name of the compilation source file
4931 matches the base name of file in which the struct was defined.
4933 This option substantially reduces the size of debugging information,
4934 but at significant potential loss in type information to the debugger.
4935 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4936 See @option{-femit-struct-debug-detailed} for more detailed control.
4938 This option works only with DWARF 2.
4940 @item -femit-struct-debug-reduced
4941 Emit debug information for struct-like types
4942 only when the base name of the compilation source file
4943 matches the base name of file in which the type was defined,
4944 unless the struct is a template or defined in a system header.
4946 This option significantly reduces the size of debugging information,
4947 with some potential loss in type information to the debugger.
4948 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4949 See @option{-femit-struct-debug-detailed} for more detailed control.
4951 This option works only with DWARF 2.
4953 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4954 Specify the struct-like types
4955 for which the compiler will generate debug information.
4956 The intent is to reduce duplicate struct debug information
4957 between different object files within the same program.
4959 This option is a detailed version of
4960 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4961 which will serve for most needs.
4963 A specification has the syntax@*
4964 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4966 The optional first word limits the specification to
4967 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4968 A struct type is used directly when it is the type of a variable, member.
4969 Indirect uses arise through pointers to structs.
4970 That is, when use of an incomplete struct would be legal, the use is indirect.
4972 @samp{struct one direct; struct two * indirect;}.
4974 The optional second word limits the specification to
4975 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4976 Generic structs are a bit complicated to explain.
4977 For C++, these are non-explicit specializations of template classes,
4978 or non-template classes within the above.
4979 Other programming languages have generics,
4980 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4982 The third word specifies the source files for those
4983 structs for which the compiler will emit debug information.
4984 The values @samp{none} and @samp{any} have the normal meaning.
4985 The value @samp{base} means that
4986 the base of name of the file in which the type declaration appears
4987 must match the base of the name of the main compilation file.
4988 In practice, this means that
4989 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4990 but types declared in other header will not.
4991 The value @samp{sys} means those types satisfying @samp{base}
4992 or declared in system or compiler headers.
4994 You may need to experiment to determine the best settings for your application.
4996 The default is @samp{-femit-struct-debug-detailed=all}.
4998 This option works only with DWARF 2.
5000 @item -fno-merge-debug-strings
5001 @opindex fmerge-debug-strings
5002 @opindex fno-merge-debug-strings
5003 Direct the linker to not merge together strings in the debugging
5004 information that are identical in different object files. Merging is
5005 not supported by all assemblers or linkers. Merging decreases the size
5006 of the debug information in the output file at the cost of increasing
5007 link processing time. Merging is enabled by default.
5009 @item -fdebug-prefix-map=@var{old}=@var{new}
5010 @opindex fdebug-prefix-map
5011 When compiling files in directory @file{@var{old}}, record debugging
5012 information describing them as in @file{@var{new}} instead.
5014 @item -fno-dwarf2-cfi-asm
5015 @opindex fdwarf2-cfi-asm
5016 @opindex fno-dwarf2-cfi-asm
5017 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
5018 instead of using GAS @code{.cfi_*} directives.
5020 @cindex @command{prof}
5023 Generate extra code to write profile information suitable for the
5024 analysis program @command{prof}. You must use this option when compiling
5025 the source files you want data about, and you must also use it when
5028 @cindex @command{gprof}
5031 Generate extra code to write profile information suitable for the
5032 analysis program @command{gprof}. You must use this option when compiling
5033 the source files you want data about, and you must also use it when
5038 Makes the compiler print out each function name as it is compiled, and
5039 print some statistics about each pass when it finishes.
5042 @opindex ftime-report
5043 Makes the compiler print some statistics about the time consumed by each
5044 pass when it finishes.
5047 @opindex fmem-report
5048 Makes the compiler print some statistics about permanent memory
5049 allocation when it finishes.
5051 @item -fpre-ipa-mem-report
5052 @opindex fpre-ipa-mem-report
5053 @item -fpost-ipa-mem-report
5054 @opindex fpost-ipa-mem-report
5055 Makes the compiler print some statistics about permanent memory
5056 allocation before or after interprocedural optimization.
5059 @opindex fstack-usage
5060 Makes the compiler output stack usage information for the program, on a
5061 per-function basis. The filename for the dump is made by appending
5062 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
5063 the output file, if explicitly specified and it is not an executable,
5064 otherwise it is the basename of the source file. An entry is made up
5069 The name of the function.
5073 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
5076 The qualifier @code{static} means that the function manipulates the stack
5077 statically: a fixed number of bytes are allocated for the frame on function
5078 entry and released on function exit; no stack adjustments are otherwise made
5079 in the function. The second field is this fixed number of bytes.
5081 The qualifier @code{dynamic} means that the function manipulates the stack
5082 dynamically: in addition to the static allocation described above, stack
5083 adjustments are made in the body of the function, for example to push/pop
5084 arguments around function calls. If the qualifier @code{bounded} is also
5085 present, the amount of these adjustments is bounded at compile time and
5086 the second field is an upper bound of the total amount of stack used by
5087 the function. If it is not present, the amount of these adjustments is
5088 not bounded at compile time and the second field only represents the
5091 @item -fprofile-arcs
5092 @opindex fprofile-arcs
5093 Add code so that program flow @dfn{arcs} are instrumented. During
5094 execution the program records how many times each branch and call is
5095 executed and how many times it is taken or returns. When the compiled
5096 program exits it saves this data to a file called
5097 @file{@var{auxname}.gcda} for each source file. The data may be used for
5098 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5099 test coverage analysis (@option{-ftest-coverage}). Each object file's
5100 @var{auxname} is generated from the name of the output file, if
5101 explicitly specified and it is not the final executable, otherwise it is
5102 the basename of the source file. In both cases any suffix is removed
5103 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5104 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5105 @xref{Cross-profiling}.
5107 @cindex @command{gcov}
5111 This option is used to compile and link code instrumented for coverage
5112 analysis. The option is a synonym for @option{-fprofile-arcs}
5113 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5114 linking). See the documentation for those options for more details.
5119 Compile the source files with @option{-fprofile-arcs} plus optimization
5120 and code generation options. For test coverage analysis, use the
5121 additional @option{-ftest-coverage} option. You do not need to profile
5122 every source file in a program.
5125 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5126 (the latter implies the former).
5129 Run the program on a representative workload to generate the arc profile
5130 information. This may be repeated any number of times. You can run
5131 concurrent instances of your program, and provided that the file system
5132 supports locking, the data files will be correctly updated. Also
5133 @code{fork} calls are detected and correctly handled (double counting
5137 For profile-directed optimizations, compile the source files again with
5138 the same optimization and code generation options plus
5139 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5140 Control Optimization}).
5143 For test coverage analysis, use @command{gcov} to produce human readable
5144 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5145 @command{gcov} documentation for further information.
5149 With @option{-fprofile-arcs}, for each function of your program GCC
5150 creates a program flow graph, then finds a spanning tree for the graph.
5151 Only arcs that are not on the spanning tree have to be instrumented: the
5152 compiler adds code to count the number of times that these arcs are
5153 executed. When an arc is the only exit or only entrance to a block, the
5154 instrumentation code can be added to the block; otherwise, a new basic
5155 block must be created to hold the instrumentation code.
5158 @item -ftest-coverage
5159 @opindex ftest-coverage
5160 Produce a notes file that the @command{gcov} code-coverage utility
5161 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5162 show program coverage. Each source file's note file is called
5163 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5164 above for a description of @var{auxname} and instructions on how to
5165 generate test coverage data. Coverage data will match the source files
5166 more closely, if you do not optimize.
5168 @item -fdbg-cnt-list
5169 @opindex fdbg-cnt-list
5170 Print the name and the counter upper bound for all debug counters.
5173 @item -fdbg-cnt=@var{counter-value-list}
5175 Set the internal debug counter upper bound. @var{counter-value-list}
5176 is a comma-separated list of @var{name}:@var{value} pairs
5177 which sets the upper bound of each debug counter @var{name} to @var{value}.
5178 All debug counters have the initial upper bound of @var{UINT_MAX},
5179 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5180 e.g. With -fdbg-cnt=dce:10,tail_call:0
5181 dbg_cnt(dce) will return true only for first 10 invocations
5183 @item -fenable-@var{kind}-@var{pass}
5184 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5188 This is a set of debugging options that are used to explicitly disable/enable
5189 optimization passes. For compiler users, regular options for enabling/disabling
5190 passes should be used instead.
5194 @item -fdisable-ipa-@var{pass}
5195 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5196 statically invoked in the compiler multiple times, the pass name should be
5197 appended with a sequential number starting from 1.
5199 @item -fdisable-rtl-@var{pass}
5200 @item -fdisable-rtl-@var{pass}=@var{range-list}
5201 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5202 statically invoked in the compiler multiple times, the pass name should be
5203 appended with a sequential number starting from 1. @var{range-list} is a comma
5204 seperated list of function ranges or assembler names. Each range is a number
5205 pair seperated by a colon. The range is inclusive in both ends. If the range
5206 is trivial, the number pair can be simplified as a single number. If the
5207 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5208 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5209 function header of a dump file, and the pass names can be dumped by using
5210 option @option{-fdump-passes}.
5212 @item -fdisable-tree-@var{pass}
5213 @item -fdisable-tree-@var{pass}=@var{range-list}
5214 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5217 @item -fenable-ipa-@var{pass}
5218 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5219 statically invoked in the compiler multiple times, the pass name should be
5220 appended with a sequential number starting from 1.
5222 @item -fenable-rtl-@var{pass}
5223 @item -fenable-rtl-@var{pass}=@var{range-list}
5224 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5225 description and examples.
5227 @item -fenable-tree-@var{pass}
5228 @item -fenable-tree-@var{pass}=@var{range-list}
5229 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5230 of option arguments.
5234 # disable ccp1 for all functions
5236 # disable complete unroll for function whose cgraph node uid is 1
5237 -fenable-tree-cunroll=1
5238 # disable gcse2 for functions at the following ranges [1,1],
5239 # [300,400], and [400,1000]
5240 # disable gcse2 for functions foo and foo2
5241 -fdisable-rtl-gcse2=foo,foo2
5242 # disable early inlining
5243 -fdisable-tree-einline
5244 # disable ipa inlining
5245 -fdisable-ipa-inline
5246 # enable tree full unroll
5247 -fenable-tree-unroll
5253 @item -d@var{letters}
5254 @itemx -fdump-rtl-@var{pass}
5256 Says to make debugging dumps during compilation at times specified by
5257 @var{letters}. This is used for debugging the RTL-based passes of the
5258 compiler. The file names for most of the dumps are made by appending
5259 a pass number and a word to the @var{dumpname}, and the files are
5260 created in the directory of the output file. Note that the pass
5261 number is computed statically as passes get registered into the pass
5262 manager. Thus the numbering is not related to the dynamic order of
5263 execution of passes. In particular, a pass installed by a plugin
5264 could have a number over 200 even if it executed quite early.
5265 @var{dumpname} is generated from the name of the output file, if
5266 explicitly specified and it is not an executable, otherwise it is the
5267 basename of the source file. These switches may have different effects
5268 when @option{-E} is used for preprocessing.
5270 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5271 @option{-d} option @var{letters}. Here are the possible
5272 letters for use in @var{pass} and @var{letters}, and their meanings:
5276 @item -fdump-rtl-alignments
5277 @opindex fdump-rtl-alignments
5278 Dump after branch alignments have been computed.
5280 @item -fdump-rtl-asmcons
5281 @opindex fdump-rtl-asmcons
5282 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5284 @item -fdump-rtl-auto_inc_dec
5285 @opindex fdump-rtl-auto_inc_dec
5286 Dump after auto-inc-dec discovery. This pass is only run on
5287 architectures that have auto inc or auto dec instructions.
5289 @item -fdump-rtl-barriers
5290 @opindex fdump-rtl-barriers
5291 Dump after cleaning up the barrier instructions.
5293 @item -fdump-rtl-bbpart
5294 @opindex fdump-rtl-bbpart
5295 Dump after partitioning hot and cold basic blocks.
5297 @item -fdump-rtl-bbro
5298 @opindex fdump-rtl-bbro
5299 Dump after block reordering.
5301 @item -fdump-rtl-btl1
5302 @itemx -fdump-rtl-btl2
5303 @opindex fdump-rtl-btl2
5304 @opindex fdump-rtl-btl2
5305 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5306 after the two branch
5307 target load optimization passes.
5309 @item -fdump-rtl-bypass
5310 @opindex fdump-rtl-bypass
5311 Dump after jump bypassing and control flow optimizations.
5313 @item -fdump-rtl-combine
5314 @opindex fdump-rtl-combine
5315 Dump after the RTL instruction combination pass.
5317 @item -fdump-rtl-compgotos
5318 @opindex fdump-rtl-compgotos
5319 Dump after duplicating the computed gotos.
5321 @item -fdump-rtl-ce1
5322 @itemx -fdump-rtl-ce2
5323 @itemx -fdump-rtl-ce3
5324 @opindex fdump-rtl-ce1
5325 @opindex fdump-rtl-ce2
5326 @opindex fdump-rtl-ce3
5327 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5328 @option{-fdump-rtl-ce3} enable dumping after the three
5329 if conversion passes.
5331 @item -fdump-rtl-cprop_hardreg
5332 @opindex fdump-rtl-cprop_hardreg
5333 Dump after hard register copy propagation.
5335 @item -fdump-rtl-csa
5336 @opindex fdump-rtl-csa
5337 Dump after combining stack adjustments.
5339 @item -fdump-rtl-cse1
5340 @itemx -fdump-rtl-cse2
5341 @opindex fdump-rtl-cse1
5342 @opindex fdump-rtl-cse2
5343 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5344 the two common sub-expression elimination passes.
5346 @item -fdump-rtl-dce
5347 @opindex fdump-rtl-dce
5348 Dump after the standalone dead code elimination passes.
5350 @item -fdump-rtl-dbr
5351 @opindex fdump-rtl-dbr
5352 Dump after delayed branch scheduling.
5354 @item -fdump-rtl-dce1
5355 @itemx -fdump-rtl-dce2
5356 @opindex fdump-rtl-dce1
5357 @opindex fdump-rtl-dce2
5358 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5359 the two dead store elimination passes.
5362 @opindex fdump-rtl-eh
5363 Dump after finalization of EH handling code.
5365 @item -fdump-rtl-eh_ranges
5366 @opindex fdump-rtl-eh_ranges
5367 Dump after conversion of EH handling range regions.
5369 @item -fdump-rtl-expand
5370 @opindex fdump-rtl-expand
5371 Dump after RTL generation.
5373 @item -fdump-rtl-fwprop1
5374 @itemx -fdump-rtl-fwprop2
5375 @opindex fdump-rtl-fwprop1
5376 @opindex fdump-rtl-fwprop2
5377 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5378 dumping after the two forward propagation passes.
5380 @item -fdump-rtl-gcse1
5381 @itemx -fdump-rtl-gcse2
5382 @opindex fdump-rtl-gcse1
5383 @opindex fdump-rtl-gcse2
5384 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5385 after global common subexpression elimination.
5387 @item -fdump-rtl-init-regs
5388 @opindex fdump-rtl-init-regs
5389 Dump after the initialization of the registers.
5391 @item -fdump-rtl-initvals
5392 @opindex fdump-rtl-initvals
5393 Dump after the computation of the initial value sets.
5395 @item -fdump-rtl-into_cfglayout
5396 @opindex fdump-rtl-into_cfglayout
5397 Dump after converting to cfglayout mode.
5399 @item -fdump-rtl-ira
5400 @opindex fdump-rtl-ira
5401 Dump after iterated register allocation.
5403 @item -fdump-rtl-jump
5404 @opindex fdump-rtl-jump
5405 Dump after the second jump optimization.
5407 @item -fdump-rtl-loop2
5408 @opindex fdump-rtl-loop2
5409 @option{-fdump-rtl-loop2} enables dumping after the rtl
5410 loop optimization passes.
5412 @item -fdump-rtl-mach
5413 @opindex fdump-rtl-mach
5414 Dump after performing the machine dependent reorganization pass, if that
5417 @item -fdump-rtl-mode_sw
5418 @opindex fdump-rtl-mode_sw
5419 Dump after removing redundant mode switches.
5421 @item -fdump-rtl-rnreg
5422 @opindex fdump-rtl-rnreg
5423 Dump after register renumbering.
5425 @item -fdump-rtl-outof_cfglayout
5426 @opindex fdump-rtl-outof_cfglayout
5427 Dump after converting from cfglayout mode.
5429 @item -fdump-rtl-peephole2
5430 @opindex fdump-rtl-peephole2
5431 Dump after the peephole pass.
5433 @item -fdump-rtl-postreload
5434 @opindex fdump-rtl-postreload
5435 Dump after post-reload optimizations.
5437 @item -fdump-rtl-pro_and_epilogue
5438 @opindex fdump-rtl-pro_and_epilogue
5439 Dump after generating the function prologues and epilogues.
5441 @item -fdump-rtl-regmove
5442 @opindex fdump-rtl-regmove
5443 Dump after the register move pass.
5445 @item -fdump-rtl-sched1
5446 @itemx -fdump-rtl-sched2
5447 @opindex fdump-rtl-sched1
5448 @opindex fdump-rtl-sched2
5449 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5450 after the basic block scheduling passes.
5452 @item -fdump-rtl-see
5453 @opindex fdump-rtl-see
5454 Dump after sign extension elimination.
5456 @item -fdump-rtl-seqabstr
5457 @opindex fdump-rtl-seqabstr
5458 Dump after common sequence discovery.
5460 @item -fdump-rtl-shorten
5461 @opindex fdump-rtl-shorten
5462 Dump after shortening branches.
5464 @item -fdump-rtl-sibling
5465 @opindex fdump-rtl-sibling
5466 Dump after sibling call optimizations.
5468 @item -fdump-rtl-split1
5469 @itemx -fdump-rtl-split2
5470 @itemx -fdump-rtl-split3
5471 @itemx -fdump-rtl-split4
5472 @itemx -fdump-rtl-split5
5473 @opindex fdump-rtl-split1
5474 @opindex fdump-rtl-split2
5475 @opindex fdump-rtl-split3
5476 @opindex fdump-rtl-split4
5477 @opindex fdump-rtl-split5
5478 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5479 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5480 @option{-fdump-rtl-split5} enable dumping after five rounds of
5481 instruction splitting.
5483 @item -fdump-rtl-sms
5484 @opindex fdump-rtl-sms
5485 Dump after modulo scheduling. This pass is only run on some
5488 @item -fdump-rtl-stack
5489 @opindex fdump-rtl-stack
5490 Dump after conversion from GCC's "flat register file" registers to the
5491 x87's stack-like registers. This pass is only run on x86 variants.
5493 @item -fdump-rtl-subreg1
5494 @itemx -fdump-rtl-subreg2
5495 @opindex fdump-rtl-subreg1
5496 @opindex fdump-rtl-subreg2
5497 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5498 the two subreg expansion passes.
5500 @item -fdump-rtl-unshare
5501 @opindex fdump-rtl-unshare
5502 Dump after all rtl has been unshared.
5504 @item -fdump-rtl-vartrack
5505 @opindex fdump-rtl-vartrack
5506 Dump after variable tracking.
5508 @item -fdump-rtl-vregs
5509 @opindex fdump-rtl-vregs
5510 Dump after converting virtual registers to hard registers.
5512 @item -fdump-rtl-web
5513 @opindex fdump-rtl-web
5514 Dump after live range splitting.
5516 @item -fdump-rtl-regclass
5517 @itemx -fdump-rtl-subregs_of_mode_init
5518 @itemx -fdump-rtl-subregs_of_mode_finish
5519 @itemx -fdump-rtl-dfinit
5520 @itemx -fdump-rtl-dfinish
5521 @opindex fdump-rtl-regclass
5522 @opindex fdump-rtl-subregs_of_mode_init
5523 @opindex fdump-rtl-subregs_of_mode_finish
5524 @opindex fdump-rtl-dfinit
5525 @opindex fdump-rtl-dfinish
5526 These dumps are defined but always produce empty files.
5529 @itemx -fdump-rtl-all
5531 @opindex fdump-rtl-all
5532 Produce all the dumps listed above.
5536 Annotate the assembler output with miscellaneous debugging information.
5540 Dump all macro definitions, at the end of preprocessing, in addition to
5545 Produce a core dump whenever an error occurs.
5549 Annotate the assembler output with a comment indicating which
5550 pattern and alternative was used. The length of each instruction is
5555 Dump the RTL in the assembler output as a comment before each instruction.
5556 Also turns on @option{-dp} annotation.
5560 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5561 dump a representation of the control flow graph suitable for viewing with VCG
5562 to @file{@var{file}.@var{pass}.vcg}.
5566 Just generate RTL for a function instead of compiling it. Usually used
5567 with @option{-fdump-rtl-expand}.
5571 @opindex fdump-noaddr
5572 When doing debugging dumps, suppress address output. This makes it more
5573 feasible to use diff on debugging dumps for compiler invocations with
5574 different compiler binaries and/or different
5575 text / bss / data / heap / stack / dso start locations.
5577 @item -fdump-unnumbered
5578 @opindex fdump-unnumbered
5579 When doing debugging dumps, suppress instruction numbers and address output.
5580 This makes it more feasible to use diff on debugging dumps for compiler
5581 invocations with different options, in particular with and without
5584 @item -fdump-unnumbered-links
5585 @opindex fdump-unnumbered-links
5586 When doing debugging dumps (see @option{-d} option above), suppress
5587 instruction numbers for the links to the previous and next instructions
5590 @item -fdump-translation-unit @r{(C++ only)}
5591 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5592 @opindex fdump-translation-unit
5593 Dump a representation of the tree structure for the entire translation
5594 unit to a file. The file name is made by appending @file{.tu} to the
5595 source file name, and the file is created in the same directory as the
5596 output file. If the @samp{-@var{options}} form is used, @var{options}
5597 controls the details of the dump as described for the
5598 @option{-fdump-tree} options.
5600 @item -fdump-class-hierarchy @r{(C++ only)}
5601 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5602 @opindex fdump-class-hierarchy
5603 Dump a representation of each class's hierarchy and virtual function
5604 table layout to a file. The file name is made by appending
5605 @file{.class} to the source file name, and the file is created in the
5606 same directory as the output file. If the @samp{-@var{options}} form
5607 is used, @var{options} controls the details of the dump as described
5608 for the @option{-fdump-tree} options.
5610 @item -fdump-ipa-@var{switch}
5612 Control the dumping at various stages of inter-procedural analysis
5613 language tree to a file. The file name is generated by appending a
5614 switch specific suffix to the source file name, and the file is created
5615 in the same directory as the output file. The following dumps are
5620 Enables all inter-procedural analysis dumps.
5623 Dumps information about call-graph optimization, unused function removal,
5624 and inlining decisions.
5627 Dump after function inlining.
5632 @opindex fdump-passes
5633 Dump the list of optimization passes that are turned on and off by
5634 the current command-line options.
5636 @item -fdump-statistics-@var{option}
5637 @opindex fdump-statistics
5638 Enable and control dumping of pass statistics in a separate file. The
5639 file name is generated by appending a suffix ending in
5640 @samp{.statistics} to the source file name, and the file is created in
5641 the same directory as the output file. If the @samp{-@var{option}}
5642 form is used, @samp{-stats} will cause counters to be summed over the
5643 whole compilation unit while @samp{-details} will dump every event as
5644 the passes generate them. The default with no option is to sum
5645 counters for each function compiled.
5647 @item -fdump-tree-@var{switch}
5648 @itemx -fdump-tree-@var{switch}-@var{options}
5650 Control the dumping at various stages of processing the intermediate
5651 language tree to a file. The file name is generated by appending a
5652 switch specific suffix to the source file name, and the file is
5653 created in the same directory as the output file. If the
5654 @samp{-@var{options}} form is used, @var{options} is a list of
5655 @samp{-} separated options which control the details of the dump. Not
5656 all options are applicable to all dumps; those that are not
5657 meaningful will be ignored. The following options are available
5661 Print the address of each node. Usually this is not meaningful as it
5662 changes according to the environment and source file. Its primary use
5663 is for tying up a dump file with a debug environment.
5665 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5666 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5667 use working backward from mangled names in the assembly file.
5669 Inhibit dumping of members of a scope or body of a function merely
5670 because that scope has been reached. Only dump such items when they
5671 are directly reachable by some other path. When dumping pretty-printed
5672 trees, this option inhibits dumping the bodies of control structures.
5674 Print a raw representation of the tree. By default, trees are
5675 pretty-printed into a C-like representation.
5677 Enable more detailed dumps (not honored by every dump option).
5679 Enable dumping various statistics about the pass (not honored by every dump
5682 Enable showing basic block boundaries (disabled in raw dumps).
5684 Enable showing virtual operands for every statement.
5686 Enable showing line numbers for statements.
5688 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5690 Enable showing the tree dump for each statement.
5692 Enable showing the EH region number holding each statement.
5694 Enable showing scalar evolution analysis details.
5696 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5697 and @option{lineno}.
5700 The following tree dumps are possible:
5704 @opindex fdump-tree-original
5705 Dump before any tree based optimization, to @file{@var{file}.original}.
5708 @opindex fdump-tree-optimized
5709 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5712 @opindex fdump-tree-gimple
5713 Dump each function before and after the gimplification pass to a file. The
5714 file name is made by appending @file{.gimple} to the source file name.
5717 @opindex fdump-tree-cfg
5718 Dump the control flow graph of each function to a file. The file name is
5719 made by appending @file{.cfg} to the source file name.
5722 @opindex fdump-tree-vcg
5723 Dump the control flow graph of each function to a file in VCG format. The
5724 file name is made by appending @file{.vcg} to the source file name. Note
5725 that if the file contains more than one function, the generated file cannot
5726 be used directly by VCG@. You will need to cut and paste each function's
5727 graph into its own separate file first.
5730 @opindex fdump-tree-ch
5731 Dump each function after copying loop headers. The file name is made by
5732 appending @file{.ch} to the source file name.
5735 @opindex fdump-tree-ssa
5736 Dump SSA related information to a file. The file name is made by appending
5737 @file{.ssa} to the source file name.
5740 @opindex fdump-tree-alias
5741 Dump aliasing information for each function. The file name is made by
5742 appending @file{.alias} to the source file name.
5745 @opindex fdump-tree-ccp
5746 Dump each function after CCP@. The file name is made by appending
5747 @file{.ccp} to the source file name.
5750 @opindex fdump-tree-storeccp
5751 Dump each function after STORE-CCP@. The file name is made by appending
5752 @file{.storeccp} to the source file name.
5755 @opindex fdump-tree-pre
5756 Dump trees after partial redundancy elimination. The file name is made
5757 by appending @file{.pre} to the source file name.
5760 @opindex fdump-tree-fre
5761 Dump trees after full redundancy elimination. The file name is made
5762 by appending @file{.fre} to the source file name.
5765 @opindex fdump-tree-copyprop
5766 Dump trees after copy propagation. The file name is made
5767 by appending @file{.copyprop} to the source file name.
5769 @item store_copyprop
5770 @opindex fdump-tree-store_copyprop
5771 Dump trees after store copy-propagation. The file name is made
5772 by appending @file{.store_copyprop} to the source file name.
5775 @opindex fdump-tree-dce
5776 Dump each function after dead code elimination. The file name is made by
5777 appending @file{.dce} to the source file name.
5780 @opindex fdump-tree-mudflap
5781 Dump each function after adding mudflap instrumentation. The file name is
5782 made by appending @file{.mudflap} to the source file name.
5785 @opindex fdump-tree-sra
5786 Dump each function after performing scalar replacement of aggregates. The
5787 file name is made by appending @file{.sra} to the source file name.
5790 @opindex fdump-tree-sink
5791 Dump each function after performing code sinking. The file name is made
5792 by appending @file{.sink} to the source file name.
5795 @opindex fdump-tree-dom
5796 Dump each function after applying dominator tree optimizations. The file
5797 name is made by appending @file{.dom} to the source file name.
5800 @opindex fdump-tree-dse
5801 Dump each function after applying dead store elimination. The file
5802 name is made by appending @file{.dse} to the source file name.
5805 @opindex fdump-tree-phiopt
5806 Dump each function after optimizing PHI nodes into straightline code. The file
5807 name is made by appending @file{.phiopt} to the source file name.
5810 @opindex fdump-tree-forwprop
5811 Dump each function after forward propagating single use variables. The file
5812 name is made by appending @file{.forwprop} to the source file name.
5815 @opindex fdump-tree-copyrename
5816 Dump each function after applying the copy rename optimization. The file
5817 name is made by appending @file{.copyrename} to the source file name.
5820 @opindex fdump-tree-nrv
5821 Dump each function after applying the named return value optimization on
5822 generic trees. The file name is made by appending @file{.nrv} to the source
5826 @opindex fdump-tree-vect
5827 Dump each function after applying vectorization of loops. The file name is
5828 made by appending @file{.vect} to the source file name.
5831 @opindex fdump-tree-slp
5832 Dump each function after applying vectorization of basic blocks. The file name
5833 is made by appending @file{.slp} to the source file name.
5836 @opindex fdump-tree-vrp
5837 Dump each function after Value Range Propagation (VRP). The file name
5838 is made by appending @file{.vrp} to the source file name.
5841 @opindex fdump-tree-all
5842 Enable all the available tree dumps with the flags provided in this option.
5845 @item -ftree-vectorizer-verbose=@var{n}
5846 @opindex ftree-vectorizer-verbose
5847 This option controls the amount of debugging output the vectorizer prints.
5848 This information is written to standard error, unless
5849 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5850 in which case it is output to the usual dump listing file, @file{.vect}.
5851 For @var{n}=0 no diagnostic information is reported.
5852 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5853 and the total number of loops that got vectorized.
5854 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5855 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5856 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5857 level that @option{-fdump-tree-vect-stats} uses.
5858 Higher verbosity levels mean either more information dumped for each
5859 reported loop, or same amount of information reported for more loops:
5860 if @var{n}=3, vectorizer cost model information is reported.
5861 If @var{n}=4, alignment related information is added to the reports.
5862 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5863 memory access-patterns) is added to the reports.
5864 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5865 that did not pass the first analysis phase (i.e., may not be countable, or
5866 may have complicated control-flow).
5867 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5868 If @var{n}=8, SLP related information is added to the reports.
5869 For @var{n}=9, all the information the vectorizer generates during its
5870 analysis and transformation is reported. This is the same verbosity level
5871 that @option{-fdump-tree-vect-details} uses.
5873 @item -frandom-seed=@var{string}
5874 @opindex frandom-seed
5875 This option provides a seed that GCC uses when it would otherwise use
5876 random numbers. It is used to generate certain symbol names
5877 that have to be different in every compiled file. It is also used to
5878 place unique stamps in coverage data files and the object files that
5879 produce them. You can use the @option{-frandom-seed} option to produce
5880 reproducibly identical object files.
5882 The @var{string} should be different for every file you compile.
5884 @item -fsched-verbose=@var{n}
5885 @opindex fsched-verbose
5886 On targets that use instruction scheduling, this option controls the
5887 amount of debugging output the scheduler prints. This information is
5888 written to standard error, unless @option{-fdump-rtl-sched1} or
5889 @option{-fdump-rtl-sched2} is specified, in which case it is output
5890 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5891 respectively. However for @var{n} greater than nine, the output is
5892 always printed to standard error.
5894 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5895 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5896 For @var{n} greater than one, it also output basic block probabilities,
5897 detailed ready list information and unit/insn info. For @var{n} greater
5898 than two, it includes RTL at abort point, control-flow and regions info.
5899 And for @var{n} over four, @option{-fsched-verbose} also includes
5903 @itemx -save-temps=cwd
5905 Store the usual ``temporary'' intermediate files permanently; place them
5906 in the current directory and name them based on the source file. Thus,
5907 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5908 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5909 preprocessed @file{foo.i} output file even though the compiler now
5910 normally uses an integrated preprocessor.
5912 When used in combination with the @option{-x} command-line option,
5913 @option{-save-temps} is sensible enough to avoid over writing an
5914 input source file with the same extension as an intermediate file.
5915 The corresponding intermediate file may be obtained by renaming the
5916 source file before using @option{-save-temps}.
5918 If you invoke GCC in parallel, compiling several different source
5919 files that share a common base name in different subdirectories or the
5920 same source file compiled for multiple output destinations, it is
5921 likely that the different parallel compilers will interfere with each
5922 other, and overwrite the temporary files. For instance:
5925 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5926 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5929 may result in @file{foo.i} and @file{foo.o} being written to
5930 simultaneously by both compilers.
5932 @item -save-temps=obj
5933 @opindex save-temps=obj
5934 Store the usual ``temporary'' intermediate files permanently. If the
5935 @option{-o} option is used, the temporary files are based on the
5936 object file. If the @option{-o} option is not used, the
5937 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5942 gcc -save-temps=obj -c foo.c
5943 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5944 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5947 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5948 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5949 @file{dir2/yfoobar.o}.
5951 @item -time@r{[}=@var{file}@r{]}
5953 Report the CPU time taken by each subprocess in the compilation
5954 sequence. For C source files, this is the compiler proper and assembler
5955 (plus the linker if linking is done).
5957 Without the specification of an output file, the output looks like this:
5964 The first number on each line is the ``user time'', that is time spent
5965 executing the program itself. The second number is ``system time'',
5966 time spent executing operating system routines on behalf of the program.
5967 Both numbers are in seconds.
5969 With the specification of an output file, the output is appended to the
5970 named file, and it looks like this:
5973 0.12 0.01 cc1 @var{options}
5974 0.00 0.01 as @var{options}
5977 The ``user time'' and the ``system time'' are moved before the program
5978 name, and the options passed to the program are displayed, so that one
5979 can later tell what file was being compiled, and with which options.
5981 @item -fvar-tracking
5982 @opindex fvar-tracking
5983 Run variable tracking pass. It computes where variables are stored at each
5984 position in code. Better debugging information is then generated
5985 (if the debugging information format supports this information).
5987 It is enabled by default when compiling with optimization (@option{-Os},
5988 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5989 the debug info format supports it.
5991 @item -fvar-tracking-assignments
5992 @opindex fvar-tracking-assignments
5993 @opindex fno-var-tracking-assignments
5994 Annotate assignments to user variables early in the compilation and
5995 attempt to carry the annotations over throughout the compilation all the
5996 way to the end, in an attempt to improve debug information while
5997 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5999 It can be enabled even if var-tracking is disabled, in which case
6000 annotations will be created and maintained, but discarded at the end.
6002 @item -fvar-tracking-assignments-toggle
6003 @opindex fvar-tracking-assignments-toggle
6004 @opindex fno-var-tracking-assignments-toggle
6005 Toggle @option{-fvar-tracking-assignments}, in the same way that
6006 @option{-gtoggle} toggles @option{-g}.
6008 @item -print-file-name=@var{library}
6009 @opindex print-file-name
6010 Print the full absolute name of the library file @var{library} that
6011 would be used when linking---and don't do anything else. With this
6012 option, GCC does not compile or link anything; it just prints the
6015 @item -print-multi-directory
6016 @opindex print-multi-directory
6017 Print the directory name corresponding to the multilib selected by any
6018 other switches present in the command line. This directory is supposed
6019 to exist in @env{GCC_EXEC_PREFIX}.
6021 @item -print-multi-lib
6022 @opindex print-multi-lib
6023 Print the mapping from multilib directory names to compiler switches
6024 that enable them. The directory name is separated from the switches by
6025 @samp{;}, and each switch starts with an @samp{@@} instead of the
6026 @samp{-}, without spaces between multiple switches. This is supposed to
6027 ease shell-processing.
6029 @item -print-multi-os-directory
6030 @opindex print-multi-os-directory
6031 Print the path to OS libraries for the selected
6032 multilib, relative to some @file{lib} subdirectory. If OS libraries are
6033 present in the @file{lib} subdirectory and no multilibs are used, this is
6034 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
6035 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
6036 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
6037 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
6039 @item -print-multiarch
6040 @opindex print-multiarch
6041 Print the path to OS libraries for the selected multiarch,
6042 relative to some @file{lib} subdirectory.
6044 @item -print-prog-name=@var{program}
6045 @opindex print-prog-name
6046 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
6048 @item -print-libgcc-file-name
6049 @opindex print-libgcc-file-name
6050 Same as @option{-print-file-name=libgcc.a}.
6052 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
6053 but you do want to link with @file{libgcc.a}. You can do
6056 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
6059 @item -print-search-dirs
6060 @opindex print-search-dirs
6061 Print the name of the configured installation directory and a list of
6062 program and library directories @command{gcc} will search---and don't do anything else.
6064 This is useful when @command{gcc} prints the error message
6065 @samp{installation problem, cannot exec cpp0: No such file or directory}.
6066 To resolve this you either need to put @file{cpp0} and the other compiler
6067 components where @command{gcc} expects to find them, or you can set the environment
6068 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
6069 Don't forget the trailing @samp{/}.
6070 @xref{Environment Variables}.
6072 @item -print-sysroot
6073 @opindex print-sysroot
6074 Print the target sysroot directory that will be used during
6075 compilation. This is the target sysroot specified either at configure
6076 time or using the @option{--sysroot} option, possibly with an extra
6077 suffix that depends on compilation options. If no target sysroot is
6078 specified, the option prints nothing.
6080 @item -print-sysroot-headers-suffix
6081 @opindex print-sysroot-headers-suffix
6082 Print the suffix added to the target sysroot when searching for
6083 headers, or give an error if the compiler is not configured with such
6084 a suffix---and don't do anything else.
6087 @opindex dumpmachine
6088 Print the compiler's target machine (for example,
6089 @samp{i686-pc-linux-gnu})---and don't do anything else.
6092 @opindex dumpversion
6093 Print the compiler version (for example, @samp{3.0})---and don't do
6098 Print the compiler's built-in specs---and don't do anything else. (This
6099 is used when GCC itself is being built.) @xref{Spec Files}.
6101 @item -feliminate-unused-debug-types
6102 @opindex feliminate-unused-debug-types
6103 Normally, when producing DWARF2 output, GCC will emit debugging
6104 information for all types declared in a compilation
6105 unit, regardless of whether or not they are actually used
6106 in that compilation unit. Sometimes this is useful, such as
6107 if, in the debugger, you want to cast a value to a type that is
6108 not actually used in your program (but is declared). More often,
6109 however, this results in a significant amount of wasted space.
6110 With this option, GCC will avoid producing debug symbol output
6111 for types that are nowhere used in the source file being compiled.
6114 @node Optimize Options
6115 @section Options That Control Optimization
6116 @cindex optimize options
6117 @cindex options, optimization
6119 These options control various sorts of optimizations.
6121 Without any optimization option, the compiler's goal is to reduce the
6122 cost of compilation and to make debugging produce the expected
6123 results. Statements are independent: if you stop the program with a
6124 breakpoint between statements, you can then assign a new value to any
6125 variable or change the program counter to any other statement in the
6126 function and get exactly the results you would expect from the source
6129 Turning on optimization flags makes the compiler attempt to improve
6130 the performance and/or code size at the expense of compilation time
6131 and possibly the ability to debug the program.
6133 The compiler performs optimization based on the knowledge it has of the
6134 program. Compiling multiple files at once to a single output file mode allows
6135 the compiler to use information gained from all of the files when compiling
6138 Not all optimizations are controlled directly by a flag. Only
6139 optimizations that have a flag are listed in this section.
6141 Most optimizations are only enabled if an @option{-O} level is set on
6142 the command line. Otherwise they are disabled, even if individual
6143 optimization flags are specified.
6145 Depending on the target and how GCC was configured, a slightly different
6146 set of optimizations may be enabled at each @option{-O} level than
6147 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6148 to find out the exact set of optimizations that are enabled at each level.
6149 @xref{Overall Options}, for examples.
6156 Optimize. Optimizing compilation takes somewhat more time, and a lot
6157 more memory for a large function.
6159 With @option{-O}, the compiler tries to reduce code size and execution
6160 time, without performing any optimizations that take a great deal of
6163 @option{-O} turns on the following optimization flags:
6167 -fcprop-registers @gol
6170 -fdelayed-branch @gol
6172 -fguess-branch-probability @gol
6173 -fif-conversion2 @gol
6174 -fif-conversion @gol
6175 -fipa-pure-const @gol
6177 -fipa-reference @gol
6179 -fsplit-wide-types @gol
6181 -ftree-builtin-call-dce @gol
6184 -ftree-copyrename @gol
6186 -ftree-dominator-opts @gol
6188 -ftree-forwprop @gol
6196 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6197 where doing so does not interfere with debugging.
6201 Optimize even more. GCC performs nearly all supported optimizations
6202 that do not involve a space-speed tradeoff.
6203 As compared to @option{-O}, this option increases both compilation time
6204 and the performance of the generated code.
6206 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6207 also turns on the following optimization flags:
6208 @gccoptlist{-fthread-jumps @gol
6209 -falign-functions -falign-jumps @gol
6210 -falign-loops -falign-labels @gol
6213 -fcse-follow-jumps -fcse-skip-blocks @gol
6214 -fdelete-null-pointer-checks @gol
6216 -fexpensive-optimizations @gol
6217 -fgcse -fgcse-lm @gol
6218 -finline-small-functions @gol
6219 -findirect-inlining @gol
6221 -foptimize-sibling-calls @gol
6222 -fpartial-inlining @gol
6225 -freorder-blocks -freorder-functions @gol
6226 -frerun-cse-after-loop @gol
6227 -fsched-interblock -fsched-spec @gol
6228 -fschedule-insns -fschedule-insns2 @gol
6229 -fstrict-aliasing -fstrict-overflow @gol
6230 -ftree-switch-conversion -ftree-tail-merge @gol
6234 Please note the warning under @option{-fgcse} about
6235 invoking @option{-O2} on programs that use computed gotos.
6239 Optimize yet more. @option{-O3} turns on all optimizations specified
6240 by @option{-O2} and also turns on the @option{-finline-functions},
6241 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6242 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6243 @option{-fipa-cp-clone} options.
6247 Reduce compilation time and make debugging produce the expected
6248 results. This is the default.
6252 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6253 do not typically increase code size. It also performs further
6254 optimizations designed to reduce code size.
6256 @option{-Os} disables the following optimization flags:
6257 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6258 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6259 -fprefetch-loop-arrays -ftree-vect-loop-version}
6263 Disregard strict standards compliance. @option{-Ofast} enables all
6264 @option{-O3} optimizations. It also enables optimizations that are not
6265 valid for all standard compliant programs.
6266 It turns on @option{-ffast-math} and the Fortran-specific
6267 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6269 If you use multiple @option{-O} options, with or without level numbers,
6270 the last such option is the one that is effective.
6273 Options of the form @option{-f@var{flag}} specify machine-independent
6274 flags. Most flags have both positive and negative forms; the negative
6275 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6276 below, only one of the forms is listed---the one you typically will
6277 use. You can figure out the other form by either removing @samp{no-}
6280 The following options control specific optimizations. They are either
6281 activated by @option{-O} options or are related to ones that are. You
6282 can use the following flags in the rare cases when ``fine-tuning'' of
6283 optimizations to be performed is desired.
6286 @item -fno-default-inline
6287 @opindex fno-default-inline
6288 Do not make member functions inline by default merely because they are
6289 defined inside the class scope (C++ only). Otherwise, when you specify
6290 @w{@option{-O}}, member functions defined inside class scope are compiled
6291 inline by default; i.e., you don't need to add @samp{inline} in front of
6292 the member function name.
6294 @item -fno-defer-pop
6295 @opindex fno-defer-pop
6296 Always pop the arguments to each function call as soon as that function
6297 returns. For machines that must pop arguments after a function call,
6298 the compiler normally lets arguments accumulate on the stack for several
6299 function calls and pops them all at once.
6301 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6303 @item -fforward-propagate
6304 @opindex fforward-propagate
6305 Perform a forward propagation pass on RTL@. The pass tries to combine two
6306 instructions and checks if the result can be simplified. If loop unrolling
6307 is active, two passes are performed and the second is scheduled after
6310 This option is enabled by default at optimization levels @option{-O},
6311 @option{-O2}, @option{-O3}, @option{-Os}.
6313 @item -ffp-contract=@var{style}
6314 @opindex ffp-contract
6315 @option{-ffp-contract=off} disables floating-point expression contraction.
6316 @option{-ffp-contract=fast} enables floating-point expression contraction
6317 such as forming of fused multiply-add operations if the target has
6318 native support for them.
6319 @option{-ffp-contract=on} enables floating-point expression contraction
6320 if allowed by the language standard. This is currently not implemented
6321 and treated equal to @option{-ffp-contract=off}.
6323 The default is @option{-ffp-contract=fast}.
6325 @item -fomit-frame-pointer
6326 @opindex fomit-frame-pointer
6327 Don't keep the frame pointer in a register for functions that
6328 don't need one. This avoids the instructions to save, set up and
6329 restore frame pointers; it also makes an extra register available
6330 in many functions. @strong{It also makes debugging impossible on
6333 On some machines, such as the VAX, this flag has no effect, because
6334 the standard calling sequence automatically handles the frame pointer
6335 and nothing is saved by pretending it doesn't exist. The
6336 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6337 whether a target machine supports this flag. @xref{Registers,,Register
6338 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6340 Starting with GCC version 4.6, the default setting (when not optimizing for
6341 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6342 @option{-fomit-frame-pointer}. The default can be reverted to
6343 @option{-fno-omit-frame-pointer} by configuring GCC with the
6344 @option{--enable-frame-pointer} configure option.
6346 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6348 @item -foptimize-sibling-calls
6349 @opindex foptimize-sibling-calls
6350 Optimize sibling and tail recursive calls.
6352 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6356 Do not expand any functions inline apart from those marked with
6357 the @code{always_inline} attribute. This is the default when not
6360 Single functions can be exempted from inlining by marking them
6361 with the @code{noinline} attribute.
6363 @item -finline-small-functions
6364 @opindex finline-small-functions
6365 Integrate functions into their callers when their body is smaller than expected
6366 function call code (so overall size of program gets smaller). The compiler
6367 heuristically decides which functions are simple enough to be worth integrating
6368 in this way. This inlining applies to all functions, even those not declared
6371 Enabled at level @option{-O2}.
6373 @item -findirect-inlining
6374 @opindex findirect-inlining
6375 Inline also indirect calls that are discovered to be known at compile
6376 time thanks to previous inlining. This option has any effect only
6377 when inlining itself is turned on by the @option{-finline-functions}
6378 or @option{-finline-small-functions} options.
6380 Enabled at level @option{-O2}.
6382 @item -finline-functions
6383 @opindex finline-functions
6384 Consider all functions for inlining, even if they are not declared inline.
6385 The compiler heuristically decides which functions are worth integrating
6388 If all calls to a given function are integrated, and the function is
6389 declared @code{static}, then the function is normally not output as
6390 assembler code in its own right.
6392 Enabled at level @option{-O3}.
6394 @item -finline-functions-called-once
6395 @opindex finline-functions-called-once
6396 Consider all @code{static} functions called once for inlining into their
6397 caller even if they are not marked @code{inline}. If a call to a given
6398 function is integrated, then the function is not output as assembler code
6401 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6403 @item -fearly-inlining
6404 @opindex fearly-inlining
6405 Inline functions marked by @code{always_inline} and functions whose body seems
6406 smaller than the function call overhead early before doing
6407 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6408 makes profiling significantly cheaper and usually inlining faster on programs
6409 having large chains of nested wrapper functions.
6415 Perform interprocedural scalar replacement of aggregates, removal of
6416 unused parameters and replacement of parameters passed by reference
6417 by parameters passed by value.
6419 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6421 @item -finline-limit=@var{n}
6422 @opindex finline-limit
6423 By default, GCC limits the size of functions that can be inlined. This flag
6424 allows coarse control of this limit. @var{n} is the size of functions that
6425 can be inlined in number of pseudo instructions.
6427 Inlining is actually controlled by a number of parameters, which may be
6428 specified individually by using @option{--param @var{name}=@var{value}}.
6429 The @option{-finline-limit=@var{n}} option sets some of these parameters
6433 @item max-inline-insns-single
6434 is set to @var{n}/2.
6435 @item max-inline-insns-auto
6436 is set to @var{n}/2.
6439 See below for a documentation of the individual
6440 parameters controlling inlining and for the defaults of these parameters.
6442 @emph{Note:} there may be no value to @option{-finline-limit} that results
6443 in default behavior.
6445 @emph{Note:} pseudo instruction represents, in this particular context, an
6446 abstract measurement of function's size. In no way does it represent a count
6447 of assembly instructions and as such its exact meaning might change from one
6448 release to an another.
6450 @item -fno-keep-inline-dllexport
6451 @opindex -fno-keep-inline-dllexport
6452 This is a more fine-grained version of @option{-fkeep-inline-functions},
6453 which applies only to functions that are declared using the @code{dllexport}
6454 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6457 @item -fkeep-inline-functions
6458 @opindex fkeep-inline-functions
6459 In C, emit @code{static} functions that are declared @code{inline}
6460 into the object file, even if the function has been inlined into all
6461 of its callers. This switch does not affect functions using the
6462 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6463 inline functions into the object file.
6465 @item -fkeep-static-consts
6466 @opindex fkeep-static-consts
6467 Emit variables declared @code{static const} when optimization isn't turned
6468 on, even if the variables aren't referenced.
6470 GCC enables this option by default. If you want to force the compiler to
6471 check if the variable was referenced, regardless of whether or not
6472 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6474 @item -fmerge-constants
6475 @opindex fmerge-constants
6476 Attempt to merge identical constants (string constants and floating-point
6477 constants) across compilation units.
6479 This option is the default for optimized compilation if the assembler and
6480 linker support it. Use @option{-fno-merge-constants} to inhibit this
6483 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6485 @item -fmerge-all-constants
6486 @opindex fmerge-all-constants
6487 Attempt to merge identical constants and identical variables.
6489 This option implies @option{-fmerge-constants}. In addition to
6490 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6491 arrays or initialized constant variables with integral or floating-point
6492 types. Languages like C or C++ require each variable, including multiple
6493 instances of the same variable in recursive calls, to have distinct locations,
6494 so using this option will result in non-conforming
6497 @item -fmodulo-sched
6498 @opindex fmodulo-sched
6499 Perform swing modulo scheduling immediately before the first scheduling
6500 pass. This pass looks at innermost loops and reorders their
6501 instructions by overlapping different iterations.
6503 @item -fmodulo-sched-allow-regmoves
6504 @opindex fmodulo-sched-allow-regmoves
6505 Perform more aggressive SMS based modulo scheduling with register moves
6506 allowed. By setting this flag certain anti-dependences edges will be
6507 deleted which will trigger the generation of reg-moves based on the
6508 life-range analysis. This option is effective only with
6509 @option{-fmodulo-sched} enabled.
6511 @item -fno-branch-count-reg
6512 @opindex fno-branch-count-reg
6513 Do not use ``decrement and branch'' instructions on a count register,
6514 but instead generate a sequence of instructions that decrement a
6515 register, compare it against zero, then branch based upon the result.
6516 This option is only meaningful on architectures that support such
6517 instructions, which include x86, PowerPC, IA-64 and S/390.
6519 The default is @option{-fbranch-count-reg}.
6521 @item -fno-function-cse
6522 @opindex fno-function-cse
6523 Do not put function addresses in registers; make each instruction that
6524 calls a constant function contain the function's address explicitly.
6526 This option results in less efficient code, but some strange hacks
6527 that alter the assembler output may be confused by the optimizations
6528 performed when this option is not used.
6530 The default is @option{-ffunction-cse}
6532 @item -fno-zero-initialized-in-bss
6533 @opindex fno-zero-initialized-in-bss
6534 If the target supports a BSS section, GCC by default puts variables that
6535 are initialized to zero into BSS@. This can save space in the resulting
6538 This option turns off this behavior because some programs explicitly
6539 rely on variables going to the data section. E.g., so that the
6540 resulting executable can find the beginning of that section and/or make
6541 assumptions based on that.
6543 The default is @option{-fzero-initialized-in-bss}.
6545 @item -fmudflap -fmudflapth -fmudflapir
6549 @cindex bounds checking
6551 For front-ends that support it (C and C++), instrument all risky
6552 pointer/array dereferencing operations, some standard library
6553 string/heap functions, and some other associated constructs with
6554 range/validity tests. Modules so instrumented should be immune to
6555 buffer overflows, invalid heap use, and some other classes of C/C++
6556 programming errors. The instrumentation relies on a separate runtime
6557 library (@file{libmudflap}), which will be linked into a program if
6558 @option{-fmudflap} is given at link time. Run-time behavior of the
6559 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6560 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6563 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6564 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6565 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6566 instrumentation should ignore pointer reads. This produces less
6567 instrumentation (and therefore faster execution) and still provides
6568 some protection against outright memory corrupting writes, but allows
6569 erroneously read data to propagate within a program.
6571 @item -fthread-jumps
6572 @opindex fthread-jumps
6573 Perform optimizations where we check to see if a jump branches to a
6574 location where another comparison subsumed by the first is found. If
6575 so, the first branch is redirected to either the destination of the
6576 second branch or a point immediately following it, depending on whether
6577 the condition is known to be true or false.
6579 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6581 @item -fsplit-wide-types
6582 @opindex fsplit-wide-types
6583 When using a type that occupies multiple registers, such as @code{long
6584 long} on a 32-bit system, split the registers apart and allocate them
6585 independently. This normally generates better code for those types,
6586 but may make debugging more difficult.
6588 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6591 @item -fcse-follow-jumps
6592 @opindex fcse-follow-jumps
6593 In common subexpression elimination (CSE), scan through jump instructions
6594 when the target of the jump is not reached by any other path. For
6595 example, when CSE encounters an @code{if} statement with an
6596 @code{else} clause, CSE will follow the jump when the condition
6599 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6601 @item -fcse-skip-blocks
6602 @opindex fcse-skip-blocks
6603 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6604 follow jumps that conditionally skip over blocks. When CSE
6605 encounters a simple @code{if} statement with no else clause,
6606 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6607 body of the @code{if}.
6609 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6611 @item -frerun-cse-after-loop
6612 @opindex frerun-cse-after-loop
6613 Re-run common subexpression elimination after loop optimizations has been
6616 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6620 Perform a global common subexpression elimination pass.
6621 This pass also performs global constant and copy propagation.
6623 @emph{Note:} When compiling a program using computed gotos, a GCC
6624 extension, you may get better run-time performance if you disable
6625 the global common subexpression elimination pass by adding
6626 @option{-fno-gcse} to the command line.
6628 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6632 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6633 attempt to move loads that are only killed by stores into themselves. This
6634 allows a loop containing a load/store sequence to be changed to a load outside
6635 the loop, and a copy/store within the loop.
6637 Enabled by default when gcse is enabled.
6641 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6642 global common subexpression elimination. This pass will attempt to move
6643 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6644 loops containing a load/store sequence can be changed to a load before
6645 the loop and a store after the loop.
6647 Not enabled at any optimization level.
6651 When @option{-fgcse-las} is enabled, the global common subexpression
6652 elimination pass eliminates redundant loads that come after stores to the
6653 same memory location (both partial and full redundancies).
6655 Not enabled at any optimization level.
6657 @item -fgcse-after-reload
6658 @opindex fgcse-after-reload
6659 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6660 pass is performed after reload. The purpose of this pass is to cleanup
6663 @item -funsafe-loop-optimizations
6664 @opindex funsafe-loop-optimizations
6665 If given, the loop optimizer will assume that loop indices do not
6666 overflow, and that the loops with nontrivial exit condition are not
6667 infinite. This enables a wider range of loop optimizations even if
6668 the loop optimizer itself cannot prove that these assumptions are valid.
6669 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6670 if it finds this kind of loop.
6672 @item -fcrossjumping
6673 @opindex fcrossjumping
6674 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6675 resulting code may or may not perform better than without cross-jumping.
6677 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6679 @item -fauto-inc-dec
6680 @opindex fauto-inc-dec
6681 Combine increments or decrements of addresses with memory accesses.
6682 This pass is always skipped on architectures that do not have
6683 instructions to support this. Enabled by default at @option{-O} and
6684 higher on architectures that support this.
6688 Perform dead code elimination (DCE) on RTL@.
6689 Enabled by default at @option{-O} and higher.
6693 Perform dead store elimination (DSE) on RTL@.
6694 Enabled by default at @option{-O} and higher.
6696 @item -fif-conversion
6697 @opindex fif-conversion
6698 Attempt to transform conditional jumps into branch-less equivalents. This
6699 include use of conditional moves, min, max, set flags and abs instructions, and
6700 some tricks doable by standard arithmetics. The use of conditional execution
6701 on chips where it is available is controlled by @code{if-conversion2}.
6703 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6705 @item -fif-conversion2
6706 @opindex fif-conversion2
6707 Use conditional execution (where available) to transform conditional jumps into
6708 branch-less equivalents.
6710 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6712 @item -fdelete-null-pointer-checks
6713 @opindex fdelete-null-pointer-checks
6714 Assume that programs cannot safely dereference null pointers, and that
6715 no code or data element resides there. This enables simple constant
6716 folding optimizations at all optimization levels. In addition, other
6717 optimization passes in GCC use this flag to control global dataflow
6718 analyses that eliminate useless checks for null pointers; these assume
6719 that if a pointer is checked after it has already been dereferenced,
6722 Note however that in some environments this assumption is not true.
6723 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6724 for programs that depend on that behavior.
6726 Some targets, especially embedded ones, disable this option at all levels.
6727 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6728 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6729 are enabled independently at different optimization levels.
6731 @item -fdevirtualize
6732 @opindex fdevirtualize
6733 Attempt to convert calls to virtual functions to direct calls. This
6734 is done both within a procedure and interprocedurally as part of
6735 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6736 propagation (@option{-fipa-cp}).
6737 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6739 @item -fexpensive-optimizations
6740 @opindex fexpensive-optimizations
6741 Perform a number of minor optimizations that are relatively expensive.
6743 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6747 Attempt to remove redundant extension instructions. This is especially
6748 helpful for the x86-64 architecture which implicitly zero-extends in 64-bit
6749 registers after writing to their lower 32-bit half.
6751 Enabled for x86 at levels @option{-O2}, @option{-O3}.
6753 @item -foptimize-register-move
6755 @opindex foptimize-register-move
6757 Attempt to reassign register numbers in move instructions and as
6758 operands of other simple instructions in order to maximize the amount of
6759 register tying. This is especially helpful on machines with two-operand
6762 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6765 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6767 @item -fira-algorithm=@var{algorithm}
6768 Use the specified coloring algorithm for the integrated register
6769 allocator. The @var{algorithm} argument can be @samp{priority}, which
6770 specifies Chow's priority coloring, or @samp{CB}, which specifies
6771 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
6772 for all architectures, but for those targets that do support it, it is
6773 the default because it generates better code.
6775 @item -fira-region=@var{region}
6776 Use specified regions for the integrated register allocator. The
6777 @var{region} argument should be one of the following:
6782 Use all loops as register allocation regions.
6783 This can give the best results for machines with a small and/or
6784 irregular register set.
6787 Use all loops except for loops with small register pressure
6788 as the regions. This value usually gives
6789 the best results in most cases and for most architectures,
6790 and is enabled by default when compiling with optimization for speed
6791 (@option{-O}, @option{-O2}, @dots{}).
6794 Use all functions as a single region.
6795 This typically results in the smallest code size, and is enabled by default for
6796 @option{-Os} or @option{-O0}.
6800 @item -fira-loop-pressure
6801 @opindex fira-loop-pressure
6802 Use IRA to evaluate register pressure in loops for decisions to move
6803 loop invariants. This option usually results in generation
6804 of faster and smaller code on machines with large register files (>= 32
6805 registers), but it can slow the compiler down.
