1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.3 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
168 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} @gol
174 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
175 -fhosted -ffreestanding -fopenmp -fms-extensions -fplan9-extensions @gol
176 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
177 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
178 -fsigned-bitfields -fsigned-char @gol
179 -funsigned-bitfields -funsigned-char}
181 @item C++ Language Options
182 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
183 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
184 -fconserve-space -fconstexpr-depth=@var{n} -ffriend-injection @gol
185 -fno-elide-constructors @gol
186 -fno-enforce-eh-specs @gol
187 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
188 -fno-implicit-templates @gol
189 -fno-implicit-inline-templates @gol
190 -fno-implement-inlines -fms-extensions @gol
191 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
192 -fno-optional-diags -fpermissive @gol
193 -fno-pretty-templates @gol
194 -frepo -fno-rtti -fstats -ftemplate-depth=@var{n} @gol
195 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
196 -fno-default-inline -fvisibility-inlines-hidden @gol
197 -fvisibility-ms-compat @gol
198 -Wabi -Wconversion-null -Wctor-dtor-privacy @gol
199 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
200 -Weffc++ -Wstrict-null-sentinel @gol
201 -Wno-non-template-friend -Wold-style-cast @gol
202 -Woverloaded-virtual -Wno-pmf-conversions @gol
205 @item Objective-C and Objective-C++ Language Options
206 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
207 Objective-C and Objective-C++ Dialects}.
208 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
209 -fgnu-runtime -fnext-runtime @gol
210 -fno-nil-receivers @gol
211 -fobjc-abi-version=@var{n} @gol
212 -fobjc-call-cxx-cdtors @gol
213 -fobjc-direct-dispatch @gol
214 -fobjc-exceptions @gol
217 -fobjc-std=objc1 @gol
218 -freplace-objc-classes @gol
221 -Wassign-intercept @gol
222 -Wno-protocol -Wselector @gol
223 -Wstrict-selector-match @gol
224 -Wundeclared-selector}
226 @item Language Independent Options
227 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
228 @gccoptlist{-fmessage-length=@var{n} @gol
229 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
230 -fno-diagnostics-show-option}
232 @item Warning Options
233 @xref{Warning Options,,Options to Request or Suppress Warnings}.
234 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -pedantic @gol
235 -pedantic-errors @gol
236 -w -Wextra -Wall -Waddress -Waggregate-return -Warray-bounds @gol
237 -Wno-attributes -Wno-builtin-macro-redefined @gol
238 -Wc++-compat -Wc++0x-compat -Wcast-align -Wcast-qual @gol
239 -Wchar-subscripts -Wclobbered -Wcomment @gol
240 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wno-deprecated @gol
241 -Wno-deprecated-declarations -Wdisabled-optimization @gol
242 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
243 -Wno-endif-labels -Werror -Werror=* @gol
244 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
245 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
246 -Wformat-security -Wformat-y2k @gol
247 -Wframe-larger-than=@var{len} -Wjump-misses-init -Wignored-qualifiers @gol
248 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
249 -Winit-self -Winline -Wmaybe-uninitialized @gol
250 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
251 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
252 -Wlogical-op -Wlong-long @gol
253 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
254 -Wmissing-format-attribute -Wmissing-include-dirs @gol
256 -Wno-multichar -Wnonnull -Wno-overflow @gol
257 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
258 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
259 -Wpointer-arith -Wno-pointer-to-int-cast @gol
260 -Wredundant-decls @gol
261 -Wreturn-type -Wsequence-point -Wshadow @gol
262 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
263 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
264 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
265 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
266 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
267 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
268 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
269 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
270 -Wunused-label -Wunused-parameter -Wno-unused-result -Wunused-value @gol
271 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
272 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
274 @item C and Objective-C-only Warning Options
275 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
276 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
277 -Wold-style-declaration -Wold-style-definition @gol
278 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
279 -Wdeclaration-after-statement -Wpointer-sign}
281 @item Debugging Options
282 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
283 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
284 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
285 -fdisable-ipa-@var{pass_name} @gol
286 -fdisable-rtl-@var{pass_name} @gol
287 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
288 -fdisable-tree-@var{pass_name} @gol
289 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
290 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
291 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
292 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
293 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
295 -fdump-statistics @gol
297 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
298 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
301 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
302 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-nrv -fdump-tree-vect @gol
311 -fdump-tree-sink @gol
312 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
313 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
316 -ftree-vectorizer-verbose=@var{n} @gol
317 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
318 -fdump-final-insns=@var{file} @gol
319 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
320 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
321 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
322 -fenable-@var{kind}-@var{pass} @gol
323 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
324 -fdebug-types-section @gol
325 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
326 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
327 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
328 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
329 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
330 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
331 -ggdb -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
332 -gvms -gxcoff -gxcoff+ @gol
333 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
334 -fdebug-prefix-map=@var{old}=@var{new} @gol
335 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
336 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
337 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
338 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
339 -print-prog-name=@var{program} -print-search-dirs -Q @gol
340 -print-sysroot -print-sysroot-headers-suffix @gol
341 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
343 @item Optimization Options
344 @xref{Optimize Options,,Options that Control Optimization}.
345 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
346 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
347 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
348 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
349 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
350 -fcompare-elim -fcprop-registers -fcrossjumping @gol
351 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
352 -fcx-limited-range @gol
353 -fdata-sections -fdce -fdce -fdelayed-branch @gol
354 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
355 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
356 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
357 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
358 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
359 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
360 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
361 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
362 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
363 -fira-algorithm=@var{algorithm} @gol
364 -fira-region=@var{region} @gol
365 -fira-loop-pressure -fno-ira-share-save-slots @gol
366 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
367 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
368 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
369 -floop-parallelize-all -flto -flto-compression-level @gol
370 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
371 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
372 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
373 -fno-default-inline @gol
374 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
375 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
376 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
377 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
378 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
379 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
380 -fprefetch-loop-arrays @gol
381 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
382 -fprofile-generate=@var{path} @gol
383 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
384 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
385 -freorder-blocks-and-partition -freorder-functions @gol
386 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
387 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
388 -fsched-spec-load -fsched-spec-load-dangerous @gol
389 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
390 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
391 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
392 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
393 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
394 -fselective-scheduling -fselective-scheduling2 @gol
395 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
396 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
397 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
398 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
400 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
401 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
402 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
403 -ftree-loop-if-convert-stores -ftree-loop-im @gol
404 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
405 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
406 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
407 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
408 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
409 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
410 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
411 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
412 -fwhole-program -fwpa -fuse-linker-plugin @gol
413 --param @var{name}=@var{value}
414 -O -O0 -O1 -O2 -O3 -Os -Ofast}
416 @item Preprocessor Options
417 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
418 @gccoptlist{-A@var{question}=@var{answer} @gol
419 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
420 -C -dD -dI -dM -dN @gol
421 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
422 -idirafter @var{dir} @gol
423 -include @var{file} -imacros @var{file} @gol
424 -iprefix @var{file} -iwithprefix @var{dir} @gol
425 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
426 -imultilib @var{dir} -isysroot @var{dir} @gol
427 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
428 -P -fworking-directory -remap @gol
429 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
430 -Xpreprocessor @var{option}}
432 @item Assembler Option
433 @xref{Assembler Options,,Passing Options to the Assembler}.
434 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
437 @xref{Link Options,,Options for Linking}.
438 @gccoptlist{@var{object-file-name} -l@var{library} @gol
439 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
440 -s -static -static-libgcc -static-libstdc++ -shared @gol
441 -shared-libgcc -symbolic @gol
442 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
445 @item Directory Options
446 @xref{Directory Options,,Options for Directory Search}.
447 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
448 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
451 @item Machine Dependent Options
452 @xref{Submodel Options,,Hardware Models and Configurations}.
453 @c This list is ordered alphanumerically by subsection name.
454 @c Try and put the significant identifier (CPU or system) first,
455 @c so users have a clue at guessing where the ones they want will be.
458 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
459 -mabi=@var{name} @gol
460 -mapcs-stack-check -mno-apcs-stack-check @gol
461 -mapcs-float -mno-apcs-float @gol
462 -mapcs-reentrant -mno-apcs-reentrant @gol
463 -msched-prolog -mno-sched-prolog @gol
464 -mlittle-endian -mbig-endian -mwords-little-endian @gol
465 -mfloat-abi=@var{name} -mfpe @gol
466 -mfp16-format=@var{name}
467 -mthumb-interwork -mno-thumb-interwork @gol
468 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
469 -mstructure-size-boundary=@var{n} @gol
470 -mabort-on-noreturn @gol
471 -mlong-calls -mno-long-calls @gol
472 -msingle-pic-base -mno-single-pic-base @gol
473 -mpic-register=@var{reg} @gol
474 -mnop-fun-dllimport @gol
475 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
476 -mpoke-function-name @gol
478 -mtpcs-frame -mtpcs-leaf-frame @gol
479 -mcaller-super-interworking -mcallee-super-interworking @gol
480 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
481 -mword-relocations @gol
482 -mfix-cortex-m3-ldrd}
485 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
486 -mcall-prologues -mtiny-stack -mint8}
488 @emph{Blackfin Options}
489 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
490 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
491 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
492 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
493 -mno-id-shared-library -mshared-library-id=@var{n} @gol
494 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
495 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
496 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
500 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
501 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
502 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
503 -mstack-align -mdata-align -mconst-align @gol
504 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
505 -melf -maout -melinux -mlinux -sim -sim2 @gol
506 -mmul-bug-workaround -mno-mul-bug-workaround}
508 @emph{Darwin Options}
509 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
510 -arch_only -bind_at_load -bundle -bundle_loader @gol
511 -client_name -compatibility_version -current_version @gol
513 -dependency-file -dylib_file -dylinker_install_name @gol
514 -dynamic -dynamiclib -exported_symbols_list @gol
515 -filelist -flat_namespace -force_cpusubtype_ALL @gol
516 -force_flat_namespace -headerpad_max_install_names @gol
518 -image_base -init -install_name -keep_private_externs @gol
519 -multi_module -multiply_defined -multiply_defined_unused @gol
520 -noall_load -no_dead_strip_inits_and_terms @gol
521 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
522 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
523 -private_bundle -read_only_relocs -sectalign @gol
524 -sectobjectsymbols -whyload -seg1addr @gol
525 -sectcreate -sectobjectsymbols -sectorder @gol
526 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
527 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
528 -segprot -segs_read_only_addr -segs_read_write_addr @gol
529 -single_module -static -sub_library -sub_umbrella @gol
530 -twolevel_namespace -umbrella -undefined @gol
531 -unexported_symbols_list -weak_reference_mismatches @gol
532 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
533 -mkernel -mone-byte-bool}
535 @emph{DEC Alpha Options}
536 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
537 -mieee -mieee-with-inexact -mieee-conformant @gol
538 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
539 -mtrap-precision=@var{mode} -mbuild-constants @gol
540 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
541 -mbwx -mmax -mfix -mcix @gol
542 -mfloat-vax -mfloat-ieee @gol
543 -mexplicit-relocs -msmall-data -mlarge-data @gol
544 -msmall-text -mlarge-text @gol
545 -mmemory-latency=@var{time}}
547 @emph{DEC Alpha/VMS Options}
548 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
551 @gccoptlist{-msmall-model -mno-lsim}
554 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
555 -mhard-float -msoft-float @gol
556 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
557 -mdouble -mno-double @gol
558 -mmedia -mno-media -mmuladd -mno-muladd @gol
559 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
560 -mlinked-fp -mlong-calls -malign-labels @gol
561 -mlibrary-pic -macc-4 -macc-8 @gol
562 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
563 -moptimize-membar -mno-optimize-membar @gol
564 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
565 -mvliw-branch -mno-vliw-branch @gol
566 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
567 -mno-nested-cond-exec -mtomcat-stats @gol
571 @emph{GNU/Linux Options}
572 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
573 -tno-android-cc -tno-android-ld}
575 @emph{H8/300 Options}
576 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
579 @gccoptlist{-march=@var{architecture-type} @gol
580 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
581 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
582 -mfixed-range=@var{register-range} @gol
583 -mjump-in-delay -mlinker-opt -mlong-calls @gol
584 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
585 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
586 -mno-jump-in-delay -mno-long-load-store @gol
587 -mno-portable-runtime -mno-soft-float @gol
588 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
589 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
590 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
591 -munix=@var{unix-std} -nolibdld -static -threads}
593 @emph{i386 and x86-64 Options}
594 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
595 -mfpmath=@var{unit} @gol
596 -masm=@var{dialect} -mno-fancy-math-387 @gol
597 -mno-fp-ret-in-387 -msoft-float @gol
598 -mno-wide-multiply -mrtd -malign-double @gol
599 -mpreferred-stack-boundary=@var{num} @gol
600 -mincoming-stack-boundary=@var{num} @gol
601 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
602 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
603 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
604 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlwp @gol
605 -mthreads -mno-align-stringops -minline-all-stringops @gol
606 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
607 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
608 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
609 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
610 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
611 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
612 -mcmodel=@var{code-model} -mabi=@var{name} @gol
613 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
614 -msse2avx -mfentry -m8bit-idiv @gol
615 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
617 @emph{i386 and x86-64 Windows Options}
618 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
619 -mnop-fun-dllimport -mthread @gol
620 -municode -mwin32 -mwindows -fno-set-stack-executable}
623 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
624 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
625 -mconstant-gp -mauto-pic -mfused-madd @gol
626 -minline-float-divide-min-latency @gol
627 -minline-float-divide-max-throughput @gol
628 -mno-inline-float-divide @gol
629 -minline-int-divide-min-latency @gol
630 -minline-int-divide-max-throughput @gol
631 -mno-inline-int-divide @gol
632 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
633 -mno-inline-sqrt @gol
634 -mdwarf2-asm -mearly-stop-bits @gol
635 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
636 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
637 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
638 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
639 -msched-spec-ldc -msched-spec-control-ldc @gol
640 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
641 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
642 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
643 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
645 @emph{IA-64/VMS Options}
646 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
649 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
650 -msign-extend-enabled -muser-enabled}
652 @emph{M32R/D Options}
653 @gccoptlist{-m32r2 -m32rx -m32r @gol
655 -malign-loops -mno-align-loops @gol
656 -missue-rate=@var{number} @gol
657 -mbranch-cost=@var{number} @gol
658 -mmodel=@var{code-size-model-type} @gol
659 -msdata=@var{sdata-type} @gol
660 -mno-flush-func -mflush-func=@var{name} @gol
661 -mno-flush-trap -mflush-trap=@var{number} @gol
665 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
667 @emph{M680x0 Options}
668 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
669 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
670 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
671 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
672 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
673 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
674 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
675 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
679 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
680 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
681 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
682 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
683 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
686 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
687 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
688 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
689 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
692 @emph{MicroBlaze Options}
693 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
694 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
695 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
696 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
697 -mxl-mode-@var{app-model}}
700 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
701 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
702 -mips64 -mips64r2 @gol
703 -mips16 -mno-mips16 -mflip-mips16 @gol
704 -minterlink-mips16 -mno-interlink-mips16 @gol
705 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
706 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
707 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
708 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
709 -mfpu=@var{fpu-type} @gol
710 -msmartmips -mno-smartmips @gol
711 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
712 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
713 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
714 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
715 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
716 -membedded-data -mno-embedded-data @gol
717 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
718 -mcode-readable=@var{setting} @gol
719 -msplit-addresses -mno-split-addresses @gol
720 -mexplicit-relocs -mno-explicit-relocs @gol
721 -mcheck-zero-division -mno-check-zero-division @gol
722 -mdivide-traps -mdivide-breaks @gol
723 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
724 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
725 -mfix-24k -mno-fix-24k @gol
726 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
727 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
728 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
729 -mflush-func=@var{func} -mno-flush-func @gol
730 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
731 -mfp-exceptions -mno-fp-exceptions @gol
732 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
733 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
736 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
737 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
738 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
739 -mno-base-addresses -msingle-exit -mno-single-exit}
741 @emph{MN10300 Options}
742 @gccoptlist{-mmult-bug -mno-mult-bug @gol
743 -mno-am33 -mam33 -mam33-2 -mam34 @gol
744 -mtune=@var{cpu-type} @gol
745 -mreturn-pointer-on-d0 @gol
746 -mno-crt0 -mrelax -mliw -msetlb}
748 @emph{PDP-11 Options}
749 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
750 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
751 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
752 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
753 -mbranch-expensive -mbranch-cheap @gol
754 -munix-asm -mdec-asm}
756 @emph{picoChip Options}
757 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
758 -msymbol-as-address -mno-inefficient-warnings}
760 @emph{PowerPC Options}
761 See RS/6000 and PowerPC Options.
763 @emph{RS/6000 and PowerPC Options}
764 @gccoptlist{-mcpu=@var{cpu-type} @gol
765 -mtune=@var{cpu-type} @gol
766 -mcmodel=@var{code-model} @gol
767 -mpower -mno-power -mpower2 -mno-power2 @gol
768 -mpowerpc -mpowerpc64 -mno-powerpc @gol
769 -maltivec -mno-altivec @gol
770 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
771 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
772 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
773 -mfprnd -mno-fprnd @gol
774 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
775 -mnew-mnemonics -mold-mnemonics @gol
776 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
777 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
778 -malign-power -malign-natural @gol
779 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
780 -msingle-float -mdouble-float -msimple-fpu @gol
781 -mstring -mno-string -mupdate -mno-update @gol
782 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
783 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
784 -mstrict-align -mno-strict-align -mrelocatable @gol
785 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
786 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
787 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
788 -mprioritize-restricted-insns=@var{priority} @gol
789 -msched-costly-dep=@var{dependence_type} @gol
790 -minsert-sched-nops=@var{scheme} @gol
791 -mcall-sysv -mcall-netbsd @gol
792 -maix-struct-return -msvr4-struct-return @gol
793 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
794 -mblock-move-inline-limit=@var{num} @gol
795 -misel -mno-isel @gol
796 -misel=yes -misel=no @gol
798 -mspe=yes -mspe=no @gol
800 -mgen-cell-microcode -mwarn-cell-microcode @gol
801 -mvrsave -mno-vrsave @gol
802 -mmulhw -mno-mulhw @gol
803 -mdlmzb -mno-dlmzb @gol
804 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
805 -mprototype -mno-prototype @gol
806 -msim -mmvme -mads -myellowknife -memb -msdata @gol
807 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
808 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
809 -mno-recip-precision @gol
810 -mveclibabi=@var{type} -mfriz -mno-friz -mr11 -mno-r11}
813 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
815 -mbig-endian-data -mlittle-endian-data @gol
818 -mas100-syntax -mno-as100-syntax@gol
820 -mmax-constant-size=@gol
822 -msave-acc-in-interrupts}
824 @emph{S/390 and zSeries Options}
825 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
826 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
827 -mlong-double-64 -mlong-double-128 @gol
828 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
829 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
830 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
831 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
832 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
835 @gccoptlist{-meb -mel @gol
839 -mscore5 -mscore5u -mscore7 -mscore7d}
842 @gccoptlist{-m1 -m2 -m2e @gol
843 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
845 -m4-nofpu -m4-single-only -m4-single -m4 @gol
846 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
847 -m5-64media -m5-64media-nofpu @gol
848 -m5-32media -m5-32media-nofpu @gol
849 -m5-compact -m5-compact-nofpu @gol
850 -mb -ml -mdalign -mrelax @gol
851 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
852 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
853 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
854 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
855 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
856 -maccumulate-outgoing-args -minvalid-symbols}
858 @emph{Solaris 2 Options}
859 @gccoptlist{-mimpure-text -mno-impure-text @gol
863 @gccoptlist{-mcpu=@var{cpu-type} @gol
864 -mtune=@var{cpu-type} @gol
865 -mcmodel=@var{code-model} @gol
866 -m32 -m64 -mapp-regs -mno-app-regs @gol
867 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
868 -mfpu -mno-fpu -mhard-float -msoft-float @gol
869 -mhard-quad-float -msoft-quad-float @gol
871 -mstack-bias -mno-stack-bias @gol
872 -munaligned-doubles -mno-unaligned-doubles @gol
873 -mv8plus -mno-v8plus -mvis -mno-vis}
876 @gccoptlist{-mwarn-reloc -merror-reloc @gol
877 -msafe-dma -munsafe-dma @gol
879 -msmall-mem -mlarge-mem -mstdmain @gol
880 -mfixed-range=@var{register-range} @gol
882 -maddress-space-conversion -mno-address-space-conversion @gol
883 -mcache-size=@var{cache-size} @gol
884 -matomic-updates -mno-atomic-updates}
886 @emph{System V Options}
887 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
890 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
891 -mprolog-function -mno-prolog-function -mspace @gol
892 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
893 -mapp-regs -mno-app-regs @gol
894 -mdisable-callt -mno-disable-callt @gol
897 -mv850e1 -mv850es @gol
902 @gccoptlist{-mg -mgnu -munix}
904 @emph{VxWorks Options}
905 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
906 -Xbind-lazy -Xbind-now}
908 @emph{x86-64 Options}
909 See i386 and x86-64 Options.
911 @emph{Xstormy16 Options}
914 @emph{Xtensa Options}
915 @gccoptlist{-mconst16 -mno-const16 @gol
916 -mfused-madd -mno-fused-madd @gol
918 -mserialize-volatile -mno-serialize-volatile @gol
919 -mtext-section-literals -mno-text-section-literals @gol
920 -mtarget-align -mno-target-align @gol
921 -mlongcalls -mno-longcalls}
923 @emph{zSeries Options}
924 See S/390 and zSeries Options.
926 @item Code Generation Options
927 @xref{Code Gen Options,,Options for Code Generation Conventions}.
928 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
929 -ffixed-@var{reg} -fexceptions @gol
930 -fnon-call-exceptions -funwind-tables @gol
931 -fasynchronous-unwind-tables @gol
932 -finhibit-size-directive -finstrument-functions @gol
933 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
934 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
935 -fno-common -fno-ident @gol
936 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
937 -fno-jump-tables @gol
938 -frecord-gcc-switches @gol
939 -freg-struct-return -fshort-enums @gol
940 -fshort-double -fshort-wchar @gol
941 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
942 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
943 -fno-stack-limit -fsplit-stack @gol
944 -fleading-underscore -ftls-model=@var{model} @gol
945 -ftrapv -fwrapv -fbounds-check @gol
946 -fvisibility -fstrict-volatile-bitfields}
950 * Overall Options:: Controlling the kind of output:
951 an executable, object files, assembler files,
952 or preprocessed source.
953 * C Dialect Options:: Controlling the variant of C language compiled.
954 * C++ Dialect Options:: Variations on C++.
955 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
957 * Language Independent Options:: Controlling how diagnostics should be
959 * Warning Options:: How picky should the compiler be?
960 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
961 * Optimize Options:: How much optimization?
962 * Preprocessor Options:: Controlling header files and macro definitions.
963 Also, getting dependency information for Make.
964 * Assembler Options:: Passing options to the assembler.
965 * Link Options:: Specifying libraries and so on.
966 * Directory Options:: Where to find header files and libraries.
967 Where to find the compiler executable files.
968 * Spec Files:: How to pass switches to sub-processes.
969 * Target Options:: Running a cross-compiler, or an old version of GCC.
972 @node Overall Options
973 @section Options Controlling the Kind of Output
975 Compilation can involve up to four stages: preprocessing, compilation
976 proper, assembly and linking, always in that order. GCC is capable of
977 preprocessing and compiling several files either into several
978 assembler input files, or into one assembler input file; then each
979 assembler input file produces an object file, and linking combines all
980 the object files (those newly compiled, and those specified as input)
981 into an executable file.
983 @cindex file name suffix
984 For any given input file, the file name suffix determines what kind of
989 C source code which must be preprocessed.
992 C source code which should not be preprocessed.
995 C++ source code which should not be preprocessed.
998 Objective-C source code. Note that you must link with the @file{libobjc}
999 library to make an Objective-C program work.
1002 Objective-C source code which should not be preprocessed.
1006 Objective-C++ source code. Note that you must link with the @file{libobjc}
1007 library to make an Objective-C++ program work. Note that @samp{.M} refers
1008 to a literal capital M@.
1010 @item @var{file}.mii
1011 Objective-C++ source code which should not be preprocessed.
1014 C, C++, Objective-C or Objective-C++ header file to be turned into a
1015 precompiled header (default), or C, C++ header file to be turned into an
1016 Ada spec (via the @option{-fdump-ada-spec} switch).
1019 @itemx @var{file}.cp
1020 @itemx @var{file}.cxx
1021 @itemx @var{file}.cpp
1022 @itemx @var{file}.CPP
1023 @itemx @var{file}.c++
1025 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1026 the last two letters must both be literally @samp{x}. Likewise,
1027 @samp{.C} refers to a literal capital C@.
1031 Objective-C++ source code which must be preprocessed.
1033 @item @var{file}.mii
1034 Objective-C++ source code which should not be preprocessed.
1038 @itemx @var{file}.hp
1039 @itemx @var{file}.hxx
1040 @itemx @var{file}.hpp
1041 @itemx @var{file}.HPP
1042 @itemx @var{file}.h++
1043 @itemx @var{file}.tcc
1044 C++ header file to be turned into a precompiled header or Ada spec.
1047 @itemx @var{file}.for
1048 @itemx @var{file}.ftn
1049 Fixed form Fortran source code which should not be preprocessed.
1052 @itemx @var{file}.FOR
1053 @itemx @var{file}.fpp
1054 @itemx @var{file}.FPP
1055 @itemx @var{file}.FTN
1056 Fixed form Fortran source code which must be preprocessed (with the traditional
1059 @item @var{file}.f90
1060 @itemx @var{file}.f95
1061 @itemx @var{file}.f03
1062 @itemx @var{file}.f08
1063 Free form Fortran source code which should not be preprocessed.
1065 @item @var{file}.F90
1066 @itemx @var{file}.F95
1067 @itemx @var{file}.F03
1068 @itemx @var{file}.F08
1069 Free form Fortran source code which must be preprocessed (with the
1070 traditional preprocessor).
1075 @c FIXME: Descriptions of Java file types.
1081 @item @var{file}.ads
1082 Ada source code file which contains a library unit declaration (a
1083 declaration of a package, subprogram, or generic, or a generic
1084 instantiation), or a library unit renaming declaration (a package,
1085 generic, or subprogram renaming declaration). Such files are also
1088 @item @var{file}.adb
1089 Ada source code file containing a library unit body (a subprogram or
1090 package body). Such files are also called @dfn{bodies}.
1092 @c GCC also knows about some suffixes for languages not yet included:
1103 @itemx @var{file}.sx
1104 Assembler code which must be preprocessed.
1107 An object file to be fed straight into linking.
1108 Any file name with no recognized suffix is treated this way.
1112 You can specify the input language explicitly with the @option{-x} option:
1115 @item -x @var{language}
1116 Specify explicitly the @var{language} for the following input files
1117 (rather than letting the compiler choose a default based on the file
1118 name suffix). This option applies to all following input files until
1119 the next @option{-x} option. Possible values for @var{language} are:
1121 c c-header cpp-output
1122 c++ c++-header c++-cpp-output
1123 objective-c objective-c-header objective-c-cpp-output
1124 objective-c++ objective-c++-header objective-c++-cpp-output
1125 assembler assembler-with-cpp
1127 f77 f77-cpp-input f95 f95-cpp-input
1133 Turn off any specification of a language, so that subsequent files are
1134 handled according to their file name suffixes (as they are if @option{-x}
1135 has not been used at all).
1137 @item -pass-exit-codes
1138 @opindex pass-exit-codes
1139 Normally the @command{gcc} program will exit with the code of 1 if any
1140 phase of the compiler returns a non-success return code. If you specify
1141 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1142 numerically highest error produced by any phase that returned an error
1143 indication. The C, C++, and Fortran frontends return 4, if an internal
1144 compiler error is encountered.
1147 If you only want some of the stages of compilation, you can use
1148 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1149 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1150 @command{gcc} is to stop. Note that some combinations (for example,
1151 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1156 Compile or assemble the source files, but do not link. The linking
1157 stage simply is not done. The ultimate output is in the form of an
1158 object file for each source file.
1160 By default, the object file name for a source file is made by replacing
1161 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1163 Unrecognized input files, not requiring compilation or assembly, are
1168 Stop after the stage of compilation proper; do not assemble. The output
1169 is in the form of an assembler code file for each non-assembler input
1172 By default, the assembler file name for a source file is made by
1173 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1175 Input files that don't require compilation are ignored.
1179 Stop after the preprocessing stage; do not run the compiler proper. The
1180 output is in the form of preprocessed source code, which is sent to the
1183 Input files which don't require preprocessing are ignored.
1185 @cindex output file option
1188 Place output in file @var{file}. This applies regardless to whatever
1189 sort of output is being produced, whether it be an executable file,
1190 an object file, an assembler file or preprocessed C code.
1192 If @option{-o} is not specified, the default is to put an executable
1193 file in @file{a.out}, the object file for
1194 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1195 assembler file in @file{@var{source}.s}, a precompiled header file in
1196 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1201 Print (on standard error output) the commands executed to run the stages
1202 of compilation. Also print the version number of the compiler driver
1203 program and of the preprocessor and the compiler proper.
1207 Like @option{-v} except the commands are not executed and arguments
1208 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1209 This is useful for shell scripts to capture the driver-generated command lines.
1213 Use pipes rather than temporary files for communication between the
1214 various stages of compilation. This fails to work on some systems where
1215 the assembler is unable to read from a pipe; but the GNU assembler has
1220 Print (on the standard output) a description of the command line options
1221 understood by @command{gcc}. If the @option{-v} option is also specified
1222 then @option{--help} will also be passed on to the various processes
1223 invoked by @command{gcc}, so that they can display the command line options
1224 they accept. If the @option{-Wextra} option has also been specified
1225 (prior to the @option{--help} option), then command line options which
1226 have no documentation associated with them will also be displayed.
1229 @opindex target-help
1230 Print (on the standard output) a description of target-specific command
1231 line options for each tool. For some targets extra target-specific
1232 information may also be printed.
1234 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1235 Print (on the standard output) a description of the command line
1236 options understood by the compiler that fit into all specified classes
1237 and qualifiers. These are the supported classes:
1240 @item @samp{optimizers}
1241 This will display all of the optimization options supported by the
1244 @item @samp{warnings}
1245 This will display all of the options controlling warning messages
1246 produced by the compiler.
1249 This will display target-specific options. Unlike the
1250 @option{--target-help} option however, target-specific options of the
1251 linker and assembler will not be displayed. This is because those
1252 tools do not currently support the extended @option{--help=} syntax.
1255 This will display the values recognized by the @option{--param}
1258 @item @var{language}
1259 This will display the options supported for @var{language}, where
1260 @var{language} is the name of one of the languages supported in this
1264 This will display the options that are common to all languages.
1267 These are the supported qualifiers:
1270 @item @samp{undocumented}
1271 Display only those options which are undocumented.
1274 Display options which take an argument that appears after an equal
1275 sign in the same continuous piece of text, such as:
1276 @samp{--help=target}.
1278 @item @samp{separate}
1279 Display options which take an argument that appears as a separate word
1280 following the original option, such as: @samp{-o output-file}.
1283 Thus for example to display all the undocumented target-specific
1284 switches supported by the compiler the following can be used:
1287 --help=target,undocumented
1290 The sense of a qualifier can be inverted by prefixing it with the
1291 @samp{^} character, so for example to display all binary warning
1292 options (i.e., ones that are either on or off and that do not take an
1293 argument), which have a description the following can be used:
1296 --help=warnings,^joined,^undocumented
1299 The argument to @option{--help=} should not consist solely of inverted
1302 Combining several classes is possible, although this usually
1303 restricts the output by so much that there is nothing to display. One
1304 case where it does work however is when one of the classes is
1305 @var{target}. So for example to display all the target-specific
1306 optimization options the following can be used:
1309 --help=target,optimizers
1312 The @option{--help=} option can be repeated on the command line. Each
1313 successive use will display its requested class of options, skipping
1314 those that have already been displayed.
1316 If the @option{-Q} option appears on the command line before the
1317 @option{--help=} option, then the descriptive text displayed by
1318 @option{--help=} is changed. Instead of describing the displayed
1319 options, an indication is given as to whether the option is enabled,
1320 disabled or set to a specific value (assuming that the compiler
1321 knows this at the point where the @option{--help=} option is used).
1323 Here is a truncated example from the ARM port of @command{gcc}:
1326 % gcc -Q -mabi=2 --help=target -c
1327 The following options are target specific:
1329 -mabort-on-noreturn [disabled]
1333 The output is sensitive to the effects of previous command line
1334 options, so for example it is possible to find out which optimizations
1335 are enabled at @option{-O2} by using:
1338 -Q -O2 --help=optimizers
1341 Alternatively you can discover which binary optimizations are enabled
1342 by @option{-O3} by using:
1345 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1346 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1347 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1350 @item -no-canonical-prefixes
1351 @opindex no-canonical-prefixes
1352 Do not expand any symbolic links, resolve references to @samp{/../}
1353 or @samp{/./}, or make the path absolute when generating a relative
1358 Display the version number and copyrights of the invoked GCC@.
1362 Invoke all subcommands under a wrapper program. The name of the
1363 wrapper program and its parameters are passed as a comma separated
1367 gcc -c t.c -wrapper gdb,--args
1370 This will invoke all subprograms of @command{gcc} under
1371 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1372 @samp{gdb --args cc1 @dots{}}.
1374 @item -fplugin=@var{name}.so
1375 Load the plugin code in file @var{name}.so, assumed to be a
1376 shared object to be dlopen'd by the compiler. The base name of
1377 the shared object file is used to identify the plugin for the
1378 purposes of argument parsing (See
1379 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1380 Each plugin should define the callback functions specified in the
1383 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1384 Define an argument called @var{key} with a value of @var{value}
1385 for the plugin called @var{name}.
1387 @item -fdump-ada-spec@r{[}-slim@r{]}
1388 For C and C++ source and include files, generate corresponding Ada
1389 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1390 GNAT User's Guide}, which provides detailed documentation on this feature.
1392 @item -fdump-go-spec=@var{file}
1393 For input files in any language, generate corresponding Go
1394 declarations in @var{file}. This generates Go @code{const},
1395 @code{type}, @code{var}, and @code{func} declarations which may be a
1396 useful way to start writing a Go interface to code written in some
1399 @include @value{srcdir}/../libiberty/at-file.texi
1403 @section Compiling C++ Programs
1405 @cindex suffixes for C++ source
1406 @cindex C++ source file suffixes
1407 C++ source files conventionally use one of the suffixes @samp{.C},
1408 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1409 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1410 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1411 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1412 files with these names and compiles them as C++ programs even if you
1413 call the compiler the same way as for compiling C programs (usually
1414 with the name @command{gcc}).
1418 However, the use of @command{gcc} does not add the C++ library.
1419 @command{g++} is a program that calls GCC and treats @samp{.c},
1420 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1421 files unless @option{-x} is used, and automatically specifies linking
1422 against the C++ library. This program is also useful when
1423 precompiling a C header file with a @samp{.h} extension for use in C++
1424 compilations. On many systems, @command{g++} is also installed with
1425 the name @command{c++}.
1427 @cindex invoking @command{g++}
1428 When you compile C++ programs, you may specify many of the same
1429 command-line options that you use for compiling programs in any
1430 language; or command-line options meaningful for C and related
1431 languages; or options that are meaningful only for C++ programs.
1432 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1433 explanations of options for languages related to C@.
1434 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1435 explanations of options that are meaningful only for C++ programs.
1437 @node C Dialect Options
1438 @section Options Controlling C Dialect
1439 @cindex dialect options
1440 @cindex language dialect options
1441 @cindex options, dialect
1443 The following options control the dialect of C (or languages derived
1444 from C, such as C++, Objective-C and Objective-C++) that the compiler
1448 @cindex ANSI support
1452 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1453 equivalent to @samp{-std=c++98}.
1455 This turns off certain features of GCC that are incompatible with ISO
1456 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1457 such as the @code{asm} and @code{typeof} keywords, and
1458 predefined macros such as @code{unix} and @code{vax} that identify the
1459 type of system you are using. It also enables the undesirable and
1460 rarely used ISO trigraph feature. For the C compiler,
1461 it disables recognition of C++ style @samp{//} comments as well as
1462 the @code{inline} keyword.
1464 The alternate keywords @code{__asm__}, @code{__extension__},
1465 @code{__inline__} and @code{__typeof__} continue to work despite
1466 @option{-ansi}. You would not want to use them in an ISO C program, of
1467 course, but it is useful to put them in header files that might be included
1468 in compilations done with @option{-ansi}. Alternate predefined macros
1469 such as @code{__unix__} and @code{__vax__} are also available, with or
1470 without @option{-ansi}.
1472 The @option{-ansi} option does not cause non-ISO programs to be
1473 rejected gratuitously. For that, @option{-pedantic} is required in
1474 addition to @option{-ansi}. @xref{Warning Options}.
1476 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1477 option is used. Some header files may notice this macro and refrain
1478 from declaring certain functions or defining certain macros that the
1479 ISO standard doesn't call for; this is to avoid interfering with any
1480 programs that might use these names for other things.
1482 Functions that would normally be built in but do not have semantics
1483 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1484 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1485 built-in functions provided by GCC}, for details of the functions
1490 Determine the language standard. @xref{Standards,,Language Standards
1491 Supported by GCC}, for details of these standard versions. This option
1492 is currently only supported when compiling C or C++.
1494 The compiler can accept several base standards, such as @samp{c90} or
1495 @samp{c++98}, and GNU dialects of those standards, such as
1496 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1497 compiler will accept all programs following that standard and those
1498 using GNU extensions that do not contradict it. For example,
1499 @samp{-std=c90} turns off certain features of GCC that are
1500 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1501 keywords, but not other GNU extensions that do not have a meaning in
1502 ISO C90, such as omitting the middle term of a @code{?:}
1503 expression. On the other hand, by specifying a GNU dialect of a
1504 standard, all features the compiler support are enabled, even when
1505 those features change the meaning of the base standard and some
1506 strict-conforming programs may be rejected. The particular standard
1507 is used by @option{-pedantic} to identify which features are GNU
1508 extensions given that version of the standard. For example
1509 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1510 comments, while @samp{-std=gnu99 -pedantic} would not.
1512 A value for this option must be provided; possible values are
1518 Support all ISO C90 programs (certain GNU extensions that conflict
1519 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1521 @item iso9899:199409
1522 ISO C90 as modified in amendment 1.
1528 ISO C99. Note that this standard is not yet fully supported; see
1529 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1530 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1533 ISO C1X, the draft of the next revision of the ISO C standard.
1534 Support is limited and experimental and features enabled by this
1535 option may be changed or removed if changed in or removed from the
1540 GNU dialect of ISO C90 (including some C99 features). This
1541 is the default for C code.
1545 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1546 this will become the default. The name @samp{gnu9x} is deprecated.
1549 GNU dialect of ISO C1X. Support is limited and experimental and
1550 features enabled by this option may be changed or removed if changed
1551 in or removed from the standard draft.
1554 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1558 GNU dialect of @option{-std=c++98}. This is the default for
1562 The working draft of the upcoming ISO C++0x standard. This option
1563 enables experimental features that are likely to be included in
1564 C++0x. The working draft is constantly changing, and any feature that is
1565 enabled by this flag may be removed from future versions of GCC if it is
1566 not part of the C++0x standard.
1569 GNU dialect of @option{-std=c++0x}. This option enables
1570 experimental features that may be removed in future versions of GCC.