6807 This option is enabled at level @option{-O3} for some targets.
6809 @item -fno-ira-share-save-slots
6810 @opindex fno-ira-share-save-slots
6811 Disable sharing of stack slots used for saving call-used hard
6812 registers living through a call. Each hard register gets a
6813 separate stack slot, and as a result function stack frames are
6816 @item -fno-ira-share-spill-slots
6817 @opindex fno-ira-share-spill-slots
6818 Disable sharing of stack slots allocated for pseudo-registers. Each
6819 pseudo-register that does not get a hard register gets a separate
6820 stack slot, and as a result function stack frames are larger.
6822 @item -fira-verbose=@var{n}
6823 @opindex fira-verbose
6824 Control the verbosity of the dump file for the integrated register allocator.
6825 The default value is 5. If the value @var{n} is greater or equal to 10,
6826 the dump output is sent to stderr using the same format as @var{n} minus 10.
6828 @item -fdelayed-branch
6829 @opindex fdelayed-branch
6830 If supported for the target machine, attempt to reorder instructions
6831 to exploit instruction slots available after delayed branch
6834 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6836 @item -fschedule-insns
6837 @opindex fschedule-insns
6838 If supported for the target machine, attempt to reorder instructions to
6839 eliminate execution stalls due to required data being unavailable. This
6840 helps machines that have slow floating point or memory load instructions
6841 by allowing other instructions to be issued until the result of the load
6842 or floating-point instruction is required.
6844 Enabled at levels @option{-O2}, @option{-O3}.
6846 @item -fschedule-insns2
6847 @opindex fschedule-insns2
6848 Similar to @option{-fschedule-insns}, but requests an additional pass of
6849 instruction scheduling after register allocation has been done. This is
6850 especially useful on machines with a relatively small number of
6851 registers and where memory load instructions take more than one cycle.
6853 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6855 @item -fno-sched-interblock
6856 @opindex fno-sched-interblock
6857 Don't schedule instructions across basic blocks. This is normally
6858 enabled by default when scheduling before register allocation, i.e.@:
6859 with @option{-fschedule-insns} or at @option{-O2} or higher.
6861 @item -fno-sched-spec
6862 @opindex fno-sched-spec
6863 Don't allow speculative motion of non-load instructions. This is normally
6864 enabled by default when scheduling before register allocation, i.e.@:
6865 with @option{-fschedule-insns} or at @option{-O2} or higher.
6867 @item -fsched-pressure
6868 @opindex fsched-pressure
6869 Enable register pressure sensitive insn scheduling before the register
6870 allocation. This only makes sense when scheduling before register
6871 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6872 @option{-O2} or higher. Usage of this option can improve the
6873 generated code and decrease its size by preventing register pressure
6874 increase above the number of available hard registers and as a
6875 consequence register spills in the register allocation.
6877 @item -fsched-spec-load
6878 @opindex fsched-spec-load
6879 Allow speculative motion of some load instructions. This only makes
6880 sense when scheduling before register allocation, i.e.@: with
6881 @option{-fschedule-insns} or at @option{-O2} or higher.
6883 @item -fsched-spec-load-dangerous
6884 @opindex fsched-spec-load-dangerous
6885 Allow speculative motion of more load instructions. This only makes
6886 sense when scheduling before register allocation, i.e.@: with
6887 @option{-fschedule-insns} or at @option{-O2} or higher.
6889 @item -fsched-stalled-insns
6890 @itemx -fsched-stalled-insns=@var{n}
6891 @opindex fsched-stalled-insns
6892 Define how many insns (if any) can be moved prematurely from the queue
6893 of stalled insns into the ready list, during the second scheduling pass.
6894 @option{-fno-sched-stalled-insns} means that no insns will be moved
6895 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6896 on how many queued insns can be moved prematurely.
6897 @option{-fsched-stalled-insns} without a value is equivalent to
6898 @option{-fsched-stalled-insns=1}.
6900 @item -fsched-stalled-insns-dep
6901 @itemx -fsched-stalled-insns-dep=@var{n}
6902 @opindex fsched-stalled-insns-dep
6903 Define how many insn groups (cycles) will be examined for a dependency
6904 on a stalled insn that is candidate for premature removal from the queue
6905 of stalled insns. This has an effect only during the second scheduling pass,
6906 and only if @option{-fsched-stalled-insns} is used.
6907 @option{-fno-sched-stalled-insns-dep} is equivalent to
6908 @option{-fsched-stalled-insns-dep=0}.
6909 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6910 @option{-fsched-stalled-insns-dep=1}.
6912 @item -fsched2-use-superblocks
6913 @opindex fsched2-use-superblocks
6914 When scheduling after register allocation, do use superblock scheduling
6915 algorithm. Superblock scheduling allows motion across basic block boundaries
6916 resulting on faster schedules. This option is experimental, as not all machine
6917 descriptions used by GCC model the CPU closely enough to avoid unreliable
6918 results from the algorithm.
6920 This only makes sense when scheduling after register allocation, i.e.@: with
6921 @option{-fschedule-insns2} or at @option{-O2} or higher.
6923 @item -fsched-group-heuristic
6924 @opindex fsched-group-heuristic
6925 Enable the group heuristic in the scheduler. This heuristic favors
6926 the instruction that belongs to a schedule group. This is enabled
6927 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6928 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6930 @item -fsched-critical-path-heuristic
6931 @opindex fsched-critical-path-heuristic
6932 Enable the critical-path heuristic in the scheduler. This heuristic favors
6933 instructions on the critical path. This is enabled by default when
6934 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6935 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6937 @item -fsched-spec-insn-heuristic
6938 @opindex fsched-spec-insn-heuristic
6939 Enable the speculative instruction heuristic in the scheduler. This
6940 heuristic favors speculative instructions with greater dependency weakness.
6941 This is enabled by default when scheduling is enabled, i.e.@:
6942 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6943 or at @option{-O2} or higher.
6945 @item -fsched-rank-heuristic
6946 @opindex fsched-rank-heuristic
6947 Enable the rank heuristic in the scheduler. This heuristic favors
6948 the instruction belonging to a basic block with greater size or frequency.
6949 This is enabled by default when scheduling is enabled, i.e.@:
6950 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6951 at @option{-O2} or higher.
6953 @item -fsched-last-insn-heuristic
6954 @opindex fsched-last-insn-heuristic
6955 Enable the last-instruction heuristic in the scheduler. This heuristic
6956 favors the instruction that is less dependent on the last instruction
6957 scheduled. This is enabled by default when scheduling is enabled,
6958 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6959 at @option{-O2} or higher.
6961 @item -fsched-dep-count-heuristic
6962 @opindex fsched-dep-count-heuristic
6963 Enable the dependent-count heuristic in the scheduler. This heuristic
6964 favors the instruction that has more instructions depending on it.
6965 This is enabled by default when scheduling is enabled, i.e.@:
6966 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6967 at @option{-O2} or higher.
6969 @item -freschedule-modulo-scheduled-loops
6970 @opindex freschedule-modulo-scheduled-loops
6971 The modulo scheduling comes before the traditional scheduling, if a loop
6972 was modulo scheduled we may want to prevent the later scheduling passes
6973 from changing its schedule, we use this option to control that.
6975 @item -fselective-scheduling
6976 @opindex fselective-scheduling
6977 Schedule instructions using selective scheduling algorithm. Selective
6978 scheduling runs instead of the first scheduler pass.
6980 @item -fselective-scheduling2
6981 @opindex fselective-scheduling2
6982 Schedule instructions using selective scheduling algorithm. Selective
6983 scheduling runs instead of the second scheduler pass.
6985 @item -fsel-sched-pipelining
6986 @opindex fsel-sched-pipelining
6987 Enable software pipelining of innermost loops during selective scheduling.
6988 This option has no effect until one of @option{-fselective-scheduling} or
6989 @option{-fselective-scheduling2} is turned on.
6991 @item -fsel-sched-pipelining-outer-loops
6992 @opindex fsel-sched-pipelining-outer-loops
6993 When pipelining loops during selective scheduling, also pipeline outer loops.
6994 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6997 @opindex fshrink-wrap
6998 Emit function prologues only before parts of the function that need it,
6999 rather than at the top of the function. This flag is enabled by default at
7000 @option{-O} and higher.
7002 @item -fcaller-saves
7003 @opindex fcaller-saves
7004 Enable values to be allocated in registers that will be clobbered by
7005 function calls, by emitting extra instructions to save and restore the
7006 registers around such calls. Such allocation is done only when it
7007 seems to result in better code than would otherwise be produced.
7009 This option is always enabled by default on certain machines, usually
7010 those which have no call-preserved registers to use instead.
7012 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7014 @item -fcombine-stack-adjustments
7015 @opindex fcombine-stack-adjustments
7016 Tracks stack adjustments (pushes and pops) and stack memory references
7017 and then tries to find ways to combine them.
7019 Enabled by default at @option{-O1} and higher.
7021 @item -fconserve-stack
7022 @opindex fconserve-stack
7023 Attempt to minimize stack usage. The compiler will attempt to use less
7024 stack space, even if that makes the program slower. This option
7025 implies setting the @option{large-stack-frame} parameter to 100
7026 and the @option{large-stack-frame-growth} parameter to 400.
7028 @item -ftree-reassoc
7029 @opindex ftree-reassoc
7030 Perform reassociation on trees. This flag is enabled by default
7031 at @option{-O} and higher.
7035 Perform partial redundancy elimination (PRE) on trees. This flag is
7036 enabled by default at @option{-O2} and @option{-O3}.
7038 @item -ftree-forwprop
7039 @opindex ftree-forwprop
7040 Perform forward propagation on trees. This flag is enabled by default
7041 at @option{-O} and higher.
7045 Perform full redundancy elimination (FRE) on trees. The difference
7046 between FRE and PRE is that FRE only considers expressions
7047 that are computed on all paths leading to the redundant computation.
7048 This analysis is faster than PRE, though it exposes fewer redundancies.
7049 This flag is enabled by default at @option{-O} and higher.
7051 @item -ftree-phiprop
7052 @opindex ftree-phiprop
7053 Perform hoisting of loads from conditional pointers on trees. This
7054 pass is enabled by default at @option{-O} and higher.
7056 @item -ftree-copy-prop
7057 @opindex ftree-copy-prop
7058 Perform copy propagation on trees. This pass eliminates unnecessary
7059 copy operations. This flag is enabled by default at @option{-O} and
7062 @item -fipa-pure-const
7063 @opindex fipa-pure-const
7064 Discover which functions are pure or constant.
7065 Enabled by default at @option{-O} and higher.
7067 @item -fipa-reference
7068 @opindex fipa-reference
7069 Discover which static variables do not escape cannot escape the
7071 Enabled by default at @option{-O} and higher.
7075 Perform interprocedural pointer analysis and interprocedural modification
7076 and reference analysis. This option can cause excessive memory and
7077 compile-time usage on large compilation units. It is not enabled by
7078 default at any optimization level.
7081 @opindex fipa-profile
7082 Perform interprocedural profile propagation. The functions called only from
7083 cold functions are marked as cold. Also functions executed once (such as
7084 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7085 functions and loop less parts of functions executed once are then optimized for
7087 Enabled by default at @option{-O} and higher.
7091 Perform interprocedural constant propagation.
7092 This optimization analyzes the program to determine when values passed
7093 to functions are constants and then optimizes accordingly.
7094 This optimization can substantially increase performance
7095 if the application has constants passed to functions.
7096 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7098 @item -fipa-cp-clone
7099 @opindex fipa-cp-clone
7100 Perform function cloning to make interprocedural constant propagation stronger.
7101 When enabled, interprocedural constant propagation will perform function cloning
7102 when externally visible function can be called with constant arguments.
7103 Because this optimization can create multiple copies of functions,
7104 it may significantly increase code size
7105 (see @option{--param ipcp-unit-growth=@var{value}}).
7106 This flag is enabled by default at @option{-O3}.
7108 @item -fipa-matrix-reorg
7109 @opindex fipa-matrix-reorg
7110 Perform matrix flattening and transposing.
7111 Matrix flattening tries to replace an @math{m}-dimensional matrix
7112 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7113 This reduces the level of indirection needed for accessing the elements
7114 of the matrix. The second optimization is matrix transposing, which
7115 attempts to change the order of the matrix's dimensions in order to
7116 improve cache locality.
7117 Both optimizations need the @option{-fwhole-program} flag.
7118 Transposing is enabled only if profiling information is available.
7122 Perform forward store motion on trees. This flag is
7123 enabled by default at @option{-O} and higher.
7125 @item -ftree-bit-ccp
7126 @opindex ftree-bit-ccp
7127 Perform sparse conditional bit constant propagation on trees and propagate
7128 pointer alignment information.
7129 This pass only operates on local scalar variables and is enabled by default
7130 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7134 Perform sparse conditional constant propagation (CCP) on trees. This
7135 pass only operates on local scalar variables and is enabled by default
7136 at @option{-O} and higher.
7138 @item -ftree-switch-conversion
7139 Perform conversion of simple initializations in a switch to
7140 initializations from a scalar array. This flag is enabled by default
7141 at @option{-O2} and higher.
7143 @item -ftree-tail-merge
7144 Look for identical code sequences. When found, replace one with a jump to the
7145 other. This optimization is known as tail merging or cross jumping. This flag
7146 is enabled by default at @option{-O2} and higher. The compilation time
7148 be limited using @option{max-tail-merge-comparisons} parameter and
7149 @option{max-tail-merge-iterations} parameter.
7153 Perform dead code elimination (DCE) on trees. This flag is enabled by
7154 default at @option{-O} and higher.
7156 @item -ftree-builtin-call-dce
7157 @opindex ftree-builtin-call-dce
7158 Perform conditional dead code elimination (DCE) for calls to builtin functions
7159 that may set @code{errno} but are otherwise side-effect free. This flag is
7160 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7163 @item -ftree-dominator-opts
7164 @opindex ftree-dominator-opts
7165 Perform a variety of simple scalar cleanups (constant/copy
7166 propagation, redundancy elimination, range propagation and expression
7167 simplification) based on a dominator tree traversal. This also
7168 performs jump threading (to reduce jumps to jumps). This flag is
7169 enabled by default at @option{-O} and higher.
7173 Perform dead store elimination (DSE) on trees. A dead store is a store into
7174 a memory location that is later overwritten by another store without
7175 any intervening loads. In this case the earlier store can be deleted. This
7176 flag is enabled by default at @option{-O} and higher.
7180 Perform loop header copying on trees. This is beneficial since it increases
7181 effectiveness of code motion optimizations. It also saves one jump. This flag
7182 is enabled by default at @option{-O} and higher. It is not enabled
7183 for @option{-Os}, since it usually increases code size.
7185 @item -ftree-loop-optimize
7186 @opindex ftree-loop-optimize
7187 Perform loop optimizations on trees. This flag is enabled by default
7188 at @option{-O} and higher.
7190 @item -ftree-loop-linear
7191 @opindex ftree-loop-linear
7192 Perform loop interchange transformations on tree. Same as
7193 @option{-floop-interchange}. To use this code transformation, GCC has
7194 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7195 enable the Graphite loop transformation infrastructure.
7197 @item -floop-interchange
7198 @opindex floop-interchange
7199 Perform loop interchange transformations on loops. Interchanging two
7200 nested loops switches the inner and outer loops. For example, given a
7205 A(J, I) = A(J, I) * C
7209 loop interchange will transform the loop as if the user had written:
7213 A(J, I) = A(J, I) * C
7217 which can be beneficial when @code{N} is larger than the caches,
7218 because in Fortran, the elements of an array are stored in memory
7219 contiguously by column, and the original loop iterates over rows,
7220 potentially creating at each access a cache miss. This optimization
7221 applies to all the languages supported by GCC and is not limited to
7222 Fortran. To use this code transformation, GCC has to be configured
7223 with @option{--with-ppl} and @option{--with-cloog} to enable the
7224 Graphite loop transformation infrastructure.
7226 @item -floop-strip-mine
7227 @opindex floop-strip-mine
7228 Perform loop strip mining transformations on loops. Strip mining
7229 splits a loop into two nested loops. The outer loop has strides
7230 equal to the strip size and the inner loop has strides of the
7231 original loop within a strip. The strip length can be changed
7232 using the @option{loop-block-tile-size} parameter. For example,
7239 loop strip mining will transform the loop as if the user had written:
7242 DO I = II, min (II + 50, N)
7247 This optimization applies to all the languages supported by GCC and is
7248 not limited to Fortran. To use this code transformation, GCC has to
7249 be configured with @option{--with-ppl} and @option{--with-cloog} to
7250 enable the Graphite loop transformation infrastructure.
7253 @opindex floop-block
7254 Perform loop blocking transformations on loops. Blocking strip mines
7255 each loop in the loop nest such that the memory accesses of the
7256 element loops fit inside caches. The strip length can be changed
7257 using the @option{loop-block-tile-size} parameter. For example, given
7262 A(J, I) = B(I) + C(J)
7266 loop blocking will transform the loop as if the user had written:
7270 DO I = II, min (II + 50, N)
7271 DO J = JJ, min (JJ + 50, M)
7272 A(J, I) = B(I) + C(J)
7278 which can be beneficial when @code{M} is larger than the caches,
7279 because the innermost loop will iterate over a smaller amount of data
7280 which can be kept in the caches. This optimization applies to all the
7281 languages supported by GCC and is not limited to Fortran. To use this
7282 code transformation, GCC has to be configured with @option{--with-ppl}
7283 and @option{--with-cloog} to enable the Graphite loop transformation
7286 @item -fgraphite-identity
7287 @opindex fgraphite-identity
7288 Enable the identity transformation for graphite. For every SCoP we generate
7289 the polyhedral representation and transform it back to gimple. Using
7290 @option{-fgraphite-identity} we can check the costs or benefits of the
7291 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7292 are also performed by the code generator CLooG, like index splitting and
7293 dead code elimination in loops.
7295 @item -floop-flatten
7296 @opindex floop-flatten
7297 Removes the loop nesting structure: transforms the loop nest into a
7298 single loop. This transformation can be useful as an enablement
7299 transform for vectorization and parallelization. This feature
7301 To use this code transformation, GCC has to be configured
7302 with @option{--with-ppl} and @option{--with-cloog} to enable the
7303 Graphite loop transformation infrastructure.
7305 @item -floop-parallelize-all
7306 @opindex floop-parallelize-all
7307 Use the Graphite data dependence analysis to identify loops that can
7308 be parallelized. Parallelize all the loops that can be analyzed to
7309 not contain loop carried dependences without checking that it is
7310 profitable to parallelize the loops.
7312 @item -fcheck-data-deps
7313 @opindex fcheck-data-deps
7314 Compare the results of several data dependence analyzers. This option
7315 is used for debugging the data dependence analyzers.
7317 @item -ftree-loop-if-convert
7318 Attempt to transform conditional jumps in the innermost loops to
7319 branch-less equivalents. The intent is to remove control-flow from
7320 the innermost loops in order to improve the ability of the
7321 vectorization pass to handle these loops. This is enabled by default
7322 if vectorization is enabled.
7324 @item -ftree-loop-if-convert-stores
7325 Attempt to also if-convert conditional jumps containing memory writes.
7326 This transformation can be unsafe for multi-threaded programs as it
7327 transforms conditional memory writes into unconditional memory writes.
7330 for (i = 0; i < N; i++)
7334 would be transformed to
7336 for (i = 0; i < N; i++)
7337 A[i] = cond ? expr : A[i];
7339 potentially producing data races.
7341 @item -ftree-loop-distribution
7342 Perform loop distribution. This flag can improve cache performance on
7343 big loop bodies and allow further loop optimizations, like
7344 parallelization or vectorization, to take place. For example, the loop
7361 @item -ftree-loop-distribute-patterns
7362 Perform loop distribution of patterns that can be code generated with
7363 calls to a library. This flag is enabled by default at @option{-O3}.
7365 This pass distributes the initialization loops and generates a call to
7366 memset zero. For example, the loop
7382 and the initialization loop is transformed into a call to memset zero.
7384 @item -ftree-loop-im
7385 @opindex ftree-loop-im
7386 Perform loop invariant motion on trees. This pass moves only invariants that
7387 would be hard to handle at RTL level (function calls, operations that expand to
7388 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7389 operands of conditions that are invariant out of the loop, so that we can use
7390 just trivial invariantness analysis in loop unswitching. The pass also includes
7393 @item -ftree-loop-ivcanon
7394 @opindex ftree-loop-ivcanon
7395 Create a canonical counter for number of iterations in loops for which
7396 determining number of iterations requires complicated analysis. Later
7397 optimizations then may determine the number easily. Useful especially
7398 in connection with unrolling.
7402 Perform induction variable optimizations (strength reduction, induction
7403 variable merging and induction variable elimination) on trees.
7405 @item -ftree-parallelize-loops=n
7406 @opindex ftree-parallelize-loops
7407 Parallelize loops, i.e., split their iteration space to run in n threads.
7408 This is only possible for loops whose iterations are independent
7409 and can be arbitrarily reordered. The optimization is only
7410 profitable on multiprocessor machines, for loops that are CPU-intensive,
7411 rather than constrained e.g.@: by memory bandwidth. This option
7412 implies @option{-pthread}, and thus is only supported on targets
7413 that have support for @option{-pthread}.
7417 Perform function-local points-to analysis on trees. This flag is
7418 enabled by default at @option{-O} and higher.
7422 Perform scalar replacement of aggregates. This pass replaces structure
7423 references with scalars to prevent committing structures to memory too
7424 early. This flag is enabled by default at @option{-O} and higher.
7426 @item -ftree-copyrename
7427 @opindex ftree-copyrename
7428 Perform copy renaming on trees. This pass attempts to rename compiler
7429 temporaries to other variables at copy locations, usually resulting in
7430 variable names which more closely resemble the original variables. This flag
7431 is enabled by default at @option{-O} and higher.
7435 Perform temporary expression replacement during the SSA->normal phase. Single
7436 use/single def temporaries are replaced at their use location with their
7437 defining expression. This results in non-GIMPLE code, but gives the expanders
7438 much more complex trees to work on resulting in better RTL generation. This is
7439 enabled by default at @option{-O} and higher.
7441 @item -ftree-vectorize
7442 @opindex ftree-vectorize
7443 Perform loop vectorization on trees. This flag is enabled by default at
7446 @item -ftree-slp-vectorize
7447 @opindex ftree-slp-vectorize
7448 Perform basic block vectorization on trees. This flag is enabled by default at
7449 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7451 @item -ftree-vect-loop-version
7452 @opindex ftree-vect-loop-version
7453 Perform loop versioning when doing loop vectorization on trees. When a loop
7454 appears to be vectorizable except that data alignment or data dependence cannot
7455 be determined at compile time, then vectorized and non-vectorized versions of
7456 the loop are generated along with run-time checks for alignment or dependence
7457 to control which version is executed. This option is enabled by default
7458 except at level @option{-Os} where it is disabled.
7460 @item -fvect-cost-model
7461 @opindex fvect-cost-model
7462 Enable cost model for vectorization.
7466 Perform Value Range Propagation on trees. This is similar to the
7467 constant propagation pass, but instead of values, ranges of values are
7468 propagated. This allows the optimizers to remove unnecessary range
7469 checks like array bound checks and null pointer checks. This is
7470 enabled by default at @option{-O2} and higher. Null pointer check
7471 elimination is only done if @option{-fdelete-null-pointer-checks} is
7476 Perform tail duplication to enlarge superblock size. This transformation
7477 simplifies the control flow of the function allowing other optimizations to do
7480 @item -funroll-loops
7481 @opindex funroll-loops
7482 Unroll loops whose number of iterations can be determined at compile
7483 time or upon entry to the loop. @option{-funroll-loops} implies
7484 @option{-frerun-cse-after-loop}. This option makes code larger,
7485 and may or may not make it run faster.
7487 @item -funroll-all-loops
7488 @opindex funroll-all-loops
7489 Unroll all loops, even if their number of iterations is uncertain when
7490 the loop is entered. This usually makes programs run more slowly.
7491 @option{-funroll-all-loops} implies the same options as
7492 @option{-funroll-loops},
7494 @item -fsplit-ivs-in-unroller
7495 @opindex fsplit-ivs-in-unroller
7496 Enables expressing of values of induction variables in later iterations
7497 of the unrolled loop using the value in the first iteration. This breaks
7498 long dependency chains, thus improving efficiency of the scheduling passes.
7500 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7501 same effect. However in cases the loop body is more complicated than
7502 a single basic block, this is not reliable. It also does not work at all
7503 on some of the architectures due to restrictions in the CSE pass.
7505 This optimization is enabled by default.
7507 @item -fvariable-expansion-in-unroller
7508 @opindex fvariable-expansion-in-unroller
7509 With this option, the compiler will create multiple copies of some
7510 local variables when unrolling a loop which can result in superior code.
7512 @item -fpartial-inlining
7513 @opindex fpartial-inlining
7514 Inline parts of functions. This option has any effect only
7515 when inlining itself is turned on by the @option{-finline-functions}
7516 or @option{-finline-small-functions} options.
7518 Enabled at level @option{-O2}.
7520 @item -fpredictive-commoning
7521 @opindex fpredictive-commoning
7522 Perform predictive commoning optimization, i.e., reusing computations
7523 (especially memory loads and stores) performed in previous
7524 iterations of loops.
7526 This option is enabled at level @option{-O3}.
7528 @item -fprefetch-loop-arrays
7529 @opindex fprefetch-loop-arrays
7530 If supported by the target machine, generate instructions to prefetch
7531 memory to improve the performance of loops that access large arrays.
7533 This option may generate better or worse code; results are highly
7534 dependent on the structure of loops within the source code.
7536 Disabled at level @option{-Os}.
7539 @itemx -fno-peephole2
7540 @opindex fno-peephole
7541 @opindex fno-peephole2
7542 Disable any machine-specific peephole optimizations. The difference
7543 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7544 are implemented in the compiler; some targets use one, some use the
7545 other, a few use both.
7547 @option{-fpeephole} is enabled by default.
7548 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7550 @item -fno-guess-branch-probability
7551 @opindex fno-guess-branch-probability
7552 Do not guess branch probabilities using heuristics.
7554 GCC will use heuristics to guess branch probabilities if they are
7555 not provided by profiling feedback (@option{-fprofile-arcs}). These
7556 heuristics are based on the control flow graph. If some branch probabilities
7557 are specified by @samp{__builtin_expect}, then the heuristics will be
7558 used to guess branch probabilities for the rest of the control flow graph,
7559 taking the @samp{__builtin_expect} info into account. The interactions
7560 between the heuristics and @samp{__builtin_expect} can be complex, and in
7561 some cases, it may be useful to disable the heuristics so that the effects
7562 of @samp{__builtin_expect} are easier to understand.
7564 The default is @option{-fguess-branch-probability} at levels
7565 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7567 @item -freorder-blocks
7568 @opindex freorder-blocks
7569 Reorder basic blocks in the compiled function in order to reduce number of
7570 taken branches and improve code locality.
7572 Enabled at levels @option{-O2}, @option{-O3}.
7574 @item -freorder-blocks-and-partition
7575 @opindex freorder-blocks-and-partition
7576 In addition to reordering basic blocks in the compiled function, in order
7577 to reduce number of taken branches, partitions hot and cold basic blocks
7578 into separate sections of the assembly and .o files, to improve
7579 paging and cache locality performance.
7581 This optimization is automatically turned off in the presence of
7582 exception handling, for linkonce sections, for functions with a user-defined
7583 section attribute and on any architecture that does not support named
7586 @item -freorder-functions
7587 @opindex freorder-functions
7588 Reorder functions in the object file in order to
7589 improve code locality. This is implemented by using special
7590 subsections @code{.text.hot} for most frequently executed functions and
7591 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7592 the linker so object file format must support named sections and linker must
7593 place them in a reasonable way.
7595 Also profile feedback must be available in to make this option effective. See
7596 @option{-fprofile-arcs} for details.
7598 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7600 @item -fstrict-aliasing
7601 @opindex fstrict-aliasing
7602 Allow the compiler to assume the strictest aliasing rules applicable to
7603 the language being compiled. For C (and C++), this activates
7604 optimizations based on the type of expressions. In particular, an
7605 object of one type is assumed never to reside at the same address as an
7606 object of a different type, unless the types are almost the same. For
7607 example, an @code{unsigned int} can alias an @code{int}, but not a
7608 @code{void*} or a @code{double}. A character type may alias any other
7611 @anchor{Type-punning}Pay special attention to code like this:
7624 The practice of reading from a different union member than the one most
7625 recently written to (called ``type-punning'') is common. Even with
7626 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7627 is accessed through the union type. So, the code above will work as
7628 expected. @xref{Structures unions enumerations and bit-fields
7629 implementation}. However, this code might not:
7640 Similarly, access by taking the address, casting the resulting pointer
7641 and dereferencing the result has undefined behavior, even if the cast
7642 uses a union type, e.g.:
7646 return ((union a_union *) &d)->i;
7650 The @option{-fstrict-aliasing} option is enabled at levels
7651 @option{-O2}, @option{-O3}, @option{-Os}.
7653 @item -fstrict-overflow
7654 @opindex fstrict-overflow
7655 Allow the compiler to assume strict signed overflow rules, depending
7656 on the language being compiled. For C (and C++) this means that
7657 overflow when doing arithmetic with signed numbers is undefined, which
7658 means that the compiler may assume that it will not happen. This
7659 permits various optimizations. For example, the compiler will assume
7660 that an expression like @code{i + 10 > i} will always be true for
7661 signed @code{i}. This assumption is only valid if signed overflow is
7662 undefined, as the expression is false if @code{i + 10} overflows when
7663 using twos complement arithmetic. When this option is in effect any
7664 attempt to determine whether an operation on signed numbers will
7665 overflow must be written carefully to not actually involve overflow.
7667 This option also allows the compiler to assume strict pointer
7668 semantics: given a pointer to an object, if adding an offset to that
7669 pointer does not produce a pointer to the same object, the addition is
7670 undefined. This permits the compiler to conclude that @code{p + u >
7671 p} is always true for a pointer @code{p} and unsigned integer
7672 @code{u}. This assumption is only valid because pointer wraparound is
7673 undefined, as the expression is false if @code{p + u} overflows using
7674 twos complement arithmetic.
7676 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7677 that integer signed overflow is fully defined: it wraps. When
7678 @option{-fwrapv} is used, there is no difference between
7679 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7680 integers. With @option{-fwrapv} certain types of overflow are
7681 permitted. For example, if the compiler gets an overflow when doing
7682 arithmetic on constants, the overflowed value can still be used with
7683 @option{-fwrapv}, but not otherwise.
7685 The @option{-fstrict-overflow} option is enabled at levels
7686 @option{-O2}, @option{-O3}, @option{-Os}.
7688 @item -falign-functions
7689 @itemx -falign-functions=@var{n}
7690 @opindex falign-functions
7691 Align the start of functions to the next power-of-two greater than
7692 @var{n}, skipping up to @var{n} bytes. For instance,
7693 @option{-falign-functions=32} aligns functions to the next 32-byte
7694 boundary, but @option{-falign-functions=24} would align to the next
7695 32-byte boundary only if this can be done by skipping 23 bytes or less.
7697 @option{-fno-align-functions} and @option{-falign-functions=1} are
7698 equivalent and mean that functions will not be aligned.
7700 Some assemblers only support this flag when @var{n} is a power of two;
7701 in that case, it is rounded up.
7703 If @var{n} is not specified or is zero, use a machine-dependent default.
7705 Enabled at levels @option{-O2}, @option{-O3}.
7707 @item -falign-labels
7708 @itemx -falign-labels=@var{n}
7709 @opindex falign-labels
7710 Align all branch targets to a power-of-two boundary, skipping up to
7711 @var{n} bytes like @option{-falign-functions}. This option can easily
7712 make code slower, because it must insert dummy operations for when the
7713 branch target is reached in the usual flow of the code.
7715 @option{-fno-align-labels} and @option{-falign-labels=1} are
7716 equivalent and mean that labels will not be aligned.
7718 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7719 are greater than this value, then their values are used instead.
7721 If @var{n} is not specified or is zero, use a machine-dependent default
7722 which is very likely to be @samp{1}, meaning no alignment.
7724 Enabled at levels @option{-O2}, @option{-O3}.
7727 @itemx -falign-loops=@var{n}
7728 @opindex falign-loops
7729 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7730 like @option{-falign-functions}. The hope is that the loop will be
7731 executed many times, which will make up for any execution of the dummy
7734 @option{-fno-align-loops} and @option{-falign-loops=1} are
7735 equivalent and mean that loops will not be aligned.
7737 If @var{n} is not specified or is zero, use a machine-dependent default.
7739 Enabled at levels @option{-O2}, @option{-O3}.
7742 @itemx -falign-jumps=@var{n}
7743 @opindex falign-jumps
7744 Align branch targets to a power-of-two boundary, for branch targets
7745 where the targets can only be reached by jumping, skipping up to @var{n}
7746 bytes like @option{-falign-functions}. In this case, no dummy operations
7749 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7750 equivalent and mean that loops will not be aligned.
7752 If @var{n} is not specified or is zero, use a machine-dependent default.
7754 Enabled at levels @option{-O2}, @option{-O3}.
7756 @item -funit-at-a-time
7757 @opindex funit-at-a-time
7758 This option is left for compatibility reasons. @option{-funit-at-a-time}
7759 has no effect, while @option{-fno-unit-at-a-time} implies
7760 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7764 @item -fno-toplevel-reorder
7765 @opindex fno-toplevel-reorder
7766 Do not reorder top-level functions, variables, and @code{asm}
7767 statements. Output them in the same order that they appear in the
7768 input file. When this option is used, unreferenced static variables
7769 will not be removed. This option is intended to support existing code
7770 that relies on a particular ordering. For new code, it is better to
7773 Enabled at level @option{-O0}. When disabled explicitly, it also implies
7774 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
7779 Constructs webs as commonly used for register allocation purposes and assign
7780 each web individual pseudo register. This allows the register allocation pass
7781 to operate on pseudos directly, but also strengthens several other optimization
7782 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7783 however, make debugging impossible, since variables will no longer stay in a
7786 Enabled by default with @option{-funroll-loops}.
7788 @item -fwhole-program
7789 @opindex fwhole-program
7790 Assume that the current compilation unit represents the whole program being
7791 compiled. All public functions and variables with the exception of @code{main}
7792 and those merged by attribute @code{externally_visible} become static functions
7793 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.
7794 While this option is equivalent to proper use of the @code{static} keyword for
7795 programs consisting of a single file, in combination with option
7796 @option{-flto} this flag can be used to
7797 compile many smaller scale programs since the functions and variables become
7798 local for the whole combined compilation unit, not for the single source file
7801 This option implies @option{-fwhole-file} for Fortran programs.
7803 @item -flto[=@var{n}]
7805 This option runs the standard link-time optimizer. When invoked
7806 with source code, it generates GIMPLE (one of GCC's internal
7807 representations) and writes it to special ELF sections in the object
7808 file. When the object files are linked together, all the function
7809 bodies are read from these ELF sections and instantiated as if they
7810 had been part of the same translation unit.
7812 To use the link-time optimizer, @option{-flto} needs to be specified at
7813 compile time and during the final link. For example:
7816 gcc -c -O2 -flto foo.c
7817 gcc -c -O2 -flto bar.c
7818 gcc -o myprog -flto -O2 foo.o bar.o
7821 The first two invocations to GCC save a bytecode representation
7822 of GIMPLE into special ELF sections inside @file{foo.o} and
7823 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
7824 @file{foo.o} and @file{bar.o}, merges the two files into a single
7825 internal image, and compiles the result as usual. Since both
7826 @file{foo.o} and @file{bar.o} are merged into a single image, this
7827 causes all the interprocedural analyses and optimizations in GCC to
7828 work across the two files as if they were a single one. This means,
7829 for example, that the inliner is able to inline functions in
7830 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7832 Another (simpler) way to enable link-time optimization is:
7835 gcc -o myprog -flto -O2 foo.c bar.c
7838 The above generates bytecode for @file{foo.c} and @file{bar.c},
7839 merges them together into a single GIMPLE representation and optimizes
7840 them as usual to produce @file{myprog}.
7842 The only important thing to keep in mind is that to enable link-time
7843 optimizations the @option{-flto} flag needs to be passed to both the
7844 compile and the link commands.
7846 To make whole program optimization effective, it is necessary to make
7847 certain whole program assumptions. The compiler needs to know
7848 what functions and variables can be accessed by libraries and runtime
7849 outside of the link-time optimized unit. When supported by the linker,
7850 the linker plugin (see @option{-fuse-linker-plugin}) passes information
7851 to the compiler about used and externally visible symbols. When
7852 the linker plugin is not available, @option{-fwhole-program} should be
7853 used to allow the compiler to make these assumptions, which leads
7854 to more aggressive optimization decisions.
7856 Note that when a file is compiled with @option{-flto}, the generated
7857 object file is larger than a regular object file because it
7858 contains GIMPLE bytecodes and the usual final code. This means that
7859 object files with LTO information can be linked as normal object
7860 files; if @option{-flto} is not passed to the linker, no
7861 interprocedural optimizations are applied.
7863 Additionally, the optimization flags used to compile individual files
7864 are not necessarily related to those used at link time. For instance,
7867 gcc -c -O0 -flto foo.c
7868 gcc -c -O0 -flto bar.c
7869 gcc -o myprog -flto -O3 foo.o bar.o
7872 This produces individual object files with unoptimized assembler
7873 code, but the resulting binary @file{myprog} is optimized at
7874 @option{-O3}. If, instead, the final binary is generated without
7875 @option{-flto}, then @file{myprog} is not optimized.
7877 When producing the final binary with @option{-flto}, GCC only
7878 applies link-time optimizations to those files that contain bytecode.
7879 Therefore, you can mix and match object files and libraries with
7880 GIMPLE bytecodes and final object code. GCC automatically selects
7881 which files to optimize in LTO mode and which files to link without
7884 There are some code generation flags preserved by GCC when
7885 generating bytecodes, as they need to be used during the final link
7886 stage. Currently, the following options are saved into the GIMPLE
7887 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7888 @option{-m} target flags.
7890 At link time, these options are read in and reapplied. Note that the
7891 current implementation makes no attempt to recognize conflicting
7892 values for these options. If different files have conflicting option
7893 values (e.g., one file is compiled with @option{-fPIC} and another
7894 isn't), the compiler simply uses the last value read from the
7895 bytecode files. It is recommended, then, that you compile all the files
7896 participating in the same link with the same options.
7898 If LTO encounters objects with C linkage declared with incompatible
7899 types in separate translation units to be linked together (undefined
7900 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7901 issued. The behavior is still undefined at run time.
7903 Another feature of LTO is that it is possible to apply interprocedural
7904 optimizations on files written in different languages. This requires
7905 support in the language front end. Currently, the C, C++ and
7906 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7907 something like this should work:
7912 gfortran -c -flto baz.f90
7913 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7916 Notice that the final link is done with @command{g++} to get the C++
7917 runtime libraries and @option{-lgfortran} is added to get the Fortran
7918 runtime libraries. In general, when mixing languages in LTO mode, you
7919 should use the same link command options as when mixing languages in a
7920 regular (non-LTO) compilation; all you need to add is @option{-flto} to
7921 all the compile and link commands.
7923 If object files containing GIMPLE bytecode are stored in a library archive, say
7924 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7925 are using a linker with plugin support. To enable this feature, use
7926 the flag @option{-fuse-linker-plugin} at link time:
7929 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7932 With the linker plugin enabled, the linker extracts the needed
7933 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
7934 to make them part of the aggregated GIMPLE image to be optimized.
7936 If you are not using a linker with plugin support and/or do not
7937 enable the linker plugin, then the objects inside @file{libfoo.a}
7938 are extracted and linked as usual, but they do not participate
7939 in the LTO optimization process.
7941 Link-time optimizations do not require the presence of the whole program to
7942 operate. If the program does not require any symbols to be exported, it is
7943 possible to combine @option{-flto} and @option{-fwhole-program} to allow
7944 the interprocedural optimizers to use more aggressive assumptions which may
7945 lead to improved optimization opportunities.
7946 Use of @option{-fwhole-program} is not needed when linker plugin is
7947 active (see @option{-fuse-linker-plugin}).
7949 The current implementation of LTO makes no
7950 attempt to generate bytecode that is portable between different
7951 types of hosts. The bytecode files are versioned and there is a
7952 strict version check, so bytecode files generated in one version of
7953 GCC will not work with an older/newer version of GCC.
7955 Link-time optimization does not work well with generation of debugging
7956 information. Combining @option{-flto} with
7957 @option{-g} is currently experimental and expected to produce wrong
7960 If you specify the optional @var{n}, the optimization and code
7961 generation done at link time is executed in parallel using @var{n}
7962 parallel jobs by utilizing an installed @command{make} program. The
7963 environment variable @env{MAKE} may be used to override the program
7964 used. The default value for @var{n} is 1.
7966 You can also specify @option{-flto=jobserver} to use GNU make's
7967 job server mode to determine the number of parallel jobs. This
7968 is useful when the Makefile calling GCC is already executing in parallel.
7969 You must prepend a @samp{+} to the command recipe in the parent Makefile
7970 for this to work. This option likely only works if @env{MAKE} is
7973 This option is disabled by default
7975 @item -flto-partition=@var{alg}
7976 @opindex flto-partition
7977 Specify the partitioning algorithm used by the link-time optimizer.
7978 The value is either @code{1to1} to specify a partitioning mirroring
7979 the original source files or @code{balanced} to specify partitioning
7980 into equally sized chunks (whenever possible). Specifying @code{none}
7981 as an algorithm disables partitioning and streaming completely. The
7982 default value is @code{balanced}.
7984 @item -flto-compression-level=@var{n}
7985 This option specifies the level of compression used for intermediate
7986 language written to LTO object files, and is only meaningful in
7987 conjunction with LTO mode (@option{-flto}). Valid
7988 values are 0 (no compression) to 9 (maximum compression). Values
7989 outside this range are clamped to either 0 or 9. If the option is not
7990 given, a default balanced compression setting is used.
7993 Prints a report with internal details on the workings of the link-time
7994 optimizer. The contents of this report vary from version to version.
7995 It is meant to be useful to GCC developers when processing object
7996 files in LTO mode (via @option{-flto}).
7998 Disabled by default.
8000 @item -fuse-linker-plugin
8001 Enables the use of a linker plugin during link-time optimization. This
8002 option relies on plugin support in the linker, which is available in gold
8003 or in GNU ld 2.21 or newer.
8005 This option enables the extraction of object files with GIMPLE bytecode out
8006 of library archives. This improves the quality of optimization by exposing
8007 more code to the link-time optimizer. This information specifies what
8008 symbols can be accessed externally (by non-LTO object or during dynamic
8009 linking). Resulting code quality improvements on binaries (and shared
8010 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
8011 See @option{-flto} for a description of the effect of this flag and how to
8014 This option is enabled by default when LTO support in GCC is enabled
8015 and GCC was configured for use with
8016 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8018 @item -ffat-lto-objects
8019 @opindex ffat-lto-objects
8020 Fat LTO objects are object files that contain both the intermediate language
8021 and the object code. This makes them usable for both LTO linking and normal
8022 linking. This option is effective only when compiling with @option{-flto}
8023 and is ignored at link time.
8025 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8026 requires the complete toolchain to be aware of LTO. It requires a linker with
8027 linker plugin support for basic functionality. Additionally, nm, ar and ranlib
8028 need to support linker plugins to allow a full-featured build environment
8029 (capable of building static libraries etc).
8031 The default is @option{-ffat-lto-objects} but this default is intended to
8032 change in future releases when linker plugin enabled environments become more
8035 @item -fcompare-elim
8036 @opindex fcompare-elim
8037 After register allocation and post-register allocation instruction splitting,
8038 identify arithmetic instructions that compute processor flags similar to a
8039 comparison operation based on that arithmetic. If possible, eliminate the
8040 explicit comparison operation.
8042 This pass only applies to certain targets that cannot explicitly represent
8043 the comparison operation before register allocation is complete.
8045 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8047 @item -fcprop-registers
8048 @opindex fcprop-registers
8049 After register allocation and post-register allocation instruction splitting,
8050 we perform a copy-propagation pass to try to reduce scheduling dependencies
8051 and occasionally eliminate the copy.
8053 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8055 @item -fprofile-correction
8056 @opindex fprofile-correction
8057 Profiles collected using an instrumented binary for multi-threaded programs may
8058 be inconsistent due to missed counter updates. When this option is specified,
8059 GCC will use heuristics to correct or smooth out such inconsistencies. By
8060 default, GCC will emit an error message when an inconsistent profile is detected.
8062 @item -fprofile-dir=@var{path}
8063 @opindex fprofile-dir
8065 Set the directory to search for the profile data files in to @var{path}.
8066 This option affects only the profile data generated by
8067 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
8068 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
8069 and its related options. Both absolute and relative paths can be used.
8070 By default, GCC will use the current directory as @var{path}, thus the
8071 profile data file will appear in the same directory as the object file.
8073 @item -fprofile-generate
8074 @itemx -fprofile-generate=@var{path}
8075 @opindex fprofile-generate
8077 Enable options usually used for instrumenting application to produce
8078 profile useful for later recompilation with profile feedback based
8079 optimization. You must use @option{-fprofile-generate} both when
8080 compiling and when linking your program.
8082 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
8084 If @var{path} is specified, GCC will look at the @var{path} to find
8085 the profile feedback data files. See @option{-fprofile-dir}.
8088 @itemx -fprofile-use=@var{path}
8089 @opindex fprofile-use
8090 Enable profile feedback directed optimizations, and optimizations
8091 generally profitable only with profile feedback available.
8093 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
8094 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
8096 By default, GCC emits an error message if the feedback profiles do not
8097 match the source code. This error can be turned into a warning by using
8098 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8101 If @var{path} is specified, GCC will look at the @var{path} to find
8102 the profile feedback data files. See @option{-fprofile-dir}.
8105 The following options control compiler behavior regarding floating-point
8106 arithmetic. These options trade off between speed and
8107 correctness. All must be specifically enabled.
8111 @opindex ffloat-store
8112 Do not store floating-point variables in registers, and inhibit other
8113 options that might change whether a floating-point value is taken from a
8116 @cindex floating-point precision
8117 This option prevents undesirable excess precision on machines such as
8118 the 68000 where the floating registers (of the 68881) keep more
8119 precision than a @code{double} is supposed to have. Similarly for the
8120 x86 architecture. For most programs, the excess precision does only
8121 good, but a few programs rely on the precise definition of IEEE floating
8122 point. Use @option{-ffloat-store} for such programs, after modifying
8123 them to store all pertinent intermediate computations into variables.
8125 @item -fexcess-precision=@var{style}
8126 @opindex fexcess-precision
8127 This option allows further control over excess precision on machines
8128 where floating-point registers have more precision than the IEEE
8129 @code{float} and @code{double} types and the processor does not
8130 support operations rounding to those types. By default,
8131 @option{-fexcess-precision=fast} is in effect; this means that
8132 operations are carried out in the precision of the registers and that
8133 it is unpredictable when rounding to the types specified in the source
8134 code takes place. When compiling C, if
8135 @option{-fexcess-precision=standard} is specified then excess
8136 precision will follow the rules specified in ISO C99; in particular,
8137 both casts and assignments cause values to be rounded to their
8138 semantic types (whereas @option{-ffloat-store} only affects
8139 assignments). This option is enabled by default for C if a strict
8140 conformance option such as @option{-std=c99} is used.
8143 @option{-fexcess-precision=standard} is not implemented for languages
8144 other than C, and has no effect if
8145 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8146 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8147 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8148 semantics apply without excess precision, and in the latter, rounding
8153 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8154 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8155 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8157 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8159 This option is not turned on by any @option{-O} option besides
8160 @option{-Ofast} since it can result in incorrect output for programs
8161 that depend on an exact implementation of IEEE or ISO rules/specifications
8162 for math functions. It may, however, yield faster code for programs
8163 that do not require the guarantees of these specifications.
8165 @item -fno-math-errno
8166 @opindex fno-math-errno
8167 Do not set ERRNO after calling math functions that are executed
8168 with a single instruction, e.g., sqrt. A program that relies on
8169 IEEE exceptions for math error handling may want to use this flag
8170 for speed while maintaining IEEE arithmetic compatibility.
8172 This option is not turned on by any @option{-O} option since
8173 it can result in incorrect output for programs that depend on
8174 an exact implementation of IEEE or ISO rules/specifications for
8175 math functions. It may, however, yield faster code for programs
8176 that do not require the guarantees of these specifications.
8178 The default is @option{-fmath-errno}.
8180 On Darwin systems, the math library never sets @code{errno}. There is
8181 therefore no reason for the compiler to consider the possibility that
8182 it might, and @option{-fno-math-errno} is the default.
8184 @item -funsafe-math-optimizations
8185 @opindex funsafe-math-optimizations
8187 Allow optimizations for floating-point arithmetic that (a) assume
8188 that arguments and results are valid and (b) may violate IEEE or
8189 ANSI standards. When used at link-time, it may include libraries
8190 or startup files that change the default FPU control word or other
8191 similar optimizations.
8193 This option is not turned on by any @option{-O} option since
8194 it can result in incorrect output for programs that depend on
8195 an exact implementation of IEEE or ISO rules/specifications for
8196 math functions. It may, however, yield faster code for programs
8197 that do not require the guarantees of these specifications.
8198 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8199 @option{-fassociative-math} and @option{-freciprocal-math}.
8201 The default is @option{-fno-unsafe-math-optimizations}.
8203 @item -fassociative-math
8204 @opindex fassociative-math
8206 Allow re-association of operands in series of floating-point operations.
8207 This violates the ISO C and C++ language standard by possibly changing
8208 computation result. NOTE: re-ordering may change the sign of zero as
8209 well as ignore NaNs and inhibit or create underflow or overflow (and
8210 thus cannot be used on code that relies on rounding behavior like
8211 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8212 and thus may not be used when ordered comparisons are required.
8213 This option requires that both @option{-fno-signed-zeros} and
8214 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8215 much sense with @option{-frounding-math}. For Fortran the option
8216 is automatically enabled when both @option{-fno-signed-zeros} and
8217 @option{-fno-trapping-math} are in effect.
8219 The default is @option{-fno-associative-math}.
8221 @item -freciprocal-math
8222 @opindex freciprocal-math
8224 Allow the reciprocal of a value to be used instead of dividing by
8225 the value if this enables optimizations. For example @code{x / y}
8226 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
8227 is subject to common subexpression elimination. Note that this loses
8228 precision and increases the number of flops operating on the value.
8230 The default is @option{-fno-reciprocal-math}.
8232 @item -ffinite-math-only
8233 @opindex ffinite-math-only
8234 Allow optimizations for floating-point arithmetic that assume
8235 that arguments and results are not NaNs or +-Infs.
8237 This option is not turned on by any @option{-O} option since
8238 it can result in incorrect output for programs that depend on
8239 an exact implementation of IEEE or ISO rules/specifications for
8240 math functions. It may, however, yield faster code for programs
8241 that do not require the guarantees of these specifications.
8243 The default is @option{-fno-finite-math-only}.
8245 @item -fno-signed-zeros
8246 @opindex fno-signed-zeros
8247 Allow optimizations for floating-point arithmetic that ignore the
8248 signedness of zero. IEEE arithmetic specifies the behavior of
8249 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8250 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8251 This option implies that the sign of a zero result isn't significant.
8253 The default is @option{-fsigned-zeros}.
8255 @item -fno-trapping-math
8256 @opindex fno-trapping-math
8257 Compile code assuming that floating-point operations cannot generate
8258 user-visible traps. These traps include division by zero, overflow,
8259 underflow, inexact result and invalid operation. This option requires
8260 that @option{-fno-signaling-nans} be in effect. Setting this option may
8261 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8263 This option should never be turned on by any @option{-O} option since
8264 it can result in incorrect output for programs that depend on
8265 an exact implementation of IEEE or ISO rules/specifications for
8268 The default is @option{-ftrapping-math}.
8270 @item -frounding-math
8271 @opindex frounding-math
8272 Disable transformations and optimizations that assume default floating-point
8273 rounding behavior. This is round-to-zero for all floating point
8274 to integer conversions, and round-to-nearest for all other arithmetic
8275 truncations. This option should be specified for programs that change
8276 the FP rounding mode dynamically, or that may be executed with a
8277 non-default rounding mode. This option disables constant folding of
8278 floating-point expressions at compile time (which may be affected by
8279 rounding mode) and arithmetic transformations that are unsafe in the
8280 presence of sign-dependent rounding modes.
8282 The default is @option{-fno-rounding-math}.
8284 This option is experimental and does not currently guarantee to
8285 disable all GCC optimizations that are affected by rounding mode.
8286 Future versions of GCC may provide finer control of this setting
8287 using C99's @code{FENV_ACCESS} pragma. This command-line option
8288 will be used to specify the default state for @code{FENV_ACCESS}.
8290 @item -fsignaling-nans
8291 @opindex fsignaling-nans
8292 Compile code assuming that IEEE signaling NaNs may generate user-visible
8293 traps during floating-point operations. Setting this option disables
8294 optimizations that may change the number of exceptions visible with
8295 signaling NaNs. This option implies @option{-ftrapping-math}.
8297 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8300 The default is @option{-fno-signaling-nans}.
8302 This option is experimental and does not currently guarantee to
8303 disable all GCC optimizations that affect signaling NaN behavior.
8305 @item -fsingle-precision-constant
8306 @opindex fsingle-precision-constant
8307 Treat floating-point constants as single precision instead of
8308 implicitly converting them to double-precision constants.
8310 @item -fcx-limited-range
8311 @opindex fcx-limited-range
8312 When enabled, this option states that a range reduction step is not
8313 needed when performing complex division. Also, there is no checking
8314 whether the result of a complex multiplication or division is @code{NaN
8315 + I*NaN}, with an attempt to rescue the situation in that case. The
8316 default is @option{-fno-cx-limited-range}, but is enabled by
8317 @option{-ffast-math}.
8319 This option controls the default setting of the ISO C99
8320 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8323 @item -fcx-fortran-rules
8324 @opindex fcx-fortran-rules
8325 Complex multiplication and division follow Fortran rules. Range
8326 reduction is done as part of complex division, but there is no checking
8327 whether the result of a complex multiplication or division is @code{NaN
8328 + I*NaN}, with an attempt to rescue the situation in that case.
8330 The default is @option{-fno-cx-fortran-rules}.
8334 The following options control optimizations that may improve
8335 performance, but are not enabled by any @option{-O} options. This
8336 section includes experimental options that may produce broken code.
8339 @item -fbranch-probabilities
8340 @opindex fbranch-probabilities
8341 After running a program compiled with @option{-fprofile-arcs}
8342 (@pxref{Debugging Options,, Options for Debugging Your Program or
8343 @command{gcc}}), you can compile it a second time using
8344 @option{-fbranch-probabilities}, to improve optimizations based on
8345 the number of times each branch was taken. When the program
8346 compiled with @option{-fprofile-arcs} exits it saves arc execution
8347 counts to a file called @file{@var{sourcename}.gcda} for each source
8348 file. The information in this data file is very dependent on the
8349 structure of the generated code, so you must use the same source code
8350 and the same optimization options for both compilations.
8352 With @option{-fbranch-probabilities}, GCC puts a
8353 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8354 These can be used to improve optimization. Currently, they are only
8355 used in one place: in @file{reorg.c}, instead of guessing which path a
8356 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8357 exactly determine which path is taken more often.
8359 @item -fprofile-values
8360 @opindex fprofile-values
8361 If combined with @option{-fprofile-arcs}, it adds code so that some
8362 data about values of expressions in the program is gathered.
8364 With @option{-fbranch-probabilities}, it reads back the data gathered
8365 from profiling values of expressions for usage in optimizations.
8367 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8371 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8372 a code to gather information about values of expressions.
8374 With @option{-fbranch-probabilities}, it reads back the data gathered
8375 and actually performs the optimizations based on them.
8376 Currently the optimizations include specialization of division operation
8377 using the knowledge about the value of the denominator.
8379 @item -frename-registers
8380 @opindex frename-registers
8381 Attempt to avoid false dependencies in scheduled code by making use
8382 of registers left over after register allocation. This optimization
8383 will most benefit processors with lots of registers. Depending on the
8384 debug information format adopted by the target, however, it can
8385 make debugging impossible, since variables will no longer stay in
8386 a ``home register''.
8388 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8392 Perform tail duplication to enlarge superblock size. This transformation
8393 simplifies the control flow of the function allowing other optimizations to do
8396 Enabled with @option{-fprofile-use}.
8398 @item -funroll-loops
8399 @opindex funroll-loops
8400 Unroll loops whose number of iterations can be determined at compile time or
8401 upon entry to the loop. @option{-funroll-loops} implies
8402 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8403 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8404 small constant number of iterations). This option makes code larger, and may
8405 or may not make it run faster.
8407 Enabled with @option{-fprofile-use}.
8409 @item -funroll-all-loops
8410 @opindex funroll-all-loops
8411 Unroll all loops, even if their number of iterations is uncertain when
8412 the loop is entered. This usually makes programs run more slowly.
8413 @option{-funroll-all-loops} implies the same options as
8414 @option{-funroll-loops}.
8417 @opindex fpeel-loops
8418 Peels loops for which there is enough information that they do not
8419 roll much (from profile feedback). It also turns on complete loop peeling
8420 (i.e.@: complete removal of loops with small constant number of iterations).
8422 Enabled with @option{-fprofile-use}.
8424 @item -fmove-loop-invariants
8425 @opindex fmove-loop-invariants
8426 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8427 at level @option{-O1}
8429 @item -funswitch-loops
8430 @opindex funswitch-loops
8431 Move branches with loop invariant conditions out of the loop, with duplicates
8432 of the loop on both branches (modified according to result of the condition).
8434 @item -ffunction-sections
8435 @itemx -fdata-sections
8436 @opindex ffunction-sections
8437 @opindex fdata-sections
8438 Place each function or data item into its own section in the output
8439 file if the target supports arbitrary sections. The name of the
8440 function or the name of the data item determines the section's name
8443 Use these options on systems where the linker can perform optimizations
8444 to improve locality of reference in the instruction space. Most systems
8445 using the ELF object format and SPARC processors running Solaris 2 have
8446 linkers with such optimizations. AIX may have these optimizations in
8449 Only use these options when there are significant benefits from doing
8450 so. When you specify these options, the assembler and linker will
8451 create larger object and executable files and will also be slower.
8452 You will not be able to use @code{gprof} on all systems if you
8453 specify this option and you may have problems with debugging if
8454 you specify both this option and @option{-g}.
8456 @item -fbranch-target-load-optimize
8457 @opindex fbranch-target-load-optimize
8458 Perform branch target register load optimization before prologue / epilogue
8460 The use of target registers can typically be exposed only during reload,
8461 thus hoisting loads out of loops and doing inter-block scheduling needs
8462 a separate optimization pass.
8464 @item -fbranch-target-load-optimize2
8465 @opindex fbranch-target-load-optimize2
8466 Perform branch target register load optimization after prologue / epilogue
8469 @item -fbtr-bb-exclusive
8470 @opindex fbtr-bb-exclusive
8471 When performing branch target register load optimization, don't reuse
8472 branch target registers in within any basic block.
8474 @item -fstack-protector
8475 @opindex fstack-protector
8476 Emit extra code to check for buffer overflows, such as stack smashing
8477 attacks. This is done by adding a guard variable to functions with
8478 vulnerable objects. This includes functions that call alloca, and
8479 functions with buffers larger than 8 bytes. The guards are initialized
8480 when a function is entered and then checked when the function exits.