1573 @item -fgnu89-inline
1574 @opindex fgnu89-inline
1575 The option @option{-fgnu89-inline} tells GCC to use the traditional
1576 GNU semantics for @code{inline} functions when in C99 mode.
1577 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1578 is accepted and ignored by GCC versions 4.1.3 up to but not including
1579 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1580 C99 mode. Using this option is roughly equivalent to adding the
1581 @code{gnu_inline} function attribute to all inline functions
1582 (@pxref{Function Attributes}).
1584 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1585 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1586 specifies the default behavior). This option was first supported in
1587 GCC 4.3. This option is not supported in @option{-std=c90} or
1588 @option{-std=gnu90} mode.
1590 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1591 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1592 in effect for @code{inline} functions. @xref{Common Predefined
1593 Macros,,,cpp,The C Preprocessor}.
1595 @item -aux-info @var{filename}
1597 Output to the given filename prototyped declarations for all functions
1598 declared and/or defined in a translation unit, including those in header
1599 files. This option is silently ignored in any language other than C@.
1601 Besides declarations, the file indicates, in comments, the origin of
1602 each declaration (source file and line), whether the declaration was
1603 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1604 @samp{O} for old, respectively, in the first character after the line
1605 number and the colon), and whether it came from a declaration or a
1606 definition (@samp{C} or @samp{F}, respectively, in the following
1607 character). In the case of function definitions, a K&R-style list of
1608 arguments followed by their declarations is also provided, inside
1609 comments, after the declaration.
1613 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1614 keyword, so that code can use these words as identifiers. You can use
1615 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1616 instead. @option{-ansi} implies @option{-fno-asm}.
1618 In C++, this switch only affects the @code{typeof} keyword, since
1619 @code{asm} and @code{inline} are standard keywords. You may want to
1620 use the @option{-fno-gnu-keywords} flag instead, which has the same
1621 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1622 switch only affects the @code{asm} and @code{typeof} keywords, since
1623 @code{inline} is a standard keyword in ISO C99.
1626 @itemx -fno-builtin-@var{function}
1627 @opindex fno-builtin
1628 @cindex built-in functions
1629 Don't recognize built-in functions that do not begin with
1630 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1631 functions provided by GCC}, for details of the functions affected,
1632 including those which are not built-in functions when @option{-ansi} or
1633 @option{-std} options for strict ISO C conformance are used because they
1634 do not have an ISO standard meaning.
1636 GCC normally generates special code to handle certain built-in functions
1637 more efficiently; for instance, calls to @code{alloca} may become single
1638 instructions that adjust the stack directly, and calls to @code{memcpy}
1639 may become inline copy loops. The resulting code is often both smaller
1640 and faster, but since the function calls no longer appear as such, you
1641 cannot set a breakpoint on those calls, nor can you change the behavior
1642 of the functions by linking with a different library. In addition,
1643 when a function is recognized as a built-in function, GCC may use
1644 information about that function to warn about problems with calls to
1645 that function, or to generate more efficient code, even if the
1646 resulting code still contains calls to that function. For example,
1647 warnings are given with @option{-Wformat} for bad calls to
1648 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1649 known not to modify global memory.
1651 With the @option{-fno-builtin-@var{function}} option
1652 only the built-in function @var{function} is
1653 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1654 function is named that is not built-in in this version of GCC, this
1655 option is ignored. There is no corresponding
1656 @option{-fbuiltin-@var{function}} option; if you wish to enable
1657 built-in functions selectively when using @option{-fno-builtin} or
1658 @option{-ffreestanding}, you may define macros such as:
1661 #define abs(n) __builtin_abs ((n))
1662 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1667 @cindex hosted environment
1669 Assert that compilation takes place in a hosted environment. This implies
1670 @option{-fbuiltin}. A hosted environment is one in which the
1671 entire standard library is available, and in which @code{main} has a return
1672 type of @code{int}. Examples are nearly everything except a kernel.
1673 This is equivalent to @option{-fno-freestanding}.
1675 @item -ffreestanding
1676 @opindex ffreestanding
1677 @cindex hosted environment
1679 Assert that compilation takes place in a freestanding environment. This
1680 implies @option{-fno-builtin}. A freestanding environment
1681 is one in which the standard library may not exist, and program startup may
1682 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1683 This is equivalent to @option{-fno-hosted}.
1685 @xref{Standards,,Language Standards Supported by GCC}, for details of
1686 freestanding and hosted environments.
1690 @cindex OpenMP parallel
1691 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1692 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1693 compiler generates parallel code according to the OpenMP Application
1694 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1695 implies @option{-pthread}, and thus is only supported on targets that
1696 have support for @option{-pthread}.
1698 @item -fms-extensions
1699 @opindex fms-extensions
1700 Accept some non-standard constructs used in Microsoft header files.
1702 In C++ code, this allows member names in structures to be similar
1703 to previous types declarations.
1712 Some cases of unnamed fields in structures and unions are only
1713 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1714 fields within structs/unions}, for details.
1716 @item -fplan9-extensions
1717 Accept some non-standard constructs used in Plan 9 code.
1719 This enables @option{-fms-extensions}, permits passing pointers to
1720 structures with anonymous fields to functions which expect pointers to
1721 elements of the type of the field, and permits referring to anonymous
1722 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1723 struct/union fields within structs/unions}, for details. This is only
1724 supported for C, not C++.
1728 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1729 options for strict ISO C conformance) implies @option{-trigraphs}.
1731 @item -no-integrated-cpp
1732 @opindex no-integrated-cpp
1733 Performs a compilation in two passes: preprocessing and compiling. This
1734 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1735 @option{-B} option. The user supplied compilation step can then add in
1736 an additional preprocessing step after normal preprocessing but before
1737 compiling. The default is to use the integrated cpp (internal cpp)
1739 The semantics of this option will change if "cc1", "cc1plus", and
1740 "cc1obj" are merged.
1742 @cindex traditional C language
1743 @cindex C language, traditional
1745 @itemx -traditional-cpp
1746 @opindex traditional-cpp
1747 @opindex traditional
1748 Formerly, these options caused GCC to attempt to emulate a pre-standard
1749 C compiler. They are now only supported with the @option{-E} switch.
1750 The preprocessor continues to support a pre-standard mode. See the GNU
1751 CPP manual for details.
1753 @item -fcond-mismatch
1754 @opindex fcond-mismatch
1755 Allow conditional expressions with mismatched types in the second and
1756 third arguments. The value of such an expression is void. This option
1757 is not supported for C++.
1759 @item -flax-vector-conversions
1760 @opindex flax-vector-conversions
1761 Allow implicit conversions between vectors with differing numbers of
1762 elements and/or incompatible element types. This option should not be
1765 @item -funsigned-char
1766 @opindex funsigned-char
1767 Let the type @code{char} be unsigned, like @code{unsigned char}.
1769 Each kind of machine has a default for what @code{char} should
1770 be. It is either like @code{unsigned char} by default or like
1771 @code{signed char} by default.
1773 Ideally, a portable program should always use @code{signed char} or
1774 @code{unsigned char} when it depends on the signedness of an object.
1775 But many programs have been written to use plain @code{char} and
1776 expect it to be signed, or expect it to be unsigned, depending on the
1777 machines they were written for. This option, and its inverse, let you
1778 make such a program work with the opposite default.
1780 The type @code{char} is always a distinct type from each of
1781 @code{signed char} or @code{unsigned char}, even though its behavior
1782 is always just like one of those two.
1785 @opindex fsigned-char
1786 Let the type @code{char} be signed, like @code{signed char}.
1788 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1789 the negative form of @option{-funsigned-char}. Likewise, the option
1790 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1792 @item -fsigned-bitfields
1793 @itemx -funsigned-bitfields
1794 @itemx -fno-signed-bitfields
1795 @itemx -fno-unsigned-bitfields
1796 @opindex fsigned-bitfields
1797 @opindex funsigned-bitfields
1798 @opindex fno-signed-bitfields
1799 @opindex fno-unsigned-bitfields
1800 These options control whether a bit-field is signed or unsigned, when the
1801 declaration does not use either @code{signed} or @code{unsigned}. By
1802 default, such a bit-field is signed, because this is consistent: the
1803 basic integer types such as @code{int} are signed types.
1806 @node C++ Dialect Options
1807 @section Options Controlling C++ Dialect
1809 @cindex compiler options, C++
1810 @cindex C++ options, command line
1811 @cindex options, C++
1812 This section describes the command-line options that are only meaningful
1813 for C++ programs; but you can also use most of the GNU compiler options
1814 regardless of what language your program is in. For example, you
1815 might compile a file @code{firstClass.C} like this:
1818 g++ -g -frepo -O -c firstClass.C
1822 In this example, only @option{-frepo} is an option meant
1823 only for C++ programs; you can use the other options with any
1824 language supported by GCC@.
1826 Here is a list of options that are @emph{only} for compiling C++ programs:
1830 @item -fabi-version=@var{n}
1831 @opindex fabi-version
1832 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1833 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1834 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1835 the version that conforms most closely to the C++ ABI specification.
1836 Therefore, the ABI obtained using version 0 will change as ABI bugs
1839 The default is version 2.
1841 Version 3 corrects an error in mangling a constant address as a
1844 Version 4 implements a standard mangling for vector types.
1846 Version 5 corrects the mangling of attribute const/volatile on
1847 function pointer types, decltype of a plain decl, and use of a
1848 function parameter in the declaration of another parameter.
1850 Version 6 corrects the promotion behavior of C++0x scoped enums.
1852 See also @option{-Wabi}.
1854 @item -fno-access-control
1855 @opindex fno-access-control
1856 Turn off all access checking. This switch is mainly useful for working
1857 around bugs in the access control code.
1861 Check that the pointer returned by @code{operator new} is non-null
1862 before attempting to modify the storage allocated. This check is
1863 normally unnecessary because the C++ standard specifies that
1864 @code{operator new} will only return @code{0} if it is declared
1865 @samp{throw()}, in which case the compiler will always check the
1866 return value even without this option. In all other cases, when
1867 @code{operator new} has a non-empty exception specification, memory
1868 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1869 @samp{new (nothrow)}.
1871 @item -fconserve-space
1872 @opindex fconserve-space
1873 Put uninitialized or runtime-initialized global variables into the
1874 common segment, as C does. This saves space in the executable at the
1875 cost of not diagnosing duplicate definitions. If you compile with this
1876 flag and your program mysteriously crashes after @code{main()} has
1877 completed, you may have an object that is being destroyed twice because
1878 two definitions were merged.
1880 This option is no longer useful on most targets, now that support has
1881 been added for putting variables into BSS without making them common.
1883 @item -fconstexpr-depth=@var{n}
1884 @opindex fconstexpr-depth
1885 Set the maximum nested evaluation depth for C++0x constexpr functions
1886 to @var{n}. A limit is needed to detect endless recursion during
1887 constant expression evaluation. The minimum specified by the standard
1890 @item -fno-deduce-init-list
1891 @opindex fno-deduce-init-list
1892 Disable deduction of a template type parameter as
1893 std::initializer_list from a brace-enclosed initializer list, i.e.
1896 template <class T> auto forward(T t) -> decltype (realfn (t))
1903 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1907 This option is present because this deduction is an extension to the
1908 current specification in the C++0x working draft, and there was
1909 some concern about potential overload resolution problems.
1911 @item -ffriend-injection
1912 @opindex ffriend-injection
1913 Inject friend functions into the enclosing namespace, so that they are
1914 visible outside the scope of the class in which they are declared.
1915 Friend functions were documented to work this way in the old Annotated
1916 C++ Reference Manual, and versions of G++ before 4.1 always worked
1917 that way. However, in ISO C++ a friend function which is not declared
1918 in an enclosing scope can only be found using argument dependent
1919 lookup. This option causes friends to be injected as they were in
1922 This option is for compatibility, and may be removed in a future
1925 @item -fno-elide-constructors
1926 @opindex fno-elide-constructors
1927 The C++ standard allows an implementation to omit creating a temporary
1928 which is only used to initialize another object of the same type.
1929 Specifying this option disables that optimization, and forces G++ to
1930 call the copy constructor in all cases.
1932 @item -fno-enforce-eh-specs
1933 @opindex fno-enforce-eh-specs
1934 Don't generate code to check for violation of exception specifications
1935 at runtime. This option violates the C++ standard, but may be useful
1936 for reducing code size in production builds, much like defining
1937 @samp{NDEBUG}. This does not give user code permission to throw
1938 exceptions in violation of the exception specifications; the compiler
1939 will still optimize based on the specifications, so throwing an
1940 unexpected exception will result in undefined behavior.
1943 @itemx -fno-for-scope
1945 @opindex fno-for-scope
1946 If @option{-ffor-scope} is specified, the scope of variables declared in
1947 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1948 as specified by the C++ standard.
1949 If @option{-fno-for-scope} is specified, the scope of variables declared in
1950 a @i{for-init-statement} extends to the end of the enclosing scope,
1951 as was the case in old versions of G++, and other (traditional)
1952 implementations of C++.
1954 The default if neither flag is given to follow the standard,
1955 but to allow and give a warning for old-style code that would
1956 otherwise be invalid, or have different behavior.
1958 @item -fno-gnu-keywords
1959 @opindex fno-gnu-keywords
1960 Do not recognize @code{typeof} as a keyword, so that code can use this
1961 word as an identifier. You can use the keyword @code{__typeof__} instead.
1962 @option{-ansi} implies @option{-fno-gnu-keywords}.
1964 @item -fno-implicit-templates
1965 @opindex fno-implicit-templates
1966 Never emit code for non-inline templates which are instantiated
1967 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1968 @xref{Template Instantiation}, for more information.
1970 @item -fno-implicit-inline-templates
1971 @opindex fno-implicit-inline-templates
1972 Don't emit code for implicit instantiations of inline templates, either.
1973 The default is to handle inlines differently so that compiles with and
1974 without optimization will need the same set of explicit instantiations.
1976 @item -fno-implement-inlines
1977 @opindex fno-implement-inlines
1978 To save space, do not emit out-of-line copies of inline functions
1979 controlled by @samp{#pragma implementation}. This will cause linker
1980 errors if these functions are not inlined everywhere they are called.
1982 @item -fms-extensions
1983 @opindex fms-extensions
1984 Disable pedantic warnings about constructs used in MFC, such as implicit
1985 int and getting a pointer to member function via non-standard syntax.
1987 @item -fno-nonansi-builtins
1988 @opindex fno-nonansi-builtins
1989 Disable built-in declarations of functions that are not mandated by
1990 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1991 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1994 @opindex fnothrow-opt
1995 Treat a @code{throw()} exception specification as though it were a
1996 @code{noexcept} specification to reduce or eliminate the text size
1997 overhead relative to a function with no exception specification. If
1998 the function has local variables of types with non-trivial
1999 destructors, the exception specification will actually make the
2000 function smaller because the EH cleanups for those variables can be
2001 optimized away. The semantic effect is that an exception thrown out of
2002 a function with such an exception specification will result in a call
2003 to @code{terminate} rather than @code{unexpected}.
2005 @item -fno-operator-names
2006 @opindex fno-operator-names
2007 Do not treat the operator name keywords @code{and}, @code{bitand},
2008 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2009 synonyms as keywords.
2011 @item -fno-optional-diags
2012 @opindex fno-optional-diags
2013 Disable diagnostics that the standard says a compiler does not need to
2014 issue. Currently, the only such diagnostic issued by G++ is the one for
2015 a name having multiple meanings within a class.
2018 @opindex fpermissive
2019 Downgrade some diagnostics about nonconformant code from errors to
2020 warnings. Thus, using @option{-fpermissive} will allow some
2021 nonconforming code to compile.
2023 @item -fno-pretty-templates
2024 @opindex fno-pretty-templates
2025 When an error message refers to a specialization of a function
2026 template, the compiler will normally print the signature of the
2027 template followed by the template arguments and any typedefs or
2028 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2029 rather than @code{void f(int)}) so that it's clear which template is
2030 involved. When an error message refers to a specialization of a class
2031 template, the compiler will omit any template arguments which match
2032 the default template arguments for that template. If either of these
2033 behaviors make it harder to understand the error message rather than
2034 easier, using @option{-fno-pretty-templates} will disable them.
2038 Enable automatic template instantiation at link time. This option also
2039 implies @option{-fno-implicit-templates}. @xref{Template
2040 Instantiation}, for more information.
2044 Disable generation of information about every class with virtual
2045 functions for use by the C++ runtime type identification features
2046 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2047 of the language, you can save some space by using this flag. Note that
2048 exception handling uses the same information, but it will generate it as
2049 needed. The @samp{dynamic_cast} operator can still be used for casts that
2050 do not require runtime type information, i.e.@: casts to @code{void *} or to
2051 unambiguous base classes.
2055 Emit statistics about front-end processing at the end of the compilation.
2056 This information is generally only useful to the G++ development team.
2058 @item -fstrict-enums
2059 @opindex fstrict-enums
2060 Allow the compiler to optimize using the assumption that a value of
2061 enumeration type can only be one of the values of the enumeration (as
2062 defined in the C++ standard; basically, a value which can be
2063 represented in the minimum number of bits needed to represent all the
2064 enumerators). This assumption may not be valid if the program uses a
2065 cast to convert an arbitrary integer value to the enumeration type.
2067 @item -ftemplate-depth=@var{n}
2068 @opindex ftemplate-depth
2069 Set the maximum instantiation depth for template classes to @var{n}.
2070 A limit on the template instantiation depth is needed to detect
2071 endless recursions during template class instantiation. ANSI/ISO C++
2072 conforming programs must not rely on a maximum depth greater than 17
2073 (changed to 1024 in C++0x). The default value is 900, as the compiler
2074 can run out of stack space before hitting 1024 in some situations.
2076 @item -fno-threadsafe-statics
2077 @opindex fno-threadsafe-statics
2078 Do not emit the extra code to use the routines specified in the C++
2079 ABI for thread-safe initialization of local statics. You can use this
2080 option to reduce code size slightly in code that doesn't need to be
2083 @item -fuse-cxa-atexit
2084 @opindex fuse-cxa-atexit
2085 Register destructors for objects with static storage duration with the
2086 @code{__cxa_atexit} function rather than the @code{atexit} function.
2087 This option is required for fully standards-compliant handling of static
2088 destructors, but will only work if your C library supports
2089 @code{__cxa_atexit}.
2091 @item -fno-use-cxa-get-exception-ptr
2092 @opindex fno-use-cxa-get-exception-ptr
2093 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2094 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2095 if the runtime routine is not available.
2097 @item -fvisibility-inlines-hidden
2098 @opindex fvisibility-inlines-hidden
2099 This switch declares that the user does not attempt to compare
2100 pointers to inline methods where the addresses of the two functions
2101 were taken in different shared objects.
2103 The effect of this is that GCC may, effectively, mark inline methods with
2104 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2105 appear in the export table of a DSO and do not require a PLT indirection
2106 when used within the DSO@. Enabling this option can have a dramatic effect
2107 on load and link times of a DSO as it massively reduces the size of the
2108 dynamic export table when the library makes heavy use of templates.
2110 The behavior of this switch is not quite the same as marking the
2111 methods as hidden directly, because it does not affect static variables
2112 local to the function or cause the compiler to deduce that
2113 the function is defined in only one shared object.
2115 You may mark a method as having a visibility explicitly to negate the
2116 effect of the switch for that method. For example, if you do want to
2117 compare pointers to a particular inline method, you might mark it as
2118 having default visibility. Marking the enclosing class with explicit
2119 visibility will have no effect.
2121 Explicitly instantiated inline methods are unaffected by this option
2122 as their linkage might otherwise cross a shared library boundary.
2123 @xref{Template Instantiation}.
2125 @item -fvisibility-ms-compat
2126 @opindex fvisibility-ms-compat
2127 This flag attempts to use visibility settings to make GCC's C++
2128 linkage model compatible with that of Microsoft Visual Studio.
2130 The flag makes these changes to GCC's linkage model:
2134 It sets the default visibility to @code{hidden}, like
2135 @option{-fvisibility=hidden}.
2138 Types, but not their members, are not hidden by default.
2141 The One Definition Rule is relaxed for types without explicit
2142 visibility specifications which are defined in more than one different
2143 shared object: those declarations are permitted if they would have
2144 been permitted when this option was not used.
2147 In new code it is better to use @option{-fvisibility=hidden} and
2148 export those classes which are intended to be externally visible.
2149 Unfortunately it is possible for code to rely, perhaps accidentally,
2150 on the Visual Studio behavior.
2152 Among the consequences of these changes are that static data members
2153 of the same type with the same name but defined in different shared
2154 objects will be different, so changing one will not change the other;
2155 and that pointers to function members defined in different shared
2156 objects may not compare equal. When this flag is given, it is a
2157 violation of the ODR to define types with the same name differently.
2161 Do not use weak symbol support, even if it is provided by the linker.
2162 By default, G++ will use weak symbols if they are available. This
2163 option exists only for testing, and should not be used by end-users;
2164 it will result in inferior code and has no benefits. This option may
2165 be removed in a future release of G++.
2169 Do not search for header files in the standard directories specific to
2170 C++, but do still search the other standard directories. (This option
2171 is used when building the C++ library.)
2174 In addition, these optimization, warning, and code generation options
2175 have meanings only for C++ programs:
2178 @item -fno-default-inline
2179 @opindex fno-default-inline
2180 Do not assume @samp{inline} for functions defined inside a class scope.
2181 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2182 functions will have linkage like inline functions; they just won't be
2185 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2188 Warn when G++ generates code that is probably not compatible with the
2189 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2190 all such cases, there are probably some cases that are not warned about,
2191 even though G++ is generating incompatible code. There may also be
2192 cases where warnings are emitted even though the code that is generated
2195 You should rewrite your code to avoid these warnings if you are
2196 concerned about the fact that code generated by G++ may not be binary
2197 compatible with code generated by other compilers.
2199 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2204 A template with a non-type template parameter of reference type is
2205 mangled incorrectly:
2208 template <int &> struct S @{@};
2212 This is fixed in @option{-fabi-version=3}.
2215 SIMD vector types declared using @code{__attribute ((vector_size))} are
2216 mangled in a non-standard way that does not allow for overloading of
2217 functions taking vectors of different sizes.
2219 The mangling is changed in @option{-fabi-version=4}.
2222 The known incompatibilities in @option{-fabi-version=1} include:
2227 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2228 pack data into the same byte as a base class. For example:
2231 struct A @{ virtual void f(); int f1 : 1; @};
2232 struct B : public A @{ int f2 : 1; @};
2236 In this case, G++ will place @code{B::f2} into the same byte
2237 as@code{A::f1}; other compilers will not. You can avoid this problem
2238 by explicitly padding @code{A} so that its size is a multiple of the
2239 byte size on your platform; that will cause G++ and other compilers to
2240 layout @code{B} identically.
2243 Incorrect handling of tail-padding for virtual bases. G++ does not use
2244 tail padding when laying out virtual bases. For example:
2247 struct A @{ virtual void f(); char c1; @};
2248 struct B @{ B(); char c2; @};
2249 struct C : public A, public virtual B @{@};
2253 In this case, G++ will not place @code{B} into the tail-padding for
2254 @code{A}; other compilers will. You can avoid this problem by
2255 explicitly padding @code{A} so that its size is a multiple of its
2256 alignment (ignoring virtual base classes); that will cause G++ and other
2257 compilers to layout @code{C} identically.
2260 Incorrect handling of bit-fields with declared widths greater than that
2261 of their underlying types, when the bit-fields appear in a union. For
2265 union U @{ int i : 4096; @};
2269 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2270 union too small by the number of bits in an @code{int}.
2273 Empty classes can be placed at incorrect offsets. For example:
2283 struct C : public B, public A @{@};
2287 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2288 it should be placed at offset zero. G++ mistakenly believes that the
2289 @code{A} data member of @code{B} is already at offset zero.
2292 Names of template functions whose types involve @code{typename} or
2293 template template parameters can be mangled incorrectly.
2296 template <typename Q>
2297 void f(typename Q::X) @{@}
2299 template <template <typename> class Q>
2300 void f(typename Q<int>::X) @{@}
2304 Instantiations of these templates may be mangled incorrectly.
2308 It also warns psABI related changes. The known psABI changes at this
2314 For SYSV/x86-64, when passing union with long double, it is changed to
2315 pass in memory as specified in psABI. For example:
2325 @code{union U} will always be passed in memory.
2329 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2330 @opindex Wctor-dtor-privacy
2331 @opindex Wno-ctor-dtor-privacy
2332 Warn when a class seems unusable because all the constructors or
2333 destructors in that class are private, and it has neither friends nor
2334 public static member functions.
2336 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2337 @opindex Wdelete-non-virtual-dtor
2338 @opindex Wno-delete-non-virtual-dtor
2339 Warn when @samp{delete} is used to destroy an instance of a class which
2340 has virtual functions and non-virtual destructor. It is unsafe to delete
2341 an instance of a derived class through a pointer to a base class if the
2342 base class does not have a virtual destructor. This warning is enabled
2345 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2347 @opindex Wno-noexcept
2348 Warn when a noexcept-expression evaluates to false because of a call
2349 to a function that does not have a non-throwing exception
2350 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2351 the compiler to never throw an exception.
2353 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2354 @opindex Wnon-virtual-dtor
2355 @opindex Wno-non-virtual-dtor
2356 Warn when a class has virtual functions and accessible non-virtual
2357 destructor, in which case it would be possible but unsafe to delete
2358 an instance of a derived class through a pointer to the base class.
2359 This warning is also enabled if -Weffc++ is specified.
2361 @item -Wreorder @r{(C++ and Objective-C++ only)}
2363 @opindex Wno-reorder
2364 @cindex reordering, warning
2365 @cindex warning for reordering of member initializers
2366 Warn when the order of member initializers given in the code does not
2367 match the order in which they must be executed. For instance:
2373 A(): j (0), i (1) @{ @}
2377 The compiler will rearrange the member initializers for @samp{i}
2378 and @samp{j} to match the declaration order of the members, emitting
2379 a warning to that effect. This warning is enabled by @option{-Wall}.
2382 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2385 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2388 Warn about violations of the following style guidelines from Scott Meyers'
2389 @cite{Effective C++} book:
2393 Item 11: Define a copy constructor and an assignment operator for classes
2394 with dynamically allocated memory.
2397 Item 12: Prefer initialization to assignment in constructors.
2400 Item 14: Make destructors virtual in base classes.
2403 Item 15: Have @code{operator=} return a reference to @code{*this}.
2406 Item 23: Don't try to return a reference when you must return an object.
2410 Also warn about violations of the following style guidelines from
2411 Scott Meyers' @cite{More Effective C++} book:
2415 Item 6: Distinguish between prefix and postfix forms of increment and
2416 decrement operators.
2419 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2423 When selecting this option, be aware that the standard library
2424 headers do not obey all of these guidelines; use @samp{grep -v}
2425 to filter out those warnings.
2427 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2428 @opindex Wstrict-null-sentinel
2429 @opindex Wno-strict-null-sentinel
2430 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2431 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2432 to @code{__null}. Although it is a null pointer constant not a null pointer,
2433 it is guaranteed to be of the same size as a pointer. But this use is
2434 not portable across different compilers.
2436 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2437 @opindex Wno-non-template-friend
2438 @opindex Wnon-template-friend
2439 Disable warnings when non-templatized friend functions are declared
2440 within a template. Since the advent of explicit template specification
2441 support in G++, if the name of the friend is an unqualified-id (i.e.,
2442 @samp{friend foo(int)}), the C++ language specification demands that the
2443 friend declare or define an ordinary, nontemplate function. (Section
2444 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2445 could be interpreted as a particular specialization of a templatized
2446 function. Because this non-conforming behavior is no longer the default
2447 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2448 check existing code for potential trouble spots and is on by default.
2449 This new compiler behavior can be turned off with
2450 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2451 but disables the helpful warning.
2453 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2454 @opindex Wold-style-cast
2455 @opindex Wno-old-style-cast
2456 Warn if an old-style (C-style) cast to a non-void type is used within
2457 a C++ program. The new-style casts (@samp{dynamic_cast},
2458 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2459 less vulnerable to unintended effects and much easier to search for.
2461 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2462 @opindex Woverloaded-virtual
2463 @opindex Wno-overloaded-virtual
2464 @cindex overloaded virtual function, warning
2465 @cindex warning for overloaded virtual function
2466 Warn when a function declaration hides virtual functions from a
2467 base class. For example, in:
2474 struct B: public A @{
2479 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2487 will fail to compile.
2489 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2490 @opindex Wno-pmf-conversions
2491 @opindex Wpmf-conversions
2492 Disable the diagnostic for converting a bound pointer to member function
2495 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2496 @opindex Wsign-promo
2497 @opindex Wno-sign-promo
2498 Warn when overload resolution chooses a promotion from unsigned or
2499 enumerated type to a signed type, over a conversion to an unsigned type of
2500 the same size. Previous versions of G++ would try to preserve
2501 unsignedness, but the standard mandates the current behavior.
2506 A& operator = (int);
2516 In this example, G++ will synthesize a default @samp{A& operator =
2517 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2520 @node Objective-C and Objective-C++ Dialect Options
2521 @section Options Controlling Objective-C and Objective-C++ Dialects
2523 @cindex compiler options, Objective-C and Objective-C++
2524 @cindex Objective-C and Objective-C++ options, command line
2525 @cindex options, Objective-C and Objective-C++
2526 (NOTE: This manual does not describe the Objective-C and Objective-C++
2527 languages themselves. @xref{Standards,,Language Standards
2528 Supported by GCC}, for references.)
2530 This section describes the command-line options that are only meaningful
2531 for Objective-C and Objective-C++ programs, but you can also use most of
2532 the language-independent GNU compiler options.
2533 For example, you might compile a file @code{some_class.m} like this:
2536 gcc -g -fgnu-runtime -O -c some_class.m
2540 In this example, @option{-fgnu-runtime} is an option meant only for
2541 Objective-C and Objective-C++ programs; you can use the other options with
2542 any language supported by GCC@.
2544 Note that since Objective-C is an extension of the C language, Objective-C
2545 compilations may also use options specific to the C front-end (e.g.,
2546 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2547 C++-specific options (e.g., @option{-Wabi}).
2549 Here is a list of options that are @emph{only} for compiling Objective-C
2550 and Objective-C++ programs:
2553 @item -fconstant-string-class=@var{class-name}
2554 @opindex fconstant-string-class
2555 Use @var{class-name} as the name of the class to instantiate for each
2556 literal string specified with the syntax @code{@@"@dots{}"}. The default
2557 class name is @code{NXConstantString} if the GNU runtime is being used, and
2558 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2559 @option{-fconstant-cfstrings} option, if also present, will override the
2560 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2561 to be laid out as constant CoreFoundation strings.
2564 @opindex fgnu-runtime
2565 Generate object code compatible with the standard GNU Objective-C
2566 runtime. This is the default for most types of systems.
2568 @item -fnext-runtime
2569 @opindex fnext-runtime
2570 Generate output compatible with the NeXT runtime. This is the default
2571 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2572 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2575 @item -fno-nil-receivers
2576 @opindex fno-nil-receivers
2577 Assume that all Objective-C message dispatches (@code{[receiver
2578 message:arg]}) in this translation unit ensure that the receiver is
2579 not @code{nil}. This allows for more efficient entry points in the
2580 runtime to be used. This option is only available in conjunction with
2581 the NeXT runtime and ABI version 0 or 1.
2583 @item -fobjc-abi-version=@var{n}
2584 @opindex fobjc-abi-version
2585 Use version @var{n} of the Objective-C ABI for the selected runtime.
2586 This option is currently supported only for the NeXT runtime. In that
2587 case, Version 0 is the traditional (32-bit) ABI without support for
2588 properties and other Objective-C 2.0 additions. Version 1 is the
2589 traditional (32-bit) ABI with support for properties and other
2590 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2591 nothing is specified, the default is Version 0 on 32-bit target
2592 machines, and Version 2 on 64-bit target machines.
2594 @item -fobjc-call-cxx-cdtors
2595 @opindex fobjc-call-cxx-cdtors
2596 For each Objective-C class, check if any of its instance variables is a
2597 C++ object with a non-trivial default constructor. If so, synthesize a
2598 special @code{- (id) .cxx_construct} instance method that will run
2599 non-trivial default constructors on any such instance variables, in order,
2600 and then return @code{self}. Similarly, check if any instance variable
2601 is a C++ object with a non-trivial destructor, and if so, synthesize a
2602 special @code{- (void) .cxx_destruct} method that will run
2603 all such default destructors, in reverse order.
2605 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2606 methods thusly generated will only operate on instance variables
2607 declared in the current Objective-C class, and not those inherited
2608 from superclasses. It is the responsibility of the Objective-C
2609 runtime to invoke all such methods in an object's inheritance
2610 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2611 by the runtime immediately after a new object instance is allocated;
2612 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2613 before the runtime deallocates an object instance.
2615 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2616 support for invoking the @code{- (id) .cxx_construct} and
2617 @code{- (void) .cxx_destruct} methods.
2619 @item -fobjc-direct-dispatch
2620 @opindex fobjc-direct-dispatch
2621 Allow fast jumps to the message dispatcher. On Darwin this is
2622 accomplished via the comm page.
2624 @item -fobjc-exceptions
2625 @opindex fobjc-exceptions
2626 Enable syntactic support for structured exception handling in
2627 Objective-C, similar to what is offered by C++ and Java. This option
2628 is required to use the Objective-C keywords @code{@@try},
2629 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2630 @code{@@synchronized}. This option is available with both the GNU
2631 runtime and the NeXT runtime (but not available in conjunction with
2632 the NeXT runtime on Mac OS X 10.2 and earlier).
2636 Enable garbage collection (GC) in Objective-C and Objective-C++
2637 programs. This option is only available with the NeXT runtime; the
2638 GNU runtime has a different garbage collection implementation that
2639 does not require special compiler flags.
2641 @item -fobjc-nilcheck
2642 @opindex fobjc-nilcheck
2643 For the NeXT runtime with version 2 of the ABI, check for a nil
2644 receiver in method invocations before doing the actual method call.
2645 This is the default and can be disabled using
2646 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2647 checked for nil in this way no matter what this flag is set to.
2648 Currently this flag does nothing when the GNU runtime, or an older
2649 version of the NeXT runtime ABI, is used.
2651 @item -fobjc-std=objc1
2653 Conform to the language syntax of Objective-C 1.0, the language
2654 recognized by GCC 4.0. This only affects the Objective-C additions to
2655 the C/C++ language; it does not affect conformance to C/C++ standards,
2656 which is controlled by the separate C/C++ dialect option flags. When
2657 this option is used with the Objective-C or Objective-C++ compiler,
2658 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2659 This is useful if you need to make sure that your Objective-C code can
2660 be compiled with older versions of GCC.
2662 @item -freplace-objc-classes
2663 @opindex freplace-objc-classes
2664 Emit a special marker instructing @command{ld(1)} not to statically link in
2665 the resulting object file, and allow @command{dyld(1)} to load it in at
2666 run time instead. This is used in conjunction with the Fix-and-Continue
2667 debugging mode, where the object file in question may be recompiled and
2668 dynamically reloaded in the course of program execution, without the need
2669 to restart the program itself. Currently, Fix-and-Continue functionality
2670 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2675 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2676 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2677 compile time) with static class references that get initialized at load time,
2678 which improves run-time performance. Specifying the @option{-fzero-link} flag
2679 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2680 to be retained. This is useful in Zero-Link debugging mode, since it allows
2681 for individual class implementations to be modified during program execution.
2682 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2683 regardless of command line options.
2687 Dump interface declarations for all classes seen in the source file to a
2688 file named @file{@var{sourcename}.decl}.
2690 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2691 @opindex Wassign-intercept
2692 @opindex Wno-assign-intercept
2693 Warn whenever an Objective-C assignment is being intercepted by the
2696 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2697 @opindex Wno-protocol
2699 If a class is declared to implement a protocol, a warning is issued for
2700 every method in the protocol that is not implemented by the class. The
2701 default behavior is to issue a warning for every method not explicitly
2702 implemented in the class, even if a method implementation is inherited
2703 from the superclass. If you use the @option{-Wno-protocol} option, then
2704 methods inherited from the superclass are considered to be implemented,
2705 and no warning is issued for them.
2707 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2709 @opindex Wno-selector
2710 Warn if multiple methods of different types for the same selector are
2711 found during compilation. The check is performed on the list of methods
2712 in the final stage of compilation. Additionally, a check is performed
2713 for each selector appearing in a @code{@@selector(@dots{})}
2714 expression, and a corresponding method for that selector has been found
2715 during compilation. Because these checks scan the method table only at
2716 the end of compilation, these warnings are not produced if the final
2717 stage of compilation is not reached, for example because an error is
2718 found during compilation, or because the @option{-fsyntax-only} option is
2721 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2722 @opindex Wstrict-selector-match
2723 @opindex Wno-strict-selector-match
2724 Warn if multiple methods with differing argument and/or return types are
2725 found for a given selector when attempting to send a message using this
2726 selector to a receiver of type @code{id} or @code{Class}. When this flag
2727 is off (which is the default behavior), the compiler will omit such warnings
2728 if any differences found are confined to types which share the same size
2731 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2732 @opindex Wundeclared-selector
2733 @opindex Wno-undeclared-selector
2734 Warn if a @code{@@selector(@dots{})} expression referring to an
2735 undeclared selector is found. A selector is considered undeclared if no
2736 method with that name has been declared before the
2737 @code{@@selector(@dots{})} expression, either explicitly in an
2738 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2739 an @code{@@implementation} section. This option always performs its
2740 checks as soon as a @code{@@selector(@dots{})} expression is found,
2741 while @option{-Wselector} only performs its checks in the final stage of
2742 compilation. This also enforces the coding style convention
2743 that methods and selectors must be declared before being used.
2745 @item -print-objc-runtime-info
2746 @opindex print-objc-runtime-info
2747 Generate C header describing the largest structure that is passed by
2752 @node Language Independent Options
2753 @section Options to Control Diagnostic Messages Formatting
2754 @cindex options to control diagnostics formatting
2755 @cindex diagnostic messages
2756 @cindex message formatting
2758 Traditionally, diagnostic messages have been formatted irrespective of
2759 the output device's aspect (e.g.@: its width, @dots{}). The options described
2760 below can be used to control the diagnostic messages formatting
2761 algorithm, e.g.@: how many characters per line, how often source location
2762 information should be reported. Right now, only the C++ front end can
2763 honor these options. However it is expected, in the near future, that
2764 the remaining front ends would be able to digest them correctly.
2767 @item -fmessage-length=@var{n}
2768 @opindex fmessage-length
2769 Try to format error messages so that they fit on lines of about @var{n}
2770 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2771 the front ends supported by GCC@. If @var{n} is zero, then no
2772 line-wrapping will be done; each error message will appear on a single
2775 @opindex fdiagnostics-show-location
2776 @item -fdiagnostics-show-location=once
2777 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2778 reporter to emit @emph{once} source location information; that is, in
2779 case the message is too long to fit on a single physical line and has to
2780 be wrapped, the source location won't be emitted (as prefix) again,
2781 over and over, in subsequent continuation lines. This is the default
2784 @item -fdiagnostics-show-location=every-line
2785 Only meaningful in line-wrapping mode. Instructs the diagnostic
2786 messages reporter to emit the same source location information (as
2787 prefix) for physical lines that result from the process of breaking
2788 a message which is too long to fit on a single line.