8481 If a guard check fails, an error message is printed and the program exits.
8483 @item -fstack-protector-all
8484 @opindex fstack-protector-all
8485 Like @option{-fstack-protector} except that all functions are protected.
8487 @item -fsection-anchors
8488 @opindex fsection-anchors
8489 Try to reduce the number of symbolic address calculations by using
8490 shared ``anchor'' symbols to address nearby objects. This transformation
8491 can help to reduce the number of GOT entries and GOT accesses on some
8494 For example, the implementation of the following function @code{foo}:
8498 int foo (void) @{ return a + b + c; @}
8501 would usually calculate the addresses of all three variables, but if you
8502 compile it with @option{-fsection-anchors}, it will access the variables
8503 from a common anchor point instead. The effect is similar to the
8504 following pseudocode (which isn't valid C):
8509 register int *xr = &x;
8510 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8514 Not all targets support this option.
8516 @item --param @var{name}=@var{value}
8518 In some places, GCC uses various constants to control the amount of
8519 optimization that is done. For example, GCC will not inline functions
8520 that contain more than a certain number of instructions. You can
8521 control some of these constants on the command line using the
8522 @option{--param} option.
8524 The names of specific parameters, and the meaning of the values, are
8525 tied to the internals of the compiler, and are subject to change
8526 without notice in future releases.
8528 In each case, the @var{value} is an integer. The allowable choices for
8529 @var{name} are given in the following table:
8532 @item predictable-branch-outcome
8533 When branch is predicted to be taken with probability lower than this threshold
8534 (in percent), then it is considered well predictable. The default is 10.
8536 @item max-crossjump-edges
8537 The maximum number of incoming edges to consider for crossjumping.
8538 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8539 the number of edges incoming to each block. Increasing values mean
8540 more aggressive optimization, making the compilation time increase with
8541 probably small improvement in executable size.
8543 @item min-crossjump-insns
8544 The minimum number of instructions that must be matched at the end
8545 of two blocks before crossjumping will be performed on them. This
8546 value is ignored in the case where all instructions in the block being
8547 crossjumped from are matched. The default value is 5.
8549 @item max-grow-copy-bb-insns
8550 The maximum code size expansion factor when copying basic blocks
8551 instead of jumping. The expansion is relative to a jump instruction.
8552 The default value is 8.
8554 @item max-goto-duplication-insns
8555 The maximum number of instructions to duplicate to a block that jumps
8556 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8557 passes, GCC factors computed gotos early in the compilation process,
8558 and unfactors them as late as possible. Only computed jumps at the
8559 end of a basic blocks with no more than max-goto-duplication-insns are
8560 unfactored. The default value is 8.
8562 @item max-delay-slot-insn-search
8563 The maximum number of instructions to consider when looking for an
8564 instruction to fill a delay slot. If more than this arbitrary number of
8565 instructions is searched, the time savings from filling the delay slot
8566 will be minimal so stop searching. Increasing values mean more
8567 aggressive optimization, making the compilation time increase with probably
8568 small improvement in execution time.
8570 @item max-delay-slot-live-search
8571 When trying to fill delay slots, the maximum number of instructions to
8572 consider when searching for a block with valid live register
8573 information. Increasing this arbitrarily chosen value means more
8574 aggressive optimization, increasing the compilation time. This parameter
8575 should be removed when the delay slot code is rewritten to maintain the
8578 @item max-gcse-memory
8579 The approximate maximum amount of memory that will be allocated in
8580 order to perform the global common subexpression elimination
8581 optimization. If more memory than specified is required, the
8582 optimization will not be done.
8584 @item max-gcse-insertion-ratio
8585 If the ratio of expression insertions to deletions is larger than this value
8586 for any expression, then RTL PRE will insert or remove the expression and thus
8587 leave partially redundant computations in the instruction stream. The default value is 20.
8589 @item max-pending-list-length
8590 The maximum number of pending dependencies scheduling will allow
8591 before flushing the current state and starting over. Large functions
8592 with few branches or calls can create excessively large lists which
8593 needlessly consume memory and resources.
8595 @item max-modulo-backtrack-attempts
8596 The maximum number of backtrack attempts the scheduler should make
8597 when modulo scheduling a loop. Larger values can exponentially increase
8600 @item max-inline-insns-single
8601 Several parameters control the tree inliner used in gcc.
8602 This number sets the maximum number of instructions (counted in GCC's
8603 internal representation) in a single function that the tree inliner
8604 will consider for inlining. This only affects functions declared
8605 inline and methods implemented in a class declaration (C++).
8606 The default value is 400.
8608 @item max-inline-insns-auto
8609 When you use @option{-finline-functions} (included in @option{-O3}),
8610 a lot of functions that would otherwise not be considered for inlining
8611 by the compiler will be investigated. To those functions, a different
8612 (more restrictive) limit compared to functions declared inline can
8614 The default value is 40.
8616 @item large-function-insns
8617 The limit specifying really large functions. For functions larger than this
8618 limit after inlining, inlining is constrained by
8619 @option{--param large-function-growth}. This parameter is useful primarily
8620 to avoid extreme compilation time caused by non-linear algorithms used by the
8622 The default value is 2700.
8624 @item large-function-growth
8625 Specifies maximal growth of large function caused by inlining in percents.
8626 The default value is 100 which limits large function growth to 2.0 times
8629 @item large-unit-insns
8630 The limit specifying large translation unit. Growth caused by inlining of
8631 units larger than this limit is limited by @option{--param inline-unit-growth}.
8632 For small units this might be too tight (consider unit consisting of function A
8633 that is inline and B that just calls A three time. If B is small relative to
8634 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8635 large units consisting of small inlineable functions however the overall unit
8636 growth limit is needed to avoid exponential explosion of code size. Thus for
8637 smaller units, the size is increased to @option{--param large-unit-insns}
8638 before applying @option{--param inline-unit-growth}. The default is 10000
8640 @item inline-unit-growth
8641 Specifies maximal overall growth of the compilation unit caused by inlining.
8642 The default value is 30 which limits unit growth to 1.3 times the original
8645 @item ipcp-unit-growth
8646 Specifies maximal overall growth of the compilation unit caused by
8647 interprocedural constant propagation. The default value is 10 which limits
8648 unit growth to 1.1 times the original size.
8650 @item large-stack-frame
8651 The limit specifying large stack frames. While inlining the algorithm is trying
8652 to not grow past this limit too much. Default value is 256 bytes.
8654 @item large-stack-frame-growth
8655 Specifies maximal growth of large stack frames caused by inlining in percents.
8656 The default value is 1000 which limits large stack frame growth to 11 times
8659 @item max-inline-insns-recursive
8660 @itemx max-inline-insns-recursive-auto
8661 Specifies maximum number of instructions out-of-line copy of self recursive inline
8662 function can grow into by performing recursive inlining.
8664 For functions declared inline @option{--param max-inline-insns-recursive} is
8665 taken into account. For function not declared inline, recursive inlining
8666 happens only when @option{-finline-functions} (included in @option{-O3}) is
8667 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8668 default value is 450.
8670 @item max-inline-recursive-depth
8671 @itemx max-inline-recursive-depth-auto
8672 Specifies maximum recursion depth used by the recursive inlining.
8674 For functions declared inline @option{--param max-inline-recursive-depth} is
8675 taken into account. For function not declared inline, recursive inlining
8676 happens only when @option{-finline-functions} (included in @option{-O3}) is
8677 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8680 @item min-inline-recursive-probability
8681 Recursive inlining is profitable only for function having deep recursion
8682 in average and can hurt for function having little recursion depth by
8683 increasing the prologue size or complexity of function body to other
8686 When profile feedback is available (see @option{-fprofile-generate}) the actual
8687 recursion depth can be guessed from probability that function will recurse via
8688 given call expression. This parameter limits inlining only to call expression
8689 whose probability exceeds given threshold (in percents). The default value is
8692 @item early-inlining-insns
8693 Specify growth that early inliner can make. In effect it increases amount of
8694 inlining for code having large abstraction penalty. The default value is 10.
8696 @item max-early-inliner-iterations
8697 @itemx max-early-inliner-iterations
8698 Limit of iterations of early inliner. This basically bounds number of nested
8699 indirect calls early inliner can resolve. Deeper chains are still handled by
8702 @item comdat-sharing-probability
8703 @itemx comdat-sharing-probability
8704 Probability (in percent) that C++ inline function with comdat visibility
8705 will be shared across multiple compilation units. The default value is 20.
8707 @item min-vect-loop-bound
8708 The minimum number of iterations under which a loop will not get vectorized
8709 when @option{-ftree-vectorize} is used. The number of iterations after
8710 vectorization needs to be greater than the value specified by this option
8711 to allow vectorization. The default value is 0.
8713 @item gcse-cost-distance-ratio
8714 Scaling factor in calculation of maximum distance an expression
8715 can be moved by GCSE optimizations. This is currently supported only in the
8716 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8717 will be with simple expressions, i.e., the expressions that have cost
8718 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8719 hoisting of simple expressions. The default value is 10.
8721 @item gcse-unrestricted-cost
8722 Cost, roughly measured as the cost of a single typical machine
8723 instruction, at which GCSE optimizations will not constrain
8724 the distance an expression can travel. This is currently
8725 supported only in the code hoisting pass. The lesser the cost,
8726 the more aggressive code hoisting will be. Specifying 0 will
8727 allow all expressions to travel unrestricted distances.
8728 The default value is 3.
8730 @item max-hoist-depth
8731 The depth of search in the dominator tree for expressions to hoist.
8732 This is used to avoid quadratic behavior in hoisting algorithm.
8733 The value of 0 will avoid limiting the search, but may slow down compilation
8734 of huge functions. The default value is 30.
8736 @item max-tail-merge-comparisons
8737 The maximum amount of similar bbs to compare a bb with. This is used to
8738 avoid quadratic behavior in tree tail merging. The default value is 10.
8740 @item max-tail-merge-iterations
8741 The maximum amount of iterations of the pass over the function. This is used to
8742 limit compilation time in tree tail merging. The default value is 2.
8744 @item max-unrolled-insns
8745 The maximum number of instructions that a loop should have if that loop
8746 is unrolled, and if the loop is unrolled, it determines how many times
8747 the loop code is unrolled.
8749 @item max-average-unrolled-insns
8750 The maximum number of instructions biased by probabilities of their execution
8751 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8752 it determines how many times the loop code is unrolled.
8754 @item max-unroll-times
8755 The maximum number of unrollings of a single loop.
8757 @item max-peeled-insns
8758 The maximum number of instructions that a loop should have if that loop
8759 is peeled, and if the loop is peeled, it determines how many times
8760 the loop code is peeled.
8762 @item max-peel-times
8763 The maximum number of peelings of a single loop.
8765 @item max-completely-peeled-insns
8766 The maximum number of insns of a completely peeled loop.
8768 @item max-completely-peel-times
8769 The maximum number of iterations of a loop to be suitable for complete peeling.
8771 @item max-completely-peel-loop-nest-depth
8772 The maximum depth of a loop nest suitable for complete peeling.
8774 @item max-unswitch-insns
8775 The maximum number of insns of an unswitched loop.
8777 @item max-unswitch-level
8778 The maximum number of branches unswitched in a single loop.
8781 The minimum cost of an expensive expression in the loop invariant motion.
8783 @item iv-consider-all-candidates-bound
8784 Bound on number of candidates for induction variables below that
8785 all candidates are considered for each use in induction variable
8786 optimizations. Only the most relevant candidates are considered
8787 if there are more candidates, to avoid quadratic time complexity.
8789 @item iv-max-considered-uses
8790 The induction variable optimizations give up on loops that contain more
8791 induction variable uses.
8793 @item iv-always-prune-cand-set-bound
8794 If number of candidates in the set is smaller than this value,
8795 we always try to remove unnecessary ivs from the set during its
8796 optimization when a new iv is added to the set.
8798 @item scev-max-expr-size
8799 Bound on size of expressions used in the scalar evolutions analyzer.
8800 Large expressions slow the analyzer.
8802 @item scev-max-expr-complexity
8803 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8804 Complex expressions slow the analyzer.
8806 @item omega-max-vars
8807 The maximum number of variables in an Omega constraint system.
8808 The default value is 128.
8810 @item omega-max-geqs
8811 The maximum number of inequalities in an Omega constraint system.
8812 The default value is 256.
8815 The maximum number of equalities in an Omega constraint system.
8816 The default value is 128.
8818 @item omega-max-wild-cards
8819 The maximum number of wildcard variables that the Omega solver will
8820 be able to insert. The default value is 18.
8822 @item omega-hash-table-size
8823 The size of the hash table in the Omega solver. The default value is
8826 @item omega-max-keys
8827 The maximal number of keys used by the Omega solver. The default
8830 @item omega-eliminate-redundant-constraints
8831 When set to 1, use expensive methods to eliminate all redundant
8832 constraints. The default value is 0.
8834 @item vect-max-version-for-alignment-checks
8835 The maximum number of run-time checks that can be performed when
8836 doing loop versioning for alignment in the vectorizer. See option
8837 ftree-vect-loop-version for more information.
8839 @item vect-max-version-for-alias-checks
8840 The maximum number of run-time checks that can be performed when
8841 doing loop versioning for alias in the vectorizer. See option
8842 ftree-vect-loop-version for more information.
8844 @item max-iterations-to-track
8846 The maximum number of iterations of a loop the brute force algorithm
8847 for analysis of # of iterations of the loop tries to evaluate.
8849 @item hot-bb-count-fraction
8850 Select fraction of the maximal count of repetitions of basic block in program
8851 given basic block needs to have to be considered hot.
8853 @item hot-bb-frequency-fraction
8854 Select fraction of the entry block frequency of executions of basic block in
8855 function given basic block needs to have to be considered hot.
8857 @item max-predicted-iterations
8858 The maximum number of loop iterations we predict statically. This is useful
8859 in cases where function contain single loop with known bound and other loop
8860 with unknown. We predict the known number of iterations correctly, while
8861 the unknown number of iterations average to roughly 10. This means that the
8862 loop without bounds would appear artificially cold relative to the other one.
8864 @item align-threshold
8866 Select fraction of the maximal frequency of executions of basic block in
8867 function given basic block will get aligned.
8869 @item align-loop-iterations
8871 A loop expected to iterate at lest the selected number of iterations will get
8874 @item tracer-dynamic-coverage
8875 @itemx tracer-dynamic-coverage-feedback
8877 This value is used to limit superblock formation once the given percentage of
8878 executed instructions is covered. This limits unnecessary code size
8881 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8882 feedback is available. The real profiles (as opposed to statically estimated
8883 ones) are much less balanced allowing the threshold to be larger value.
8885 @item tracer-max-code-growth
8886 Stop tail duplication once code growth has reached given percentage. This is
8887 rather hokey argument, as most of the duplicates will be eliminated later in
8888 cross jumping, so it may be set to much higher values than is the desired code
8891 @item tracer-min-branch-ratio
8893 Stop reverse growth when the reverse probability of best edge is less than this
8894 threshold (in percent).
8896 @item tracer-min-branch-ratio
8897 @itemx tracer-min-branch-ratio-feedback
8899 Stop forward growth if the best edge do have probability lower than this
8902 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8903 compilation for profile feedback and one for compilation without. The value
8904 for compilation with profile feedback needs to be more conservative (higher) in
8905 order to make tracer effective.
8907 @item max-cse-path-length
8909 Maximum number of basic blocks on path that cse considers. The default is 10.
8912 The maximum instructions CSE process before flushing. The default is 1000.
8914 @item ggc-min-expand
8916 GCC uses a garbage collector to manage its own memory allocation. This
8917 parameter specifies the minimum percentage by which the garbage
8918 collector's heap should be allowed to expand between collections.
8919 Tuning this may improve compilation speed; it has no effect on code
8922 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8923 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8924 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8925 GCC is not able to calculate RAM on a particular platform, the lower
8926 bound of 30% is used. Setting this parameter and
8927 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8928 every opportunity. This is extremely slow, but can be useful for
8931 @item ggc-min-heapsize
8933 Minimum size of the garbage collector's heap before it begins bothering
8934 to collect garbage. The first collection occurs after the heap expands
8935 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8936 tuning this may improve compilation speed, and has no effect on code
8939 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
8940 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8941 with a lower bound of 4096 (four megabytes) and an upper bound of
8942 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8943 particular platform, the lower bound is used. Setting this parameter
8944 very large effectively disables garbage collection. Setting this
8945 parameter and @option{ggc-min-expand} to zero causes a full collection
8946 to occur at every opportunity.
8948 @item max-reload-search-insns
8949 The maximum number of instruction reload should look backward for equivalent
8950 register. Increasing values mean more aggressive optimization, making the
8951 compilation time increase with probably slightly better performance.
8952 The default value is 100.
8954 @item max-cselib-memory-locations
8955 The maximum number of memory locations cselib should take into account.
8956 Increasing values mean more aggressive optimization, making the compilation time
8957 increase with probably slightly better performance. The default value is 500.
8959 @item reorder-blocks-duplicate
8960 @itemx reorder-blocks-duplicate-feedback
8962 Used by basic block reordering pass to decide whether to use unconditional
8963 branch or duplicate the code on its destination. Code is duplicated when its
8964 estimated size is smaller than this value multiplied by the estimated size of
8965 unconditional jump in the hot spots of the program.
8967 The @option{reorder-block-duplicate-feedback} is used only when profile
8968 feedback is available and may be set to higher values than
8969 @option{reorder-block-duplicate} since information about the hot spots is more
8972 @item max-sched-ready-insns
8973 The maximum number of instructions ready to be issued the scheduler should
8974 consider at any given time during the first scheduling pass. Increasing
8975 values mean more thorough searches, making the compilation time increase
8976 with probably little benefit. The default value is 100.
8978 @item max-sched-region-blocks
8979 The maximum number of blocks in a region to be considered for
8980 interblock scheduling. The default value is 10.
8982 @item max-pipeline-region-blocks
8983 The maximum number of blocks in a region to be considered for
8984 pipelining in the selective scheduler. The default value is 15.
8986 @item max-sched-region-insns
8987 The maximum number of insns in a region to be considered for
8988 interblock scheduling. The default value is 100.
8990 @item max-pipeline-region-insns
8991 The maximum number of insns in a region to be considered for
8992 pipelining in the selective scheduler. The default value is 200.
8995 The minimum probability (in percents) of reaching a source block
8996 for interblock speculative scheduling. The default value is 40.
8998 @item max-sched-extend-regions-iters
8999 The maximum number of iterations through CFG to extend regions.
9000 0 - disable region extension,
9001 N - do at most N iterations.
9002 The default value is 0.
9004 @item max-sched-insn-conflict-delay
9005 The maximum conflict delay for an insn to be considered for speculative motion.
9006 The default value is 3.
9008 @item sched-spec-prob-cutoff
9009 The minimal probability of speculation success (in percents), so that
9010 speculative insn will be scheduled.
9011 The default value is 40.
9013 @item sched-mem-true-dep-cost
9014 Minimal distance (in CPU cycles) between store and load targeting same
9015 memory locations. The default value is 1.
9017 @item selsched-max-lookahead
9018 The maximum size of the lookahead window of selective scheduling. It is a
9019 depth of search for available instructions.
9020 The default value is 50.
9022 @item selsched-max-sched-times
9023 The maximum number of times that an instruction will be scheduled during
9024 selective scheduling. This is the limit on the number of iterations
9025 through which the instruction may be pipelined. The default value is 2.
9027 @item selsched-max-insns-to-rename
9028 The maximum number of best instructions in the ready list that are considered
9029 for renaming in the selective scheduler. The default value is 2.
9032 The minimum value of stage count that swing modulo scheduler will
9033 generate. The default value is 2.
9035 @item max-last-value-rtl
9036 The maximum size measured as number of RTLs that can be recorded in an expression
9037 in combiner for a pseudo register as last known value of that register. The default
9040 @item integer-share-limit
9041 Small integer constants can use a shared data structure, reducing the
9042 compiler's memory usage and increasing its speed. This sets the maximum
9043 value of a shared integer constant. The default value is 256.
9045 @item min-virtual-mappings
9046 Specifies the minimum number of virtual mappings in the incremental
9047 SSA updater that should be registered to trigger the virtual mappings
9048 heuristic defined by virtual-mappings-ratio. The default value is
9051 @item virtual-mappings-ratio
9052 If the number of virtual mappings is virtual-mappings-ratio bigger
9053 than the number of virtual symbols to be updated, then the incremental
9054 SSA updater switches to a full update for those symbols. The default
9057 @item ssp-buffer-size
9058 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
9059 protection when @option{-fstack-protection} is used.
9061 @item max-jump-thread-duplication-stmts
9062 Maximum number of statements allowed in a block that needs to be
9063 duplicated when threading jumps.
9065 @item max-fields-for-field-sensitive
9066 Maximum number of fields in a structure we will treat in
9067 a field sensitive manner during pointer analysis. The default is zero
9068 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
9070 @item prefetch-latency
9071 Estimate on average number of instructions that are executed before
9072 prefetch finishes. The distance we prefetch ahead is proportional
9073 to this constant. Increasing this number may also lead to less
9074 streams being prefetched (see @option{simultaneous-prefetches}).
9076 @item simultaneous-prefetches
9077 Maximum number of prefetches that can run at the same time.
9079 @item l1-cache-line-size
9080 The size of cache line in L1 cache, in bytes.
9083 The size of L1 cache, in kilobytes.
9086 The size of L2 cache, in kilobytes.
9088 @item min-insn-to-prefetch-ratio
9089 The minimum ratio between the number of instructions and the
9090 number of prefetches to enable prefetching in a loop.
9092 @item prefetch-min-insn-to-mem-ratio
9093 The minimum ratio between the number of instructions and the
9094 number of memory references to enable prefetching in a loop.
9096 @item use-canonical-types
9097 Whether the compiler should use the ``canonical'' type system. By
9098 default, this should always be 1, which uses a more efficient internal
9099 mechanism for comparing types in C++ and Objective-C++. However, if
9100 bugs in the canonical type system are causing compilation failures,
9101 set this value to 0 to disable canonical types.
9103 @item switch-conversion-max-branch-ratio
9104 Switch initialization conversion will refuse to create arrays that are
9105 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9106 branches in the switch.
9108 @item max-partial-antic-length
9109 Maximum length of the partial antic set computed during the tree
9110 partial redundancy elimination optimization (@option{-ftree-pre}) when
9111 optimizing at @option{-O3} and above. For some sorts of source code
9112 the enhanced partial redundancy elimination optimization can run away,
9113 consuming all of the memory available on the host machine. This
9114 parameter sets a limit on the length of the sets that are computed,
9115 which prevents the runaway behavior. Setting a value of 0 for
9116 this parameter will allow an unlimited set length.
9118 @item sccvn-max-scc-size
9119 Maximum size of a strongly connected component (SCC) during SCCVN
9120 processing. If this limit is hit, SCCVN processing for the whole
9121 function will not be done and optimizations depending on it will
9122 be disabled. The default maximum SCC size is 10000.
9124 @item ira-max-loops-num
9125 IRA uses regional register allocation by default. If a function
9126 contains more loops than the number given by this parameter, only at most
9127 the given number of the most frequently-executed loops form regions
9128 for regional register allocation. The default value of the
9131 @item ira-max-conflict-table-size
9132 Although IRA uses a sophisticated algorithm to compress the conflict
9133 table, the table can still require excessive amounts of memory for
9134 huge functions. If the conflict table for a function could be more
9135 than the size in MB given by this parameter, the register allocator
9136 instead uses a faster, simpler, and lower-quality
9137 algorithm that does not require building a pseudo-register conflict table.
9138 The default value of the parameter is 2000.
9140 @item ira-loop-reserved-regs
9141 IRA can be used to evaluate more accurate register pressure in loops
9142 for decisions to move loop invariants (see @option{-O3}). The number
9143 of available registers reserved for some other purposes is given
9144 by this parameter. The default value of the parameter is 2, which is
9145 the minimal number of registers needed by typical instructions.
9146 This value is the best found from numerous experiments.
9148 @item loop-invariant-max-bbs-in-loop
9149 Loop invariant motion can be very expensive, both in compilation time and
9150 in amount of needed compile-time memory, with very large loops. Loops
9151 with more basic blocks than this parameter won't have loop invariant
9152 motion optimization performed on them. The default value of the
9153 parameter is 1000 for -O1 and 10000 for -O2 and above.
9155 @item loop-max-datarefs-for-datadeps
9156 Building data dapendencies is expensive for very large loops. This
9157 parameter limits the number of data references in loops that are
9158 considered for data dependence analysis. These large loops will not
9159 be handled then by the optimizations using loop data dependencies.
9160 The default value is 1000.
9162 @item max-vartrack-size
9163 Sets a maximum number of hash table slots to use during variable
9164 tracking dataflow analysis of any function. If this limit is exceeded
9165 with variable tracking at assignments enabled, analysis for that
9166 function is retried without it, after removing all debug insns from
9167 the function. If the limit is exceeded even without debug insns, var
9168 tracking analysis is completely disabled for the function. Setting
9169 the parameter to zero makes it unlimited.
9171 @item max-vartrack-expr-depth
9172 Sets a maximum number of recursion levels when attempting to map
9173 variable names or debug temporaries to value expressions. This trades
9174 compilation time for more complete debug information. If this is set too
9175 low, value expressions that are available and could be represented in
9176 debug information may end up not being used; setting this higher may
9177 enable the compiler to find more complex debug expressions, but compile
9178 time and memory use may grow. The default is 12.
9180 @item min-nondebug-insn-uid
9181 Use uids starting at this parameter for nondebug insns. The range below
9182 the parameter is reserved exclusively for debug insns created by
9183 @option{-fvar-tracking-assignments}, but debug insns may get
9184 (non-overlapping) uids above it if the reserved range is exhausted.
9186 @item ipa-sra-ptr-growth-factor
9187 IPA-SRA will replace a pointer to an aggregate with one or more new
9188 parameters only when their cumulative size is less or equal to
9189 @option{ipa-sra-ptr-growth-factor} times the size of the original
9192 @item tm-max-aggregate-size
9193 When making copies of thread-local variables in a transaction, this
9194 parameter specifies the size in bytes after which variables will be
9195 saved with the logging functions as opposed to save/restore code
9196 sequence pairs. This option only applies when using
9199 @item graphite-max-nb-scop-params
9200 To avoid exponential effects in the Graphite loop transforms, the
9201 number of parameters in a Static Control Part (SCoP) is bounded. The
9202 default value is 10 parameters. A variable whose value is unknown at
9203 compilation time and defined outside a SCoP is a parameter of the SCoP.
9205 @item graphite-max-bbs-per-function
9206 To avoid exponential effects in the detection of SCoPs, the size of
9207 the functions analyzed by Graphite is bounded. The default value is
9210 @item loop-block-tile-size
9211 Loop blocking or strip mining transforms, enabled with
9212 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9213 loop in the loop nest by a given number of iterations. The strip
9214 length can be changed using the @option{loop-block-tile-size}
9215 parameter. The default value is 51 iterations.
9217 @item ipa-cp-value-list-size
9218 IPA-CP attempts to track all possible values and types passed to a function's
9219 parameter in order to propagate them and perform devirtualization.
9220 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9221 stores per one formal parameter of a function.
9223 @item lto-partitions
9224 Specify desired number of partitions produced during WHOPR compilation.
9225 The number of partitions should exceed the number of CPUs used for compilation.
9226 The default value is 32.
9228 @item lto-minpartition
9229 Size of minimal partition for WHOPR (in estimated instructions).
9230 This prevents expenses of splitting very small programs into too many
9233 @item cxx-max-namespaces-for-diagnostic-help
9234 The maximum number of namespaces to consult for suggestions when C++
9235 name lookup fails for an identifier. The default is 1000.
9237 @item sink-frequency-threshold
9238 The maximum relative execution frequency (in percents) of the target block
9239 relative to a statement's original block to allow statement sinking of a
9240 statement. Larger numbers result in more aggressive statement sinking.
9241 The default value is 75. A small positive adjustment is applied for
9242 statements with memory operands as those are even more profitable so sink.
9244 @item max-stores-to-sink
9245 The maximum number of conditional stores paires that can be sunk. Set to 0
9246 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9247 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9249 @item allow-load-data-races
9250 Allow optimizers to introduce new data races on loads.
9251 Set to 1 to allow, otherwise to 0. This option is enabled by default
9252 unless implicitly set by the @option{-fmemory-model=} option.
9254 @item allow-store-data-races
9255 Allow optimizers to introduce new data races on stores.
9256 Set to 1 to allow, otherwise to 0. This option is enabled by default
9257 unless implicitly set by the @option{-fmemory-model=} option.
9259 @item allow-packed-load-data-races
9260 Allow optimizers to introduce new data races on packed data loads.
9261 Set to 1 to allow, otherwise to 0. This option is enabled by default
9262 unless implicitly set by the @option{-fmemory-model=} option.
9264 @item allow-packed-store-data-races
9265 Allow optimizers to introduce new data races on packed data stores.
9266 Set to 1 to allow, otherwise to 0. This option is enabled by default
9267 unless implicitly set by the @option{-fmemory-model=} option.
9269 @item case-values-threshold
9270 The smallest number of different values for which it is best to use a
9271 jump-table instead of a tree of conditional branches. If the value is
9272 0, use the default for the machine. The default is 0.
9274 @item tree-reassoc-width
9275 Set the maximum number of instructions executed in parallel in
9276 reassociated tree. This parameter overrides target dependent
9277 heuristics used by default if has non zero value.
9282 @node Preprocessor Options
9283 @section Options Controlling the Preprocessor
9284 @cindex preprocessor options
9285 @cindex options, preprocessor
9287 These options control the C preprocessor, which is run on each C source
9288 file before actual compilation.
9290 If you use the @option{-E} option, nothing is done except preprocessing.
9291 Some of these options make sense only together with @option{-E} because
9292 they cause the preprocessor output to be unsuitable for actual
9296 @item -Wp,@var{option}
9298 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9299 and pass @var{option} directly through to the preprocessor. If
9300 @var{option} contains commas, it is split into multiple options at the
9301 commas. However, many options are modified, translated or interpreted
9302 by the compiler driver before being passed to the preprocessor, and
9303 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9304 interface is undocumented and subject to change, so whenever possible
9305 you should avoid using @option{-Wp} and let the driver handle the
9308 @item -Xpreprocessor @var{option}
9309 @opindex Xpreprocessor
9310 Pass @var{option} as an option to the preprocessor. You can use this to
9311 supply system-specific preprocessor options that GCC does not know how to
9314 If you want to pass an option that takes an argument, you must use
9315 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9318 @include cppopts.texi
9320 @node Assembler Options
9321 @section Passing Options to the Assembler
9323 @c prevent bad page break with this line
9324 You can pass options to the assembler.
9327 @item -Wa,@var{option}
9329 Pass @var{option} as an option to the assembler. If @var{option}
9330 contains commas, it is split into multiple options at the commas.
9332 @item -Xassembler @var{option}
9334 Pass @var{option} as an option to the assembler. You can use this to
9335 supply system-specific assembler options that GCC does not know how to
9338 If you want to pass an option that takes an argument, you must use
9339 @option{-Xassembler} twice, once for the option and once for the argument.
9344 @section Options for Linking
9345 @cindex link options
9346 @cindex options, linking
9348 These options come into play when the compiler links object files into
9349 an executable output file. They are meaningless if the compiler is
9350 not doing a link step.
9354 @item @var{object-file-name}
9355 A file name that does not end in a special recognized suffix is
9356 considered to name an object file or library. (Object files are
9357 distinguished from libraries by the linker according to the file
9358 contents.) If linking is done, these object files are used as input
9367 If any of these options is used, then the linker is not run, and
9368 object file names should not be used as arguments. @xref{Overall
9372 @item -l@var{library}
9373 @itemx -l @var{library}
9375 Search the library named @var{library} when linking. (The second
9376 alternative with the library as a separate argument is only for
9377 POSIX compliance and is not recommended.)
9379 It makes a difference where in the command you write this option; the
9380 linker searches and processes libraries and object files in the order they
9381 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9382 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9383 to functions in @samp{z}, those functions may not be loaded.
9385 The linker searches a standard list of directories for the library,
9386 which is actually a file named @file{lib@var{library}.a}. The linker
9387 then uses this file as if it had been specified precisely by name.
9389 The directories searched include several standard system directories
9390 plus any that you specify with @option{-L}.
9392 Normally the files found this way are library files---archive files
9393 whose members are object files. The linker handles an archive file by
9394 scanning through it for members which define symbols that have so far
9395 been referenced but not defined. But if the file that is found is an
9396 ordinary object file, it is linked in the usual fashion. The only
9397 difference between using an @option{-l} option and specifying a file name
9398 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9399 and searches several directories.
9403 You need this special case of the @option{-l} option in order to
9404 link an Objective-C or Objective-C++ program.
9407 @opindex nostartfiles
9408 Do not use the standard system startup files when linking.
9409 The standard system libraries are used normally, unless @option{-nostdlib}
9410 or @option{-nodefaultlibs} is used.
9412 @item -nodefaultlibs
9413 @opindex nodefaultlibs
9414 Do not use the standard system libraries when linking.
9415 Only the libraries you specify will be passed to the linker, options
9416 specifying linkage of the system libraries, such as @code{-static-libgcc}
9417 or @code{-shared-libgcc}, will be ignored.
9418 The standard startup files are used normally, unless @option{-nostartfiles}
9419 is used. The compiler may generate calls to @code{memcmp},
9420 @code{memset}, @code{memcpy} and @code{memmove}.
9421 These entries are usually resolved by entries in
9422 libc. These entry points should be supplied through some other
9423 mechanism when this option is specified.
9427 Do not use the standard system startup files or libraries when linking.
9428 No startup files and only the libraries you specify will be passed to
9429 the linker, options specifying linkage of the system libraries, such as
9430 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9431 The compiler may generate calls to @code{memcmp}, @code{memset},
9432 @code{memcpy} and @code{memmove}.
9433 These entries are usually resolved by entries in
9434 libc. These entry points should be supplied through some other
9435 mechanism when this option is specified.
9437 @cindex @option{-lgcc}, use with @option{-nostdlib}
9438 @cindex @option{-nostdlib} and unresolved references
9439 @cindex unresolved references and @option{-nostdlib}
9440 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9441 @cindex @option{-nodefaultlibs} and unresolved references
9442 @cindex unresolved references and @option{-nodefaultlibs}
9443 One of the standard libraries bypassed by @option{-nostdlib} and
9444 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9445 which GCC uses to overcome shortcomings of particular machines, or special
9446 needs for some languages.
9447 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9448 Collection (GCC) Internals},
9449 for more discussion of @file{libgcc.a}.)
9450 In most cases, you need @file{libgcc.a} even when you want to avoid
9451 other standard libraries. In other words, when you specify @option{-nostdlib}
9452 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9453 This ensures that you have no unresolved references to internal GCC
9454 library subroutines. (For example, @samp{__main}, used to ensure C++
9455 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9456 GNU Compiler Collection (GCC) Internals}.)
9460 Produce a position independent executable on targets that support it.
9461 For predictable results, you must also specify the same set of options
9462 that were used to generate code (@option{-fpie}, @option{-fPIE},
9463 or model suboptions) when you specify this option.
9467 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9468 that support it. This instructs the linker to add all symbols, not
9469 only used ones, to the dynamic symbol table. This option is needed
9470 for some uses of @code{dlopen} or to allow obtaining backtraces
9471 from within a program.
9475 Remove all symbol table and relocation information from the executable.
9479 On systems that support dynamic linking, this prevents linking with the shared
9480 libraries. On other systems, this option has no effect.
9484 Produce a shared object which can then be linked with other objects to
9485 form an executable. Not all systems support this option. For predictable
9486 results, you must also specify the same set of options that were used to
9487 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9488 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9489 needs to build supplementary stub code for constructors to work. On
9490 multi-libbed systems, @samp{gcc -shared} must select the correct support
9491 libraries to link against. Failing to supply the correct flags may lead
9492 to subtle defects. Supplying them in cases where they are not necessary
9495 @item -shared-libgcc
9496 @itemx -static-libgcc
9497 @opindex shared-libgcc
9498 @opindex static-libgcc
9499 On systems that provide @file{libgcc} as a shared library, these options
9500 force the use of either the shared or static version respectively.
9501 If no shared version of @file{libgcc} was built when the compiler was
9502 configured, these options have no effect.
9504 There are several situations in which an application should use the
9505 shared @file{libgcc} instead of the static version. The most common
9506 of these is when the application wishes to throw and catch exceptions
9507 across different shared libraries. In that case, each of the libraries
9508 as well as the application itself should use the shared @file{libgcc}.
9510 Therefore, the G++ and GCJ drivers automatically add
9511 @option{-shared-libgcc} whenever you build a shared library or a main
9512 executable, because C++ and Java programs typically use exceptions, so
9513 this is the right thing to do.
9515 If, instead, you use the GCC driver to create shared libraries, you may
9516 find that they will not always be linked with the shared @file{libgcc}.
9517 If GCC finds, at its configuration time, that you have a non-GNU linker
9518 or a GNU linker that does not support option @option{--eh-frame-hdr},
9519 it will link the shared version of @file{libgcc} into shared libraries
9520 by default. Otherwise, it will take advantage of the linker and optimize
9521 away the linking with the shared version of @file{libgcc}, linking with
9522 the static version of libgcc by default. This allows exceptions to
9523 propagate through such shared libraries, without incurring relocation
9524 costs at library load time.
9526 However, if a library or main executable is supposed to throw or catch
9527 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9528 for the languages used in the program, or using the option
9529 @option{-shared-libgcc}, such that it is linked with the shared
9532 @item -static-libstdc++
9533 When the @command{g++} program is used to link a C++ program, it will
9534 normally automatically link against @option{libstdc++}. If
9535 @file{libstdc++} is available as a shared library, and the
9536 @option{-static} option is not used, then this will link against the
9537 shared version of @file{libstdc++}. That is normally fine. However, it
9538 is sometimes useful to freeze the version of @file{libstdc++} used by
9539 the program without going all the way to a fully static link. The
9540 @option{-static-libstdc++} option directs the @command{g++} driver to
9541 link @file{libstdc++} statically, without necessarily linking other
9542 libraries statically.
9546 Bind references to global symbols when building a shared object. Warn
9547 about any unresolved references (unless overridden by the link editor
9548 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9551 @item -T @var{script}
9553 @cindex linker script
9554 Use @var{script} as the linker script. This option is supported by most
9555 systems using the GNU linker. On some targets, such as bare-board
9556 targets without an operating system, the @option{-T} option may be required
9557 when linking to avoid references to undefined symbols.
9559 @item -Xlinker @var{option}
9561 Pass @var{option} as an option to the linker. You can use this to
9562 supply system-specific linker options that GCC does not recognize.
9564 If you want to pass an option that takes a separate argument, you must use
9565 @option{-Xlinker} twice, once for the option and once for the argument.
9566 For example, to pass @option{-assert definitions}, you must write
9567 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9568 @option{-Xlinker "-assert definitions"}, because this passes the entire
9569 string as a single argument, which is not what the linker expects.
9571 When using the GNU linker, it is usually more convenient to pass
9572 arguments to linker options using the @option{@var{option}=@var{value}}
9573 syntax than as separate arguments. For example, you can specify
9574 @samp{-Xlinker -Map=output.map} rather than
9575 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9576 this syntax for command-line options.
9578 @item -Wl,@var{option}
9580 Pass @var{option} as an option to the linker. If @var{option} contains
9581 commas, it is split into multiple options at the commas. You can use this
9582 syntax to pass an argument to the option.
9583 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9584 linker. When using the GNU linker, you can also get the same effect with
9585 @samp{-Wl,-Map=output.map}.
9587 @item -u @var{symbol}
9589 Pretend the symbol @var{symbol} is undefined, to force linking of
9590 library modules to define it. You can use @option{-u} multiple times with
9591 different symbols to force loading of additional library modules.
9594 @node Directory Options
9595 @section Options for Directory Search
9596 @cindex directory options
9597 @cindex options, directory search
9600 These options specify directories to search for header files, for
9601 libraries and for parts of the compiler:
9606 Add the directory @var{dir} to the head of the list of directories to be
9607 searched for header files. This can be used to override a system header
9608 file, substituting your own version, since these directories are
9609 searched before the system header file directories. However, you should
9610 not use this option to add directories that contain vendor-supplied
9611 system header files (use @option{-isystem} for that). If you use more than
9612 one @option{-I} option, the directories are scanned in left-to-right
9613 order; the standard system directories come after.
9615 If a standard system include directory, or a directory specified with
9616 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9617 option will be ignored. The directory will still be searched but as a
9618 system directory at its normal position in the system include chain.
9619 This is to ensure that GCC's procedure to fix buggy system headers and
9620 the ordering for the include_next directive are not inadvertently changed.
9621 If you really need to change the search order for system directories,
9622 use the @option{-nostdinc} and/or @option{-isystem} options.
9624 @item -iplugindir=@var{dir}
9625 Set the directory to search for plugins that are passed
9626 by @option{-fplugin=@var{name}} instead of
9627 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9628 to be used by the user, but only passed by the driver.
9630 @item -iquote@var{dir}
9632 Add the directory @var{dir} to the head of the list of directories to
9633 be searched for header files only for the case of @samp{#include
9634 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9635 otherwise just like @option{-I}.
9639 Add directory @var{dir} to the list of directories to be searched
9642 @item -B@var{prefix}
9644 This option specifies where to find the executables, libraries,
9645 include files, and data files of the compiler itself.
9647 The compiler driver program runs one or more of the subprograms
9648 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9649 @var{prefix} as a prefix for each program it tries to run, both with and
9650 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9652 For each subprogram to be run, the compiler driver first tries the
9653 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9654 was not specified, the driver tries two standard prefixes,
9655 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9656 those results in a file name that is found, the unmodified program
9657 name is searched for using the directories specified in your
9658 @env{PATH} environment variable.
9660 The compiler will check to see if the path provided by the @option{-B}
9661 refers to a directory, and if necessary it will add a directory
9662 separator character at the end of the path.
9664 @option{-B} prefixes that effectively specify directory names also apply
9665 to libraries in the linker, because the compiler translates these
9666 options into @option{-L} options for the linker. They also apply to
9667 includes files in the preprocessor, because the compiler translates these
9668 options into @option{-isystem} options for the preprocessor. In this case,
9669 the compiler appends @samp{include} to the prefix.
9671 The runtime support file @file{libgcc.a} can also be searched for using
9672 the @option{-B} prefix, if needed. If it is not found there, the two
9673 standard prefixes above are tried, and that is all. The file is left
9674 out of the link if it is not found by those means.
9676 Another way to specify a prefix much like the @option{-B} prefix is to use
9677 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9680 As a special kludge, if the path provided by @option{-B} is
9681 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9682 9, then it will be replaced by @file{[dir/]include}. This is to help
9683 with boot-strapping the compiler.
9685 @item -specs=@var{file}
9687 Process @var{file} after the compiler reads in the standard @file{specs}
9688 file, in order to override the defaults which the @file{gcc} driver
9689 program uses when determining what switches to pass to @file{cc1},
9690 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9691 @option{-specs=@var{file}} can be specified on the command line, and they
9692 are processed in order, from left to right.
9694 @item --sysroot=@var{dir}
9696 Use @var{dir} as the logical root directory for headers and libraries.
9697 For example, if the compiler would normally search for headers in
9698 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9699 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9701 If you use both this option and the @option{-isysroot} option, then
9702 the @option{--sysroot} option will apply to libraries, but the
9703 @option{-isysroot} option will apply to header files.
9705 The GNU linker (beginning with version 2.16) has the necessary support
9706 for this option. If your linker does not support this option, the
9707 header file aspect of @option{--sysroot} will still work, but the
9708 library aspect will not.
9712 This option has been deprecated. Please use @option{-iquote} instead for
9713 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9714 Any directories you specify with @option{-I} options before the @option{-I-}
9715 option are searched only for the case of @samp{#include "@var{file}"};
9716 they are not searched for @samp{#include <@var{file}>}.
9718 If additional directories are specified with @option{-I} options after
9719 the @option{-I-}, these directories are searched for all @samp{#include}
9720 directives. (Ordinarily @emph{all} @option{-I} directories are used
9723 In addition, the @option{-I-} option inhibits the use of the current
9724 directory (where the current input file came from) as the first search
9725 directory for @samp{#include "@var{file}"}. There is no way to
9726 override this effect of @option{-I-}. With @option{-I.} you can specify
9727 searching the directory that was current when the compiler was
9728 invoked. That is not exactly the same as what the preprocessor does
9729 by default, but it is often satisfactory.
9731 @option{-I-} does not inhibit the use of the standard system directories
9732 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9739 @section Specifying subprocesses and the switches to pass to them
9742 @command{gcc} is a driver program. It performs its job by invoking a
9743 sequence of other programs to do the work of compiling, assembling and
9744 linking. GCC interprets its command-line parameters and uses these to
9745 deduce which programs it should invoke, and which command-line options
9746 it ought to place on their command lines. This behavior is controlled
9747 by @dfn{spec strings}. In most cases there is one spec string for each
9748 program that GCC can invoke, but a few programs have multiple spec
9749 strings to control their behavior. The spec strings built into GCC can
9750 be overridden by using the @option{-specs=} command-line switch to specify
9753 @dfn{Spec files} are plaintext files that are used to construct spec
9754 strings. They consist of a sequence of directives separated by blank
9755 lines. The type of directive is determined by the first non-whitespace
9756 character on the line, which can be one of the following:
9759 @item %@var{command}
9760 Issues a @var{command} to the spec file processor. The commands that can
9764 @item %include <@var{file}>
9765 @cindex @code{%include}
9766 Search for @var{file} and insert its text at the current point in the
9769 @item %include_noerr <@var{file}>
9770 @cindex @code{%include_noerr}
9771 Just like @samp{%include}, but do not generate an error message if the include
9772 file cannot be found.
9774 @item %rename @var{old_name} @var{new_name}
9775 @cindex @code{%rename}
9776 Rename the spec string @var{old_name} to @var{new_name}.
9780 @item *[@var{spec_name}]:
9781 This tells the compiler to create, override or delete the named spec
9782 string. All lines after this directive up to the next directive or
9783 blank line are considered to be the text for the spec string. If this
9784 results in an empty string then the spec will be deleted. (Or, if the
9785 spec did not exist, then nothing will happen.) Otherwise, if the spec
9786 does not currently exist a new spec will be created. If the spec does
9787 exist then its contents will be overridden by the text of this
9788 directive, unless the first character of that text is the @samp{+}
9789 character, in which case the text will be appended to the spec.
9791 @item [@var{suffix}]:
9792 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9793 and up to the next directive or blank line are considered to make up the
9794 spec string for the indicated suffix. When the compiler encounters an
9795 input file with the named suffix, it will processes the spec string in
9796 order to work out how to compile that file. For example:
9803 This says that any input file whose name ends in @samp{.ZZ} should be
9804 passed to the program @samp{z-compile}, which should be invoked with the
9805 command-line switch @option{-input} and with the result of performing the
9806 @samp{%i} substitution. (See below.)
9808 As an alternative to providing a spec string, the text that follows a
9809 suffix directive can be one of the following:
9812 @item @@@var{language}
9813 This says that the suffix is an alias for a known @var{language}. This is
9814 similar to using the @option{-x} command-line switch to GCC to specify a
9815 language explicitly. For example:
9822 Says that .ZZ files are, in fact, C++ source files.
9825 This causes an error messages saying:
9828 @var{name} compiler not installed on this system.
9832 GCC already has an extensive list of suffixes built into it.
9833 This directive will add an entry to the end of the list of suffixes, but
9834 since the list is searched from the end backwards, it is effectively
9835 possible to override earlier entries using this technique.
9839 GCC has the following spec strings built into it. Spec files can
9840 override these strings or create their own. Note that individual
9841 targets can also add their own spec strings to this list.
9844 asm Options to pass to the assembler
9845 asm_final Options to pass to the assembler post-processor
9846 cpp Options to pass to the C preprocessor
9847 cc1 Options to pass to the C compiler
9848 cc1plus Options to pass to the C++ compiler
9849 endfile Object files to include at the end of the link
9850 link Options to pass to the linker
9851 lib Libraries to include on the command line to the linker
9852 libgcc Decides which GCC support library to pass to the linker
9853 linker Sets the name of the linker
9854 predefines Defines to be passed to the C preprocessor
9855 signed_char Defines to pass to CPP to say whether @code{char} is signed
9857 startfile Object files to include at the start of the link
9860 Here is a small example of a spec file:
9866 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9869 This example renames the spec called @samp{lib} to @samp{old_lib} and
9870 then overrides the previous definition of @samp{lib} with a new one.
9871 The new definition adds in some extra command-line options before
9872 including the text of the old definition.
9874 @dfn{Spec strings} are a list of command-line options to be passed to their
9875 corresponding program. In addition, the spec strings can contain
9876 @samp{%}-prefixed sequences to substitute variable text or to
9877 conditionally insert text into the command line. Using these constructs
9878 it is possible to generate quite complex command lines.
9880 Here is a table of all defined @samp{%}-sequences for spec
9881 strings. Note that spaces are not generated automatically around the
9882 results of expanding these sequences. Therefore you can concatenate them
9883 together or combine them with constant text in a single argument.
9887 Substitute one @samp{%} into the program name or argument.
9890 Substitute the name of the input file being processed.
9893 Substitute the basename of the input file being processed.
9894 This is the substring up to (and not including) the last period
9895 and not including the directory.
9898 This is the same as @samp{%b}, but include the file suffix (text after
9902 Marks the argument containing or following the @samp{%d} as a
9903 temporary file name, so that that file will be deleted if GCC exits
9904 successfully. Unlike @samp{%g}, this contributes no text to the
9907 @item %g@var{suffix}
9908 Substitute a file name that has suffix @var{suffix} and is chosen
9909 once per compilation, and mark the argument in the same way as
9910 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9911 name is now chosen in a way that is hard to predict even when previously
9912 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9913 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9914 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9915 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9916 was simply substituted with a file name chosen once per compilation,
9917 without regard to any appended suffix (which was therefore treated
9918 just like ordinary text), making such attacks more likely to succeed.
9920 @item %u@var{suffix}
9921 Like @samp{%g}, but generates a new temporary file name even if
9922 @samp{%u@var{suffix}} was already seen.
9924 @item %U@var{suffix}
9925 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9926 new one if there is no such last file name. In the absence of any
9927 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9928 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9929 would involve the generation of two distinct file names, one
9930 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9931 simply substituted with a file name chosen for the previous @samp{%u},
9932 without regard to any appended suffix.
9934 @item %j@var{suffix}
9935 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9936 writable, and if save-temps is off; otherwise, substitute the name
9937 of a temporary file, just like @samp{%u}. This temporary file is not
9938 meant for communication between processes, but rather as a junk
9941 @item %|@var{suffix}
9942 @itemx %m@var{suffix}
9943 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9944 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9945 all. These are the two most common ways to instruct a program that it
9946 should read from standard input or write to standard output. If you
9947 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9948 construct: see for example @file{f/lang-specs.h}.
9950 @item %.@var{SUFFIX}
9951 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9952 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9953 terminated by the next space or %.
9956 Marks the argument containing or following the @samp{%w} as the
9957 designated output file of this compilation. This puts the argument
9958 into the sequence of arguments that @samp{%o} will substitute later.
9961 Substitutes the names of all the output files, with spaces
9962 automatically placed around them. You should write spaces
9963 around the @samp{%o} as well or the results are undefined.
9964 @samp{%o} is for use in the specs for running the linker.
9965 Input files whose names have no recognized suffix are not compiled
9966 at all, but they are included among the output files, so they will
9970 Substitutes the suffix for object files. Note that this is
9971 handled specially when it immediately follows @samp{%g, %u, or %U},
9972 because of the need for those to form complete file names. The
9973 handling is such that @samp{%O} is treated exactly as if it had already
9974 been substituted, except that @samp{%g, %u, and %U} do not currently
9975 support additional @var{suffix} characters following @samp{%O} as they would
9976 following, for example, @samp{.o}.
9979 Substitutes the standard macro predefinitions for the
9980 current target machine. Use this when running @code{cpp}.
9983 Like @samp{%p}, but puts @samp{__} before and after the name of each
9984 predefined macro, except for macros that start with @samp{__} or with
9985 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9989 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9990 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9991 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9992 and @option{-imultilib} as necessary.
9995 Current argument is the name of a library or startup file of some sort.
9996 Search for that file in a standard list of directories and substitute
9997 the full name found. The current working directory is included in the
9998 list of directories scanned.
10001 Current argument is the name of a linker script. Search for that file
10002 in the current list of directories to scan for libraries. If the file
10003 is located insert a @option{--script} option into the command line
10004 followed by the full path name found. If the file is not found then
10005 generate an error message. Note: the current working directory is not
10009 Print @var{str} as an error message. @var{str} is terminated by a newline.
10010 Use this when inconsistent options are detected.
10012 @item %(@var{name})
10013 Substitute the contents of spec string @var{name} at this point.
10015 @item %x@{@var{option}@}
10016 Accumulate an option for @samp{%X}.
10019 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
10023 Output the accumulated assembler options specified by @option{-Wa}.
10026 Output the accumulated preprocessor options specified by @option{-Wp}.
10029 Process the @code{asm} spec. This is used to compute the
10030 switches to be passed to the assembler.
10033 Process the @code{asm_final} spec. This is a spec string for
10034 passing switches to an assembler post-processor, if such a program is
10038 Process the @code{link} spec. This is the spec for computing the
10039 command line passed to the linker. Typically it will make use of the
10040 @samp{%L %G %S %D and %E} sequences.
10043 Dump out a @option{-L} option for each directory that GCC believes might
10044 contain startup files. If the target supports multilibs then the
10045 current multilib directory will be prepended to each of these paths.
10048 Process the @code{lib} spec. This is a spec string for deciding which
10049 libraries should be included on the command line to the linker.
10052 Process the @code{libgcc} spec. This is a spec string for deciding
10053 which GCC support library should be included on the command line to the linker.
10056 Process the @code{startfile} spec. This is a spec for deciding which
10057 object files should be the first ones passed to the linker. Typically
10058 this might be a file named @file{crt0.o}.
10061 Process the @code{endfile} spec. This is a spec string that specifies
10062 the last object files that will be passed to the linker.
10065 Process the @code{cpp} spec. This is used to construct the arguments
10066 to be passed to the C preprocessor.
10069 Process the @code{cc1} spec. This is used to construct the options to be
10070 passed to the actual C compiler (@samp{cc1}).
10073 Process the @code{cc1plus} spec. This is used to construct the options to be
10074 passed to the actual C++ compiler (@samp{cc1plus}).
10077 Substitute the variable part of a matched option. See below.
10078 Note that each comma in the substituted string is replaced by
10082 Remove all occurrences of @code{-S} from the command line. Note---this
10083 command is position dependent. @samp{%} commands in the spec string
10084 before this one will see @code{-S}, @samp{%} commands in the spec string
10085 after this one will not.
10087 @item %:@var{function}(@var{args})
10088 Call the named function @var{function}, passing it @var{args}.
10089 @var{args} is first processed as a nested spec string, then split
10090 into an argument vector in the usual fashion. The function returns
10091 a string which is processed as if it had appeared literally as part
10092 of the current spec.
10094 The following built-in spec functions are provided:
10097 @item @code{getenv}
10098 The @code{getenv} spec function takes two arguments: an environment
10099 variable name and a string. If the environment variable is not
10100 defined, a fatal error is issued. Otherwise, the return value is the
10101 value of the environment variable concatenated with the string. For
10102 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
10105 %:getenv(TOPDIR /include)
10108 expands to @file{/path/to/top/include}.
10110 @item @code{if-exists}
10111 The @code{if-exists} spec function takes one argument, an absolute
10112 pathname to a file. If the file exists, @code{if-exists} returns the
10113 pathname. Here is a small example of its usage:
10117 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
10120 @item @code{if-exists-else}
10121 The @code{if-exists-else} spec function is similar to the @code{if-exists}
10122 spec function, except that it takes two arguments. The first argument is
10123 an absolute pathname to a file. If the file exists, @code{if-exists-else}
10124 returns the pathname. If it does not exist, it returns the second argument.
10125 This way, @code{if-exists-else} can be used to select one file or another,
10126 based on the existence of the first. Here is a small example of its usage:
10130 crt0%O%s %:if-exists(crti%O%s) \
10131 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
10134 @item @code{replace-outfile}
10135 The @code{replace-outfile} spec function takes two arguments. It looks for the
10136 first argument in the outfiles array and replaces it with the second argument. Here
10137 is a small example of its usage:
10140 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10143 @item @code{remove-outfile}
10144 The @code{remove-outfile} spec function takes one argument. It looks for the
10145 first argument in the outfiles array and removes it. Here is a small example
10149 %:remove-outfile(-lm)
10152 @item @code{pass-through-libs}
10153 The @code{pass-through-libs} spec function takes any number of arguments. It
10154 finds any @option{-l} options and any non-options ending in ".a" (which it
10155 assumes are the names of linker input library archive files) and returns a
10156 result containing all the found arguments each prepended by
10157 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10158 intended to be passed to the LTO linker plugin.
10161 %:pass-through-libs(%G %L %G)
10164 @item @code{print-asm-header}
10165 The @code{print-asm-header} function takes no arguments and simply
10166 prints a banner like:
10172 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10175 It is used to separate compiler options from assembler options
10176 in the @option{--target-help} output.
10179 @item %@{@code{S}@}
10180 Substitutes the @code{-S} switch, if that switch was given to GCC@.
10181 If that switch was not specified, this substitutes nothing. Note that
10182 the leading dash is omitted when specifying this option, and it is
10183 automatically inserted if the substitution is performed. Thus the spec
10184 string @samp{%@{foo@}} would match the command-line option @option{-foo}
10185 and would output the command-line option @option{-foo}.
10187 @item %W@{@code{S}@}
10188 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10189 deleted on failure.
10191 @item %@{@code{S}*@}
10192 Substitutes all the switches specified to GCC whose names start
10193 with @code{-S}, but which also take an argument. This is used for
10194 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10195 GCC considers @option{-o foo} as being
10196 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10197 text, including the space. Thus two arguments would be generated.
10199 @item %@{@code{S}*&@code{T}*@}
10200 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10201 (the order of @code{S} and @code{T} in the spec is not significant).
10202 There can be any number of ampersand-separated variables; for each the
10203 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10205 @item %@{@code{S}:@code{X}@}
10206 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10208 @item %@{!@code{S}:@code{X}@}
10209 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10211 @item %@{@code{S}*:@code{X}@}
10212 Substitutes @code{X} if one or more switches whose names start with
10213 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10214 once, no matter how many such switches appeared. However, if @code{%*}
10215 appears somewhere in @code{X}, then @code{X} will be substituted once
10216 for each matching switch, with the @code{%*} replaced by the part of
10217 that switch that matched the @code{*}.
10219 @item %@{.@code{S}:@code{X}@}
10220 Substitutes @code{X}, if processing a file with suffix @code{S}.
10222 @item %@{!.@code{S}:@code{X}@}
10223 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10225 @item %@{,@code{S}:@code{X}@}
10226 Substitutes @code{X}, if processing a file for language @code{S}.
10228 @item %@{!,@code{S}:@code{X}@}
10229 Substitutes @code{X}, if not processing a file for language @code{S}.
10231 @item %@{@code{S}|@code{P}:@code{X}@}
10232 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10233 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10234 @code{*} sequences as well, although they have a stronger binding than
10235 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10236 alternatives must be starred, and only the first matching alternative
10239 For example, a spec string like this:
10242 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10245 will output the following command-line options from the following input
10246 command-line options:
10251 -d fred.c -foo -baz -boggle
10252 -d jim.d -bar -baz -boggle
10255 @item %@{S:X; T:Y; :D@}
10257 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10258 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10259 be as many clauses as you need. This may be combined with @code{.},
10260 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10265 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10266 construct may contain other nested @samp{%} constructs or spaces, or
10267 even newlines. They are processed as usual, as described above.