2790 @item -fno-diagnostics-show-option
2791 @opindex fno-diagnostics-show-option
2792 @opindex fdiagnostics-show-option
2793 By default, each diagnostic emitted includes text which indicates the
2794 command line option that directly controls the diagnostic (if such an
2795 option is known to the diagnostic machinery). Specifying the
2796 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2798 @item -Wcoverage-mismatch
2799 @opindex Wcoverage-mismatch
2800 Warn if feedback profiles do not match when using the
2801 @option{-fprofile-use} option.
2802 If a source file was changed between @option{-fprofile-gen} and
2803 @option{-fprofile-use}, the files with the profile feedback can fail
2804 to match the source file and GCC can not use the profile feedback
2805 information. By default, this warning is enabled and is treated as an
2806 error. @option{-Wno-coverage-mismatch} can be used to disable the
2807 warning or @option{-Wno-error=coverage-mismatch} can be used to
2808 disable the error. Disable the error for this warning can result in
2809 poorly optimized code, so disabling the error is useful only in the
2810 case of very minor changes such as bug fixes to an existing code-base.
2811 Completely disabling the warning is not recommended.
2815 @node Warning Options
2816 @section Options to Request or Suppress Warnings
2817 @cindex options to control warnings
2818 @cindex warning messages
2819 @cindex messages, warning
2820 @cindex suppressing warnings
2822 Warnings are diagnostic messages that report constructions which
2823 are not inherently erroneous but which are risky or suggest there
2824 may have been an error.
2826 The following language-independent options do not enable specific
2827 warnings but control the kinds of diagnostics produced by GCC.
2830 @cindex syntax checking
2832 @opindex fsyntax-only
2833 Check the code for syntax errors, but don't do anything beyond that.
2835 @item -fmax-errors=@var{n}
2836 @opindex fmax-errors
2837 Limits the maximum number of error messages to @var{n}, at which point
2838 GCC bails out rather than attempting to continue processing the source
2839 code. If @var{n} is 0 (the default), there is no limit on the number
2840 of error messages produced. If @option{-Wfatal-errors} is also
2841 specified, then @option{-Wfatal-errors} takes precedence over this
2846 Inhibit all warning messages.
2851 Make all warnings into errors.
2856 Make the specified warning into an error. The specifier for a warning
2857 is appended, for example @option{-Werror=switch} turns the warnings
2858 controlled by @option{-Wswitch} into errors. This switch takes a
2859 negative form, to be used to negate @option{-Werror} for specific
2860 warnings, for example @option{-Wno-error=switch} makes
2861 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2864 The warning message for each controllable warning includes the
2865 option which controls the warning. That option can then be used with
2866 @option{-Werror=} and @option{-Wno-error=} as described above.
2867 (Printing of the option in the warning message can be disabled using the
2868 @option{-fno-diagnostics-show-option} flag.)
2870 Note that specifying @option{-Werror=}@var{foo} automatically implies
2871 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2874 @item -Wfatal-errors
2875 @opindex Wfatal-errors
2876 @opindex Wno-fatal-errors
2877 This option causes the compiler to abort compilation on the first error
2878 occurred rather than trying to keep going and printing further error
2883 You can request many specific warnings with options beginning
2884 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2885 implicit declarations. Each of these specific warning options also
2886 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2887 example, @option{-Wno-implicit}. This manual lists only one of the
2888 two forms, whichever is not the default. For further,
2889 language-specific options also refer to @ref{C++ Dialect Options} and
2890 @ref{Objective-C and Objective-C++ Dialect Options}.
2892 When an unrecognized warning option is requested (e.g.,
2893 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2894 that the option is not recognized. However, if the @option{-Wno-} form
2895 is used, the behavior is slightly different: No diagnostic will be
2896 produced for @option{-Wno-unknown-warning} unless other diagnostics
2897 are being produced. This allows the use of new @option{-Wno-} options
2898 with old compilers, but if something goes wrong, the compiler will
2899 warn that an unrecognized option was used.
2904 Issue all the warnings demanded by strict ISO C and ISO C++;
2905 reject all programs that use forbidden extensions, and some other
2906 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2907 version of the ISO C standard specified by any @option{-std} option used.
2909 Valid ISO C and ISO C++ programs should compile properly with or without
2910 this option (though a rare few will require @option{-ansi} or a
2911 @option{-std} option specifying the required version of ISO C)@. However,
2912 without this option, certain GNU extensions and traditional C and C++
2913 features are supported as well. With this option, they are rejected.
2915 @option{-pedantic} does not cause warning messages for use of the
2916 alternate keywords whose names begin and end with @samp{__}. Pedantic
2917 warnings are also disabled in the expression that follows
2918 @code{__extension__}. However, only system header files should use
2919 these escape routes; application programs should avoid them.
2920 @xref{Alternate Keywords}.
2922 Some users try to use @option{-pedantic} to check programs for strict ISO
2923 C conformance. They soon find that it does not do quite what they want:
2924 it finds some non-ISO practices, but not all---only those for which
2925 ISO C @emph{requires} a diagnostic, and some others for which
2926 diagnostics have been added.
2928 A feature to report any failure to conform to ISO C might be useful in
2929 some instances, but would require considerable additional work and would
2930 be quite different from @option{-pedantic}. We don't have plans to
2931 support such a feature in the near future.
2933 Where the standard specified with @option{-std} represents a GNU
2934 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2935 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2936 extended dialect is based. Warnings from @option{-pedantic} are given
2937 where they are required by the base standard. (It would not make sense
2938 for such warnings to be given only for features not in the specified GNU
2939 C dialect, since by definition the GNU dialects of C include all
2940 features the compiler supports with the given option, and there would be
2941 nothing to warn about.)
2943 @item -pedantic-errors
2944 @opindex pedantic-errors
2945 Like @option{-pedantic}, except that errors are produced rather than
2951 This enables all the warnings about constructions that some users
2952 consider questionable, and that are easy to avoid (or modify to
2953 prevent the warning), even in conjunction with macros. This also
2954 enables some language-specific warnings described in @ref{C++ Dialect
2955 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2957 @option{-Wall} turns on the following warning flags:
2959 @gccoptlist{-Waddress @gol
2960 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2962 -Wchar-subscripts @gol
2963 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2964 -Wimplicit-int @r{(C and Objective-C only)} @gol
2965 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2968 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2969 -Wmaybe-uninitialized @gol
2970 -Wmissing-braces @gol
2976 -Wsequence-point @gol
2977 -Wsign-compare @r{(only in C++)} @gol
2978 -Wstrict-aliasing @gol
2979 -Wstrict-overflow=1 @gol
2982 -Wuninitialized @gol
2983 -Wunknown-pragmas @gol
2984 -Wunused-function @gol
2987 -Wunused-variable @gol
2988 -Wvolatile-register-var @gol
2991 Note that some warning flags are not implied by @option{-Wall}. Some of
2992 them warn about constructions that users generally do not consider
2993 questionable, but which occasionally you might wish to check for;
2994 others warn about constructions that are necessary or hard to avoid in
2995 some cases, and there is no simple way to modify the code to suppress
2996 the warning. Some of them are enabled by @option{-Wextra} but many of
2997 them must be enabled individually.
3003 This enables some extra warning flags that are not enabled by
3004 @option{-Wall}. (This option used to be called @option{-W}. The older
3005 name is still supported, but the newer name is more descriptive.)
3007 @gccoptlist{-Wclobbered @gol
3009 -Wignored-qualifiers @gol
3010 -Wmissing-field-initializers @gol
3011 -Wmissing-parameter-type @r{(C only)} @gol
3012 -Wold-style-declaration @r{(C only)} @gol
3013 -Woverride-init @gol
3016 -Wuninitialized @gol
3017 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3018 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3021 The option @option{-Wextra} also prints warning messages for the
3027 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3028 @samp{>}, or @samp{>=}.
3031 (C++ only) An enumerator and a non-enumerator both appear in a
3032 conditional expression.
3035 (C++ only) Ambiguous virtual bases.
3038 (C++ only) Subscripting an array which has been declared @samp{register}.
3041 (C++ only) Taking the address of a variable which has been declared
3045 (C++ only) A base class is not initialized in a derived class' copy
3050 @item -Wchar-subscripts
3051 @opindex Wchar-subscripts
3052 @opindex Wno-char-subscripts
3053 Warn if an array subscript has type @code{char}. This is a common cause
3054 of error, as programmers often forget that this type is signed on some
3056 This warning is enabled by @option{-Wall}.
3060 @opindex Wno-comment
3061 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3062 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3063 This warning is enabled by @option{-Wall}.
3066 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3068 Suppress warning messages emitted by @code{#warning} directives.
3070 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3071 @opindex Wdouble-promotion
3072 @opindex Wno-double-promotion
3073 Give a warning when a value of type @code{float} is implicitly
3074 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3075 floating-point unit implement @code{float} in hardware, but emulate
3076 @code{double} in software. On such a machine, doing computations
3077 using @code{double} values is much more expensive because of the
3078 overhead required for software emulation.
3080 It is easy to accidentally do computations with @code{double} because
3081 floating-point literals are implicitly of type @code{double}. For
3085 float area(float radius)
3087 return 3.14159 * radius * radius;
3091 the compiler will perform the entire computation with @code{double}
3092 because the floating-point literal is a @code{double}.
3097 @opindex ffreestanding
3098 @opindex fno-builtin
3099 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3100 the arguments supplied have types appropriate to the format string
3101 specified, and that the conversions specified in the format string make
3102 sense. This includes standard functions, and others specified by format
3103 attributes (@pxref{Function Attributes}), in the @code{printf},
3104 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3105 not in the C standard) families (or other target-specific families).
3106 Which functions are checked without format attributes having been
3107 specified depends on the standard version selected, and such checks of
3108 functions without the attribute specified are disabled by
3109 @option{-ffreestanding} or @option{-fno-builtin}.
3111 The formats are checked against the format features supported by GNU
3112 libc version 2.2. These include all ISO C90 and C99 features, as well
3113 as features from the Single Unix Specification and some BSD and GNU
3114 extensions. Other library implementations may not support all these
3115 features; GCC does not support warning about features that go beyond a
3116 particular library's limitations. However, if @option{-pedantic} is used
3117 with @option{-Wformat}, warnings will be given about format features not
3118 in the selected standard version (but not for @code{strfmon} formats,
3119 since those are not in any version of the C standard). @xref{C Dialect
3120 Options,,Options Controlling C Dialect}.
3122 Since @option{-Wformat} also checks for null format arguments for
3123 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3125 @option{-Wformat} is included in @option{-Wall}. For more control over some
3126 aspects of format checking, the options @option{-Wformat-y2k},
3127 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3128 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3129 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3132 @opindex Wformat-y2k
3133 @opindex Wno-format-y2k
3134 If @option{-Wformat} is specified, also warn about @code{strftime}
3135 formats which may yield only a two-digit year.
3137 @item -Wno-format-contains-nul
3138 @opindex Wno-format-contains-nul
3139 @opindex Wformat-contains-nul
3140 If @option{-Wformat} is specified, do not warn about format strings that
3143 @item -Wno-format-extra-args
3144 @opindex Wno-format-extra-args
3145 @opindex Wformat-extra-args
3146 If @option{-Wformat} is specified, do not warn about excess arguments to a
3147 @code{printf} or @code{scanf} format function. The C standard specifies
3148 that such arguments are ignored.
3150 Where the unused arguments lie between used arguments that are
3151 specified with @samp{$} operand number specifications, normally
3152 warnings are still given, since the implementation could not know what
3153 type to pass to @code{va_arg} to skip the unused arguments. However,
3154 in the case of @code{scanf} formats, this option will suppress the
3155 warning if the unused arguments are all pointers, since the Single
3156 Unix Specification says that such unused arguments are allowed.
3158 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3159 @opindex Wno-format-zero-length
3160 @opindex Wformat-zero-length
3161 If @option{-Wformat} is specified, do not warn about zero-length formats.
3162 The C standard specifies that zero-length formats are allowed.
3164 @item -Wformat-nonliteral
3165 @opindex Wformat-nonliteral
3166 @opindex Wno-format-nonliteral
3167 If @option{-Wformat} is specified, also warn if the format string is not a
3168 string literal and so cannot be checked, unless the format function
3169 takes its format arguments as a @code{va_list}.
3171 @item -Wformat-security
3172 @opindex Wformat-security
3173 @opindex Wno-format-security
3174 If @option{-Wformat} is specified, also warn about uses of format
3175 functions that represent possible security problems. At present, this
3176 warns about calls to @code{printf} and @code{scanf} functions where the
3177 format string is not a string literal and there are no format arguments,
3178 as in @code{printf (foo);}. This may be a security hole if the format
3179 string came from untrusted input and contains @samp{%n}. (This is
3180 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3181 in future warnings may be added to @option{-Wformat-security} that are not
3182 included in @option{-Wformat-nonliteral}.)
3186 @opindex Wno-format=2
3187 Enable @option{-Wformat} plus format checks not included in
3188 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3189 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3191 @item -Wnonnull @r{(C and Objective-C only)}
3193 @opindex Wno-nonnull
3194 Warn about passing a null pointer for arguments marked as
3195 requiring a non-null value by the @code{nonnull} function attribute.
3197 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3198 can be disabled with the @option{-Wno-nonnull} option.
3200 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3202 @opindex Wno-init-self
3203 Warn about uninitialized variables which are initialized with themselves.
3204 Note this option can only be used with the @option{-Wuninitialized} option.
3206 For example, GCC will warn about @code{i} being uninitialized in the
3207 following snippet only when @option{-Winit-self} has been specified:
3218 @item -Wimplicit-int @r{(C and Objective-C only)}
3219 @opindex Wimplicit-int
3220 @opindex Wno-implicit-int
3221 Warn when a declaration does not specify a type.
3222 This warning is enabled by @option{-Wall}.
3224 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3225 @opindex Wimplicit-function-declaration
3226 @opindex Wno-implicit-function-declaration
3227 Give a warning whenever a function is used before being declared. In
3228 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3229 enabled by default and it is made into an error by
3230 @option{-pedantic-errors}. This warning is also enabled by
3233 @item -Wimplicit @r{(C and Objective-C only)}
3235 @opindex Wno-implicit
3236 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3237 This warning is enabled by @option{-Wall}.
3239 @item -Wignored-qualifiers @r{(C and C++ only)}
3240 @opindex Wignored-qualifiers
3241 @opindex Wno-ignored-qualifiers
3242 Warn if the return type of a function has a type qualifier
3243 such as @code{const}. For ISO C such a type qualifier has no effect,
3244 since the value returned by a function is not an lvalue.
3245 For C++, the warning is only emitted for scalar types or @code{void}.
3246 ISO C prohibits qualified @code{void} return types on function
3247 definitions, so such return types always receive a warning
3248 even without this option.
3250 This warning is also enabled by @option{-Wextra}.
3255 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3256 a function with external linkage, returning int, taking either zero
3257 arguments, two, or three arguments of appropriate types. This warning
3258 is enabled by default in C++ and is enabled by either @option{-Wall}
3259 or @option{-pedantic}.
3261 @item -Wmissing-braces
3262 @opindex Wmissing-braces
3263 @opindex Wno-missing-braces
3264 Warn if an aggregate or union initializer is not fully bracketed. In
3265 the following example, the initializer for @samp{a} is not fully
3266 bracketed, but that for @samp{b} is fully bracketed.
3269 int a[2][2] = @{ 0, 1, 2, 3 @};
3270 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3273 This warning is enabled by @option{-Wall}.
3275 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3276 @opindex Wmissing-include-dirs
3277 @opindex Wno-missing-include-dirs
3278 Warn if a user-supplied include directory does not exist.
3281 @opindex Wparentheses
3282 @opindex Wno-parentheses
3283 Warn if parentheses are omitted in certain contexts, such
3284 as when there is an assignment in a context where a truth value
3285 is expected, or when operators are nested whose precedence people
3286 often get confused about.
3288 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3289 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3290 interpretation from that of ordinary mathematical notation.
3292 Also warn about constructions where there may be confusion to which
3293 @code{if} statement an @code{else} branch belongs. Here is an example of
3308 In C/C++, every @code{else} branch belongs to the innermost possible
3309 @code{if} statement, which in this example is @code{if (b)}. This is
3310 often not what the programmer expected, as illustrated in the above
3311 example by indentation the programmer chose. When there is the
3312 potential for this confusion, GCC will issue a warning when this flag
3313 is specified. To eliminate the warning, add explicit braces around
3314 the innermost @code{if} statement so there is no way the @code{else}
3315 could belong to the enclosing @code{if}. The resulting code would
3332 Also warn for dangerous uses of the
3333 ?: with omitted middle operand GNU extension. When the condition
3334 in the ?: operator is a boolean expression the omitted value will
3335 be always 1. Often the user expects it to be a value computed
3336 inside the conditional expression instead.
3338 This warning is enabled by @option{-Wall}.
3340 @item -Wsequence-point
3341 @opindex Wsequence-point
3342 @opindex Wno-sequence-point
3343 Warn about code that may have undefined semantics because of violations
3344 of sequence point rules in the C and C++ standards.
3346 The C and C++ standards defines the order in which expressions in a C/C++
3347 program are evaluated in terms of @dfn{sequence points}, which represent
3348 a partial ordering between the execution of parts of the program: those
3349 executed before the sequence point, and those executed after it. These
3350 occur after the evaluation of a full expression (one which is not part
3351 of a larger expression), after the evaluation of the first operand of a
3352 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3353 function is called (but after the evaluation of its arguments and the
3354 expression denoting the called function), and in certain other places.
3355 Other than as expressed by the sequence point rules, the order of
3356 evaluation of subexpressions of an expression is not specified. All
3357 these rules describe only a partial order rather than a total order,
3358 since, for example, if two functions are called within one expression
3359 with no sequence point between them, the order in which the functions
3360 are called is not specified. However, the standards committee have
3361 ruled that function calls do not overlap.
3363 It is not specified when between sequence points modifications to the
3364 values of objects take effect. Programs whose behavior depends on this
3365 have undefined behavior; the C and C++ standards specify that ``Between
3366 the previous and next sequence point an object shall have its stored
3367 value modified at most once by the evaluation of an expression.
3368 Furthermore, the prior value shall be read only to determine the value
3369 to be stored.''. If a program breaks these rules, the results on any
3370 particular implementation are entirely unpredictable.
3372 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3373 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3374 diagnosed by this option, and it may give an occasional false positive
3375 result, but in general it has been found fairly effective at detecting
3376 this sort of problem in programs.
3378 The standard is worded confusingly, therefore there is some debate
3379 over the precise meaning of the sequence point rules in subtle cases.
3380 Links to discussions of the problem, including proposed formal
3381 definitions, may be found on the GCC readings page, at
3382 @uref{http://gcc.gnu.org/@/readings.html}.
3384 This warning is enabled by @option{-Wall} for C and C++.
3387 @opindex Wreturn-type
3388 @opindex Wno-return-type
3389 Warn whenever a function is defined with a return-type that defaults
3390 to @code{int}. Also warn about any @code{return} statement with no
3391 return-value in a function whose return-type is not @code{void}
3392 (falling off the end of the function body is considered returning
3393 without a value), and about a @code{return} statement with an
3394 expression in a function whose return-type is @code{void}.
3396 For C++, a function without return type always produces a diagnostic
3397 message, even when @option{-Wno-return-type} is specified. The only
3398 exceptions are @samp{main} and functions defined in system headers.
3400 This warning is enabled by @option{-Wall}.
3405 Warn whenever a @code{switch} statement has an index of enumerated type
3406 and lacks a @code{case} for one or more of the named codes of that
3407 enumeration. (The presence of a @code{default} label prevents this
3408 warning.) @code{case} labels outside the enumeration range also
3409 provoke warnings when this option is used (even if there is a
3410 @code{default} label).
3411 This warning is enabled by @option{-Wall}.
3413 @item -Wswitch-default
3414 @opindex Wswitch-default
3415 @opindex Wno-switch-default
3416 Warn whenever a @code{switch} statement does not have a @code{default}
3420 @opindex Wswitch-enum
3421 @opindex Wno-switch-enum
3422 Warn whenever a @code{switch} statement has an index of enumerated type
3423 and lacks a @code{case} for one or more of the named codes of that
3424 enumeration. @code{case} labels outside the enumeration range also
3425 provoke warnings when this option is used. The only difference
3426 between @option{-Wswitch} and this option is that this option gives a
3427 warning about an omitted enumeration code even if there is a
3428 @code{default} label.
3430 @item -Wsync-nand @r{(C and C++ only)}
3432 @opindex Wno-sync-nand
3433 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3434 built-in functions are used. These functions changed semantics in GCC 4.4.
3438 @opindex Wno-trigraphs
3439 Warn if any trigraphs are encountered that might change the meaning of
3440 the program (trigraphs within comments are not warned about).
3441 This warning is enabled by @option{-Wall}.
3443 @item -Wunused-but-set-parameter
3444 @opindex Wunused-but-set-parameter
3445 @opindex Wno-unused-but-set-parameter
3446 Warn whenever a function parameter is assigned to, but otherwise unused
3447 (aside from its declaration).
3449 To suppress this warning use the @samp{unused} attribute
3450 (@pxref{Variable Attributes}).
3452 This warning is also enabled by @option{-Wunused} together with
3455 @item -Wunused-but-set-variable
3456 @opindex Wunused-but-set-variable
3457 @opindex Wno-unused-but-set-variable
3458 Warn whenever a local variable is assigned to, but otherwise unused
3459 (aside from its declaration).
3460 This warning is enabled by @option{-Wall}.
3462 To suppress this warning use the @samp{unused} attribute
3463 (@pxref{Variable Attributes}).
3465 This warning is also enabled by @option{-Wunused}, which is enabled
3468 @item -Wunused-function
3469 @opindex Wunused-function
3470 @opindex Wno-unused-function
3471 Warn whenever a static function is declared but not defined or a
3472 non-inline static function is unused.
3473 This warning is enabled by @option{-Wall}.
3475 @item -Wunused-label
3476 @opindex Wunused-label
3477 @opindex Wno-unused-label
3478 Warn whenever a label is declared but not used.
3479 This warning is enabled by @option{-Wall}.
3481 To suppress this warning use the @samp{unused} attribute
3482 (@pxref{Variable Attributes}).
3484 @item -Wunused-parameter
3485 @opindex Wunused-parameter
3486 @opindex Wno-unused-parameter
3487 Warn whenever a function parameter is unused aside from its declaration.
3489 To suppress this warning use the @samp{unused} attribute
3490 (@pxref{Variable Attributes}).
3492 @item -Wno-unused-result
3493 @opindex Wunused-result
3494 @opindex Wno-unused-result
3495 Do not warn if a caller of a function marked with attribute
3496 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3497 its return value. The default is @option{-Wunused-result}.
3499 @item -Wunused-variable
3500 @opindex Wunused-variable
3501 @opindex Wno-unused-variable
3502 Warn whenever a local variable or non-constant static variable is unused
3503 aside from its declaration.
3504 This warning is enabled by @option{-Wall}.
3506 To suppress this warning use the @samp{unused} attribute
3507 (@pxref{Variable Attributes}).
3509 @item -Wunused-value
3510 @opindex Wunused-value
3511 @opindex Wno-unused-value
3512 Warn whenever a statement computes a result that is explicitly not
3513 used. To suppress this warning cast the unused expression to
3514 @samp{void}. This includes an expression-statement or the left-hand
3515 side of a comma expression that contains no side effects. For example,
3516 an expression such as @samp{x[i,j]} will cause a warning, while
3517 @samp{x[(void)i,j]} will not.
3519 This warning is enabled by @option{-Wall}.
3524 All the above @option{-Wunused} options combined.
3526 In order to get a warning about an unused function parameter, you must
3527 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3528 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3530 @item -Wuninitialized
3531 @opindex Wuninitialized
3532 @opindex Wno-uninitialized
3533 Warn if an automatic variable is used without first being initialized
3534 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3535 warn if a non-static reference or non-static @samp{const} member
3536 appears in a class without constructors.
3538 If you want to warn about code which uses the uninitialized value of the
3539 variable in its own initializer, use the @option{-Winit-self} option.
3541 These warnings occur for individual uninitialized or clobbered
3542 elements of structure, union or array variables as well as for
3543 variables which are uninitialized or clobbered as a whole. They do
3544 not occur for variables or elements declared @code{volatile}. Because
3545 these warnings depend on optimization, the exact variables or elements
3546 for which there are warnings will depend on the precise optimization
3547 options and version of GCC used.
3549 Note that there may be no warning about a variable that is used only
3550 to compute a value that itself is never used, because such
3551 computations may be deleted by data flow analysis before the warnings
3554 @item -Wmaybe-uninitialized
3555 @opindex Wmaybe-uninitialized
3556 @opindex Wno-maybe-uninitialized
3557 For an automatic variable, if there exists a path from the function
3558 entry to a use of the variable that is initialized, but there exist
3559 some other paths the variable is not initialized, the compiler will
3560 emit a warning if it can not prove the uninitialized paths do not
3561 happen at runtime. These warnings are made optional because GCC is
3562 not smart enough to see all the reasons why the code might be correct
3563 despite appearing to have an error. Here is one example of how
3584 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3585 always initialized, but GCC doesn't know this. To suppress the
3586 warning, the user needs to provide a default case with assert(0) or
3589 @cindex @code{longjmp} warnings
3590 This option also warns when a non-volatile automatic variable might be
3591 changed by a call to @code{longjmp}. These warnings as well are possible
3592 only in optimizing compilation.
3594 The compiler sees only the calls to @code{setjmp}. It cannot know
3595 where @code{longjmp} will be called; in fact, a signal handler could
3596 call it at any point in the code. As a result, you may get a warning
3597 even when there is in fact no problem because @code{longjmp} cannot
3598 in fact be called at the place which would cause a problem.
3600 Some spurious warnings can be avoided if you declare all the functions
3601 you use that never return as @code{noreturn}. @xref{Function
3604 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3606 @item -Wunknown-pragmas
3607 @opindex Wunknown-pragmas
3608 @opindex Wno-unknown-pragmas
3609 @cindex warning for unknown pragmas
3610 @cindex unknown pragmas, warning
3611 @cindex pragmas, warning of unknown
3612 Warn when a #pragma directive is encountered which is not understood by
3613 GCC@. If this command line option is used, warnings will even be issued
3614 for unknown pragmas in system header files. This is not the case if
3615 the warnings were only enabled by the @option{-Wall} command line option.
3618 @opindex Wno-pragmas
3620 Do not warn about misuses of pragmas, such as incorrect parameters,
3621 invalid syntax, or conflicts between pragmas. See also
3622 @samp{-Wunknown-pragmas}.
3624 @item -Wstrict-aliasing
3625 @opindex Wstrict-aliasing
3626 @opindex Wno-strict-aliasing
3627 This option is only active when @option{-fstrict-aliasing} is active.
3628 It warns about code which might break the strict aliasing rules that the
3629 compiler is using for optimization. The warning does not catch all
3630 cases, but does attempt to catch the more common pitfalls. It is
3631 included in @option{-Wall}.
3632 It is equivalent to @option{-Wstrict-aliasing=3}
3634 @item -Wstrict-aliasing=n
3635 @opindex Wstrict-aliasing=n
3636 @opindex Wno-strict-aliasing=n
3637 This option is only active when @option{-fstrict-aliasing} is active.
3638 It warns about code which might break the strict aliasing rules that the
3639 compiler is using for optimization.
3640 Higher levels correspond to higher accuracy (fewer false positives).
3641 Higher levels also correspond to more effort, similar to the way -O works.
3642 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3645 Level 1: Most aggressive, quick, least accurate.
3646 Possibly useful when higher levels
3647 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3648 false negatives. However, it has many false positives.
3649 Warns for all pointer conversions between possibly incompatible types,
3650 even if never dereferenced. Runs in the frontend only.
3652 Level 2: Aggressive, quick, not too precise.
3653 May still have many false positives (not as many as level 1 though),
3654 and few false negatives (but possibly more than level 1).
3655 Unlike level 1, it only warns when an address is taken. Warns about
3656 incomplete types. Runs in the frontend only.
3658 Level 3 (default for @option{-Wstrict-aliasing}):
3659 Should have very few false positives and few false
3660 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3661 Takes care of the common pun+dereference pattern in the frontend:
3662 @code{*(int*)&some_float}.
3663 If optimization is enabled, it also runs in the backend, where it deals
3664 with multiple statement cases using flow-sensitive points-to information.
3665 Only warns when the converted pointer is dereferenced.
3666 Does not warn about incomplete types.
3668 @item -Wstrict-overflow
3669 @itemx -Wstrict-overflow=@var{n}
3670 @opindex Wstrict-overflow
3671 @opindex Wno-strict-overflow
3672 This option is only active when @option{-fstrict-overflow} is active.
3673 It warns about cases where the compiler optimizes based on the
3674 assumption that signed overflow does not occur. Note that it does not
3675 warn about all cases where the code might overflow: it only warns
3676 about cases where the compiler implements some optimization. Thus
3677 this warning depends on the optimization level.
3679 An optimization which assumes that signed overflow does not occur is
3680 perfectly safe if the values of the variables involved are such that
3681 overflow never does, in fact, occur. Therefore this warning can
3682 easily give a false positive: a warning about code which is not
3683 actually a problem. To help focus on important issues, several
3684 warning levels are defined. No warnings are issued for the use of
3685 undefined signed overflow when estimating how many iterations a loop
3686 will require, in particular when determining whether a loop will be
3690 @item -Wstrict-overflow=1
3691 Warn about cases which are both questionable and easy to avoid. For
3692 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3693 compiler will simplify this to @code{1}. This level of
3694 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3695 are not, and must be explicitly requested.
3697 @item -Wstrict-overflow=2
3698 Also warn about other cases where a comparison is simplified to a
3699 constant. For example: @code{abs (x) >= 0}. This can only be
3700 simplified when @option{-fstrict-overflow} is in effect, because
3701 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3702 zero. @option{-Wstrict-overflow} (with no level) is the same as
3703 @option{-Wstrict-overflow=2}.
3705 @item -Wstrict-overflow=3
3706 Also warn about other cases where a comparison is simplified. For
3707 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3709 @item -Wstrict-overflow=4
3710 Also warn about other simplifications not covered by the above cases.
3711 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3713 @item -Wstrict-overflow=5
3714 Also warn about cases where the compiler reduces the magnitude of a
3715 constant involved in a comparison. For example: @code{x + 2 > y} will
3716 be simplified to @code{x + 1 >= y}. This is reported only at the
3717 highest warning level because this simplification applies to many
3718 comparisons, so this warning level will give a very large number of
3722 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3723 @opindex Wsuggest-attribute=
3724 @opindex Wno-suggest-attribute=
3725 Warn for cases where adding an attribute may be beneficial. The
3726 attributes currently supported are listed below.
3729 @item -Wsuggest-attribute=pure
3730 @itemx -Wsuggest-attribute=const
3731 @itemx -Wsuggest-attribute=noreturn
3732 @opindex Wsuggest-attribute=pure
3733 @opindex Wno-suggest-attribute=pure
3734 @opindex Wsuggest-attribute=const
3735 @opindex Wno-suggest-attribute=const
3736 @opindex Wsuggest-attribute=noreturn
3737 @opindex Wno-suggest-attribute=noreturn
3739 Warn about functions which might be candidates for attributes
3740 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3741 functions visible in other compilation units or (in the case of @code{pure} and
3742 @code{const}) if it cannot prove that the function returns normally. A function
3743 returns normally if it doesn't contain an infinite loop nor returns abnormally
3744 by throwing, calling @code{abort()} or trapping. This analysis requires option
3745 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3746 higher. Higher optimization levels improve the accuracy of the analysis.
3749 @item -Warray-bounds
3750 @opindex Wno-array-bounds
3751 @opindex Warray-bounds
3752 This option is only active when @option{-ftree-vrp} is active
3753 (default for @option{-O2} and above). It warns about subscripts to arrays
3754 that are always out of bounds. This warning is enabled by @option{-Wall}.
3756 @item -Wno-div-by-zero
3757 @opindex Wno-div-by-zero
3758 @opindex Wdiv-by-zero
3759 Do not warn about compile-time integer division by zero. Floating point
3760 division by zero is not warned about, as it can be a legitimate way of
3761 obtaining infinities and NaNs.
3763 @item -Wsystem-headers
3764 @opindex Wsystem-headers
3765 @opindex Wno-system-headers
3766 @cindex warnings from system headers
3767 @cindex system headers, warnings from
3768 Print warning messages for constructs found in system header files.
3769 Warnings from system headers are normally suppressed, on the assumption
3770 that they usually do not indicate real problems and would only make the
3771 compiler output harder to read. Using this command line option tells
3772 GCC to emit warnings from system headers as if they occurred in user
3773 code. However, note that using @option{-Wall} in conjunction with this
3774 option will @emph{not} warn about unknown pragmas in system
3775 headers---for that, @option{-Wunknown-pragmas} must also be used.
3778 @opindex Wtrampolines
3779 @opindex Wno-trampolines
3780 Warn about trampolines generated for pointers to nested functions.
3782 A trampoline is a small piece of data or code that is created at run
3783 time on the stack when the address of a nested function is taken, and
3784 is used to call the nested function indirectly. For some targets, it
3785 is made up of data only and thus requires no special treatment. But,
3786 for most targets, it is made up of code and thus requires the stack
3787 to be made executable in order for the program to work properly.
3790 @opindex Wfloat-equal
3791 @opindex Wno-float-equal
3792 Warn if floating point values are used in equality comparisons.
3794 The idea behind this is that sometimes it is convenient (for the
3795 programmer) to consider floating-point values as approximations to
3796 infinitely precise real numbers. If you are doing this, then you need
3797 to compute (by analyzing the code, or in some other way) the maximum or
3798 likely maximum error that the computation introduces, and allow for it
3799 when performing comparisons (and when producing output, but that's a
3800 different problem). In particular, instead of testing for equality, you
3801 would check to see whether the two values have ranges that overlap; and
3802 this is done with the relational operators, so equality comparisons are
3805 @item -Wtraditional @r{(C and Objective-C only)}
3806 @opindex Wtraditional
3807 @opindex Wno-traditional
3808 Warn about certain constructs that behave differently in traditional and
3809 ISO C@. Also warn about ISO C constructs that have no traditional C
3810 equivalent, and/or problematic constructs which should be avoided.
3814 Macro parameters that appear within string literals in the macro body.
3815 In traditional C macro replacement takes place within string literals,
3816 but does not in ISO C@.
3819 In traditional C, some preprocessor directives did not exist.
3820 Traditional preprocessors would only consider a line to be a directive
3821 if the @samp{#} appeared in column 1 on the line. Therefore
3822 @option{-Wtraditional} warns about directives that traditional C
3823 understands but would ignore because the @samp{#} does not appear as the
3824 first character on the line. It also suggests you hide directives like
3825 @samp{#pragma} not understood by traditional C by indenting them. Some
3826 traditional implementations would not recognize @samp{#elif}, so it
3827 suggests avoiding it altogether.
3830 A function-like macro that appears without arguments.
3833 The unary plus operator.
3836 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3837 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3838 constants.) Note, these suffixes appear in macros defined in the system
3839 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3840 Use of these macros in user code might normally lead to spurious
3841 warnings, however GCC's integrated preprocessor has enough context to
3842 avoid warning in these cases.
3845 A function declared external in one block and then used after the end of
3849 A @code{switch} statement has an operand of type @code{long}.
3852 A non-@code{static} function declaration follows a @code{static} one.
3853 This construct is not accepted by some traditional C compilers.
3856 The ISO type of an integer constant has a different width or
3857 signedness from its traditional type. This warning is only issued if
3858 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3859 typically represent bit patterns, are not warned about.
3862 Usage of ISO string concatenation is detected.
3865 Initialization of automatic aggregates.
3868 Identifier conflicts with labels. Traditional C lacks a separate
3869 namespace for labels.
3872 Initialization of unions. If the initializer is zero, the warning is
3873 omitted. This is done under the assumption that the zero initializer in
3874 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3875 initializer warnings and relies on default initialization to zero in the
3879 Conversions by prototypes between fixed/floating point values and vice
3880 versa. The absence of these prototypes when compiling with traditional
3881 C would cause serious problems. This is a subset of the possible
3882 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3885 Use of ISO C style function definitions. This warning intentionally is
3886 @emph{not} issued for prototype declarations or variadic functions
3887 because these ISO C features will appear in your code when using
3888 libiberty's traditional C compatibility macros, @code{PARAMS} and
3889 @code{VPARAMS}. This warning is also bypassed for nested functions
3890 because that feature is already a GCC extension and thus not relevant to
3891 traditional C compatibility.
3894 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3895 @opindex Wtraditional-conversion
3896 @opindex Wno-traditional-conversion
3897 Warn if a prototype causes a type conversion that is different from what
3898 would happen to the same argument in the absence of a prototype. This
3899 includes conversions of fixed point to floating and vice versa, and
3900 conversions changing the width or signedness of a fixed point argument
3901 except when the same as the default promotion.
3903 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3904 @opindex Wdeclaration-after-statement
3905 @opindex Wno-declaration-after-statement
3906 Warn when a declaration is found after a statement in a block. This
3907 construct, known from C++, was introduced with ISO C99 and is by default
3908 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3909 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3914 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3916 @item -Wno-endif-labels
3917 @opindex Wno-endif-labels
3918 @opindex Wendif-labels
3919 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3924 Warn whenever a local variable or type declaration shadows another variable,
3925 parameter, type, or class member (in C++), or whenever a built-in function
3926 is shadowed. Note that in C++, the compiler will not warn if a local variable
3927 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3929 @item -Wlarger-than=@var{len}
3930 @opindex Wlarger-than=@var{len}
3931 @opindex Wlarger-than-@var{len}
3932 Warn whenever an object of larger than @var{len} bytes is defined.
3934 @item -Wframe-larger-than=@var{len}
3935 @opindex Wframe-larger-than
3936 Warn if the size of a function frame is larger than @var{len} bytes.
3937 The computation done to determine the stack frame size is approximate
3938 and not conservative.
3939 The actual requirements may be somewhat greater than @var{len}
3940 even if you do not get a warning. In addition, any space allocated
3941 via @code{alloca}, variable-length arrays, or related constructs
3942 is not included by the compiler when determining
3943 whether or not to issue a warning.
3945 @item -Wstack-usage=@var{len}
3946 @opindex Wstack-usage
3947 Warn if the stack usage of a function might be larger than @var{len} bytes.
3948 The computation done to determine the stack usage is conservative.
3949 Any space allocated via @code{alloca}, variable-length arrays, or related
3950 constructs is included by the compiler when determining whether or not to
3953 The message is in keeping with the output of @option{-fstack-usage}.
3957 If the stack usage is fully static but exceeds the specified amount, it's:
3960 Â warning: stack usage is 1120 bytes
3963 If the stack usage is (partly) dynamic but bounded, it's:
3966 Â warning: stack usage might be 1648 bytes
3969 If the stack usage is (partly) dynamic and not bounded, it's:
3972 Â warning: stack usage might be unbounded
3976 @item -Wunsafe-loop-optimizations
3977 @opindex Wunsafe-loop-optimizations
3978 @opindex Wno-unsafe-loop-optimizations
3979 Warn if the loop cannot be optimized because the compiler could not
3980 assume anything on the bounds of the loop indices. With
3981 @option{-funsafe-loop-optimizations} warn if the compiler made
3984 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
3985 @opindex Wno-pedantic-ms-format
3986 @opindex Wpedantic-ms-format
3987 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
3988 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
3989 depending on the MS runtime, when you are using the options @option{-Wformat}
3990 and @option{-pedantic} without gnu-extensions.
3992 @item -Wpointer-arith
3993 @opindex Wpointer-arith
3994 @opindex Wno-pointer-arith
3995 Warn about anything that depends on the ``size of'' a function type or
3996 of @code{void}. GNU C assigns these types a size of 1, for
3997 convenience in calculations with @code{void *} pointers and pointers
3998 to functions. In C++, warn also when an arithmetic operation involves
3999 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4002 @opindex Wtype-limits
4003 @opindex Wno-type-limits
4004 Warn if a comparison is always true or always false due to the limited
4005 range of the data type, but do not warn for constant expressions. For
4006 example, warn if an unsigned variable is compared against zero with
4007 @samp{<} or @samp{>=}. This warning is also enabled by
4010 @item -Wbad-function-cast @r{(C and Objective-C only)}
4011 @opindex Wbad-function-cast
4012 @opindex Wno-bad-function-cast
4013 Warn whenever a function call is cast to a non-matching type.
4014 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4016 @item -Wc++-compat @r{(C and Objective-C only)}
4017 Warn about ISO C constructs that are outside of the common subset of
4018 ISO C and ISO C++, e.g.@: request for implicit conversion from
4019 @code{void *} to a pointer to non-@code{void} type.
4021 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
4022 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
4023 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
4024 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
4028 @opindex Wno-cast-qual
4029 Warn whenever a pointer is cast so as to remove a type qualifier from
4030 the target type. For example, warn if a @code{const char *} is cast
4031 to an ordinary @code{char *}.
4033 Also warn when making a cast which introduces a type qualifier in an
4034 unsafe way. For example, casting @code{char **} to @code{const char **}
4035 is unsafe, as in this example:
4038 /* p is char ** value. */
4039 const char **q = (const char **) p;
4040 /* Assignment of readonly string to const char * is OK. */
4042 /* Now char** pointer points to read-only memory. */
4047 @opindex Wcast-align
4048 @opindex Wno-cast-align
4049 Warn whenever a pointer is cast such that the required alignment of the
4050 target is increased. For example, warn if a @code{char *} is cast to
4051 an @code{int *} on machines where integers can only be accessed at
4052 two- or four-byte boundaries.
4054 @item -Wwrite-strings
4055 @opindex Wwrite-strings
4056 @opindex Wno-write-strings
4057 When compiling C, give string constants the type @code{const
4058 char[@var{length}]} so that copying the address of one into a
4059 non-@code{const} @code{char *} pointer will get a warning. These
4060 warnings will help you find at compile time code that can try to write
4061 into a string constant, but only if you have been very careful about
4062 using @code{const} in declarations and prototypes. Otherwise, it will
4063 just be a nuisance. This is why we did not make @option{-Wall} request
4066 When compiling C++, warn about the deprecated conversion from string
4067 literals to @code{char *}. This warning is enabled by default for C++
4072 @opindex Wno-clobbered
4073 Warn for variables that might be changed by @samp{longjmp} or
4074 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4077 @opindex Wconversion
4078 @opindex Wno-conversion
4079 Warn for implicit conversions that may alter a value. This includes
4080 conversions between real and integer, like @code{abs (x)} when
4081 @code{x} is @code{double}; conversions between signed and unsigned,
4082 like @code{unsigned ui = -1}; and conversions to smaller types, like
4083 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4084 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4085 changed by the conversion like in @code{abs (2.0)}. Warnings about
4086 conversions between signed and unsigned integers can be disabled by
4087 using @option{-Wno-sign-conversion}.
4089 For C++, also warn for confusing overload resolution for user-defined
4090 conversions; and conversions that will never use a type conversion
4091 operator: conversions to @code{void}, the same type, a base class or a
4092 reference to them. Warnings about conversions between signed and
4093 unsigned integers are disabled by default in C++ unless
4094 @option{-Wsign-conversion} is explicitly enabled.
4096 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4097 @opindex Wconversion-null
4098 @opindex Wno-conversion-null
4099 Do not warn for conversions between @code{NULL} and non-pointer
4100 types. @option{-Wconversion-null} is enabled by default.
4103 @opindex Wempty-body
4104 @opindex Wno-empty-body
4105 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4106 while} statement. This warning is also enabled by @option{-Wextra}.
4108 @item -Wenum-compare
4109 @opindex Wenum-compare
4110 @opindex Wno-enum-compare
4111 Warn about a comparison between values of different enum types. In C++
4112 this warning is enabled by default. In C this warning is enabled by
4115 @item -Wjump-misses-init @r{(C, Objective-C only)}
4116 @opindex Wjump-misses-init
4117 @opindex Wno-jump-misses-init
4118 Warn if a @code{goto} statement or a @code{switch} statement jumps
4119 forward across the initialization of a variable, or jumps backward to a
4120 label after the variable has been initialized. This only warns about
4121 variables which are initialized when they are declared. This warning is
4122 only supported for C and Objective C; in C++ this sort of branch is an
4125 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4126 can be disabled with the @option{-Wno-jump-misses-init} option.
4128 @item -Wsign-compare
4129 @opindex Wsign-compare
4130 @opindex Wno-sign-compare
4131 @cindex warning for comparison of signed and unsigned values
4132 @cindex comparison of signed and unsigned values, warning
4133 @cindex signed and unsigned values, comparison warning
4134 Warn when a comparison between signed and unsigned values could produce
4135 an incorrect result when the signed value is converted to unsigned.
4136 This warning is also enabled by @option{-Wextra}; to get the other warnings
4137 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4139 @item -Wsign-conversion
4140 @opindex Wsign-conversion
4141 @opindex Wno-sign-conversion
4142 Warn for implicit conversions that may change the sign of an integer
4143 value, like assigning a signed integer expression to an unsigned
4144 integer variable. An explicit cast silences the warning. In C, this
4145 option is enabled also by @option{-Wconversion}.
4149 @opindex Wno-address
4150 Warn about suspicious uses of memory addresses. These include using
4151 the address of a function in a conditional expression, such as
4152 @code{void func(void); if (func)}, and comparisons against the memory
4153 address of a string literal, such as @code{if (x == "abc")}. Such
4154 uses typically indicate a programmer error: the address of a function
4155 always evaluates to true, so their use in a conditional usually
4156 indicate that the programmer forgot the parentheses in a function
4157 call; and comparisons against string literals result in unspecified
4158 behavior and are not portable in C, so they usually indicate that the
4159 programmer intended to use @code{strcmp}. This warning is enabled by
4163 @opindex Wlogical-op
4164 @opindex Wno-logical-op
4165 Warn about suspicious uses of logical operators in expressions.
4166 This includes using logical operators in contexts where a
4167 bit-wise operator is likely to be expected.
4169 @item -Waggregate-return
4170 @opindex Waggregate-return
4171 @opindex Wno-aggregate-return
4172 Warn if any functions that return structures or unions are defined or
4173 called. (In languages where you can return an array, this also elicits
4176 @item -Wno-attributes
4177 @opindex Wno-attributes
4178 @opindex Wattributes
4179 Do not warn if an unexpected @code{__attribute__} is used, such as
4180 unrecognized attributes, function attributes applied to variables,
4181 etc. This will not stop errors for incorrect use of supported
4184 @item -Wno-builtin-macro-redefined
4185 @opindex Wno-builtin-macro-redefined
4186 @opindex Wbuiltin-macro-redefined
4187 Do not warn if certain built-in macros are redefined. This suppresses
4188 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4189 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4191 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4192 @opindex Wstrict-prototypes
4193 @opindex Wno-strict-prototypes
4194 Warn if a function is declared or defined without specifying the
4195 argument types. (An old-style function definition is permitted without
4196 a warning if preceded by a declaration which specifies the argument
4199 @item -Wold-style-declaration @r{(C and Objective-C only)}
4200 @opindex Wold-style-declaration
4201 @opindex Wno-old-style-declaration
4202 Warn for obsolescent usages, according to the C Standard, in a
4203 declaration. For example, warn if storage-class specifiers like
4204 @code{static} are not the first things in a declaration. This warning
4205 is also enabled by @option{-Wextra}.
4207 @item -Wold-style-definition @r{(C and Objective-C only)}
4208 @opindex Wold-style-definition
4209 @opindex Wno-old-style-definition
4210 Warn if an old-style function definition is used. A warning is given
4211 even if there is a previous prototype.
4213 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4214 @opindex Wmissing-parameter-type
4215 @opindex Wno-missing-parameter-type
4216 A function parameter is declared without a type specifier in K&R-style
4223 This warning is also enabled by @option{-Wextra}.
4225 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4226 @opindex Wmissing-prototypes
4227 @opindex Wno-missing-prototypes
4228 Warn if a global function is defined without a previous prototype
4229 declaration. This warning is issued even if the definition itself
4230 provides a prototype. The aim is to detect global functions that fail
4231 to be declared in header files.
4233 @item -Wmissing-declarations
4234 @opindex Wmissing-declarations
4235 @opindex Wno-missing-declarations
4236 Warn if a global function is defined without a previous declaration.
4237 Do so even if the definition itself provides a prototype.
4238 Use this option to detect global functions that are not declared in
4239 header files. In C++, no warnings are issued for function templates,
4240 or for inline functions, or for functions in anonymous namespaces.
4242 @item -Wmissing-field-initializers
4243 @opindex Wmissing-field-initializers
4244 @opindex Wno-missing-field-initializers
4248 Warn if a structure's initializer has some fields missing. For
4249 example, the following code would cause such a warning, because
4250 @code{x.h} is implicitly zero:
4253 struct s @{ int f, g, h; @};
4254 struct s x = @{ 3, 4 @};
4257 This option does not warn about designated initializers, so the following
4258 modification would not trigger a warning:
4261 struct s @{ int f, g, h; @};
4262 struct s x = @{ .f = 3, .g = 4 @};
4265 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4266 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4268 @item -Wmissing-format-attribute
4269 @opindex Wmissing-format-attribute
4270 @opindex Wno-missing-format-attribute
4273 Warn about function pointers which might be candidates for @code{format}
4274 attributes. Note these are only possible candidates, not absolute ones.
4275 GCC will guess that function pointers with @code{format} attributes that
4276 are used in assignment, initialization, parameter passing or return
4277 statements should have a corresponding @code{format} attribute in the
4278 resulting type. I.e.@: the left-hand side of the assignment or
4279 initialization, the type of the parameter variable, or the return type
4280 of the containing function respectively should also have a @code{format}
4281 attribute to avoid the warning.
4283 GCC will also warn about function definitions which might be
4284 candidates for @code{format} attributes. Again, these are only
4285 possible candidates. GCC will guess that @code{format} attributes
4286 might be appropriate for any function that calls a function like
4287 @code{vprintf} or @code{vscanf}, but this might not always be the
4288 case, and some functions for which @code{format} attributes are
4289 appropriate may not be detected.
4291 @item -Wno-multichar
4292 @opindex Wno-multichar
4294 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4295 Usually they indicate a typo in the user's code, as they have
4296 implementation-defined values, and should not be used in portable code.
4298 @item -Wnormalized=<none|id|nfc|nfkc>
4299 @opindex Wnormalized=
4302 @cindex character set, input normalization
4303 In ISO C and ISO C++, two identifiers are different if they are
4304 different sequences of characters. However, sometimes when characters
4305 outside the basic ASCII character set are used, you can have two
4306 different character sequences that look the same. To avoid confusion,
4307 the ISO 10646 standard sets out some @dfn{normalization rules} which
4308 when applied ensure that two sequences that look the same are turned into
4309 the same sequence. GCC can warn you if you are using identifiers which
4310 have not been normalized; this option controls that warning.
4312 There are four levels of warning that GCC supports. The default is
4313 @option{-Wnormalized=nfc}, which warns about any identifier which is
4314 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4315 recommended form for most uses.
4317 Unfortunately, there are some characters which ISO C and ISO C++ allow
4318 in identifiers that when turned into NFC aren't allowable as
4319 identifiers. That is, there's no way to use these symbols in portable
4320 ISO C or C++ and have all your identifiers in NFC@.
4321 @option{-Wnormalized=id} suppresses the warning for these characters.
4322 It is hoped that future versions of the standards involved will correct
4323 this, which is why this option is not the default.
4325 You can switch the warning off for all characters by writing
4326 @option{-Wnormalized=none}. You would only want to do this if you
4327 were using some other normalization scheme (like ``D''), because
4328 otherwise you can easily create bugs that are literally impossible to see.
4330 Some characters in ISO 10646 have distinct meanings but look identical
4331 in some fonts or display methodologies, especially once formatting has
4332 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4333 LETTER N'', will display just like a regular @code{n} which has been
4334 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4335 normalization scheme to convert all these into a standard form as
4336 well, and GCC will warn if your code is not in NFKC if you use
4337 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4338 about every identifier that contains the letter O because it might be
4339 confused with the digit 0, and so is not the default, but may be
4340 useful as a local coding convention if the programming environment is
4341 unable to be fixed to display these characters distinctly.
4343 @item -Wno-deprecated
4344 @opindex Wno-deprecated
4345 @opindex Wdeprecated
4346 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4348 @item -Wno-deprecated-declarations
4349 @opindex Wno-deprecated-declarations
4350 @opindex Wdeprecated-declarations
4351 Do not warn about uses of functions (@pxref{Function Attributes}),
4352 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4353 Attributes}) marked as deprecated by using the @code{deprecated}
4357 @opindex Wno-overflow
4359 Do not warn about compile-time overflow in constant expressions.
4361 @item -Woverride-init @r{(C and Objective-C only)}
4362 @opindex Woverride-init
4363 @opindex Wno-override-init
4367 Warn if an initialized field without side effects is overridden when
4368 using designated initializers (@pxref{Designated Inits, , Designated
4371 This warning is included in @option{-Wextra}. To get other
4372 @option{-Wextra} warnings without this one, use @samp{-Wextra
4373 -Wno-override-init}.
4378 Warn if a structure is given the packed attribute, but the packed
4379 attribute has no effect on the layout or size of the structure.
4380 Such structures may be mis-aligned for little benefit. For
4381 instance, in this code, the variable @code{f.x} in @code{struct bar}
4382 will be misaligned even though @code{struct bar} does not itself
4383 have the packed attribute:
4390 @} __attribute__((packed));
4398 @item -Wpacked-bitfield-compat
4399 @opindex Wpacked-bitfield-compat
4400 @opindex Wno-packed-bitfield-compat
4401 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4402 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4403 the change can lead to differences in the structure layout. GCC
4404 informs you when the offset of such a field has changed in GCC 4.4.
4405 For example there is no longer a 4-bit padding between field @code{a}
4406 and @code{b} in this structure:
4413 @} __attribute__ ((packed));
4416 This warning is enabled by default. Use
4417 @option{-Wno-packed-bitfield-compat} to disable this warning.
4422 Warn if padding is included in a structure, either to align an element
4423 of the structure or to align the whole structure. Sometimes when this
4424 happens it is possible to rearrange the fields of the structure to
4425 reduce the padding and so make the structure smaller.
4427 @item -Wredundant-decls
4428 @opindex Wredundant-decls
4429 @opindex Wno-redundant-decls
4430 Warn if anything is declared more than once in the same scope, even in
4431 cases where multiple declaration is valid and changes nothing.
4433 @item -Wnested-externs @r{(C and Objective-C only)}
4434 @opindex Wnested-externs
4435 @opindex Wno-nested-externs
4436 Warn if an @code{extern} declaration is encountered within a function.
4441 Warn if a function can not be inlined and it was declared as inline.
4442 Even with this option, the compiler will not warn about failures to
4443 inline functions declared in system headers.
4445 The compiler uses a variety of heuristics to determine whether or not
4446 to inline a function. For example, the compiler takes into account
4447 the size of the function being inlined and the amount of inlining
4448 that has already been done in the current function. Therefore,
4449 seemingly insignificant changes in the source program can cause the
4450 warnings produced by @option{-Winline} to appear or disappear.
4452 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4453 @opindex Wno-invalid-offsetof
4454 @opindex Winvalid-offsetof
4455 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4456 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4457 to a non-POD type is undefined. In existing C++ implementations,
4458 however, @samp{offsetof} typically gives meaningful results even when
4459 applied to certain kinds of non-POD types. (Such as a simple
4460 @samp{struct} that fails to be a POD type only by virtue of having a
4461 constructor.) This flag is for users who are aware that they are
4462 writing nonportable code and who have deliberately chosen to ignore the
4465 The restrictions on @samp{offsetof} may be relaxed in a future version
4466 of the C++ standard.
4468 @item -Wno-int-to-pointer-cast
4469 @opindex Wno-int-to-pointer-cast
4470 @opindex Wint-to-pointer-cast
4471 Suppress warnings from casts to pointer type of an integer of a
4472 different size. In C++, casting to a pointer type of smaller size is
4473 an error. @option{Wint-to-pointer-cast} is enabled by default.
4476 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4477 @opindex Wno-pointer-to-int-cast
4478 @opindex Wpointer-to-int-cast
4479 Suppress warnings from casts from a pointer to an integer type of a
4483 @opindex Winvalid-pch
4484 @opindex Wno-invalid-pch
4485 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4486 the search path but can't be used.
4490 @opindex Wno-long-long
4491 Warn if @samp{long long} type is used. This is enabled by either
4492 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4493 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4495 @item -Wvariadic-macros
4496 @opindex Wvariadic-macros
4497 @opindex Wno-variadic-macros
4498 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4499 alternate syntax when in pedantic ISO C99 mode. This is default.
4500 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4505 Warn if variable length array is used in the code.
4506 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4507 the variable length array.
4509 @item -Wvolatile-register-var
4510 @opindex Wvolatile-register-var
4511 @opindex Wno-volatile-register-var
4512 Warn if a register variable is declared volatile. The volatile
4513 modifier does not inhibit all optimizations that may eliminate reads
4514 and/or writes to register variables. This warning is enabled by
4517 @item -Wdisabled-optimization
4518 @opindex Wdisabled-optimization
4519 @opindex Wno-disabled-optimization
4520 Warn if a requested optimization pass is disabled. This warning does
4521 not generally indicate that there is anything wrong with your code; it
4522 merely indicates that GCC's optimizers were unable to handle the code
4523 effectively. Often, the problem is that your code is too big or too
4524 complex; GCC will refuse to optimize programs when the optimization
4525 itself is likely to take inordinate amounts of time.
4527 @item -Wpointer-sign @r{(C and Objective-C only)}
4528 @opindex Wpointer-sign
4529 @opindex Wno-pointer-sign
4530 Warn for pointer argument passing or assignment with different signedness.
4531 This option is only supported for C and Objective-C@. It is implied by
4532 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4533 @option{-Wno-pointer-sign}.
4535 @item -Wstack-protector
4536 @opindex Wstack-protector
4537 @opindex Wno-stack-protector
4538 This option is only active when @option{-fstack-protector} is active. It
4539 warns about functions that will not be protected against stack smashing.
4542 @opindex Wno-mudflap
4543 Suppress warnings about constructs that cannot be instrumented by
4546 @item -Woverlength-strings
4547 @opindex Woverlength-strings
4548 @opindex Wno-overlength-strings
4549 Warn about string constants which are longer than the ``minimum
4550 maximum'' length specified in the C standard. Modern compilers
4551 generally allow string constants which are much longer than the
4552 standard's minimum limit, but very portable programs should avoid
4553 using longer strings.
4555 The limit applies @emph{after} string constant concatenation, and does
4556 not count the trailing NUL@. In C90, the limit was 509 characters; in
4557 C99, it was raised to 4095. C++98 does not specify a normative
4558 minimum maximum, so we do not diagnose overlength strings in C++@.
4560 This option is implied by @option{-pedantic}, and can be disabled with
4561 @option{-Wno-overlength-strings}.
4563 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4564 @opindex Wunsuffixed-float-constants
4566 GCC will issue a warning for any floating constant that does not have
4567 a suffix. When used together with @option{-Wsystem-headers} it will
4568 warn about such constants in system header files. This can be useful
4569 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4570 from the decimal floating-point extension to C99.
4573 @node Debugging Options
4574 @section Options for Debugging Your Program or GCC
4575 @cindex options, debugging
4576 @cindex debugging information options
4578 GCC has various special options that are used for debugging
4579 either your program or GCC:
4584 Produce debugging information in the operating system's native format
4585 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4588 On most systems that use stabs format, @option{-g} enables use of extra
4589 debugging information that only GDB can use; this extra information
4590 makes debugging work better in GDB but will probably make other debuggers
4592 refuse to read the program. If you want to control for certain whether
4593 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4594 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4596 GCC allows you to use @option{-g} with
4597 @option{-O}. The shortcuts taken by optimized code may occasionally
4598 produce surprising results: some variables you declared may not exist
4599 at all; flow of control may briefly move where you did not expect it;
4600 some statements may not be executed because they compute constant
4601 results or their values were already at hand; some statements may
4602 execute in different places because they were moved out of loops.
4604 Nevertheless it proves possible to debug optimized output. This makes
4605 it reasonable to use the optimizer for programs that might have bugs.
4607 The following options are useful when GCC is generated with the
4608 capability for more than one debugging format.
4612 Produce debugging information for use by GDB@. This means to use the
4613 most expressive format available (DWARF 2, stabs, or the native format
4614 if neither of those are supported), including GDB extensions if at all
4619 Produce debugging information in stabs format (if that is supported),
4620 without GDB extensions. This is the format used by DBX on most BSD
4621 systems. On MIPS, Alpha and System V Release 4 systems this option
4622 produces stabs debugging output which is not understood by DBX or SDB@.
4623 On System V Release 4 systems this option requires the GNU assembler.
4625 @item -feliminate-unused-debug-symbols
4626 @opindex feliminate-unused-debug-symbols
4627 Produce debugging information in stabs format (if that is supported),
4628 for only symbols that are actually used.
4630 @item -femit-class-debug-always
4631 Instead of emitting debugging information for a C++ class in only one
4632 object file, emit it in all object files using the class. This option
4633 should be used only with debuggers that are unable to handle the way GCC
4634 normally emits debugging information for classes because using this
4635 option will increase the size of debugging information by as much as a
4638 @item -fno-debug-types-section
4639 @opindex fno-debug-types-section
4640 @opindex fdebug-types-section
4641 By default when using DWARF v4 or higher type DIEs will be put into
4642 their own .debug_types section instead of making them part of the
4643 .debug_info section. It is more efficient to put them in a separate
4644 comdat sections since the linker will then be able to remove duplicates.
4645 But not all DWARF consumers support .debug_types sections yet.
4649 Produce debugging information in stabs format (if that is supported),
4650 using GNU extensions understood only by the GNU debugger (GDB)@. The
4651 use of these extensions is likely to make other debuggers crash or
4652 refuse to read the program.
4656 Produce debugging information in COFF format (if that is supported).
4657 This is the format used by SDB on most System V systems prior to
4662 Produce debugging information in XCOFF format (if that is supported).
4663 This is the format used by the DBX debugger on IBM RS/6000 systems.
4667 Produce debugging information in XCOFF format (if that is supported),
4668 using GNU extensions understood only by the GNU debugger (GDB)@. The
4669 use of these extensions is likely to make other debuggers crash or
4670 refuse to read the program, and may cause assemblers other than the GNU
4671 assembler (GAS) to fail with an error.
4673 @item -gdwarf-@var{version}
4674 @opindex gdwarf-@var{version}
4675 Produce debugging information in DWARF format (if that is
4676 supported). This is the format used by DBX on IRIX 6. The value
4677 of @var{version} may be either 2, 3 or 4; the default version is 2.
4679 Note that with DWARF version 2 some ports require, and will always
4680 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4682 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4683 for maximum benefit.
4685 @item -gstrict-dwarf
4686 @opindex gstrict-dwarf
4687 Disallow using extensions of later DWARF standard version than selected
4688 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4689 DWARF extensions from later standard versions is allowed.
4691 @item -gno-strict-dwarf
4692 @opindex gno-strict-dwarf
4693 Allow using extensions of later DWARF standard version than selected with
4694 @option{-gdwarf-@var{version}}.
4698 Produce debugging information in VMS debug format (if that is
4699 supported). This is the format used by DEBUG on VMS systems.
4702 @itemx -ggdb@var{level}
4703 @itemx -gstabs@var{level}
4704 @itemx -gcoff@var{level}
4705 @itemx -gxcoff@var{level}
4706 @itemx -gvms@var{level}
4707 Request debugging information and also use @var{level} to specify how
4708 much information. The default level is 2.
4710 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4713 Level 1 produces minimal information, enough for making backtraces in
4714 parts of the program that you don't plan to debug. This includes
4715 descriptions of functions and external variables, but no information
4716 about local variables and no line numbers.
4718 Level 3 includes extra information, such as all the macro definitions
4719 present in the program. Some debuggers support macro expansion when
4720 you use @option{-g3}.
4722 @option{-gdwarf-2} does not accept a concatenated debug level, because
4723 GCC used to support an option @option{-gdwarf} that meant to generate
4724 debug information in version 1 of the DWARF format (which is very
4725 different from version 2), and it would have been too confusing. That
4726 debug format is long obsolete, but the option cannot be changed now.
4727 Instead use an additional @option{-g@var{level}} option to change the
4728 debug level for DWARF.
4732 Turn off generation of debug info, if leaving out this option would have
4733 generated it, or turn it on at level 2 otherwise. The position of this
4734 argument in the command line does not matter, it takes effect after all
4735 other options are processed, and it does so only once, no matter how
4736 many times it is given. This is mainly intended to be used with
4737 @option{-fcompare-debug}.
4739 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4740 @opindex fdump-final-insns
4741 Dump the final internal representation (RTL) to @var{file}. If the
4742 optional argument is omitted (or if @var{file} is @code{.}), the name
4743 of the dump file will be determined by appending @code{.gkd} to the
4744 compilation output file name.
4746 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4747 @opindex fcompare-debug
4748 @opindex fno-compare-debug
4749 If no error occurs during compilation, run the compiler a second time,
4750 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4751 passed to the second compilation. Dump the final internal
4752 representation in both compilations, and print an error if they differ.
4754 If the equal sign is omitted, the default @option{-gtoggle} is used.
4756 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4757 and nonzero, implicitly enables @option{-fcompare-debug}. If
4758 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4759 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4762 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4763 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4764 of the final representation and the second compilation, preventing even
4765 @env{GCC_COMPARE_DEBUG} from taking effect.
4767 To verify full coverage during @option{-fcompare-debug} testing, set
4768 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4769 which GCC will reject as an invalid option in any actual compilation
4770 (rather than preprocessing, assembly or linking). To get just a
4771 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4772 not overridden} will do.
4774 @item -fcompare-debug-second
4775 @opindex fcompare-debug-second
4776 This option is implicitly passed to the compiler for the second
4777 compilation requested by @option{-fcompare-debug}, along with options to
4778 silence warnings, and omitting other options that would cause
4779 side-effect compiler outputs to files or to the standard output. Dump
4780 files and preserved temporary files are renamed so as to contain the
4781 @code{.gk} additional extension during the second compilation, to avoid
4782 overwriting those generated by the first.
4784 When this option is passed to the compiler driver, it causes the
4785 @emph{first} compilation to be skipped, which makes it useful for little
4786 other than debugging the compiler proper.
4788 @item -feliminate-dwarf2-dups
4789 @opindex feliminate-dwarf2-dups
4790 Compress DWARF2 debugging information by eliminating duplicated
4791 information about each symbol. This option only makes sense when
4792 generating DWARF2 debugging information with @option{-gdwarf-2}.
4794 @item -femit-struct-debug-baseonly
4795 Emit debug information for struct-like types
4796 only when the base name of the compilation source file
4797 matches the base name of file in which the struct was defined.
4799 This option substantially reduces the size of debugging information,
4800 but at significant potential loss in type information to the debugger.
4801 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4802 See @option{-femit-struct-debug-detailed} for more detailed control.
4804 This option works only with DWARF 2.
4806 @item -femit-struct-debug-reduced
4807 Emit debug information for struct-like types
4808 only when the base name of the compilation source file
4809 matches the base name of file in which the type was defined,
4810 unless the struct is a template or defined in a system header.
4812 This option significantly reduces the size of debugging information,
4813 with some potential loss in type information to the debugger.
4814 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4815 See @option{-femit-struct-debug-detailed} for more detailed control.
4817 This option works only with DWARF 2.
4819 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4820 Specify the struct-like types
4821 for which the compiler will generate debug information.
4822 The intent is to reduce duplicate struct debug information
4823 between different object files within the same program.
4825 This option is a detailed version of
4826 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4827 which will serve for most needs.
4829 A specification has the syntax@*
4830 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4832 The optional first word limits the specification to
4833 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4834 A struct type is used directly when it is the type of a variable, member.
4835 Indirect uses arise through pointers to structs.
4836 That is, when use of an incomplete struct would be legal, the use is indirect.
4838 @samp{struct one direct; struct two * indirect;}.
4840 The optional second word limits the specification to
4841 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4842 Generic structs are a bit complicated to explain.
4843 For C++, these are non-explicit specializations of template classes,
4844 or non-template classes within the above.
4845 Other programming languages have generics,
4846 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4848 The third word specifies the source files for those
4849 structs for which the compiler will emit debug information.
4850 The values @samp{none} and @samp{any} have the normal meaning.
4851 The value @samp{base} means that
4852 the base of name of the file in which the type declaration appears
4853 must match the base of the name of the main compilation file.
4854 In practice, this means that
4855 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4856 but types declared in other header will not.
4857 The value @samp{sys} means those types satisfying @samp{base}
4858 or declared in system or compiler headers.
4860 You may need to experiment to determine the best settings for your application.
4862 The default is @samp{-femit-struct-debug-detailed=all}.
4864 This option works only with DWARF 2.
4866 @item -fno-merge-debug-strings
4867 @opindex fmerge-debug-strings
4868 @opindex fno-merge-debug-strings
4869 Direct the linker to not merge together strings in the debugging
4870 information which are identical in different object files. Merging is
4871 not supported by all assemblers or linkers. Merging decreases the size
4872 of the debug information in the output file at the cost of increasing
4873 link processing time. Merging is enabled by default.
4875 @item -fdebug-prefix-map=@var{old}=@var{new}
4876 @opindex fdebug-prefix-map
4877 When compiling files in directory @file{@var{old}}, record debugging
4878 information describing them as in @file{@var{new}} instead.
4880 @item -fno-dwarf2-cfi-asm
4881 @opindex fdwarf2-cfi-asm
4882 @opindex fno-dwarf2-cfi-asm
4883 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4884 instead of using GAS @code{.cfi_*} directives.
4886 @cindex @command{prof}
4889 Generate extra code to write profile information suitable for the
4890 analysis program @command{prof}. You must use this option when compiling
4891 the source files you want data about, and you must also use it when
4894 @cindex @command{gprof}
4897 Generate extra code to write profile information suitable for the
4898 analysis program @command{gprof}. You must use this option when compiling
4899 the source files you want data about, and you must also use it when
4904 Makes the compiler print out each function name as it is compiled, and
4905 print some statistics about each pass when it finishes.
4908 @opindex ftime-report
4909 Makes the compiler print some statistics about the time consumed by each
4910 pass when it finishes.
4913 @opindex fmem-report
4914 Makes the compiler print some statistics about permanent memory
4915 allocation when it finishes.
4917 @item -fpre-ipa-mem-report
4918 @opindex fpre-ipa-mem-report
4919 @item -fpost-ipa-mem-report
4920 @opindex fpost-ipa-mem-report
4921 Makes the compiler print some statistics about permanent memory
4922 allocation before or after interprocedural optimization.
4925 @opindex fstack-usage
4926 Makes the compiler output stack usage information for the program, on a
4927 per-function basis. The filename for the dump is made by appending
4928 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4929 the output file, if explicitly specified and it is not an executable,
4930 otherwise it is the basename of the source file. An entry is made up
4935 The name of the function.
4939 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4942 The qualifier @code{static} means that the function manipulates the stack
4943 statically: a fixed number of bytes are allocated for the frame on function
4944 entry and released on function exit; no stack adjustments are otherwise made
4945 in the function. The second field is this fixed number of bytes.
4947 The qualifier @code{dynamic} means that the function manipulates the stack
4948 dynamically: in addition to the static allocation described above, stack
4949 adjustments are made in the body of the function, for example to push/pop
4950 arguments around function calls. If the qualifier @code{bounded} is also
4951 present, the amount of these adjustments is bounded at compile-time and
4952 the second field is an upper bound of the total amount of stack used by
4953 the function. If it is not present, the amount of these adjustments is
4954 not bounded at compile-time and the second field only represents the
4957 @item -fprofile-arcs
4958 @opindex fprofile-arcs
4959 Add code so that program flow @dfn{arcs} are instrumented. During
4960 execution the program records how many times each branch and call is
4961 executed and how many times it is taken or returns. When the compiled
4962 program exits it saves this data to a file called
4963 @file{@var{auxname}.gcda} for each source file. The data may be used for
4964 profile-directed optimizations (@option{-fbranch-probabilities}), or for
4965 test coverage analysis (@option{-ftest-coverage}). Each object file's
4966 @var{auxname} is generated from the name of the output file, if
4967 explicitly specified and it is not the final executable, otherwise it is
4968 the basename of the source file. In both cases any suffix is removed
4969 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
4970 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
4971 @xref{Cross-profiling}.
4973 @cindex @command{gcov}
4977 This option is used to compile and link code instrumented for coverage
4978 analysis. The option is a synonym for @option{-fprofile-arcs}
4979 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
4980 linking). See the documentation for those options for more details.
4985 Compile the source files with @option{-fprofile-arcs} plus optimization
4986 and code generation options. For test coverage analysis, use the
4987 additional @option{-ftest-coverage} option. You do not need to profile
4988 every source file in a program.
4991 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
4992 (the latter implies the former).
4995 Run the program on a representative workload to generate the arc profile
4996 information. This may be repeated any number of times. You can run
4997 concurrent instances of your program, and provided that the file system
4998 supports locking, the data files will be correctly updated. Also
4999 @code{fork} calls are detected and correctly handled (double counting
5003 For profile-directed optimizations, compile the source files again with
5004 the same optimization and code generation options plus
5005 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5006 Control Optimization}).
5009 For test coverage analysis, use @command{gcov} to produce human readable
5010 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5011 @command{gcov} documentation for further information.
5015 With @option{-fprofile-arcs}, for each function of your program GCC
5016 creates a program flow graph, then finds a spanning tree for the graph.
5017 Only arcs that are not on the spanning tree have to be instrumented: the
5018 compiler adds code to count the number of times that these arcs are
5019 executed. When an arc is the only exit or only entrance to a block, the
5020 instrumentation code can be added to the block; otherwise, a new basic
5021 block must be created to hold the instrumentation code.
5024 @item -ftest-coverage
5025 @opindex ftest-coverage
5026 Produce a notes file that the @command{gcov} code-coverage utility
5027 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5028 show program coverage. Each source file's note file is called
5029 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5030 above for a description of @var{auxname} and instructions on how to
5031 generate test coverage data. Coverage data will match the source files
5032 more closely, if you do not optimize.
5034 @item -fdbg-cnt-list
5035 @opindex fdbg-cnt-list
5036 Print the name and the counter upper bound for all debug counters.
5039 @item -fdbg-cnt=@var{counter-value-list}
5041 Set the internal debug counter upper bound. @var{counter-value-list}
5042 is a comma-separated list of @var{name}:@var{value} pairs
5043 which sets the upper bound of each debug counter @var{name} to @var{value}.
5044 All debug counters have the initial upper bound of @var{UINT_MAX},
5045 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5046 e.g. With -fdbg-cnt=dce:10,tail_call:0
5047 dbg_cnt(dce) will return true only for first 10 invocations
5049 @itemx -fenable-@var{kind}-@var{pass}
5050 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5054 This is a set of debugging options that are used to explicitly disable/enable
5055 optimization passes. For compiler users, regular options for enabling/disabling
5056 passes should be used instead.
5060 @item -fdisable-ipa-@var{pass}
5061 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5062 statically invoked in the compiler multiple times, the pass name should be
5063 appended with a sequential number starting from 1.
5065 @item -fdisable-rtl-@var{pass}
5066 @item -fdisable-rtl-@var{pass}=@var{range-list}
5067 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5068 statically invoked in the compiler multiple times, the pass name should be
5069 appended with a sequential number starting from 1. @var{range-list} is a comma
5070 seperated list of function ranges or assembler names. Each range is a number
5071 pair seperated by a colon. The range is inclusive in both ends. If the range
5072 is trivial, the number pair can be simplified as a single number. If the
5073 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5074 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5075 function header of a dump file, and the pass names can be dumped by using
5076 option @option{-fdump-passes}.
5078 @item -fdisable-tree-@var{pass}
5079 @item -fdisable-tree-@var{pass}=@var{range-list}
5080 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5083 @item -fenable-ipa-@var{pass}
5084 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5085 statically invoked in the compiler multiple times, the pass name should be
5086 appended with a sequential number starting from 1.
5088 @item -fenable-rtl-@var{pass}
5089 @item -fenable-rtl-@var{pass}=@var{range-list}
5090 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5091 description and examples.
5093 @item -fenable-tree-@var{pass}
5094 @item -fenable-tree-@var{pass}=@var{range-list}
5095 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5096 of option arguments.
5100 # disable ccp1 for all functions
5102 # disable complete unroll for function whose cgraph node uid is 1
5103 -fenable-tree-cunroll=1
5104 # disable gcse2 for functions at the following ranges [1,1],
5105 # [300,400], and [400,1000]
5106 # disable gcse2 for functions foo and foo2
5107 -fdisable-rtl-gcse2=foo,foo2
5108 # disable early inlining
5109 -fdisable-tree-einline
5110 # disable ipa inlining
5111 -fdisable-ipa-inline
5112 # enable tree full unroll
5113 -fenable-tree-unroll
5119 @item -d@var{letters}
5120 @itemx -fdump-rtl-@var{pass}
5122 Says to make debugging dumps during compilation at times specified by
5123 @var{letters}. This is used for debugging the RTL-based passes of the
5124 compiler. The file names for most of the dumps are made by appending
5125 a pass number and a word to the @var{dumpname}, and the files are
5126 created in the directory of the output file. Note that the pass
5127 number is computed statically as passes get registered into the pass
5128 manager. Thus the numbering is not related to the dynamic order of
5129 execution of passes. In particular, a pass installed by a plugin
5130 could have a number over 200 even if it executed quite early.
5131 @var{dumpname} is generated from the name of the output file, if
5132 explicitly specified and it is not an executable, otherwise it is the
5133 basename of the source file. These switches may have different effects
5134 when @option{-E} is used for preprocessing.
5136 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5137 @option{-d} option @var{letters}. Here are the possible
5138 letters for use in @var{pass} and @var{letters}, and their meanings:
5142 @item -fdump-rtl-alignments
5143 @opindex fdump-rtl-alignments
5144 Dump after branch alignments have been computed.
5146 @item -fdump-rtl-asmcons
5147 @opindex fdump-rtl-asmcons
5148 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5150 @item -fdump-rtl-auto_inc_dec
5151 @opindex fdump-rtl-auto_inc_dec
5152 Dump after auto-inc-dec discovery. This pass is only run on
5153 architectures that have auto inc or auto dec instructions.
5155 @item -fdump-rtl-barriers
5156 @opindex fdump-rtl-barriers
5157 Dump after cleaning up the barrier instructions.
5159 @item -fdump-rtl-bbpart
5160 @opindex fdump-rtl-bbpart
5161 Dump after partitioning hot and cold basic blocks.
5163 @item -fdump-rtl-bbro
5164 @opindex fdump-rtl-bbro
5165 Dump after block reordering.
5167 @item -fdump-rtl-btl1
5168 @itemx -fdump-rtl-btl2
5169 @opindex fdump-rtl-btl2
5170 @opindex fdump-rtl-btl2
5171 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5172 after the two branch
5173 target load optimization passes.