10268 Trailing white space in @code{X} is ignored. White space may also
10269 appear anywhere on the left side of the colon in these constructs,
10270 except between @code{.} or @code{*} and the corresponding word.
10272 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10273 handled specifically in these constructs. If another value of
10274 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10275 @option{-W} switch is found later in the command line, the earlier
10276 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10277 just one letter, which passes all matching options.
10279 The character @samp{|} at the beginning of the predicate text is used to
10280 indicate that a command should be piped to the following command, but
10281 only if @option{-pipe} is specified.
10283 It is built into GCC which switches take arguments and which do not.
10284 (You might think it would be useful to generalize this to allow each
10285 compiler's spec to say which switches take arguments. But this cannot
10286 be done in a consistent fashion. GCC cannot even decide which input
10287 files have been specified without knowing which switches take arguments,
10288 and it must know which input files to compile in order to tell which
10291 GCC also knows implicitly that arguments starting in @option{-l} are to be
10292 treated as compiler output files, and passed to the linker in their
10293 proper position among the other output files.
10295 @c man begin OPTIONS
10297 @node Target Options
10298 @section Specifying Target Machine and Compiler Version
10299 @cindex target options
10300 @cindex cross compiling
10301 @cindex specifying machine version
10302 @cindex specifying compiler version and target machine
10303 @cindex compiler version, specifying
10304 @cindex target machine, specifying
10306 The usual way to run GCC is to run the executable called @command{gcc}, or
10307 @command{@var{machine}-gcc} when cross-compiling, or
10308 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10309 one that was installed last.
10311 @node Submodel Options
10312 @section Hardware Models and Configurations
10313 @cindex submodel options
10314 @cindex specifying hardware config
10315 @cindex hardware models and configurations, specifying
10316 @cindex machine dependent options
10318 Each target machine types can have its own
10319 special options, starting with @samp{-m}, to choose among various
10320 hardware models or configurations---for example, 68010 vs 68020,
10321 floating coprocessor or none. A single installed version of the
10322 compiler can compile for any model or configuration, according to the
10325 Some configurations of the compiler also support additional special
10326 options, usually for compatibility with other compilers on the same
10329 @c This list is ordered alphanumerically by subsection name.
10330 @c It should be the same order and spelling as these options are listed
10331 @c in Machine Dependent Options
10334 * Adapteva Epiphany Options::
10337 * Blackfin Options::
10342 * DEC Alpha Options::
10343 * DEC Alpha/VMS Options::
10346 * GNU/Linux Options::
10349 * i386 and x86-64 Options::
10350 * i386 and x86-64 Windows Options::
10352 * IA-64/VMS Options::
10359 * MicroBlaze Options::
10362 * MN10300 Options::
10364 * picoChip Options::
10365 * PowerPC Options::
10367 * RS/6000 and PowerPC Options::
10369 * S/390 and zSeries Options::
10372 * Solaris 2 Options::
10375 * System V Options::
10376 * TILE-Gx Options::
10377 * TILEPro Options::
10380 * VxWorks Options::
10382 * Xstormy16 Options::
10384 * zSeries Options::
10387 @node Adapteva Epiphany Options
10388 @subsection Adapteva Epiphany Options
10390 These @samp{-m} options are defined for Adapteva Epiphany:
10393 @item -mhalf-reg-file
10394 @opindex mhalf-reg-file
10395 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
10396 That allows code to run on hardware variants that lack these registers.
10398 @item -mprefer-short-insn-regs
10399 @opindex mprefer-short-insn-regs
10400 Preferrentially allocate registers that allow short instruction generation.
10401 This can result in increasesd instruction count, so if this reduces or
10402 increases code size might vary from case to case.
10404 @item -mbranch-cost=@var{num}
10405 @opindex mbranch-cost
10406 Set the cost of branches to roughly @var{num} ``simple'' instructions.
10407 This cost is only a heuristic and is not guaranteed to produce
10408 consistent results across releases.
10412 Enable the generation of conditional moves.
10414 @item -mnops=@var{num}
10416 Emit @var{num} nops before every other generated instruction.
10418 @item -mno-soft-cmpsf
10419 @opindex mno-soft-cmpsf
10420 For single-precision floating-point comparisons, emit an fsub instruction
10421 and test the flags. This is faster than a software comparison, but can
10422 get incorrect results in the presence of NaNs, or when two different small
10423 numbers are compared such that their difference is calculated as zero.
10424 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
10425 software comparisons.
10427 @item -mstack-offset=@var{num}
10428 @opindex mstack-offset
10429 Set the offset between the top of the stack and the stack pointer.
10430 E.g., a value of 8 means that the eight bytes in the range sp+0@dots{}sp+7
10431 can be used by leaf functions without stack allocation.
10432 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
10433 Note also that this option changes the ABI, compiling a program with a
10434 different stack offset than the libraries have been compiled with
10435 will generally not work.
10436 This option can be useful if you want to evaluate if a different stack
10437 offset would give you better code, but to actually use a different stack
10438 offset to build working programs, it is recommended to configure the
10439 toolchain with the appropriate @samp{--with-stack-offset=@var{num}} option.
10441 @item -mno-round-nearest
10442 @opindex mno-round-nearest
10443 Make the scheduler assume that the rounding mode has been set to
10444 truncating. The default is @option{-mround-nearest}.
10447 @opindex mlong-calls
10448 If not otherwise specified by an attribute, assume all calls might be beyond
10449 the offset range of the b / bl instructions, and therefore load the
10450 function address into a register before performing a (otherwise direct) call.
10451 This is the default.
10453 @item -mshort-calls
10454 @opindex short-calls
10455 If not otherwise specified by an attribute, assume all direct calls are
10456 in the range of the b / bl instructions, so use these instructions
10457 for direct calls. The default is @option{-mlong-calls}.
10461 Assume addresses can be loaded as 16-bit unsigned values. This does not
10462 apply to function addresses for which @option{-mlong-calls} semantics
10465 @item -mfp-mode=@var{mode}
10467 Set the prevailing mode of the floating-point unit.
10468 This determines the floating-point mode that is provided and expected
10469 at function call and return time. Making this mode match the mode you
10470 predominantly need at function start can make your programs smaller and
10471 faster by avoiding unnecessary mode switches.
10473 @var{mode} can be set to one the following values:
10477 Any mode at function entry is valid, and retained or restored when
10478 the function returns, and when it calls other functions.
10479 This mode is useful for compiling libraries or other compilation units
10480 you might want to incorporate into different programs with different
10481 prevailing FPU modes, and the convenience of being able to use a single
10482 object file outweighs the size and speed overhead for any extra
10483 mode switching that might be needed, compared with what would be needed
10484 with a more specific choice of prevailing FPU mode.
10487 This is the mode used for floating-point calculations with
10488 truncating (i.e.@: round towards zero) rounding mode. That includes
10489 conversion from floating point to integer.
10491 @item round-nearest
10492 This is the mode used for floating-point calculations with
10493 round-to-nearest-or-even rounding mode.
10496 This is the mode used to perform integer calculations in the FPU, e.g.@:
10497 integer multiply, or integer multiply-and-accumulate.
10500 The default is @option{-mfp-mode=caller}
10502 @item -mnosplit-lohi
10503 @opindex mnosplit-lohi
10505 @opindex mno-postinc
10506 @item -mno-postmodify
10507 @opindex mno-postmodify
10508 Code generation tweaks that disable, respectively, splitting of 32-bit
10509 loads, generation of post-increment addresses, and generation of
10510 post-modify addresses. The defaults are @option{msplit-lohi},
10511 @option{-mpost-inc}, and @option{-mpost-modify}.
10513 @item -mnovect-double
10514 @opindex mno-vect-double
10515 Change the preferred SIMD mode to SImode. The default is
10516 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
10518 @item -max-vect-align=@var{num}
10519 @opindex max-vect-align
10520 The maximum alignment for SIMD vector mode types.
10521 @var{num} may be 4 or 8. The default is 8.
10522 Note that this is an ABI change, even though many library function
10523 interfaces will be unaffected, if they don't use SIMD vector modes
10524 in places where they affect size and/or alignment of relevant types.
10526 @item -msplit-vecmove-early
10527 @opindex msplit-vecmove-early
10528 Split vector moves into single word moves before reload. In theory this
10529 could give better register allocation, but so far the reverse seems to be
10530 generally the case.
10532 @item -m1reg-@var{reg}
10534 Specify a register to hold the constant @minus{}1, which makes loading small negative
10535 constants and certain bitmasks faster.
10536 Allowable values for reg are r43 and r63, which specify to use that register
10537 as a fixed register, and none, which means that no register is used for this
10538 purpose. The default is @option{-m1reg-none}.
10543 @subsection ARM Options
10544 @cindex ARM options
10546 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10550 @item -mabi=@var{name}
10552 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10553 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10556 @opindex mapcs-frame
10557 Generate a stack frame that is compliant with the ARM Procedure Call
10558 Standard for all functions, even if this is not strictly necessary for
10559 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10560 with this option will cause the stack frames not to be generated for
10561 leaf functions. The default is @option{-mno-apcs-frame}.
10565 This is a synonym for @option{-mapcs-frame}.
10568 @c not currently implemented
10569 @item -mapcs-stack-check
10570 @opindex mapcs-stack-check
10571 Generate code to check the amount of stack space available upon entry to
10572 every function (that actually uses some stack space). If there is
10573 insufficient space available then either the function
10574 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10575 called, depending upon the amount of stack space required. The runtime
10576 system is required to provide these functions. The default is
10577 @option{-mno-apcs-stack-check}, since this produces smaller code.
10579 @c not currently implemented
10581 @opindex mapcs-float
10582 Pass floating-point arguments using the floating-point registers. This is
10583 one of the variants of the APCS@. This option is recommended if the
10584 target hardware has a floating-point unit or if a lot of floating-point
10585 arithmetic is going to be performed by the code. The default is
10586 @option{-mno-apcs-float}, since integer only code is slightly increased in
10587 size if @option{-mapcs-float} is used.
10589 @c not currently implemented
10590 @item -mapcs-reentrant
10591 @opindex mapcs-reentrant
10592 Generate reentrant, position independent code. The default is
10593 @option{-mno-apcs-reentrant}.
10596 @item -mthumb-interwork
10597 @opindex mthumb-interwork
10598 Generate code that supports calling between the ARM and Thumb
10599 instruction sets. Without this option, on pre-v5 architectures, the
10600 two instruction sets cannot be reliably used inside one program. The
10601 default is @option{-mno-thumb-interwork}, since slightly larger code
10602 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10603 configurations this option is meaningless.
10605 @item -mno-sched-prolog
10606 @opindex mno-sched-prolog
10607 Prevent the reordering of instructions in the function prologue, or the
10608 merging of those instruction with the instructions in the function's
10609 body. This means that all functions will start with a recognizable set
10610 of instructions (or in fact one of a choice from a small set of
10611 different function prologues), and this information can be used to
10612 locate the start if functions inside an executable piece of code. The
10613 default is @option{-msched-prolog}.
10615 @item -mfloat-abi=@var{name}
10616 @opindex mfloat-abi
10617 Specifies which floating-point ABI to use. Permissible values
10618 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10620 Specifying @samp{soft} causes GCC to generate output containing
10621 library calls for floating-point operations.
10622 @samp{softfp} allows the generation of code using hardware floating-point
10623 instructions, but still uses the soft-float calling conventions.
10624 @samp{hard} allows generation of floating-point instructions
10625 and uses FPU-specific calling conventions.
10627 The default depends on the specific target configuration. Note that
10628 the hard-float and soft-float ABIs are not link-compatible; you must
10629 compile your entire program with the same ABI, and link with a
10630 compatible set of libraries.
10632 @item -mlittle-endian
10633 @opindex mlittle-endian
10634 Generate code for a processor running in little-endian mode. This is
10635 the default for all standard configurations.
10638 @opindex mbig-endian
10639 Generate code for a processor running in big-endian mode; the default is
10640 to compile code for a little-endian processor.
10642 @item -mwords-little-endian
10643 @opindex mwords-little-endian
10644 This option only applies when generating code for big-endian processors.
10645 Generate code for a little-endian word order but a big-endian byte
10646 order. That is, a byte order of the form @samp{32107654}. Note: this
10647 option should only be used if you require compatibility with code for
10648 big-endian ARM processors generated by versions of the compiler prior to
10649 2.8. This option is now deprecated.
10651 @item -march=@var{name}
10653 This specifies the name of the target ARM architecture. GCC uses this
10654 name to determine what kind of instructions it can emit when generating
10655 assembly code. This option can be used in conjunction with or instead
10656 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10657 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10658 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10659 @samp{armv6}, @samp{armv6j},
10660 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10661 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
10662 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10664 @option{-march=native} causes the compiler to auto-detect the architecture
10665 of the build computer. At present, this feature is only supported on
10666 Linux, and not all architectures are recognized. If the auto-detect is
10667 unsuccessful the option has no effect.
10669 @item -mtune=@var{name}
10671 This option specifies the name of the target ARM processor for
10672 which GCC should tune the performance of the code.
10673 For some ARM implementations better performance can be obtained by using
10675 Permissible names are: @samp{arm2}, @samp{arm250},
10676 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10677 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10678 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10679 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10681 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10682 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10683 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10684 @samp{strongarm1110},
10685 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10686 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10687 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10688 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10689 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10690 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10691 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10692 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
10693 @samp{cortex-a15}, @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10694 @samp{cortex-m4}, @samp{cortex-m3},
10697 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10698 @samp{fa526}, @samp{fa626},
10699 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10701 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10702 performance for a blend of processors within architecture @var{arch}.
10703 The aim is to generate code that run well on the current most popular
10704 processors, balancing between optimizations that benefit some CPUs in the
10705 range, and avoiding performance pitfalls of other CPUs. The effects of
10706 this option may change in future GCC versions as CPU models come and go.
10708 @option{-mtune=native} causes the compiler to auto-detect the CPU
10709 of the build computer. At present, this feature is only supported on
10710 Linux, and not all architectures are recognized. If the auto-detect is
10711 unsuccessful the option has no effect.
10713 @item -mcpu=@var{name}
10715 This specifies the name of the target ARM processor. GCC uses this name
10716 to derive the name of the target ARM architecture (as if specified
10717 by @option{-march}) and the ARM processor type for which to tune for
10718 performance (as if specified by @option{-mtune}). Where this option
10719 is used in conjunction with @option{-march} or @option{-mtune},
10720 those options take precedence over the appropriate part of this option.
10722 Permissible names for this option are the same as those for
10725 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10726 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10727 See @option{-mtune} for more information.
10729 @option{-mcpu=native} causes the compiler to auto-detect the CPU
10730 of the build computer. At present, this feature is only supported on
10731 Linux, and not all architectures are recognized. If the auto-detect is
10732 unsuccessful the option has no effect.
10734 @item -mfpu=@var{name}
10735 @itemx -mfpe=@var{number}
10736 @itemx -mfp=@var{number}
10740 This specifies what floating-point hardware (or hardware emulation) is
10741 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10742 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10743 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10744 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10745 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10746 @option{-mfp} and @option{-mfpe} are synonyms for
10747 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10750 If @option{-msoft-float} is specified this specifies the format of
10751 floating-point values.
10753 If the selected floating-point hardware includes the NEON extension
10754 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10755 operations will not be used by GCC's auto-vectorization pass unless
10756 @option{-funsafe-math-optimizations} is also specified. This is
10757 because NEON hardware does not fully implement the IEEE 754 standard for
10758 floating-point arithmetic (in particular denormal values are treated as
10759 zero), so the use of NEON instructions may lead to a loss of precision.
10761 @item -mfp16-format=@var{name}
10762 @opindex mfp16-format
10763 Specify the format of the @code{__fp16} half-precision floating-point type.
10764 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10765 the default is @samp{none}, in which case the @code{__fp16} type is not
10766 defined. @xref{Half-Precision}, for more information.
10768 @item -mstructure-size-boundary=@var{n}
10769 @opindex mstructure-size-boundary
10770 The size of all structures and unions will be rounded up to a multiple
10771 of the number of bits set by this option. Permissible values are 8, 32
10772 and 64. The default value varies for different toolchains. For the COFF
10773 targeted toolchain the default value is 8. A value of 64 is only allowed
10774 if the underlying ABI supports it.
10776 Specifying the larger number can produce faster, more efficient code, but
10777 can also increase the size of the program. Different values are potentially
10778 incompatible. Code compiled with one value cannot necessarily expect to
10779 work with code or libraries compiled with another value, if they exchange
10780 information using structures or unions.
10782 @item -mabort-on-noreturn
10783 @opindex mabort-on-noreturn
10784 Generate a call to the function @code{abort} at the end of a
10785 @code{noreturn} function. It will be executed if the function tries to
10789 @itemx -mno-long-calls
10790 @opindex mlong-calls
10791 @opindex mno-long-calls
10792 Tells the compiler to perform function calls by first loading the
10793 address of the function into a register and then performing a subroutine
10794 call on this register. This switch is needed if the target function
10795 will lie outside of the 64 megabyte addressing range of the offset based
10796 version of subroutine call instruction.
10798 Even if this switch is enabled, not all function calls will be turned
10799 into long calls. The heuristic is that static functions, functions
10800 that have the @samp{short-call} attribute, functions that are inside
10801 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10802 definitions have already been compiled within the current compilation
10803 unit, will not be turned into long calls. The exception to this rule is
10804 that weak function definitions, functions with the @samp{long-call}
10805 attribute or the @samp{section} attribute, and functions that are within
10806 the scope of a @samp{#pragma long_calls} directive, will always be
10807 turned into long calls.
10809 This feature is not enabled by default. Specifying
10810 @option{-mno-long-calls} will restore the default behavior, as will
10811 placing the function calls within the scope of a @samp{#pragma
10812 long_calls_off} directive. Note these switches have no effect on how
10813 the compiler generates code to handle function calls via function
10816 @item -msingle-pic-base
10817 @opindex msingle-pic-base
10818 Treat the register used for PIC addressing as read-only, rather than
10819 loading it in the prologue for each function. The runtime system is
10820 responsible for initializing this register with an appropriate value
10821 before execution begins.
10823 @item -mpic-register=@var{reg}
10824 @opindex mpic-register
10825 Specify the register to be used for PIC addressing. The default is R10
10826 unless stack-checking is enabled, when R9 is used.
10828 @item -mcirrus-fix-invalid-insns
10829 @opindex mcirrus-fix-invalid-insns
10830 @opindex mno-cirrus-fix-invalid-insns
10831 Insert NOPs into the instruction stream to in order to work around
10832 problems with invalid Maverick instruction combinations. This option
10833 is only valid if the @option{-mcpu=ep9312} option has been used to
10834 enable generation of instructions for the Cirrus Maverick floating-point
10835 co-processor. This option is not enabled by default, since the
10836 problem is only present in older Maverick implementations. The default
10837 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10840 @item -mpoke-function-name
10841 @opindex mpoke-function-name
10842 Write the name of each function into the text section, directly
10843 preceding the function prologue. The generated code is similar to this:
10847 .ascii "arm_poke_function_name", 0
10850 .word 0xff000000 + (t1 - t0)
10851 arm_poke_function_name
10853 stmfd sp!, @{fp, ip, lr, pc@}
10857 When performing a stack backtrace, code can inspect the value of
10858 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10859 location @code{pc - 12} and the top 8 bits are set, then we know that
10860 there is a function name embedded immediately preceding this location
10861 and has length @code{((pc[-3]) & 0xff000000)}.
10868 Select between generating code that executes in ARM and Thumb
10869 states. The default for most configurations is to generate code
10870 that executes in ARM state, but the default can be changed by
10871 configuring GCC with the @option{--with-mode=}@var{state}
10875 @opindex mtpcs-frame
10876 Generate a stack frame that is compliant with the Thumb Procedure Call
10877 Standard for all non-leaf functions. (A leaf function is one that does
10878 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10880 @item -mtpcs-leaf-frame
10881 @opindex mtpcs-leaf-frame
10882 Generate a stack frame that is compliant with the Thumb Procedure Call
10883 Standard for all leaf functions. (A leaf function is one that does
10884 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10886 @item -mcallee-super-interworking
10887 @opindex mcallee-super-interworking
10888 Gives all externally visible functions in the file being compiled an ARM
10889 instruction set header which switches to Thumb mode before executing the
10890 rest of the function. This allows these functions to be called from
10891 non-interworking code. This option is not valid in AAPCS configurations
10892 because interworking is enabled by default.
10894 @item -mcaller-super-interworking
10895 @opindex mcaller-super-interworking
10896 Allows calls via function pointers (including virtual functions) to
10897 execute correctly regardless of whether the target code has been
10898 compiled for interworking or not. There is a small overhead in the cost
10899 of executing a function pointer if this option is enabled. This option
10900 is not valid in AAPCS configurations because interworking is enabled
10903 @item -mtp=@var{name}
10905 Specify the access model for the thread local storage pointer. The valid
10906 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10907 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10908 (supported in the arm6k architecture), and @option{auto}, which uses the
10909 best available method for the selected processor. The default setting is
10912 @item -mtls-dialect=@var{dialect}
10913 @opindex mtls-dialect
10914 Specify the dialect to use for accessing thread local storage. Two
10915 dialects are supported --- @option{gnu} and @option{gnu2}. The
10916 @option{gnu} dialect selects the original GNU scheme for supporting
10917 local and global dynamic TLS models. The @option{gnu2} dialect
10918 selects the GNU descriptor scheme, which provides better performance
10919 for shared libraries. The GNU descriptor scheme is compatible with
10920 the original scheme, but does require new assembler, linker and
10921 library support. Initial and local exec TLS models are unaffected by
10922 this option and always use the original scheme.
10924 @item -mword-relocations
10925 @opindex mword-relocations
10926 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
10927 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10928 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10931 @item -mfix-cortex-m3-ldrd
10932 @opindex mfix-cortex-m3-ldrd
10933 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10934 with overlapping destination and base registers are used. This option avoids
10935 generating these instructions. This option is enabled by default when
10936 @option{-mcpu=cortex-m3} is specified.
10938 @item -munaligned-access
10939 @itemx -mno-unaligned-access
10940 @opindex munaligned-access
10941 @opindex mno-unaligned-access
10942 Enables (or disables) reading and writing of 16- and 32- bit values
10943 from addresses that are not 16- or 32- bit aligned. By default
10944 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
10945 architectures, and enabled for all other architectures. If unaligned
10946 access is not enabled then words in packed data structures will be
10947 accessed a byte at a time.
10949 The ARM attribute @code{Tag_CPU_unaligned_access} will be set in the
10950 generated object file to either true or false, depending upon the
10951 setting of this option. If unaligned access is enabled then the
10952 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} will also be
10958 @subsection AVR Options
10959 @cindex AVR Options
10962 @item -mmcu=@var{mcu}
10964 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
10966 The default for this option is@tie{}@code{avr2}.
10968 GCC supports the following AVR devices and ISAs:
10973 ``Classic'' devices with up to 8@tie{}KiB of program memory.
10974 @*@var{mcu}@tie{}= @code{attiny22}, @code{attiny26}, @code{at90c8534},
10975 @code{at90s2313}, @code{at90s2323}, @code{at90s2333},
10976 @code{at90s2343}, @code{at90s4414}, @code{at90s4433},
10977 @code{at90s4434}, @code{at90s8515}, @code{at90s8535}.
10980 ``Classic'' devices with up to 8@tie{}KiB of program memory and with
10981 the @code{MOVW} instruction.
10982 @*@var{mcu}@tie{}= @code{ata6289}, @code{attiny13}, @code{attiny13a},
10983 @code{attiny2313}, @code{attiny2313a}, @code{attiny24},
10984 @code{attiny24a}, @code{attiny25}, @code{attiny261},
10985 @code{attiny261a}, @code{attiny43u}, @code{attiny4313},
10986 @code{attiny44}, @code{attiny44a}, @code{attiny45}, @code{attiny461},
10987 @code{attiny461a}, @code{attiny48}, @code{attiny84}, @code{attiny84a},
10988 @code{attiny85}, @code{attiny861}, @code{attiny861a}, @code{attiny87},
10989 @code{attiny88}, @code{at86rf401}.
10992 ``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory.
10993 @*@var{mcu}@tie{}= @code{at43usb355}, @code{at76c711}.
10996 ``Classic'' devices with 128@tie{}KiB of program memory.
10997 @*@var{mcu}@tie{}= @code{atmega103}, @code{at43usb320}.
11000 ``Classic'' devices with 16@tie{}KiB up to 64@tie{}KiB of program
11001 memory and with the @code{MOVW} instruction.
11002 @*@var{mcu}@tie{}= @code{atmega16u2}, @code{atmega32u2},
11003 @code{atmega8u2}, @code{attiny167}, @code{at90usb162},
11007 ``Enhanced'' devices with up to 8@tie{}KiB of program memory.
11008 @*@var{mcu}@tie{}= @code{atmega48}, @code{atmega48a},
11009 @code{atmega48p}, @code{atmega8}, @code{atmega8hva},
11010 @code{atmega8515}, @code{atmega8535}, @code{atmega88},
11011 @code{atmega88a}, @code{atmega88p}, @code{atmega88pa},
11012 @code{at90pwm1}, @code{at90pwm2}, @code{at90pwm2b}, @code{at90pwm3},
11013 @code{at90pwm3b}, @code{at90pwm81}.
11016 ``Enhanced'' devices with 16@tie{}KiB up to 64@tie{}KiB of program memory.
11017 @*@var{mcu}@tie{}= @code{atmega16}, @code{atmega16a},
11018 @code{atmega16hva}, @code{atmega16hva2}, @code{atmega16hvb},
11019 @code{atmega16m1}, @code{atmega16u4}, @code{atmega161},
11020 @code{atmega162}, @code{atmega163}, @code{atmega164a},
11021 @code{atmega164p}, @code{atmega165}, @code{atmega165a},
11022 @code{atmega165p}, @code{atmega168}, @code{atmega168a},
11023 @code{atmega168p}, @code{atmega169}, @code{atmega169a},
11024 @code{atmega169p}, @code{atmega169pa}, @code{atmega32},
11025 @code{atmega32c1}, @code{atmega32hvb}, @code{atmega32m1},
11026 @code{atmega32u4}, @code{atmega32u6}, @code{atmega323},
11027 @code{atmega324a}, @code{atmega324p}, @code{atmega324pa},
11028 @code{atmega325}, @code{atmega325a}, @code{atmega325p},
11029 @code{atmega3250}, @code{atmega3250a}, @code{atmega3250p},
11030 @code{atmega328}, @code{atmega328p}, @code{atmega329},
11031 @code{atmega329a}, @code{atmega329p}, @code{atmega329pa},
11032 @code{atmega3290}, @code{atmega3290a}, @code{atmega3290p},
11033 @code{atmega406}, @code{atmega64}, @code{atmega64c1},
11034 @code{atmega64hve}, @code{atmega64m1}, @code{atmega640},
11035 @code{atmega644}, @code{atmega644a}, @code{atmega644p},
11036 @code{atmega644pa}, @code{atmega645}, @code{atmega645a},
11037 @code{atmega645p}, @code{atmega6450}, @code{atmega6450a},
11038 @code{atmega6450p}, @code{atmega649}, @code{atmega649a},
11039 @code{atmega649p}, @code{atmega6490}, @code{at90can32},
11040 @code{at90can64}, @code{at90pwm216}, @code{at90pwm316},
11041 @code{at90scr100}, @code{at90usb646}, @code{at90usb647}, @code{at94k},
11045 ``Enhanced'' devices with 128@tie{}KiB of program memory.
11046 @*@var{mcu}@tie{}= @code{atmega128}, @code{atmega128rfa1},
11047 @code{atmega1280}, @code{atmega1281}, @code{atmega1284p},
11048 @code{at90can128}, @code{at90usb1286}, @code{at90usb1287}.
11051 ``Enhanced'' devices with 3-byte PC, i.e.@: with more than
11052 128@tie{}KiB of program memory.
11053 @*@var{mcu}@tie{}= @code{atmega2560}, @code{atmega2561}.
11056 ``XMEGA'' devices with more than 8@tie{}KiB and up to 64@tie{}KiB of
11058 @*@var{mcu}@tie{}= @code{atxmega16a4}, @code{atxmega16d4},
11059 @code{atxmega16x1}, @code{atxmega32a4}, @code{atxmega32d4},
11060 @code{atxmega32x1}.
11063 ``XMEGA'' devices with more than 64@tie{}KiB and up to 128@tie{}KiB of
11065 @*@var{mcu}@tie{}= @code{atxmega64a3}, @code{atxmega64d3}.
11068 ``XMEGA'' devices with more than 64@tie{}KiB and up to 128@tie{}KiB of
11069 program memory and more than 64@tie{}KiB of RAM.
11070 @*@var{mcu}@tie{}= @code{atxmega64a1}, @code{atxmega64a1u}.
11073 ``XMEGA'' devices with more than 128@tie{}KiB of program memory.
11074 @*@var{mcu}@tie{}= @code{atxmega128a3}, @code{atxmega128d3},
11075 @code{atxmega192a3}, @code{atxmega192d3}, @code{atxmega256a3},
11076 @code{atxmega256a3b}, @code{atxmega256a3bu}, @code{atxmega256d3}.
11079 ``XMEGA'' devices with more than 128@tie{}KiB of program memory and
11080 more than 64@tie{}KiB of RAM.
11081 @*@var{mcu}@tie{}= @code{atxmega128a1}, @code{atxmega128a1u}.
11084 This ISA is implemented by the minimal AVR core and supported for
11086 @*@var{mcu}@tie{}= @code{attiny11}, @code{attiny12}, @code{attiny15},
11087 @code{attiny28}, @code{at90s1200}.
11091 @item -maccumulate-args
11092 @opindex maccumulate-args
11093 Accumulate outgoing function arguments and acquire/release the needed
11094 stack space for outgoing function arguments once in function
11095 prologue/epilogue. Without this option, outgoing arguments are pushed
11096 before calling a function and popped afterwards.
11098 Popping the arguments after the function call can be expensive on
11099 AVR so that accumulating the stack space might lead to smaller
11100 executables because arguments need not to be removed from the
11101 stack after such a function call.
11103 This option can lead to reduced code size for functions that perform
11104 several calls to functions that get their arguments on the stack like
11105 calls to printf-like functions.
11107 @item -mbranch-cost=@var{cost}
11108 @opindex mbranch-cost
11109 Set the branch costs for conditional branch instructions to
11110 @var{cost}. Reasonable values for @var{cost} are small, non-negative
11111 integers. The default branch cost is 0.
11113 @item -mcall-prologues
11114 @opindex mcall-prologues
11115 Functions prologues/epilogues are expanded as calls to appropriate
11116 subroutines. Code size is smaller.
11120 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
11121 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
11122 and @code{long long} is 4 bytes. Please note that this option does not
11123 conform to the C standards, but it results in smaller code
11126 @item -mno-interrupts
11127 @opindex mno-interrupts
11128 Generated code is not compatible with hardware interrupts.
11129 Code size is smaller.
11133 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
11134 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
11135 Setting @code{-mrelax} just adds the @code{--relax} option to the
11136 linker command line when the linker is called.
11138 Jump relaxing is performed by the linker because jump offsets are not
11139 known before code is located. Therefore, the assembler code generated by the
11140 compiler is the same, but the instructions in the executable may
11141 differ from instructions in the assembler code.
11143 Relaxing must be turned on if linker stubs are needed, see the
11144 section on @code{EIND} and linker stubs below.
11146 @item -mshort-calls
11147 @opindex mshort-calls
11148 This option has been deprecated and will be removed in GCC 4.8.
11149 See @code{-mrelax} for a replacement.
11151 Use @code{RCALL}/@code{RJMP} instructions even on devices with
11152 16@tie{}KiB or more of program memory, i.e.@: on devices that
11153 have the @code{CALL} and @code{JMP} instructions.
11157 Treat the stack pointer register as an 8-bit register,
11158 i.e.@: assume the high byte of the stack pointer is zero.
11159 In general, you don't need to set this option by hand.
11161 This option is used internally by the compiler to select and
11162 build multilibs for architectures @code{avr2} and @code{avr25}.
11163 These architectures mix devices with and without @code{SPH}.
11164 For any setting other than @code{-mmcu=avr2} or @code{-mmcu=avr25}
11165 the compiler driver will add or remove this option from the compiler
11166 proper's command line, because the compiler then knows if the device
11167 or architecture has an 8-bit stack pointer and thus no @code{SPH}
11172 Use address register @code{X} in a way proposed by the hardware. This means
11173 that @code{X} is only used in indirect, post-increment or
11174 pre-decrement addressing.
11176 Without this option, the @code{X} register may be used in the same way
11177 as @code{Y} or @code{Z} which then is emulated by additional
11179 For example, loading a value with @code{X+const} addressing with a
11180 small non-negative @code{const < 64} to a register @var{Rn} is
11184 adiw r26, const ; X += const
11185 ld @var{Rn}, X ; @var{Rn} = *X
11186 sbiw r26, const ; X -= const
11190 @opindex mtiny-stack
11191 Only change the lower 8@tie{}bits of the stack pointer.
11194 @subsubsection @code{EIND} and Devices with more than 128 Ki Bytes of Flash
11195 @cindex @code{EIND}
11196 Pointers in the implementation are 16@tie{}bits wide.
11197 The address of a function or label is represented as word address so
11198 that indirect jumps and calls can target any code address in the
11199 range of 64@tie{}Ki words.
11201 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
11202 bytes of program memory space, there is a special function register called
11203 @code{EIND} that serves as most significant part of the target address
11204 when @code{EICALL} or @code{EIJMP} instructions are used.
11206 Indirect jumps and calls on these devices are handled as follows by
11207 the compiler and are subject to some limitations:
11212 The compiler never sets @code{EIND}.
11215 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
11216 instructions or might read @code{EIND} directly in order to emulate an
11217 indirect call/jump by means of a @code{RET} instruction.
11220 The compiler assumes that @code{EIND} never changes during the startup
11221 code or during the application. In particular, @code{EIND} is not
11222 saved/restored in function or interrupt service routine
11226 For indirect calls to functions and computed goto, the linker
11227 generates @emph{stubs}. Stubs are jump pads sometimes also called
11228 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
11229 The stub contains a direct jump to the desired address.
11232 Linker relaxation must be turned on so that the linker will generate
11233 the stubs correctly an all situaltion. See the compiler option
11234 @code{-mrelax} and the linler option @code{--relax}.
11235 There are corner cases where the linker is supposed to generate stubs
11236 but aborts without relaxation and without a helpful error message.
11239 The default linker script is arranged for code with @code{EIND = 0}.
11240 If code is supposed to work for a setup with @code{EIND != 0}, a custom
11241 linker script has to be used in order to place the sections whose
11242 name start with @code{.trampolines} into the segment where @code{EIND}
11246 The startup code from libgcc never sets @code{EIND}.
11247 Notice that startup code is a blend of code from libgcc and AVR-LibC.
11248 For the impact of AVR-LibC on @code{EIND}, see the
11249 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
11252 It is legitimate for user-specific startup code to set up @code{EIND}
11253 early, for example by means of initialization code located in
11254 section @code{.init3}. Such code runs prior to general startup code
11255 that initializes RAM and calls constructors, but after the bit
11256 of startup code from AVR-LibC that sets @code{EIND} to the segment
11257 where the vector table is located.
11259 #include <avr/io.h>
11262 __attribute__((section(".init3"),naked,used,no_instrument_function))
11263 init3_set_eind (void)
11265 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
11266 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
11271 The @code{__trampolines_start} symbol is defined in the linker script.
11274 Stubs are generated automatically by the linker if
11275 the following two conditions are met:
11278 @item The address of a label is taken by means of the @code{gs} modifier
11279 (short for @emph{generate stubs}) like so:
11281 LDI r24, lo8(gs(@var{func}))
11282 LDI r25, hi8(gs(@var{func}))
11284 @item The final location of that label is in a code segment
11285 @emph{outside} the segment where the stubs are located.
11289 The compiler emits such @code{gs} modifiers for code labels in the
11290 following situations:
11292 @item Taking address of a function or code label.
11293 @item Computed goto.
11294 @item If prologue-save function is used, see @option{-mcall-prologues}
11295 command-line option.
11296 @item Switch/case dispatch tables. If you do not want such dispatch
11297 tables you can specify the @option{-fno-jump-tables} command-line option.
11298 @item C and C++ constructors/destructors called during startup/shutdown.
11299 @item If the tools hit a @code{gs()} modifier explained above.
11303 Jumping to non-symbolic addresses like so is @emph{not} supported:
11308 /* Call function at word address 0x2 */
11309 return ((int(*)(void)) 0x2)();
11313 Instead, a stub has to be set up, i.e.@: the function has to be called
11314 through a symbol (@code{func_4} in the example):
11319 extern int func_4 (void);
11321 /* Call function at byte address 0x4 */
11326 and the application be linked with @code{-Wl,--defsym,func_4=0x4}.
11327 Alternatively, @code{func_4} can be defined in the linker script.
11330 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
11331 @cindex @code{RAMPD}
11332 @cindex @code{RAMPX}
11333 @cindex @code{RAMPY}
11334 @cindex @code{RAMPZ}
11335 Some AVR devices support memories larger than the 64@tie{}KiB range
11336 that can be accessed with 16-bit pointers. To access memory locations
11337 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
11338 register is used as high part of the address:
11339 The @code{X}, @code{Y}, @code{Z} address register is concatenated
11340 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
11341 register, respectively, to get a wide address. Similarly,
11342 @code{RAMPD} is used together with direct addressing.
11346 The startup code initializes the @code{RAMP} special function
11347 registers with zero.
11350 If a @ref{AVR Named Address Spaces,named address space} other than
11351 generic or @code{__flash} is used, then @code{RAMPZ} is set
11352 as needed before the operation.
11355 If the device supports RAM larger than 64@tie{KiB} and the compiler
11356 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
11357 is reset to zero after the operation.
11360 If the device comes with a specific @code{RAMP} register, the ISR
11361 prologue/epilogue saves/restores that SFR and initializes it with
11362 zero in case the ISR code might (implicitly) use it.
11365 RAM larger than 64@tie{KiB} is not supported by GCC for AVR targets.
11366 If you use inline assembler to read from locations outside the
11367 16-bit address range and change one of the @code{RAMP} registers,
11368 you must reset it to zero after the access.
11372 @subsubsection AVR Built-in Macros
11374 GCC defines several built-in macros so that the user code can test
11375 for the presence or absence of features. Almost any of the following
11376 built-in macros are deduced from device capabilities and thus
11377 triggered by the @code{-mmcu=} command-line option.
11379 For even more AVR-specific built-in macros see
11380 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
11385 Build-in macro that resolves to a decimal number that identifies the
11386 architecture and depends on the @code{-mmcu=@var{mcu}} option.
11387 Possible values are:
11389 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
11390 @code{4}, @code{5}, @code{51}, @code{6}, @code{102}, @code{104},
11391 @code{105}, @code{106}, @code{107}
11393 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3},
11394 @code{avr31}, @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51},
11395 @code{avr6}, @code{avrxmega2}, @code{avrxmega4}, @code{avrxmega5},
11396 @code{avrxmega6}, @code{avrxmega7}, respectively.
11397 If @var{mcu} specifies a device, this built-in macro is set
11398 accordingly. For example, with @code{-mmcu=atmega8} the macro will be
11399 defined to @code{4}.
11401 @item __AVR_@var{Device}__
11402 Setting @code{-mmcu=@var{device}} defines this built-in macro which reflects
11403 the device's name. For example, @code{-mmcu=atmega8} defines the
11404 built-in macro @code{__AVR_ATmega8__}, @code{-mmcu=attiny261a} defines
11405 @code{__AVR_ATtiny261A__}, etc.
11407 The built-in macros' names follow
11408 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
11409 the device name as from the AVR user manual. The difference between
11410 @var{Device} in the built-in macro and @var{device} in
11411 @code{-mmcu=@var{device}} is that the latter is always lowercase.
11413 If @var{device} is not a device but only a core architecture like
11414 @code{avr51}, this macro will not be defined.
11416 @item __AVR_XMEGA__
11417 The device/architecture belongs to the XMEGA family of devices.
11419 @item __AVR_HAVE_ELPM__
11420 The device has the the @code{ELPM} instruction.
11422 @item __AVR_HAVE_ELPMX__
11423 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
11424 R@var{n},Z+} instructions.
11426 @item __AVR_HAVE_MOVW__
11427 The device has the @code{MOVW} instruction to perform 16-bit
11428 register-register moves.
11430 @item __AVR_HAVE_LPMX__
11431 The device has the @code{LPM R@var{n},Z} and
11432 @code{LPM R@var{n},Z+} instructions.
11434 @item __AVR_HAVE_MUL__
11435 The device has a hardware multiplier.
11437 @item __AVR_HAVE_JMP_CALL__
11438 The device has the @code{JMP} and @code{CALL} instructions.
11439 This is the case for devices with at least 16@tie{}KiB of program
11440 memory and if @code{-mshort-calls} is not set.
11442 @item __AVR_HAVE_EIJMP_EICALL__
11443 @item __AVR_3_BYTE_PC__
11444 The device has the @code{EIJMP} and @code{EICALL} instructions.
11445 This is the case for devices with more than 128@tie{}KiB of program memory.
11446 This also means that the program counter
11447 (PC) is 3@tie{}bytes wide.
11449 @item __AVR_2_BYTE_PC__
11450 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
11451 with up to 128@tie{}KiB of program memory.
11453 @item __AVR_HAVE_8BIT_SP__
11454 @item __AVR_HAVE_16BIT_SP__
11455 The stack pointer (SP) register is treated as 8-bit respectively
11456 16-bit register by the compiler.
11457 The definition of these macros is affected by @code{-mtiny-stack}.
11459 @item __AVR_HAVE_SPH__
11461 The device has the SPH (high part of stack pointer) special function
11462 register or has an 8-bit stack pointer, respectively.
11463 The definition of these macros is affected by @code{-mmcu=} and
11464 in the cases of @code{-mmcu=avr2} and @code{-mmcu=avr25} also
11467 @item __AVR_HAVE_RAMPD__
11468 @item __AVR_HAVE_RAMPX__
11469 @item __AVR_HAVE_RAMPY__
11470 @item __AVR_HAVE_RAMPZ__
11471 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
11472 @code{RAMPZ} special function register, respectively.
11474 @item __NO_INTERRUPTS__
11475 This macro reflects the @code{-mno-interrupts} command line option.
11477 @item __AVR_ERRATA_SKIP__
11478 @item __AVR_ERRATA_SKIP_JMP_CALL__
11479 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
11480 instructions because of a hardware erratum. Skip instructions are
11481 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
11482 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
11485 @item __AVR_SFR_OFFSET__=@var{offset}
11486 Instructions that can address I/O special function registers directly
11487 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
11488 address as if addressed by an instruction to access RAM like @code{LD}
11489 or @code{STS}. This offset depends on the device architecture and has
11490 to be subtracted from the RAM address in order to get the
11491 respective I/O@tie{}address.
11493 @item __WITH_AVRLIBC__
11494 The compiler is configured to be used together with AVR-Libc.
11495 See the @code{--with-avrlibc} configure option.
11499 @node Blackfin Options
11500 @subsection Blackfin Options
11501 @cindex Blackfin Options
11504 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
11506 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
11507 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
11508 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
11509 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
11510 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
11511 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
11512 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
11513 @samp{bf561}, @samp{bf592}.
11514 The optional @var{sirevision} specifies the silicon revision of the target
11515 Blackfin processor. Any workarounds available for the targeted silicon revision
11516 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
11517 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
11518 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
11519 hexadecimal digits representing the major and minor numbers in the silicon
11520 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
11521 is not defined. If @var{sirevision} is @samp{any}, the
11522 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
11523 If this optional @var{sirevision} is not used, GCC assumes the latest known
11524 silicon revision of the targeted Blackfin processor.
11526 Support for @samp{bf561} is incomplete. For @samp{bf561},
11527 Only the processor macro is defined.
11528 Without this option, @samp{bf532} is used as the processor by default.
11529 The corresponding predefined processor macros for @var{cpu} is to
11530 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
11531 provided by libgloss to be linked in if @option{-msim} is not given.
11535 Specifies that the program will be run on the simulator. This causes
11536 the simulator BSP provided by libgloss to be linked in. This option
11537 has effect only for @samp{bfin-elf} toolchain.
11538 Certain other options, such as @option{-mid-shared-library} and
11539 @option{-mfdpic}, imply @option{-msim}.
11541 @item -momit-leaf-frame-pointer
11542 @opindex momit-leaf-frame-pointer
11543 Don't keep the frame pointer in a register for leaf functions. This
11544 avoids the instructions to save, set up and restore frame pointers and
11545 makes an extra register available in leaf functions. The option
11546 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
11547 which might make debugging harder.
11549 @item -mspecld-anomaly
11550 @opindex mspecld-anomaly
11551 When enabled, the compiler will ensure that the generated code does not
11552 contain speculative loads after jump instructions. If this option is used,
11553 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
11555 @item -mno-specld-anomaly
11556 @opindex mno-specld-anomaly
11557 Don't generate extra code to prevent speculative loads from occurring.
11559 @item -mcsync-anomaly
11560 @opindex mcsync-anomaly
11561 When enabled, the compiler will ensure that the generated code does not
11562 contain CSYNC or SSYNC instructions too soon after conditional branches.
11563 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
11565 @item -mno-csync-anomaly
11566 @opindex mno-csync-anomaly
11567 Don't generate extra code to prevent CSYNC or SSYNC instructions from
11568 occurring too soon after a conditional branch.
11572 When enabled, the compiler is free to take advantage of the knowledge that
11573 the entire program fits into the low 64k of memory.
11576 @opindex mno-low-64k
11577 Assume that the program is arbitrarily large. This is the default.
11579 @item -mstack-check-l1
11580 @opindex mstack-check-l1
11581 Do stack checking using information placed into L1 scratchpad memory by the
11584 @item -mid-shared-library
11585 @opindex mid-shared-library
11586 Generate code that supports shared libraries via the library ID method.
11587 This allows for execute in place and shared libraries in an environment
11588 without virtual memory management. This option implies @option{-fPIC}.
11589 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11591 @item -mno-id-shared-library
11592 @opindex mno-id-shared-library
11593 Generate code that doesn't assume ID based shared libraries are being used.
11594 This is the default.
11596 @item -mleaf-id-shared-library
11597 @opindex mleaf-id-shared-library
11598 Generate code that supports shared libraries via the library ID method,
11599 but assumes that this library or executable won't link against any other
11600 ID shared libraries. That allows the compiler to use faster code for jumps
11603 @item -mno-leaf-id-shared-library
11604 @opindex mno-leaf-id-shared-library
11605 Do not assume that the code being compiled won't link against any ID shared
11606 libraries. Slower code will be generated for jump and call insns.
11608 @item -mshared-library-id=n
11609 @opindex mshared-library-id
11610 Specified the identification number of the ID based shared library being
11611 compiled. Specifying a value of 0 will generate more compact code, specifying
11612 other values will force the allocation of that number to the current
11613 library but is no more space or time efficient than omitting this option.
11617 Generate code that allows the data segment to be located in a different
11618 area of memory from the text segment. This allows for execute in place in
11619 an environment without virtual memory management by eliminating relocations
11620 against the text section.
11622 @item -mno-sep-data
11623 @opindex mno-sep-data
11624 Generate code that assumes that the data segment follows the text segment.
11625 This is the default.
11628 @itemx -mno-long-calls
11629 @opindex mlong-calls
11630 @opindex mno-long-calls
11631 Tells the compiler to perform function calls by first loading the
11632 address of the function into a register and then performing a subroutine
11633 call on this register. This switch is needed if the target function
11634 lies outside of the 24-bit addressing range of the offset-based
11635 version of subroutine call instruction.
11637 This feature is not enabled by default. Specifying
11638 @option{-mno-long-calls} will restore the default behavior. Note these
11639 switches have no effect on how the compiler generates code to handle
11640 function calls via function pointers.
11644 Link with the fast floating-point library. This library relaxes some of
11645 the IEEE floating-point standard's rules for checking inputs against
11646 Not-a-Number (NAN), in the interest of performance.
11649 @opindex minline-plt
11650 Enable inlining of PLT entries in function calls to functions that are
11651 not known to bind locally. It has no effect without @option{-mfdpic}.
11654 @opindex mmulticore
11655 Build standalone application for multicore Blackfin processor. Proper
11656 start files and link scripts will be used to support multicore.
11657 This option defines @code{__BFIN_MULTICORE}. It can only be used with
11658 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
11659 @option{-mcorea} or @option{-mcoreb}. If it's used without
11660 @option{-mcorea} or @option{-mcoreb}, single application/dual core
11661 programming model is used. In this model, the main function of Core B
11662 should be named as coreb_main. If it's used with @option{-mcorea} or
11663 @option{-mcoreb}, one application per core programming model is used.
11664 If this option is not used, single core application programming
11669 Build standalone application for Core A of BF561 when using
11670 one application per core programming model. Proper start files
11671 and link scripts will be used to support Core A. This option
11672 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
11676 Build standalone application for Core B of BF561 when using
11677 one application per core programming model. Proper start files
11678 and link scripts will be used to support Core B. This option
11679 defines @code{__BFIN_COREB}. When this option is used, coreb_main
11680 should be used instead of main. It must be used with
11681 @option{-mmulticore}.
11685 Build standalone application for SDRAM. Proper start files and
11686 link scripts will be used to put the application into SDRAM.
11687 Loader should initialize SDRAM before loading the application
11688 into SDRAM. This option defines @code{__BFIN_SDRAM}.
11692 Assume that ICPLBs are enabled at run time. This has an effect on certain
11693 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
11694 are enabled; for standalone applications the default is off.
11698 @subsection C6X Options
11699 @cindex C6X Options
11702 @item -march=@var{name}
11704 This specifies the name of the target architecture. GCC uses this
11705 name to determine what kind of instructions it can emit when generating
11706 assembly code. Permissible names are: @samp{c62x},
11707 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
11710 @opindex mbig-endian
11711 Generate code for a big-endian target.
11713 @item -mlittle-endian
11714 @opindex mlittle-endian
11715 Generate code for a little-endian target. This is the default.
11719 Choose startup files and linker script suitable for the simulator.
11721 @item -msdata=default
11722 @opindex msdata=default
11723 Put small global and static data in the @samp{.neardata} section,
11724 which is pointed to by register @code{B14}. Put small uninitialized
11725 global and static data in the @samp{.bss} section, which is adjacent
11726 to the @samp{.neardata} section. Put small read-only data into the
11727 @samp{.rodata} section. The corresponding sections used for large
11728 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
11731 @opindex msdata=all
11732 Put all data, not just small objets, into the sections reserved for
11733 small data, and use addressing relative to the @code{B14} register to
11737 @opindex msdata=none
11738 Make no use of the sections reserved for small data, and use absolute
11739 addresses to access all data. Put all initialized global and static
11740 data in the @samp{.fardata} section, and all uninitialized data in the
11741 @samp{.far} section. Put all constant data into the @samp{.const}
11746 @subsection CRIS Options
11747 @cindex CRIS Options
11749 These options are defined specifically for the CRIS ports.
11752 @item -march=@var{architecture-type}
11753 @itemx -mcpu=@var{architecture-type}
11756 Generate code for the specified architecture. The choices for
11757 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
11758 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
11759 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
11762 @item -mtune=@var{architecture-type}
11764 Tune to @var{architecture-type} everything applicable about the generated
11765 code, except for the ABI and the set of available instructions. The
11766 choices for @var{architecture-type} are the same as for
11767 @option{-march=@var{architecture-type}}.
11769 @item -mmax-stack-frame=@var{n}
11770 @opindex mmax-stack-frame
11771 Warn when the stack frame of a function exceeds @var{n} bytes.
11777 The options @option{-metrax4} and @option{-metrax100} are synonyms for
11778 @option{-march=v3} and @option{-march=v8} respectively.
11780 @item -mmul-bug-workaround
11781 @itemx -mno-mul-bug-workaround
11782 @opindex mmul-bug-workaround
11783 @opindex mno-mul-bug-workaround
11784 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
11785 models where it applies. This option is active by default.
11789 Enable CRIS-specific verbose debug-related information in the assembly
11790 code. This option also has the effect to turn off the @samp{#NO_APP}
11791 formatted-code indicator to the assembler at the beginning of the
11796 Do not use condition-code results from previous instruction; always emit
11797 compare and test instructions before use of condition codes.
11799 @item -mno-side-effects
11800 @opindex mno-side-effects
11801 Do not emit instructions with side-effects in addressing modes other than
11804 @item -mstack-align
11805 @itemx -mno-stack-align
11806 @itemx -mdata-align
11807 @itemx -mno-data-align
11808 @itemx -mconst-align
11809 @itemx -mno-const-align
11810 @opindex mstack-align
11811 @opindex mno-stack-align
11812 @opindex mdata-align
11813 @opindex mno-data-align
11814 @opindex mconst-align
11815 @opindex mno-const-align
11816 These options (no-options) arranges (eliminate arrangements) for the
11817 stack-frame, individual data and constants to be aligned for the maximum
11818 single data access size for the chosen CPU model. The default is to
11819 arrange for 32-bit alignment. ABI details such as structure layout are
11820 not affected by these options.
11828 Similar to the stack- data- and const-align options above, these options
11829 arrange for stack-frame, writable data and constants to all be 32-bit,
11830 16-bit or 8-bit aligned. The default is 32-bit alignment.
11832 @item -mno-prologue-epilogue
11833 @itemx -mprologue-epilogue
11834 @opindex mno-prologue-epilogue
11835 @opindex mprologue-epilogue
11836 With @option{-mno-prologue-epilogue}, the normal function prologue and
11837 epilogue which set up the stack frame are omitted and no return
11838 instructions or return sequences are generated in the code. Use this
11839 option only together with visual inspection of the compiled code: no
11840 warnings or errors are generated when call-saved registers must be saved,
11841 or storage for local variable needs to be allocated.
11845 @opindex mno-gotplt
11847 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11848 instruction sequences that load addresses for functions from the PLT part
11849 of the GOT rather than (traditional on other architectures) calls to the
11850 PLT@. The default is @option{-mgotplt}.
11854 Legacy no-op option only recognized with the cris-axis-elf and
11855 cris-axis-linux-gnu targets.
11859 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11863 This option, recognized for the cris-axis-elf arranges
11864 to link with input-output functions from a simulator library. Code,
11865 initialized data and zero-initialized data are allocated consecutively.
11869 Like @option{-sim}, but pass linker options to locate initialized data at
11870 0x40000000 and zero-initialized data at 0x80000000.
11874 @subsection CR16 Options
11875 @cindex CR16 Options
11877 These options are defined specifically for the CR16 ports.
11883 Enable the use of multiply-accumulate instructions. Disabled by default.
11887 @opindex mcr16cplus
11889 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
11894 Links the library libsim.a which is in compatible with simulator. Applicable
11895 to elf compiler only.
11899 Choose integer type as 32-bit wide.
11903 Generates sbit/cbit instructions for bit manipulations.
11905 @item -mdata-model=@var{model}
11906 @opindex mdata-model
11907 Choose a data model. The choices for @var{model} are @samp{near},
11908 @samp{far} or @samp{medium}. @samp{medium} is default.
11909 However, @samp{far} is not valid when -mcr16c option is chosen as
11910 CR16C architecture does not support far data model.
11913 @node Darwin Options
11914 @subsection Darwin Options
11915 @cindex Darwin options
11917 These options are defined for all architectures running the Darwin operating
11920 FSF GCC on Darwin does not create ``fat'' object files; it will create
11921 an object file for the single architecture that it was built to
11922 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11923 @option{-arch} options are used; it does so by running the compiler or
11924 linker multiple times and joining the results together with
11927 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11928 @samp{i686}) is determined by the flags that specify the ISA
11929 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11930 @option{-force_cpusubtype_ALL} option can be used to override this.
11932 The Darwin tools vary in their behavior when presented with an ISA
11933 mismatch. The assembler, @file{as}, will only permit instructions to
11934 be used that are valid for the subtype of the file it is generating,
11935 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11936 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11937 and print an error if asked to create a shared library with a less
11938 restrictive subtype than its input files (for instance, trying to put
11939 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11940 for executables, @file{ld}, will quietly give the executable the most
11941 restrictive subtype of any of its input files.
11946 Add the framework directory @var{dir} to the head of the list of
11947 directories to be searched for header files. These directories are
11948 interleaved with those specified by @option{-I} options and are
11949 scanned in a left-to-right order.
11951 A framework directory is a directory with frameworks in it. A
11952 framework is a directory with a @samp{"Headers"} and/or
11953 @samp{"PrivateHeaders"} directory contained directly in it that ends
11954 in @samp{".framework"}. The name of a framework is the name of this
11955 directory excluding the @samp{".framework"}. Headers associated with
11956 the framework are found in one of those two directories, with
11957 @samp{"Headers"} being searched first. A subframework is a framework
11958 directory that is in a framework's @samp{"Frameworks"} directory.
11959 Includes of subframework headers can only appear in a header of a
11960 framework that contains the subframework, or in a sibling subframework
11961 header. Two subframeworks are siblings if they occur in the same
11962 framework. A subframework should not have the same name as a
11963 framework, a warning will be issued if this is violated. Currently a
11964 subframework cannot have subframeworks, in the future, the mechanism
11965 may be extended to support this. The standard frameworks can be found
11966 in @samp{"/System/Library/Frameworks"} and
11967 @samp{"/Library/Frameworks"}. An example include looks like
11968 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11969 the name of the framework and header.h is found in the
11970 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11972 @item -iframework@var{dir}
11973 @opindex iframework
11974 Like @option{-F} except the directory is a treated as a system
11975 directory. The main difference between this @option{-iframework} and
11976 @option{-F} is that with @option{-iframework} the compiler does not
11977 warn about constructs contained within header files found via
11978 @var{dir}. This option is valid only for the C family of languages.
11982 Emit debugging information for symbols that are used. For STABS
11983 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11984 This is by default ON@.
11988 Emit debugging information for all symbols and types.
11990 @item -mmacosx-version-min=@var{version}
11991 The earliest version of MacOS X that this executable will run on
11992 is @var{version}. Typical values of @var{version} include @code{10.1},
11993 @code{10.2}, and @code{10.3.9}.
11995 If the compiler was built to use the system's headers by default,
11996 then the default for this option is the system version on which the
11997 compiler is running, otherwise the default is to make choices that
11998 are compatible with as many systems and code bases as possible.
12002 Enable kernel development mode. The @option{-mkernel} option sets
12003 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
12004 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
12005 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
12006 applicable. This mode also sets @option{-mno-altivec},
12007 @option{-msoft-float}, @option{-fno-builtin} and
12008 @option{-mlong-branch} for PowerPC targets.
12010 @item -mone-byte-bool
12011 @opindex mone-byte-bool
12012 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
12013 By default @samp{sizeof(bool)} is @samp{4} when compiling for
12014 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
12015 option has no effect on x86.
12017 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
12018 to generate code that is not binary compatible with code generated
12019 without that switch. Using this switch may require recompiling all
12020 other modules in a program, including system libraries. Use this
12021 switch to conform to a non-default data model.
12023 @item -mfix-and-continue
12024 @itemx -ffix-and-continue
12025 @itemx -findirect-data
12026 @opindex mfix-and-continue
12027 @opindex ffix-and-continue
12028 @opindex findirect-data
12029 Generate code suitable for fast turn around development. Needed to
12030 enable gdb to dynamically load @code{.o} files into already running
12031 programs. @option{-findirect-data} and @option{-ffix-and-continue}
12032 are provided for backwards compatibility.
12036 Loads all members of static archive libraries.
12037 See man ld(1) for more information.
12039 @item -arch_errors_fatal
12040 @opindex arch_errors_fatal
12041 Cause the errors having to do with files that have the wrong architecture
12044 @item -bind_at_load
12045 @opindex bind_at_load
12046 Causes the output file to be marked such that the dynamic linker will
12047 bind all undefined references when the file is loaded or launched.