5175 @item -fdump-rtl-bypass
5176 @opindex fdump-rtl-bypass
5177 Dump after jump bypassing and control flow optimizations.
5179 @item -fdump-rtl-combine
5180 @opindex fdump-rtl-combine
5181 Dump after the RTL instruction combination pass.
5183 @item -fdump-rtl-compgotos
5184 @opindex fdump-rtl-compgotos
5185 Dump after duplicating the computed gotos.
5187 @item -fdump-rtl-ce1
5188 @itemx -fdump-rtl-ce2
5189 @itemx -fdump-rtl-ce3
5190 @opindex fdump-rtl-ce1
5191 @opindex fdump-rtl-ce2
5192 @opindex fdump-rtl-ce3
5193 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5194 @option{-fdump-rtl-ce3} enable dumping after the three
5195 if conversion passes.
5197 @itemx -fdump-rtl-cprop_hardreg
5198 @opindex fdump-rtl-cprop_hardreg
5199 Dump after hard register copy propagation.
5201 @itemx -fdump-rtl-csa
5202 @opindex fdump-rtl-csa
5203 Dump after combining stack adjustments.
5205 @item -fdump-rtl-cse1
5206 @itemx -fdump-rtl-cse2
5207 @opindex fdump-rtl-cse1
5208 @opindex fdump-rtl-cse2
5209 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5210 the two common sub-expression elimination passes.
5212 @itemx -fdump-rtl-dce
5213 @opindex fdump-rtl-dce
5214 Dump after the standalone dead code elimination passes.
5216 @itemx -fdump-rtl-dbr
5217 @opindex fdump-rtl-dbr
5218 Dump after delayed branch scheduling.
5220 @item -fdump-rtl-dce1
5221 @itemx -fdump-rtl-dce2
5222 @opindex fdump-rtl-dce1
5223 @opindex fdump-rtl-dce2
5224 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5225 the two dead store elimination passes.
5228 @opindex fdump-rtl-eh
5229 Dump after finalization of EH handling code.
5231 @item -fdump-rtl-eh_ranges
5232 @opindex fdump-rtl-eh_ranges
5233 Dump after conversion of EH handling range regions.
5235 @item -fdump-rtl-expand
5236 @opindex fdump-rtl-expand
5237 Dump after RTL generation.
5239 @item -fdump-rtl-fwprop1
5240 @itemx -fdump-rtl-fwprop2
5241 @opindex fdump-rtl-fwprop1
5242 @opindex fdump-rtl-fwprop2
5243 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5244 dumping after the two forward propagation passes.
5246 @item -fdump-rtl-gcse1
5247 @itemx -fdump-rtl-gcse2
5248 @opindex fdump-rtl-gcse1
5249 @opindex fdump-rtl-gcse2
5250 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5251 after global common subexpression elimination.
5253 @item -fdump-rtl-init-regs
5254 @opindex fdump-rtl-init-regs
5255 Dump after the initialization of the registers.
5257 @item -fdump-rtl-initvals
5258 @opindex fdump-rtl-initvals
5259 Dump after the computation of the initial value sets.
5261 @itemx -fdump-rtl-into_cfglayout
5262 @opindex fdump-rtl-into_cfglayout
5263 Dump after converting to cfglayout mode.
5265 @item -fdump-rtl-ira
5266 @opindex fdump-rtl-ira
5267 Dump after iterated register allocation.
5269 @item -fdump-rtl-jump
5270 @opindex fdump-rtl-jump
5271 Dump after the second jump optimization.
5273 @item -fdump-rtl-loop2
5274 @opindex fdump-rtl-loop2
5275 @option{-fdump-rtl-loop2} enables dumping after the rtl
5276 loop optimization passes.
5278 @item -fdump-rtl-mach
5279 @opindex fdump-rtl-mach
5280 Dump after performing the machine dependent reorganization pass, if that
5283 @item -fdump-rtl-mode_sw
5284 @opindex fdump-rtl-mode_sw
5285 Dump after removing redundant mode switches.
5287 @item -fdump-rtl-rnreg
5288 @opindex fdump-rtl-rnreg
5289 Dump after register renumbering.
5291 @itemx -fdump-rtl-outof_cfglayout
5292 @opindex fdump-rtl-outof_cfglayout
5293 Dump after converting from cfglayout mode.
5295 @item -fdump-rtl-peephole2
5296 @opindex fdump-rtl-peephole2
5297 Dump after the peephole pass.
5299 @item -fdump-rtl-postreload
5300 @opindex fdump-rtl-postreload
5301 Dump after post-reload optimizations.
5303 @itemx -fdump-rtl-pro_and_epilogue
5304 @opindex fdump-rtl-pro_and_epilogue
5305 Dump after generating the function pro and epilogues.
5307 @item -fdump-rtl-regmove
5308 @opindex fdump-rtl-regmove
5309 Dump after the register move pass.
5311 @item -fdump-rtl-sched1
5312 @itemx -fdump-rtl-sched2
5313 @opindex fdump-rtl-sched1
5314 @opindex fdump-rtl-sched2
5315 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5316 after the basic block scheduling passes.
5318 @item -fdump-rtl-see
5319 @opindex fdump-rtl-see
5320 Dump after sign extension elimination.
5322 @item -fdump-rtl-seqabstr
5323 @opindex fdump-rtl-seqabstr
5324 Dump after common sequence discovery.
5326 @item -fdump-rtl-shorten
5327 @opindex fdump-rtl-shorten
5328 Dump after shortening branches.
5330 @item -fdump-rtl-sibling
5331 @opindex fdump-rtl-sibling
5332 Dump after sibling call optimizations.
5334 @item -fdump-rtl-split1
5335 @itemx -fdump-rtl-split2
5336 @itemx -fdump-rtl-split3
5337 @itemx -fdump-rtl-split4
5338 @itemx -fdump-rtl-split5
5339 @opindex fdump-rtl-split1
5340 @opindex fdump-rtl-split2
5341 @opindex fdump-rtl-split3
5342 @opindex fdump-rtl-split4
5343 @opindex fdump-rtl-split5
5344 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5345 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5346 @option{-fdump-rtl-split5} enable dumping after five rounds of
5347 instruction splitting.
5349 @item -fdump-rtl-sms
5350 @opindex fdump-rtl-sms
5351 Dump after modulo scheduling. This pass is only run on some
5354 @item -fdump-rtl-stack
5355 @opindex fdump-rtl-stack
5356 Dump after conversion from GCC's "flat register file" registers to the
5357 x87's stack-like registers. This pass is only run on x86 variants.
5359 @item -fdump-rtl-subreg1
5360 @itemx -fdump-rtl-subreg2
5361 @opindex fdump-rtl-subreg1
5362 @opindex fdump-rtl-subreg2
5363 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5364 the two subreg expansion passes.
5366 @item -fdump-rtl-unshare
5367 @opindex fdump-rtl-unshare
5368 Dump after all rtl has been unshared.
5370 @item -fdump-rtl-vartrack
5371 @opindex fdump-rtl-vartrack
5372 Dump after variable tracking.
5374 @item -fdump-rtl-vregs
5375 @opindex fdump-rtl-vregs
5376 Dump after converting virtual registers to hard registers.
5378 @item -fdump-rtl-web
5379 @opindex fdump-rtl-web
5380 Dump after live range splitting.
5382 @item -fdump-rtl-regclass
5383 @itemx -fdump-rtl-subregs_of_mode_init
5384 @itemx -fdump-rtl-subregs_of_mode_finish
5385 @itemx -fdump-rtl-dfinit
5386 @itemx -fdump-rtl-dfinish
5387 @opindex fdump-rtl-regclass
5388 @opindex fdump-rtl-subregs_of_mode_init
5389 @opindex fdump-rtl-subregs_of_mode_finish
5390 @opindex fdump-rtl-dfinit
5391 @opindex fdump-rtl-dfinish
5392 These dumps are defined but always produce empty files.
5394 @item -fdump-rtl-all
5395 @opindex fdump-rtl-all
5396 Produce all the dumps listed above.
5400 Annotate the assembler output with miscellaneous debugging information.
5404 Dump all macro definitions, at the end of preprocessing, in addition to
5409 Produce a core dump whenever an error occurs.
5413 Print statistics on memory usage, at the end of the run, to
5418 Annotate the assembler output with a comment indicating which
5419 pattern and alternative was used. The length of each instruction is
5424 Dump the RTL in the assembler output as a comment before each instruction.
5425 Also turns on @option{-dp} annotation.
5429 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5430 dump a representation of the control flow graph suitable for viewing with VCG
5431 to @file{@var{file}.@var{pass}.vcg}.
5435 Just generate RTL for a function instead of compiling it. Usually used
5436 with @option{-fdump-rtl-expand}.
5440 @opindex fdump-noaddr
5441 When doing debugging dumps, suppress address output. This makes it more
5442 feasible to use diff on debugging dumps for compiler invocations with
5443 different compiler binaries and/or different
5444 text / bss / data / heap / stack / dso start locations.
5446 @item -fdump-unnumbered
5447 @opindex fdump-unnumbered
5448 When doing debugging dumps, suppress instruction numbers and address output.
5449 This makes it more feasible to use diff on debugging dumps for compiler
5450 invocations with different options, in particular with and without
5453 @item -fdump-unnumbered-links
5454 @opindex fdump-unnumbered-links
5455 When doing debugging dumps (see @option{-d} option above), suppress
5456 instruction numbers for the links to the previous and next instructions
5459 @item -fdump-translation-unit @r{(C++ only)}
5460 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5461 @opindex fdump-translation-unit
5462 Dump a representation of the tree structure for the entire translation
5463 unit to a file. The file name is made by appending @file{.tu} to the
5464 source file name, and the file is created in the same directory as the
5465 output file. If the @samp{-@var{options}} form is used, @var{options}
5466 controls the details of the dump as described for the
5467 @option{-fdump-tree} options.
5469 @item -fdump-class-hierarchy @r{(C++ only)}
5470 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5471 @opindex fdump-class-hierarchy
5472 Dump a representation of each class's hierarchy and virtual function
5473 table layout to a file. The file name is made by appending
5474 @file{.class} to the source file name, and the file is created in the
5475 same directory as the output file. If the @samp{-@var{options}} form
5476 is used, @var{options} controls the details of the dump as described
5477 for the @option{-fdump-tree} options.
5479 @item -fdump-ipa-@var{switch}
5481 Control the dumping at various stages of inter-procedural analysis
5482 language tree to a file. The file name is generated by appending a
5483 switch specific suffix to the source file name, and the file is created
5484 in the same directory as the output file. The following dumps are
5489 Enables all inter-procedural analysis dumps.
5492 Dumps information about call-graph optimization, unused function removal,
5493 and inlining decisions.
5496 Dump after function inlining.
5501 @opindex fdump-passes
5502 Dump the list of optimization passes that are turned on and off by
5503 the current command line options.
5505 @item -fdump-statistics-@var{option}
5506 @opindex fdump-statistics
5507 Enable and control dumping of pass statistics in a separate file. The
5508 file name is generated by appending a suffix ending in
5509 @samp{.statistics} to the source file name, and the file is created in
5510 the same directory as the output file. If the @samp{-@var{option}}
5511 form is used, @samp{-stats} will cause counters to be summed over the
5512 whole compilation unit while @samp{-details} will dump every event as
5513 the passes generate them. The default with no option is to sum
5514 counters for each function compiled.
5516 @item -fdump-tree-@var{switch}
5517 @itemx -fdump-tree-@var{switch}-@var{options}
5519 Control the dumping at various stages of processing the intermediate
5520 language tree to a file. The file name is generated by appending a
5521 switch specific suffix to the source file name, and the file is
5522 created in the same directory as the output file. If the
5523 @samp{-@var{options}} form is used, @var{options} is a list of
5524 @samp{-} separated options that control the details of the dump. Not
5525 all options are applicable to all dumps, those which are not
5526 meaningful will be ignored. The following options are available
5530 Print the address of each node. Usually this is not meaningful as it
5531 changes according to the environment and source file. Its primary use
5532 is for tying up a dump file with a debug environment.
5534 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5535 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5536 use working backward from mangled names in the assembly file.
5538 Inhibit dumping of members of a scope or body of a function merely
5539 because that scope has been reached. Only dump such items when they
5540 are directly reachable by some other path. When dumping pretty-printed
5541 trees, this option inhibits dumping the bodies of control structures.
5543 Print a raw representation of the tree. By default, trees are
5544 pretty-printed into a C-like representation.
5546 Enable more detailed dumps (not honored by every dump option).
5548 Enable dumping various statistics about the pass (not honored by every dump
5551 Enable showing basic block boundaries (disabled in raw dumps).
5553 Enable showing virtual operands for every statement.
5555 Enable showing line numbers for statements.
5557 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5559 Enable showing the tree dump for each statement.
5561 Enable showing the EH region number holding each statement.
5563 Enable showing scalar evolution analysis details.
5565 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5566 and @option{lineno}.
5569 The following tree dumps are possible:
5573 @opindex fdump-tree-original
5574 Dump before any tree based optimization, to @file{@var{file}.original}.
5577 @opindex fdump-tree-optimized
5578 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5581 @opindex fdump-tree-gimple
5582 Dump each function before and after the gimplification pass to a file. The
5583 file name is made by appending @file{.gimple} to the source file name.
5586 @opindex fdump-tree-cfg
5587 Dump the control flow graph of each function to a file. The file name is
5588 made by appending @file{.cfg} to the source file name.
5591 @opindex fdump-tree-vcg
5592 Dump the control flow graph of each function to a file in VCG format. The
5593 file name is made by appending @file{.vcg} to the source file name. Note
5594 that if the file contains more than one function, the generated file cannot
5595 be used directly by VCG@. You will need to cut and paste each function's
5596 graph into its own separate file first.
5599 @opindex fdump-tree-ch
5600 Dump each function after copying loop headers. The file name is made by
5601 appending @file{.ch} to the source file name.
5604 @opindex fdump-tree-ssa
5605 Dump SSA related information to a file. The file name is made by appending
5606 @file{.ssa} to the source file name.
5609 @opindex fdump-tree-alias
5610 Dump aliasing information for each function. The file name is made by
5611 appending @file{.alias} to the source file name.
5614 @opindex fdump-tree-ccp
5615 Dump each function after CCP@. The file name is made by appending
5616 @file{.ccp} to the source file name.
5619 @opindex fdump-tree-storeccp
5620 Dump each function after STORE-CCP@. The file name is made by appending
5621 @file{.storeccp} to the source file name.
5624 @opindex fdump-tree-pre
5625 Dump trees after partial redundancy elimination. The file name is made
5626 by appending @file{.pre} to the source file name.
5629 @opindex fdump-tree-fre
5630 Dump trees after full redundancy elimination. The file name is made
5631 by appending @file{.fre} to the source file name.
5634 @opindex fdump-tree-copyprop
5635 Dump trees after copy propagation. The file name is made
5636 by appending @file{.copyprop} to the source file name.
5638 @item store_copyprop
5639 @opindex fdump-tree-store_copyprop
5640 Dump trees after store copy-propagation. The file name is made
5641 by appending @file{.store_copyprop} to the source file name.
5644 @opindex fdump-tree-dce
5645 Dump each function after dead code elimination. The file name is made by
5646 appending @file{.dce} to the source file name.
5649 @opindex fdump-tree-mudflap
5650 Dump each function after adding mudflap instrumentation. The file name is
5651 made by appending @file{.mudflap} to the source file name.
5654 @opindex fdump-tree-sra
5655 Dump each function after performing scalar replacement of aggregates. The
5656 file name is made by appending @file{.sra} to the source file name.
5659 @opindex fdump-tree-sink
5660 Dump each function after performing code sinking. The file name is made
5661 by appending @file{.sink} to the source file name.
5664 @opindex fdump-tree-dom
5665 Dump each function after applying dominator tree optimizations. The file
5666 name is made by appending @file{.dom} to the source file name.
5669 @opindex fdump-tree-dse
5670 Dump each function after applying dead store elimination. The file
5671 name is made by appending @file{.dse} to the source file name.
5674 @opindex fdump-tree-phiopt
5675 Dump each function after optimizing PHI nodes into straightline code. The file
5676 name is made by appending @file{.phiopt} to the source file name.
5679 @opindex fdump-tree-forwprop
5680 Dump each function after forward propagating single use variables. The file
5681 name is made by appending @file{.forwprop} to the source file name.
5684 @opindex fdump-tree-copyrename
5685 Dump each function after applying the copy rename optimization. The file
5686 name is made by appending @file{.copyrename} to the source file name.
5689 @opindex fdump-tree-nrv
5690 Dump each function after applying the named return value optimization on
5691 generic trees. The file name is made by appending @file{.nrv} to the source
5695 @opindex fdump-tree-vect
5696 Dump each function after applying vectorization of loops. The file name is
5697 made by appending @file{.vect} to the source file name.
5700 @opindex fdump-tree-slp
5701 Dump each function after applying vectorization of basic blocks. The file name
5702 is made by appending @file{.slp} to the source file name.
5705 @opindex fdump-tree-vrp
5706 Dump each function after Value Range Propagation (VRP). The file name
5707 is made by appending @file{.vrp} to the source file name.
5710 @opindex fdump-tree-all
5711 Enable all the available tree dumps with the flags provided in this option.
5714 @item -ftree-vectorizer-verbose=@var{n}
5715 @opindex ftree-vectorizer-verbose
5716 This option controls the amount of debugging output the vectorizer prints.
5717 This information is written to standard error, unless
5718 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5719 in which case it is output to the usual dump listing file, @file{.vect}.
5720 For @var{n}=0 no diagnostic information is reported.
5721 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5722 and the total number of loops that got vectorized.
5723 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5724 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5725 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5726 level that @option{-fdump-tree-vect-stats} uses.
5727 Higher verbosity levels mean either more information dumped for each
5728 reported loop, or same amount of information reported for more loops:
5729 if @var{n}=3, vectorizer cost model information is reported.
5730 If @var{n}=4, alignment related information is added to the reports.
5731 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5732 memory access-patterns) is added to the reports.
5733 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5734 that did not pass the first analysis phase (i.e., may not be countable, or
5735 may have complicated control-flow).
5736 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5737 If @var{n}=8, SLP related information is added to the reports.
5738 For @var{n}=9, all the information the vectorizer generates during its
5739 analysis and transformation is reported. This is the same verbosity level
5740 that @option{-fdump-tree-vect-details} uses.
5742 @item -frandom-seed=@var{string}
5743 @opindex frandom-seed
5744 This option provides a seed that GCC uses when it would otherwise use
5745 random numbers. It is used to generate certain symbol names
5746 that have to be different in every compiled file. It is also used to
5747 place unique stamps in coverage data files and the object files that
5748 produce them. You can use the @option{-frandom-seed} option to produce
5749 reproducibly identical object files.
5751 The @var{string} should be different for every file you compile.
5753 @item -fsched-verbose=@var{n}
5754 @opindex fsched-verbose
5755 On targets that use instruction scheduling, this option controls the
5756 amount of debugging output the scheduler prints. This information is
5757 written to standard error, unless @option{-fdump-rtl-sched1} or
5758 @option{-fdump-rtl-sched2} is specified, in which case it is output
5759 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5760 respectively. However for @var{n} greater than nine, the output is
5761 always printed to standard error.
5763 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5764 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5765 For @var{n} greater than one, it also output basic block probabilities,
5766 detailed ready list information and unit/insn info. For @var{n} greater
5767 than two, it includes RTL at abort point, control-flow and regions info.
5768 And for @var{n} over four, @option{-fsched-verbose} also includes
5772 @itemx -save-temps=cwd
5774 Store the usual ``temporary'' intermediate files permanently; place them
5775 in the current directory and name them based on the source file. Thus,
5776 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5777 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5778 preprocessed @file{foo.i} output file even though the compiler now
5779 normally uses an integrated preprocessor.
5781 When used in combination with the @option{-x} command line option,
5782 @option{-save-temps} is sensible enough to avoid over writing an
5783 input source file with the same extension as an intermediate file.
5784 The corresponding intermediate file may be obtained by renaming the
5785 source file before using @option{-save-temps}.
5787 If you invoke GCC in parallel, compiling several different source
5788 files that share a common base name in different subdirectories or the
5789 same source file compiled for multiple output destinations, it is
5790 likely that the different parallel compilers will interfere with each
5791 other, and overwrite the temporary files. For instance:
5794 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5795 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5798 may result in @file{foo.i} and @file{foo.o} being written to
5799 simultaneously by both compilers.
5801 @item -save-temps=obj
5802 @opindex save-temps=obj
5803 Store the usual ``temporary'' intermediate files permanently. If the
5804 @option{-o} option is used, the temporary files are based on the
5805 object file. If the @option{-o} option is not used, the
5806 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5811 gcc -save-temps=obj -c foo.c
5812 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5813 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5816 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5817 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5818 @file{dir2/yfoobar.o}.
5820 @item -time@r{[}=@var{file}@r{]}
5822 Report the CPU time taken by each subprocess in the compilation
5823 sequence. For C source files, this is the compiler proper and assembler
5824 (plus the linker if linking is done).
5826 Without the specification of an output file, the output looks like this:
5833 The first number on each line is the ``user time'', that is time spent
5834 executing the program itself. The second number is ``system time'',
5835 time spent executing operating system routines on behalf of the program.
5836 Both numbers are in seconds.
5838 With the specification of an output file, the output is appended to the
5839 named file, and it looks like this:
5842 0.12 0.01 cc1 @var{options}
5843 0.00 0.01 as @var{options}
5846 The ``user time'' and the ``system time'' are moved before the program
5847 name, and the options passed to the program are displayed, so that one
5848 can later tell what file was being compiled, and with which options.
5850 @item -fvar-tracking
5851 @opindex fvar-tracking
5852 Run variable tracking pass. It computes where variables are stored at each
5853 position in code. Better debugging information is then generated
5854 (if the debugging information format supports this information).
5856 It is enabled by default when compiling with optimization (@option{-Os},
5857 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5858 the debug info format supports it.
5860 @item -fvar-tracking-assignments
5861 @opindex fvar-tracking-assignments
5862 @opindex fno-var-tracking-assignments
5863 Annotate assignments to user variables early in the compilation and
5864 attempt to carry the annotations over throughout the compilation all the
5865 way to the end, in an attempt to improve debug information while
5866 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5868 It can be enabled even if var-tracking is disabled, in which case
5869 annotations will be created and maintained, but discarded at the end.
5871 @item -fvar-tracking-assignments-toggle
5872 @opindex fvar-tracking-assignments-toggle
5873 @opindex fno-var-tracking-assignments-toggle
5874 Toggle @option{-fvar-tracking-assignments}, in the same way that
5875 @option{-gtoggle} toggles @option{-g}.
5877 @item -print-file-name=@var{library}
5878 @opindex print-file-name
5879 Print the full absolute name of the library file @var{library} that
5880 would be used when linking---and don't do anything else. With this
5881 option, GCC does not compile or link anything; it just prints the
5884 @item -print-multi-directory
5885 @opindex print-multi-directory
5886 Print the directory name corresponding to the multilib selected by any
5887 other switches present in the command line. This directory is supposed
5888 to exist in @env{GCC_EXEC_PREFIX}.
5890 @item -print-multi-lib
5891 @opindex print-multi-lib
5892 Print the mapping from multilib directory names to compiler switches
5893 that enable them. The directory name is separated from the switches by
5894 @samp{;}, and each switch starts with an @samp{@@} instead of the
5895 @samp{-}, without spaces between multiple switches. This is supposed to
5896 ease shell-processing.
5898 @item -print-multi-os-directory
5899 @opindex print-multi-os-directory
5900 Print the path to OS libraries for the selected
5901 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5902 present in the @file{lib} subdirectory and no multilibs are used, this is
5903 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5904 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5905 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5906 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5908 @item -print-prog-name=@var{program}
5909 @opindex print-prog-name
5910 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5912 @item -print-libgcc-file-name
5913 @opindex print-libgcc-file-name
5914 Same as @option{-print-file-name=libgcc.a}.
5916 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5917 but you do want to link with @file{libgcc.a}. You can do
5920 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5923 @item -print-search-dirs
5924 @opindex print-search-dirs
5925 Print the name of the configured installation directory and a list of
5926 program and library directories @command{gcc} will search---and don't do anything else.
5928 This is useful when @command{gcc} prints the error message
5929 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5930 To resolve this you either need to put @file{cpp0} and the other compiler
5931 components where @command{gcc} expects to find them, or you can set the environment
5932 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5933 Don't forget the trailing @samp{/}.
5934 @xref{Environment Variables}.
5936 @item -print-sysroot
5937 @opindex print-sysroot
5938 Print the target sysroot directory that will be used during
5939 compilation. This is the target sysroot specified either at configure
5940 time or using the @option{--sysroot} option, possibly with an extra
5941 suffix that depends on compilation options. If no target sysroot is
5942 specified, the option prints nothing.
5944 @item -print-sysroot-headers-suffix
5945 @opindex print-sysroot-headers-suffix
5946 Print the suffix added to the target sysroot when searching for
5947 headers, or give an error if the compiler is not configured with such
5948 a suffix---and don't do anything else.
5951 @opindex dumpmachine
5952 Print the compiler's target machine (for example,
5953 @samp{i686-pc-linux-gnu})---and don't do anything else.
5956 @opindex dumpversion
5957 Print the compiler version (for example, @samp{3.0})---and don't do
5962 Print the compiler's built-in specs---and don't do anything else. (This
5963 is used when GCC itself is being built.) @xref{Spec Files}.
5965 @item -feliminate-unused-debug-types
5966 @opindex feliminate-unused-debug-types
5967 Normally, when producing DWARF2 output, GCC will emit debugging
5968 information for all types declared in a compilation
5969 unit, regardless of whether or not they are actually used
5970 in that compilation unit. Sometimes this is useful, such as
5971 if, in the debugger, you want to cast a value to a type that is
5972 not actually used in your program (but is declared). More often,
5973 however, this results in a significant amount of wasted space.
5974 With this option, GCC will avoid producing debug symbol output
5975 for types that are nowhere used in the source file being compiled.
5978 @node Optimize Options
5979 @section Options That Control Optimization
5980 @cindex optimize options
5981 @cindex options, optimization
5983 These options control various sorts of optimizations.
5985 Without any optimization option, the compiler's goal is to reduce the
5986 cost of compilation and to make debugging produce the expected
5987 results. Statements are independent: if you stop the program with a
5988 breakpoint between statements, you can then assign a new value to any
5989 variable or change the program counter to any other statement in the
5990 function and get exactly the results you would expect from the source
5993 Turning on optimization flags makes the compiler attempt to improve
5994 the performance and/or code size at the expense of compilation time
5995 and possibly the ability to debug the program.
5997 The compiler performs optimization based on the knowledge it has of the
5998 program. Compiling multiple files at once to a single output file mode allows
5999 the compiler to use information gained from all of the files when compiling
6002 Not all optimizations are controlled directly by a flag. Only
6003 optimizations that have a flag are listed in this section.
6005 Most optimizations are only enabled if an @option{-O} level is set on
6006 the command line. Otherwise they are disabled, even if individual
6007 optimization flags are specified.
6009 Depending on the target and how GCC was configured, a slightly different
6010 set of optimizations may be enabled at each @option{-O} level than
6011 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6012 to find out the exact set of optimizations that are enabled at each level.
6013 @xref{Overall Options}, for examples.
6020 Optimize. Optimizing compilation takes somewhat more time, and a lot
6021 more memory for a large function.
6023 With @option{-O}, the compiler tries to reduce code size and execution
6024 time, without performing any optimizations that take a great deal of
6027 @option{-O} turns on the following optimization flags:
6031 -fcprop-registers @gol
6034 -fdelayed-branch @gol
6036 -fguess-branch-probability @gol
6037 -fif-conversion2 @gol
6038 -fif-conversion @gol
6039 -fipa-pure-const @gol
6041 -fipa-reference @gol
6043 -fsplit-wide-types @gol
6045 -ftree-builtin-call-dce @gol
6048 -ftree-copyrename @gol
6050 -ftree-dominator-opts @gol
6052 -ftree-forwprop @gol
6060 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6061 where doing so does not interfere with debugging.
6065 Optimize even more. GCC performs nearly all supported optimizations
6066 that do not involve a space-speed tradeoff.
6067 As compared to @option{-O}, this option increases both compilation time
6068 and the performance of the generated code.
6070 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6071 also turns on the following optimization flags:
6072 @gccoptlist{-fthread-jumps @gol
6073 -falign-functions -falign-jumps @gol
6074 -falign-loops -falign-labels @gol
6077 -fcse-follow-jumps -fcse-skip-blocks @gol
6078 -fdelete-null-pointer-checks @gol
6080 -fexpensive-optimizations @gol
6081 -fgcse -fgcse-lm @gol
6082 -finline-small-functions @gol
6083 -findirect-inlining @gol
6085 -foptimize-sibling-calls @gol
6086 -fpartial-inlining @gol
6089 -freorder-blocks -freorder-functions @gol
6090 -frerun-cse-after-loop @gol
6091 -fsched-interblock -fsched-spec @gol
6092 -fschedule-insns -fschedule-insns2 @gol
6093 -fstrict-aliasing -fstrict-overflow @gol
6094 -ftree-switch-conversion @gol
6098 Please note the warning under @option{-fgcse} about
6099 invoking @option{-O2} on programs that use computed gotos.
6103 Optimize yet more. @option{-O3} turns on all optimizations specified
6104 by @option{-O2} and also turns on the @option{-finline-functions},
6105 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6106 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6107 @option{-fipa-cp-clone} options.
6111 Reduce compilation time and make debugging produce the expected
6112 results. This is the default.
6116 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6117 do not typically increase code size. It also performs further
6118 optimizations designed to reduce code size.
6120 @option{-Os} disables the following optimization flags:
6121 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6122 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6123 -fprefetch-loop-arrays -ftree-vect-loop-version}
6127 Disregard strict standards compliance. @option{-Ofast} enables all
6128 @option{-O3} optimizations. It also enables optimizations that are not
6129 valid for all standard compliant programs.
6130 It turns on @option{-ffast-math} and the Fortran-specific
6131 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6133 If you use multiple @option{-O} options, with or without level numbers,
6134 the last such option is the one that is effective.
6137 Options of the form @option{-f@var{flag}} specify machine-independent
6138 flags. Most flags have both positive and negative forms; the negative
6139 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6140 below, only one of the forms is listed---the one you typically will
6141 use. You can figure out the other form by either removing @samp{no-}
6144 The following options control specific optimizations. They are either
6145 activated by @option{-O} options or are related to ones that are. You
6146 can use the following flags in the rare cases when ``fine-tuning'' of
6147 optimizations to be performed is desired.
6150 @item -fno-default-inline
6151 @opindex fno-default-inline
6152 Do not make member functions inline by default merely because they are
6153 defined inside the class scope (C++ only). Otherwise, when you specify
6154 @w{@option{-O}}, member functions defined inside class scope are compiled
6155 inline by default; i.e., you don't need to add @samp{inline} in front of
6156 the member function name.
6158 @item -fno-defer-pop
6159 @opindex fno-defer-pop
6160 Always pop the arguments to each function call as soon as that function
6161 returns. For machines which must pop arguments after a function call,
6162 the compiler normally lets arguments accumulate on the stack for several
6163 function calls and pops them all at once.
6165 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6167 @item -fforward-propagate
6168 @opindex fforward-propagate
6169 Perform a forward propagation pass on RTL@. The pass tries to combine two
6170 instructions and checks if the result can be simplified. If loop unrolling
6171 is active, two passes are performed and the second is scheduled after
6174 This option is enabled by default at optimization levels @option{-O},
6175 @option{-O2}, @option{-O3}, @option{-Os}.
6177 @item -ffp-contract=@var{style}
6178 @opindex ffp-contract
6179 @option{-ffp-contract=off} disables floating-point expression contraction.
6180 @option{-ffp-contract=fast} enables floating-point expression contraction
6181 such as forming of fused multiply-add operations if the target has
6182 native support for them.
6183 @option{-ffp-contract=on} enables floating-point expression contraction
6184 if allowed by the language standard. This is currently not implemented
6185 and treated equal to @option{-ffp-contract=off}.
6187 The default is @option{-ffp-contract=fast}.
6189 @item -fomit-frame-pointer
6190 @opindex fomit-frame-pointer
6191 Don't keep the frame pointer in a register for functions that
6192 don't need one. This avoids the instructions to save, set up and
6193 restore frame pointers; it also makes an extra register available
6194 in many functions. @strong{It also makes debugging impossible on
6197 On some machines, such as the VAX, this flag has no effect, because
6198 the standard calling sequence automatically handles the frame pointer
6199 and nothing is saved by pretending it doesn't exist. The
6200 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6201 whether a target machine supports this flag. @xref{Registers,,Register
6202 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6204 Starting with GCC version 4.6, the default setting (when not optimizing for
6205 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6206 @option{-fomit-frame-pointer}. The default can be reverted to
6207 @option{-fno-omit-frame-pointer} by configuring GCC with the
6208 @option{--enable-frame-pointer} configure option.
6210 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6212 @item -foptimize-sibling-calls
6213 @opindex foptimize-sibling-calls
6214 Optimize sibling and tail recursive calls.
6216 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6220 Don't pay attention to the @code{inline} keyword. Normally this option
6221 is used to keep the compiler from expanding any functions inline.
6222 Note that if you are not optimizing, no functions can be expanded inline.
6224 @item -finline-small-functions
6225 @opindex finline-small-functions
6226 Integrate functions into their callers when their body is smaller than expected
6227 function call code (so overall size of program gets smaller). The compiler
6228 heuristically decides which functions are simple enough to be worth integrating
6231 Enabled at level @option{-O2}.
6233 @item -findirect-inlining
6234 @opindex findirect-inlining
6235 Inline also indirect calls that are discovered to be known at compile
6236 time thanks to previous inlining. This option has any effect only
6237 when inlining itself is turned on by the @option{-finline-functions}
6238 or @option{-finline-small-functions} options.
6240 Enabled at level @option{-O2}.
6242 @item -finline-functions
6243 @opindex finline-functions
6244 Integrate all simple functions into their callers. The compiler
6245 heuristically decides which functions are simple enough to be worth
6246 integrating in this way.
6248 If all calls to a given function are integrated, and the function is
6249 declared @code{static}, then the function is normally not output as
6250 assembler code in its own right.
6252 Enabled at level @option{-O3}.
6254 @item -finline-functions-called-once
6255 @opindex finline-functions-called-once
6256 Consider all @code{static} functions called once for inlining into their
6257 caller even if they are not marked @code{inline}. If a call to a given
6258 function is integrated, then the function is not output as assembler code
6261 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6263 @item -fearly-inlining
6264 @opindex fearly-inlining
6265 Inline functions marked by @code{always_inline} and functions whose body seems
6266 smaller than the function call overhead early before doing
6267 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6268 makes profiling significantly cheaper and usually inlining faster on programs
6269 having large chains of nested wrapper functions.
6275 Perform interprocedural scalar replacement of aggregates, removal of
6276 unused parameters and replacement of parameters passed by reference
6277 by parameters passed by value.
6279 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6281 @item -finline-limit=@var{n}
6282 @opindex finline-limit
6283 By default, GCC limits the size of functions that can be inlined. This flag
6284 allows coarse control of this limit. @var{n} is the size of functions that
6285 can be inlined in number of pseudo instructions.
6287 Inlining is actually controlled by a number of parameters, which may be
6288 specified individually by using @option{--param @var{name}=@var{value}}.
6289 The @option{-finline-limit=@var{n}} option sets some of these parameters
6293 @item max-inline-insns-single
6294 is set to @var{n}/2.
6295 @item max-inline-insns-auto
6296 is set to @var{n}/2.
6299 See below for a documentation of the individual
6300 parameters controlling inlining and for the defaults of these parameters.
6302 @emph{Note:} there may be no value to @option{-finline-limit} that results
6303 in default behavior.
6305 @emph{Note:} pseudo instruction represents, in this particular context, an
6306 abstract measurement of function's size. In no way does it represent a count
6307 of assembly instructions and as such its exact meaning might change from one
6308 release to an another.
6310 @item -fno-keep-inline-dllexport
6311 @opindex -fno-keep-inline-dllexport
6312 This is a more fine-grained version of @option{-fkeep-inline-functions},
6313 which applies only to functions that are declared using the @code{dllexport}
6314 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6317 @item -fkeep-inline-functions
6318 @opindex fkeep-inline-functions
6319 In C, emit @code{static} functions that are declared @code{inline}
6320 into the object file, even if the function has been inlined into all
6321 of its callers. This switch does not affect functions using the
6322 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6323 inline functions into the object file.
6325 @item -fkeep-static-consts
6326 @opindex fkeep-static-consts
6327 Emit variables declared @code{static const} when optimization isn't turned
6328 on, even if the variables aren't referenced.
6330 GCC enables this option by default. If you want to force the compiler to
6331 check if the variable was referenced, regardless of whether or not
6332 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6334 @item -fmerge-constants
6335 @opindex fmerge-constants
6336 Attempt to merge identical constants (string constants and floating point
6337 constants) across compilation units.
6339 This option is the default for optimized compilation if the assembler and
6340 linker support it. Use @option{-fno-merge-constants} to inhibit this
6343 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6345 @item -fmerge-all-constants
6346 @opindex fmerge-all-constants
6347 Attempt to merge identical constants and identical variables.
6349 This option implies @option{-fmerge-constants}. In addition to
6350 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6351 arrays or initialized constant variables with integral or floating point
6352 types. Languages like C or C++ require each variable, including multiple
6353 instances of the same variable in recursive calls, to have distinct locations,
6354 so using this option will result in non-conforming
6357 @item -fmodulo-sched
6358 @opindex fmodulo-sched
6359 Perform swing modulo scheduling immediately before the first scheduling
6360 pass. This pass looks at innermost loops and reorders their
6361 instructions by overlapping different iterations.
6363 @item -fmodulo-sched-allow-regmoves
6364 @opindex fmodulo-sched-allow-regmoves
6365 Perform more aggressive SMS based modulo scheduling with register moves
6366 allowed. By setting this flag certain anti-dependences edges will be
6367 deleted which will trigger the generation of reg-moves based on the
6368 life-range analysis. This option is effective only with
6369 @option{-fmodulo-sched} enabled.
6371 @item -fno-branch-count-reg
6372 @opindex fno-branch-count-reg
6373 Do not use ``decrement and branch'' instructions on a count register,
6374 but instead generate a sequence of instructions that decrement a
6375 register, compare it against zero, then branch based upon the result.
6376 This option is only meaningful on architectures that support such
6377 instructions, which include x86, PowerPC, IA-64 and S/390.
6379 The default is @option{-fbranch-count-reg}.
6381 @item -fno-function-cse
6382 @opindex fno-function-cse
6383 Do not put function addresses in registers; make each instruction that
6384 calls a constant function contain the function's address explicitly.
6386 This option results in less efficient code, but some strange hacks
6387 that alter the assembler output may be confused by the optimizations
6388 performed when this option is not used.
6390 The default is @option{-ffunction-cse}
6392 @item -fno-zero-initialized-in-bss
6393 @opindex fno-zero-initialized-in-bss
6394 If the target supports a BSS section, GCC by default puts variables that
6395 are initialized to zero into BSS@. This can save space in the resulting
6398 This option turns off this behavior because some programs explicitly
6399 rely on variables going to the data section. E.g., so that the
6400 resulting executable can find the beginning of that section and/or make
6401 assumptions based on that.
6403 The default is @option{-fzero-initialized-in-bss}.
6405 @item -fmudflap -fmudflapth -fmudflapir
6409 @cindex bounds checking
6411 For front-ends that support it (C and C++), instrument all risky
6412 pointer/array dereferencing operations, some standard library
6413 string/heap functions, and some other associated constructs with
6414 range/validity tests. Modules so instrumented should be immune to
6415 buffer overflows, invalid heap use, and some other classes of C/C++
6416 programming errors. The instrumentation relies on a separate runtime
6417 library (@file{libmudflap}), which will be linked into a program if
6418 @option{-fmudflap} is given at link time. Run-time behavior of the
6419 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6420 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6423 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6424 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6425 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6426 instrumentation should ignore pointer reads. This produces less
6427 instrumentation (and therefore faster execution) and still provides
6428 some protection against outright memory corrupting writes, but allows
6429 erroneously read data to propagate within a program.
6431 @item -fthread-jumps
6432 @opindex fthread-jumps
6433 Perform optimizations where we check to see if a jump branches to a
6434 location where another comparison subsumed by the first is found. If
6435 so, the first branch is redirected to either the destination of the
6436 second branch or a point immediately following it, depending on whether
6437 the condition is known to be true or false.
6439 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6441 @item -fsplit-wide-types
6442 @opindex fsplit-wide-types
6443 When using a type that occupies multiple registers, such as @code{long
6444 long} on a 32-bit system, split the registers apart and allocate them
6445 independently. This normally generates better code for those types,
6446 but may make debugging more difficult.
6448 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6451 @item -fcse-follow-jumps
6452 @opindex fcse-follow-jumps
6453 In common subexpression elimination (CSE), scan through jump instructions
6454 when the target of the jump is not reached by any other path. For
6455 example, when CSE encounters an @code{if} statement with an
6456 @code{else} clause, CSE will follow the jump when the condition
6459 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6461 @item -fcse-skip-blocks
6462 @opindex fcse-skip-blocks
6463 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6464 follow jumps which conditionally skip over blocks. When CSE
6465 encounters a simple @code{if} statement with no else clause,
6466 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6467 body of the @code{if}.
6469 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6471 @item -frerun-cse-after-loop
6472 @opindex frerun-cse-after-loop
6473 Re-run common subexpression elimination after loop optimizations has been
6476 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6480 Perform a global common subexpression elimination pass.
6481 This pass also performs global constant and copy propagation.
6483 @emph{Note:} When compiling a program using computed gotos, a GCC
6484 extension, you may get better runtime performance if you disable
6485 the global common subexpression elimination pass by adding
6486 @option{-fno-gcse} to the command line.
6488 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6492 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6493 attempt to move loads which are only killed by stores into themselves. This
6494 allows a loop containing a load/store sequence to be changed to a load outside
6495 the loop, and a copy/store within the loop.
6497 Enabled by default when gcse is enabled.
6501 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6502 global common subexpression elimination. This pass will attempt to move
6503 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6504 loops containing a load/store sequence can be changed to a load before
6505 the loop and a store after the loop.
6507 Not enabled at any optimization level.
6511 When @option{-fgcse-las} is enabled, the global common subexpression
6512 elimination pass eliminates redundant loads that come after stores to the
6513 same memory location (both partial and full redundancies).
6515 Not enabled at any optimization level.
6517 @item -fgcse-after-reload
6518 @opindex fgcse-after-reload
6519 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6520 pass is performed after reload. The purpose of this pass is to cleanup
6523 @item -funsafe-loop-optimizations
6524 @opindex funsafe-loop-optimizations
6525 If given, the loop optimizer will assume that loop indices do not
6526 overflow, and that the loops with nontrivial exit condition are not
6527 infinite. This enables a wider range of loop optimizations even if
6528 the loop optimizer itself cannot prove that these assumptions are valid.
6529 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6530 if it finds this kind of loop.
6532 @item -fcrossjumping
6533 @opindex fcrossjumping
6534 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6535 resulting code may or may not perform better than without cross-jumping.
6537 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6539 @item -fauto-inc-dec
6540 @opindex fauto-inc-dec
6541 Combine increments or decrements of addresses with memory accesses.
6542 This pass is always skipped on architectures that do not have
6543 instructions to support this. Enabled by default at @option{-O} and
6544 higher on architectures that support this.
6548 Perform dead code elimination (DCE) on RTL@.
6549 Enabled by default at @option{-O} and higher.
6553 Perform dead store elimination (DSE) on RTL@.
6554 Enabled by default at @option{-O} and higher.
6556 @item -fif-conversion
6557 @opindex fif-conversion
6558 Attempt to transform conditional jumps into branch-less equivalents. This
6559 include use of conditional moves, min, max, set flags and abs instructions, and
6560 some tricks doable by standard arithmetics. The use of conditional execution
6561 on chips where it is available is controlled by @code{if-conversion2}.
6563 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6565 @item -fif-conversion2
6566 @opindex fif-conversion2
6567 Use conditional execution (where available) to transform conditional jumps into
6568 branch-less equivalents.
6570 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6572 @item -fdelete-null-pointer-checks
6573 @opindex fdelete-null-pointer-checks
6574 Assume that programs cannot safely dereference null pointers, and that
6575 no code or data element resides there. This enables simple constant
6576 folding optimizations at all optimization levels. In addition, other
6577 optimization passes in GCC use this flag to control global dataflow
6578 analyses that eliminate useless checks for null pointers; these assume
6579 that if a pointer is checked after it has already been dereferenced,
6582 Note however that in some environments this assumption is not true.
6583 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6584 for programs which depend on that behavior.
6586 Some targets, especially embedded ones, disable this option at all levels.
6587 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6588 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6589 are enabled independently at different optimization levels.
6591 @item -fdevirtualize
6592 @opindex fdevirtualize
6593 Attempt to convert calls to virtual functions to direct calls. This
6594 is done both within a procedure and interprocedurally as part of
6595 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6596 propagation (@option{-fipa-cp}).
6597 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6599 @item -fexpensive-optimizations
6600 @opindex fexpensive-optimizations
6601 Perform a number of minor optimizations that are relatively expensive.
6603 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6605 @item -foptimize-register-move
6607 @opindex foptimize-register-move
6609 Attempt to reassign register numbers in move instructions and as
6610 operands of other simple instructions in order to maximize the amount of
6611 register tying. This is especially helpful on machines with two-operand
6614 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6617 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6619 @item -fira-algorithm=@var{algorithm}
6620 Use specified coloring algorithm for the integrated register
6621 allocator. The @var{algorithm} argument should be @code{priority} or
6622 @code{CB}. The first algorithm specifies Chow's priority coloring,
6623 the second one specifies Chaitin-Briggs coloring. The second
6624 algorithm can be unimplemented for some architectures. If it is
6625 implemented, it is the default because Chaitin-Briggs coloring as a
6626 rule generates a better code.
6628 @item -fira-region=@var{region}
6629 Use specified regions for the integrated register allocator. The
6630 @var{region} argument should be one of @code{all}, @code{mixed}, or
6631 @code{one}. The first value means using all loops as register
6632 allocation regions, the second value which is the default means using
6633 all loops except for loops with small register pressure as the
6634 regions, and third one means using all function as a single region.
6635 The first value can give best result for machines with small size and
6636 irregular register set, the third one results in faster and generates
6637 decent code and the smallest size code, and the default value usually
6638 give the best results in most cases and for most architectures.
6640 @item -fira-loop-pressure
6641 @opindex fira-loop-pressure
6642 Use IRA to evaluate register pressure in loops for decision to move
6643 loop invariants. Usage of this option usually results in generation
6644 of faster and smaller code on machines with big register files (>= 32
6645 registers) but it can slow compiler down.
6647 This option is enabled at level @option{-O3} for some targets.
6649 @item -fno-ira-share-save-slots
6650 @opindex fno-ira-share-save-slots
6651 Switch off sharing stack slots used for saving call used hard
6652 registers living through a call. Each hard register will get a
6653 separate stack slot and as a result function stack frame will be
6656 @item -fno-ira-share-spill-slots
6657 @opindex fno-ira-share-spill-slots
6658 Switch off sharing stack slots allocated for pseudo-registers. Each
6659 pseudo-register which did not get a hard register will get a separate
6660 stack slot and as a result function stack frame will be bigger.
6662 @item -fira-verbose=@var{n}
6663 @opindex fira-verbose
6664 Set up how verbose dump file for the integrated register allocator
6665 will be. Default value is 5. If the value is greater or equal to 10,
6666 the dump file will be stderr as if the value were @var{n} minus 10.
6668 @item -fdelayed-branch
6669 @opindex fdelayed-branch
6670 If supported for the target machine, attempt to reorder instructions
6671 to exploit instruction slots available after delayed branch
6674 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6676 @item -fschedule-insns
6677 @opindex fschedule-insns
6678 If supported for the target machine, attempt to reorder instructions to
6679 eliminate execution stalls due to required data being unavailable. This
6680 helps machines that have slow floating point or memory load instructions
6681 by allowing other instructions to be issued until the result of the load
6682 or floating point instruction is required.
6684 Enabled at levels @option{-O2}, @option{-O3}.
6686 @item -fschedule-insns2
6687 @opindex fschedule-insns2
6688 Similar to @option{-fschedule-insns}, but requests an additional pass of
6689 instruction scheduling after register allocation has been done. This is
6690 especially useful on machines with a relatively small number of
6691 registers and where memory load instructions take more than one cycle.
6693 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6695 @item -fno-sched-interblock
6696 @opindex fno-sched-interblock
6697 Don't schedule instructions across basic blocks. This is normally
6698 enabled by default when scheduling before register allocation, i.e.@:
6699 with @option{-fschedule-insns} or at @option{-O2} or higher.
6701 @item -fno-sched-spec
6702 @opindex fno-sched-spec
6703 Don't allow speculative motion of non-load instructions. This is normally
6704 enabled by default when scheduling before register allocation, i.e.@:
6705 with @option{-fschedule-insns} or at @option{-O2} or higher.
6707 @item -fsched-pressure
6708 @opindex fsched-pressure
6709 Enable register pressure sensitive insn scheduling before the register
6710 allocation. This only makes sense when scheduling before register
6711 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6712 @option{-O2} or higher. Usage of this option can improve the
6713 generated code and decrease its size by preventing register pressure
6714 increase above the number of available hard registers and as a
6715 consequence register spills in the register allocation.
6717 @item -fsched-spec-load
6718 @opindex fsched-spec-load
6719 Allow speculative motion of some load instructions. This only makes
6720 sense when scheduling before register allocation, i.e.@: with
6721 @option{-fschedule-insns} or at @option{-O2} or higher.
6723 @item -fsched-spec-load-dangerous
6724 @opindex fsched-spec-load-dangerous
6725 Allow speculative motion of more load instructions. This only makes
6726 sense when scheduling before register allocation, i.e.@: with
6727 @option{-fschedule-insns} or at @option{-O2} or higher.
6729 @item -fsched-stalled-insns
6730 @itemx -fsched-stalled-insns=@var{n}
6731 @opindex fsched-stalled-insns
6732 Define how many insns (if any) can be moved prematurely from the queue
6733 of stalled insns into the ready list, during the second scheduling pass.
6734 @option{-fno-sched-stalled-insns} means that no insns will be moved
6735 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6736 on how many queued insns can be moved prematurely.
6737 @option{-fsched-stalled-insns} without a value is equivalent to
6738 @option{-fsched-stalled-insns=1}.
6740 @item -fsched-stalled-insns-dep
6741 @itemx -fsched-stalled-insns-dep=@var{n}
6742 @opindex fsched-stalled-insns-dep
6743 Define how many insn groups (cycles) will be examined for a dependency
6744 on a stalled insn that is candidate for premature removal from the queue
6745 of stalled insns. This has an effect only during the second scheduling pass,
6746 and only if @option{-fsched-stalled-insns} is used.
6747 @option{-fno-sched-stalled-insns-dep} is equivalent to
6748 @option{-fsched-stalled-insns-dep=0}.
6749 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6750 @option{-fsched-stalled-insns-dep=1}.
6752 @item -fsched2-use-superblocks
6753 @opindex fsched2-use-superblocks
6754 When scheduling after register allocation, do use superblock scheduling
6755 algorithm. Superblock scheduling allows motion across basic block boundaries
6756 resulting on faster schedules. This option is experimental, as not all machine
6757 descriptions used by GCC model the CPU closely enough to avoid unreliable
6758 results from the algorithm.
6760 This only makes sense when scheduling after register allocation, i.e.@: with
6761 @option{-fschedule-insns2} or at @option{-O2} or higher.
6763 @item -fsched-group-heuristic
6764 @opindex fsched-group-heuristic
6765 Enable the group heuristic in the scheduler. This heuristic favors
6766 the instruction that belongs to a schedule group. This is enabled
6767 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6768 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6770 @item -fsched-critical-path-heuristic
6771 @opindex fsched-critical-path-heuristic
6772 Enable the critical-path heuristic in the scheduler. This heuristic favors
6773 instructions on the critical path. This is enabled by default when
6774 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6775 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6777 @item -fsched-spec-insn-heuristic
6778 @opindex fsched-spec-insn-heuristic
6779 Enable the speculative instruction heuristic in the scheduler. This
6780 heuristic favors speculative instructions with greater dependency weakness.
6781 This is enabled by default when scheduling is enabled, i.e.@:
6782 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6783 or at @option{-O2} or higher.
6785 @item -fsched-rank-heuristic
6786 @opindex fsched-rank-heuristic
6787 Enable the rank heuristic in the scheduler. This heuristic favors
6788 the instruction belonging to a basic block with greater size or frequency.
6789 This is enabled by default when scheduling is enabled, i.e.@:
6790 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6791 at @option{-O2} or higher.
6793 @item -fsched-last-insn-heuristic
6794 @opindex fsched-last-insn-heuristic
6795 Enable the last-instruction heuristic in the scheduler. This heuristic
6796 favors the instruction that is less dependent on the last instruction
6797 scheduled. This is enabled by default when scheduling is enabled,
6798 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6799 at @option{-O2} or higher.
6801 @item -fsched-dep-count-heuristic
6802 @opindex fsched-dep-count-heuristic
6803 Enable the dependent-count heuristic in the scheduler. This heuristic
6804 favors the instruction that has more instructions depending on it.
6805 This is enabled by default when scheduling is enabled, i.e.@:
6806 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6807 at @option{-O2} or higher.
6809 @item -freschedule-modulo-scheduled-loops
6810 @opindex freschedule-modulo-scheduled-loops
6811 The modulo scheduling comes before the traditional scheduling, if a loop
6812 was modulo scheduled we may want to prevent the later scheduling passes
6813 from changing its schedule, we use this option to control that.
6815 @item -fselective-scheduling
6816 @opindex fselective-scheduling
6817 Schedule instructions using selective scheduling algorithm. Selective
6818 scheduling runs instead of the first scheduler pass.
6820 @item -fselective-scheduling2
6821 @opindex fselective-scheduling2
6822 Schedule instructions using selective scheduling algorithm. Selective
6823 scheduling runs instead of the second scheduler pass.
6825 @item -fsel-sched-pipelining
6826 @opindex fsel-sched-pipelining
6827 Enable software pipelining of innermost loops during selective scheduling.
6828 This option has no effect until one of @option{-fselective-scheduling} or
6829 @option{-fselective-scheduling2} is turned on.
6831 @item -fsel-sched-pipelining-outer-loops
6832 @opindex fsel-sched-pipelining-outer-loops
6833 When pipelining loops during selective scheduling, also pipeline outer loops.
6834 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6836 @item -fcaller-saves
6837 @opindex fcaller-saves
6838 Enable values to be allocated in registers that will be clobbered by
6839 function calls, by emitting extra instructions to save and restore the
6840 registers around such calls. Such allocation is done only when it
6841 seems to result in better code than would otherwise be produced.
6843 This option is always enabled by default on certain machines, usually
6844 those which have no call-preserved registers to use instead.
6846 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6848 @item -fcombine-stack-adjustments
6849 @opindex fcombine-stack-adjustments
6850 Tracks stack adjustments (pushes and pops) and stack memory references
6851 and then tries to find ways to combine them.
6853 Enabled by default at @option{-O1} and higher.
6855 @item -fconserve-stack
6856 @opindex fconserve-stack
6857 Attempt to minimize stack usage. The compiler will attempt to use less
6858 stack space, even if that makes the program slower. This option
6859 implies setting the @option{large-stack-frame} parameter to 100
6860 and the @option{large-stack-frame-growth} parameter to 400.
6862 @item -ftree-reassoc
6863 @opindex ftree-reassoc
6864 Perform reassociation on trees. This flag is enabled by default
6865 at @option{-O} and higher.
6869 Perform partial redundancy elimination (PRE) on trees. This flag is
6870 enabled by default at @option{-O2} and @option{-O3}.
6872 @item -ftree-forwprop
6873 @opindex ftree-forwprop
6874 Perform forward propagation on trees. This flag is enabled by default
6875 at @option{-O} and higher.
6879 Perform full redundancy elimination (FRE) on trees. The difference
6880 between FRE and PRE is that FRE only considers expressions
6881 that are computed on all paths leading to the redundant computation.
6882 This analysis is faster than PRE, though it exposes fewer redundancies.
6883 This flag is enabled by default at @option{-O} and higher.
6885 @item -ftree-phiprop
6886 @opindex ftree-phiprop
6887 Perform hoisting of loads from conditional pointers on trees. This
6888 pass is enabled by default at @option{-O} and higher.
6890 @item -ftree-copy-prop
6891 @opindex ftree-copy-prop
6892 Perform copy propagation on trees. This pass eliminates unnecessary
6893 copy operations. This flag is enabled by default at @option{-O} and
6896 @item -fipa-pure-const
6897 @opindex fipa-pure-const
6898 Discover which functions are pure or constant.
6899 Enabled by default at @option{-O} and higher.
6901 @item -fipa-reference
6902 @opindex fipa-reference
6903 Discover which static variables do not escape cannot escape the
6905 Enabled by default at @option{-O} and higher.
6909 Perform interprocedural pointer analysis and interprocedural modification
6910 and reference analysis. This option can cause excessive memory and
6911 compile-time usage on large compilation units. It is not enabled by
6912 default at any optimization level.
6915 @opindex fipa-profile
6916 Perform interprocedural profile propagation. The functions called only from
6917 cold functions are marked as cold. Also functions executed once (such as
6918 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6919 functions and loop less parts of functions executed once are then optimized for
6921 Enabled by default at @option{-O} and higher.
6925 Perform interprocedural constant propagation.
6926 This optimization analyzes the program to determine when values passed
6927 to functions are constants and then optimizes accordingly.
6928 This optimization can substantially increase performance
6929 if the application has constants passed to functions.
6930 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6932 @item -fipa-cp-clone
6933 @opindex fipa-cp-clone
6934 Perform function cloning to make interprocedural constant propagation stronger.
6935 When enabled, interprocedural constant propagation will perform function cloning
6936 when externally visible function can be called with constant arguments.
6937 Because this optimization can create multiple copies of functions,
6938 it may significantly increase code size
6939 (see @option{--param ipcp-unit-growth=@var{value}}).
6940 This flag is enabled by default at @option{-O3}.
6942 @item -fipa-matrix-reorg
6943 @opindex fipa-matrix-reorg
6944 Perform matrix flattening and transposing.
6945 Matrix flattening tries to replace an @math{m}-dimensional matrix
6946 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6947 This reduces the level of indirection needed for accessing the elements
6948 of the matrix. The second optimization is matrix transposing that
6949 attempts to change the order of the matrix's dimensions in order to
6950 improve cache locality.
6951 Both optimizations need the @option{-fwhole-program} flag.
6952 Transposing is enabled only if profiling information is available.
6956 Perform forward store motion on trees. This flag is
6957 enabled by default at @option{-O} and higher.
6959 @item -ftree-bit-ccp
6960 @opindex ftree-bit-ccp
6961 Perform sparse conditional bit constant propagation on trees and propagate
6962 pointer alignment information.
6963 This pass only operates on local scalar variables and is enabled by default
6964 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
6968 Perform sparse conditional constant propagation (CCP) on trees. This
6969 pass only operates on local scalar variables and is enabled by default
6970 at @option{-O} and higher.
6972 @item -ftree-switch-conversion
6973 Perform conversion of simple initializations in a switch to
6974 initializations from a scalar array. This flag is enabled by default
6975 at @option{-O2} and higher.
6979 Perform dead code elimination (DCE) on trees. This flag is enabled by
6980 default at @option{-O} and higher.
6982 @item -ftree-builtin-call-dce
6983 @opindex ftree-builtin-call-dce
6984 Perform conditional dead code elimination (DCE) for calls to builtin functions
6985 that may set @code{errno} but are otherwise side-effect free. This flag is
6986 enabled by default at @option{-O2} and higher if @option{-Os} is not also
6989 @item -ftree-dominator-opts
6990 @opindex ftree-dominator-opts
6991 Perform a variety of simple scalar cleanups (constant/copy
6992 propagation, redundancy elimination, range propagation and expression
6993 simplification) based on a dominator tree traversal. This also
6994 performs jump threading (to reduce jumps to jumps). This flag is
6995 enabled by default at @option{-O} and higher.
6999 Perform dead store elimination (DSE) on trees. A dead store is a store into
7000 a memory location which will later be overwritten by another store without
7001 any intervening loads. In this case the earlier store can be deleted. This
7002 flag is enabled by default at @option{-O} and higher.
7006 Perform loop header copying on trees. This is beneficial since it increases
7007 effectiveness of code motion optimizations. It also saves one jump. This flag
7008 is enabled by default at @option{-O} and higher. It is not enabled
7009 for @option{-Os}, since it usually increases code size.
7011 @item -ftree-loop-optimize
7012 @opindex ftree-loop-optimize
7013 Perform loop optimizations on trees. This flag is enabled by default
7014 at @option{-O} and higher.
7016 @item -ftree-loop-linear
7017 @opindex ftree-loop-linear
7018 Perform loop interchange transformations on tree. Same as
7019 @option{-floop-interchange}. To use this code transformation, GCC has
7020 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7021 enable the Graphite loop transformation infrastructure.
7023 @item -floop-interchange
7024 @opindex floop-interchange
7025 Perform loop interchange transformations on loops. Interchanging two
7026 nested loops switches the inner and outer loops. For example, given a
7031 A(J, I) = A(J, I) * C
7035 loop interchange will transform the loop as if the user had written:
7039 A(J, I) = A(J, I) * C
7043 which can be beneficial when @code{N} is larger than the caches,
7044 because in Fortran, the elements of an array are stored in memory
7045 contiguously by column, and the original loop iterates over rows,
7046 potentially creating at each access a cache miss. This optimization
7047 applies to all the languages supported by GCC and is not limited to
7048 Fortran. To use this code transformation, GCC has to be configured
7049 with @option{--with-ppl} and @option{--with-cloog} to enable the
7050 Graphite loop transformation infrastructure.
7052 @item -floop-strip-mine
7053 @opindex floop-strip-mine
7054 Perform loop strip mining transformations on loops. Strip mining
7055 splits a loop into two nested loops. The outer loop has strides
7056 equal to the strip size and the inner loop has strides of the
7057 original loop within a strip. The strip length can be changed
7058 using the @option{loop-block-tile-size} parameter. For example,
7065 loop strip mining will transform the loop as if the user had written:
7068 DO I = II, min (II + 50, N)
7073 This optimization applies to all the languages supported by GCC and is
7074 not limited to Fortran. To use this code transformation, GCC has to
7075 be configured with @option{--with-ppl} and @option{--with-cloog} to
7076 enable the Graphite loop transformation infrastructure.
7079 @opindex floop-block
7080 Perform loop blocking transformations on loops. Blocking strip mines
7081 each loop in the loop nest such that the memory accesses of the
7082 element loops fit inside caches. The strip length can be changed
7083 using the @option{loop-block-tile-size} parameter. For example, given
7088 A(J, I) = B(I) + C(J)
7092 loop blocking will transform the loop as if the user had written:
7096 DO I = II, min (II + 50, N)
7097 DO J = JJ, min (JJ + 50, M)
7098 A(J, I) = B(I) + C(J)
7104 which can be beneficial when @code{M} is larger than the caches,
7105 because the innermost loop will iterate over a smaller amount of data
7106 that can be kept in the caches. This optimization applies to all the
7107 languages supported by GCC and is not limited to Fortran. To use this
7108 code transformation, GCC has to be configured with @option{--with-ppl}
7109 and @option{--with-cloog} to enable the Graphite loop transformation
7112 @item -fgraphite-identity
7113 @opindex fgraphite-identity
7114 Enable the identity transformation for graphite. For every SCoP we generate
7115 the polyhedral representation and transform it back to gimple. Using
7116 @option{-fgraphite-identity} we can check the costs or benefits of the
7117 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7118 are also performed by the code generator CLooG, like index splitting and
7119 dead code elimination in loops.
7121 @item -floop-flatten
7122 @opindex floop-flatten
7123 Removes the loop nesting structure: transforms the loop nest into a
7124 single loop. This transformation can be useful to vectorize all the
7125 levels of the loop nest.
7127 @item -floop-parallelize-all
7128 @opindex floop-parallelize-all
7129 Use the Graphite data dependence analysis to identify loops that can
7130 be parallelized. Parallelize all the loops that can be analyzed to
7131 not contain loop carried dependences without checking that it is
7132 profitable to parallelize the loops.
7134 @item -fcheck-data-deps
7135 @opindex fcheck-data-deps
7136 Compare the results of several data dependence analyzers. This option
7137 is used for debugging the data dependence analyzers.
7139 @item -ftree-loop-if-convert
7140 Attempt to transform conditional jumps in the innermost loops to
7141 branch-less equivalents. The intent is to remove control-flow from
7142 the innermost loops in order to improve the ability of the
7143 vectorization pass to handle these loops. This is enabled by default
7144 if vectorization is enabled.
7146 @item -ftree-loop-if-convert-stores
7147 Attempt to also if-convert conditional jumps containing memory writes.
7148 This transformation can be unsafe for multi-threaded programs as it
7149 transforms conditional memory writes into unconditional memory writes.
7152 for (i = 0; i < N; i++)
7156 would be transformed to
7158 for (i = 0; i < N; i++)
7159 A[i] = cond ? expr : A[i];
7161 potentially producing data races.
7163 @item -ftree-loop-distribution
7164 Perform loop distribution. This flag can improve cache performance on
7165 big loop bodies and allow further loop optimizations, like
7166 parallelization or vectorization, to take place. For example, the loop
7183 @item -ftree-loop-distribute-patterns
7184 Perform loop distribution of patterns that can be code generated with
7185 calls to a library. This flag is enabled by default at @option{-O3}.
7187 This pass distributes the initialization loops and generates a call to
7188 memset zero. For example, the loop
7204 and the initialization loop is transformed into a call to memset zero.
7206 @item -ftree-loop-im
7207 @opindex ftree-loop-im
7208 Perform loop invariant motion on trees. This pass moves only invariants that
7209 would be hard to handle at RTL level (function calls, operations that expand to
7210 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7211 operands of conditions that are invariant out of the loop, so that we can use
7212 just trivial invariantness analysis in loop unswitching. The pass also includes
7215 @item -ftree-loop-ivcanon
7216 @opindex ftree-loop-ivcanon
7217 Create a canonical counter for number of iterations in the loop for that
7218 determining number of iterations requires complicated analysis. Later
7219 optimizations then may determine the number easily. Useful especially
7220 in connection with unrolling.
7224 Perform induction variable optimizations (strength reduction, induction
7225 variable merging and induction variable elimination) on trees.
7227 @item -ftree-parallelize-loops=n
7228 @opindex ftree-parallelize-loops
7229 Parallelize loops, i.e., split their iteration space to run in n threads.
7230 This is only possible for loops whose iterations are independent
7231 and can be arbitrarily reordered. The optimization is only
7232 profitable on multiprocessor machines, for loops that are CPU-intensive,
7233 rather than constrained e.g.@: by memory bandwidth. This option
7234 implies @option{-pthread}, and thus is only supported on targets
7235 that have support for @option{-pthread}.
7239 Perform function-local points-to analysis on trees. This flag is
7240 enabled by default at @option{-O} and higher.
7244 Perform scalar replacement of aggregates. This pass replaces structure
7245 references with scalars to prevent committing structures to memory too
7246 early. This flag is enabled by default at @option{-O} and higher.
7248 @item -ftree-copyrename
7249 @opindex ftree-copyrename
7250 Perform copy renaming on trees. This pass attempts to rename compiler
7251 temporaries to other variables at copy locations, usually resulting in
7252 variable names which more closely resemble the original variables. This flag
7253 is enabled by default at @option{-O} and higher.
7257 Perform temporary expression replacement during the SSA->normal phase. Single
7258 use/single def temporaries are replaced at their use location with their
7259 defining expression. This results in non-GIMPLE code, but gives the expanders
7260 much more complex trees to work on resulting in better RTL generation. This is
7261 enabled by default at @option{-O} and higher.
7263 @item -ftree-vectorize
7264 @opindex ftree-vectorize
7265 Perform loop vectorization on trees. This flag is enabled by default at
7268 @item -ftree-slp-vectorize
7269 @opindex ftree-slp-vectorize
7270 Perform basic block vectorization on trees. This flag is enabled by default at
7271 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7273 @item -ftree-vect-loop-version
7274 @opindex ftree-vect-loop-version
7275 Perform loop versioning when doing loop vectorization on trees. When a loop
7276 appears to be vectorizable except that data alignment or data dependence cannot
7277 be determined at compile time then vectorized and non-vectorized versions of
7278 the loop are generated along with runtime checks for alignment or dependence
7279 to control which version is executed. This option is enabled by default
7280 except at level @option{-Os} where it is disabled.
7282 @item -fvect-cost-model
7283 @opindex fvect-cost-model
7284 Enable cost model for vectorization.
7288 Perform Value Range Propagation on trees. This is similar to the
7289 constant propagation pass, but instead of values, ranges of values are
7290 propagated. This allows the optimizers to remove unnecessary range
7291 checks like array bound checks and null pointer checks. This is
7292 enabled by default at @option{-O2} and higher. Null pointer check
7293 elimination is only done if @option{-fdelete-null-pointer-checks} is
7298 Perform tail duplication to enlarge superblock size. This transformation
7299 simplifies the control flow of the function allowing other optimizations to do
7302 @item -funroll-loops
7303 @opindex funroll-loops
7304 Unroll loops whose number of iterations can be determined at compile
7305 time or upon entry to the loop. @option{-funroll-loops} implies
7306 @option{-frerun-cse-after-loop}. This option makes code larger,
7307 and may or may not make it run faster.
7309 @item -funroll-all-loops
7310 @opindex funroll-all-loops
7311 Unroll all loops, even if their number of iterations is uncertain when
7312 the loop is entered. This usually makes programs run more slowly.
7313 @option{-funroll-all-loops} implies the same options as
7314 @option{-funroll-loops},
7316 @item -fsplit-ivs-in-unroller
7317 @opindex fsplit-ivs-in-unroller
7318 Enables expressing of values of induction variables in later iterations
7319 of the unrolled loop using the value in the first iteration. This breaks
7320 long dependency chains, thus improving efficiency of the scheduling passes.
7322 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7323 same effect. However in cases the loop body is more complicated than
7324 a single basic block, this is not reliable. It also does not work at all
7325 on some of the architectures due to restrictions in the CSE pass.
7327 This optimization is enabled by default.
7329 @item -fvariable-expansion-in-unroller
7330 @opindex fvariable-expansion-in-unroller
7331 With this option, the compiler will create multiple copies of some
7332 local variables when unrolling a loop which can result in superior code.
7334 @item -fpartial-inlining
7335 @opindex fpartial-inlining
7336 Inline parts of functions. This option has any effect only
7337 when inlining itself is turned on by the @option{-finline-functions}
7338 or @option{-finline-small-functions} options.
7340 Enabled at level @option{-O2}.
7342 @item -fpredictive-commoning
7343 @opindex fpredictive-commoning
7344 Perform predictive commoning optimization, i.e., reusing computations
7345 (especially memory loads and stores) performed in previous
7346 iterations of loops.
7348 This option is enabled at level @option{-O3}.
7350 @item -fprefetch-loop-arrays
7351 @opindex fprefetch-loop-arrays
7352 If supported by the target machine, generate instructions to prefetch
7353 memory to improve the performance of loops that access large arrays.
7355 This option may generate better or worse code; results are highly
7356 dependent on the structure of loops within the source code.
7358 Disabled at level @option{-Os}.
7361 @itemx -fno-peephole2
7362 @opindex fno-peephole
7363 @opindex fno-peephole2
7364 Disable any machine-specific peephole optimizations. The difference
7365 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7366 are implemented in the compiler; some targets use one, some use the
7367 other, a few use both.
7369 @option{-fpeephole} is enabled by default.
7370 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7372 @item -fno-guess-branch-probability
7373 @opindex fno-guess-branch-probability
7374 Do not guess branch probabilities using heuristics.
7376 GCC will use heuristics to guess branch probabilities if they are
7377 not provided by profiling feedback (@option{-fprofile-arcs}). These
7378 heuristics are based on the control flow graph. If some branch probabilities
7379 are specified by @samp{__builtin_expect}, then the heuristics will be
7380 used to guess branch probabilities for the rest of the control flow graph,
7381 taking the @samp{__builtin_expect} info into account. The interactions
7382 between the heuristics and @samp{__builtin_expect} can be complex, and in
7383 some cases, it may be useful to disable the heuristics so that the effects
7384 of @samp{__builtin_expect} are easier to understand.
7386 The default is @option{-fguess-branch-probability} at levels
7387 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7389 @item -freorder-blocks
7390 @opindex freorder-blocks
7391 Reorder basic blocks in the compiled function in order to reduce number of
7392 taken branches and improve code locality.
7394 Enabled at levels @option{-O2}, @option{-O3}.
7396 @item -freorder-blocks-and-partition
7397 @opindex freorder-blocks-and-partition
7398 In addition to reordering basic blocks in the compiled function, in order
7399 to reduce number of taken branches, partitions hot and cold basic blocks
7400 into separate sections of the assembly and .o files, to improve
7401 paging and cache locality performance.
7403 This optimization is automatically turned off in the presence of
7404 exception handling, for linkonce sections, for functions with a user-defined
7405 section attribute and on any architecture that does not support named
7408 @item -freorder-functions
7409 @opindex freorder-functions
7410 Reorder functions in the object file in order to
7411 improve code locality. This is implemented by using special
7412 subsections @code{.text.hot} for most frequently executed functions and
7413 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7414 the linker so object file format must support named sections and linker must
7415 place them in a reasonable way.
7417 Also profile feedback must be available in to make this option effective. See
7418 @option{-fprofile-arcs} for details.
7420 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7422 @item -fstrict-aliasing
7423 @opindex fstrict-aliasing
7424 Allow the compiler to assume the strictest aliasing rules applicable to
7425 the language being compiled. For C (and C++), this activates
7426 optimizations based on the type of expressions. In particular, an
7427 object of one type is assumed never to reside at the same address as an
7428 object of a different type, unless the types are almost the same. For
7429 example, an @code{unsigned int} can alias an @code{int}, but not a
7430 @code{void*} or a @code{double}. A character type may alias any other
7433 @anchor{Type-punning}Pay special attention to code like this:
7446 The practice of reading from a different union member than the one most
7447 recently written to (called ``type-punning'') is common. Even with
7448 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7449 is accessed through the union type. So, the code above will work as
7450 expected. @xref{Structures unions enumerations and bit-fields
7451 implementation}. However, this code might not:
7462 Similarly, access by taking the address, casting the resulting pointer
7463 and dereferencing the result has undefined behavior, even if the cast
7464 uses a union type, e.g.:
7468 return ((union a_union *) &d)->i;
7472 The @option{-fstrict-aliasing} option is enabled at levels
7473 @option{-O2}, @option{-O3}, @option{-Os}.
7475 @item -fstrict-overflow
7476 @opindex fstrict-overflow
7477 Allow the compiler to assume strict signed overflow rules, depending
7478 on the language being compiled. For C (and C++) this means that
7479 overflow when doing arithmetic with signed numbers is undefined, which
7480 means that the compiler may assume that it will not happen. This
7481 permits various optimizations. For example, the compiler will assume
7482 that an expression like @code{i + 10 > i} will always be true for
7483 signed @code{i}. This assumption is only valid if signed overflow is
7484 undefined, as the expression is false if @code{i + 10} overflows when
7485 using twos complement arithmetic. When this option is in effect any
7486 attempt to determine whether an operation on signed numbers will
7487 overflow must be written carefully to not actually involve overflow.
7489 This option also allows the compiler to assume strict pointer
7490 semantics: given a pointer to an object, if adding an offset to that
7491 pointer does not produce a pointer to the same object, the addition is
7492 undefined. This permits the compiler to conclude that @code{p + u >
7493 p} is always true for a pointer @code{p} and unsigned integer
7494 @code{u}. This assumption is only valid because pointer wraparound is
7495 undefined, as the expression is false if @code{p + u} overflows using
7496 twos complement arithmetic.
7498 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7499 that integer signed overflow is fully defined: it wraps. When
7500 @option{-fwrapv} is used, there is no difference between
7501 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7502 integers. With @option{-fwrapv} certain types of overflow are
7503 permitted. For example, if the compiler gets an overflow when doing
7504 arithmetic on constants, the overflowed value can still be used with
7505 @option{-fwrapv}, but not otherwise.
7507 The @option{-fstrict-overflow} option is enabled at levels
7508 @option{-O2}, @option{-O3}, @option{-Os}.
7510 @item -falign-functions
7511 @itemx -falign-functions=@var{n}
7512 @opindex falign-functions
7513 Align the start of functions to the next power-of-two greater than
7514 @var{n}, skipping up to @var{n} bytes. For instance,
7515 @option{-falign-functions=32} aligns functions to the next 32-byte
7516 boundary, but @option{-falign-functions=24} would align to the next
7517 32-byte boundary only if this can be done by skipping 23 bytes or less.
7519 @option{-fno-align-functions} and @option{-falign-functions=1} are
7520 equivalent and mean that functions will not be aligned.
7522 Some assemblers only support this flag when @var{n} is a power of two;
7523 in that case, it is rounded up.
7525 If @var{n} is not specified or is zero, use a machine-dependent default.
7527 Enabled at levels @option{-O2}, @option{-O3}.
7529 @item -falign-labels
7530 @itemx -falign-labels=@var{n}
7531 @opindex falign-labels
7532 Align all branch targets to a power-of-two boundary, skipping up to
7533 @var{n} bytes like @option{-falign-functions}. This option can easily
7534 make code slower, because it must insert dummy operations for when the
7535 branch target is reached in the usual flow of the code.
7537 @option{-fno-align-labels} and @option{-falign-labels=1} are
7538 equivalent and mean that labels will not be aligned.
7540 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7541 are greater than this value, then their values are used instead.
7543 If @var{n} is not specified or is zero, use a machine-dependent default
7544 which is very likely to be @samp{1}, meaning no alignment.
7546 Enabled at levels @option{-O2}, @option{-O3}.
7549 @itemx -falign-loops=@var{n}
7550 @opindex falign-loops
7551 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7552 like @option{-falign-functions}. The hope is that the loop will be
7553 executed many times, which will make up for any execution of the dummy
7556 @option{-fno-align-loops} and @option{-falign-loops=1} are
7557 equivalent and mean that loops will not be aligned.