12051 Produce a Mach-o bundle format file.
12052 See man ld(1) for more information.
12054 @item -bundle_loader @var{executable}
12055 @opindex bundle_loader
12056 This option specifies the @var{executable} that will be loading the build
12057 output file being linked. See man ld(1) for more information.
12060 @opindex dynamiclib
12061 When passed this option, GCC will produce a dynamic library instead of
12062 an executable when linking, using the Darwin @file{libtool} command.
12064 @item -force_cpusubtype_ALL
12065 @opindex force_cpusubtype_ALL
12066 This causes GCC's output file to have the @var{ALL} subtype, instead of
12067 one controlled by the @option{-mcpu} or @option{-march} option.
12069 @item -allowable_client @var{client_name}
12070 @itemx -client_name
12071 @itemx -compatibility_version
12072 @itemx -current_version
12074 @itemx -dependency-file
12076 @itemx -dylinker_install_name
12078 @itemx -exported_symbols_list
12081 @itemx -flat_namespace
12082 @itemx -force_flat_namespace
12083 @itemx -headerpad_max_install_names
12086 @itemx -install_name
12087 @itemx -keep_private_externs
12088 @itemx -multi_module
12089 @itemx -multiply_defined
12090 @itemx -multiply_defined_unused
12093 @itemx -no_dead_strip_inits_and_terms
12094 @itemx -nofixprebinding
12095 @itemx -nomultidefs
12097 @itemx -noseglinkedit
12098 @itemx -pagezero_size
12100 @itemx -prebind_all_twolevel_modules
12101 @itemx -private_bundle
12103 @itemx -read_only_relocs
12105 @itemx -sectobjectsymbols
12109 @itemx -sectobjectsymbols
12112 @itemx -segs_read_only_addr
12114 @itemx -segs_read_write_addr
12115 @itemx -seg_addr_table
12116 @itemx -seg_addr_table_filename
12117 @itemx -seglinkedit
12119 @itemx -segs_read_only_addr
12120 @itemx -segs_read_write_addr
12121 @itemx -single_module
12123 @itemx -sub_library
12125 @itemx -sub_umbrella
12126 @itemx -twolevel_namespace
12129 @itemx -unexported_symbols_list
12130 @itemx -weak_reference_mismatches
12131 @itemx -whatsloaded
12132 @opindex allowable_client
12133 @opindex client_name
12134 @opindex compatibility_version
12135 @opindex current_version
12136 @opindex dead_strip
12137 @opindex dependency-file
12138 @opindex dylib_file
12139 @opindex dylinker_install_name
12141 @opindex exported_symbols_list
12143 @opindex flat_namespace
12144 @opindex force_flat_namespace
12145 @opindex headerpad_max_install_names
12146 @opindex image_base
12148 @opindex install_name
12149 @opindex keep_private_externs
12150 @opindex multi_module
12151 @opindex multiply_defined
12152 @opindex multiply_defined_unused
12153 @opindex noall_load
12154 @opindex no_dead_strip_inits_and_terms
12155 @opindex nofixprebinding
12156 @opindex nomultidefs
12158 @opindex noseglinkedit
12159 @opindex pagezero_size
12161 @opindex prebind_all_twolevel_modules
12162 @opindex private_bundle
12163 @opindex read_only_relocs
12165 @opindex sectobjectsymbols
12168 @opindex sectcreate
12169 @opindex sectobjectsymbols
12172 @opindex segs_read_only_addr
12173 @opindex segs_read_write_addr
12174 @opindex seg_addr_table
12175 @opindex seg_addr_table_filename
12176 @opindex seglinkedit
12178 @opindex segs_read_only_addr
12179 @opindex segs_read_write_addr
12180 @opindex single_module
12182 @opindex sub_library
12183 @opindex sub_umbrella
12184 @opindex twolevel_namespace
12187 @opindex unexported_symbols_list
12188 @opindex weak_reference_mismatches
12189 @opindex whatsloaded
12190 These options are passed to the Darwin linker. The Darwin linker man page
12191 describes them in detail.
12194 @node DEC Alpha Options
12195 @subsection DEC Alpha Options
12197 These @samp{-m} options are defined for the DEC Alpha implementations:
12200 @item -mno-soft-float
12201 @itemx -msoft-float
12202 @opindex mno-soft-float
12203 @opindex msoft-float
12204 Use (do not use) the hardware floating-point instructions for
12205 floating-point operations. When @option{-msoft-float} is specified,
12206 functions in @file{libgcc.a} will be used to perform floating-point
12207 operations. Unless they are replaced by routines that emulate the
12208 floating-point operations, or compiled in such a way as to call such
12209 emulations routines, these routines will issue floating-point
12210 operations. If you are compiling for an Alpha without floating-point
12211 operations, you must ensure that the library is built so as not to call
12214 Note that Alpha implementations without floating-point operations are
12215 required to have floating-point registers.
12218 @itemx -mno-fp-regs
12220 @opindex mno-fp-regs
12221 Generate code that uses (does not use) the floating-point register set.
12222 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
12223 register set is not used, floating-point operands are passed in integer
12224 registers as if they were integers and floating-point results are passed
12225 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
12226 so any function with a floating-point argument or return value called by code
12227 compiled with @option{-mno-fp-regs} must also be compiled with that
12230 A typical use of this option is building a kernel that does not use,
12231 and hence need not save and restore, any floating-point registers.
12235 The Alpha architecture implements floating-point hardware optimized for
12236 maximum performance. It is mostly compliant with the IEEE floating-point
12237 standard. However, for full compliance, software assistance is
12238 required. This option generates code fully IEEE-compliant code
12239 @emph{except} that the @var{inexact-flag} is not maintained (see below).
12240 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
12241 defined during compilation. The resulting code is less efficient but is
12242 able to correctly support denormalized numbers and exceptional IEEE
12243 values such as not-a-number and plus/minus infinity. Other Alpha
12244 compilers call this option @option{-ieee_with_no_inexact}.
12246 @item -mieee-with-inexact
12247 @opindex mieee-with-inexact
12248 This is like @option{-mieee} except the generated code also maintains
12249 the IEEE @var{inexact-flag}. Turning on this option causes the
12250 generated code to implement fully-compliant IEEE math. In addition to
12251 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
12252 macro. On some Alpha implementations the resulting code may execute
12253 significantly slower than the code generated by default. Since there is
12254 very little code that depends on the @var{inexact-flag}, you should
12255 normally not specify this option. Other Alpha compilers call this
12256 option @option{-ieee_with_inexact}.
12258 @item -mfp-trap-mode=@var{trap-mode}
12259 @opindex mfp-trap-mode
12260 This option controls what floating-point related traps are enabled.
12261 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
12262 The trap mode can be set to one of four values:
12266 This is the default (normal) setting. The only traps that are enabled
12267 are the ones that cannot be disabled in software (e.g., division by zero
12271 In addition to the traps enabled by @samp{n}, underflow traps are enabled
12275 Like @samp{u}, but the instructions are marked to be safe for software
12276 completion (see Alpha architecture manual for details).
12279 Like @samp{su}, but inexact traps are enabled as well.
12282 @item -mfp-rounding-mode=@var{rounding-mode}
12283 @opindex mfp-rounding-mode
12284 Selects the IEEE rounding mode. Other Alpha compilers call this option
12285 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
12290 Normal IEEE rounding mode. Floating-point numbers are rounded towards
12291 the nearest machine number or towards the even machine number in case
12295 Round towards minus infinity.
12298 Chopped rounding mode. Floating-point numbers are rounded towards zero.
12301 Dynamic rounding mode. A field in the floating-point control register
12302 (@var{fpcr}, see Alpha architecture reference manual) controls the
12303 rounding mode in effect. The C library initializes this register for
12304 rounding towards plus infinity. Thus, unless your program modifies the
12305 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
12308 @item -mtrap-precision=@var{trap-precision}
12309 @opindex mtrap-precision
12310 In the Alpha architecture, floating-point traps are imprecise. This
12311 means without software assistance it is impossible to recover from a
12312 floating trap and program execution normally needs to be terminated.
12313 GCC can generate code that can assist operating system trap handlers
12314 in determining the exact location that caused a floating-point trap.
12315 Depending on the requirements of an application, different levels of
12316 precisions can be selected:
12320 Program precision. This option is the default and means a trap handler
12321 can only identify which program caused a floating-point exception.
12324 Function precision. The trap handler can determine the function that
12325 caused a floating-point exception.
12328 Instruction precision. The trap handler can determine the exact
12329 instruction that caused a floating-point exception.
12332 Other Alpha compilers provide the equivalent options called
12333 @option{-scope_safe} and @option{-resumption_safe}.
12335 @item -mieee-conformant
12336 @opindex mieee-conformant
12337 This option marks the generated code as IEEE conformant. You must not
12338 use this option unless you also specify @option{-mtrap-precision=i} and either
12339 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
12340 is to emit the line @samp{.eflag 48} in the function prologue of the
12341 generated assembly file. Under DEC Unix, this has the effect that
12342 IEEE-conformant math library routines will be linked in.
12344 @item -mbuild-constants
12345 @opindex mbuild-constants
12346 Normally GCC examines a 32- or 64-bit integer constant to
12347 see if it can construct it from smaller constants in two or three
12348 instructions. If it cannot, it will output the constant as a literal and
12349 generate code to load it from the data segment at run time.
12351 Use this option to require GCC to construct @emph{all} integer constants
12352 using code, even if it takes more instructions (the maximum is six).
12354 You would typically use this option to build a shared library dynamic
12355 loader. Itself a shared library, it must relocate itself in memory
12356 before it can find the variables and constants in its own data segment.
12362 Select whether to generate code to be assembled by the vendor-supplied
12363 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
12381 Indicate whether GCC should generate code to use the optional BWX,
12382 CIX, FIX and MAX instruction sets. The default is to use the instruction
12383 sets supported by the CPU type specified via @option{-mcpu=} option or that
12384 of the CPU on which GCC was built if none was specified.
12387 @itemx -mfloat-ieee
12388 @opindex mfloat-vax
12389 @opindex mfloat-ieee
12390 Generate code that uses (does not use) VAX F and G floating-point
12391 arithmetic instead of IEEE single and double precision.
12393 @item -mexplicit-relocs
12394 @itemx -mno-explicit-relocs
12395 @opindex mexplicit-relocs
12396 @opindex mno-explicit-relocs
12397 Older Alpha assemblers provided no way to generate symbol relocations
12398 except via assembler macros. Use of these macros does not allow
12399 optimal instruction scheduling. GNU binutils as of version 2.12
12400 supports a new syntax that allows the compiler to explicitly mark
12401 which relocations should apply to which instructions. This option
12402 is mostly useful for debugging, as GCC detects the capabilities of
12403 the assembler when it is built and sets the default accordingly.
12406 @itemx -mlarge-data
12407 @opindex msmall-data
12408 @opindex mlarge-data
12409 When @option{-mexplicit-relocs} is in effect, static data is
12410 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
12411 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
12412 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
12413 16-bit relocations off of the @code{$gp} register. This limits the
12414 size of the small data area to 64KB, but allows the variables to be
12415 directly accessed via a single instruction.
12417 The default is @option{-mlarge-data}. With this option the data area
12418 is limited to just below 2GB@. Programs that require more than 2GB of
12419 data must use @code{malloc} or @code{mmap} to allocate the data in the
12420 heap instead of in the program's data segment.
12422 When generating code for shared libraries, @option{-fpic} implies
12423 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
12426 @itemx -mlarge-text
12427 @opindex msmall-text
12428 @opindex mlarge-text
12429 When @option{-msmall-text} is used, the compiler assumes that the
12430 code of the entire program (or shared library) fits in 4MB, and is
12431 thus reachable with a branch instruction. When @option{-msmall-data}
12432 is used, the compiler can assume that all local symbols share the
12433 same @code{$gp} value, and thus reduce the number of instructions
12434 required for a function call from 4 to 1.
12436 The default is @option{-mlarge-text}.
12438 @item -mcpu=@var{cpu_type}
12440 Set the instruction set and instruction scheduling parameters for
12441 machine type @var{cpu_type}. You can specify either the @samp{EV}
12442 style name or the corresponding chip number. GCC supports scheduling
12443 parameters for the EV4, EV5 and EV6 family of processors and will
12444 choose the default values for the instruction set from the processor
12445 you specify. If you do not specify a processor type, GCC will default
12446 to the processor on which the compiler was built.
12448 Supported values for @var{cpu_type} are
12454 Schedules as an EV4 and has no instruction set extensions.
12458 Schedules as an EV5 and has no instruction set extensions.
12462 Schedules as an EV5 and supports the BWX extension.
12467 Schedules as an EV5 and supports the BWX and MAX extensions.
12471 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
12475 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
12478 Native toolchains also support the value @samp{native},
12479 which selects the best architecture option for the host processor.
12480 @option{-mcpu=native} has no effect if GCC does not recognize
12483 @item -mtune=@var{cpu_type}
12485 Set only the instruction scheduling parameters for machine type
12486 @var{cpu_type}. The instruction set is not changed.
12488 Native toolchains also support the value @samp{native},
12489 which selects the best architecture option for the host processor.
12490 @option{-mtune=native} has no effect if GCC does not recognize
12493 @item -mmemory-latency=@var{time}
12494 @opindex mmemory-latency
12495 Sets the latency the scheduler should assume for typical memory
12496 references as seen by the application. This number is highly
12497 dependent on the memory access patterns used by the application
12498 and the size of the external cache on the machine.
12500 Valid options for @var{time} are
12504 A decimal number representing clock cycles.
12510 The compiler contains estimates of the number of clock cycles for
12511 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
12512 (also called Dcache, Scache, and Bcache), as well as to main memory.
12513 Note that L3 is only valid for EV5.
12518 @node DEC Alpha/VMS Options
12519 @subsection DEC Alpha/VMS Options
12521 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
12524 @item -mvms-return-codes
12525 @opindex mvms-return-codes
12526 Return VMS condition codes from main. The default is to return POSIX
12527 style condition (e.g.@: error) codes.
12529 @item -mdebug-main=@var{prefix}
12530 @opindex mdebug-main=@var{prefix}
12531 Flag the first routine whose name starts with @var{prefix} as the main
12532 routine for the debugger.
12536 Default to 64-bit memory allocation routines.
12540 @subsection FR30 Options
12541 @cindex FR30 Options
12543 These options are defined specifically for the FR30 port.
12547 @item -msmall-model
12548 @opindex msmall-model
12549 Use the small address space model. This can produce smaller code, but
12550 it does assume that all symbolic values and addresses will fit into a
12555 Assume that runtime support has been provided and so there is no need
12556 to include the simulator library (@file{libsim.a}) on the linker
12562 @subsection FRV Options
12563 @cindex FRV Options
12569 Only use the first 32 general-purpose registers.
12574 Use all 64 general-purpose registers.
12579 Use only the first 32 floating-point registers.
12584 Use all 64 floating-point registers.
12587 @opindex mhard-float
12589 Use hardware instructions for floating-point operations.
12592 @opindex msoft-float
12594 Use library routines for floating-point operations.
12599 Dynamically allocate condition code registers.
12604 Do not try to dynamically allocate condition code registers, only
12605 use @code{icc0} and @code{fcc0}.
12610 Change ABI to use double word insns.
12615 Do not use double word instructions.
12620 Use floating-point double instructions.
12623 @opindex mno-double
12625 Do not use floating-point double instructions.
12630 Use media instructions.
12635 Do not use media instructions.
12640 Use multiply and add/subtract instructions.
12643 @opindex mno-muladd
12645 Do not use multiply and add/subtract instructions.
12650 Select the FDPIC ABI, which uses function descriptors to represent
12651 pointers to functions. Without any PIC/PIE-related options, it
12652 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
12653 assumes GOT entries and small data are within a 12-bit range from the
12654 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
12655 are computed with 32 bits.
12656 With a @samp{bfin-elf} target, this option implies @option{-msim}.
12659 @opindex minline-plt
12661 Enable inlining of PLT entries in function calls to functions that are
12662 not known to bind locally. It has no effect without @option{-mfdpic}.
12663 It's enabled by default if optimizing for speed and compiling for
12664 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
12665 optimization option such as @option{-O3} or above is present in the
12671 Assume a large TLS segment when generating thread-local code.
12676 Do not assume a large TLS segment when generating thread-local code.
12681 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
12682 that is known to be in read-only sections. It's enabled by default,
12683 except for @option{-fpic} or @option{-fpie}: even though it may help
12684 make the global offset table smaller, it trades 1 instruction for 4.
12685 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
12686 one of which may be shared by multiple symbols, and it avoids the need
12687 for a GOT entry for the referenced symbol, so it's more likely to be a
12688 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
12690 @item -multilib-library-pic
12691 @opindex multilib-library-pic
12693 Link with the (library, not FD) pic libraries. It's implied by
12694 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
12695 @option{-fpic} without @option{-mfdpic}. You should never have to use
12699 @opindex mlinked-fp
12701 Follow the EABI requirement of always creating a frame pointer whenever
12702 a stack frame is allocated. This option is enabled by default and can
12703 be disabled with @option{-mno-linked-fp}.
12706 @opindex mlong-calls
12708 Use indirect addressing to call functions outside the current
12709 compilation unit. This allows the functions to be placed anywhere
12710 within the 32-bit address space.
12712 @item -malign-labels
12713 @opindex malign-labels
12715 Try to align labels to an 8-byte boundary by inserting nops into the
12716 previous packet. This option only has an effect when VLIW packing
12717 is enabled. It doesn't create new packets; it merely adds nops to
12720 @item -mlibrary-pic
12721 @opindex mlibrary-pic
12723 Generate position-independent EABI code.
12728 Use only the first four media accumulator registers.
12733 Use all eight media accumulator registers.
12738 Pack VLIW instructions.
12743 Do not pack VLIW instructions.
12746 @opindex mno-eflags
12748 Do not mark ABI switches in e_flags.
12751 @opindex mcond-move
12753 Enable the use of conditional-move instructions (default).
12755 This switch is mainly for debugging the compiler and will likely be removed
12756 in a future version.
12758 @item -mno-cond-move
12759 @opindex mno-cond-move
12761 Disable the use of conditional-move instructions.
12763 This switch is mainly for debugging the compiler and will likely be removed
12764 in a future version.
12769 Enable the use of conditional set instructions (default).
12771 This switch is mainly for debugging the compiler and will likely be removed
12772 in a future version.
12777 Disable the use of conditional set instructions.
12779 This switch is mainly for debugging the compiler and will likely be removed
12780 in a future version.
12783 @opindex mcond-exec
12785 Enable the use of conditional execution (default).
12787 This switch is mainly for debugging the compiler and will likely be removed
12788 in a future version.
12790 @item -mno-cond-exec
12791 @opindex mno-cond-exec
12793 Disable the use of conditional execution.
12795 This switch is mainly for debugging the compiler and will likely be removed
12796 in a future version.
12798 @item -mvliw-branch
12799 @opindex mvliw-branch
12801 Run a pass to pack branches into VLIW instructions (default).
12803 This switch is mainly for debugging the compiler and will likely be removed
12804 in a future version.
12806 @item -mno-vliw-branch
12807 @opindex mno-vliw-branch
12809 Do not run a pass to pack branches into VLIW instructions.
12811 This switch is mainly for debugging the compiler and will likely be removed
12812 in a future version.
12814 @item -mmulti-cond-exec
12815 @opindex mmulti-cond-exec
12817 Enable optimization of @code{&&} and @code{||} in conditional execution
12820 This switch is mainly for debugging the compiler and will likely be removed
12821 in a future version.
12823 @item -mno-multi-cond-exec
12824 @opindex mno-multi-cond-exec
12826 Disable optimization of @code{&&} and @code{||} in conditional execution.
12828 This switch is mainly for debugging the compiler and will likely be removed
12829 in a future version.
12831 @item -mnested-cond-exec
12832 @opindex mnested-cond-exec
12834 Enable nested conditional execution optimizations (default).
12836 This switch is mainly for debugging the compiler and will likely be removed
12837 in a future version.
12839 @item -mno-nested-cond-exec
12840 @opindex mno-nested-cond-exec
12842 Disable nested conditional execution optimizations.
12844 This switch is mainly for debugging the compiler and will likely be removed
12845 in a future version.
12847 @item -moptimize-membar
12848 @opindex moptimize-membar
12850 This switch removes redundant @code{membar} instructions from the
12851 compiler generated code. It is enabled by default.
12853 @item -mno-optimize-membar
12854 @opindex mno-optimize-membar
12856 This switch disables the automatic removal of redundant @code{membar}
12857 instructions from the generated code.
12859 @item -mtomcat-stats
12860 @opindex mtomcat-stats
12862 Cause gas to print out tomcat statistics.
12864 @item -mcpu=@var{cpu}
12867 Select the processor type for which to generate code. Possible values are
12868 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12869 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12873 @node GNU/Linux Options
12874 @subsection GNU/Linux Options
12876 These @samp{-m} options are defined for GNU/Linux targets:
12881 Use the GNU C library. This is the default except
12882 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12886 Use uClibc C library. This is the default on
12887 @samp{*-*-linux-*uclibc*} targets.
12891 Use Bionic C library. This is the default on
12892 @samp{*-*-linux-*android*} targets.
12896 Compile code compatible with Android platform. This is the default on
12897 @samp{*-*-linux-*android*} targets.
12899 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12900 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12901 this option makes the GCC driver pass Android-specific options to the linker.
12902 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12905 @item -tno-android-cc
12906 @opindex tno-android-cc
12907 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12908 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12909 @option{-fno-rtti} by default.
12911 @item -tno-android-ld
12912 @opindex tno-android-ld
12913 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12914 linking options to the linker.
12918 @node H8/300 Options
12919 @subsection H8/300 Options
12921 These @samp{-m} options are defined for the H8/300 implementations:
12926 Shorten some address references at link time, when possible; uses the
12927 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12928 ld, Using ld}, for a fuller description.
12932 Generate code for the H8/300H@.
12936 Generate code for the H8S@.
12940 Generate code for the H8S and H8/300H in the normal mode. This switch
12941 must be used either with @option{-mh} or @option{-ms}.
12945 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12949 Make @code{int} data 32 bits by default.
12952 @opindex malign-300
12953 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12954 The default for the H8/300H and H8S is to align longs and floats on
12956 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
12957 This option has no effect on the H8/300.
12961 @subsection HPPA Options
12962 @cindex HPPA Options
12964 These @samp{-m} options are defined for the HPPA family of computers:
12967 @item -march=@var{architecture-type}
12969 Generate code for the specified architecture. The choices for
12970 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12971 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12972 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12973 architecture option for your machine. Code compiled for lower numbered
12974 architectures will run on higher numbered architectures, but not the
12977 @item -mpa-risc-1-0
12978 @itemx -mpa-risc-1-1
12979 @itemx -mpa-risc-2-0
12980 @opindex mpa-risc-1-0
12981 @opindex mpa-risc-1-1
12982 @opindex mpa-risc-2-0
12983 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12986 @opindex mbig-switch
12987 Generate code suitable for big switch tables. Use this option only if
12988 the assembler/linker complain about out of range branches within a switch
12991 @item -mjump-in-delay
12992 @opindex mjump-in-delay
12993 Fill delay slots of function calls with unconditional jump instructions
12994 by modifying the return pointer for the function call to be the target
12995 of the conditional jump.
12997 @item -mdisable-fpregs
12998 @opindex mdisable-fpregs
12999 Prevent floating-point registers from being used in any manner. This is
13000 necessary for compiling kernels that perform lazy context switching of
13001 floating-point registers. If you use this option and attempt to perform
13002 floating-point operations, the compiler aborts.
13004 @item -mdisable-indexing
13005 @opindex mdisable-indexing
13006 Prevent the compiler from using indexing address modes. This avoids some
13007 rather obscure problems when compiling MIG generated code under MACH@.
13009 @item -mno-space-regs
13010 @opindex mno-space-regs
13011 Generate code that assumes the target has no space registers. This allows
13012 GCC to generate faster indirect calls and use unscaled index address modes.
13014 Such code is suitable for level 0 PA systems and kernels.
13016 @item -mfast-indirect-calls
13017 @opindex mfast-indirect-calls
13018 Generate code that assumes calls never cross space boundaries. This
13019 allows GCC to emit code that performs faster indirect calls.
13021 This option will not work in the presence of shared libraries or nested
13024 @item -mfixed-range=@var{register-range}
13025 @opindex mfixed-range
13026 Generate code treating the given register range as fixed registers.
13027 A fixed register is one that the register allocator can not use. This is
13028 useful when compiling kernel code. A register range is specified as
13029 two registers separated by a dash. Multiple register ranges can be
13030 specified separated by a comma.
13032 @item -mlong-load-store
13033 @opindex mlong-load-store
13034 Generate 3-instruction load and store sequences as sometimes required by
13035 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
13038 @item -mportable-runtime
13039 @opindex mportable-runtime
13040 Use the portable calling conventions proposed by HP for ELF systems.
13044 Enable the use of assembler directives only GAS understands.
13046 @item -mschedule=@var{cpu-type}
13048 Schedule code according to the constraints for the machine type
13049 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
13050 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
13051 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
13052 proper scheduling option for your machine. The default scheduling is
13056 @opindex mlinker-opt
13057 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
13058 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
13059 linkers in which they give bogus error messages when linking some programs.
13062 @opindex msoft-float
13063 Generate output containing library calls for floating point.
13064 @strong{Warning:} the requisite libraries are not available for all HPPA
13065 targets. Normally the facilities of the machine's usual C compiler are
13066 used, but this cannot be done directly in cross-compilation. You must make
13067 your own arrangements to provide suitable library functions for
13070 @option{-msoft-float} changes the calling convention in the output file;
13071 therefore, it is only useful if you compile @emph{all} of a program with
13072 this option. In particular, you need to compile @file{libgcc.a}, the
13073 library that comes with GCC, with @option{-msoft-float} in order for
13078 Generate the predefine, @code{_SIO}, for server IO@. The default is
13079 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
13080 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
13081 options are available under HP-UX and HI-UX@.
13085 Use GNU ld specific options. This passes @option{-shared} to ld when
13086 building a shared library. It is the default when GCC is configured,
13087 explicitly or implicitly, with the GNU linker. This option does not
13088 have any affect on which ld is called, it only changes what parameters
13089 are passed to that ld. The ld that is called is determined by the
13090 @option{--with-ld} configure option, GCC's program search path, and
13091 finally by the user's @env{PATH}. The linker used by GCC can be printed
13092 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
13093 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
13097 Use HP ld specific options. This passes @option{-b} to ld when building
13098 a shared library and passes @option{+Accept TypeMismatch} to ld on all
13099 links. It is the default when GCC is configured, explicitly or
13100 implicitly, with the HP linker. This option does not have any affect on
13101 which ld is called, it only changes what parameters are passed to that
13102 ld. The ld that is called is determined by the @option{--with-ld}
13103 configure option, GCC's program search path, and finally by the user's
13104 @env{PATH}. The linker used by GCC can be printed using @samp{which
13105 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
13106 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
13109 @opindex mno-long-calls
13110 Generate code that uses long call sequences. This ensures that a call
13111 is always able to reach linker generated stubs. The default is to generate
13112 long calls only when the distance from the call site to the beginning
13113 of the function or translation unit, as the case may be, exceeds a
13114 predefined limit set by the branch type being used. The limits for
13115 normal calls are 7,600,000 and 240,000 bytes, respectively for the
13116 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
13119 Distances are measured from the beginning of functions when using the
13120 @option{-ffunction-sections} option, or when using the @option{-mgas}
13121 and @option{-mno-portable-runtime} options together under HP-UX with
13124 It is normally not desirable to use this option as it will degrade
13125 performance. However, it may be useful in large applications,
13126 particularly when partial linking is used to build the application.
13128 The types of long calls used depends on the capabilities of the
13129 assembler and linker, and the type of code being generated. The
13130 impact on systems that support long absolute calls, and long pic
13131 symbol-difference or pc-relative calls should be relatively small.
13132 However, an indirect call is used on 32-bit ELF systems in pic code
13133 and it is quite long.
13135 @item -munix=@var{unix-std}
13137 Generate compiler predefines and select a startfile for the specified
13138 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
13139 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
13140 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
13141 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
13142 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
13145 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
13146 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
13147 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
13148 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
13149 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
13150 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
13152 It is @emph{important} to note that this option changes the interfaces
13153 for various library routines. It also affects the operational behavior
13154 of the C library. Thus, @emph{extreme} care is needed in using this
13157 Library code that is intended to operate with more than one UNIX
13158 standard must test, set and restore the variable @var{__xpg4_extended_mask}
13159 as appropriate. Most GNU software doesn't provide this capability.
13163 Suppress the generation of link options to search libdld.sl when the
13164 @option{-static} option is specified on HP-UX 10 and later.
13168 The HP-UX implementation of setlocale in libc has a dependency on
13169 libdld.sl. There isn't an archive version of libdld.sl. Thus,
13170 when the @option{-static} option is specified, special link options
13171 are needed to resolve this dependency.
13173 On HP-UX 10 and later, the GCC driver adds the necessary options to
13174 link with libdld.sl when the @option{-static} option is specified.
13175 This causes the resulting binary to be dynamic. On the 64-bit port,
13176 the linkers generate dynamic binaries by default in any case. The
13177 @option{-nolibdld} option can be used to prevent the GCC driver from
13178 adding these link options.
13182 Add support for multithreading with the @dfn{dce thread} library
13183 under HP-UX@. This option sets flags for both the preprocessor and
13187 @node i386 and x86-64 Options
13188 @subsection Intel 386 and AMD x86-64 Options
13189 @cindex i386 Options
13190 @cindex x86-64 Options
13191 @cindex Intel 386 Options
13192 @cindex AMD x86-64 Options
13194 These @samp{-m} options are defined for the i386 and x86-64 family of
13198 @item -mtune=@var{cpu-type}
13200 Tune to @var{cpu-type} everything applicable about the generated code, except
13201 for the ABI and the set of available instructions. The choices for
13202 @var{cpu-type} are:
13205 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
13206 If you know the CPU on which your code will run, then you should use
13207 the corresponding @option{-mtune} option instead of
13208 @option{-mtune=generic}. But, if you do not know exactly what CPU users
13209 of your application will have, then you should use this option.
13211 As new processors are deployed in the marketplace, the behavior of this
13212 option will change. Therefore, if you upgrade to a newer version of
13213 GCC, the code generated option will change to reflect the processors
13214 that were most common when that version of GCC was released.
13216 There is no @option{-march=generic} option because @option{-march}
13217 indicates the instruction set the compiler can use, and there is no
13218 generic instruction set applicable to all processors. In contrast,
13219 @option{-mtune} indicates the processor (or, in this case, collection of
13220 processors) for which the code is optimized.
13222 This selects the CPU to tune for at compilation time by determining
13223 the processor type of the compiling machine. Using @option{-mtune=native}
13224 will produce code optimized for the local machine under the constraints
13225 of the selected instruction set. Using @option{-march=native} will
13226 enable all instruction subsets supported by the local machine (hence
13227 the result might not run on different machines).
13229 Original Intel's i386 CPU@.
13231 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
13232 @item i586, pentium
13233 Intel Pentium CPU with no MMX support.
13235 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
13237 Intel PentiumPro CPU@.
13239 Same as @code{generic}, but when used as @code{march} option, PentiumPro
13240 instruction set will be used, so the code will run on all i686 family chips.
13242 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
13243 @item pentium3, pentium3m
13244 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
13247 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
13248 support. Used by Centrino notebooks.
13249 @item pentium4, pentium4m
13250 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
13252 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
13255 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
13256 SSE2 and SSE3 instruction set support.
13258 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13259 instruction set support.
13261 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
13262 and SSE4.2 instruction set support.
13264 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13265 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
13267 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
13268 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
13271 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
13272 instruction set support.
13274 AMD K6 CPU with MMX instruction set support.
13276 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
13277 @item athlon, athlon-tbird
13278 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
13280 @item athlon-4, athlon-xp, athlon-mp
13281 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
13282 instruction set support.
13283 @item k8, opteron, athlon64, athlon-fx
13284 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
13285 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
13286 @item k8-sse3, opteron-sse3, athlon64-sse3
13287 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
13288 @item amdfam10, barcelona
13289 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
13290 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
13291 instruction set extensions.)
13293 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
13294 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
13295 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
13297 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
13298 supersets BMI, TBM, F16C, FMA, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE,
13299 SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
13302 AMD Family 14h core based CPUs with x86-64 instruction set support. (This
13303 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
13304 instruction set extensions.)
13306 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
13309 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
13310 instruction set support.
13312 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
13313 implemented for this chip.)
13315 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
13316 implemented for this chip.)
13318 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
13321 While picking a specific @var{cpu-type} will schedule things appropriately
13322 for that particular chip, the compiler will not generate any code that
13323 does not run on the default machine type without the @option{-march=@var{cpu-type}}
13324 option being used. For example, if GCC is configured for i686-pc-linux-gnu
13325 then @option{-mtune=pentium4} will generate code that is tuned for Pentium4
13326 but will still run on i686 machines.
13328 @item -march=@var{cpu-type}
13330 Generate instructions for the machine type @var{cpu-type}. The choices
13331 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
13332 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
13334 @item -mcpu=@var{cpu-type}
13336 A deprecated synonym for @option{-mtune}.
13338 @item -mfpmath=@var{unit}
13340 Generate floating-point arithmetic for selected unit @var{unit}. The choices
13341 for @var{unit} are:
13345 Use the standard 387 floating-point coprocessor present on the majority of chips and
13346 emulated otherwise. Code compiled with this option runs almost everywhere.
13347 The temporary results are computed in 80-bit precision instead of the precision
13348 specified by the type, resulting in slightly different results compared to most
13349 of other chips. See @option{-ffloat-store} for more detailed description.
13351 This is the default choice for i386 compiler.
13354 Use scalar floating-point instructions present in the SSE instruction set.
13355 This instruction set is supported by Pentium3 and newer chips, in the AMD line
13356 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
13357 instruction set supports only single-precision arithmetic, thus the double and
13358 extended-precision arithmetic are still done using 387. A later version, present
13359 only in Pentium4 and the future AMD x86-64 chips, supports double-precision
13362 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
13363 or @option{-msse2} switches to enable SSE extensions and make this option
13364 effective. For the x86-64 compiler, these extensions are enabled by default.
13366 The resulting code should be considerably faster in the majority of cases and avoid
13367 the numerical instability problems of 387 code, but may break some existing
13368 code that expects temporaries to be 80 bits.
13370 This is the default choice for the x86-64 compiler.
13375 Attempt to utilize both instruction sets at once. This effectively double the
13376 amount of available registers and on chips with separate execution units for
13377 387 and SSE the execution resources too. Use this option with care, as it is
13378 still experimental, because the GCC register allocator does not model separate
13379 functional units well resulting in instable performance.
13382 @item -masm=@var{dialect}
13383 @opindex masm=@var{dialect}
13384 Output asm instructions using selected @var{dialect}. Supported
13385 choices are @samp{intel} or @samp{att} (the default one). Darwin does
13386 not support @samp{intel}.
13389 @itemx -mno-ieee-fp
13391 @opindex mno-ieee-fp
13392 Control whether or not the compiler uses IEEE floating-point
13393 comparisons. These handle correctly the case where the result of a
13394 comparison is unordered.
13397 @opindex msoft-float
13398 Generate output containing library calls for floating point.
13399 @strong{Warning:} the requisite libraries are not part of GCC@.
13400 Normally the facilities of the machine's usual C compiler are used, but
13401 this can't be done directly in cross-compilation. You must make your
13402 own arrangements to provide suitable library functions for
13405 On machines where a function returns floating-point results in the 80387
13406 register stack, some floating-point opcodes may be emitted even if
13407 @option{-msoft-float} is used.
13409 @item -mno-fp-ret-in-387
13410 @opindex mno-fp-ret-in-387
13411 Do not use the FPU registers for return values of functions.
13413 The usual calling convention has functions return values of types
13414 @code{float} and @code{double} in an FPU register, even if there
13415 is no FPU@. The idea is that the operating system should emulate
13418 The option @option{-mno-fp-ret-in-387} causes such values to be returned
13419 in ordinary CPU registers instead.
13421 @item -mno-fancy-math-387
13422 @opindex mno-fancy-math-387
13423 Some 387 emulators do not support the @code{sin}, @code{cos} and
13424 @code{sqrt} instructions for the 387. Specify this option to avoid
13425 generating those instructions. This option is the default on FreeBSD,
13426 OpenBSD and NetBSD@. This option is overridden when @option{-march}
13427 indicates that the target CPU will always have an FPU and so the
13428 instruction will not need emulation. As of revision 2.6.1, these
13429 instructions are not generated unless you also use the
13430 @option{-funsafe-math-optimizations} switch.
13432 @item -malign-double
13433 @itemx -mno-align-double
13434 @opindex malign-double
13435 @opindex mno-align-double
13436 Control whether GCC aligns @code{double}, @code{long double}, and
13437 @code{long long} variables on a two-word boundary or a one-word
13438 boundary. Aligning @code{double} variables on a two-word boundary
13439 produces code that runs somewhat faster on a @samp{Pentium} at the
13440 expense of more memory.
13442 On x86-64, @option{-malign-double} is enabled by default.
13444 @strong{Warning:} if you use the @option{-malign-double} switch,
13445 structures containing the above types will be aligned differently than
13446 the published application binary interface specifications for the 386
13447 and will not be binary compatible with structures in code compiled
13448 without that switch.
13450 @item -m96bit-long-double
13451 @itemx -m128bit-long-double
13452 @opindex m96bit-long-double
13453 @opindex m128bit-long-double
13454 These switches control the size of @code{long double} type. The i386
13455 application binary interface specifies the size to be 96 bits,
13456 so @option{-m96bit-long-double} is the default in 32-bit mode.
13458 Modern architectures (Pentium and newer) prefer @code{long double}
13459 to be aligned to an 8- or 16-byte boundary. In arrays or structures
13460 conforming to the ABI, this is not possible. So specifying
13461 @option{-m128bit-long-double} aligns @code{long double}
13462 to a 16-byte boundary by padding the @code{long double} with an additional
13465 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
13466 its ABI specifies that @code{long double} is to be aligned on 16-byte boundary.
13468 Notice that neither of these options enable any extra precision over the x87
13469 standard of 80 bits for a @code{long double}.
13471 @strong{Warning:} if you override the default value for your target ABI, the
13472 structures and arrays containing @code{long double} variables will change
13473 their size as well as function calling convention for function taking
13474 @code{long double} will be modified. Hence they will not be binary
13475 compatible with arrays or structures in code compiled without that switch.
13477 @item -mlarge-data-threshold=@var{number}
13478 @opindex mlarge-data-threshold=@var{number}
13479 When @option{-mcmodel=medium} is specified, the data greater than
13480 @var{threshold} are placed in large data section. This value must be the
13481 same across all object linked into the binary and defaults to 65535.
13485 Use a different function-calling convention, in which functions that
13486 take a fixed number of arguments return with the @code{ret} @var{num}
13487 instruction, which pops their arguments while returning. This saves one
13488 instruction in the caller since there is no need to pop the arguments
13491 You can specify that an individual function is called with this calling
13492 sequence with the function attribute @samp{stdcall}. You can also
13493 override the @option{-mrtd} option by using the function attribute
13494 @samp{cdecl}. @xref{Function Attributes}.
13496 @strong{Warning:} this calling convention is incompatible with the one
13497 normally used on Unix, so you cannot use it if you need to call
13498 libraries compiled with the Unix compiler.
13500 Also, you must provide function prototypes for all functions that
13501 take variable numbers of arguments (including @code{printf});
13502 otherwise incorrect code will be generated for calls to those
13505 In addition, seriously incorrect code will result if you call a
13506 function with too many arguments. (Normally, extra arguments are
13507 harmlessly ignored.)
13509 @item -mregparm=@var{num}
13511 Control how many registers are used to pass integer arguments. By
13512 default, no registers are used to pass arguments, and at most 3
13513 registers can be used. You can control this behavior for a specific
13514 function by using the function attribute @samp{regparm}.
13515 @xref{Function Attributes}.
13517 @strong{Warning:} if you use this switch, and
13518 @var{num} is nonzero, then you must build all modules with the same
13519 value, including any libraries. This includes the system libraries and
13523 @opindex msseregparm
13524 Use SSE register passing conventions for float and double arguments
13525 and return values. You can control this behavior for a specific
13526 function by using the function attribute @samp{sseregparm}.
13527 @xref{Function Attributes}.
13529 @strong{Warning:} if you use this switch then you must build all
13530 modules with the same value, including any libraries. This includes
13531 the system libraries and startup modules.
13533 @item -mvect8-ret-in-mem
13534 @opindex mvect8-ret-in-mem
13535 Return 8-byte vectors in memory instead of MMX registers. This is the
13536 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
13537 Studio compilers until version 12. Later compiler versions (starting
13538 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
13539 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
13540 you need to remain compatible with existing code produced by those
13541 previous compiler versions or older versions of GCC.
13550 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
13551 is specified, the significands of results of floating-point operations are
13552 rounded to 24 bits (single precision); @option{-mpc64} rounds the
13553 significands of results of floating-point operations to 53 bits (double
13554 precision) and @option{-mpc80} rounds the significands of results of
13555 floating-point operations to 64 bits (extended double precision), which is
13556 the default. When this option is used, floating-point operations in higher
13557 precisions are not available to the programmer without setting the FPU
13558 control word explicitly.
13560 Setting the rounding of floating-point operations to less than the default
13561 80 bits can speed some programs by 2% or more. Note that some mathematical
13562 libraries assume that extended-precision (80-bit) floating-point operations
13563 are enabled by default; routines in such libraries could suffer significant
13564 loss of accuracy, typically through so-called "catastrophic cancellation",
13565 when this option is used to set the precision to less than extended precision.
13567 @item -mstackrealign
13568 @opindex mstackrealign
13569 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
13570 option will generate an alternate prologue and epilogue that realigns the
13571 run-time stack if necessary. This supports mixing legacy codes that keep
13572 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
13573 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
13574 applicable to individual functions.
13576 @item -mpreferred-stack-boundary=@var{num}
13577 @opindex mpreferred-stack-boundary
13578 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
13579 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
13580 the default is 4 (16 bytes or 128 bits).
13582 @item -mincoming-stack-boundary=@var{num}
13583 @opindex mincoming-stack-boundary
13584 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
13585 boundary. If @option{-mincoming-stack-boundary} is not specified,
13586 the one specified by @option{-mpreferred-stack-boundary} will be used.
13588 On Pentium and PentiumPro, @code{double} and @code{long double} values
13589 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
13590 suffer significant run time performance penalties. On Pentium III, the
13591 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
13592 properly if it is not 16-byte aligned.
13594 To ensure proper alignment of this values on the stack, the stack boundary
13595 must be as aligned as that required by any value stored on the stack.
13596 Further, every function must be generated such that it keeps the stack
13597 aligned. Thus calling a function compiled with a higher preferred
13598 stack boundary from a function compiled with a lower preferred stack
13599 boundary will most likely misalign the stack. It is recommended that
13600 libraries that use callbacks always use the default setting.
13602 This extra alignment does consume extra stack space, and generally
13603 increases code size. Code that is sensitive to stack space usage, such
13604 as embedded systems and operating system kernels, may want to reduce the
13605 preferred alignment to @option{-mpreferred-stack-boundary=2}.
13634 @itemx -mno-fsgsbase
13670 These switches enable or disable the use of instructions in the MMX, SSE,
13671 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
13672 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
13673 @: extended instruction sets.
13674 These extensions are also available as built-in functions: see
13675 @ref{X86 Built-in Functions}, for details of the functions enabled and
13676 disabled by these switches.
13678 To have SSE/SSE2 instructions generated automatically from floating-point
13679 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
13681 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
13682 generates new AVX instructions or AVX equivalence for all SSEx instructions
13685 These options will enable GCC to use these extended instructions in
13686 generated code, even without @option{-mfpmath=sse}. Applications that
13687 perform run-time CPU detection must compile separate files for each
13688 supported architecture, using the appropriate flags. In particular,
13689 the file containing the CPU detection code should be compiled without
13694 This option instructs GCC to emit a @code{cld} instruction in the prologue
13695 of functions that use string instructions. String instructions depend on
13696 the DF flag to select between autoincrement or autodecrement mode. While the
13697 ABI specifies the DF flag to be cleared on function entry, some operating
13698 systems violate this specification by not clearing the DF flag in their
13699 exception dispatchers. The exception handler can be invoked with the DF flag
13700 set, which leads to wrong direction mode when string instructions are used.
13701 This option can be enabled by default on 32-bit x86 targets by configuring
13702 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
13703 instructions can be suppressed with the @option{-mno-cld} compiler option
13707 @opindex mvzeroupper
13708 This option instructs GCC to emit a @code{vzeroupper} instruction
13709 before a transfer of control flow out of the function to minimize
13710 AVX to SSE transition penalty as well as remove unnecessary zeroupper
13713 @item -mprefer-avx128
13714 @opindex mprefer-avx128
13715 This option instructs GCC to use 128-bit AVX instructions instead of
13716 256-bit AVX instructions in the auto-vectorizer.
13720 This option will enable GCC to use CMPXCHG16B instruction in generated code.
13721 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
13722 data types. This is useful for high resolution counters that could be updated
13723 by multiple processors (or cores). This instruction is generated as part of
13724 atomic built-in functions: see @ref{__sync Builtins} or
13725 @ref{__atomic Builtins} for details.
13729 This option will enable GCC to use SAHF instruction in generated 64-bit code.
13730 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
13731 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
13732 SAHF are load and store instructions, respectively, for certain status flags.
13733 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
13734 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
13738 This option will enable GCC to use movbe instruction to implement
13739 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
13743 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
13744 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
13745 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
13749 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
13750 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
13751 to increase precision instead of DIVSS and SQRTSS (and their vectorized
13752 variants) for single-precision floating-point arguments. These instructions
13753 are generated only when @option{-funsafe-math-optimizations} is enabled
13754 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
13755 Note that while the throughput of the sequence is higher than the throughput
13756 of the non-reciprocal instruction, the precision of the sequence can be
13757 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
13759 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of RSQRTSS
13760 (or RSQRTPS) already with @option{-ffast-math} (or the above option
13761 combination), and doesn't need @option{-mrecip}.
13763 Also note that GCC emits the above sequence with additional Newton-Raphson step
13764 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
13765 already with @option{-ffast-math} (or the above option combination), and
13766 doesn't need @option{-mrecip}.
13768 @item -mrecip=@var{opt}
13769 @opindex mrecip=opt
13770 This option allows to control which reciprocal estimate instructions
13771 may be used. @var{opt} is a comma separated list of options, which may
13772 be preceded by a @code{!} to invert the option:
13773 @code{all}: enable all estimate instructions,
13774 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
13775 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip},
13776 @code{div}: enable the approximation for scalar division,
13777 @code{vec-div}: enable the approximation for vectorized division,
13778 @code{sqrt}: enable the approximation for scalar square root,
13779 @code{vec-sqrt}: enable the approximation for vectorized square root.
13781 So for example, @option{-mrecip=all,!sqrt} would enable
13782 all of the reciprocal approximations, except for square root.
13784 @item -mveclibabi=@var{type}
13785 @opindex mveclibabi
13786 Specifies the ABI type to use for vectorizing intrinsics using an
13787 external library. Supported types are @code{svml} for the Intel short
13788 vector math library and @code{acml} for the AMD math core library style
13789 of interfacing. GCC will currently emit calls to @code{vmldExp2},
13790 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
13791 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
13792 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
13793 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
13794 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
13795 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
13796 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
13797 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
13798 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
13799 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
13800 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
13801 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
13802 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
13803 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
13804 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
13805 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
13806 compatible library will have to be specified at link time.
13808 @item -mabi=@var{name}
13810 Generate code for the specified calling convention. Permissible values
13811 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
13812 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
13813 ABI when targeting Windows. On all other systems, the default is the
13814 SYSV ABI. You can control this behavior for a specific function by
13815 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
13816 @xref{Function Attributes}.
13818 @item -mtls-dialect=@var{type}
13819 @opindex mtls-dialect
13820 Generate code to access thread-local storage using the @samp{gnu} or
13821 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
13822 @samp{gnu2} is more efficient, but it may add compile- and run-time
13823 requirements that cannot be satisfied on all systems.
13826 @itemx -mno-push-args
13827 @opindex mpush-args
13828 @opindex mno-push-args
13829 Use PUSH operations to store outgoing parameters. This method is shorter
13830 and usually equally fast as method using SUB/MOV operations and is enabled
13831 by default. In some cases disabling it may improve performance because of
13832 improved scheduling and reduced dependencies.
13834 @item -maccumulate-outgoing-args
13835 @opindex maccumulate-outgoing-args
13836 If enabled, the maximum amount of space required for outgoing arguments will be
13837 computed in the function prologue. This is faster on most modern CPUs
13838 because of reduced dependencies, improved scheduling and reduced stack usage
13839 when preferred stack boundary is not equal to 2. The drawback is a notable
13840 increase in code size. This switch implies @option{-mno-push-args}.
13844 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
13845 on thread-safe exception handling must compile and link all code with the
13846 @option{-mthreads} option. When compiling, @option{-mthreads} defines
13847 @option{-D_MT}; when linking, it links in a special thread helper library
13848 @option{-lmingwthrd} which cleans up per thread exception handling data.
13850 @item -mno-align-stringops
13851 @opindex mno-align-stringops
13852 Do not align destination of inlined string operations. This switch reduces
13853 code size and improves performance in case the destination is already aligned,
13854 but GCC doesn't know about it.
13856 @item -minline-all-stringops
13857 @opindex minline-all-stringops
13858 By default GCC inlines string operations only when the destination is
13859 known to be aligned to least a 4-byte boundary.
13860 This enables more inlining, increase code
13861 size, but may improve performance of code that depends on fast memcpy, strlen
13862 and memset for short lengths.
13864 @item -minline-stringops-dynamically
13865 @opindex minline-stringops-dynamically
13866 For string operations of unknown size, use run-time checks with
13867 inline code for small blocks and a library call for large blocks.
13869 @item -mstringop-strategy=@var{alg}
13870 @opindex mstringop-strategy=@var{alg}
13871 Overwrite internal decision heuristic about particular algorithm to inline
13872 string operation with. The allowed values are @code{rep_byte},
13873 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
13874 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
13875 expanding inline loop, @code{libcall} for always expanding library call.
13877 @item -momit-leaf-frame-pointer
13878 @opindex momit-leaf-frame-pointer
13879 Don't keep the frame pointer in a register for leaf functions. This
13880 avoids the instructions to save, set up and restore frame pointers and
13881 makes an extra register available in leaf functions. The option
13882 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
13883 which might make debugging harder.
13885 @item -mtls-direct-seg-refs
13886 @itemx -mno-tls-direct-seg-refs
13887 @opindex mtls-direct-seg-refs
13888 Controls whether TLS variables may be accessed with offsets from the
13889 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
13890 or whether the thread base pointer must be added. Whether or not this
13891 is legal depends on the operating system, and whether it maps the
13892 segment to cover the entire TLS area.
13894 For systems that use GNU libc, the default is on.
13897 @itemx -mno-sse2avx
13899 Specify that the assembler should encode SSE instructions with VEX
13900 prefix. The option @option{-mavx} turns this on by default.
13905 If profiling is active @option{-pg} put the profiling
13906 counter call before prologue.
13907 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13908 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13911 @itemx -mno-8bit-idiv
13913 On some processors, like Intel Atom, 8-bit unsigned integer divide is
13914 much faster than 32-bit/64-bit integer divide. This option generates a
13915 run-time check. If both dividend and divisor are within range of 0
13916 to 255, 8-bit unsigned integer divide is used instead of
13917 32-bit/64-bit integer divide.
13919 @item -mavx256-split-unaligned-load
13920 @item -mavx256-split-unaligned-store
13921 @opindex avx256-split-unaligned-load
13922 @opindex avx256-split-unaligned-store
13923 Split 32-byte AVX unaligned load and store.
13927 These @samp{-m} switches are supported in addition to the above
13928 on AMD x86-64 processors in 64-bit environments.
13937 Generate code for a 32-bit or 64-bit environment.
13938 The @option{-m32} option sets int, long and pointer to 32 bits and
13939 generates code that runs on any i386 system.
13940 The @option{-m64} option sets int to 32 bits and long and pointer
13941 to 64 bits and generates code for AMD's x86-64 architecture.
13942 The @option{-mx32} option sets int, long and pointer to 32 bits and
13943 generates code for AMD's x86-64 architecture.
13944 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
13945 and @option{-mdynamic-no-pic} options.
13947 @item -mno-red-zone
13948 @opindex mno-red-zone
13949 Do not use a so called red zone for x86-64 code. The red zone is mandated
13950 by the x86-64 ABI, it is a 128-byte area beyond the location of the
13951 stack pointer that will not be modified by signal or interrupt handlers
13952 and therefore can be used for temporary data without adjusting the stack
13953 pointer. The flag @option{-mno-red-zone} disables this red zone.
13955 @item -mcmodel=small
13956 @opindex mcmodel=small
13957 Generate code for the small code model: the program and its symbols must
13958 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13959 Programs can be statically or dynamically linked. This is the default
13962 @item -mcmodel=kernel
13963 @opindex mcmodel=kernel
13964 Generate code for the kernel code model. The kernel runs in the
13965 negative 2 GB of the address space.
13966 This model has to be used for Linux kernel code.
13968 @item -mcmodel=medium
13969 @opindex mcmodel=medium
13970 Generate code for the medium model: The program is linked in the lower 2
13971 GB of the address space. Small symbols are also placed there. Symbols
13972 with sizes larger than @option{-mlarge-data-threshold} are put into
13973 large data or bss sections and can be located above 2GB. Programs can
13974 be statically or dynamically linked.
13976 @item -mcmodel=large
13977 @opindex mcmodel=large
13978 Generate code for the large model: This model makes no assumptions
13979 about addresses and sizes of sections.
13982 @node i386 and x86-64 Windows Options
13983 @subsection i386 and x86-64 Windows Options
13984 @cindex i386 and x86-64 Windows Options
13986 These additional options are available for Windows targets:
13991 This option is available for Cygwin and MinGW targets. It
13992 specifies that a console application is to be generated, by
13993 instructing the linker to set the PE header subsystem type
13994 required for console applications.
13995 This is the default behavior for Cygwin and MinGW targets.
13999 This option is available for Cygwin and MinGW targets. It
14000 specifies that a DLL - a dynamic link library - is to be
14001 generated, enabling the selection of the required runtime
14002 startup object and entry point.
14004 @item -mnop-fun-dllimport
14005 @opindex mnop-fun-dllimport
14006 This option is available for Cygwin and MinGW targets. It
14007 specifies that the dllimport attribute should be ignored.
14011 This option is available for MinGW targets. It specifies
14012 that MinGW-specific thread support is to be used.
14016 This option is available for mingw-w64 targets. It specifies
14017 that the UNICODE macro is getting pre-defined and that the
14018 unicode capable runtime startup code is chosen.
14022 This option is available for Cygwin and MinGW targets. It
14023 specifies that the typical Windows pre-defined macros are to
14024 be set in the pre-processor, but does not influence the choice
14025 of runtime library/startup code.
14029 This option is available for Cygwin and MinGW targets. It
14030 specifies that a GUI application is to be generated by
14031 instructing the linker to set the PE header subsystem type
14034 @item -fno-set-stack-executable
14035 @opindex fno-set-stack-executable
14036 This option is available for MinGW targets. It specifies that
14037 the executable flag for stack used by nested functions isn't
14038 set. This is necessary for binaries running in kernel mode of
14039 Windows, as there the user32 API, which is used to set executable
14040 privileges, isn't available.
14042 @item -mpe-aligned-commons
14043 @opindex mpe-aligned-commons
14044 This option is available for Cygwin and MinGW targets. It
14045 specifies that the GNU extension to the PE file format that
14046 permits the correct alignment of COMMON variables should be
14047 used when generating code. It will be enabled by default if
14048 GCC detects that the target assembler found during configuration
14049 supports the feature.
14052 See also under @ref{i386 and x86-64 Options} for standard options.
14054 @node IA-64 Options
14055 @subsection IA-64 Options
14056 @cindex IA-64 Options
14058 These are the @samp{-m} options defined for the Intel IA-64 architecture.
14062 @opindex mbig-endian
14063 Generate code for a big-endian target. This is the default for HP-UX@.
14065 @item -mlittle-endian
14066 @opindex mlittle-endian
14067 Generate code for a little-endian target. This is the default for AIX5
14073 @opindex mno-gnu-as
14074 Generate (or don't) code for the GNU assembler. This is the default.
14075 @c Also, this is the default if the configure option @option{--with-gnu-as}
14081 @opindex mno-gnu-ld
14082 Generate (or don't) code for the GNU linker. This is the default.
14083 @c Also, this is the default if the configure option @option{--with-gnu-ld}
14088 Generate code that does not use a global pointer register. The result
14089 is not position independent code, and violates the IA-64 ABI@.
14091 @item -mvolatile-asm-stop
14092 @itemx -mno-volatile-asm-stop
14093 @opindex mvolatile-asm-stop
14094 @opindex mno-volatile-asm-stop
14095 Generate (or don't) a stop bit immediately before and after volatile asm
14098 @item -mregister-names
14099 @itemx -mno-register-names
14100 @opindex mregister-names
14101 @opindex mno-register-names
14102 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
14103 the stacked registers. This may make assembler output more readable.
14109 Disable (or enable) optimizations that use the small data section. This may
14110 be useful for working around optimizer bugs.
14112 @item -mconstant-gp
14113 @opindex mconstant-gp
14114 Generate code that uses a single constant global pointer value. This is
14115 useful when compiling kernel code.
14119 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
14120 This is useful when compiling firmware code.
14122 @item -minline-float-divide-min-latency
14123 @opindex minline-float-divide-min-latency
14124 Generate code for inline divides of floating-point values
14125 using the minimum latency algorithm.
14127 @item -minline-float-divide-max-throughput
14128 @opindex minline-float-divide-max-throughput
14129 Generate code for inline divides of floating-point values
14130 using the maximum throughput algorithm.
14132 @item -mno-inline-float-divide
14133 @opindex mno-inline-float-divide
14134 Do not generate inline code for divides of floating-point values.
14136 @item -minline-int-divide-min-latency
14137 @opindex minline-int-divide-min-latency
14138 Generate code for inline divides of integer values
14139 using the minimum latency algorithm.
14141 @item -minline-int-divide-max-throughput
14142 @opindex minline-int-divide-max-throughput
14143 Generate code for inline divides of integer values
14144 using the maximum throughput algorithm.
14146 @item -mno-inline-int-divide
14147 @opindex mno-inline-int-divide
14148 Do not generate inline code for divides of integer values.
14150 @item -minline-sqrt-min-latency
14151 @opindex minline-sqrt-min-latency
14152 Generate code for inline square roots
14153 using the minimum latency algorithm.
14155 @item -minline-sqrt-max-throughput
14156 @opindex minline-sqrt-max-throughput
14157 Generate code for inline square roots
14158 using the maximum throughput algorithm.
14160 @item -mno-inline-sqrt
14161 @opindex mno-inline-sqrt
14162 Do not generate inline code for sqrt.
14165 @itemx -mno-fused-madd
14166 @opindex mfused-madd
14167 @opindex mno-fused-madd
14168 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
14169 instructions. The default is to use these instructions.
14171 @item -mno-dwarf2-asm
14172 @itemx -mdwarf2-asm
14173 @opindex mno-dwarf2-asm
14174 @opindex mdwarf2-asm
14175 Don't (or do) generate assembler code for the DWARF2 line number debugging
14176 info. This may be useful when not using the GNU assembler.