7559 If @var{n} is not specified or is zero, use a machine-dependent default.
7561 Enabled at levels @option{-O2}, @option{-O3}.
7564 @itemx -falign-jumps=@var{n}
7565 @opindex falign-jumps
7566 Align branch targets to a power-of-two boundary, for branch targets
7567 where the targets can only be reached by jumping, skipping up to @var{n}
7568 bytes like @option{-falign-functions}. In this case, no dummy operations
7571 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7572 equivalent and mean that loops will not be aligned.
7574 If @var{n} is not specified or is zero, use a machine-dependent default.
7576 Enabled at levels @option{-O2}, @option{-O3}.
7578 @item -funit-at-a-time
7579 @opindex funit-at-a-time
7580 This option is left for compatibility reasons. @option{-funit-at-a-time}
7581 has no effect, while @option{-fno-unit-at-a-time} implies
7582 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7586 @item -fno-toplevel-reorder
7587 @opindex fno-toplevel-reorder
7588 Do not reorder top-level functions, variables, and @code{asm}
7589 statements. Output them in the same order that they appear in the
7590 input file. When this option is used, unreferenced static variables
7591 will not be removed. This option is intended to support existing code
7592 which relies on a particular ordering. For new code, it is better to
7595 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7596 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7601 Constructs webs as commonly used for register allocation purposes and assign
7602 each web individual pseudo register. This allows the register allocation pass
7603 to operate on pseudos directly, but also strengthens several other optimization
7604 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7605 however, make debugging impossible, since variables will no longer stay in a
7608 Enabled by default with @option{-funroll-loops}.
7610 @item -fwhole-program
7611 @opindex fwhole-program
7612 Assume that the current compilation unit represents the whole program being
7613 compiled. All public functions and variables with the exception of @code{main}
7614 and those merged by attribute @code{externally_visible} become static functions
7615 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.
7616 While this option is equivalent to proper use of the @code{static} keyword for
7617 programs consisting of a single file, in combination with option
7618 @option{-flto} this flag can be used to
7619 compile many smaller scale programs since the functions and variables become
7620 local for the whole combined compilation unit, not for the single source file
7623 This option implies @option{-fwhole-file} for Fortran programs.
7625 @item -flto[=@var{n}]
7627 This option runs the standard link-time optimizer. When invoked
7628 with source code, it generates GIMPLE (one of GCC's internal
7629 representations) and writes it to special ELF sections in the object
7630 file. When the object files are linked together, all the function
7631 bodies are read from these ELF sections and instantiated as if they
7632 had been part of the same translation unit.
7634 To use the link-timer optimizer, @option{-flto} needs to be specified at
7635 compile time and during the final link. For example,
7638 gcc -c -O2 -flto foo.c
7639 gcc -c -O2 -flto bar.c
7640 gcc -o myprog -flto -O2 foo.o bar.o
7643 The first two invocations to GCC will save a bytecode representation
7644 of GIMPLE into special ELF sections inside @file{foo.o} and
7645 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7646 @file{foo.o} and @file{bar.o}, merge the two files into a single
7647 internal image, and compile the result as usual. Since both
7648 @file{foo.o} and @file{bar.o} are merged into a single image, this
7649 causes all the inter-procedural analyses and optimizations in GCC to
7650 work across the two files as if they were a single one. This means,
7651 for example, that the inliner will be able to inline functions in
7652 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7654 Another (simpler) way to enable link-time optimization is,
7657 gcc -o myprog -flto -O2 foo.c bar.c
7660 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7661 merge them together into a single GIMPLE representation and optimize
7662 them as usual to produce @file{myprog}.
7664 The only important thing to keep in mind is that to enable link-time
7665 optimizations the @option{-flto} flag needs to be passed to both the
7666 compile and the link commands.
7668 To make whole program optimization effective, it is necessary to make
7669 certain whole program assumptions. The compiler needs to know
7670 what functions and variables can be accessed by libraries and runtime
7671 outside of the link time optimized unit. When supported by the linker,
7672 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7673 compiler information about used and externally visible symbols. When
7674 the linker plugin is not available, @option{-fwhole-program} should be
7675 used to allow the compiler to make these assumptions, which will lead
7676 to more aggressive optimization decisions.
7678 Note that when a file is compiled with @option{-flto}, the generated
7679 object file will be larger than a regular object file because it will
7680 contain GIMPLE bytecodes and the usual final code. This means that
7681 object files with LTO information can be linked as a normal object
7682 file. So, in the previous example, if the final link is done with
7685 gcc -o myprog foo.o bar.o
7688 The only difference will be that no inter-procedural optimizations
7689 will be applied to produce @file{myprog}. The two object files
7690 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7693 Additionally, the optimization flags used to compile individual files
7694 are not necessarily related to those used at link-time. For instance,
7697 gcc -c -O0 -flto foo.c
7698 gcc -c -O0 -flto bar.c
7699 gcc -o myprog -flto -O3 foo.o bar.o
7702 This will produce individual object files with unoptimized assembler
7703 code, but the resulting binary @file{myprog} will be optimized at
7704 @option{-O3}. Now, if the final binary is generated without
7705 @option{-flto}, then @file{myprog} will not be optimized.
7707 When producing the final binary with @option{-flto}, GCC will only
7708 apply link-time optimizations to those files that contain bytecode.
7709 Therefore, you can mix and match object files and libraries with
7710 GIMPLE bytecodes and final object code. GCC will automatically select
7711 which files to optimize in LTO mode and which files to link without
7714 There are some code generation flags that GCC will preserve when
7715 generating bytecodes, as they need to be used during the final link
7716 stage. Currently, the following options are saved into the GIMPLE
7717 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7718 @option{-m} target flags.
7720 At link time, these options are read-in and reapplied. Note that the
7721 current implementation makes no attempt at recognizing conflicting
7722 values for these options. If two or more files have a conflicting
7723 value (e.g., one file is compiled with @option{-fPIC} and another
7724 isn't), the compiler will simply use the last value read from the
7725 bytecode files. It is recommended, then, that all the files
7726 participating in the same link be compiled with the same options.
7728 Another feature of LTO is that it is possible to apply interprocedural
7729 optimizations on files written in different languages. This requires
7730 some support in the language front end. Currently, the C, C++ and
7731 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7732 something like this should work
7737 gfortran -c -flto baz.f90
7738 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7741 Notice that the final link is done with @command{g++} to get the C++
7742 runtime libraries and @option{-lgfortran} is added to get the Fortran
7743 runtime libraries. In general, when mixing languages in LTO mode, you
7744 should use the same link command used when mixing languages in a
7745 regular (non-LTO) compilation. This means that if your build process
7746 was mixing languages before, all you need to add is @option{-flto} to
7747 all the compile and link commands.
7749 If LTO encounters objects with C linkage declared with incompatible
7750 types in separate translation units to be linked together (undefined
7751 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7752 issued. The behavior is still undefined at runtime.
7754 If object files containing GIMPLE bytecode are stored in a library archive, say
7755 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7756 are using a linker with linker plugin support. To enable this feature, use
7757 the flag @option{-fuse-linker-plugin} at link-time:
7760 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7763 With the linker plugin enabled, the linker will extract the needed
7764 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7765 to make them part of the aggregated GIMPLE image to be optimized.
7767 If you are not using a linker with linker plugin support and/or do not
7768 enable linker plugin then the objects inside @file{libfoo.a}
7769 will be extracted and linked as usual, but they will not participate
7770 in the LTO optimization process.
7772 Link time optimizations do not require the presence of the whole program to
7773 operate. If the program does not require any symbols to be exported, it is
7774 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7775 the interprocedural optimizers to use more aggressive assumptions which may
7776 lead to improved optimization opportunities.
7777 Use of @option{-fwhole-program} is not needed when linker plugin is
7778 active (see @option{-fuse-linker-plugin}).
7780 Regarding portability: the current implementation of LTO makes no
7781 attempt at generating bytecode that can be ported between different
7782 types of hosts. The bytecode files are versioned and there is a
7783 strict version check, so bytecode files generated in one version of
7784 GCC will not work with an older/newer version of GCC.
7786 Link time optimization does not play well with generating debugging
7787 information. Combining @option{-flto} with
7788 @option{-g} is currently experimental and expected to produce wrong
7791 If you specify the optional @var{n}, the optimization and code
7792 generation done at link time is executed in parallel using @var{n}
7793 parallel jobs by utilizing an installed @command{make} program. The
7794 environment variable @env{MAKE} may be used to override the program
7795 used. The default value for @var{n} is 1.
7797 You can also specify @option{-flto=jobserver} to use GNU make's
7798 job server mode to determine the number of parallel jobs. This
7799 is useful when the Makefile calling GCC is already executing in parallel.
7800 The parent Makefile will need a @samp{+} prepended to the command recipe
7801 for this to work. This will likely only work if @env{MAKE} is
7804 This option is disabled by default.
7806 @item -flto-partition=@var{alg}
7807 @opindex flto-partition
7808 Specify the partitioning algorithm used by the link time optimizer.
7809 The value is either @code{1to1} to specify a partitioning mirroring
7810 the original source files or @code{balanced} to specify partitioning
7811 into equally sized chunks (whenever possible). Specifying @code{none}
7812 as an algorithm disables partitioning and streaming completely. The
7813 default value is @code{balanced}.
7815 @item -flto-compression-level=@var{n}
7816 This option specifies the level of compression used for intermediate
7817 language written to LTO object files, and is only meaningful in
7818 conjunction with LTO mode (@option{-flto}). Valid
7819 values are 0 (no compression) to 9 (maximum compression). Values
7820 outside this range are clamped to either 0 or 9. If the option is not
7821 given, a default balanced compression setting is used.
7824 Prints a report with internal details on the workings of the link-time
7825 optimizer. The contents of this report vary from version to version,
7826 it is meant to be useful to GCC developers when processing object
7827 files in LTO mode (via @option{-flto}).
7829 Disabled by default.
7831 @item -fuse-linker-plugin
7832 Enables the use of a linker plugin during link time optimization. This
7833 option relies on plugin support in the linker, which is available in gold
7834 or in GNU ld 2.21 or newer.
7836 This option enables the extraction of object files with GIMPLE bytecode out
7837 of library archives. This improves the quality of optimization by exposing
7838 more code to the link time optimizer. This information specifies what
7839 symbols can be accessed externally (by non-LTO object or during dynamic
7840 linking). Resulting code quality improvements on binaries (and shared
7841 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7842 See @option{-flto} for a description of the effect of this flag and how to
7845 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7846 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7848 @item -fcompare-elim
7849 @opindex fcompare-elim
7850 After register allocation and post-register allocation instruction splitting,
7851 identify arithmetic instructions that compute processor flags similar to a
7852 comparison operation based on that arithmetic. If possible, eliminate the
7853 explicit comparison operation.
7855 This pass only applies to certain targets that cannot explicitly represent
7856 the comparison operation before register allocation is complete.
7858 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7860 @item -fcprop-registers
7861 @opindex fcprop-registers
7862 After register allocation and post-register allocation instruction splitting,
7863 we perform a copy-propagation pass to try to reduce scheduling dependencies
7864 and occasionally eliminate the copy.
7866 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7868 @item -fprofile-correction
7869 @opindex fprofile-correction
7870 Profiles collected using an instrumented binary for multi-threaded programs may
7871 be inconsistent due to missed counter updates. When this option is specified,
7872 GCC will use heuristics to correct or smooth out such inconsistencies. By
7873 default, GCC will emit an error message when an inconsistent profile is detected.
7875 @item -fprofile-dir=@var{path}
7876 @opindex fprofile-dir
7878 Set the directory to search for the profile data files in to @var{path}.
7879 This option affects only the profile data generated by
7880 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7881 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7882 and its related options. Both absolute and relative paths can be used.
7883 By default, GCC will use the current directory as @var{path}, thus the
7884 profile data file will appear in the same directory as the object file.
7886 @item -fprofile-generate
7887 @itemx -fprofile-generate=@var{path}
7888 @opindex fprofile-generate
7890 Enable options usually used for instrumenting application to produce
7891 profile useful for later recompilation with profile feedback based
7892 optimization. You must use @option{-fprofile-generate} both when
7893 compiling and when linking your program.
7895 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7897 If @var{path} is specified, GCC will look at the @var{path} to find
7898 the profile feedback data files. See @option{-fprofile-dir}.
7901 @itemx -fprofile-use=@var{path}
7902 @opindex fprofile-use
7903 Enable profile feedback directed optimizations, and optimizations
7904 generally profitable only with profile feedback available.
7906 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7907 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7909 By default, GCC emits an error message if the feedback profiles do not
7910 match the source code. This error can be turned into a warning by using
7911 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7914 If @var{path} is specified, GCC will look at the @var{path} to find
7915 the profile feedback data files. See @option{-fprofile-dir}.
7918 The following options control compiler behavior regarding floating
7919 point arithmetic. These options trade off between speed and
7920 correctness. All must be specifically enabled.
7924 @opindex ffloat-store
7925 Do not store floating point variables in registers, and inhibit other
7926 options that might change whether a floating point value is taken from a
7929 @cindex floating point precision
7930 This option prevents undesirable excess precision on machines such as
7931 the 68000 where the floating registers (of the 68881) keep more
7932 precision than a @code{double} is supposed to have. Similarly for the
7933 x86 architecture. For most programs, the excess precision does only
7934 good, but a few programs rely on the precise definition of IEEE floating
7935 point. Use @option{-ffloat-store} for such programs, after modifying
7936 them to store all pertinent intermediate computations into variables.
7938 @item -fexcess-precision=@var{style}
7939 @opindex fexcess-precision
7940 This option allows further control over excess precision on machines
7941 where floating-point registers have more precision than the IEEE
7942 @code{float} and @code{double} types and the processor does not
7943 support operations rounding to those types. By default,
7944 @option{-fexcess-precision=fast} is in effect; this means that
7945 operations are carried out in the precision of the registers and that
7946 it is unpredictable when rounding to the types specified in the source
7947 code takes place. When compiling C, if
7948 @option{-fexcess-precision=standard} is specified then excess
7949 precision will follow the rules specified in ISO C99; in particular,
7950 both casts and assignments cause values to be rounded to their
7951 semantic types (whereas @option{-ffloat-store} only affects
7952 assignments). This option is enabled by default for C if a strict
7953 conformance option such as @option{-std=c99} is used.
7956 @option{-fexcess-precision=standard} is not implemented for languages
7957 other than C, and has no effect if
7958 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7959 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
7960 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
7961 semantics apply without excess precision, and in the latter, rounding
7966 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
7967 @option{-ffinite-math-only}, @option{-fno-rounding-math},
7968 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
7970 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
7972 This option is not turned on by any @option{-O} option besides
7973 @option{-Ofast} since it can result in incorrect output for programs
7974 which depend on an exact implementation of IEEE or ISO rules/specifications
7975 for math functions. It may, however, yield faster code for programs
7976 that do not require the guarantees of these specifications.
7978 @item -fno-math-errno
7979 @opindex fno-math-errno
7980 Do not set ERRNO after calling math functions that are executed
7981 with a single instruction, e.g., sqrt. A program that relies on
7982 IEEE exceptions for math error handling may want to use this flag
7983 for speed while maintaining IEEE arithmetic compatibility.
7985 This option is not turned on by any @option{-O} option since
7986 it can result in incorrect output for programs which depend on
7987 an exact implementation of IEEE or ISO rules/specifications for
7988 math functions. It may, however, yield faster code for programs
7989 that do not require the guarantees of these specifications.
7991 The default is @option{-fmath-errno}.
7993 On Darwin systems, the math library never sets @code{errno}. There is
7994 therefore no reason for the compiler to consider the possibility that
7995 it might, and @option{-fno-math-errno} is the default.
7997 @item -funsafe-math-optimizations
7998 @opindex funsafe-math-optimizations
8000 Allow optimizations for floating-point arithmetic that (a) assume
8001 that arguments and results are valid and (b) may violate IEEE or
8002 ANSI standards. When used at link-time, it may include libraries
8003 or startup files that change the default FPU control word or other
8004 similar optimizations.
8006 This option is not turned on by any @option{-O} option since
8007 it can result in incorrect output for programs which depend on
8008 an exact implementation of IEEE or ISO rules/specifications for
8009 math functions. It may, however, yield faster code for programs
8010 that do not require the guarantees of these specifications.
8011 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8012 @option{-fassociative-math} and @option{-freciprocal-math}.
8014 The default is @option{-fno-unsafe-math-optimizations}.
8016 @item -fassociative-math
8017 @opindex fassociative-math
8019 Allow re-association of operands in series of floating-point operations.
8020 This violates the ISO C and C++ language standard by possibly changing
8021 computation result. NOTE: re-ordering may change the sign of zero as
8022 well as ignore NaNs and inhibit or create underflow or overflow (and
8023 thus cannot be used on a code which relies on rounding behavior like
8024 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
8025 and thus may not be used when ordered comparisons are required.
8026 This option requires that both @option{-fno-signed-zeros} and
8027 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8028 much sense with @option{-frounding-math}. For Fortran the option
8029 is automatically enabled when both @option{-fno-signed-zeros} and
8030 @option{-fno-trapping-math} are in effect.
8032 The default is @option{-fno-associative-math}.
8034 @item -freciprocal-math
8035 @opindex freciprocal-math
8037 Allow the reciprocal of a value to be used instead of dividing by
8038 the value if this enables optimizations. For example @code{x / y}
8039 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8040 is subject to common subexpression elimination. Note that this loses
8041 precision and increases the number of flops operating on the value.
8043 The default is @option{-fno-reciprocal-math}.
8045 @item -ffinite-math-only
8046 @opindex ffinite-math-only
8047 Allow optimizations for floating-point arithmetic that assume
8048 that arguments and results are not NaNs or +-Infs.
8050 This option is not turned on by any @option{-O} option since
8051 it can result in incorrect output for programs which depend on
8052 an exact implementation of IEEE or ISO rules/specifications for
8053 math functions. It may, however, yield faster code for programs
8054 that do not require the guarantees of these specifications.
8056 The default is @option{-fno-finite-math-only}.
8058 @item -fno-signed-zeros
8059 @opindex fno-signed-zeros
8060 Allow optimizations for floating point arithmetic that ignore the
8061 signedness of zero. IEEE arithmetic specifies the behavior of
8062 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8063 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8064 This option implies that the sign of a zero result isn't significant.
8066 The default is @option{-fsigned-zeros}.
8068 @item -fno-trapping-math
8069 @opindex fno-trapping-math
8070 Compile code assuming that floating-point operations cannot generate
8071 user-visible traps. These traps include division by zero, overflow,
8072 underflow, inexact result and invalid operation. This option requires
8073 that @option{-fno-signaling-nans} be in effect. Setting this option may
8074 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8076 This option should never be turned on by any @option{-O} option since
8077 it can result in incorrect output for programs which depend on
8078 an exact implementation of IEEE or ISO rules/specifications for
8081 The default is @option{-ftrapping-math}.
8083 @item -frounding-math
8084 @opindex frounding-math
8085 Disable transformations and optimizations that assume default floating
8086 point rounding behavior. This is round-to-zero for all floating point
8087 to integer conversions, and round-to-nearest for all other arithmetic
8088 truncations. This option should be specified for programs that change
8089 the FP rounding mode dynamically, or that may be executed with a
8090 non-default rounding mode. This option disables constant folding of
8091 floating point expressions at compile-time (which may be affected by
8092 rounding mode) and arithmetic transformations that are unsafe in the
8093 presence of sign-dependent rounding modes.
8095 The default is @option{-fno-rounding-math}.
8097 This option is experimental and does not currently guarantee to
8098 disable all GCC optimizations that are affected by rounding mode.
8099 Future versions of GCC may provide finer control of this setting
8100 using C99's @code{FENV_ACCESS} pragma. This command line option
8101 will be used to specify the default state for @code{FENV_ACCESS}.
8103 @item -fsignaling-nans
8104 @opindex fsignaling-nans
8105 Compile code assuming that IEEE signaling NaNs may generate user-visible
8106 traps during floating-point operations. Setting this option disables
8107 optimizations that may change the number of exceptions visible with
8108 signaling NaNs. This option implies @option{-ftrapping-math}.
8110 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8113 The default is @option{-fno-signaling-nans}.
8115 This option is experimental and does not currently guarantee to
8116 disable all GCC optimizations that affect signaling NaN behavior.
8118 @item -fsingle-precision-constant
8119 @opindex fsingle-precision-constant
8120 Treat floating point constant as single precision constant instead of
8121 implicitly converting it to double precision constant.
8123 @item -fcx-limited-range
8124 @opindex fcx-limited-range
8125 When enabled, this option states that a range reduction step is not
8126 needed when performing complex division. Also, there is no checking
8127 whether the result of a complex multiplication or division is @code{NaN
8128 + I*NaN}, with an attempt to rescue the situation in that case. The
8129 default is @option{-fno-cx-limited-range}, but is enabled by
8130 @option{-ffast-math}.
8132 This option controls the default setting of the ISO C99
8133 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8136 @item -fcx-fortran-rules
8137 @opindex fcx-fortran-rules
8138 Complex multiplication and division follow Fortran rules. Range
8139 reduction is done as part of complex division, but there is no checking
8140 whether the result of a complex multiplication or division is @code{NaN
8141 + I*NaN}, with an attempt to rescue the situation in that case.
8143 The default is @option{-fno-cx-fortran-rules}.
8147 The following options control optimizations that may improve
8148 performance, but are not enabled by any @option{-O} options. This
8149 section includes experimental options that may produce broken code.
8152 @item -fbranch-probabilities
8153 @opindex fbranch-probabilities
8154 After running a program compiled with @option{-fprofile-arcs}
8155 (@pxref{Debugging Options,, Options for Debugging Your Program or
8156 @command{gcc}}), you can compile it a second time using
8157 @option{-fbranch-probabilities}, to improve optimizations based on
8158 the number of times each branch was taken. When the program
8159 compiled with @option{-fprofile-arcs} exits it saves arc execution
8160 counts to a file called @file{@var{sourcename}.gcda} for each source
8161 file. The information in this data file is very dependent on the
8162 structure of the generated code, so you must use the same source code
8163 and the same optimization options for both compilations.
8165 With @option{-fbranch-probabilities}, GCC puts a
8166 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8167 These can be used to improve optimization. Currently, they are only
8168 used in one place: in @file{reorg.c}, instead of guessing which path a
8169 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8170 exactly determine which path is taken more often.
8172 @item -fprofile-values
8173 @opindex fprofile-values
8174 If combined with @option{-fprofile-arcs}, it adds code so that some
8175 data about values of expressions in the program is gathered.
8177 With @option{-fbranch-probabilities}, it reads back the data gathered
8178 from profiling values of expressions for usage in optimizations.
8180 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8184 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8185 a code to gather information about values of expressions.
8187 With @option{-fbranch-probabilities}, it reads back the data gathered
8188 and actually performs the optimizations based on them.
8189 Currently the optimizations include specialization of division operation
8190 using the knowledge about the value of the denominator.
8192 @item -frename-registers
8193 @opindex frename-registers
8194 Attempt to avoid false dependencies in scheduled code by making use
8195 of registers left over after register allocation. This optimization
8196 will most benefit processors with lots of registers. Depending on the
8197 debug information format adopted by the target, however, it can
8198 make debugging impossible, since variables will no longer stay in
8199 a ``home register''.
8201 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8205 Perform tail duplication to enlarge superblock size. This transformation
8206 simplifies the control flow of the function allowing other optimizations to do
8209 Enabled with @option{-fprofile-use}.
8211 @item -funroll-loops
8212 @opindex funroll-loops
8213 Unroll loops whose number of iterations can be determined at compile time or
8214 upon entry to the loop. @option{-funroll-loops} implies
8215 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8216 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8217 small constant number of iterations). This option makes code larger, and may
8218 or may not make it run faster.
8220 Enabled with @option{-fprofile-use}.
8222 @item -funroll-all-loops
8223 @opindex funroll-all-loops
8224 Unroll all loops, even if their number of iterations is uncertain when
8225 the loop is entered. This usually makes programs run more slowly.
8226 @option{-funroll-all-loops} implies the same options as
8227 @option{-funroll-loops}.
8230 @opindex fpeel-loops
8231 Peels the loops for that there is enough information that they do not
8232 roll much (from profile feedback). It also turns on complete loop peeling
8233 (i.e.@: complete removal of loops with small constant number of iterations).
8235 Enabled with @option{-fprofile-use}.
8237 @item -fmove-loop-invariants
8238 @opindex fmove-loop-invariants
8239 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8240 at level @option{-O1}
8242 @item -funswitch-loops
8243 @opindex funswitch-loops
8244 Move branches with loop invariant conditions out of the loop, with duplicates
8245 of the loop on both branches (modified according to result of the condition).
8247 @item -ffunction-sections
8248 @itemx -fdata-sections
8249 @opindex ffunction-sections
8250 @opindex fdata-sections
8251 Place each function or data item into its own section in the output
8252 file if the target supports arbitrary sections. The name of the
8253 function or the name of the data item determines the section's name
8256 Use these options on systems where the linker can perform optimizations
8257 to improve locality of reference in the instruction space. Most systems
8258 using the ELF object format and SPARC processors running Solaris 2 have
8259 linkers with such optimizations. AIX may have these optimizations in
8262 Only use these options when there are significant benefits from doing
8263 so. When you specify these options, the assembler and linker will
8264 create larger object and executable files and will also be slower.
8265 You will not be able to use @code{gprof} on all systems if you
8266 specify this option and you may have problems with debugging if
8267 you specify both this option and @option{-g}.
8269 @item -fbranch-target-load-optimize
8270 @opindex fbranch-target-load-optimize
8271 Perform branch target register load optimization before prologue / epilogue
8273 The use of target registers can typically be exposed only during reload,
8274 thus hoisting loads out of loops and doing inter-block scheduling needs
8275 a separate optimization pass.
8277 @item -fbranch-target-load-optimize2
8278 @opindex fbranch-target-load-optimize2
8279 Perform branch target register load optimization after prologue / epilogue
8282 @item -fbtr-bb-exclusive
8283 @opindex fbtr-bb-exclusive
8284 When performing branch target register load optimization, don't reuse
8285 branch target registers in within any basic block.
8287 @item -fstack-protector
8288 @opindex fstack-protector
8289 Emit extra code to check for buffer overflows, such as stack smashing
8290 attacks. This is done by adding a guard variable to functions with
8291 vulnerable objects. This includes functions that call alloca, and
8292 functions with buffers larger than 8 bytes. The guards are initialized
8293 when a function is entered and then checked when the function exits.
8294 If a guard check fails, an error message is printed and the program exits.
8296 @item -fstack-protector-all
8297 @opindex fstack-protector-all
8298 Like @option{-fstack-protector} except that all functions are protected.
8300 @item -fsection-anchors
8301 @opindex fsection-anchors
8302 Try to reduce the number of symbolic address calculations by using
8303 shared ``anchor'' symbols to address nearby objects. This transformation
8304 can help to reduce the number of GOT entries and GOT accesses on some
8307 For example, the implementation of the following function @code{foo}:
8311 int foo (void) @{ return a + b + c; @}
8314 would usually calculate the addresses of all three variables, but if you
8315 compile it with @option{-fsection-anchors}, it will access the variables
8316 from a common anchor point instead. The effect is similar to the
8317 following pseudocode (which isn't valid C):
8322 register int *xr = &x;
8323 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8327 Not all targets support this option.
8329 @item --param @var{name}=@var{value}
8331 In some places, GCC uses various constants to control the amount of
8332 optimization that is done. For example, GCC will not inline functions
8333 that contain more that a certain number of instructions. You can
8334 control some of these constants on the command-line using the
8335 @option{--param} option.
8337 The names of specific parameters, and the meaning of the values, are
8338 tied to the internals of the compiler, and are subject to change
8339 without notice in future releases.
8341 In each case, the @var{value} is an integer. The allowable choices for
8342 @var{name} are given in the following table:
8345 @item predictable-branch-outcome
8346 When branch is predicted to be taken with probability lower than this threshold
8347 (in percent), then it is considered well predictable. The default is 10.
8349 @item max-crossjump-edges
8350 The maximum number of incoming edges to consider for crossjumping.
8351 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8352 the number of edges incoming to each block. Increasing values mean
8353 more aggressive optimization, making the compile time increase with
8354 probably small improvement in executable size.
8356 @item min-crossjump-insns
8357 The minimum number of instructions which must be matched at the end
8358 of two blocks before crossjumping will be performed on them. This
8359 value is ignored in the case where all instructions in the block being
8360 crossjumped from are matched. The default value is 5.
8362 @item max-grow-copy-bb-insns
8363 The maximum code size expansion factor when copying basic blocks
8364 instead of jumping. The expansion is relative to a jump instruction.
8365 The default value is 8.
8367 @item max-goto-duplication-insns
8368 The maximum number of instructions to duplicate to a block that jumps
8369 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8370 passes, GCC factors computed gotos early in the compilation process,
8371 and unfactors them as late as possible. Only computed jumps at the
8372 end of a basic blocks with no more than max-goto-duplication-insns are
8373 unfactored. The default value is 8.
8375 @item max-delay-slot-insn-search
8376 The maximum number of instructions to consider when looking for an
8377 instruction to fill a delay slot. If more than this arbitrary number of
8378 instructions is searched, the time savings from filling the delay slot
8379 will be minimal so stop searching. Increasing values mean more
8380 aggressive optimization, making the compile time increase with probably
8381 small improvement in executable run time.
8383 @item max-delay-slot-live-search
8384 When trying to fill delay slots, the maximum number of instructions to
8385 consider when searching for a block with valid live register
8386 information. Increasing this arbitrarily chosen value means more
8387 aggressive optimization, increasing the compile time. This parameter
8388 should be removed when the delay slot code is rewritten to maintain the
8391 @item max-gcse-memory
8392 The approximate maximum amount of memory that will be allocated in
8393 order to perform the global common subexpression elimination
8394 optimization. If more memory than specified is required, the
8395 optimization will not be done.
8397 @item max-gcse-insertion-ratio
8398 If the ratio of expression insertions to deletions is larger than this value
8399 for any expression, then RTL PRE will insert or remove the expression and thus
8400 leave partially redundant computations in the instruction stream. The default value is 20.
8402 @item max-pending-list-length
8403 The maximum number of pending dependencies scheduling will allow
8404 before flushing the current state and starting over. Large functions
8405 with few branches or calls can create excessively large lists which
8406 needlessly consume memory and resources.
8408 @item max-inline-insns-single
8409 Several parameters control the tree inliner used in gcc.
8410 This number sets the maximum number of instructions (counted in GCC's
8411 internal representation) in a single function that the tree inliner
8412 will consider for inlining. This only affects functions declared
8413 inline and methods implemented in a class declaration (C++).
8414 The default value is 400.
8416 @item max-inline-insns-auto
8417 When you use @option{-finline-functions} (included in @option{-O3}),
8418 a lot of functions that would otherwise not be considered for inlining
8419 by the compiler will be investigated. To those functions, a different
8420 (more restrictive) limit compared to functions declared inline can
8422 The default value is 40.
8424 @item large-function-insns
8425 The limit specifying really large functions. For functions larger than this
8426 limit after inlining, inlining is constrained by
8427 @option{--param large-function-growth}. This parameter is useful primarily
8428 to avoid extreme compilation time caused by non-linear algorithms used by the
8430 The default value is 2700.
8432 @item large-function-growth
8433 Specifies maximal growth of large function caused by inlining in percents.
8434 The default value is 100 which limits large function growth to 2.0 times
8437 @item large-unit-insns
8438 The limit specifying large translation unit. Growth caused by inlining of
8439 units larger than this limit is limited by @option{--param inline-unit-growth}.
8440 For small units this might be too tight (consider unit consisting of function A
8441 that is inline and B that just calls A three time. If B is small relative to
8442 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8443 large units consisting of small inlineable functions however the overall unit
8444 growth limit is needed to avoid exponential explosion of code size. Thus for
8445 smaller units, the size is increased to @option{--param large-unit-insns}
8446 before applying @option{--param inline-unit-growth}. The default is 10000
8448 @item inline-unit-growth
8449 Specifies maximal overall growth of the compilation unit caused by inlining.
8450 The default value is 30 which limits unit growth to 1.3 times the original
8453 @item ipcp-unit-growth
8454 Specifies maximal overall growth of the compilation unit caused by
8455 interprocedural constant propagation. The default value is 10 which limits
8456 unit growth to 1.1 times the original size.
8458 @item large-stack-frame
8459 The limit specifying large stack frames. While inlining the algorithm is trying
8460 to not grow past this limit too much. Default value is 256 bytes.
8462 @item large-stack-frame-growth
8463 Specifies maximal growth of large stack frames caused by inlining in percents.
8464 The default value is 1000 which limits large stack frame growth to 11 times
8467 @item max-inline-insns-recursive
8468 @itemx max-inline-insns-recursive-auto
8469 Specifies maximum number of instructions out-of-line copy of self recursive inline
8470 function can grow into by performing recursive inlining.
8472 For functions declared inline @option{--param max-inline-insns-recursive} is
8473 taken into account. For function not declared inline, recursive inlining
8474 happens only when @option{-finline-functions} (included in @option{-O3}) is
8475 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8476 default value is 450.
8478 @item max-inline-recursive-depth
8479 @itemx max-inline-recursive-depth-auto
8480 Specifies maximum recursion depth used by the recursive inlining.
8482 For functions declared inline @option{--param max-inline-recursive-depth} is
8483 taken into account. For function not declared inline, recursive inlining
8484 happens only when @option{-finline-functions} (included in @option{-O3}) is
8485 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8488 @item min-inline-recursive-probability
8489 Recursive inlining is profitable only for function having deep recursion
8490 in average and can hurt for function having little recursion depth by
8491 increasing the prologue size or complexity of function body to other
8494 When profile feedback is available (see @option{-fprofile-generate}) the actual
8495 recursion depth can be guessed from probability that function will recurse via
8496 given call expression. This parameter limits inlining only to call expression
8497 whose probability exceeds given threshold (in percents). The default value is
8500 @item early-inlining-insns
8501 Specify growth that early inliner can make. In effect it increases amount of
8502 inlining for code having large abstraction penalty. The default value is 10.
8504 @item max-early-inliner-iterations
8505 @itemx max-early-inliner-iterations
8506 Limit of iterations of early inliner. This basically bounds number of nested
8507 indirect calls early inliner can resolve. Deeper chains are still handled by
8510 @item comdat-sharing-probability
8511 @itemx comdat-sharing-probability
8512 Probability (in percent) that C++ inline function with comdat visibility
8513 will be shared across multiple compilation units. The default value is 20.
8515 @item min-vect-loop-bound
8516 The minimum number of iterations under which a loop will not get vectorized
8517 when @option{-ftree-vectorize} is used. The number of iterations after
8518 vectorization needs to be greater than the value specified by this option
8519 to allow vectorization. The default value is 0.
8521 @item gcse-cost-distance-ratio
8522 Scaling factor in calculation of maximum distance an expression
8523 can be moved by GCSE optimizations. This is currently supported only in the
8524 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8525 will be with simple expressions, i.e., the expressions which have cost
8526 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8527 hoisting of simple expressions. The default value is 10.
8529 @item gcse-unrestricted-cost
8530 Cost, roughly measured as the cost of a single typical machine
8531 instruction, at which GCSE optimizations will not constrain
8532 the distance an expression can travel. This is currently
8533 supported only in the code hoisting pass. The lesser the cost,
8534 the more aggressive code hoisting will be. Specifying 0 will
8535 allow all expressions to travel unrestricted distances.
8536 The default value is 3.
8538 @item max-hoist-depth
8539 The depth of search in the dominator tree for expressions to hoist.
8540 This is used to avoid quadratic behavior in hoisting algorithm.
8541 The value of 0 will avoid limiting the search, but may slow down compilation
8542 of huge functions. The default value is 30.
8544 @item max-unrolled-insns
8545 The maximum number of instructions that a loop should have if that loop
8546 is unrolled, and if the loop is unrolled, it determines how many times
8547 the loop code is unrolled.
8549 @item max-average-unrolled-insns
8550 The maximum number of instructions biased by probabilities of their execution
8551 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8552 it determines how many times the loop code is unrolled.
8554 @item max-unroll-times
8555 The maximum number of unrollings of a single loop.
8557 @item max-peeled-insns
8558 The maximum number of instructions that a loop should have if that loop
8559 is peeled, and if the loop is peeled, it determines how many times
8560 the loop code is peeled.
8562 @item max-peel-times
8563 The maximum number of peelings of a single loop.
8565 @item max-completely-peeled-insns
8566 The maximum number of insns of a completely peeled loop.
8568 @item max-completely-peel-times
8569 The maximum number of iterations of a loop to be suitable for complete peeling.
8571 @item max-completely-peel-loop-nest-depth
8572 The maximum depth of a loop nest suitable for complete peeling.
8574 @item max-unswitch-insns
8575 The maximum number of insns of an unswitched loop.
8577 @item max-unswitch-level
8578 The maximum number of branches unswitched in a single loop.
8581 The minimum cost of an expensive expression in the loop invariant motion.
8583 @item iv-consider-all-candidates-bound
8584 Bound on number of candidates for induction variables below that
8585 all candidates are considered for each use in induction variable
8586 optimizations. Only the most relevant candidates are considered
8587 if there are more candidates, to avoid quadratic time complexity.
8589 @item iv-max-considered-uses
8590 The induction variable optimizations give up on loops that contain more
8591 induction variable uses.
8593 @item iv-always-prune-cand-set-bound
8594 If number of candidates in the set is smaller than this value,
8595 we always try to remove unnecessary ivs from the set during its
8596 optimization when a new iv is added to the set.
8598 @item scev-max-expr-size
8599 Bound on size of expressions used in the scalar evolutions analyzer.
8600 Large expressions slow the analyzer.
8602 @item scev-max-expr-complexity
8603 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8604 Complex expressions slow the analyzer.
8606 @item omega-max-vars
8607 The maximum number of variables in an Omega constraint system.
8608 The default value is 128.
8610 @item omega-max-geqs
8611 The maximum number of inequalities in an Omega constraint system.
8612 The default value is 256.
8615 The maximum number of equalities in an Omega constraint system.
8616 The default value is 128.
8618 @item omega-max-wild-cards
8619 The maximum number of wildcard variables that the Omega solver will
8620 be able to insert. The default value is 18.
8622 @item omega-hash-table-size
8623 The size of the hash table in the Omega solver. The default value is
8626 @item omega-max-keys
8627 The maximal number of keys used by the Omega solver. The default
8630 @item omega-eliminate-redundant-constraints
8631 When set to 1, use expensive methods to eliminate all redundant
8632 constraints. The default value is 0.
8634 @item vect-max-version-for-alignment-checks
8635 The maximum number of runtime checks that can be performed when
8636 doing loop versioning for alignment in the vectorizer. See option
8637 ftree-vect-loop-version for more information.
8639 @item vect-max-version-for-alias-checks
8640 The maximum number of runtime checks that can be performed when
8641 doing loop versioning for alias in the vectorizer. See option
8642 ftree-vect-loop-version for more information.
8644 @item max-iterations-to-track
8646 The maximum number of iterations of a loop the brute force algorithm
8647 for analysis of # of iterations of the loop tries to evaluate.
8649 @item hot-bb-count-fraction
8650 Select fraction of the maximal count of repetitions of basic block in program
8651 given basic block needs to have to be considered hot.
8653 @item hot-bb-frequency-fraction
8654 Select fraction of the entry block frequency of executions of basic block in
8655 function given basic block needs to have to be considered hot.