14178 @item -mearly-stop-bits
14179 @itemx -mno-early-stop-bits
14180 @opindex mearly-stop-bits
14181 @opindex mno-early-stop-bits
14182 Allow stop bits to be placed earlier than immediately preceding the
14183 instruction that triggered the stop bit. This can improve instruction
14184 scheduling, but does not always do so.
14186 @item -mfixed-range=@var{register-range}
14187 @opindex mfixed-range
14188 Generate code treating the given register range as fixed registers.
14189 A fixed register is one that the register allocator can not use. This is
14190 useful when compiling kernel code. A register range is specified as
14191 two registers separated by a dash. Multiple register ranges can be
14192 specified separated by a comma.
14194 @item -mtls-size=@var{tls-size}
14196 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
14199 @item -mtune=@var{cpu-type}
14201 Tune the instruction scheduling for a particular CPU, Valid values are
14202 itanium, itanium1, merced, itanium2, and mckinley.
14208 Generate code for a 32-bit or 64-bit environment.
14209 The 32-bit environment sets int, long and pointer to 32 bits.
14210 The 64-bit environment sets int to 32 bits and long and pointer
14211 to 64 bits. These are HP-UX specific flags.
14213 @item -mno-sched-br-data-spec
14214 @itemx -msched-br-data-spec
14215 @opindex mno-sched-br-data-spec
14216 @opindex msched-br-data-spec
14217 (Dis/En)able data speculative scheduling before reload.
14218 This will result in generation of the ld.a instructions and
14219 the corresponding check instructions (ld.c / chk.a).
14220 The default is 'disable'.
14222 @item -msched-ar-data-spec
14223 @itemx -mno-sched-ar-data-spec
14224 @opindex msched-ar-data-spec
14225 @opindex mno-sched-ar-data-spec
14226 (En/Dis)able data speculative scheduling after reload.
14227 This will result in generation of the ld.a instructions and
14228 the corresponding check instructions (ld.c / chk.a).
14229 The default is 'enable'.
14231 @item -mno-sched-control-spec
14232 @itemx -msched-control-spec
14233 @opindex mno-sched-control-spec
14234 @opindex msched-control-spec
14235 (Dis/En)able control speculative scheduling. This feature is
14236 available only during region scheduling (i.e.@: before reload).
14237 This will result in generation of the ld.s instructions and
14238 the corresponding check instructions chk.s .
14239 The default is 'disable'.
14241 @item -msched-br-in-data-spec
14242 @itemx -mno-sched-br-in-data-spec
14243 @opindex msched-br-in-data-spec
14244 @opindex mno-sched-br-in-data-spec
14245 (En/Dis)able speculative scheduling of the instructions that
14246 are dependent on the data speculative loads before reload.
14247 This is effective only with @option{-msched-br-data-spec} enabled.
14248 The default is 'enable'.
14250 @item -msched-ar-in-data-spec
14251 @itemx -mno-sched-ar-in-data-spec
14252 @opindex msched-ar-in-data-spec
14253 @opindex mno-sched-ar-in-data-spec
14254 (En/Dis)able speculative scheduling of the instructions that
14255 are dependent on the data speculative loads after reload.
14256 This is effective only with @option{-msched-ar-data-spec} enabled.
14257 The default is 'enable'.
14259 @item -msched-in-control-spec
14260 @itemx -mno-sched-in-control-spec
14261 @opindex msched-in-control-spec
14262 @opindex mno-sched-in-control-spec
14263 (En/Dis)able speculative scheduling of the instructions that
14264 are dependent on the control speculative loads.
14265 This is effective only with @option{-msched-control-spec} enabled.
14266 The default is 'enable'.
14268 @item -mno-sched-prefer-non-data-spec-insns
14269 @itemx -msched-prefer-non-data-spec-insns
14270 @opindex mno-sched-prefer-non-data-spec-insns
14271 @opindex msched-prefer-non-data-spec-insns
14272 If enabled, data speculative instructions will be chosen for schedule
14273 only if there are no other choices at the moment. This will make
14274 the use of the data speculation much more conservative.
14275 The default is 'disable'.
14277 @item -mno-sched-prefer-non-control-spec-insns
14278 @itemx -msched-prefer-non-control-spec-insns
14279 @opindex mno-sched-prefer-non-control-spec-insns
14280 @opindex msched-prefer-non-control-spec-insns
14281 If enabled, control speculative instructions will be chosen for schedule
14282 only if there are no other choices at the moment. This will make
14283 the use of the control speculation much more conservative.
14284 The default is 'disable'.
14286 @item -mno-sched-count-spec-in-critical-path
14287 @itemx -msched-count-spec-in-critical-path
14288 @opindex mno-sched-count-spec-in-critical-path
14289 @opindex msched-count-spec-in-critical-path
14290 If enabled, speculative dependencies will be considered during
14291 computation of the instructions priorities. This will make the use of the
14292 speculation a bit more conservative.
14293 The default is 'disable'.
14295 @item -msched-spec-ldc
14296 @opindex msched-spec-ldc
14297 Use a simple data speculation check. This option is on by default.
14299 @item -msched-control-spec-ldc
14300 @opindex msched-spec-ldc
14301 Use a simple check for control speculation. This option is on by default.
14303 @item -msched-stop-bits-after-every-cycle
14304 @opindex msched-stop-bits-after-every-cycle
14305 Place a stop bit after every cycle when scheduling. This option is on
14308 @item -msched-fp-mem-deps-zero-cost
14309 @opindex msched-fp-mem-deps-zero-cost
14310 Assume that floating-point stores and loads are not likely to cause a conflict
14311 when placed into the same instruction group. This option is disabled by
14314 @item -msel-sched-dont-check-control-spec
14315 @opindex msel-sched-dont-check-control-spec
14316 Generate checks for control speculation in selective scheduling.
14317 This flag is disabled by default.
14319 @item -msched-max-memory-insns=@var{max-insns}
14320 @opindex msched-max-memory-insns
14321 Limit on the number of memory insns per instruction group, giving lower
14322 priority to subsequent memory insns attempting to schedule in the same
14323 instruction group. Frequently useful to prevent cache bank conflicts.
14324 The default value is 1.
14326 @item -msched-max-memory-insns-hard-limit
14327 @opindex msched-max-memory-insns-hard-limit
14328 Disallow more than `msched-max-memory-insns' in instruction group.
14329 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
14330 when limit is reached but may still schedule memory operations.
14334 @node IA-64/VMS Options
14335 @subsection IA-64/VMS Options
14337 These @samp{-m} options are defined for the IA-64/VMS implementations:
14340 @item -mvms-return-codes
14341 @opindex mvms-return-codes
14342 Return VMS condition codes from main. The default is to return POSIX
14343 style condition (e.g.@ error) codes.
14345 @item -mdebug-main=@var{prefix}
14346 @opindex mdebug-main=@var{prefix}
14347 Flag the first routine whose name starts with @var{prefix} as the main
14348 routine for the debugger.
14352 Default to 64-bit memory allocation routines.
14356 @subsection LM32 Options
14357 @cindex LM32 options
14359 These @option{-m} options are defined for the Lattice Mico32 architecture:
14362 @item -mbarrel-shift-enabled
14363 @opindex mbarrel-shift-enabled
14364 Enable barrel-shift instructions.
14366 @item -mdivide-enabled
14367 @opindex mdivide-enabled
14368 Enable divide and modulus instructions.
14370 @item -mmultiply-enabled
14371 @opindex multiply-enabled
14372 Enable multiply instructions.
14374 @item -msign-extend-enabled
14375 @opindex msign-extend-enabled
14376 Enable sign extend instructions.
14378 @item -muser-enabled
14379 @opindex muser-enabled
14380 Enable user-defined instructions.
14385 @subsection M32C Options
14386 @cindex M32C options
14389 @item -mcpu=@var{name}
14391 Select the CPU for which code is generated. @var{name} may be one of
14392 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
14393 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
14394 the M32C/80 series.
14398 Specifies that the program will be run on the simulator. This causes
14399 an alternate runtime library to be linked in which supports, for
14400 example, file I/O@. You must not use this option when generating
14401 programs that will run on real hardware; you must provide your own
14402 runtime library for whatever I/O functions are needed.
14404 @item -memregs=@var{number}
14406 Specifies the number of memory-based pseudo-registers GCC will use
14407 during code generation. These pseudo-registers will be used like real
14408 registers, so there is a tradeoff between GCC's ability to fit the
14409 code into available registers, and the performance penalty of using
14410 memory instead of registers. Note that all modules in a program must
14411 be compiled with the same value for this option. Because of that, you
14412 must not use this option with the default runtime libraries gcc
14417 @node M32R/D Options
14418 @subsection M32R/D Options
14419 @cindex M32R/D options
14421 These @option{-m} options are defined for Renesas M32R/D architectures:
14426 Generate code for the M32R/2@.
14430 Generate code for the M32R/X@.
14434 Generate code for the M32R@. This is the default.
14436 @item -mmodel=small
14437 @opindex mmodel=small
14438 Assume all objects live in the lower 16MB of memory (so that their addresses
14439 can be loaded with the @code{ld24} instruction), and assume all subroutines
14440 are reachable with the @code{bl} instruction.
14441 This is the default.
14443 The addressability of a particular object can be set with the
14444 @code{model} attribute.
14446 @item -mmodel=medium
14447 @opindex mmodel=medium
14448 Assume objects may be anywhere in the 32-bit address space (the compiler
14449 will generate @code{seth/add3} instructions to load their addresses), and
14450 assume all subroutines are reachable with the @code{bl} instruction.
14452 @item -mmodel=large
14453 @opindex mmodel=large
14454 Assume objects may be anywhere in the 32-bit address space (the compiler
14455 will generate @code{seth/add3} instructions to load their addresses), and
14456 assume subroutines may not be reachable with the @code{bl} instruction
14457 (the compiler will generate the much slower @code{seth/add3/jl}
14458 instruction sequence).
14461 @opindex msdata=none
14462 Disable use of the small data area. Variables will be put into
14463 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
14464 @code{section} attribute has been specified).
14465 This is the default.
14467 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
14468 Objects may be explicitly put in the small data area with the
14469 @code{section} attribute using one of these sections.
14471 @item -msdata=sdata
14472 @opindex msdata=sdata
14473 Put small global and static data in the small data area, but do not
14474 generate special code to reference them.
14477 @opindex msdata=use
14478 Put small global and static data in the small data area, and generate
14479 special instructions to reference them.
14483 @cindex smaller data references
14484 Put global and static objects less than or equal to @var{num} bytes
14485 into the small data or bss sections instead of the normal data or bss
14486 sections. The default value of @var{num} is 8.
14487 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
14488 for this option to have any effect.
14490 All modules should be compiled with the same @option{-G @var{num}} value.
14491 Compiling with different values of @var{num} may or may not work; if it
14492 doesn't the linker will give an error message---incorrect code will not be
14497 Makes the M32R specific code in the compiler display some statistics
14498 that might help in debugging programs.
14500 @item -malign-loops
14501 @opindex malign-loops
14502 Align all loops to a 32-byte boundary.
14504 @item -mno-align-loops
14505 @opindex mno-align-loops
14506 Do not enforce a 32-byte alignment for loops. This is the default.
14508 @item -missue-rate=@var{number}
14509 @opindex missue-rate=@var{number}
14510 Issue @var{number} instructions per cycle. @var{number} can only be 1
14513 @item -mbranch-cost=@var{number}
14514 @opindex mbranch-cost=@var{number}
14515 @var{number} can only be 1 or 2. If it is 1 then branches will be
14516 preferred over conditional code, if it is 2, then the opposite will
14519 @item -mflush-trap=@var{number}
14520 @opindex mflush-trap=@var{number}
14521 Specifies the trap number to use to flush the cache. The default is
14522 12. Valid numbers are between 0 and 15 inclusive.
14524 @item -mno-flush-trap
14525 @opindex mno-flush-trap
14526 Specifies that the cache cannot be flushed by using a trap.
14528 @item -mflush-func=@var{name}
14529 @opindex mflush-func=@var{name}
14530 Specifies the name of the operating system function to call to flush
14531 the cache. The default is @emph{_flush_cache}, but a function call
14532 will only be used if a trap is not available.
14534 @item -mno-flush-func
14535 @opindex mno-flush-func
14536 Indicates that there is no OS function for flushing the cache.
14540 @node M680x0 Options
14541 @subsection M680x0 Options
14542 @cindex M680x0 options
14544 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
14545 The default settings depend on which architecture was selected when
14546 the compiler was configured; the defaults for the most common choices
14550 @item -march=@var{arch}
14552 Generate code for a specific M680x0 or ColdFire instruction set
14553 architecture. Permissible values of @var{arch} for M680x0
14554 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
14555 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
14556 architectures are selected according to Freescale's ISA classification
14557 and the permissible values are: @samp{isaa}, @samp{isaaplus},
14558 @samp{isab} and @samp{isac}.
14560 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
14561 code for a ColdFire target. The @var{arch} in this macro is one of the
14562 @option{-march} arguments given above.
14564 When used together, @option{-march} and @option{-mtune} select code
14565 that runs on a family of similar processors but that is optimized
14566 for a particular microarchitecture.
14568 @item -mcpu=@var{cpu}
14570 Generate code for a specific M680x0 or ColdFire processor.
14571 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
14572 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
14573 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
14574 below, which also classifies the CPUs into families:
14576 @multitable @columnfractions 0.20 0.80
14577 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
14578 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
14579 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
14580 @item @samp{5206e} @tab @samp{5206e}
14581 @item @samp{5208} @tab @samp{5207} @samp{5208}
14582 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
14583 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
14584 @item @samp{5216} @tab @samp{5214} @samp{5216}
14585 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
14586 @item @samp{5225} @tab @samp{5224} @samp{5225}
14587 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
14588 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
14589 @item @samp{5249} @tab @samp{5249}
14590 @item @samp{5250} @tab @samp{5250}
14591 @item @samp{5271} @tab @samp{5270} @samp{5271}
14592 @item @samp{5272} @tab @samp{5272}
14593 @item @samp{5275} @tab @samp{5274} @samp{5275}
14594 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
14595 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
14596 @item @samp{5307} @tab @samp{5307}
14597 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
14598 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
14599 @item @samp{5407} @tab @samp{5407}
14600 @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}
14603 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
14604 @var{arch} is compatible with @var{cpu}. Other combinations of
14605 @option{-mcpu} and @option{-march} are rejected.
14607 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
14608 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
14609 where the value of @var{family} is given by the table above.
14611 @item -mtune=@var{tune}
14613 Tune the code for a particular microarchitecture, within the
14614 constraints set by @option{-march} and @option{-mcpu}.
14615 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
14616 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
14617 and @samp{cpu32}. The ColdFire microarchitectures
14618 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
14620 You can also use @option{-mtune=68020-40} for code that needs
14621 to run relatively well on 68020, 68030 and 68040 targets.
14622 @option{-mtune=68020-60} is similar but includes 68060 targets
14623 as well. These two options select the same tuning decisions as
14624 @option{-m68020-40} and @option{-m68020-60} respectively.
14626 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
14627 when tuning for 680x0 architecture @var{arch}. It also defines
14628 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
14629 option is used. If gcc is tuning for a range of architectures,
14630 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
14631 it defines the macros for every architecture in the range.
14633 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
14634 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
14635 of the arguments given above.
14641 Generate output for a 68000. This is the default
14642 when the compiler is configured for 68000-based systems.
14643 It is equivalent to @option{-march=68000}.
14645 Use this option for microcontrollers with a 68000 or EC000 core,
14646 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
14650 Generate output for a 68010. This is the default
14651 when the compiler is configured for 68010-based systems.
14652 It is equivalent to @option{-march=68010}.
14658 Generate output for a 68020. This is the default
14659 when the compiler is configured for 68020-based systems.
14660 It is equivalent to @option{-march=68020}.
14664 Generate output for a 68030. This is the default when the compiler is
14665 configured for 68030-based systems. It is equivalent to
14666 @option{-march=68030}.
14670 Generate output for a 68040. This is the default when the compiler is
14671 configured for 68040-based systems. It is equivalent to
14672 @option{-march=68040}.
14674 This option inhibits the use of 68881/68882 instructions that have to be
14675 emulated by software on the 68040. Use this option if your 68040 does not
14676 have code to emulate those instructions.
14680 Generate output for a 68060. This is the default when the compiler is
14681 configured for 68060-based systems. It is equivalent to
14682 @option{-march=68060}.
14684 This option inhibits the use of 68020 and 68881/68882 instructions that
14685 have to be emulated by software on the 68060. Use this option if your 68060
14686 does not have code to emulate those instructions.
14690 Generate output for a CPU32. This is the default
14691 when the compiler is configured for CPU32-based systems.
14692 It is equivalent to @option{-march=cpu32}.
14694 Use this option for microcontrollers with a
14695 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
14696 68336, 68340, 68341, 68349 and 68360.
14700 Generate output for a 520X ColdFire CPU@. This is the default
14701 when the compiler is configured for 520X-based systems.
14702 It is equivalent to @option{-mcpu=5206}, and is now deprecated
14703 in favor of that option.
14705 Use this option for microcontroller with a 5200 core, including
14706 the MCF5202, MCF5203, MCF5204 and MCF5206.
14710 Generate output for a 5206e ColdFire CPU@. The option is now
14711 deprecated in favor of the equivalent @option{-mcpu=5206e}.
14715 Generate output for a member of the ColdFire 528X family.
14716 The option is now deprecated in favor of the equivalent
14717 @option{-mcpu=528x}.
14721 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
14722 in favor of the equivalent @option{-mcpu=5307}.
14726 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
14727 in favor of the equivalent @option{-mcpu=5407}.
14731 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
14732 This includes use of hardware floating-point instructions.
14733 The option is equivalent to @option{-mcpu=547x}, and is now
14734 deprecated in favor of that option.
14738 Generate output for a 68040, without using any of the new instructions.
14739 This results in code that can run relatively efficiently on either a
14740 68020/68881 or a 68030 or a 68040. The generated code does use the
14741 68881 instructions that are emulated on the 68040.
14743 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
14747 Generate output for a 68060, without using any of the new instructions.
14748 This results in code that can run relatively efficiently on either a
14749 68020/68881 or a 68030 or a 68040. The generated code does use the
14750 68881 instructions that are emulated on the 68060.
14752 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
14756 @opindex mhard-float
14758 Generate floating-point instructions. This is the default for 68020
14759 and above, and for ColdFire devices that have an FPU@. It defines the
14760 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
14761 on ColdFire targets.
14764 @opindex msoft-float
14765 Do not generate floating-point instructions; use library calls instead.
14766 This is the default for 68000, 68010, and 68832 targets. It is also
14767 the default for ColdFire devices that have no FPU.
14773 Generate (do not generate) ColdFire hardware divide and remainder
14774 instructions. If @option{-march} is used without @option{-mcpu},
14775 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
14776 architectures. Otherwise, the default is taken from the target CPU
14777 (either the default CPU, or the one specified by @option{-mcpu}). For
14778 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
14779 @option{-mcpu=5206e}.
14781 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
14785 Consider type @code{int} to be 16 bits wide, like @code{short int}.
14786 Additionally, parameters passed on the stack are also aligned to a
14787 16-bit boundary even on targets whose API mandates promotion to 32-bit.
14791 Do not consider type @code{int} to be 16 bits wide. This is the default.
14794 @itemx -mno-bitfield
14795 @opindex mnobitfield
14796 @opindex mno-bitfield
14797 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
14798 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
14802 Do use the bit-field instructions. The @option{-m68020} option implies
14803 @option{-mbitfield}. This is the default if you use a configuration
14804 designed for a 68020.
14808 Use a different function-calling convention, in which functions
14809 that take a fixed number of arguments return with the @code{rtd}
14810 instruction, which pops their arguments while returning. This
14811 saves one instruction in the caller since there is no need to pop
14812 the arguments there.
14814 This calling convention is incompatible with the one normally
14815 used on Unix, so you cannot use it if you need to call libraries
14816 compiled with the Unix compiler.
14818 Also, you must provide function prototypes for all functions that
14819 take variable numbers of arguments (including @code{printf});
14820 otherwise incorrect code will be generated for calls to those
14823 In addition, seriously incorrect code will result if you call a
14824 function with too many arguments. (Normally, extra arguments are
14825 harmlessly ignored.)
14827 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
14828 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
14832 Do not use the calling conventions selected by @option{-mrtd}.
14833 This is the default.
14836 @itemx -mno-align-int
14837 @opindex malign-int
14838 @opindex mno-align-int
14839 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
14840 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
14841 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
14842 Aligning variables on 32-bit boundaries produces code that runs somewhat
14843 faster on processors with 32-bit busses at the expense of more memory.
14845 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
14846 align structures containing the above types differently than
14847 most published application binary interface specifications for the m68k.
14851 Use the pc-relative addressing mode of the 68000 directly, instead of
14852 using a global offset table. At present, this option implies @option{-fpic},
14853 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
14854 not presently supported with @option{-mpcrel}, though this could be supported for
14855 68020 and higher processors.
14857 @item -mno-strict-align
14858 @itemx -mstrict-align
14859 @opindex mno-strict-align
14860 @opindex mstrict-align
14861 Do not (do) assume that unaligned memory references will be handled by
14865 Generate code that allows the data segment to be located in a different
14866 area of memory from the text segment. This allows for execute in place in
14867 an environment without virtual memory management. This option implies
14870 @item -mno-sep-data
14871 Generate code that assumes that the data segment follows the text segment.
14872 This is the default.
14874 @item -mid-shared-library
14875 Generate code that supports shared libraries via the library ID method.
14876 This allows for execute in place and shared libraries in an environment
14877 without virtual memory management. This option implies @option{-fPIC}.
14879 @item -mno-id-shared-library
14880 Generate code that doesn't assume ID based shared libraries are being used.
14881 This is the default.
14883 @item -mshared-library-id=n
14884 Specified the identification number of the ID based shared library being
14885 compiled. Specifying a value of 0 will generate more compact code, specifying
14886 other values will force the allocation of that number to the current
14887 library but is no more space or time efficient than omitting this option.
14893 When generating position-independent code for ColdFire, generate code
14894 that works if the GOT has more than 8192 entries. This code is
14895 larger and slower than code generated without this option. On M680x0
14896 processors, this option is not needed; @option{-fPIC} suffices.
14898 GCC normally uses a single instruction to load values from the GOT@.
14899 While this is relatively efficient, it only works if the GOT
14900 is smaller than about 64k. Anything larger causes the linker
14901 to report an error such as:
14903 @cindex relocation truncated to fit (ColdFire)
14905 relocation truncated to fit: R_68K_GOT16O foobar
14908 If this happens, you should recompile your code with @option{-mxgot}.
14909 It should then work with very large GOTs. However, code generated with
14910 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14911 the value of a global symbol.
14913 Note that some linkers, including newer versions of the GNU linker,
14914 can create multiple GOTs and sort GOT entries. If you have such a linker,
14915 you should only need to use @option{-mxgot} when compiling a single
14916 object file that accesses more than 8192 GOT entries. Very few do.
14918 These options have no effect unless GCC is generating
14919 position-independent code.
14923 @node MCore Options
14924 @subsection MCore Options
14925 @cindex MCore options
14927 These are the @samp{-m} options defined for the Motorola M*Core
14933 @itemx -mno-hardlit
14935 @opindex mno-hardlit
14936 Inline constants into the code stream if it can be done in two
14937 instructions or less.
14943 Use the divide instruction. (Enabled by default).
14945 @item -mrelax-immediate
14946 @itemx -mno-relax-immediate
14947 @opindex mrelax-immediate
14948 @opindex mno-relax-immediate
14949 Allow arbitrary sized immediates in bit operations.
14951 @item -mwide-bitfields
14952 @itemx -mno-wide-bitfields
14953 @opindex mwide-bitfields
14954 @opindex mno-wide-bitfields
14955 Always treat bit-fields as int-sized.
14957 @item -m4byte-functions
14958 @itemx -mno-4byte-functions
14959 @opindex m4byte-functions
14960 @opindex mno-4byte-functions
14961 Force all functions to be aligned to a 4-byte boundary.
14963 @item -mcallgraph-data
14964 @itemx -mno-callgraph-data
14965 @opindex mcallgraph-data
14966 @opindex mno-callgraph-data
14967 Emit callgraph information.
14970 @itemx -mno-slow-bytes
14971 @opindex mslow-bytes
14972 @opindex mno-slow-bytes
14973 Prefer word access when reading byte quantities.
14975 @item -mlittle-endian
14976 @itemx -mbig-endian
14977 @opindex mlittle-endian
14978 @opindex mbig-endian
14979 Generate code for a little-endian target.
14985 Generate code for the 210 processor.
14989 Assume that runtime support has been provided and so omit the
14990 simulator library (@file{libsim.a)} from the linker command line.
14992 @item -mstack-increment=@var{size}
14993 @opindex mstack-increment
14994 Set the maximum amount for a single stack increment operation. Large
14995 values can increase the speed of programs that contain functions
14996 that need a large amount of stack space, but they can also trigger a
14997 segmentation fault if the stack is extended too much. The default
15003 @subsection MeP Options
15004 @cindex MeP options
15010 Enables the @code{abs} instruction, which is the absolute difference
15011 between two registers.
15015 Enables all the optional instructions - average, multiply, divide, bit
15016 operations, leading zero, absolute difference, min/max, clip, and
15022 Enables the @code{ave} instruction, which computes the average of two
15025 @item -mbased=@var{n}
15027 Variables of size @var{n} bytes or smaller will be placed in the
15028 @code{.based} section by default. Based variables use the @code{$tp}
15029 register as a base register, and there is a 128-byte limit to the
15030 @code{.based} section.
15034 Enables the bit operation instructions - bit test (@code{btstm}), set
15035 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
15036 test-and-set (@code{tas}).
15038 @item -mc=@var{name}
15040 Selects which section constant data will be placed in. @var{name} may
15041 be @code{tiny}, @code{near}, or @code{far}.
15045 Enables the @code{clip} instruction. Note that @code{-mclip} is not
15046 useful unless you also provide @code{-mminmax}.
15048 @item -mconfig=@var{name}
15050 Selects one of the build-in core configurations. Each MeP chip has
15051 one or more modules in it; each module has a core CPU and a variety of
15052 coprocessors, optional instructions, and peripherals. The
15053 @code{MeP-Integrator} tool, not part of GCC, provides these
15054 configurations through this option; using this option is the same as
15055 using all the corresponding command-line options. The default
15056 configuration is @code{default}.
15060 Enables the coprocessor instructions. By default, this is a 32-bit
15061 coprocessor. Note that the coprocessor is normally enabled via the
15062 @code{-mconfig=} option.
15066 Enables the 32-bit coprocessor's instructions.
15070 Enables the 64-bit coprocessor's instructions.
15074 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
15078 Causes constant variables to be placed in the @code{.near} section.
15082 Enables the @code{div} and @code{divu} instructions.
15086 Generate big-endian code.
15090 Generate little-endian code.
15092 @item -mio-volatile
15093 @opindex mio-volatile
15094 Tells the compiler that any variable marked with the @code{io}
15095 attribute is to be considered volatile.
15099 Causes variables to be assigned to the @code{.far} section by default.
15103 Enables the @code{leadz} (leading zero) instruction.
15107 Causes variables to be assigned to the @code{.near} section by default.
15111 Enables the @code{min} and @code{max} instructions.
15115 Enables the multiplication and multiply-accumulate instructions.
15119 Disables all the optional instructions enabled by @code{-mall-opts}.
15123 Enables the @code{repeat} and @code{erepeat} instructions, used for
15124 low-overhead looping.
15128 Causes all variables to default to the @code{.tiny} section. Note
15129 that there is a 65536-byte limit to this section. Accesses to these
15130 variables use the @code{%gp} base register.
15134 Enables the saturation instructions. Note that the compiler does not
15135 currently generate these itself, but this option is included for
15136 compatibility with other tools, like @code{as}.
15140 Link the SDRAM-based runtime instead of the default ROM-based runtime.
15144 Link the simulator runtime libraries.
15148 Link the simulator runtime libraries, excluding built-in support
15149 for reset and exception vectors and tables.
15153 Causes all functions to default to the @code{.far} section. Without
15154 this option, functions default to the @code{.near} section.
15156 @item -mtiny=@var{n}
15158 Variables that are @var{n} bytes or smaller will be allocated to the
15159 @code{.tiny} section. These variables use the @code{$gp} base
15160 register. The default for this option is 4, but note that there's a
15161 65536-byte limit to the @code{.tiny} section.
15165 @node MicroBlaze Options
15166 @subsection MicroBlaze Options
15167 @cindex MicroBlaze Options
15172 @opindex msoft-float
15173 Use software emulation for floating point (default).
15176 @opindex mhard-float
15177 Use hardware floating-point instructions.
15181 Do not optimize block moves, use @code{memcpy}.
15183 @item -mno-clearbss
15184 @opindex mno-clearbss
15185 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
15187 @item -mcpu=@var{cpu-type}
15189 Use features of and schedule code for given CPU.
15190 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
15191 where @var{X} is a major version, @var{YY} is the minor version, and
15192 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
15193 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
15195 @item -mxl-soft-mul
15196 @opindex mxl-soft-mul
15197 Use software multiply emulation (default).
15199 @item -mxl-soft-div
15200 @opindex mxl-soft-div
15201 Use software emulation for divides (default).
15203 @item -mxl-barrel-shift
15204 @opindex mxl-barrel-shift
15205 Use the hardware barrel shifter.
15207 @item -mxl-pattern-compare
15208 @opindex mxl-pattern-compare
15209 Use pattern compare instructions.
15211 @item -msmall-divides
15212 @opindex msmall-divides
15213 Use table lookup optimization for small signed integer divisions.
15215 @item -mxl-stack-check
15216 @opindex mxl-stack-check
15217 This option is deprecated. Use -fstack-check instead.
15220 @opindex mxl-gp-opt
15221 Use GP relative sdata/sbss sections.
15223 @item -mxl-multiply-high
15224 @opindex mxl-multiply-high
15225 Use multiply high instructions for high part of 32x32 multiply.
15227 @item -mxl-float-convert
15228 @opindex mxl-float-convert
15229 Use hardware floating-point conversion instructions.
15231 @item -mxl-float-sqrt
15232 @opindex mxl-float-sqrt
15233 Use hardware floating-point square root instruction.
15235 @item -mxl-mode-@var{app-model}
15236 Select application model @var{app-model}. Valid models are
15239 normal executable (default), uses startup code @file{crt0.o}.
15242 for use with Xilinx Microprocessor Debugger (XMD) based
15243 software intrusive debug agent called xmdstub. This uses startup file
15244 @file{crt1.o} and sets the start address of the program to be 0x800.
15247 for applications that are loaded using a bootloader.
15248 This model uses startup file @file{crt2.o} which does not contain a processor
15249 reset vector handler. This is suitable for transferring control on a
15250 processor reset to the bootloader rather than the application.
15253 for applications that do not require any of the
15254 MicroBlaze vectors. This option may be useful for applications running
15255 within a monitoring application. This model uses @file{crt3.o} as a startup file.
15258 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
15259 @option{-mxl-mode-@var{app-model}}.
15264 @subsection MIPS Options
15265 @cindex MIPS options
15271 Generate big-endian code.
15275 Generate little-endian code. This is the default for @samp{mips*el-*-*}
15278 @item -march=@var{arch}
15280 Generate code that will run on @var{arch}, which can be the name of a
15281 generic MIPS ISA, or the name of a particular processor.
15283 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
15284 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
15285 The processor names are:
15286 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
15287 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
15288 @samp{5kc}, @samp{5kf},
15290 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
15291 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
15292 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
15293 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
15294 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
15295 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
15297 @samp{octeon}, @samp{octeon+}, @samp{octeon2},
15299 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
15300 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
15301 @samp{rm7000}, @samp{rm9000},
15302 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
15305 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
15306 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
15308 The special value @samp{from-abi} selects the
15309 most compatible architecture for the selected ABI (that is,
15310 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
15312 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
15313 which selects the best architecture option for the host processor.
15314 @option{-march=native} has no effect if GCC does not recognize
15317 In processor names, a final @samp{000} can be abbreviated as @samp{k}
15318 (for example, @samp{-march=r2k}). Prefixes are optional, and
15319 @samp{vr} may be written @samp{r}.
15321 Names of the form @samp{@var{n}f2_1} refer to processors with
15322 FPUs clocked at half the rate of the core, names of the form
15323 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
15324 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
15325 processors with FPUs clocked a ratio of 3:2 with respect to the core.
15326 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
15327 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
15328 accepted as synonyms for @samp{@var{n}f1_1}.
15330 GCC defines two macros based on the value of this option. The first
15331 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
15332 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
15333 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
15334 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
15335 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
15337 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
15338 above. In other words, it will have the full prefix and will not
15339 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
15340 the macro names the resolved architecture (either @samp{"mips1"} or
15341 @samp{"mips3"}). It names the default architecture when no
15342 @option{-march} option is given.
15344 @item -mtune=@var{arch}
15346 Optimize for @var{arch}. Among other things, this option controls
15347 the way instructions are scheduled, and the perceived cost of arithmetic
15348 operations. The list of @var{arch} values is the same as for
15351 When this option is not used, GCC will optimize for the processor
15352 specified by @option{-march}. By using @option{-march} and
15353 @option{-mtune} together, it is possible to generate code that will
15354 run on a family of processors, but optimize the code for one
15355 particular member of that family.
15357 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
15358 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
15359 @samp{-march} ones described above.
15363 Equivalent to @samp{-march=mips1}.
15367 Equivalent to @samp{-march=mips2}.
15371 Equivalent to @samp{-march=mips3}.
15375 Equivalent to @samp{-march=mips4}.
15379 Equivalent to @samp{-march=mips32}.
15383 Equivalent to @samp{-march=mips32r2}.
15387 Equivalent to @samp{-march=mips64}.
15391 Equivalent to @samp{-march=mips64r2}.
15396 @opindex mno-mips16
15397 Generate (do not generate) MIPS16 code. If GCC is targetting a
15398 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
15400 MIPS16 code generation can also be controlled on a per-function basis
15401 by means of @code{mips16} and @code{nomips16} attributes.
15402 @xref{Function Attributes}, for more information.
15404 @item -mflip-mips16
15405 @opindex mflip-mips16
15406 Generate MIPS16 code on alternating functions. This option is provided
15407 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
15408 not intended for ordinary use in compiling user code.
15410 @item -minterlink-mips16
15411 @itemx -mno-interlink-mips16
15412 @opindex minterlink-mips16
15413 @opindex mno-interlink-mips16
15414 Require (do not require) that non-MIPS16 code be link-compatible with
15417 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
15418 it must either use a call or an indirect jump. @option{-minterlink-mips16}
15419 therefore disables direct jumps unless GCC knows that the target of the
15420 jump is not MIPS16.
15432 Generate code for the given ABI@.
15434 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
15435 generates 64-bit code when you select a 64-bit architecture, but you
15436 can use @option{-mgp32} to get 32-bit code instead.
15438 For information about the O64 ABI, see
15439 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
15441 GCC supports a variant of the o32 ABI in which floating-point registers
15442 are 64 rather than 32 bits wide. You can select this combination with
15443 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
15444 and @samp{mfhc1} instructions and is therefore only supported for
15445 MIPS32R2 processors.
15447 The register assignments for arguments and return values remain the
15448 same, but each scalar value is passed in a single 64-bit register
15449 rather than a pair of 32-bit registers. For example, scalar
15450 floating-point values are returned in @samp{$f0} only, not a
15451 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
15452 remains the same, but all 64 bits are saved.
15455 @itemx -mno-abicalls
15457 @opindex mno-abicalls
15458 Generate (do not generate) code that is suitable for SVR4-style
15459 dynamic objects. @option{-mabicalls} is the default for SVR4-based
15464 Generate (do not generate) code that is fully position-independent,
15465 and that can therefore be linked into shared libraries. This option
15466 only affects @option{-mabicalls}.
15468 All @option{-mabicalls} code has traditionally been position-independent,
15469 regardless of options like @option{-fPIC} and @option{-fpic}. However,
15470 as an extension, the GNU toolchain allows executables to use absolute
15471 accesses for locally-binding symbols. It can also use shorter GP
15472 initialization sequences and generate direct calls to locally-defined
15473 functions. This mode is selected by @option{-mno-shared}.
15475 @option{-mno-shared} depends on binutils 2.16 or higher and generates
15476 objects that can only be linked by the GNU linker. However, the option
15477 does not affect the ABI of the final executable; it only affects the ABI
15478 of relocatable objects. Using @option{-mno-shared} will generally make
15479 executables both smaller and quicker.
15481 @option{-mshared} is the default.
15487 Assume (do not assume) that the static and dynamic linkers
15488 support PLTs and copy relocations. This option only affects
15489 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
15490 has no effect without @samp{-msym32}.
15492 You can make @option{-mplt} the default by configuring
15493 GCC with @option{--with-mips-plt}. The default is
15494 @option{-mno-plt} otherwise.
15500 Lift (do not lift) the usual restrictions on the size of the global
15503 GCC normally uses a single instruction to load values from the GOT@.
15504 While this is relatively efficient, it will only work if the GOT
15505 is smaller than about 64k. Anything larger will cause the linker
15506 to report an error such as:
15508 @cindex relocation truncated to fit (MIPS)
15510 relocation truncated to fit: R_MIPS_GOT16 foobar
15513 If this happens, you should recompile your code with @option{-mxgot}.
15514 It should then work with very large GOTs, although it will also be
15515 less efficient, since it will take three instructions to fetch the
15516 value of a global symbol.
15518 Note that some linkers can create multiple GOTs. If you have such a
15519 linker, you should only need to use @option{-mxgot} when a single object
15520 file accesses more than 64k's worth of GOT entries. Very few do.
15522 These options have no effect unless GCC is generating position
15527 Assume that general-purpose registers are 32 bits wide.
15531 Assume that general-purpose registers are 64 bits wide.
15535 Assume that floating-point registers are 32 bits wide.
15539 Assume that floating-point registers are 64 bits wide.
15542 @opindex mhard-float
15543 Use floating-point coprocessor instructions.
15546 @opindex msoft-float
15547 Do not use floating-point coprocessor instructions. Implement
15548 floating-point calculations using library calls instead.
15550 @item -msingle-float
15551 @opindex msingle-float
15552 Assume that the floating-point coprocessor only supports single-precision
15555 @item -mdouble-float
15556 @opindex mdouble-float
15557 Assume that the floating-point coprocessor supports double-precision
15558 operations. This is the default.
15564 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
15565 implement atomic memory built-in functions. When neither option is
15566 specified, GCC will use the instructions if the target architecture
15569 @option{-mllsc} is useful if the runtime environment can emulate the
15570 instructions and @option{-mno-llsc} can be useful when compiling for
15571 nonstandard ISAs. You can make either option the default by
15572 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
15573 respectively. @option{--with-llsc} is the default for some
15574 configurations; see the installation documentation for details.
15580 Use (do not use) revision 1 of the MIPS DSP ASE@.
15581 @xref{MIPS DSP Built-in Functions}. This option defines the
15582 preprocessor macro @samp{__mips_dsp}. It also defines
15583 @samp{__mips_dsp_rev} to 1.
15589 Use (do not use) revision 2 of the MIPS DSP ASE@.
15590 @xref{MIPS DSP Built-in Functions}. This option defines the
15591 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
15592 It also defines @samp{__mips_dsp_rev} to 2.
15595 @itemx -mno-smartmips
15596 @opindex msmartmips
15597 @opindex mno-smartmips
15598 Use (do not use) the MIPS SmartMIPS ASE.
15600 @item -mpaired-single
15601 @itemx -mno-paired-single
15602 @opindex mpaired-single
15603 @opindex mno-paired-single
15604 Use (do not use) paired-single floating-point instructions.
15605 @xref{MIPS Paired-Single Support}. This option requires
15606 hardware floating-point support to be enabled.
15612 Use (do not use) MIPS Digital Media Extension instructions.
15613 This option can only be used when generating 64-bit code and requires
15614 hardware floating-point support to be enabled.
15619 @opindex mno-mips3d
15620 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
15621 The option @option{-mips3d} implies @option{-mpaired-single}.
15627 Use (do not use) MT Multithreading instructions.
15631 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
15632 an explanation of the default and the way that the pointer size is
15637 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
15639 The default size of @code{int}s, @code{long}s and pointers depends on
15640 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
15641 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
15642 32-bit @code{long}s. Pointers are the same size as @code{long}s,
15643 or the same size as integer registers, whichever is smaller.
15649 Assume (do not assume) that all symbols have 32-bit values, regardless
15650 of the selected ABI@. This option is useful in combination with
15651 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
15652 to generate shorter and faster references to symbolic addresses.
15656 Put definitions of externally-visible data in a small data section
15657 if that data is no bigger than @var{num} bytes. GCC can then access
15658 the data more efficiently; see @option{-mgpopt} for details.
15660 The default @option{-G} option depends on the configuration.
15662 @item -mlocal-sdata
15663 @itemx -mno-local-sdata
15664 @opindex mlocal-sdata
15665 @opindex mno-local-sdata
15666 Extend (do not extend) the @option{-G} behavior to local data too,
15667 such as to static variables in C@. @option{-mlocal-sdata} is the
15668 default for all configurations.
15670 If the linker complains that an application is using too much small data,
15671 you might want to try rebuilding the less performance-critical parts with
15672 @option{-mno-local-sdata}. You might also want to build large
15673 libraries with @option{-mno-local-sdata}, so that the libraries leave
15674 more room for the main program.
15676 @item -mextern-sdata
15677 @itemx -mno-extern-sdata
15678 @opindex mextern-sdata
15679 @opindex mno-extern-sdata
15680 Assume (do not assume) that externally-defined data will be in
15681 a small data section if that data is within the @option{-G} limit.
15682 @option{-mextern-sdata} is the default for all configurations.
15684 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
15685 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
15686 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
15687 is placed in a small data section. If @var{Var} is defined by another
15688 module, you must either compile that module with a high-enough
15689 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
15690 definition. If @var{Var} is common, you must link the application
15691 with a high-enough @option{-G} setting.
15693 The easiest way of satisfying these restrictions is to compile
15694 and link every module with the same @option{-G} option. However,
15695 you may wish to build a library that supports several different
15696 small data limits. You can do this by compiling the library with
15697 the highest supported @option{-G} setting and additionally using
15698 @option{-mno-extern-sdata} to stop the library from making assumptions
15699 about externally-defined data.
15705 Use (do not use) GP-relative accesses for symbols that are known to be
15706 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
15707 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
15710 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
15711 might not hold the value of @code{_gp}. For example, if the code is
15712 part of a library that might be used in a boot monitor, programs that
15713 call boot monitor routines will pass an unknown value in @code{$gp}.
15714 (In such situations, the boot monitor itself would usually be compiled
15715 with @option{-G0}.)
15717 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
15718 @option{-mno-extern-sdata}.
15720 @item -membedded-data
15721 @itemx -mno-embedded-data
15722 @opindex membedded-data
15723 @opindex mno-embedded-data
15724 Allocate variables to the read-only data section first if possible, then
15725 next in the small data section if possible, otherwise in data. This gives
15726 slightly slower code than the default, but reduces the amount of RAM required
15727 when executing, and thus may be preferred for some embedded systems.
15729 @item -muninit-const-in-rodata
15730 @itemx -mno-uninit-const-in-rodata
15731 @opindex muninit-const-in-rodata
15732 @opindex mno-uninit-const-in-rodata
15733 Put uninitialized @code{const} variables in the read-only data section.
15734 This option is only meaningful in conjunction with @option{-membedded-data}.
15736 @item -mcode-readable=@var{setting}
15737 @opindex mcode-readable
15738 Specify whether GCC may generate code that reads from executable sections.
15739 There are three possible settings:
15742 @item -mcode-readable=yes
15743 Instructions may freely access executable sections. This is the
15746 @item -mcode-readable=pcrel
15747 MIPS16 PC-relative load instructions can access executable sections,
15748 but other instructions must not do so. This option is useful on 4KSc
15749 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
15750 It is also useful on processors that can be configured to have a dual
15751 instruction/data SRAM interface and that, like the M4K, automatically
15752 redirect PC-relative loads to the instruction RAM.
15754 @item -mcode-readable=no
15755 Instructions must not access executable sections. This option can be
15756 useful on targets that are configured to have a dual instruction/data
15757 SRAM interface but that (unlike the M4K) do not automatically redirect
15758 PC-relative loads to the instruction RAM.
15761 @item -msplit-addresses
15762 @itemx -mno-split-addresses
15763 @opindex msplit-addresses
15764 @opindex mno-split-addresses
15765 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
15766 relocation operators. This option has been superseded by
15767 @option{-mexplicit-relocs} but is retained for backwards compatibility.
15769 @item -mexplicit-relocs
15770 @itemx -mno-explicit-relocs
15771 @opindex mexplicit-relocs
15772 @opindex mno-explicit-relocs
15773 Use (do not use) assembler relocation operators when dealing with symbolic
15774 addresses. The alternative, selected by @option{-mno-explicit-relocs},
15775 is to use assembler macros instead.
15777 @option{-mexplicit-relocs} is the default if GCC was configured
15778 to use an assembler that supports relocation operators.
15780 @item -mcheck-zero-division
15781 @itemx -mno-check-zero-division
15782 @opindex mcheck-zero-division
15783 @opindex mno-check-zero-division
15784 Trap (do not trap) on integer division by zero.
15786 The default is @option{-mcheck-zero-division}.
15788 @item -mdivide-traps
15789 @itemx -mdivide-breaks
15790 @opindex mdivide-traps
15791 @opindex mdivide-breaks
15792 MIPS systems check for division by zero by generating either a
15793 conditional trap or a break instruction. Using traps results in
15794 smaller code, but is only supported on MIPS II and later. Also, some
15795 versions of the Linux kernel have a bug that prevents trap from
15796 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
15797 allow conditional traps on architectures that support them and
15798 @option{-mdivide-breaks} to force the use of breaks.
15800 The default is usually @option{-mdivide-traps}, but this can be
15801 overridden at configure time using @option{--with-divide=breaks}.
15802 Divide-by-zero checks can be completely disabled using
15803 @option{-mno-check-zero-division}.
15808 @opindex mno-memcpy
15809 Force (do not force) the use of @code{memcpy()} for non-trivial block
15810 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
15811 most constant-sized copies.
15814 @itemx -mno-long-calls
15815 @opindex mlong-calls
15816 @opindex mno-long-calls
15817 Disable (do not disable) use of the @code{jal} instruction. Calling
15818 functions using @code{jal} is more efficient but requires the caller
15819 and callee to be in the same 256 megabyte segment.
15821 This option has no effect on abicalls code. The default is
15822 @option{-mno-long-calls}.
15828 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
15829 instructions, as provided by the R4650 ISA@.
15832 @itemx -mno-fused-madd
15833 @opindex mfused-madd
15834 @opindex mno-fused-madd
15835 Enable (disable) use of the floating-point multiply-accumulate
15836 instructions, when they are available. The default is
15837 @option{-mfused-madd}.
15839 When multiply-accumulate instructions are used, the intermediate
15840 product is calculated to infinite precision and is not subject to
15841 the FCSR Flush to Zero bit. This may be undesirable in some
15846 Tell the MIPS assembler to not run its preprocessor over user
15847 assembler files (with a @samp{.s} suffix) when assembling them.
15852 @opindex mno-fix-24k
15853 Work around the 24K E48 (lost data on stores during refill) errata.
15854 The workarounds are implemented by the assembler rather than by GCC.
15857 @itemx -mno-fix-r4000
15858 @opindex mfix-r4000
15859 @opindex mno-fix-r4000
15860 Work around certain R4000 CPU errata:
15863 A double-word or a variable shift may give an incorrect result if executed
15864 immediately after starting an integer division.
15866 A double-word or a variable shift may give an incorrect result if executed
15867 while an integer multiplication is in progress.
15869 An integer division may give an incorrect result if started in a delay slot
15870 of a taken branch or a jump.
15874 @itemx -mno-fix-r4400
15875 @opindex mfix-r4400
15876 @opindex mno-fix-r4400
15877 Work around certain R4400 CPU errata:
15880 A double-word or a variable shift may give an incorrect result if executed
15881 immediately after starting an integer division.
15885 @itemx -mno-fix-r10000
15886 @opindex mfix-r10000
15887 @opindex mno-fix-r10000
15888 Work around certain R10000 errata:
15891 @code{ll}/@code{sc} sequences may not behave atomically on revisions
15892 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15895 This option can only be used if the target architecture supports
15896 branch-likely instructions. @option{-mfix-r10000} is the default when
15897 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15901 @itemx -mno-fix-vr4120
15902 @opindex mfix-vr4120
15903 Work around certain VR4120 errata:
15906 @code{dmultu} does not always produce the correct result.
15908 @code{div} and @code{ddiv} do not always produce the correct result if one
15909 of the operands is negative.
15911 The workarounds for the division errata rely on special functions in
15912 @file{libgcc.a}. At present, these functions are only provided by
15913 the @code{mips64vr*-elf} configurations.
15915 Other VR4120 errata require a nop to be inserted between certain pairs of
15916 instructions. These errata are handled by the assembler, not by GCC itself.
15919 @opindex mfix-vr4130
15920 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15921 workarounds are implemented by the assembler rather than by GCC,
15922 although GCC will avoid using @code{mflo} and @code{mfhi} if the
15923 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15924 instructions are available instead.
15927 @itemx -mno-fix-sb1
15929 Work around certain SB-1 CPU core errata.
15930 (This flag currently works around the SB-1 revision 2
15931 ``F1'' and ``F2'' floating-point errata.)
15933 @item -mr10k-cache-barrier=@var{setting}
15934 @opindex mr10k-cache-barrier
15935 Specify whether GCC should insert cache barriers to avoid the
15936 side-effects of speculation on R10K processors.
15938 In common with many processors, the R10K tries to predict the outcome
15939 of a conditional branch and speculatively executes instructions from
15940 the ``taken'' branch. It later aborts these instructions if the
15941 predicted outcome was wrong. However, on the R10K, even aborted
15942 instructions can have side effects.
15944 This problem only affects kernel stores and, depending on the system,
15945 kernel loads. As an example, a speculatively-executed store may load
15946 the target memory into cache and mark the cache line as dirty, even if
15947 the store itself is later aborted. If a DMA operation writes to the
15948 same area of memory before the ``dirty'' line is flushed, the cached
15949 data will overwrite the DMA-ed data. See the R10K processor manual
15950 for a full description, including other potential problems.
15952 One workaround is to insert cache barrier instructions before every memory
15953 access that might be speculatively executed and that might have side
15954 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15955 controls GCC's implementation of this workaround. It assumes that
15956 aborted accesses to any byte in the following regions will not have
15961 the memory occupied by the current function's stack frame;
15964 the memory occupied by an incoming stack argument;
15967 the memory occupied by an object with a link-time-constant address.
15970 It is the kernel's responsibility to ensure that speculative
15971 accesses to these regions are indeed safe.
15973 If the input program contains a function declaration such as:
15979 then the implementation of @code{foo} must allow @code{j foo} and
15980 @code{jal foo} to be executed speculatively. GCC honors this
15981 restriction for functions it compiles itself. It expects non-GCC
15982 functions (such as hand-written assembly code) to do the same.
15984 The option has three forms:
15987 @item -mr10k-cache-barrier=load-store
15988 Insert a cache barrier before a load or store that might be
15989 speculatively executed and that might have side effects even
15992 @item -mr10k-cache-barrier=store
15993 Insert a cache barrier before a store that might be speculatively
15994 executed and that might have side effects even if aborted.
15996 @item -mr10k-cache-barrier=none
15997 Disable the insertion of cache barriers. This is the default setting.
16000 @item -mflush-func=@var{func}
16001 @itemx -mno-flush-func
16002 @opindex mflush-func
16003 Specifies the function to call to flush the I and D caches, or to not
16004 call any such function. If called, the function must take the same
16005 arguments as the common @code{_flush_func()}, that is, the address of the
16006 memory range for which the cache is being flushed, the size of the
16007 memory range, and the number 3 (to flush both caches). The default
16008 depends on the target GCC was configured for, but commonly is either
16009 @samp{_flush_func} or @samp{__cpu_flush}.
16011 @item mbranch-cost=@var{num}
16012 @opindex mbranch-cost
16013 Set the cost of branches to roughly @var{num} ``simple'' instructions.
16014 This cost is only a heuristic and is not guaranteed to produce
16015 consistent results across releases. A zero cost redundantly selects
16016 the default, which is based on the @option{-mtune} setting.
16018 @item -mbranch-likely
16019 @itemx -mno-branch-likely
16020 @opindex mbranch-likely
16021 @opindex mno-branch-likely
16022 Enable or disable use of Branch Likely instructions, regardless of the
16023 default for the selected architecture. By default, Branch Likely
16024 instructions may be generated if they are supported by the selected
16025 architecture. An exception is for the MIPS32 and MIPS64 architectures
16026 and processors that implement those architectures; for those, Branch
16027 Likely instructions will not be generated by default because the MIPS32
16028 and MIPS64 architectures specifically deprecate their use.
16030 @item -mfp-exceptions
16031 @itemx -mno-fp-exceptions
16032 @opindex mfp-exceptions
16033 Specifies whether FP exceptions are enabled. This affects how we schedule
16034 FP instructions for some processors. The default is that FP exceptions are
16037 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
16038 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
16041 @item -mvr4130-align
16042 @itemx -mno-vr4130-align
16043 @opindex mvr4130-align
16044 The VR4130 pipeline is two-way superscalar, but can only issue two
16045 instructions together if the first one is 8-byte aligned. When this
16046 option is enabled, GCC will align pairs of instructions that it
16047 thinks should execute in parallel.
16049 This option only has an effect when optimizing for the VR4130.
16050 It normally makes code faster, but at the expense of making it bigger.
16051 It is enabled by default at optimization level @option{-O3}.
16056 Enable (disable) generation of @code{synci} instructions on
16057 architectures that support it. The @code{synci} instructions (if
16058 enabled) will be generated when @code{__builtin___clear_cache()} is
16061 This option defaults to @code{-mno-synci}, but the default can be
16062 overridden by configuring with @code{--with-synci}.
16064 When compiling code for single processor systems, it is generally safe
16065 to use @code{synci}. However, on many multi-core (SMP) systems, it
16066 will not invalidate the instruction caches on all cores and may lead
16067 to undefined behavior.
16069 @item -mrelax-pic-calls
16070 @itemx -mno-relax-pic-calls
16071 @opindex mrelax-pic-calls
16072 Try to turn PIC calls that are normally dispatched via register
16073 @code{$25} into direct calls. This is only possible if the linker can
16074 resolve the destination at link-time and if the destination is within
16075 range for a direct call.
16077 @option{-mrelax-pic-calls} is the default if GCC was configured to use
16078 an assembler and a linker that supports the @code{.reloc} assembly
16079 directive and @code{-mexplicit-relocs} is in effect. With
16080 @code{-mno-explicit-relocs}, this optimization can be performed by the
16081 assembler and the linker alone without help from the compiler.
16083 @item -mmcount-ra-address
16084 @itemx -mno-mcount-ra-address
16085 @opindex mmcount-ra-address
16086 @opindex mno-mcount-ra-address
16087 Emit (do not emit) code that allows @code{_mcount} to modify the
16088 calling function's return address. When enabled, this option extends
16089 the usual @code{_mcount} interface with a new @var{ra-address}
16090 parameter, which has type @code{intptr_t *} and is passed in register
16091 @code{$12}. @code{_mcount} can then modify the return address by
16092 doing both of the following:
16095 Returning the new address in register @code{$31}.
16097 Storing the new address in @code{*@var{ra-address}},
16098 if @var{ra-address} is nonnull.
16101 The default is @option{-mno-mcount-ra-address}.
16106 @subsection MMIX Options
16107 @cindex MMIX Options
16109 These options are defined for the MMIX:
16113 @itemx -mno-libfuncs
16115 @opindex mno-libfuncs
16116 Specify that intrinsic library functions are being compiled, passing all
16117 values in registers, no matter the size.
16120 @itemx -mno-epsilon
16122 @opindex mno-epsilon
16123 Generate floating-point comparison instructions that compare with respect
16124 to the @code{rE} epsilon register.
16126 @item -mabi=mmixware
16128 @opindex mabi=mmixware
16130 Generate code that passes function parameters and return values that (in
16131 the called function) are seen as registers @code{$0} and up, as opposed to
16132 the GNU ABI which uses global registers @code{$231} and up.
16134 @item -mzero-extend
16135 @itemx -mno-zero-extend
16136 @opindex mzero-extend
16137 @opindex mno-zero-extend
16138 When reading data from memory in sizes shorter than 64 bits, use (do not
16139 use) zero-extending load instructions by default, rather than
16140 sign-extending ones.
16143 @itemx -mno-knuthdiv
16145 @opindex mno-knuthdiv
16146 Make the result of a division yielding a remainder have the same sign as
16147 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
16148 remainder follows the sign of the dividend. Both methods are
16149 arithmetically valid, the latter being almost exclusively used.
16151 @item -mtoplevel-symbols
16152 @itemx -mno-toplevel-symbols
16153 @opindex mtoplevel-symbols
16154 @opindex mno-toplevel-symbols
16155 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
16156 code can be used with the @code{PREFIX} assembly directive.
16160 Generate an executable in the ELF format, rather than the default
16161 @samp{mmo} format used by the @command{mmix} simulator.
16163 @item -mbranch-predict
16164 @itemx -mno-branch-predict
16165 @opindex mbranch-predict
16166 @opindex mno-branch-predict
16167 Use (do not use) the probable-branch instructions, when static branch
16168 prediction indicates a probable branch.
16170 @item -mbase-addresses
16171 @itemx -mno-base-addresses
16172 @opindex mbase-addresses
16173 @opindex mno-base-addresses
16174 Generate (do not generate) code that uses @emph{base addresses}. Using a
16175 base address automatically generates a request (handled by the assembler
16176 and the linker) for a constant to be set up in a global register. The
16177 register is used for one or more base address requests within the range 0
16178 to 255 from the value held in the register. The generally leads to short
16179 and fast code, but the number of different data items that can be
16180 addressed is limited. This means that a program that uses lots of static
16181 data may require @option{-mno-base-addresses}.
16183 @item -msingle-exit
16184 @itemx -mno-single-exit
16185 @opindex msingle-exit
16186 @opindex mno-single-exit
16187 Force (do not force) generated code to have a single exit point in each
16191 @node MN10300 Options
16192 @subsection MN10300 Options
16193 @cindex MN10300 options
16195 These @option{-m} options are defined for Matsushita MN10300 architectures:
16200 Generate code to avoid bugs in the multiply instructions for the MN10300
16201 processors. This is the default.
16203 @item -mno-mult-bug
16204 @opindex mno-mult-bug
16205 Do not generate code to avoid bugs in the multiply instructions for the
16206 MN10300 processors.
16210 Generate code using features specific to the AM33 processor.
16214 Do not generate code using features specific to the AM33 processor. This
16219 Generate code using features specific to the AM33/2.0 processor.
16223 Generate code using features specific to the AM34 processor.
16225 @item -mtune=@var{cpu-type}
16227 Use the timing characteristics of the indicated CPU type when
16228 scheduling instructions. This does not change the targeted processor
16229 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
16230 @samp{am33-2} or @samp{am34}.
16232 @item -mreturn-pointer-on-d0
16233 @opindex mreturn-pointer-on-d0
16234 When generating a function that returns a pointer, return the pointer
16235 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
16236 only in a0, and attempts to call such functions without a prototype
16237 would result in errors. Note that this option is on by default; use
16238 @option{-mno-return-pointer-on-d0} to disable it.
16242 Do not link in the C run-time initialization object file.
16246 Indicate to the linker that it should perform a relaxation optimization pass
16247 to shorten branches, calls and absolute memory addresses. This option only
16248 has an effect when used on the command line for the final link step.
16250 This option makes symbolic debugging impossible.
16254 Allow the compiler to generate @emph{Long Instruction Word}
16255 instructions if the target is the @samp{AM33} or later. This is the
16256 default. This option defines the preprocessor macro @samp{__LIW__}.
16260 Do not allow the compiler to generate @emph{Long Instruction Word}
16261 instructions. This option defines the preprocessor macro
16266 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
16267 instructions if the target is the @samp{AM33} or later. This is the
16268 default. This option defines the preprocessor macro @samp{__SETLB__}.
16272 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
16273 instructions. This option defines the preprocessor macro
16274 @samp{__NO_SETLB__}.
16278 @node PDP-11 Options
16279 @subsection PDP-11 Options
16280 @cindex PDP-11 Options
16282 These options are defined for the PDP-11:
16287 Use hardware FPP floating point. This is the default. (FIS floating
16288 point on the PDP-11/40 is not supported.)
16291 @opindex msoft-float
16292 Do not use hardware floating point.
16296 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
16300 Return floating-point results in memory. This is the default.
16304 Generate code for a PDP-11/40.
16308 Generate code for a PDP-11/45. This is the default.
16312 Generate code for a PDP-11/10.
16314 @item -mbcopy-builtin
16315 @opindex mbcopy-builtin
16316 Use inline @code{movmemhi} patterns for copying memory. This is the
16321 Do not use inline @code{movmemhi} patterns for copying memory.
16327 Use 16-bit @code{int}. This is the default.
16333 Use 32-bit @code{int}.
16336 @itemx -mno-float32
16338 @opindex mno-float32
16339 Use 64-bit @code{float}. This is the default.
16342 @itemx -mno-float64
16344 @opindex mno-float64
16345 Use 32-bit @code{float}.
16349 Use @code{abshi2} pattern. This is the default.
16353 Do not use @code{abshi2} pattern.
16355 @item -mbranch-expensive
16356 @opindex mbranch-expensive
16357 Pretend that branches are expensive. This is for experimenting with
16358 code generation only.
16360 @item -mbranch-cheap
16361 @opindex mbranch-cheap
16362 Do not pretend that branches are expensive. This is the default.
16366 Use Unix assembler syntax. This is the default when configured for
16367 @samp{pdp11-*-bsd}.
16371 Use DEC assembler syntax. This is the default when configured for any
16372 PDP-11 target other than @samp{pdp11-*-bsd}.
16375 @node picoChip Options
16376 @subsection picoChip Options
16377 @cindex picoChip options
16379 These @samp{-m} options are defined for picoChip implementations:
16383 @item -mae=@var{ae_type}
16385 Set the instruction set, register set, and instruction scheduling
16386 parameters for array element type @var{ae_type}. Supported values
16387 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
16389 @option{-mae=ANY} selects a completely generic AE type. Code
16390 generated with this option will run on any of the other AE types. The
16391 code will not be as efficient as it would be if compiled for a specific
16392 AE type, and some types of operation (e.g., multiplication) will not
16393 work properly on all types of AE.
16395 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
16396 for compiled code, and is the default.
16398 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
16399 option may suffer from poor performance of byte (char) manipulation,
16400 since the DSP AE does not provide hardware support for byte load/stores.
16402 @item -msymbol-as-address
16403 Enable the compiler to directly use a symbol name as an address in a
16404 load/store instruction, without first loading it into a
16405 register. Typically, the use of this option will generate larger
16406 programs, which run faster than when the option isn't used. However, the
16407 results vary from program to program, so it is left as a user option,
16408 rather than being permanently enabled.
16410 @item -mno-inefficient-warnings
16411 Disables warnings about the generation of inefficient code. These
16412 warnings can be generated, for example, when compiling code that
16413 performs byte-level memory operations on the MAC AE type. The MAC AE has
16414 no hardware support for byte-level memory operations, so all byte
16415 load/stores must be synthesized from word load/store operations. This is
16416 inefficient and a warning will be generated indicating to the programmer
16417 that they should rewrite the code to avoid byte operations, or to target
16418 an AE type that has the necessary hardware support. This option enables
16419 the warning to be turned off.
16423 @node PowerPC Options
16424 @subsection PowerPC Options
16425 @cindex PowerPC options
16427 These are listed under @xref{RS/6000 and PowerPC Options}.
16430 @subsection RL78 Options
16431 @cindex RL78 Options
16437 Links in additional target libraries to support operation within a
16444 Specifies the type of hardware multiplication support to be used. The
16445 default is @code{none}, which uses software multiplication functions.
16446 The @code{g13} option is for the hardware multiply/divide peripheral
16447 only on the RL78/G13 targets. The @code{rl78} option is for the
16448 standard hardware multiplication defined in the RL78 software manual.
16452 @node RS/6000 and PowerPC Options
16453 @subsection IBM RS/6000 and PowerPC Options
16454 @cindex RS/6000 and PowerPC Options
16455 @cindex IBM RS/6000 and PowerPC Options
16457 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
16464 @itemx -mno-powerpc
16465 @itemx -mpowerpc-gpopt
16466 @itemx -mno-powerpc-gpopt
16467 @itemx -mpowerpc-gfxopt
16468 @itemx -mno-powerpc-gfxopt
16471 @itemx -mno-powerpc64
16475 @itemx -mno-popcntb
16477 @itemx -mno-popcntd
16486 @itemx -mno-hard-dfp
16490 @opindex mno-power2
16492 @opindex mno-powerpc
16493 @opindex mpowerpc-gpopt
16494 @opindex mno-powerpc-gpopt
16495 @opindex mpowerpc-gfxopt
16496 @opindex mno-powerpc-gfxopt
16497 @opindex mpowerpc64
16498 @opindex mno-powerpc64
16502 @opindex mno-popcntb
16504 @opindex mno-popcntd
16510 @opindex mno-mfpgpr
16512 @opindex mno-hard-dfp
16513 GCC supports two related instruction set architectures for the
16514 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
16515 instructions supported by the @samp{rios} chip set used in the original
16516 RS/6000 systems and the @dfn{PowerPC} instruction set is the
16517 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
16518 the IBM 4xx, 6xx, and follow-on microprocessors.
16520 Neither architecture is a subset of the other. However there is a
16521 large common subset of instructions supported by both. An MQ
16522 register is included in processors supporting the POWER architecture.
16524 You use these options to specify which instructions are available on the
16525 processor you are using. The default value of these options is
16526 determined when configuring GCC@. Specifying the
16527 @option{-mcpu=@var{cpu_type}} overrides the specification of these
16528 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
16529 rather than the options listed above.
16531 The @option{-mpower} option allows GCC to generate instructions that
16532 are found only in the POWER architecture and to use the MQ register.
16533 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
16534 to generate instructions that are present in the POWER2 architecture but
16535 not the original POWER architecture.
16537 The @option{-mpowerpc} option allows GCC to generate instructions that
16538 are found only in the 32-bit subset of the PowerPC architecture.
16539 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
16540 GCC to use the optional PowerPC architecture instructions in the
16541 General Purpose group, including floating-point square root. Specifying
16542 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
16543 use the optional PowerPC architecture instructions in the Graphics
16544 group, including floating-point select.
16546 The @option{-mmfcrf} option allows GCC to generate the move from
16547 condition register field instruction implemented on the POWER4
16548 processor and other processors that support the PowerPC V2.01
16550 The @option{-mpopcntb} option allows GCC to generate the popcount and
16551 double-precision FP reciprocal estimate instruction implemented on the
16552 POWER5 processor and other processors that support the PowerPC V2.02
16554 The @option{-mpopcntd} option allows GCC to generate the popcount
16555 instruction implemented on the POWER7 processor and other processors
16556 that support the PowerPC V2.06 architecture.
16557 The @option{-mfprnd} option allows GCC to generate the FP round to
16558 integer instructions implemented on the POWER5+ processor and other
16559 processors that support the PowerPC V2.03 architecture.
16560 The @option{-mcmpb} option allows GCC to generate the compare bytes
16561 instruction implemented on the POWER6 processor and other processors
16562 that support the PowerPC V2.05 architecture.
16563 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
16564 general-purpose register instructions implemented on the POWER6X
16565 processor and other processors that support the extended PowerPC V2.05
16567 The @option{-mhard-dfp} option allows GCC to generate the decimal
16568 floating-point instructions implemented on some POWER processors.
16570 The @option{-mpowerpc64} option allows GCC to generate the additional
16571 64-bit instructions that are found in the full PowerPC64 architecture
16572 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
16573 @option{-mno-powerpc64}.
16575 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
16576 will use only the instructions in the common subset of both
16577 architectures plus some special AIX common-mode calls, and will not use
16578 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
16579 permits GCC to use any instruction from either architecture and to
16580 allow use of the MQ register; specify this for the Motorola MPC601.
16582 @item -mnew-mnemonics
16583 @itemx -mold-mnemonics
16584 @opindex mnew-mnemonics
16585 @opindex mold-mnemonics
16586 Select which mnemonics to use in the generated assembler code. With
16587 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
16588 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
16589 assembler mnemonics defined for the POWER architecture. Instructions
16590 defined in only one architecture have only one mnemonic; GCC uses that
16591 mnemonic irrespective of which of these options is specified.
16593 GCC defaults to the mnemonics appropriate for the architecture in
16594 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
16595 value of these option. Unless you are building a cross-compiler, you
16596 should normally not specify either @option{-mnew-mnemonics} or
16597 @option{-mold-mnemonics}, but should instead accept the default.
16599 @item -mcpu=@var{cpu_type}
16601 Set architecture type, register usage, choice of mnemonics, and
16602 instruction scheduling parameters for machine type @var{cpu_type}.
16603 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
16604 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
16605 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
16606 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
16607 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
16608 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
16609 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
16610 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
16611 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
16612 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
16613 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
16615 @option{-mcpu=common} selects a completely generic processor. Code
16616 generated under this option will run on any POWER or PowerPC processor.
16617 GCC will use only the instructions in the common subset of both
16618 architectures, and will not use the MQ register. GCC assumes a generic
16619 processor model for scheduling purposes.
16621 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
16622 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
16623 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
16624 types, with an appropriate, generic processor model assumed for
16625 scheduling purposes.
16627 The other options specify a specific processor. Code generated under
16628 those options will run best on that processor, and may not run at all on
16631 The @option{-mcpu} options automatically enable or disable the
16634 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
16635 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
16636 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
16637 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
16639 The particular options set for any particular CPU will vary between
16640 compiler versions, depending on what setting seems to produce optimal
16641 code for that CPU; it doesn't necessarily reflect the actual hardware's
16642 capabilities. If you wish to set an individual option to a particular
16643 value, you may specify it after the @option{-mcpu} option, like
16644 @samp{-mcpu=970 -mno-altivec}.
16646 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
16647 not enabled or disabled by the @option{-mcpu} option at present because
16648 AIX does not have full support for these options. You may still
16649 enable or disable them individually if you're sure it'll work in your
16652 @item -mtune=@var{cpu_type}
16654 Set the instruction scheduling parameters for machine type
16655 @var{cpu_type}, but do not set the architecture type, register usage, or
16656 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
16657 values for @var{cpu_type} are used for @option{-mtune} as for
16658 @option{-mcpu}. If both are specified, the code generated will use the
16659 architecture, registers, and mnemonics set by @option{-mcpu}, but the
16660 scheduling parameters set by @option{-mtune}.
16662 @item -mcmodel=small
16663 @opindex mcmodel=small
16664 Generate PowerPC64 code for the small model: The TOC is limited to
16667 @item -mcmodel=medium
16668 @opindex mcmodel=medium
16669 Generate PowerPC64 code for the medium model: The TOC and other static
16670 data may be up to a total of 4G in size.
16672 @item -mcmodel=large
16673 @opindex mcmodel=large
16674 Generate PowerPC64 code for the large model: The TOC may be up to 4G
16675 in size. Other data and code is only limited by the 64-bit address
16679 @itemx -mno-altivec
16681 @opindex mno-altivec
16682 Generate code that uses (does not use) AltiVec instructions, and also
16683 enable the use of built-in functions that allow more direct access to
16684 the AltiVec instruction set. You may also need to set
16685 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
16691 @opindex mno-vrsave
16692 Generate VRSAVE instructions when generating AltiVec code.
16694 @item -mgen-cell-microcode
16695 @opindex mgen-cell-microcode
16696 Generate Cell microcode instructions
16698 @item -mwarn-cell-microcode
16699 @opindex mwarn-cell-microcode
16700 Warning when a Cell microcode instruction is going to emitted. An example
16701 of a Cell microcode instruction is a variable shift.
16704 @opindex msecure-plt
16705 Generate code that allows ld and ld.so to build executables and shared
16706 libraries with non-exec .plt and .got sections. This is a PowerPC
16707 32-bit SYSV ABI option.
16711 Generate code that uses a BSS .plt section that ld.so fills in, and
16712 requires .plt and .got sections that are both writable and executable.
16713 This is a PowerPC 32-bit SYSV ABI option.
16719 This switch enables or disables the generation of ISEL instructions.
16721 @item -misel=@var{yes/no}
16722 This switch has been deprecated. Use @option{-misel} and
16723 @option{-mno-isel} instead.
16729 This switch enables or disables the generation of SPE simd
16735 @opindex mno-paired
16736 This switch enables or disables the generation of PAIRED simd
16739 @item -mspe=@var{yes/no}
16740 This option has been deprecated. Use @option{-mspe} and
16741 @option{-mno-spe} instead.
16747 Generate code that uses (does not use) vector/scalar (VSX)
16748 instructions, and also enable the use of built-in functions that allow
16749 more direct access to the VSX instruction set.
16751 @item -mfloat-gprs=@var{yes/single/double/no}
16752 @itemx -mfloat-gprs
16753 @opindex mfloat-gprs
16754 This switch enables or disables the generation of floating-point
16755 operations on the general-purpose registers for architectures that
16758 The argument @var{yes} or @var{single} enables the use of
16759 single-precision floating-point operations.
16761 The argument @var{double} enables the use of single and
16762 double-precision floating-point operations.
16764 The argument @var{no} disables floating-point operations on the
16765 general-purpose registers.
16767 This option is currently only available on the MPC854x.
16773 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
16774 targets (including GNU/Linux). The 32-bit environment sets int, long
16775 and pointer to 32 bits and generates code that runs on any PowerPC
16776 variant. The 64-bit environment sets int to 32 bits and long and
16777 pointer to 64 bits, and generates code for PowerPC64, as for
16778 @option{-mpowerpc64}.
16781 @itemx -mno-fp-in-toc
16782 @itemx -mno-sum-in-toc
16783 @itemx -mminimal-toc
16785 @opindex mno-fp-in-toc
16786 @opindex mno-sum-in-toc
16787 @opindex mminimal-toc
16788 Modify generation of the TOC (Table Of Contents), which is created for
16789 every executable file. The @option{-mfull-toc} option is selected by
16790 default. In that case, GCC will allocate at least one TOC entry for
16791 each unique non-automatic variable reference in your program. GCC
16792 will also place floating-point constants in the TOC@. However, only
16793 16,384 entries are available in the TOC@.
16795 If you receive a linker error message that saying you have overflowed
16796 the available TOC space, you can reduce the amount of TOC space used
16797 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
16798 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
16799 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
16800 generate code to calculate the sum of an address and a constant at
16801 run time instead of putting that sum into the TOC@. You may specify one
16802 or both of these options. Each causes GCC to produce very slightly
16803 slower and larger code at the expense of conserving TOC space.
16805 If you still run out of space in the TOC even when you specify both of
16806 these options, specify @option{-mminimal-toc} instead. This option causes
16807 GCC to make only one TOC entry for every file. When you specify this
16808 option, GCC will produce code that is slower and larger but which
16809 uses extremely little TOC space. You may wish to use this option
16810 only on files that contain less frequently executed code.
16816 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
16817 @code{long} type, and the infrastructure needed to support them.
16818 Specifying @option{-maix64} implies @option{-mpowerpc64} and
16819 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
16820 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
16823 @itemx -mno-xl-compat
16824 @opindex mxl-compat
16825 @opindex mno-xl-compat
16826 Produce code that conforms more closely to IBM XL compiler semantics
16827 when using AIX-compatible ABI@. Pass floating-point arguments to
16828 prototyped functions beyond the register save area (RSA) on the stack
16829 in addition to argument FPRs. Do not assume that most significant
16830 double in 128-bit long double value is properly rounded when comparing
16831 values and converting to double. Use XL symbol names for long double
16834 The AIX calling convention was extended but not initially documented to
16835 handle an obscure K&R C case of calling a function that takes the
16836 address of its arguments with fewer arguments than declared. IBM XL
16837 compilers access floating-point arguments that do not fit in the
16838 RSA from the stack when a subroutine is compiled without
16839 optimization. Because always storing floating-point arguments on the
16840 stack is inefficient and rarely needed, this option is not enabled by
16841 default and only is necessary when calling subroutines compiled by IBM
16842 XL compilers without optimization.
16846 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
16847 application written to use message passing with special startup code to
16848 enable the application to run. The system must have PE installed in the
16849 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
16850 must be overridden with the @option{-specs=} option to specify the
16851 appropriate directory location. The Parallel Environment does not
16852 support threads, so the @option{-mpe} option and the @option{-pthread}
16853 option are incompatible.
16855 @item -malign-natural
16856 @itemx -malign-power
16857 @opindex malign-natural
16858 @opindex malign-power
16859 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
16860 @option{-malign-natural} overrides the ABI-defined alignment of larger
16861 types, such as floating-point doubles, on their natural size-based boundary.
16862 The option @option{-malign-power} instructs GCC to follow the ABI-specified
16863 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
16865 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
16869 @itemx -mhard-float
16870 @opindex msoft-float
16871 @opindex mhard-float
16872 Generate code that does not use (uses) the floating-point register set.
16873 Software floating-point emulation is provided if you use the
16874 @option{-msoft-float} option, and pass the option to GCC when linking.
16876 @item -msingle-float
16877 @itemx -mdouble-float
16878 @opindex msingle-float
16879 @opindex mdouble-float
16880 Generate code for single- or double-precision floating-point operations.
16881 @option{-mdouble-float} implies @option{-msingle-float}.
16884 @opindex msimple-fpu
16885 Do not generate sqrt and div instructions for hardware floating-point unit.
16889 Specify type of floating-point unit. Valid values are @var{sp_lite}
16890 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
16891 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
16892 and @var{dp_full} (equivalent to -mdouble-float).
16895 @opindex mxilinx-fpu
16896 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
16899 @itemx -mno-multiple
16901 @opindex mno-multiple
16902 Generate code that uses (does not use) the load multiple word
16903 instructions and the store multiple word instructions. These
16904 instructions are generated by default on POWER systems, and not
16905 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
16906 PowerPC systems, since those instructions do not work when the
16907 processor is in little-endian mode. The exceptions are PPC740 and
16908 PPC750 which permit these instructions in little-endian mode.
16913 @opindex mno-string
16914 Generate code that uses (does not use) the load string instructions
16915 and the store string word instructions to save multiple registers and
16916 do small block moves. These instructions are generated by default on
16917 POWER systems, and not generated on PowerPC systems. Do not use
16918 @option{-mstring} on little-endian PowerPC systems, since those
16919 instructions do not work when the processor is in little-endian mode.
16920 The exceptions are PPC740 and PPC750 which permit these instructions
16921 in little-endian mode.
16926 @opindex mno-update
16927 Generate code that uses (does not use) the load or store instructions
16928 that update the base register to the address of the calculated memory
16929 location. These instructions are generated by default. If you use
16930 @option{-mno-update}, there is a small window between the time that the
16931 stack pointer is updated and the address of the previous frame is
16932 stored, which means code that walks the stack frame across interrupts or
16933 signals may get corrupted data.
16935 @item -mavoid-indexed-addresses
16936 @itemx -mno-avoid-indexed-addresses
16937 @opindex mavoid-indexed-addresses
16938 @opindex mno-avoid-indexed-addresses
16939 Generate code that tries to avoid (not avoid) the use of indexed load
16940 or store instructions. These instructions can incur a performance
16941 penalty on Power6 processors in certain situations, such as when
16942 stepping through large arrays that cross a 16M boundary. This option
16943 is enabled by default when targetting Power6 and disabled otherwise.
16946 @itemx -mno-fused-madd
16947 @opindex mfused-madd
16948 @opindex mno-fused-madd
16949 Generate code that uses (does not use) the floating-point multiply and
16950 accumulate instructions. These instructions are generated by default
16951 if hardware floating point is used. The machine-dependent
16952 @option{-mfused-madd} option is now mapped to the machine-independent
16953 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16954 mapped to @option{-ffp-contract=off}.
16960 Generate code that uses (does not use) the half-word multiply and
16961 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
16962 These instructions are generated by default when targetting those
16969 Generate code that uses (does not use) the string-search @samp{dlmzb}
16970 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
16971 generated by default when targetting those processors.
16973 @item -mno-bit-align
16975 @opindex mno-bit-align
16976 @opindex mbit-align
16977 On System V.4 and embedded PowerPC systems do not (do) force structures
16978 and unions that contain bit-fields to be aligned to the base type of the
16981 For example, by default a structure containing nothing but 8
16982 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
16983 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
16984 the structure is aligned to a 1-byte boundary and is 1 byte in
16987 @item -mno-strict-align
16988 @itemx -mstrict-align
16989 @opindex mno-strict-align
16990 @opindex mstrict-align
16991 On System V.4 and embedded PowerPC systems do not (do) assume that
16992 unaligned memory references will be handled by the system.
16994 @item -mrelocatable
16995 @itemx -mno-relocatable
16996 @opindex mrelocatable
16997 @opindex mno-relocatable
16998 Generate code that allows (does not allow) a static executable to be
16999 relocated to a different address at run time. A simple embedded
17000 PowerPC system loader should relocate the entire contents of
17001 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
17002 a table of 32-bit addresses generated by this option. For this to
17003 work, all objects linked together must be compiled with
17004 @option{-mrelocatable} or @option{-mrelocatable-lib}.
17005 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
17007 @item -mrelocatable-lib
17008 @itemx -mno-relocatable-lib
17009 @opindex mrelocatable-lib
17010 @opindex mno-relocatable-lib
17011 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
17012 @code{.fixup} section to allow static executables to be relocated at
17013 run time, but @option{-mrelocatable-lib} does not use the smaller stack
17014 alignment of @option{-mrelocatable}. Objects compiled with
17015 @option{-mrelocatable-lib} may be linked with objects compiled with
17016 any combination of the @option{-mrelocatable} options.
17022 On System V.4 and embedded PowerPC systems do not (do) assume that
17023 register 2 contains a pointer to a global area pointing to the addresses
17024 used in the program.
17027 @itemx -mlittle-endian
17029 @opindex mlittle-endian
17030 On System V.4 and embedded PowerPC systems compile code for the
17031 processor in little-endian mode. The @option{-mlittle-endian} option is
17032 the same as @option{-mlittle}.
17035 @itemx -mbig-endian
17037 @opindex mbig-endian
17038 On System V.4 and embedded PowerPC systems compile code for the
17039 processor in big-endian mode. The @option{-mbig-endian} option is
17040 the same as @option{-mbig}.
17042 @item -mdynamic-no-pic
17043 @opindex mdynamic-no-pic
17044 On Darwin and Mac OS X systems, compile code so that it is not
17045 relocatable, but that its external references are relocatable. The
17046 resulting code is suitable for applications, but not shared
17049 @item -msingle-pic-base
17050 @opindex msingle-pic-base
17051 Treat the register used for PIC addressing as read-only, rather than
17052 loading it in the prologue for each function. The runtime system is
17053 responsible for initializing this register with an appropriate value
17054 before execution begins.
17056 @item -mprioritize-restricted-insns=@var{priority}
17057 @opindex mprioritize-restricted-insns
17058 This option controls the priority that is assigned to
17059 dispatch-slot restricted instructions during the second scheduling
17060 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
17061 @var{no/highest/second-highest} priority to dispatch slot restricted
17064 @item -msched-costly-dep=@var{dependence_type}
17065 @opindex msched-costly-dep
17066 This option controls which dependences are considered costly
17067 by the target during instruction scheduling. The argument
17068 @var{dependence_type} takes one of the following values:
17069 @var{no}: no dependence is costly,
17070 @var{all}: all dependences are costly,
17071 @var{true_store_to_load}: a true dependence from store to load is costly,
17072 @var{store_to_load}: any dependence from store to load is costly,
17073 @var{number}: any dependence for which latency >= @var{number} is costly.
17075 @item -minsert-sched-nops=@var{scheme}
17076 @opindex minsert-sched-nops
17077 This option controls which nop insertion scheme will be used during
17078 the second scheduling pass. The argument @var{scheme} takes one of the
17080 @var{no}: Don't insert nops.
17081 @var{pad}: Pad with nops any dispatch group that has vacant issue slots,
17082 according to the scheduler's grouping.
17083 @var{regroup_exact}: Insert nops to force costly dependent insns into
17084 separate groups. Insert exactly as many nops as needed to force an insn
17085 to a new group, according to the estimated processor grouping.
17086 @var{number}: Insert nops to force costly dependent insns into
17087 separate groups. Insert @var{number} nops to force an insn to a new group.
17090 @opindex mcall-sysv
17091 On System V.4 and embedded PowerPC systems compile code using calling
17092 conventions that adheres to the March 1995 draft of the System V
17093 Application Binary Interface, PowerPC processor supplement. This is the
17094 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
17096 @item -mcall-sysv-eabi
17098 @opindex mcall-sysv-eabi
17099 @opindex mcall-eabi
17100 Specify both @option{-mcall-sysv} and @option{-meabi} options.
17102 @item -mcall-sysv-noeabi
17103 @opindex mcall-sysv-noeabi
17104 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
17106 @item -mcall-aixdesc
17108 On System V.4 and embedded PowerPC systems compile code for the AIX
17112 @opindex mcall-linux
17113 On System V.4 and embedded PowerPC systems compile code for the
17114 Linux-based GNU system.
17116 @item -mcall-freebsd
17117 @opindex mcall-freebsd
17118 On System V.4 and embedded PowerPC systems compile code for the
17119 FreeBSD operating system.
17121 @item -mcall-netbsd
17122 @opindex mcall-netbsd
17123 On System V.4 and embedded PowerPC systems compile code for the
17124 NetBSD operating system.
17126 @item -mcall-openbsd
17127 @opindex mcall-netbsd
17128 On System V.4 and embedded PowerPC systems compile code for the
17129 OpenBSD operating system.
17131 @item -maix-struct-return
17132 @opindex maix-struct-return
17133 Return all structures in memory (as specified by the AIX ABI)@.
17135 @item -msvr4-struct-return
17136 @opindex msvr4-struct-return
17137 Return structures smaller than 8 bytes in registers (as specified by the
17140 @item -mabi=@var{abi-type}
17142 Extend the current ABI with a particular extension, or remove such extension.
17143 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
17144 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
17148 Extend the current ABI with SPE ABI extensions. This does not change
17149 the default ABI, instead it adds the SPE ABI extensions to the current
17153 @opindex mabi=no-spe
17154 Disable Booke SPE ABI extensions for the current ABI@.
17156 @item -mabi=ibmlongdouble
17157 @opindex mabi=ibmlongdouble
17158 Change the current ABI to use IBM extended-precision long double.
17159 This is a PowerPC 32-bit SYSV ABI option.
17161 @item -mabi=ieeelongdouble
17162 @opindex mabi=ieeelongdouble
17163 Change the current ABI to use IEEE extended-precision long double.
17164 This is a PowerPC 32-bit Linux ABI option.
17167 @itemx -mno-prototype
17168 @opindex mprototype
17169 @opindex mno-prototype
17170 On System V.4 and embedded PowerPC systems assume that all calls to
17171 variable argument functions are properly prototyped. Otherwise, the
17172 compiler must insert an instruction before every non prototyped call to
17173 set or clear bit 6 of the condition code register (@var{CR}) to
17174 indicate whether floating-point values were passed in the floating-point
17175 registers in case the function takes variable arguments. With
17176 @option{-mprototype}, only calls to prototyped variable argument functions
17177 will set or clear the bit.
17181 On embedded PowerPC systems, assume that the startup module is called
17182 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
17183 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
17188 On embedded PowerPC systems, assume that the startup module is called
17189 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
17194 On embedded PowerPC systems, assume that the startup module is called
17195 @file{crt0.o} and the standard C libraries are @file{libads.a} and
17198 @item -myellowknife
17199 @opindex myellowknife
17200 On embedded PowerPC systems, assume that the startup module is called
17201 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
17206 On System V.4 and embedded PowerPC systems, specify that you are
17207 compiling for a VxWorks system.
17211 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
17212 header to indicate that @samp{eabi} extended relocations are used.
17218 On System V.4 and embedded PowerPC systems do (do not) adhere to the
17219 Embedded Applications Binary Interface (eabi) which is a set of
17220 modifications to the System V.4 specifications. Selecting @option{-meabi}
17221 means that the stack is aligned to an 8-byte boundary, a function
17222 @code{__eabi} is called to from @code{main} to set up the eabi
17223 environment, and the @option{-msdata} option can use both @code{r2} and
17224 @code{r13} to point to two separate small data areas. Selecting
17225 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
17226 do not call an initialization function from @code{main}, and the
17227 @option{-msdata} option will only use @code{r13} to point to a single
17228 small data area. The @option{-meabi} option is on by default if you
17229 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
17232 @opindex msdata=eabi
17233 On System V.4 and embedded PowerPC systems, put small initialized
17234 @code{const} global and static data in the @samp{.sdata2} section, which
17235 is pointed to by register @code{r2}. Put small initialized
17236 non-@code{const} global and static data in the @samp{.sdata} section,
17237 which is pointed to by register @code{r13}. Put small uninitialized
17238 global and static data in the @samp{.sbss} section, which is adjacent to
17239 the @samp{.sdata} section. The @option{-msdata=eabi} option is
17240 incompatible with the @option{-mrelocatable} option. The
17241 @option{-msdata=eabi} option also sets the @option{-memb} option.
17244 @opindex msdata=sysv
17245 On System V.4 and embedded PowerPC systems, put small global and static
17246 data in the @samp{.sdata} section, which is pointed to by register
17247 @code{r13}. Put small uninitialized global and static data in the
17248 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
17249 The @option{-msdata=sysv} option is incompatible with the
17250 @option{-mrelocatable} option.
17252 @item -msdata=default
17254 @opindex msdata=default
17256 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
17257 compile code the same as @option{-msdata=eabi}, otherwise compile code the
17258 same as @option{-msdata=sysv}.
17261 @opindex msdata=data
17262 On System V.4 and embedded PowerPC systems, put small global
17263 data in the @samp{.sdata} section. Put small uninitialized global
17264 data in the @samp{.sbss} section. Do not use register @code{r13}
17265 to address small data however. This is the default behavior unless
17266 other @option{-msdata} options are used.
17270 @opindex msdata=none
17272 On embedded PowerPC systems, put all initialized global and static data
17273 in the @samp{.data} section, and all uninitialized data in the
17274 @samp{.bss} section.
17276 @item -mblock-move-inline-limit=@var{num}
17277 @opindex mblock-move-inline-limit
17278 Inline all block moves (such as calls to @code{memcpy} or structure
17279 copies) less than or equal to @var{num} bytes. The minimum value for
17280 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
17281 targets. The default value is target-specific.
17285 @cindex smaller data references (PowerPC)
17286 @cindex .sdata/.sdata2 references (PowerPC)
17287 On embedded PowerPC systems, put global and static items less than or
17288 equal to @var{num} bytes into the small data or bss sections instead of
17289 the normal data or bss section. By default, @var{num} is 8. The
17290 @option{-G @var{num}} switch is also passed to the linker.
17291 All modules should be compiled with the same @option{-G @var{num}} value.
17294 @itemx -mno-regnames
17296 @opindex mno-regnames
17297 On System V.4 and embedded PowerPC systems do (do not) emit register
17298 names in the assembly language output using symbolic forms.
17301 @itemx -mno-longcall
17303 @opindex mno-longcall
17304 By default assume that all calls are far away so that a longer more
17305 expensive calling sequence is required. This is required for calls
17306 further than 32 megabytes (33,554,432 bytes) from the current location.
17307 A short call will be generated if the compiler knows
17308 the call cannot be that far away. This setting can be overridden by
17309 the @code{shortcall} function attribute, or by @code{#pragma
17312 Some linkers are capable of detecting out-of-range calls and generating
17313 glue code on the fly. On these systems, long calls are unnecessary and
17314 generate slower code. As of this writing, the AIX linker can do this,
17315 as can the GNU linker for PowerPC/64. It is planned to add this feature
17316 to the GNU linker for 32-bit PowerPC systems as well.
17318 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
17319 callee, L42'', plus a ``branch island'' (glue code). The two target
17320 addresses represent the callee and the ``branch island''. The
17321 Darwin/PPC linker will prefer the first address and generate a ``bl
17322 callee'' if the PPC ``bl'' instruction will reach the callee directly;
17323 otherwise, the linker will generate ``bl L42'' to call the ``branch
17324 island''. The ``branch island'' is appended to the body of the
17325 calling function; it computes the full 32-bit address of the callee
17328 On Mach-O (Darwin) systems, this option directs the compiler emit to
17329 the glue for every direct call, and the Darwin linker decides whether
17330 to use or discard it.
17332 In the future, we may cause GCC to ignore all longcall specifications
17333 when the linker is known to generate glue.
17335 @item -mtls-markers
17336 @itemx -mno-tls-markers
17337 @opindex mtls-markers
17338 @opindex mno-tls-markers
17339 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
17340 specifying the function argument. The relocation allows ld to
17341 reliably associate function call with argument setup instructions for
17342 TLS optimization, which in turn allows gcc to better schedule the
17347 Adds support for multithreading with the @dfn{pthreads} library.
17348 This option sets flags for both the preprocessor and linker.
17353 This option will enable GCC to use the reciprocal estimate and
17354 reciprocal square root estimate instructions with additional
17355 Newton-Raphson steps to increase precision instead of doing a divide or
17356 square root and divide for floating-point arguments. You should use
17357 the @option{-ffast-math} option when using @option{-mrecip} (or at
17358 least @option{-funsafe-math-optimizations},
17359 @option{-finite-math-only}, @option{-freciprocal-math} and
17360 @option{-fno-trapping-math}). Note that while the throughput of the
17361 sequence is generally higher than the throughput of the non-reciprocal
17362 instruction, the precision of the sequence can be decreased by up to 2
17363 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
17366 @item -mrecip=@var{opt}
17367 @opindex mrecip=opt
17368 This option allows to control which reciprocal estimate instructions
17369 may be used. @var{opt} is a comma separated list of options, which may
17370 be preceded by a @code{!} to invert the option:
17371 @code{all}: enable all estimate instructions,
17372 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
17373 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
17374 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
17375 @code{divf}: enable the single-precision reciprocal approximation instructions;
17376 @code{divd}: enable the double-precision reciprocal approximation instructions;
17377 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
17378 @code{rsqrtf}: enable the single-precision reciprocal square root approximation instructions;
17379 @code{rsqrtd}: enable the double-precision reciprocal square root approximation instructions;
17381 So for example, @option{-mrecip=all,!rsqrtd} would enable the
17382 all of the reciprocal estimate instructions, except for the
17383 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
17384 which handle the double-precision reciprocal square root calculations.
17386 @item -mrecip-precision
17387 @itemx -mno-recip-precision
17388 @opindex mrecip-precision
17389 Assume (do not assume) that the reciprocal estimate instructions
17390 provide higher-precision estimates than is mandated by the PowerPC
17391 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
17392 automatically selects @option{-mrecip-precision}. The double-precision
17393 square root estimate instructions are not generated by
17394 default on low-precision machines, since they do not provide an
17395 estimate that converges after three steps.
17397 @item -mveclibabi=@var{type}
17398 @opindex mveclibabi
17399 Specifies the ABI type to use for vectorizing intrinsics using an
17400 external library. The only type supported at present is @code{mass},
17401 which specifies to use IBM's Mathematical Acceleration Subsystem
17402 (MASS) libraries for vectorizing intrinsics using external libraries.
17403 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
17404 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
17405 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
17406 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
17407 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
17408 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
17409 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
17410 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
17411 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
17412 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
17413 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
17414 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
17415 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
17416 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
17417 for power7. Both @option{-ftree-vectorize} and
17418 @option{-funsafe-math-optimizations} have to be enabled. The MASS
17419 libraries will have to be specified at link time.
17424 Generate (do not generate) the @code{friz} instruction when the
17425 @option{-funsafe-math-optimizations} option is used to optimize
17426 rounding of floating-point values to 64-bit integer and back to floating
17427 point. The @code{friz} instruction does not return the same value if
17428 the floating-point number is too large to fit in an integer.
17430 @item -mpointers-to-nested-functions
17431 @itemx -mno-pointers-to-nested-functions
17432 @opindex mpointers-to-nested-functions
17433 Generate (do not generate) code to load up the static chain register
17434 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
17435 systems where a function pointer points to a 3-word descriptor giving
17436 the function address, TOC value to be loaded in register @var{r2}, and
17437 static chain value to be loaded in register @var{r11}. The
17438 @option{-mpointers-to-nested-functions} is on by default. You will
17439 not be able to call through pointers to nested functions or pointers
17440 to functions compiled in other languages that use the static chain if
17441 you use the @option{-mno-pointers-to-nested-functions}.
17443 @item -msave-toc-indirect
17444 @itemx -mno-save-toc-indirect
17445 @opindex msave-toc-indirect
17446 Generate (do not generate) code to save the TOC value in the reserved
17447 stack location in the function prologue if the function calls through
17448 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
17449 saved in the prologue, it is saved just before the call through the
17450 pointer. The @option{-mno-save-toc-indirect} option is the default.
17454 @subsection RX Options
17457 These command-line options are defined for RX targets:
17460 @item -m64bit-doubles
17461 @itemx -m32bit-doubles
17462 @opindex m64bit-doubles
17463 @opindex m32bit-doubles
17464 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
17465 or 32 bits (@option{-m32bit-doubles}) in size. The default is
17466 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
17467 works on 32-bit values, which is why the default is
17468 @option{-m32bit-doubles}.
17474 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
17475 floating-point hardware. The default is enabled for the @var{RX600}
17476 series and disabled for the @var{RX200} series.
17478 Floating-point instructions will only be generated for 32-bit floating-point
17479 values however, so if the @option{-m64bit-doubles} option is in
17480 use then the FPU hardware will not be used for doubles.
17482 @emph{Note} If the @option{-fpu} option is enabled then
17483 @option{-funsafe-math-optimizations} is also enabled automatically.
17484 This is because the RX FPU instructions are themselves unsafe.
17486 @item -mcpu=@var{name}
17488 Selects the type of RX CPU to be targeted. Currently three types are
17489 supported, the generic @var{RX600} and @var{RX200} series hardware and
17490 the specific @var{RX610} CPU. The default is @var{RX600}.
17492 The only difference between @var{RX600} and @var{RX610} is that the
17493 @var{RX610} does not support the @code{MVTIPL} instruction.
17495 The @var{RX200} series does not have a hardware floating-point unit
17496 and so @option{-nofpu} is enabled by default when this type is
17499 @item -mbig-endian-data
17500 @itemx -mlittle-endian-data
17501 @opindex mbig-endian-data
17502 @opindex mlittle-endian-data
17503 Store data (but not code) in the big-endian format. The default is
17504 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
17507 @item -msmall-data-limit=@var{N}
17508 @opindex msmall-data-limit
17509 Specifies the maximum size in bytes of global and static variables
17510 which can be placed into the small data area. Using the small data
17511 area can lead to smaller and faster code, but the size of area is
17512 limited and it is up to the programmer to ensure that the area does
17513 not overflow. Also when the small data area is used one of the RX's
17514 registers (usually @code{r13}) is reserved for use pointing to this
17515 area, so it is no longer available for use by the compiler. This
17516 could result in slower and/or larger code if variables which once
17517 could have been held in the reserved register are now pushed onto the
17520 Note, common variables (variables that have not been initialized) and
17521 constants are not placed into the small data area as they are assigned
17522 to other sections in the output executable.
17524 The default value is zero, which disables this feature. Note, this
17525 feature is not enabled by default with higher optimization levels
17526 (@option{-O2} etc) because of the potentially detrimental effects of
17527 reserving a register. It is up to the programmer to experiment and
17528 discover whether this feature is of benefit to their program. See the
17529 description of the @option{-mpid} option for a description of how the
17530 actual register to hold the small data area pointer is chosen.
17536 Use the simulator runtime. The default is to use the libgloss board
17539 @item -mas100-syntax
17540 @itemx -mno-as100-syntax
17541 @opindex mas100-syntax
17542 @opindex mno-as100-syntax
17543 When generating assembler output use a syntax that is compatible with
17544 Renesas's AS100 assembler. This syntax can also be handled by the GAS
17545 assembler but it has some restrictions so generating it is not the
17548 @item -mmax-constant-size=@var{N}
17549 @opindex mmax-constant-size
17550 Specifies the maximum size, in bytes, of a constant that can be used as
17551 an operand in a RX instruction. Although the RX instruction set does
17552 allow constants of up to 4 bytes in length to be used in instructions,
17553 a longer value equates to a longer instruction. Thus in some
17554 circumstances it can be beneficial to restrict the size of constants
17555 that are used in instructions. Constants that are too big are instead
17556 placed into a constant pool and referenced via register indirection.
17558 The value @var{N} can be between 0 and 4. A value of 0 (the default)
17559 or 4 means that constants of any size are allowed.
17563 Enable linker relaxation. Linker relaxation is a process whereby the
17564 linker will attempt to reduce the size of a program by finding shorter
17565 versions of various instructions. Disabled by default.
17567 @item -mint-register=@var{N}
17568 @opindex mint-register
17569 Specify the number of registers to reserve for fast interrupt handler
17570 functions. The value @var{N} can be between 0 and 4. A value of 1
17571 means that register @code{r13} will be reserved for the exclusive use
17572 of fast interrupt handlers. A value of 2 reserves @code{r13} and
17573 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
17574 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
17575 A value of 0, the default, does not reserve any registers.
17577 @item -msave-acc-in-interrupts
17578 @opindex msave-acc-in-interrupts
17579 Specifies that interrupt handler functions should preserve the
17580 accumulator register. This is only necessary if normal code might use
17581 the accumulator register, for example because it performs 64-bit
17582 multiplications. The default is to ignore the accumulator as this
17583 makes the interrupt handlers faster.
17589 Enables the generation of position independent data. When enabled any
17590 access to constant data will done via an offset from a base address
17591 held in a register. This allows the location of constant data to be
17592 determined at run time without requiring the executable to be
17593 relocated, which is a benefit to embedded applications with tight
17594 memory constraints. Data that can be modified is not affected by this
17597 Note, using this feature reserves a register, usually @code{r13}, for
17598 the constant data base address. This can result in slower and/or
17599 larger code, especially in complicated functions.
17601 The actual register chosen to hold the constant data base address
17602 depends upon whether the @option{-msmall-data-limit} and/or the
17603 @option{-mint-register} command-line options are enabled. Starting
17604 with register @code{r13} and proceeding downwards, registers are
17605 allocated first to satisfy the requirements of @option{-mint-register},
17606 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
17607 is possible for the small data area register to be @code{r8} if both
17608 @option{-mint-register=4} and @option{-mpid} are specified on the
17611 By default this feature is not enabled. The default can be restored
17612 via the @option{-mno-pid} command-line option.
17616 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
17617 has special significance to the RX port when used with the
17618 @code{interrupt} function attribute. This attribute indicates a
17619 function intended to process fast interrupts. GCC will will ensure
17620 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
17621 and/or @code{r13} and only provided that the normal use of the
17622 corresponding registers have been restricted via the
17623 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
17626 @node S/390 and zSeries Options
17627 @subsection S/390 and zSeries Options
17628 @cindex S/390 and zSeries Options
17630 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
17634 @itemx -msoft-float
17635 @opindex mhard-float
17636 @opindex msoft-float
17637 Use (do not use) the hardware floating-point instructions and registers
17638 for floating-point operations. When @option{-msoft-float} is specified,
17639 functions in @file{libgcc.a} will be used to perform floating-point
17640 operations. When @option{-mhard-float} is specified, the compiler
17641 generates IEEE floating-point instructions. This is the default.
17644 @itemx -mno-hard-dfp
17646 @opindex mno-hard-dfp
17647 Use (do not use) the hardware decimal-floating-point instructions for
17648 decimal-floating-point operations. When @option{-mno-hard-dfp} is
17649 specified, functions in @file{libgcc.a} will be used to perform
17650 decimal-floating-point operations. When @option{-mhard-dfp} is
17651 specified, the compiler generates decimal-floating-point hardware
17652 instructions. This is the default for @option{-march=z9-ec} or higher.
17654 @item -mlong-double-64
17655 @itemx -mlong-double-128
17656 @opindex mlong-double-64
17657 @opindex mlong-double-128
17658 These switches control the size of @code{long double} type. A size
17659 of 64 bits makes the @code{long double} type equivalent to the @code{double}
17660 type. This is the default.
17663 @itemx -mno-backchain
17664 @opindex mbackchain
17665 @opindex mno-backchain
17666 Store (do not store) the address of the caller's frame as backchain pointer
17667 into the callee's stack frame.
17668 A backchain may be needed to allow debugging using tools that do not understand
17669 DWARF-2 call frame information.
17670 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
17671 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
17672 the backchain is placed into the topmost word of the 96/160 byte register
17675 In general, code compiled with @option{-mbackchain} is call-compatible with
17676 code compiled with @option{-mmo-backchain}; however, use of the backchain
17677 for debugging purposes usually requires that the whole binary is built with
17678 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
17679 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17680 to build a linux kernel use @option{-msoft-float}.
17682 The default is to not maintain the backchain.
17684 @item -mpacked-stack
17685 @itemx -mno-packed-stack
17686 @opindex mpacked-stack
17687 @opindex mno-packed-stack
17688 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
17689 specified, the compiler uses the all fields of the 96/160 byte register save
17690 area only for their default purpose; unused fields still take up stack space.
17691 When @option{-mpacked-stack} is specified, register save slots are densely
17692 packed at the top of the register save area; unused space is reused for other
17693 purposes, allowing for more efficient use of the available stack space.
17694 However, when @option{-mbackchain} is also in effect, the topmost word of
17695 the save area is always used to store the backchain, and the return address
17696 register is always saved two words below the backchain.
17698 As long as the stack frame backchain is not used, code generated with
17699 @option{-mpacked-stack} is call-compatible with code generated with
17700 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
17701 S/390 or zSeries generated code that uses the stack frame backchain at run
17702 time, not just for debugging purposes. Such code is not call-compatible
17703 with code compiled with @option{-mpacked-stack}. Also, note that the
17704 combination of @option{-mbackchain},
17705 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
17706 to build a linux kernel use @option{-msoft-float}.
17708 The default is to not use the packed stack layout.
17711 @itemx -mno-small-exec
17712 @opindex msmall-exec
17713 @opindex mno-small-exec
17714 Generate (or do not generate) code using the @code{bras} instruction
17715 to do subroutine calls.
17716 This only works reliably if the total executable size does not
17717 exceed 64k. The default is to use the @code{basr} instruction instead,
17718 which does not have this limitation.
17724 When @option{-m31} is specified, generate code compliant to the
17725 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
17726 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
17727 particular to generate 64-bit instructions. For the @samp{s390}
17728 targets, the default is @option{-m31}, while the @samp{s390x}
17729 targets default to @option{-m64}.
17735 When @option{-mzarch} is specified, generate code using the
17736 instructions available on z/Architecture.
17737 When @option{-mesa} is specified, generate code using the
17738 instructions available on ESA/390. Note that @option{-mesa} is
17739 not possible with @option{-m64}.
17740 When generating code compliant to the GNU/Linux for S/390 ABI,
17741 the default is @option{-mesa}. When generating code compliant
17742 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
17748 Generate (or do not generate) code using the @code{mvcle} instruction
17749 to perform block moves. When @option{-mno-mvcle} is specified,
17750 use a @code{mvc} loop instead. This is the default unless optimizing for
17757 Print (or do not print) additional debug information when compiling.
17758 The default is to not print debug information.
17760 @item -march=@var{cpu-type}
17762 Generate code that will run on @var{cpu-type}, which is the name of a system
17763 representing a certain processor type. Possible values for
17764 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
17765 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
17766 When generating code using the instructions available on z/Architecture,
17767 the default is @option{-march=z900}. Otherwise, the default is
17768 @option{-march=g5}.
17770 @item -mtune=@var{cpu-type}
17772 Tune to @var{cpu-type} everything applicable about the generated code,
17773 except for the ABI and the set of available instructions.
17774 The list of @var{cpu-type} values is the same as for @option{-march}.
17775 The default is the value used for @option{-march}.
17778 @itemx -mno-tpf-trace
17779 @opindex mtpf-trace
17780 @opindex mno-tpf-trace
17781 Generate code that adds (does not add) in TPF OS specific branches to trace
17782 routines in the operating system. This option is off by default, even
17783 when compiling for the TPF OS@.
17786 @itemx -mno-fused-madd
17787 @opindex mfused-madd
17788 @opindex mno-fused-madd
17789 Generate code that uses (does not use) the floating-point multiply and
17790 accumulate instructions. These instructions are generated by default if
17791 hardware floating point is used.
17793 @item -mwarn-framesize=@var{framesize}
17794 @opindex mwarn-framesize
17795 Emit a warning if the current function exceeds the given frame size. Because
17796 this is a compile-time check it doesn't need to be a real problem when the program
17797 runs. It is intended to identify functions that most probably cause
17798 a stack overflow. It is useful to be used in an environment with limited stack
17799 size e.g.@: the linux kernel.
17801 @item -mwarn-dynamicstack
17802 @opindex mwarn-dynamicstack
17803 Emit a warning if the function calls alloca or uses dynamically
17804 sized arrays. This is generally a bad idea with a limited stack size.
17806 @item -mstack-guard=@var{stack-guard}
17807 @itemx -mstack-size=@var{stack-size}
17808 @opindex mstack-guard
17809 @opindex mstack-size
17810 If these options are provided the s390 back end emits additional instructions in
17811 the function prologue which trigger a trap if the stack size is @var{stack-guard}
17812 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
17813 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
17814 the frame size of the compiled function is chosen.
17815 These options are intended to be used to help debugging stack overflow problems.
17816 The additionally emitted code causes only little overhead and hence can also be
17817 used in production like systems without greater performance degradation. The given
17818 values have to be exact powers of 2 and @var{stack-size} has to be greater than
17819 @var{stack-guard} without exceeding 64k.
17820 In order to be efficient the extra code makes the assumption that the stack starts
17821 at an address aligned to the value given by @var{stack-size}.
17822 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
17825 @node Score Options
17826 @subsection Score Options
17827 @cindex Score Options
17829 These options are defined for Score implementations:
17834 Compile code for big-endian mode. This is the default.
17838 Compile code for little-endian mode.
17842 Disable generate bcnz instruction.
17846 Enable generate unaligned load and store instruction.
17850 Enable the use of multiply-accumulate instructions. Disabled by default.
17854 Specify the SCORE5 as the target architecture.
17858 Specify the SCORE5U of the target architecture.
17862 Specify the SCORE7 as the target architecture. This is the default.
17866 Specify the SCORE7D as the target architecture.
17870 @subsection SH Options
17872 These @samp{-m} options are defined for the SH implementations:
17877 Generate code for the SH1.
17881 Generate code for the SH2.
17884 Generate code for the SH2e.
17888 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
17889 that the floating-point unit is not used.
17891 @item -m2a-single-only
17892 @opindex m2a-single-only
17893 Generate code for the SH2a-FPU, in such a way that no double-precision
17894 floating-point operations are used.
17897 @opindex m2a-single
17898 Generate code for the SH2a-FPU assuming the floating-point unit is in
17899 single-precision mode by default.
17903 Generate code for the SH2a-FPU assuming the floating-point unit is in
17904 double-precision mode by default.
17908 Generate code for the SH3.
17912 Generate code for the SH3e.
17916 Generate code for the SH4 without a floating-point unit.
17918 @item -m4-single-only
17919 @opindex m4-single-only
17920 Generate code for the SH4 with a floating-point unit that only
17921 supports single-precision arithmetic.
17925 Generate code for the SH4 assuming the floating-point unit is in
17926 single-precision mode by default.
17930 Generate code for the SH4.
17934 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
17935 floating-point unit is not used.
17937 @item -m4a-single-only
17938 @opindex m4a-single-only
17939 Generate code for the SH4a, in such a way that no double-precision
17940 floating-point operations are used.
17943 @opindex m4a-single
17944 Generate code for the SH4a assuming the floating-point unit is in
17945 single-precision mode by default.
17949 Generate code for the SH4a.
17953 Same as @option{-m4a-nofpu}, except that it implicitly passes
17954 @option{-dsp} to the assembler. GCC doesn't generate any DSP
17955 instructions at the moment.
17959 Compile code for the processor in big-endian mode.
17963 Compile code for the processor in little-endian mode.
17967 Align doubles at 64-bit boundaries. Note that this changes the calling
17968 conventions, and thus some functions from the standard C library will
17969 not work unless you recompile it first with @option{-mdalign}.
17973 Shorten some address references at link time, when possible; uses the
17974 linker option @option{-relax}.
17978 Use 32-bit offsets in @code{switch} tables. The default is to use
17983 Enable the use of bit manipulation instructions on SH2A.
17987 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
17988 alignment constraints.
17992 Comply with the calling conventions defined by Renesas.
17996 Comply with the calling conventions defined by Renesas.
18000 Comply with the calling conventions defined for GCC before the Renesas
18001 conventions were available. This option is the default for all
18002 targets of the SH toolchain.
18005 @opindex mnomacsave
18006 Mark the @code{MAC} register as call-clobbered, even if
18007 @option{-mhitachi} is given.
18013 Control the IEEE compliance of floating-point comparisons, which affects the
18014 handling of cases where the result of a comparison is unordered. By default
18015 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
18016 enabled @option{-mno-ieee} is implicitly set, which results in faster
18017 floating-point greater-equal and less-equal comparisons. The implcit settings
18018 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
18020 @item -minline-ic_invalidate
18021 @opindex minline-ic_invalidate
18022 Inline code to invalidate instruction cache entries after setting up
18023 nested function trampolines.
18024 This option has no effect if -musermode is in effect and the selected
18025 code generation option (e.g. -m4) does not allow the use of the icbi
18027 If the selected code generation option does not allow the use of the icbi
18028 instruction, and -musermode is not in effect, the inlined code will
18029 manipulate the instruction cache address array directly with an associative
18030 write. This not only requires privileged mode, but it will also
18031 fail if the cache line had been mapped via the TLB and has become unmapped.
18035 Dump instruction size and location in the assembly code.
18038 @opindex mpadstruct
18039 This option is deprecated. It pads structures to multiple of 4 bytes,
18040 which is incompatible with the SH ABI@.
18042 @item -msoft-atomic
18043 @opindex msoft-atomic
18044 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
18045 built-in functions. The generated atomic sequences require support from the
18046 interrupt / exception handling code of the system and are only suitable for
18047 single-core systems. They will not perform correctly on multi-core systems.
18048 This option is enabled by default when the target is @code{sh-*-linux*}.
18049 For details on the atomic built-in functions see @ref{__atomic Builtins}.
18053 Optimize for space instead of speed. Implied by @option{-Os}.
18056 @opindex mprefergot
18057 When generating position-independent code, emit function calls using
18058 the Global Offset Table instead of the Procedure Linkage Table.
18062 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
18063 if the inlined code would not work in user mode.
18064 This is the default when the target is @code{sh-*-linux*}.
18066 @item -multcost=@var{number}
18067 @opindex multcost=@var{number}
18068 Set the cost to assume for a multiply insn.
18070 @item -mdiv=@var{strategy}
18071 @opindex mdiv=@var{strategy}
18072 Set the division strategy to be used for integer division operations.
18073 For SHmedia @var{strategy} can be one of:
18078 Performs the operation in floating point. This has a very high latency,
18079 but needs only a few instructions, so it might be a good choice if
18080 your code has enough easily-exploitable ILP to allow the compiler to
18081 schedule the floating-point instructions together with other instructions.
18082 Division by zero causes a floating-point exception.
18085 Uses integer operations to calculate the inverse of the divisor,
18086 and then multiplies the dividend with the inverse. This strategy allows
18087 CSE and hoisting of the inverse calculation. Division by zero calculates
18088 an unspecified result, but does not trap.
18091 A variant of @samp{inv} where, if no CSE or hoisting opportunities
18092 have been found, or if the entire operation has been hoisted to the same
18093 place, the last stages of the inverse calculation are intertwined with the
18094 final multiply to reduce the overall latency, at the expense of using a few
18095 more instructions, and thus offering fewer scheduling opportunities with
18099 Calls a library function that usually implements the @samp{inv:minlat}
18101 This gives high code density for @code{m5-*media-nofpu} compilations.
18104 Uses a different entry point of the same library function, where it
18105 assumes that a pointer to a lookup table has already been set up, which
18106 exposes the pointer load to CSE and code hoisting optimizations.
18111 Use the @samp{inv} algorithm for initial
18112 code generation, but if the code stays unoptimized, revert to the @samp{call},
18113 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
18114 potentially-trapping side effect of division by zero is carried by a
18115 separate instruction, so it is possible that all the integer instructions
18116 are hoisted out, but the marker for the side effect stays where it is.
18117 A recombination to floating-point operations or a call is not possible
18122 Variants of the @samp{inv:minlat} strategy. In the case
18123 that the inverse calculation is not separated from the multiply, they speed
18124 up division where the dividend fits into 20 bits (plus sign where applicable)
18125 by inserting a test to skip a number of operations in this case; this test
18126 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
18127 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
18131 For targets other than SHmedia @var{strategy} can be one of:
18136 Calls a library function that uses the single-step division instruction
18137 @code{div1} to perform the operation. Division by zero calculates an
18138 unspecified result and does not trap. This is the default except for SH4,
18139 SH2A and SHcompact.
18142 Calls a library function that performs the operation in double precision
18143 floating point. Division by zero causes a floating-point exception. This is
18144 the default for SHcompact with FPU. Specifying this for targets that do not
18145 have a double precision FPU will default to @code{call-div1}.
18148 Calls a library function that uses a lookup table for small divisors and
18149 the @code{div1} instruction with case distinction for larger divisors. Division
18150 by zero calculates an unspecified result and does not trap. This is the default
18151 for SH4. Specifying this for targets that do not have dynamic shift
18152 instructions will default to @code{call-div1}.
18156 When a division strategy has not been specified the default strategy will be
18157 selected based on the current target. For SH2A the default strategy is to
18158 use the @code{divs} and @code{divu} instructions instead of library function
18161 @item -maccumulate-outgoing-args
18162 @opindex maccumulate-outgoing-args
18163 Reserve space once for outgoing arguments in the function prologue rather
18164 than around each call. Generally beneficial for performance and size. Also
18165 needed for unwinding to avoid changing the stack frame around conditional code.
18167 @item -mdivsi3_libfunc=@var{name}
18168 @opindex mdivsi3_libfunc=@var{name}
18169 Set the name of the library function used for 32-bit signed division to
18170 @var{name}. This only affect the name used in the call and inv:call
18171 division strategies, and the compiler will still expect the same
18172 sets of input/output/clobbered registers as if this option was not present.
18174 @item -mfixed-range=@var{register-range}
18175 @opindex mfixed-range
18176 Generate code treating the given register range as fixed registers.
18177 A fixed register is one that the register allocator can not use. This is
18178 useful when compiling kernel code. A register range is specified as
18179 two registers separated by a dash. Multiple register ranges can be
18180 specified separated by a comma.
18182 @item -madjust-unroll
18183 @opindex madjust-unroll
18184 Throttle unrolling to avoid thrashing target registers.
18185 This option only has an effect if the gcc code base supports the
18186 TARGET_ADJUST_UNROLL_MAX target hook.
18188 @item -mindexed-addressing
18189 @opindex mindexed-addressing
18190 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
18191 This is only safe if the hardware and/or OS implement 32-bit wrap-around
18192 semantics for the indexed addressing mode. The architecture allows the
18193 implementation of processors with 64-bit MMU, which the OS could use to
18194 get 32-bit addressing, but since no current hardware implementation supports
18195 this or any other way to make the indexed addressing mode safe to use in
18196 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
18198 @item -mgettrcost=@var{number}
18199 @opindex mgettrcost=@var{number}
18200 Set the cost assumed for the gettr instruction to @var{number}.
18201 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
18205 Assume pt* instructions won't trap. This will generally generate better
18206 scheduled code, but is unsafe on current hardware. The current architecture
18207 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
18208 This has the unintentional effect of making it unsafe to schedule ptabs /
18209 ptrel before a branch, or hoist it out of a loop. For example,
18210 __do_global_ctors, a part of libgcc that runs constructors at program
18211 startup, calls functions in a list which is delimited by @minus{}1. With the
18212 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
18213 That means that all the constructors will be run a bit quicker, but when
18214 the loop comes to the end of the list, the program crashes because ptabs
18215 loads @minus{}1 into a target register. Since this option is unsafe for any
18216 hardware implementing the current architecture specification, the default
18217 is -mno-pt-fixed. Unless the user specifies a specific cost with
18218 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
18219 this deters register allocation using target registers for storing
18222 @item -minvalid-symbols
18223 @opindex minvalid-symbols
18224 Assume symbols might be invalid. Ordinary function symbols generated by
18225 the compiler will always be valid to load with movi/shori/ptabs or
18226 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
18227 to generate symbols that will cause ptabs / ptrel to trap.
18228 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
18229 It will then prevent cross-basic-block cse, hoisting and most scheduling
18230 of symbol loads. The default is @option{-mno-invalid-symbols}.
18232 @item -mbranch-cost=@var{num}
18233 @opindex mbranch-cost=@var{num}
18234 Assume @var{num} to be the cost for a branch instruction. Higher numbers
18235 will make the compiler try to generate more branch-free code if possible.
18236 If not specified the value is selected depending on the processor type that
18237 is being compiled for.
18240 @opindex mcbranchdi
18241 Enable the @code{cbranchdi4} instruction pattern.
18245 Emit the @code{cmpeqdi_t} instruction pattern even when @option{-mcbranchdi}
18249 @opindex mfused-madd
18250 Allow the usage of the @code{fmac} instruction (floating-point
18251 multiply-accumulate) if the processor type supports it. Enabling this
18252 option might generate code that produces different numeric floating-point
18253 results compared to strict IEEE 754 arithmetic.
18255 @item -mpretend-cmove
18256 @opindex mpretend-cmove
18257 Prefer zero-displacement conditional branches for conditional move instruction
18258 patterns. This can result in faster code on the SH4 processor.
18262 @node Solaris 2 Options
18263 @subsection Solaris 2 Options
18264 @cindex Solaris 2 options
18266 These @samp{-m} options are supported on Solaris 2:
18269 @item -mimpure-text
18270 @opindex mimpure-text
18271 @option{-mimpure-text}, used in addition to @option{-shared}, tells
18272 the compiler to not pass @option{-z text} to the linker when linking a
18273 shared object. Using this option, you can link position-dependent
18274 code into a shared object.
18276 @option{-mimpure-text} suppresses the ``relocations remain against
18277 allocatable but non-writable sections'' linker error message.
18278 However, the necessary relocations will trigger copy-on-write, and the
18279 shared object is not actually shared across processes. Instead of
18280 using @option{-mimpure-text}, you should compile all source code with
18281 @option{-fpic} or @option{-fPIC}.
18285 These switches are supported in addition to the above on Solaris 2:
18290 Add support for multithreading using the POSIX threads library. This
18291 option sets flags for both the preprocessor and linker. This option does
18292 not affect the thread safety of object code produced by the compiler or
18293 that of libraries supplied with it.
18297 This is a synonym for @option{-pthreads}.
18300 @node SPARC Options
18301 @subsection SPARC Options
18302 @cindex SPARC options
18304 These @samp{-m} options are supported on the SPARC:
18307 @item -mno-app-regs
18309 @opindex mno-app-regs
18311 Specify @option{-mapp-regs} to generate output using the global registers
18312 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
18315 To be fully SVR4 ABI compliant at the cost of some performance loss,
18316 specify @option{-mno-app-regs}. You should compile libraries and system
18317 software with this option.
18323 With @option{-mflat}, the compiler does not generate save/restore instructions
18324 and uses a ``flat'' or single register window model. This model is compatible
18325 with the regular register window model. The local registers and the input
18326 registers (0--5) are still treated as ``call-saved'' registers and will be
18327 saved on the stack as needed.
18329 With @option{-mno-flat} (the default), the compiler generates save/restore
18330 instructions (except for leaf functions). This is the normal operating mode.
18333 @itemx -mhard-float
18335 @opindex mhard-float
18336 Generate output containing floating-point instructions. This is the
18340 @itemx -msoft-float
18342 @opindex msoft-float
18343 Generate output containing library calls for floating point.
18344 @strong{Warning:} the requisite libraries are not available for all SPARC
18345 targets. Normally the facilities of the machine's usual C compiler are
18346 used, but this cannot be done directly in cross-compilation. You must make
18347 your own arrangements to provide suitable library functions for
18348 cross-compilation. The embedded targets @samp{sparc-*-aout} and
18349 @samp{sparclite-*-*} do provide software floating-point support.
18351 @option{-msoft-float} changes the calling convention in the output file;
18352 therefore, it is only useful if you compile @emph{all} of a program with
18353 this option. In particular, you need to compile @file{libgcc.a}, the
18354 library that comes with GCC, with @option{-msoft-float} in order for
18357 @item -mhard-quad-float
18358 @opindex mhard-quad-float
18359 Generate output containing quad-word (long double) floating-point
18362 @item -msoft-quad-float
18363 @opindex msoft-quad-float
18364 Generate output containing library calls for quad-word (long double)
18365 floating-point instructions. The functions called are those specified
18366 in the SPARC ABI@. This is the default.
18368 As of this writing, there are no SPARC implementations that have hardware
18369 support for the quad-word floating-point instructions. They all invoke
18370 a trap handler for one of these instructions, and then the trap handler
18371 emulates the effect of the instruction. Because of the trap handler overhead,
18372 this is much slower than calling the ABI library routines. Thus the
18373 @option{-msoft-quad-float} option is the default.
18375 @item -mno-unaligned-doubles
18376 @itemx -munaligned-doubles
18377 @opindex mno-unaligned-doubles
18378 @opindex munaligned-doubles
18379 Assume that doubles have 8-byte alignment. This is the default.
18381 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
18382 alignment only if they are contained in another type, or if they have an
18383 absolute address. Otherwise, it assumes they have 4-byte alignment.
18384 Specifying this option avoids some rare compatibility problems with code
18385 generated by other compilers. It is not the default because it results
18386 in a performance loss, especially for floating-point code.
18388 @item -mno-faster-structs
18389 @itemx -mfaster-structs
18390 @opindex mno-faster-structs
18391 @opindex mfaster-structs
18392 With @option{-mfaster-structs}, the compiler assumes that structures
18393 should have 8-byte alignment. This enables the use of pairs of
18394 @code{ldd} and @code{std} instructions for copies in structure
18395 assignment, in place of twice as many @code{ld} and @code{st} pairs.
18396 However, the use of this changed alignment directly violates the SPARC
18397 ABI@. Thus, it's intended only for use on targets where the developer
18398 acknowledges that their resulting code will not be directly in line with
18399 the rules of the ABI@.
18401 @item -mcpu=@var{cpu_type}
18403 Set the instruction set, register set, and instruction scheduling parameters
18404 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
18405 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
18406 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
18407 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
18408 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
18409 and @samp{niagara4}.
18411 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
18412 which selects the best architecture option for the host processor.
18413 @option{-mcpu=native} has no effect if GCC does not recognize
18416 Default instruction scheduling parameters are used for values that select
18417 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
18418 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
18420 Here is a list of each supported architecture and their supported
18428 supersparc, hypersparc, leon
18431 f930, f934, sparclite86x
18437 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
18440 By default (unless configured otherwise), GCC generates code for the V7
18441 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
18442 additionally optimizes it for the Cypress CY7C602 chip, as used in the
18443 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
18444 SPARCStation 1, 2, IPX etc.
18446 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
18447 architecture. The only difference from V7 code is that the compiler emits
18448 the integer multiply and integer divide instructions which exist in SPARC-V8
18449 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
18450 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
18453 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
18454 the SPARC architecture. This adds the integer multiply, integer divide step
18455 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
18456 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
18457 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
18458 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
18459 MB86934 chip, which is the more recent SPARClite with FPU@.
18461 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
18462 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
18463 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
18464 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
18465 optimizes it for the TEMIC SPARClet chip.
18467 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
18468 architecture. This adds 64-bit integer and floating-point move instructions,
18469 3 additional floating-point condition code registers and conditional move
18470 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
18471 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
18472 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
18473 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
18474 @option{-mcpu=niagara}, the compiler additionally optimizes it for
18475 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
18476 additionally optimizes it for Sun UltraSPARC T2 chips. With
18477 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
18478 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
18479 additionally optimizes it for Sun UltraSPARC T4 chips.
18481 @item -mtune=@var{cpu_type}
18483 Set the instruction scheduling parameters for machine type
18484 @var{cpu_type}, but do not set the instruction set or register set that the
18485 option @option{-mcpu=@var{cpu_type}} would.
18487 The same values for @option{-mcpu=@var{cpu_type}} can be used for
18488 @option{-mtune=@var{cpu_type}}, but the only useful values are those
18489 that select a particular CPU implementation. Those are @samp{cypress},
18490 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
18491 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
18492 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
18493 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
18498 @opindex mno-v8plus
18499 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
18500 difference from the V8 ABI is that the global and out registers are
18501 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
18502 mode for all SPARC-V9 processors.
18508 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
18509 Visual Instruction Set extensions. The default is @option{-mno-vis}.
18515 With @option{-mvis2}, GCC generates code that takes advantage of
18516 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
18517 default is @option{-mvis2} when targetting a cpu that supports such
18518 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
18519 also sets @option{-mvis}.
18525 With @option{-mvis3}, GCC generates code that takes advantage of
18526 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
18527 default is @option{-mvis3} when targetting a cpu that supports such
18528 instructions, such as niagara-3 and later. Setting @option{-mvis3}
18529 also sets @option{-mvis2} and @option{-mvis}.
18535 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
18536 population count instruction. The default is @option{-mpopc}
18537 when targetting a cpu that supports such instructions, such as Niagara-2 and
18544 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
18545 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
18546 when targetting a cpu that supports such instructions, such as Niagara-3 and
18550 @opindex mfix-at697f
18551 Enable the documented workaround for the single erratum of the Atmel AT697F
18552 processor (which corresponds to erratum #13 of the AT697E processor).
18555 These @samp{-m} options are supported in addition to the above
18556 on SPARC-V9 processors in 64-bit environments:
18563 Generate code for a 32-bit or 64-bit environment.
18564 The 32-bit environment sets int, long and pointer to 32 bits.
18565 The 64-bit environment sets int to 32 bits and long and pointer
18568 @item -mcmodel=@var{which}
18570 Set the code model to one of
18574 The Medium/Low code model: 64-bit addresses, programs
18575 must be linked in the low 32 bits of memory. Programs can be statically
18576 or dynamically linked.
18579 The Medium/Middle code model: 64-bit addresses, programs
18580 must be linked in the low 44 bits of memory, the text and data segments must
18581 be less than 2GB in size and the data segment must be located within 2GB of
18585 The Medium/Anywhere code model: 64-bit addresses, programs
18586 may be linked anywhere in memory, the text and data segments must be less
18587 than 2GB in size and the data segment must be located within 2GB of the
18591 The Medium/Anywhere code model for embedded systems:
18592 64-bit addresses, the text and data segments must be less than 2GB in
18593 size, both starting anywhere in memory (determined at link time). The
18594 global register %g4 points to the base of the data segment. Programs
18595 are statically linked and PIC is not supported.
18598 @item -mmemory-model=@var{mem-model}
18599 @opindex mmemory-model
18600 Set the memory model in force on the processor to one of
18604 The default memory model for the processor and operating system.
18607 Relaxed Memory Order
18610 Partial Store Order
18616 Sequential Consistency
18619 These memory models are formally defined in Appendix D of the Sparc V9
18620 architecture manual, as set in the processor's @code{PSTATE.MM} field.
18623 @itemx -mno-stack-bias
18624 @opindex mstack-bias
18625 @opindex mno-stack-bias
18626 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
18627 frame pointer if present, are offset by @minus{}2047 which must be added back
18628 when making stack frame references. This is the default in 64-bit mode.
18629 Otherwise, assume no such offset is present.
18633 @subsection SPU Options
18634 @cindex SPU options
18636 These @samp{-m} options are supported on the SPU:
18640 @itemx -merror-reloc
18641 @opindex mwarn-reloc
18642 @opindex merror-reloc
18644 The loader for SPU does not handle dynamic relocations. By default, GCC
18645 will give an error when it generates code that requires a dynamic
18646 relocation. @option{-mno-error-reloc} disables the error,
18647 @option{-mwarn-reloc} will generate a warning instead.
18650 @itemx -munsafe-dma
18652 @opindex munsafe-dma
18654 Instructions that initiate or test completion of DMA must not be
18655 reordered with respect to loads and stores of the memory that is being
18656 accessed. Users typically address this problem using the volatile
18657 keyword, but that can lead to inefficient code in places where the
18658 memory is known to not change. Rather than mark the memory as volatile
18659 we treat the DMA instructions as potentially effecting all memory. With
18660 @option{-munsafe-dma} users must use the volatile keyword to protect
18663 @item -mbranch-hints
18664 @opindex mbranch-hints
18666 By default, GCC will generate a branch hint instruction to avoid
18667 pipeline stalls for always taken or probably taken branches. A hint
18668 will not be generated closer than 8 instructions away from its branch.
18669 There is little reason to disable them, except for debugging purposes,
18670 or to make an object a little bit smaller.
18674 @opindex msmall-mem
18675 @opindex mlarge-mem
18677 By default, GCC generates code assuming that addresses are never larger
18678 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
18679 a full 32-bit address.
18684 By default, GCC links against startup code that assumes the SPU-style
18685 main function interface (which has an unconventional parameter list).
18686 With @option{-mstdmain}, GCC will link your program against startup
18687 code that assumes a C99-style interface to @code{main}, including a
18688 local copy of @code{argv} strings.
18690 @item -mfixed-range=@var{register-range}
18691 @opindex mfixed-range
18692 Generate code treating the given register range as fixed registers.
18693 A fixed register is one that the register allocator can not use. This is
18694 useful when compiling kernel code. A register range is specified as
18695 two registers separated by a dash. Multiple register ranges can be
18696 specified separated by a comma.
18702 Compile code assuming that pointers to the PPU address space accessed
18703 via the @code{__ea} named address space qualifier are either 32 or 64
18704 bits wide. The default is 32 bits. As this is an ABI changing option,
18705 all object code in an executable must be compiled with the same setting.
18707 @item -maddress-space-conversion
18708 @itemx -mno-address-space-conversion
18709 @opindex maddress-space-conversion
18710 @opindex mno-address-space-conversion
18711 Allow/disallow treating the @code{__ea} address space as superset
18712 of the generic address space. This enables explicit type casts
18713 between @code{__ea} and generic pointer as well as implicit
18714 conversions of generic pointers to @code{__ea} pointers. The
18715 default is to allow address space pointer conversions.
18717 @item -mcache-size=@var{cache-size}
18718 @opindex mcache-size
18719 This option controls the version of libgcc that the compiler links to an
18720 executable and selects a software-managed cache for accessing variables
18721 in the @code{__ea} address space with a particular cache size. Possible
18722 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
18723 and @samp{128}. The default cache size is 64KB.
18725 @item -matomic-updates
18726 @itemx -mno-atomic-updates
18727 @opindex matomic-updates
18728 @opindex mno-atomic-updates
18729 This option controls the version of libgcc that the compiler links to an
18730 executable and selects whether atomic updates to the software-managed
18731 cache of PPU-side variables are used. If you use atomic updates, changes
18732 to a PPU variable from SPU code using the @code{__ea} named address space
18733 qualifier will not interfere with changes to other PPU variables residing
18734 in the same cache line from PPU code. If you do not use atomic updates,
18735 such interference may occur; however, writing back cache lines will be
18736 more efficient. The default behavior is to use atomic updates.
18739 @itemx -mdual-nops=@var{n}
18740 @opindex mdual-nops
18741 By default, GCC will insert nops to increase dual issue when it expects
18742 it to increase performance. @var{n} can be a value from 0 to 10. A
18743 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
18744 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
18746 @item -mhint-max-nops=@var{n}
18747 @opindex mhint-max-nops
18748 Maximum number of nops to insert for a branch hint. A branch hint must
18749 be at least 8 instructions away from the branch it is effecting. GCC
18750 will insert up to @var{n} nops to enforce this, otherwise it will not
18751 generate the branch hint.
18753 @item -mhint-max-distance=@var{n}
18754 @opindex mhint-max-distance
18755 The encoding of the branch hint instruction limits the hint to be within
18756 256 instructions of the branch it is effecting. By default, GCC makes
18757 sure it is within 125.
18760 @opindex msafe-hints
18761 Work around a hardware bug that causes the SPU to stall indefinitely.
18762 By default, GCC will insert the @code{hbrp} instruction to make sure
18763 this stall won't happen.
18767 @node System V Options
18768 @subsection Options for System V
18770 These additional options are available on System V Release 4 for
18771 compatibility with other compilers on those systems:
18776 Create a shared object.
18777 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
18781 Identify the versions of each tool used by the compiler, in a
18782 @code{.ident} assembler directive in the output.
18786 Refrain from adding @code{.ident} directives to the output file (this is
18789 @item -YP,@var{dirs}
18791 Search the directories @var{dirs}, and no others, for libraries
18792 specified with @option{-l}.
18794 @item -Ym,@var{dir}
18796 Look in the directory @var{dir} to find the M4 preprocessor.
18797 The assembler uses this option.
18798 @c This is supposed to go with a -Yd for predefined M4 macro files, but
18799 @c the generic assembler that comes with Solaris takes just -Ym.
18802 @node TILE-Gx Options
18803 @subsection TILE-Gx Options
18804 @cindex TILE-Gx options
18806 These @samp{-m} options are supported on the TILE-Gx:
18809 @item -mcpu=@var{name}
18811 Selects the type of CPU to be targeted. Currently the only supported
18812 type is @samp{tilegx}.
18818 Generate code for a 32-bit or 64-bit environment. The 32-bit
18819 environment sets int, long, and pointer to 32 bits. The 64-bit
18820 environment sets int to 32 bits and long and pointer to 64 bits.
18823 @node TILEPro Options
18824 @subsection TILEPro Options
18825 @cindex TILEPro options
18827 These @samp{-m} options are supported on the TILEPro:
18830 @item -mcpu=@var{name}
18832 Selects the type of CPU to be targeted. Currently the only supported
18833 type is @samp{tilepro}.
18837 Generate code for a 32-bit environment, which sets int, long, and
18838 pointer to 32 bits. This is the only supported behavior so the flag
18839 is essentially ignored.
18843 @subsection V850 Options
18844 @cindex V850 Options
18846 These @samp{-m} options are defined for V850 implementations:
18850 @itemx -mno-long-calls
18851 @opindex mlong-calls
18852 @opindex mno-long-calls
18853 Treat all calls as being far away (near). If calls are assumed to be
18854 far away, the compiler will always load the functions address up into a
18855 register, and call indirect through the pointer.
18861 Do not optimize (do optimize) basic blocks that use the same index
18862 pointer 4 or more times to copy pointer into the @code{ep} register, and
18863 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
18864 option is on by default if you optimize.
18866 @item -mno-prolog-function
18867 @itemx -mprolog-function
18868 @opindex mno-prolog-function
18869 @opindex mprolog-function
18870 Do not use (do use) external functions to save and restore registers
18871 at the prologue and epilogue of a function. The external functions
18872 are slower, but use less code space if more than one function saves
18873 the same number of registers. The @option{-mprolog-function} option
18874 is on by default if you optimize.
18878 Try to make the code as small as possible. At present, this just turns
18879 on the @option{-mep} and @option{-mprolog-function} options.
18881 @item -mtda=@var{n}
18883 Put static or global variables whose size is @var{n} bytes or less into
18884 the tiny data area that register @code{ep} points to. The tiny data
18885 area can hold up to 256 bytes in total (128 bytes for byte references).
18887 @item -msda=@var{n}
18889 Put static or global variables whose size is @var{n} bytes or less into
18890 the small data area that register @code{gp} points to. The small data
18891 area can hold up to 64 kilobytes.
18893 @item -mzda=@var{n}
18895 Put static or global variables whose size is @var{n} bytes or less into
18896 the first 32 kilobytes of memory.
18900 Specify that the target processor is the V850.
18903 @opindex mbig-switch
18904 Generate code suitable for big switch tables. Use this option only if
18905 the assembler/linker complain about out of range branches within a switch
18910 This option will cause r2 and r5 to be used in the code generated by
18911 the compiler. This setting is the default.
18913 @item -mno-app-regs
18914 @opindex mno-app-regs
18915 This option will cause r2 and r5 to be treated as fixed registers.
18919 Specify that the target processor is the V850E2V3. The preprocessor
18920 constants @samp{__v850e2v3__} will be defined if
18921 this option is used.
18925 Specify that the target processor is the V850E2. The preprocessor
18926 constants @samp{__v850e2__} will be defined if this option is used.
18930 Specify that the target processor is the V850E1. The preprocessor
18931 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
18932 this option is used.
18936 Specify that the target processor is the V850ES. This is an alias for
18937 the @option{-mv850e1} option.
18941 Specify that the target processor is the V850E@. The preprocessor
18942 constant @samp{__v850e__} will be defined if this option is used.
18944 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
18945 nor @option{-mv850e2} nor @option{-mv850e2v3}
18946 are defined then a default target processor will be chosen and the
18947 relevant @samp{__v850*__} preprocessor constant will be defined.
18949 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
18950 defined, regardless of which processor variant is the target.
18952 @item -mdisable-callt
18953 @opindex mdisable-callt
18954 This option will suppress generation of the CALLT instruction for the
18955 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
18956 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
18961 @subsection VAX Options
18962 @cindex VAX options
18964 These @samp{-m} options are defined for the VAX:
18969 Do not output certain jump instructions (@code{aobleq} and so on)
18970 that the Unix assembler for the VAX cannot handle across long
18975 Do output those jump instructions, on the assumption that you
18976 will assemble with the GNU assembler.
18980 Output code for G-format floating-point numbers instead of D-format.
18983 @node VxWorks Options
18984 @subsection VxWorks Options
18985 @cindex VxWorks Options
18987 The options in this section are defined for all VxWorks targets.
18988 Options specific to the target hardware are listed with the other
18989 options for that target.
18994 GCC can generate code for both VxWorks kernels and real time processes
18995 (RTPs). This option switches from the former to the latter. It also
18996 defines the preprocessor macro @code{__RTP__}.
18999 @opindex non-static
19000 Link an RTP executable against shared libraries rather than static
19001 libraries. The options @option{-static} and @option{-shared} can
19002 also be used for RTPs (@pxref{Link Options}); @option{-static}
19009 These options are passed down to the linker. They are defined for
19010 compatibility with Diab.
19013 @opindex Xbind-lazy
19014 Enable lazy binding of function calls. This option is equivalent to
19015 @option{-Wl,-z,now} and is defined for compatibility with Diab.
19019 Disable lazy binding of function calls. This option is the default and
19020 is defined for compatibility with Diab.
19023 @node x86-64 Options
19024 @subsection x86-64 Options
19025 @cindex x86-64 options
19027 These are listed under @xref{i386 and x86-64 Options}.
19029 @node Xstormy16 Options
19030 @subsection Xstormy16 Options
19031 @cindex Xstormy16 Options
19033 These options are defined for Xstormy16:
19038 Choose startup files and linker script suitable for the simulator.
19041 @node Xtensa Options
19042 @subsection Xtensa Options
19043 @cindex Xtensa Options
19045 These options are supported for Xtensa targets:
19049 @itemx -mno-const16
19051 @opindex mno-const16
19052 Enable or disable use of @code{CONST16} instructions for loading
19053 constant values. The @code{CONST16} instruction is currently not a
19054 standard option from Tensilica. When enabled, @code{CONST16}
19055 instructions are always used in place of the standard @code{L32R}
19056 instructions. The use of @code{CONST16} is enabled by default only if
19057 the @code{L32R} instruction is not available.
19060 @itemx -mno-fused-madd
19061 @opindex mfused-madd
19062 @opindex mno-fused-madd
19063 Enable or disable use of fused multiply/add and multiply/subtract
19064 instructions in the floating-point option. This has no effect if the
19065 floating-point option is not also enabled. Disabling fused multiply/add
19066 and multiply/subtract instructions forces the compiler to use separate
19067 instructions for the multiply and add/subtract operations. This may be
19068 desirable in some cases where strict IEEE 754-compliant results are
19069 required: the fused multiply add/subtract instructions do not round the
19070 intermediate result, thereby producing results with @emph{more} bits of
19071 precision than specified by the IEEE standard. Disabling fused multiply
19072 add/subtract instructions also ensures that the program output is not
19073 sensitive to the compiler's ability to combine multiply and add/subtract
19076 @item -mserialize-volatile
19077 @itemx -mno-serialize-volatile
19078 @opindex mserialize-volatile
19079 @opindex mno-serialize-volatile
19080 When this option is enabled, GCC inserts @code{MEMW} instructions before
19081 @code{volatile} memory references to guarantee sequential consistency.
19082 The default is @option{-mserialize-volatile}. Use
19083 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
19085 @item -mforce-no-pic
19086 @opindex mforce-no-pic
19087 For targets, like GNU/Linux, where all user-mode Xtensa code must be
19088 position-independent code (PIC), this option disables PIC for compiling
19091 @item -mtext-section-literals
19092 @itemx -mno-text-section-literals
19093 @opindex mtext-section-literals
19094 @opindex mno-text-section-literals
19095 Control the treatment of literal pools. The default is
19096 @option{-mno-text-section-literals}, which places literals in a separate
19097 section in the output file. This allows the literal pool to be placed
19098 in a data RAM/ROM, and it also allows the linker to combine literal
19099 pools from separate object files to remove redundant literals and
19100 improve code size. With @option{-mtext-section-literals}, the literals
19101 are interspersed in the text section in order to keep them as close as
19102 possible to their references. This may be necessary for large assembly
19105 @item -mtarget-align
19106 @itemx -mno-target-align
19107 @opindex mtarget-align
19108 @opindex mno-target-align
19109 When this option is enabled, GCC instructs the assembler to
19110 automatically align instructions to reduce branch penalties at the
19111 expense of some code density. The assembler attempts to widen density
19112 instructions to align branch targets and the instructions following call
19113 instructions. If there are not enough preceding safe density
19114 instructions to align a target, no widening will be performed. The
19115 default is @option{-mtarget-align}. These options do not affect the
19116 treatment of auto-aligned instructions like @code{LOOP}, which the
19117 assembler will always align, either by widening density instructions or
19118 by inserting no-op instructions.
19121 @itemx -mno-longcalls
19122 @opindex mlongcalls
19123 @opindex mno-longcalls
19124 When this option is enabled, GCC instructs the assembler to translate
19125 direct calls to indirect calls unless it can determine that the target
19126 of a direct call is in the range allowed by the call instruction. This
19127 translation typically occurs for calls to functions in other source
19128 files. Specifically, the assembler translates a direct @code{CALL}
19129 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
19130 The default is @option{-mno-longcalls}. This option should be used in
19131 programs where the call target can potentially be out of range. This
19132 option is implemented in the assembler, not the compiler, so the
19133 assembly code generated by GCC will still show direct call
19134 instructions---look at the disassembled object code to see the actual
19135 instructions. Note that the assembler will use an indirect call for
19136 every cross-file call, not just those that really will be out of range.
19139 @node zSeries Options
19140 @subsection zSeries Options
19141 @cindex zSeries options
19143 These are listed under @xref{S/390 and zSeries Options}.
19145 @node Code Gen Options
19146 @section Options for Code Generation Conventions
19147 @cindex code generation conventions
19148 @cindex options, code generation
19149 @cindex run-time options
19151 These machine-independent options control the interface conventions
19152 used in code generation.
19154 Most of them have both positive and negative forms; the negative form
19155 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
19156 one of the forms is listed---the one that is not the default. You
19157 can figure out the other form by either removing @samp{no-} or adding
19161 @item -fbounds-check
19162 @opindex fbounds-check
19163 For front ends that support it, generate additional code to check that
19164 indices used to access arrays are within the declared range. This is
19165 currently only supported by the Java and Fortran front ends, where
19166 this option defaults to true and false respectively.
19170 This option generates traps for signed overflow on addition, subtraction,
19171 multiplication operations.
19175 This option instructs the compiler to assume that signed arithmetic
19176 overflow of addition, subtraction and multiplication wraps around
19177 using twos-complement representation. This flag enables some optimizations
19178 and disables others. This option is enabled by default for the Java
19179 front end, as required by the Java language specification.
19182 @opindex fexceptions
19183 Enable exception handling. Generates extra code needed to propagate
19184 exceptions. For some targets, this implies GCC will generate frame
19185 unwind information for all functions, which can produce significant data
19186 size overhead, although it does not affect execution. If you do not
19187 specify this option, GCC will enable it by default for languages like
19188 C++ that normally require exception handling, and disable it for
19189 languages like C that do not normally require it. However, you may need
19190 to enable this option when compiling C code that needs to interoperate
19191 properly with exception handlers written in C++. You may also wish to
19192 disable this option if you are compiling older C++ programs that don't
19193 use exception handling.
19195 @item -fnon-call-exceptions
19196 @opindex fnon-call-exceptions
19197 Generate code that allows trapping instructions to throw exceptions.
19198 Note that this requires platform-specific runtime support that does
19199 not exist everywhere. Moreover, it only allows @emph{trapping}
19200 instructions to throw exceptions, i.e.@: memory references or floating-point
19201 instructions. It does not allow exceptions to be thrown from
19202 arbitrary signal handlers such as @code{SIGALRM}.
19204 @item -funwind-tables
19205 @opindex funwind-tables
19206 Similar to @option{-fexceptions}, except that it will just generate any needed
19207 static data, but will not affect the generated code in any other way.
19208 You will normally not enable this option; instead, a language processor
19209 that needs this handling would enable it on your behalf.
19211 @item -fasynchronous-unwind-tables
19212 @opindex fasynchronous-unwind-tables
19213 Generate unwind table in dwarf2 format, if supported by target machine. The
19214 table is exact at each instruction boundary, so it can be used for stack
19215 unwinding from asynchronous events (such as debugger or garbage collector).
19217 @item -fpcc-struct-return
19218 @opindex fpcc-struct-return
19219 Return ``short'' @code{struct} and @code{union} values in memory like
19220 longer ones, rather than in registers. This convention is less
19221 efficient, but it has the advantage of allowing intercallability between
19222 GCC-compiled files and files compiled with other compilers, particularly
19223 the Portable C Compiler (pcc).
19225 The precise convention for returning structures in memory depends
19226 on the target configuration macros.
19228 Short structures and unions are those whose size and alignment match
19229 that of some integer type.
19231 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
19232 switch is not binary compatible with code compiled with the
19233 @option{-freg-struct-return} switch.
19234 Use it to conform to a non-default application binary interface.
19236 @item -freg-struct-return
19237 @opindex freg-struct-return
19238 Return @code{struct} and @code{union} values in registers when possible.
19239 This is more efficient for small structures than
19240 @option{-fpcc-struct-return}.
19242 If you specify neither @option{-fpcc-struct-return} nor
19243 @option{-freg-struct-return}, GCC defaults to whichever convention is
19244 standard for the target. If there is no standard convention, GCC
19245 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
19246 the principal compiler. In those cases, we can choose the standard, and
19247 we chose the more efficient register return alternative.
19249 @strong{Warning:} code compiled with the @option{-freg-struct-return}
19250 switch is not binary compatible with code compiled with the
19251 @option{-fpcc-struct-return} switch.
19252 Use it to conform to a non-default application binary interface.
19254 @item -fshort-enums
19255 @opindex fshort-enums
19256 Allocate to an @code{enum} type only as many bytes as it needs for the
19257 declared range of possible values. Specifically, the @code{enum} type
19258 will be equivalent to the smallest integer type that has enough room.
19260 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
19261 code that is not binary compatible with code generated without that switch.
19262 Use it to conform to a non-default application binary interface.
19264 @item -fshort-double
19265 @opindex fshort-double
19266 Use the same size for @code{double} as for @code{float}.
19268 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
19269 code that is not binary compatible with code generated without that switch.
19270 Use it to conform to a non-default application binary interface.
19272 @item -fshort-wchar
19273 @opindex fshort-wchar
19274 Override the underlying type for @samp{wchar_t} to be @samp{short
19275 unsigned int} instead of the default for the target. This option is
19276 useful for building programs to run under WINE@.
19278 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
19279 code that is not binary compatible with code generated without that switch.
19280 Use it to conform to a non-default application binary interface.
19283 @opindex fno-common
19284 In C code, controls the placement of uninitialized global variables.
19285 Unix C compilers have traditionally permitted multiple definitions of
19286 such variables in different compilation units by placing the variables
19288 This is the behavior specified by @option{-fcommon}, and is the default
19289 for GCC on most targets.
19290 On the other hand, this behavior is not required by ISO C, and on some
19291 targets may carry a speed or code size penalty on variable references.
19292 The @option{-fno-common} option specifies that the compiler should place
19293 uninitialized global variables in the data section of the object file,
19294 rather than generating them as common blocks.
19295 This has the effect that if the same variable is declared
19296 (without @code{extern}) in two different compilations,
19297 you will get a multiple-definition error when you link them.
19298 In this case, you must compile with @option{-fcommon} instead.
19299 Compiling with @option{-fno-common} is useful on targets for which
19300 it provides better performance, or if you wish to verify that the
19301 program will work on other systems that always treat uninitialized
19302 variable declarations this way.
19306 Ignore the @samp{#ident} directive.
19308 @item -finhibit-size-directive
19309 @opindex finhibit-size-directive
19310 Don't output a @code{.size} assembler directive, or anything else that
19311 would cause trouble if the function is split in the middle, and the
19312 two halves are placed at locations far apart in memory. This option is
19313 used when compiling @file{crtstuff.c}; you should not need to use it
19316 @item -fverbose-asm
19317 @opindex fverbose-asm
19318 Put extra commentary information in the generated assembly code to
19319 make it more readable. This option is generally only of use to those
19320 who actually need to read the generated assembly code (perhaps while
19321 debugging the compiler itself).
19323 @option{-fno-verbose-asm}, the default, causes the
19324 extra information to be omitted and is useful when comparing two assembler
19327 @item -frecord-gcc-switches
19328 @opindex frecord-gcc-switches
19329 This switch causes the command line that was used to invoke the
19330 compiler to be recorded into the object file that is being created.
19331 This switch is only implemented on some targets and the exact format
19332 of the recording is target and binary file format dependent, but it
19333 usually takes the form of a section containing ASCII text. This
19334 switch is related to the @option{-fverbose-asm} switch, but that
19335 switch only records information in the assembler output file as
19336 comments, so it never reaches the object file.
19337 See also @option{-grecord-gcc-switches} for another
19338 way of storing compiler options into the object file.
19342 @cindex global offset table
19344 Generate position-independent code (PIC) suitable for use in a shared
19345 library, if supported for the target machine. Such code accesses all
19346 constant addresses through a global offset table (GOT)@. The dynamic
19347 loader resolves the GOT entries when the program starts (the dynamic
19348 loader is not part of GCC; it is part of the operating system). If
19349 the GOT size for the linked executable exceeds a machine-specific
19350 maximum size, you get an error message from the linker indicating that
19351 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
19352 instead. (These maximums are 8k on the SPARC and 32k
19353 on the m68k and RS/6000. The 386 has no such limit.)
19355 Position-independent code requires special support, and therefore works
19356 only on certain machines. For the 386, GCC supports PIC for System V
19357 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
19358 position-independent.
19360 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19365 If supported for the target machine, emit position-independent code,
19366 suitable for dynamic linking and avoiding any limit on the size of the
19367 global offset table. This option makes a difference on the m68k,
19368 PowerPC and SPARC@.
19370 Position-independent code requires special support, and therefore works
19371 only on certain machines.
19373 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
19380 These options are similar to @option{-fpic} and @option{-fPIC}, but
19381 generated position independent code can be only linked into executables.
19382 Usually these options are used when @option{-pie} GCC option will be
19383 used during linking.
19385 @option{-fpie} and @option{-fPIE} both define the macros
19386 @code{__pie__} and @code{__PIE__}. The macros have the value 1
19387 for @option{-fpie} and 2 for @option{-fPIE}.
19389 @item -fno-jump-tables
19390 @opindex fno-jump-tables
19391 Do not use jump tables for switch statements even where it would be
19392 more efficient than other code generation strategies. This option is
19393 of use in conjunction with @option{-fpic} or @option{-fPIC} for
19394 building code that forms part of a dynamic linker and cannot
19395 reference the address of a jump table. On some targets, jump tables
19396 do not require a GOT and this option is not needed.
19398 @item -ffixed-@var{reg}
19400 Treat the register named @var{reg} as a fixed register; generated code
19401 should never refer to it (except perhaps as a stack pointer, frame
19402 pointer or in some other fixed role).
19404 @var{reg} must be the name of a register. The register names accepted
19405 are machine-specific and are defined in the @code{REGISTER_NAMES}
19406 macro in the machine description macro file.
19408 This flag does not have a negative form, because it specifies a
19411 @item -fcall-used-@var{reg}
19412 @opindex fcall-used
19413 Treat the register named @var{reg} as an allocable register that is
19414 clobbered by function calls. It may be allocated for temporaries or
19415 variables that do not live across a call. Functions compiled this way
19416 will not save and restore the register @var{reg}.
19418 It is an error to used this flag with the frame pointer or stack pointer.
19419 Use of this flag for other registers that have fixed pervasive roles in
19420 the machine's execution model will produce disastrous results.
19422 This flag does not have a negative form, because it specifies a
19425 @item -fcall-saved-@var{reg}
19426 @opindex fcall-saved
19427 Treat the register named @var{reg} as an allocable register saved by
19428 functions. It may be allocated even for temporaries or variables that
19429 live across a call. Functions compiled this way will save and restore
19430 the register @var{reg} if they use it.
19432 It is an error to used this flag with the frame pointer or stack pointer.
19433 Use of this flag for other registers that have fixed pervasive roles in
19434 the machine's execution model will produce disastrous results.
19436 A different sort of disaster will result from the use of this flag for
19437 a register in which function values may be returned.
19439 This flag does not have a negative form, because it specifies a
19442 @item -fpack-struct[=@var{n}]
19443 @opindex fpack-struct
19444 Without a value specified, pack all structure members together without
19445 holes. When a value is specified (which must be a small power of two), pack
19446 structure members according to this value, representing the maximum
19447 alignment (that is, objects with default alignment requirements larger than
19448 this will be output potentially unaligned at the next fitting location.
19450 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
19451 code that is not binary compatible with code generated without that switch.
19452 Additionally, it makes the code suboptimal.
19453 Use it to conform to a non-default application binary interface.
19455 @item -finstrument-functions
19456 @opindex finstrument-functions
19457 Generate instrumentation calls for entry and exit to functions. Just
19458 after function entry and just before function exit, the following
19459 profiling functions will be called with the address of the current
19460 function and its call site. (On some platforms,
19461 @code{__builtin_return_address} does not work beyond the current
19462 function, so the call site information may not be available to the
19463 profiling functions otherwise.)
19466 void __cyg_profile_func_enter (void *this_fn,
19468 void __cyg_profile_func_exit (void *this_fn,
19472 The first argument is the address of the start of the current function,
19473 which may be looked up exactly in the symbol table.
19475 This instrumentation is also done for functions expanded inline in other
19476 functions. The profiling calls will indicate where, conceptually, the
19477 inline function is entered and exited. This means that addressable
19478 versions of such functions must be available. If all your uses of a
19479 function are expanded inline, this may mean an additional expansion of
19480 code size. If you use @samp{extern inline} in your C code, an
19481 addressable version of such functions must be provided. (This is
19482 normally the case anyways, but if you get lucky and the optimizer always
19483 expands the functions inline, you might have gotten away without
19484 providing static copies.)
19486 A function may be given the attribute @code{no_instrument_function}, in
19487 which case this instrumentation will not be done. This can be used, for
19488 example, for the profiling functions listed above, high-priority
19489 interrupt routines, and any functions from which the profiling functions
19490 cannot safely be called (perhaps signal handlers, if the profiling
19491 routines generate output or allocate memory).
19493 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
19494 @opindex finstrument-functions-exclude-file-list
19496 Set the list of functions that are excluded from instrumentation (see
19497 the description of @code{-finstrument-functions}). If the file that
19498 contains a function definition matches with one of @var{file}, then
19499 that function is not instrumented. The match is done on substrings:
19500 if the @var{file} parameter is a substring of the file name, it is
19501 considered to be a match.
19506 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
19510 will exclude any inline function defined in files whose pathnames
19511 contain @code{/bits/stl} or @code{include/sys}.
19513 If, for some reason, you want to include letter @code{','} in one of
19514 @var{sym}, write @code{'\,'}. For example,
19515 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
19516 (note the single quote surrounding the option).
19518 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
19519 @opindex finstrument-functions-exclude-function-list
19521 This is similar to @code{-finstrument-functions-exclude-file-list},
19522 but this option sets the list of function names to be excluded from
19523 instrumentation. The function name to be matched is its user-visible
19524 name, such as @code{vector<int> blah(const vector<int> &)}, not the
19525 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
19526 match is done on substrings: if the @var{sym} parameter is a substring
19527 of the function name, it is considered to be a match. For C99 and C++
19528 extended identifiers, the function name must be given in UTF-8, not
19529 using universal character names.
19531 @item -fstack-check
19532 @opindex fstack-check
19533 Generate code to verify that you do not go beyond the boundary of the
19534 stack. You should specify this flag if you are running in an
19535 environment with multiple threads, but only rarely need to specify it in
19536 a single-threaded environment since stack overflow is automatically
19537 detected on nearly all systems if there is only one stack.
19539 Note that this switch does not actually cause checking to be done; the
19540 operating system or the language runtime must do that. The switch causes
19541 generation of code to ensure that they see the stack being extended.
19543 You can additionally specify a string parameter: @code{no} means no
19544 checking, @code{generic} means force the use of old-style checking,
19545 @code{specific} means use the best checking method and is equivalent
19546 to bare @option{-fstack-check}.
19548 Old-style checking is a generic mechanism that requires no specific
19549 target support in the compiler but comes with the following drawbacks:
19553 Modified allocation strategy for large objects: they will always be
19554 allocated dynamically if their size exceeds a fixed threshold.
19557 Fixed limit on the size of the static frame of functions: when it is
19558 topped by a particular function, stack checking is not reliable and
19559 a warning is issued by the compiler.
19562 Inefficiency: because of both the modified allocation strategy and the
19563 generic implementation, the performances of the code are hampered.
19566 Note that old-style stack checking is also the fallback method for
19567 @code{specific} if no target support has been added in the compiler.
19569 @item -fstack-limit-register=@var{reg}
19570 @itemx -fstack-limit-symbol=@var{sym}
19571 @itemx -fno-stack-limit
19572 @opindex fstack-limit-register
19573 @opindex fstack-limit-symbol
19574 @opindex fno-stack-limit
19575 Generate code to ensure that the stack does not grow beyond a certain value,
19576 either the value of a register or the address of a symbol. If the stack
19577 would grow beyond the value, a signal is raised. For most targets,
19578 the signal is raised before the stack overruns the boundary, so
19579 it is possible to catch the signal without taking special precautions.
19581 For instance, if the stack starts at absolute address @samp{0x80000000}
19582 and grows downwards, you can use the flags
19583 @option{-fstack-limit-symbol=__stack_limit} and
19584 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
19585 of 128KB@. Note that this may only work with the GNU linker.
19587 @item -fsplit-stack
19588 @opindex fsplit-stack
19589 Generate code to automatically split the stack before it overflows.
19590 The resulting program has a discontiguous stack which can only
19591 overflow if the program is unable to allocate any more memory. This
19592 is most useful when running threaded programs, as it is no longer
19593 necessary to calculate a good stack size to use for each thread. This
19594 is currently only implemented for the i386 and x86_64 back ends running
19597 When code compiled with @option{-fsplit-stack} calls code compiled
19598 without @option{-fsplit-stack}, there may not be much stack space
19599 available for the latter code to run. If compiling all code,
19600 including library code, with @option{-fsplit-stack} is not an option,
19601 then the linker can fix up these calls so that the code compiled
19602 without @option{-fsplit-stack} always has a large stack. Support for
19603 this is implemented in the gold linker in GNU binutils release 2.21
19606 @item -fleading-underscore
19607 @opindex fleading-underscore
19608 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
19609 change the way C symbols are represented in the object file. One use
19610 is to help link with legacy assembly code.
19612 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
19613 generate code that is not binary compatible with code generated without that
19614 switch. Use it to conform to a non-default application binary interface.
19615 Not all targets provide complete support for this switch.
19617 @item -ftls-model=@var{model}
19618 @opindex ftls-model
19619 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
19620 The @var{model} argument should be one of @code{global-dynamic},
19621 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
19623 The default without @option{-fpic} is @code{initial-exec}; with
19624 @option{-fpic} the default is @code{global-dynamic}.
19626 @item -fvisibility=@var{default|internal|hidden|protected}
19627 @opindex fvisibility
19628 Set the default ELF image symbol visibility to the specified option---all
19629 symbols will be marked with this unless overridden within the code.
19630 Using this feature can very substantially improve linking and
19631 load times of shared object libraries, produce more optimized
19632 code, provide near-perfect API export and prevent symbol clashes.
19633 It is @strong{strongly} recommended that you use this in any shared objects
19636 Despite the nomenclature, @code{default} always means public; i.e.,
19637 available to be linked against from outside the shared object.
19638 @code{protected} and @code{internal} are pretty useless in real-world
19639 usage so the only other commonly used option will be @code{hidden}.
19640 The default if @option{-fvisibility} isn't specified is
19641 @code{default}, i.e., make every
19642 symbol public---this causes the same behavior as previous versions of
19645 A good explanation of the benefits offered by ensuring ELF
19646 symbols have the correct visibility is given by ``How To Write
19647 Shared Libraries'' by Ulrich Drepper (which can be found at
19648 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
19649 solution made possible by this option to marking things hidden when
19650 the default is public is to make the default hidden and mark things
19651 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
19652 and @code{__attribute__ ((visibility("default")))} instead of
19653 @code{__declspec(dllexport)} you get almost identical semantics with
19654 identical syntax. This is a great boon to those working with
19655 cross-platform projects.
19657 For those adding visibility support to existing code, you may find
19658 @samp{#pragma GCC visibility} of use. This works by you enclosing
19659 the declarations you wish to set visibility for with (for example)
19660 @samp{#pragma GCC visibility push(hidden)} and
19661 @samp{#pragma GCC visibility pop}.
19662 Bear in mind that symbol visibility should be viewed @strong{as
19663 part of the API interface contract} and thus all new code should
19664 always specify visibility when it is not the default; i.e., declarations
19665 only for use within the local DSO should @strong{always} be marked explicitly
19666 as hidden as so to avoid PLT indirection overheads---making this
19667 abundantly clear also aids readability and self-documentation of the code.
19668 Note that due to ISO C++ specification requirements, operator new and
19669 operator delete must always be of default visibility.
19671 Be aware that headers from outside your project, in particular system
19672 headers and headers from any other library you use, may not be
19673 expecting to be compiled with visibility other than the default. You
19674 may need to explicitly say @samp{#pragma GCC visibility push(default)}
19675 before including any such headers.
19677 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
19678 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
19679 no modifications. However, this means that calls to @samp{extern}
19680 functions with no explicit visibility will use the PLT, so it is more
19681 effective to use @samp{__attribute ((visibility))} and/or
19682 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
19683 declarations should be treated as hidden.
19685 Note that @samp{-fvisibility} does affect C++ vague linkage
19686 entities. This means that, for instance, an exception class that will
19687 be thrown between DSOs must be explicitly marked with default
19688 visibility so that the @samp{type_info} nodes will be unified between
19691 An overview of these techniques, their benefits and how to use them
19692 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
19694 @item -fstrict-volatile-bitfields
19695 @opindex fstrict-volatile-bitfields
19696 This option should be used if accesses to volatile bit-fields (or other
19697 structure fields, although the compiler usually honors those types
19698 anyway) should use a single access of the width of the
19699 field's type, aligned to a natural alignment if possible. For
19700 example, targets with memory-mapped peripheral registers might require
19701 all such accesses to be 16 bits wide; with this flag the user could
19702 declare all peripheral bit-fields as ``unsigned short'' (assuming short
19703 is 16 bits on these targets) to force GCC to use 16-bit accesses
19704 instead of, perhaps, a more efficient 32-bit access.
19706 If this option is disabled, the compiler will use the most efficient
19707 instruction. In the previous example, that might be a 32-bit load
19708 instruction, even though that will access bytes that do not contain
19709 any portion of the bit-field, or memory-mapped registers unrelated to
19710 the one being updated.
19712 If the target requires strict alignment, and honoring the field
19713 type would require violating this alignment, a warning is issued.
19714 If the field has @code{packed} attribute, the access is done without
19715 honoring the field type. If the field doesn't have @code{packed}
19716 attribute, the access is done honoring the field type. In both cases,
19717 GCC assumes that the user knows something about the target hardware
19718 that it is unaware of.
19720 The default value of this option is determined by the application binary
19721 interface for the target processor.
19727 @node Environment Variables
19728 @section Environment Variables Affecting GCC
19729 @cindex environment variables
19731 @c man begin ENVIRONMENT
19732 This section describes several environment variables that affect how GCC
19733 operates. Some of them work by specifying directories or prefixes to use
19734 when searching for various kinds of files. Some are used to specify other
19735 aspects of the compilation environment.
19737 Note that you can also specify places to search using options such as
19738 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
19739 take precedence over places specified using environment variables, which
19740 in turn take precedence over those specified by the configuration of GCC@.
19741 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
19742 GNU Compiler Collection (GCC) Internals}.
19747 @c @itemx LC_COLLATE
19749 @c @itemx LC_MONETARY
19750 @c @itemx LC_NUMERIC
19755 @c @findex LC_COLLATE
19756 @findex LC_MESSAGES
19757 @c @findex LC_MONETARY
19758 @c @findex LC_NUMERIC
19762 These environment variables control the way that GCC uses
19763 localization information which allows GCC to work with different
19764 national conventions. GCC inspects the locale categories
19765 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
19766 so. These locale categories can be set to any value supported by your
19767 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
19768 Kingdom encoded in UTF-8.
19770 The @env{LC_CTYPE} environment variable specifies character
19771 classification. GCC uses it to determine the character boundaries in
19772 a string; this is needed for some multibyte encodings that contain quote
19773 and escape characters that would otherwise be interpreted as a string
19776 The @env{LC_MESSAGES} environment variable specifies the language to
19777 use in diagnostic messages.
19779 If the @env{LC_ALL} environment variable is set, it overrides the value
19780 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
19781 and @env{LC_MESSAGES} default to the value of the @env{LANG}
19782 environment variable. If none of these variables are set, GCC
19783 defaults to traditional C English behavior.
19787 If @env{TMPDIR} is set, it specifies the directory to use for temporary
19788 files. GCC uses temporary files to hold the output of one stage of
19789 compilation which is to be used as input to the next stage: for example,
19790 the output of the preprocessor, which is the input to the compiler
19793 @item GCC_COMPARE_DEBUG
19794 @findex GCC_COMPARE_DEBUG
19795 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
19796 @option{-fcompare-debug} to the compiler driver. See the documentation
19797 of this option for more details.
19799 @item GCC_EXEC_PREFIX
19800 @findex GCC_EXEC_PREFIX
19801 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
19802 names of the subprograms executed by the compiler. No slash is added
19803 when this prefix is combined with the name of a subprogram, but you can
19804 specify a prefix that ends with a slash if you wish.
19806 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
19807 an appropriate prefix to use based on the pathname it was invoked with.
19809 If GCC cannot find the subprogram using the specified prefix, it
19810 tries looking in the usual places for the subprogram.
19812 The default value of @env{GCC_EXEC_PREFIX} is
19813 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
19814 the installed compiler. In many cases @var{prefix} is the value
19815 of @code{prefix} when you ran the @file{configure} script.
19817 Other prefixes specified with @option{-B} take precedence over this prefix.
19819 This prefix is also used for finding files such as @file{crt0.o} that are
19822 In addition, the prefix is used in an unusual way in finding the
19823 directories to search for header files. For each of the standard
19824 directories whose name normally begins with @samp{/usr/local/lib/gcc}
19825 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
19826 replacing that beginning with the specified prefix to produce an
19827 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
19828 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
19829 These alternate directories are searched first; the standard directories
19830 come next. If a standard directory begins with the configured
19831 @var{prefix} then the value of @var{prefix} is replaced by
19832 @env{GCC_EXEC_PREFIX} when looking for header files.
19834 @item COMPILER_PATH
19835 @findex COMPILER_PATH
19836 The value of @env{COMPILER_PATH} is a colon-separated list of
19837 directories, much like @env{PATH}. GCC tries the directories thus
19838 specified when searching for subprograms, if it can't find the
19839 subprograms using @env{GCC_EXEC_PREFIX}.
19842 @findex LIBRARY_PATH
19843 The value of @env{LIBRARY_PATH} is a colon-separated list of
19844 directories, much like @env{PATH}. When configured as a native compiler,
19845 GCC tries the directories thus specified when searching for special
19846 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
19847 using GCC also uses these directories when searching for ordinary
19848 libraries for the @option{-l} option (but directories specified with
19849 @option{-L} come first).
19853 @cindex locale definition
19854 This variable is used to pass locale information to the compiler. One way in
19855 which this information is used is to determine the character set to be used
19856 when character literals, string literals and comments are parsed in C and C++.
19857 When the compiler is configured to allow multibyte characters,
19858 the following values for @env{LANG} are recognized:
19862 Recognize JIS characters.
19864 Recognize SJIS characters.
19866 Recognize EUCJP characters.
19869 If @env{LANG} is not defined, or if it has some other value, then the
19870 compiler will use mblen and mbtowc as defined by the default locale to
19871 recognize and translate multibyte characters.
19875 Some additional environments variables affect the behavior of the
19878 @include cppenv.texi
19882 @node Precompiled Headers
19883 @section Using Precompiled Headers
19884 @cindex precompiled headers
19885 @cindex speed of compilation
19887 Often large projects have many header files that are included in every
19888 source file. The time the compiler takes to process these header files
19889 over and over again can account for nearly all of the time required to
19890 build the project. To make builds faster, GCC allows users to
19891 `precompile' a header file; then, if builds can use the precompiled
19892 header file they will be much faster.
19894 To create a precompiled header file, simply compile it as you would any
19895 other file, if necessary using the @option{-x} option to make the driver
19896 treat it as a C or C++ header file. You will probably want to use a
19897 tool like @command{make} to keep the precompiled header up-to-date when
19898 the headers it contains change.
19900 A precompiled header file will be searched for when @code{#include} is
19901 seen in the compilation. As it searches for the included file
19902 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
19903 compiler looks for a precompiled header in each directory just before it
19904 looks for the include file in that directory. The name searched for is
19905 the name specified in the @code{#include} with @samp{.gch} appended. If
19906 the precompiled header file can't be used, it is ignored.
19908 For instance, if you have @code{#include "all.h"}, and you have
19909 @file{all.h.gch} in the same directory as @file{all.h}, then the
19910 precompiled header file will be used if possible, and the original
19911 header will be used otherwise.
19913 Alternatively, you might decide to put the precompiled header file in a
19914 directory and use @option{-I} to ensure that directory is searched
19915 before (or instead of) the directory containing the original header.
19916 Then, if you want to check that the precompiled header file is always
19917 used, you can put a file of the same name as the original header in this
19918 directory containing an @code{#error} command.
19920 This also works with @option{-include}. So yet another way to use
19921 precompiled headers, good for projects not designed with precompiled
19922 header files in mind, is to simply take most of the header files used by
19923 a project, include them from another header file, precompile that header
19924 file, and @option{-include} the precompiled header. If the header files
19925 have guards against multiple inclusion, they will be skipped because
19926 they've already been included (in the precompiled header).
19928 If you need to precompile the same header file for different
19929 languages, targets, or compiler options, you can instead make a
19930 @emph{directory} named like @file{all.h.gch}, and put each precompiled
19931 header in the directory, perhaps using @option{-o}. It doesn't matter
19932 what you call the files in the directory, every precompiled header in
19933 the directory will be considered. The first precompiled header
19934 encountered in the directory that is valid for this compilation will
19935 be used; they're searched in no particular order.
19937 There are many other possibilities, limited only by your imagination,
19938 good sense, and the constraints of your build system.
19940 A precompiled header file can be used only when these conditions apply:
19944 Only one precompiled header can be used in a particular compilation.
19947 A precompiled header can't be used once the first C token is seen. You
19948 can have preprocessor directives before a precompiled header; you can
19949 even include a precompiled header from inside another header, so long as
19950 there are no C tokens before the @code{#include}.
19953 The precompiled header file must be produced for the same language as
19954 the current compilation. You can't use a C precompiled header for a C++
19958 The precompiled header file must have been produced by the same compiler
19959 binary as the current compilation is using.
19962 Any macros defined before the precompiled header is included must
19963 either be defined in the same way as when the precompiled header was
19964 generated, or must not affect the precompiled header, which usually
19965 means that they don't appear in the precompiled header at all.
19967 The @option{-D} option is one way to define a macro before a
19968 precompiled header is included; using a @code{#define} can also do it.
19969 There are also some options that define macros implicitly, like
19970 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
19973 @item If debugging information is output when using the precompiled
19974 header, using @option{-g} or similar, the same kind of debugging information
19975 must have been output when building the precompiled header. However,
19976 a precompiled header built using @option{-g} can be used in a compilation
19977 when no debugging information is being output.
19979 @item The same @option{-m} options must generally be used when building
19980 and using the precompiled header. @xref{Submodel Options},
19981 for any cases where this rule is relaxed.
19983 @item Each of the following options must be the same when building and using
19984 the precompiled header:
19986 @gccoptlist{-fexceptions}
19989 Some other command-line options starting with @option{-f},
19990 @option{-p}, or @option{-O} must be defined in the same way as when
19991 the precompiled header was generated. At present, it's not clear
19992 which options are safe to change and which are not; the safest choice
19993 is to use exactly the same options when generating and using the
19994 precompiled header. The following are known to be safe:
19996 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
19997 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
19998 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
20003 For all of these except the last, the compiler will automatically
20004 ignore the precompiled header if the conditions aren't met. If you
20005 find an option combination that doesn't work and doesn't cause the
20006 precompiled header to be ignored, please consider filing a bug report,
20009 If you do use differing options when generating and using the
20010 precompiled header, the actual behavior will be a mixture of the
20011 behavior for the options. For instance, if you use @option{-g} to
20012 generate the precompiled header but not when using it, you may or may
20013 not get debugging information for routines in the precompiled header.