8657 @item max-predicted-iterations
8658 The maximum number of loop iterations we predict statically. This is useful
8659 in cases where function contain single loop with known bound and other loop
8660 with unknown. We predict the known number of iterations correctly, while
8661 the unknown number of iterations average to roughly 10. This means that the
8662 loop without bounds would appear artificially cold relative to the other one.
8664 @item align-threshold
8666 Select fraction of the maximal frequency of executions of basic block in
8667 function given basic block will get aligned.
8669 @item align-loop-iterations
8671 A loop expected to iterate at lest the selected number of iterations will get
8674 @item tracer-dynamic-coverage
8675 @itemx tracer-dynamic-coverage-feedback
8677 This value is used to limit superblock formation once the given percentage of
8678 executed instructions is covered. This limits unnecessary code size
8681 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8682 feedback is available. The real profiles (as opposed to statically estimated
8683 ones) are much less balanced allowing the threshold to be larger value.
8685 @item tracer-max-code-growth
8686 Stop tail duplication once code growth has reached given percentage. This is
8687 rather hokey argument, as most of the duplicates will be eliminated later in
8688 cross jumping, so it may be set to much higher values than is the desired code
8691 @item tracer-min-branch-ratio
8693 Stop reverse growth when the reverse probability of best edge is less than this
8694 threshold (in percent).
8696 @item tracer-min-branch-ratio
8697 @itemx tracer-min-branch-ratio-feedback
8699 Stop forward growth if the best edge do have probability lower than this
8702 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8703 compilation for profile feedback and one for compilation without. The value
8704 for compilation with profile feedback needs to be more conservative (higher) in
8705 order to make tracer effective.
8707 @item max-cse-path-length
8709 Maximum number of basic blocks on path that cse considers. The default is 10.
8712 The maximum instructions CSE process before flushing. The default is 1000.
8714 @item ggc-min-expand
8716 GCC uses a garbage collector to manage its own memory allocation. This
8717 parameter specifies the minimum percentage by which the garbage
8718 collector's heap should be allowed to expand between collections.
8719 Tuning this may improve compilation speed; it has no effect on code
8722 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8723 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8724 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8725 GCC is not able to calculate RAM on a particular platform, the lower
8726 bound of 30% is used. Setting this parameter and
8727 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8728 every opportunity. This is extremely slow, but can be useful for
8731 @item ggc-min-heapsize
8733 Minimum size of the garbage collector's heap before it begins bothering
8734 to collect garbage. The first collection occurs after the heap expands
8735 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8736 tuning this may improve compilation speed, and has no effect on code
8739 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8740 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8741 with a lower bound of 4096 (four megabytes) and an upper bound of
8742 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8743 particular platform, the lower bound is used. Setting this parameter
8744 very large effectively disables garbage collection. Setting this
8745 parameter and @option{ggc-min-expand} to zero causes a full collection
8746 to occur at every opportunity.
8748 @item max-reload-search-insns
8749 The maximum number of instruction reload should look backward for equivalent
8750 register. Increasing values mean more aggressive optimization, making the
8751 compile time increase with probably slightly better performance. The default
8754 @item max-cselib-memory-locations
8755 The maximum number of memory locations cselib should take into account.
8756 Increasing values mean more aggressive optimization, making the compile time
8757 increase with probably slightly better performance. The default value is 500.
8759 @item reorder-blocks-duplicate
8760 @itemx reorder-blocks-duplicate-feedback
8762 Used by basic block reordering pass to decide whether to use unconditional
8763 branch or duplicate the code on its destination. Code is duplicated when its
8764 estimated size is smaller than this value multiplied by the estimated size of
8765 unconditional jump in the hot spots of the program.
8767 The @option{reorder-block-duplicate-feedback} is used only when profile
8768 feedback is available and may be set to higher values than
8769 @option{reorder-block-duplicate} since information about the hot spots is more
8772 @item max-sched-ready-insns
8773 The maximum number of instructions ready to be issued the scheduler should
8774 consider at any given time during the first scheduling pass. Increasing
8775 values mean more thorough searches, making the compilation time increase
8776 with probably little benefit. The default value is 100.
8778 @item max-sched-region-blocks
8779 The maximum number of blocks in a region to be considered for
8780 interblock scheduling. The default value is 10.
8782 @item max-pipeline-region-blocks
8783 The maximum number of blocks in a region to be considered for
8784 pipelining in the selective scheduler. The default value is 15.
8786 @item max-sched-region-insns
8787 The maximum number of insns in a region to be considered for
8788 interblock scheduling. The default value is 100.
8790 @item max-pipeline-region-insns
8791 The maximum number of insns in a region to be considered for
8792 pipelining in the selective scheduler. The default value is 200.
8795 The minimum probability (in percents) of reaching a source block
8796 for interblock speculative scheduling. The default value is 40.
8798 @item max-sched-extend-regions-iters
8799 The maximum number of iterations through CFG to extend regions.
8800 0 - disable region extension,
8801 N - do at most N iterations.
8802 The default value is 0.
8804 @item max-sched-insn-conflict-delay
8805 The maximum conflict delay for an insn to be considered for speculative motion.
8806 The default value is 3.
8808 @item sched-spec-prob-cutoff
8809 The minimal probability of speculation success (in percents), so that
8810 speculative insn will be scheduled.
8811 The default value is 40.
8813 @item sched-mem-true-dep-cost
8814 Minimal distance (in CPU cycles) between store and load targeting same
8815 memory locations. The default value is 1.
8817 @item selsched-max-lookahead
8818 The maximum size of the lookahead window of selective scheduling. It is a
8819 depth of search for available instructions.
8820 The default value is 50.
8822 @item selsched-max-sched-times
8823 The maximum number of times that an instruction will be scheduled during
8824 selective scheduling. This is the limit on the number of iterations
8825 through which the instruction may be pipelined. The default value is 2.
8827 @item selsched-max-insns-to-rename
8828 The maximum number of best instructions in the ready list that are considered
8829 for renaming in the selective scheduler. The default value is 2.
8832 The minimum value of stage count that swing modulo scheduler will
8833 generate. The default value is 2.
8835 @item max-last-value-rtl
8836 The maximum size measured as number of RTLs that can be recorded in an expression
8837 in combiner for a pseudo register as last known value of that register. The default
8840 @item integer-share-limit
8841 Small integer constants can use a shared data structure, reducing the
8842 compiler's memory usage and increasing its speed. This sets the maximum
8843 value of a shared integer constant. The default value is 256.
8845 @item min-virtual-mappings
8846 Specifies the minimum number of virtual mappings in the incremental
8847 SSA updater that should be registered to trigger the virtual mappings
8848 heuristic defined by virtual-mappings-ratio. The default value is
8851 @item virtual-mappings-ratio
8852 If the number of virtual mappings is virtual-mappings-ratio bigger
8853 than the number of virtual symbols to be updated, then the incremental
8854 SSA updater switches to a full update for those symbols. The default
8857 @item ssp-buffer-size
8858 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8859 protection when @option{-fstack-protection} is used.
8861 @item max-jump-thread-duplication-stmts
8862 Maximum number of statements allowed in a block that needs to be
8863 duplicated when threading jumps.
8865 @item max-fields-for-field-sensitive
8866 Maximum number of fields in a structure we will treat in
8867 a field sensitive manner during pointer analysis. The default is zero
8868 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8870 @item prefetch-latency
8871 Estimate on average number of instructions that are executed before
8872 prefetch finishes. The distance we prefetch ahead is proportional
8873 to this constant. Increasing this number may also lead to less
8874 streams being prefetched (see @option{simultaneous-prefetches}).
8876 @item simultaneous-prefetches
8877 Maximum number of prefetches that can run at the same time.
8879 @item l1-cache-line-size
8880 The size of cache line in L1 cache, in bytes.
8883 The size of L1 cache, in kilobytes.
8886 The size of L2 cache, in kilobytes.
8888 @item min-insn-to-prefetch-ratio
8889 The minimum ratio between the number of instructions and the
8890 number of prefetches to enable prefetching in a loop.
8892 @item prefetch-min-insn-to-mem-ratio
8893 The minimum ratio between the number of instructions and the
8894 number of memory references to enable prefetching in a loop.
8896 @item use-canonical-types
8897 Whether the compiler should use the ``canonical'' type system. By
8898 default, this should always be 1, which uses a more efficient internal
8899 mechanism for comparing types in C++ and Objective-C++. However, if
8900 bugs in the canonical type system are causing compilation failures,
8901 set this value to 0 to disable canonical types.
8903 @item switch-conversion-max-branch-ratio
8904 Switch initialization conversion will refuse to create arrays that are
8905 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8906 branches in the switch.
8908 @item max-partial-antic-length
8909 Maximum length of the partial antic set computed during the tree
8910 partial redundancy elimination optimization (@option{-ftree-pre}) when
8911 optimizing at @option{-O3} and above. For some sorts of source code
8912 the enhanced partial redundancy elimination optimization can run away,
8913 consuming all of the memory available on the host machine. This
8914 parameter sets a limit on the length of the sets that are computed,
8915 which prevents the runaway behavior. Setting a value of 0 for
8916 this parameter will allow an unlimited set length.
8918 @item sccvn-max-scc-size
8919 Maximum size of a strongly connected component (SCC) during SCCVN
8920 processing. If this limit is hit, SCCVN processing for the whole
8921 function will not be done and optimizations depending on it will
8922 be disabled. The default maximum SCC size is 10000.
8924 @item ira-max-loops-num
8925 IRA uses a regional register allocation by default. If a function
8926 contains loops more than number given by the parameter, only at most
8927 given number of the most frequently executed loops will form regions
8928 for the regional register allocation. The default value of the
8931 @item ira-max-conflict-table-size
8932 Although IRA uses a sophisticated algorithm of compression conflict
8933 table, the table can be still big for huge functions. If the conflict
8934 table for a function could be more than size in MB given by the
8935 parameter, the conflict table is not built and faster, simpler, and
8936 lower quality register allocation algorithm will be used. The
8937 algorithm do not use pseudo-register conflicts. The default value of
8938 the parameter is 2000.
8940 @item ira-loop-reserved-regs
8941 IRA can be used to evaluate more accurate register pressure in loops
8942 for decision to move loop invariants (see @option{-O3}). The number
8943 of available registers reserved for some other purposes is described
8944 by this parameter. The default value of the parameter is 2 which is
8945 minimal number of registers needed for execution of typical
8946 instruction. This value is the best found from numerous experiments.
8948 @item loop-invariant-max-bbs-in-loop
8949 Loop invariant motion can be very expensive, both in compile time and
8950 in amount of needed compile time memory, with very large loops. Loops
8951 with more basic blocks than this parameter won't have loop invariant
8952 motion optimization performed on them. The default value of the
8953 parameter is 1000 for -O1 and 10000 for -O2 and above.
8955 @item max-vartrack-size
8956 Sets a maximum number of hash table slots to use during variable
8957 tracking dataflow analysis of any function. If this limit is exceeded
8958 with variable tracking at assignments enabled, analysis for that
8959 function is retried without it, after removing all debug insns from
8960 the function. If the limit is exceeded even without debug insns, var
8961 tracking analysis is completely disabled for the function. Setting
8962 the parameter to zero makes it unlimited.
8964 @item max-vartrack-expr-depth
8965 Sets a maximum number of recursion levels when attempting to map
8966 variable names or debug temporaries to value expressions. This trades
8967 compile time for more complete debug information. If this is set too
8968 low, value expressions that are available and could be represented in
8969 debug information may end up not being used; setting this higher may
8970 enable the compiler to find more complex debug expressions, but compile
8971 time may grow exponentially, and even then, it may fail to find more
8972 usable expressions. The default is 10.
8974 @item min-nondebug-insn-uid
8975 Use uids starting at this parameter for nondebug insns. The range below
8976 the parameter is reserved exclusively for debug insns created by
8977 @option{-fvar-tracking-assignments}, but debug insns may get
8978 (non-overlapping) uids above it if the reserved range is exhausted.
8980 @item ipa-sra-ptr-growth-factor
8981 IPA-SRA will replace a pointer to an aggregate with one or more new
8982 parameters only when their cumulative size is less or equal to
8983 @option{ipa-sra-ptr-growth-factor} times the size of the original
8986 @item graphite-max-nb-scop-params
8987 To avoid exponential effects in the Graphite loop transforms, the
8988 number of parameters in a Static Control Part (SCoP) is bounded. The
8989 default value is 10 parameters. A variable whose value is unknown at
8990 compile time and defined outside a SCoP is a parameter of the SCoP.
8992 @item graphite-max-bbs-per-function
8993 To avoid exponential effects in the detection of SCoPs, the size of
8994 the functions analyzed by Graphite is bounded. The default value is
8997 @item loop-block-tile-size
8998 Loop blocking or strip mining transforms, enabled with
8999 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9000 loop in the loop nest by a given number of iterations. The strip
9001 length can be changed using the @option{loop-block-tile-size}
9002 parameter. The default value is 51 iterations.
9004 @item devirt-type-list-size
9005 IPA-CP attempts to track all possible types passed to a function's
9006 parameter in order to perform devirtualization.
9007 @option{devirt-type-list-size} is the maximum number of types it
9008 stores per a single formal parameter of a function.
9010 @item lto-partitions
9011 Specify desired number of partitions produced during WHOPR compilation.
9012 The number of partitions should exceed the number of CPUs used for compilation.
9013 The default value is 32.
9015 @item lto-minpartition
9016 Size of minimal partition for WHOPR (in estimated instructions).
9017 This prevents expenses of splitting very small programs into too many
9020 @item cxx-max-namespaces-for-diagnostic-help
9021 The maximum number of namespaces to consult for suggestions when C++
9022 name lookup fails for an identifier. The default is 1000.
9024 @item max-stores-to-sink
9025 The maximum number of conditional stores paires that can be sunk. Set to 0
9026 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9027 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9029 @item case-values-threshold
9030 The smallest number of different values for which it is best to use a
9031 jump-table instead of a tree of conditional branches. If the value is
9032 0, use the default for the machine. The default is 0.
9037 @node Preprocessor Options
9038 @section Options Controlling the Preprocessor
9039 @cindex preprocessor options
9040 @cindex options, preprocessor
9042 These options control the C preprocessor, which is run on each C source
9043 file before actual compilation.
9045 If you use the @option{-E} option, nothing is done except preprocessing.
9046 Some of these options make sense only together with @option{-E} because
9047 they cause the preprocessor output to be unsuitable for actual
9051 @item -Wp,@var{option}
9053 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9054 and pass @var{option} directly through to the preprocessor. If
9055 @var{option} contains commas, it is split into multiple options at the
9056 commas. However, many options are modified, translated or interpreted
9057 by the compiler driver before being passed to the preprocessor, and
9058 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9059 interface is undocumented and subject to change, so whenever possible
9060 you should avoid using @option{-Wp} and let the driver handle the
9063 @item -Xpreprocessor @var{option}
9064 @opindex Xpreprocessor
9065 Pass @var{option} as an option to the preprocessor. You can use this to
9066 supply system-specific preprocessor options which GCC does not know how to
9069 If you want to pass an option that takes an argument, you must use
9070 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9073 @include cppopts.texi
9075 @node Assembler Options
9076 @section Passing Options to the Assembler
9078 @c prevent bad page break with this line
9079 You can pass options to the assembler.
9082 @item -Wa,@var{option}
9084 Pass @var{option} as an option to the assembler. If @var{option}
9085 contains commas, it is split into multiple options at the commas.
9087 @item -Xassembler @var{option}
9089 Pass @var{option} as an option to the assembler. You can use this to
9090 supply system-specific assembler options which GCC does not know how to
9093 If you want to pass an option that takes an argument, you must use
9094 @option{-Xassembler} twice, once for the option and once for the argument.
9099 @section Options for Linking
9100 @cindex link options
9101 @cindex options, linking
9103 These options come into play when the compiler links object files into
9104 an executable output file. They are meaningless if the compiler is
9105 not doing a link step.
9109 @item @var{object-file-name}
9110 A file name that does not end in a special recognized suffix is
9111 considered to name an object file or library. (Object files are
9112 distinguished from libraries by the linker according to the file
9113 contents.) If linking is done, these object files are used as input
9122 If any of these options is used, then the linker is not run, and
9123 object file names should not be used as arguments. @xref{Overall
9127 @item -l@var{library}
9128 @itemx -l @var{library}
9130 Search the library named @var{library} when linking. (The second
9131 alternative with the library as a separate argument is only for
9132 POSIX compliance and is not recommended.)
9134 It makes a difference where in the command you write this option; the
9135 linker searches and processes libraries and object files in the order they
9136 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9137 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9138 to functions in @samp{z}, those functions may not be loaded.
9140 The linker searches a standard list of directories for the library,
9141 which is actually a file named @file{lib@var{library}.a}. The linker
9142 then uses this file as if it had been specified precisely by name.
9144 The directories searched include several standard system directories
9145 plus any that you specify with @option{-L}.
9147 Normally the files found this way are library files---archive files
9148 whose members are object files. The linker handles an archive file by
9149 scanning through it for members which define symbols that have so far
9150 been referenced but not defined. But if the file that is found is an
9151 ordinary object file, it is linked in the usual fashion. The only
9152 difference between using an @option{-l} option and specifying a file name
9153 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9154 and searches several directories.
9158 You need this special case of the @option{-l} option in order to
9159 link an Objective-C or Objective-C++ program.
9162 @opindex nostartfiles
9163 Do not use the standard system startup files when linking.
9164 The standard system libraries are used normally, unless @option{-nostdlib}
9165 or @option{-nodefaultlibs} is used.
9167 @item -nodefaultlibs
9168 @opindex nodefaultlibs
9169 Do not use the standard system libraries when linking.
9170 Only the libraries you specify will be passed to the linker, options
9171 specifying linkage of the system libraries, such as @code{-static-libgcc}
9172 or @code{-shared-libgcc}, will be ignored.
9173 The standard startup files are used normally, unless @option{-nostartfiles}
9174 is used. The compiler may generate calls to @code{memcmp},
9175 @code{memset}, @code{memcpy} and @code{memmove}.
9176 These entries are usually resolved by entries in
9177 libc. These entry points should be supplied through some other
9178 mechanism when this option is specified.
9182 Do not use the standard system startup files or libraries when linking.
9183 No startup files and only the libraries you specify will be passed to
9184 the linker, options specifying linkage of the system libraries, such as
9185 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9186 The compiler may generate calls to @code{memcmp}, @code{memset},
9187 @code{memcpy} and @code{memmove}.
9188 These entries are usually resolved by entries in
9189 libc. These entry points should be supplied through some other
9190 mechanism when this option is specified.
9192 @cindex @option{-lgcc}, use with @option{-nostdlib}
9193 @cindex @option{-nostdlib} and unresolved references
9194 @cindex unresolved references and @option{-nostdlib}
9195 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9196 @cindex @option{-nodefaultlibs} and unresolved references
9197 @cindex unresolved references and @option{-nodefaultlibs}
9198 One of the standard libraries bypassed by @option{-nostdlib} and
9199 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9200 that GCC uses to overcome shortcomings of particular machines, or special
9201 needs for some languages.
9202 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9203 Collection (GCC) Internals},
9204 for more discussion of @file{libgcc.a}.)
9205 In most cases, you need @file{libgcc.a} even when you want to avoid
9206 other standard libraries. In other words, when you specify @option{-nostdlib}
9207 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9208 This ensures that you have no unresolved references to internal GCC
9209 library subroutines. (For example, @samp{__main}, used to ensure C++
9210 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9211 GNU Compiler Collection (GCC) Internals}.)
9215 Produce a position independent executable on targets which support it.
9216 For predictable results, you must also specify the same set of options
9217 that were used to generate code (@option{-fpie}, @option{-fPIE},
9218 or model suboptions) when you specify this option.
9222 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9223 that support it. This instructs the linker to add all symbols, not
9224 only used ones, to the dynamic symbol table. This option is needed
9225 for some uses of @code{dlopen} or to allow obtaining backtraces
9226 from within a program.
9230 Remove all symbol table and relocation information from the executable.
9234 On systems that support dynamic linking, this prevents linking with the shared
9235 libraries. On other systems, this option has no effect.
9239 Produce a shared object which can then be linked with other objects to
9240 form an executable. Not all systems support this option. For predictable
9241 results, you must also specify the same set of options that were used to
9242 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9243 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9244 needs to build supplementary stub code for constructors to work. On
9245 multi-libbed systems, @samp{gcc -shared} must select the correct support
9246 libraries to link against. Failing to supply the correct flags may lead
9247 to subtle defects. Supplying them in cases where they are not necessary
9250 @item -shared-libgcc
9251 @itemx -static-libgcc
9252 @opindex shared-libgcc
9253 @opindex static-libgcc
9254 On systems that provide @file{libgcc} as a shared library, these options
9255 force the use of either the shared or static version respectively.
9256 If no shared version of @file{libgcc} was built when the compiler was
9257 configured, these options have no effect.
9259 There are several situations in which an application should use the
9260 shared @file{libgcc} instead of the static version. The most common
9261 of these is when the application wishes to throw and catch exceptions
9262 across different shared libraries. In that case, each of the libraries
9263 as well as the application itself should use the shared @file{libgcc}.
9265 Therefore, the G++ and GCJ drivers automatically add
9266 @option{-shared-libgcc} whenever you build a shared library or a main
9267 executable, because C++ and Java programs typically use exceptions, so
9268 this is the right thing to do.
9270 If, instead, you use the GCC driver to create shared libraries, you may
9271 find that they will not always be linked with the shared @file{libgcc}.
9272 If GCC finds, at its configuration time, that you have a non-GNU linker
9273 or a GNU linker that does not support option @option{--eh-frame-hdr},
9274 it will link the shared version of @file{libgcc} into shared libraries
9275 by default. Otherwise, it will take advantage of the linker and optimize
9276 away the linking with the shared version of @file{libgcc}, linking with
9277 the static version of libgcc by default. This allows exceptions to
9278 propagate through such shared libraries, without incurring relocation
9279 costs at library load time.
9281 However, if a library or main executable is supposed to throw or catch
9282 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9283 for the languages used in the program, or using the option
9284 @option{-shared-libgcc}, such that it is linked with the shared
9287 @item -static-libstdc++
9288 When the @command{g++} program is used to link a C++ program, it will
9289 normally automatically link against @option{libstdc++}. If
9290 @file{libstdc++} is available as a shared library, and the
9291 @option{-static} option is not used, then this will link against the
9292 shared version of @file{libstdc++}. That is normally fine. However, it
9293 is sometimes useful to freeze the version of @file{libstdc++} used by
9294 the program without going all the way to a fully static link. The
9295 @option{-static-libstdc++} option directs the @command{g++} driver to
9296 link @file{libstdc++} statically, without necessarily linking other
9297 libraries statically.
9301 Bind references to global symbols when building a shared object. Warn
9302 about any unresolved references (unless overridden by the link editor
9303 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9306 @item -T @var{script}
9308 @cindex linker script
9309 Use @var{script} as the linker script. This option is supported by most
9310 systems using the GNU linker. On some targets, such as bare-board
9311 targets without an operating system, the @option{-T} option may be required
9312 when linking to avoid references to undefined symbols.
9314 @item -Xlinker @var{option}
9316 Pass @var{option} as an option to the linker. You can use this to
9317 supply system-specific linker options which GCC does not know how to
9320 If you want to pass an option that takes a separate argument, you must use
9321 @option{-Xlinker} twice, once for the option and once for the argument.
9322 For example, to pass @option{-assert definitions}, you must write
9323 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9324 @option{-Xlinker "-assert definitions"}, because this passes the entire
9325 string as a single argument, which is not what the linker expects.
9327 When using the GNU linker, it is usually more convenient to pass
9328 arguments to linker options using the @option{@var{option}=@var{value}}
9329 syntax than as separate arguments. For example, you can specify
9330 @samp{-Xlinker -Map=output.map} rather than
9331 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9332 this syntax for command-line options.
9334 @item -Wl,@var{option}
9336 Pass @var{option} as an option to the linker. If @var{option} contains
9337 commas, it is split into multiple options at the commas. You can use this
9338 syntax to pass an argument to the option.
9339 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9340 linker. When using the GNU linker, you can also get the same effect with
9341 @samp{-Wl,-Map=output.map}.
9343 @item -u @var{symbol}
9345 Pretend the symbol @var{symbol} is undefined, to force linking of
9346 library modules to define it. You can use @option{-u} multiple times with
9347 different symbols to force loading of additional library modules.
9350 @node Directory Options
9351 @section Options for Directory Search
9352 @cindex directory options
9353 @cindex options, directory search
9356 These options specify directories to search for header files, for
9357 libraries and for parts of the compiler:
9362 Add the directory @var{dir} to the head of the list of directories to be
9363 searched for header files. This can be used to override a system header
9364 file, substituting your own version, since these directories are
9365 searched before the system header file directories. However, you should
9366 not use this option to add directories that contain vendor-supplied
9367 system header files (use @option{-isystem} for that). If you use more than
9368 one @option{-I} option, the directories are scanned in left-to-right
9369 order; the standard system directories come after.
9371 If a standard system include directory, or a directory specified with
9372 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9373 option will be ignored. The directory will still be searched but as a
9374 system directory at its normal position in the system include chain.
9375 This is to ensure that GCC's procedure to fix buggy system headers and
9376 the ordering for the include_next directive are not inadvertently changed.
9377 If you really need to change the search order for system directories,
9378 use the @option{-nostdinc} and/or @option{-isystem} options.
9380 @item -iplugindir=@var{dir}
9381 Set the directory to search for plugins which are passed
9382 by @option{-fplugin=@var{name}} instead of
9383 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9384 to be used by the user, but only passed by the driver.
9386 @item -iquote@var{dir}
9388 Add the directory @var{dir} to the head of the list of directories to
9389 be searched for header files only for the case of @samp{#include
9390 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9391 otherwise just like @option{-I}.
9395 Add directory @var{dir} to the list of directories to be searched
9398 @item -B@var{prefix}
9400 This option specifies where to find the executables, libraries,
9401 include files, and data files of the compiler itself.
9403 The compiler driver program runs one or more of the subprograms
9404 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9405 @var{prefix} as a prefix for each program it tries to run, both with and
9406 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9408 For each subprogram to be run, the compiler driver first tries the
9409 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9410 was not specified, the driver tries two standard prefixes, which are
9411 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9412 those results in a file name that is found, the unmodified program
9413 name is searched for using the directories specified in your
9414 @env{PATH} environment variable.
9416 The compiler will check to see if the path provided by the @option{-B}
9417 refers to a directory, and if necessary it will add a directory
9418 separator character at the end of the path.
9420 @option{-B} prefixes that effectively specify directory names also apply
9421 to libraries in the linker, because the compiler translates these
9422 options into @option{-L} options for the linker. They also apply to
9423 includes files in the preprocessor, because the compiler translates these
9424 options into @option{-isystem} options for the preprocessor. In this case,
9425 the compiler appends @samp{include} to the prefix.
9427 The run-time support file @file{libgcc.a} can also be searched for using
9428 the @option{-B} prefix, if needed. If it is not found there, the two
9429 standard prefixes above are tried, and that is all. The file is left
9430 out of the link if it is not found by those means.
9432 Another way to specify a prefix much like the @option{-B} prefix is to use
9433 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9436 As a special kludge, if the path provided by @option{-B} is
9437 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9438 9, then it will be replaced by @file{[dir/]include}. This is to help
9439 with boot-strapping the compiler.
9441 @item -specs=@var{file}
9443 Process @var{file} after the compiler reads in the standard @file{specs}
9444 file, in order to override the defaults that the @file{gcc} driver
9445 program uses when determining what switches to pass to @file{cc1},
9446 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9447 @option{-specs=@var{file}} can be specified on the command line, and they
9448 are processed in order, from left to right.
9450 @item --sysroot=@var{dir}
9452 Use @var{dir} as the logical root directory for headers and libraries.
9453 For example, if the compiler would normally search for headers in
9454 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9455 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9457 If you use both this option and the @option{-isysroot} option, then
9458 the @option{--sysroot} option will apply to libraries, but the
9459 @option{-isysroot} option will apply to header files.
9461 The GNU linker (beginning with version 2.16) has the necessary support
9462 for this option. If your linker does not support this option, the
9463 header file aspect of @option{--sysroot} will still work, but the
9464 library aspect will not.
9468 This option has been deprecated. Please use @option{-iquote} instead for
9469 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9470 Any directories you specify with @option{-I} options before the @option{-I-}
9471 option are searched only for the case of @samp{#include "@var{file}"};
9472 they are not searched for @samp{#include <@var{file}>}.
9474 If additional directories are specified with @option{-I} options after
9475 the @option{-I-}, these directories are searched for all @samp{#include}
9476 directives. (Ordinarily @emph{all} @option{-I} directories are used
9479 In addition, the @option{-I-} option inhibits the use of the current
9480 directory (where the current input file came from) as the first search
9481 directory for @samp{#include "@var{file}"}. There is no way to
9482 override this effect of @option{-I-}. With @option{-I.} you can specify
9483 searching the directory which was current when the compiler was
9484 invoked. That is not exactly the same as what the preprocessor does
9485 by default, but it is often satisfactory.
9487 @option{-I-} does not inhibit the use of the standard system directories
9488 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9495 @section Specifying subprocesses and the switches to pass to them
9498 @command{gcc} is a driver program. It performs its job by invoking a
9499 sequence of other programs to do the work of compiling, assembling and
9500 linking. GCC interprets its command-line parameters and uses these to
9501 deduce which programs it should invoke, and which command-line options
9502 it ought to place on their command lines. This behavior is controlled
9503 by @dfn{spec strings}. In most cases there is one spec string for each
9504 program that GCC can invoke, but a few programs have multiple spec
9505 strings to control their behavior. The spec strings built into GCC can
9506 be overridden by using the @option{-specs=} command-line switch to specify
9509 @dfn{Spec files} are plaintext files that are used to construct spec
9510 strings. They consist of a sequence of directives separated by blank
9511 lines. The type of directive is determined by the first non-whitespace
9512 character on the line and it can be one of the following:
9515 @item %@var{command}
9516 Issues a @var{command} to the spec file processor. The commands that can
9520 @item %include <@var{file}>
9521 @cindex @code{%include}
9522 Search for @var{file} and insert its text at the current point in the
9525 @item %include_noerr <@var{file}>
9526 @cindex @code{%include_noerr}
9527 Just like @samp{%include}, but do not generate an error message if the include
9528 file cannot be found.
9530 @item %rename @var{old_name} @var{new_name}
9531 @cindex @code{%rename}
9532 Rename the spec string @var{old_name} to @var{new_name}.
9536 @item *[@var{spec_name}]:
9537 This tells the compiler to create, override or delete the named spec
9538 string. All lines after this directive up to the next directive or
9539 blank line are considered to be the text for the spec string. If this
9540 results in an empty string then the spec will be deleted. (Or, if the
9541 spec did not exist, then nothing will happen.) Otherwise, if the spec
9542 does not currently exist a new spec will be created. If the spec does
9543 exist then its contents will be overridden by the text of this
9544 directive, unless the first character of that text is the @samp{+}
9545 character, in which case the text will be appended to the spec.
9547 @item [@var{suffix}]:
9548 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9549 and up to the next directive or blank line are considered to make up the
9550 spec string for the indicated suffix. When the compiler encounters an
9551 input file with the named suffix, it will processes the spec string in
9552 order to work out how to compile that file. For example:
9559 This says that any input file whose name ends in @samp{.ZZ} should be
9560 passed to the program @samp{z-compile}, which should be invoked with the
9561 command-line switch @option{-input} and with the result of performing the
9562 @samp{%i} substitution. (See below.)
9564 As an alternative to providing a spec string, the text that follows a
9565 suffix directive can be one of the following:
9568 @item @@@var{language}
9569 This says that the suffix is an alias for a known @var{language}. This is
9570 similar to using the @option{-x} command-line switch to GCC to specify a
9571 language explicitly. For example:
9578 Says that .ZZ files are, in fact, C++ source files.
9581 This causes an error messages saying:
9584 @var{name} compiler not installed on this system.
9588 GCC already has an extensive list of suffixes built into it.
9589 This directive will add an entry to the end of the list of suffixes, but
9590 since the list is searched from the end backwards, it is effectively
9591 possible to override earlier entries using this technique.
9595 GCC has the following spec strings built into it. Spec files can
9596 override these strings or create their own. Note that individual
9597 targets can also add their own spec strings to this list.
9600 asm Options to pass to the assembler
9601 asm_final Options to pass to the assembler post-processor
9602 cpp Options to pass to the C preprocessor
9603 cc1 Options to pass to the C compiler
9604 cc1plus Options to pass to the C++ compiler
9605 endfile Object files to include at the end of the link
9606 link Options to pass to the linker
9607 lib Libraries to include on the command line to the linker
9608 libgcc Decides which GCC support library to pass to the linker
9609 linker Sets the name of the linker
9610 predefines Defines to be passed to the C preprocessor
9611 signed_char Defines to pass to CPP to say whether @code{char} is signed
9613 startfile Object files to include at the start of the link
9616 Here is a small example of a spec file:
9622 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9625 This example renames the spec called @samp{lib} to @samp{old_lib} and
9626 then overrides the previous definition of @samp{lib} with a new one.
9627 The new definition adds in some extra command-line options before
9628 including the text of the old definition.
9630 @dfn{Spec strings} are a list of command-line options to be passed to their
9631 corresponding program. In addition, the spec strings can contain
9632 @samp{%}-prefixed sequences to substitute variable text or to
9633 conditionally insert text into the command line. Using these constructs
9634 it is possible to generate quite complex command lines.
9636 Here is a table of all defined @samp{%}-sequences for spec
9637 strings. Note that spaces are not generated automatically around the
9638 results of expanding these sequences. Therefore you can concatenate them
9639 together or combine them with constant text in a single argument.
9643 Substitute one @samp{%} into the program name or argument.
9646 Substitute the name of the input file being processed.
9649 Substitute the basename of the input file being processed.
9650 This is the substring up to (and not including) the last period
9651 and not including the directory.
9654 This is the same as @samp{%b}, but include the file suffix (text after
9658 Marks the argument containing or following the @samp{%d} as a
9659 temporary file name, so that that file will be deleted if GCC exits
9660 successfully. Unlike @samp{%g}, this contributes no text to the
9663 @item %g@var{suffix}
9664 Substitute a file name that has suffix @var{suffix} and is chosen
9665 once per compilation, and mark the argument in the same way as
9666 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9667 name is now chosen in a way that is hard to predict even when previously
9668 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9669 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9670 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9671 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9672 was simply substituted with a file name chosen once per compilation,
9673 without regard to any appended suffix (which was therefore treated
9674 just like ordinary text), making such attacks more likely to succeed.
9676 @item %u@var{suffix}
9677 Like @samp{%g}, but generates a new temporary file name even if
9678 @samp{%u@var{suffix}} was already seen.
9680 @item %U@var{suffix}
9681 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9682 new one if there is no such last file name. In the absence of any
9683 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9684 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9685 would involve the generation of two distinct file names, one
9686 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9687 simply substituted with a file name chosen for the previous @samp{%u},
9688 without regard to any appended suffix.
9690 @item %j@var{suffix}
9691 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9692 writable, and if save-temps is off; otherwise, substitute the name
9693 of a temporary file, just like @samp{%u}. This temporary file is not
9694 meant for communication between processes, but rather as a junk
9697 @item %|@var{suffix}
9698 @itemx %m@var{suffix}
9699 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9700 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9701 all. These are the two most common ways to instruct a program that it
9702 should read from standard input or write to standard output. If you
9703 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9704 construct: see for example @file{f/lang-specs.h}.
9706 @item %.@var{SUFFIX}
9707 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9708 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9709 terminated by the next space or %.
9712 Marks the argument containing or following the @samp{%w} as the
9713 designated output file of this compilation. This puts the argument
9714 into the sequence of arguments that @samp{%o} will substitute later.
9717 Substitutes the names of all the output files, with spaces
9718 automatically placed around them. You should write spaces
9719 around the @samp{%o} as well or the results are undefined.
9720 @samp{%o} is for use in the specs for running the linker.
9721 Input files whose names have no recognized suffix are not compiled
9722 at all, but they are included among the output files, so they will
9726 Substitutes the suffix for object files. Note that this is
9727 handled specially when it immediately follows @samp{%g, %u, or %U},
9728 because of the need for those to form complete file names. The
9729 handling is such that @samp{%O} is treated exactly as if it had already
9730 been substituted, except that @samp{%g, %u, and %U} do not currently
9731 support additional @var{suffix} characters following @samp{%O} as they would
9732 following, for example, @samp{.o}.
9735 Substitutes the standard macro predefinitions for the
9736 current target machine. Use this when running @code{cpp}.
9739 Like @samp{%p}, but puts @samp{__} before and after the name of each
9740 predefined macro, except for macros that start with @samp{__} or with
9741 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9745 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9746 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9747 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9748 and @option{-imultilib} as necessary.
9751 Current argument is the name of a library or startup file of some sort.
9752 Search for that file in a standard list of directories and substitute
9753 the full name found. The current working directory is included in the
9754 list of directories scanned.
9757 Current argument is the name of a linker script. Search for that file
9758 in the current list of directories to scan for libraries. If the file
9759 is located insert a @option{--script} option into the command line
9760 followed by the full path name found. If the file is not found then
9761 generate an error message. Note: the current working directory is not
9765 Print @var{str} as an error message. @var{str} is terminated by a newline.
9766 Use this when inconsistent options are detected.
9769 Substitute the contents of spec string @var{name} at this point.
9771 @item %x@{@var{option}@}
9772 Accumulate an option for @samp{%X}.
9775 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9779 Output the accumulated assembler options specified by @option{-Wa}.
9782 Output the accumulated preprocessor options specified by @option{-Wp}.
9785 Process the @code{asm} spec. This is used to compute the
9786 switches to be passed to the assembler.
9789 Process the @code{asm_final} spec. This is a spec string for
9790 passing switches to an assembler post-processor, if such a program is
9794 Process the @code{link} spec. This is the spec for computing the
9795 command line passed to the linker. Typically it will make use of the
9796 @samp{%L %G %S %D and %E} sequences.
9799 Dump out a @option{-L} option for each directory that GCC believes might
9800 contain startup files. If the target supports multilibs then the
9801 current multilib directory will be prepended to each of these paths.
9804 Process the @code{lib} spec. This is a spec string for deciding which
9805 libraries should be included on the command line to the linker.
9808 Process the @code{libgcc} spec. This is a spec string for deciding
9809 which GCC support library should be included on the command line to the linker.
9812 Process the @code{startfile} spec. This is a spec for deciding which
9813 object files should be the first ones passed to the linker. Typically
9814 this might be a file named @file{crt0.o}.
9817 Process the @code{endfile} spec. This is a spec string that specifies
9818 the last object files that will be passed to the linker.
9821 Process the @code{cpp} spec. This is used to construct the arguments
9822 to be passed to the C preprocessor.
9825 Process the @code{cc1} spec. This is used to construct the options to be
9826 passed to the actual C compiler (@samp{cc1}).
9829 Process the @code{cc1plus} spec. This is used to construct the options to be
9830 passed to the actual C++ compiler (@samp{cc1plus}).
9833 Substitute the variable part of a matched option. See below.
9834 Note that each comma in the substituted string is replaced by
9838 Remove all occurrences of @code{-S} from the command line. Note---this
9839 command is position dependent. @samp{%} commands in the spec string
9840 before this one will see @code{-S}, @samp{%} commands in the spec string
9841 after this one will not.
9843 @item %:@var{function}(@var{args})
9844 Call the named function @var{function}, passing it @var{args}.
9845 @var{args} is first processed as a nested spec string, then split
9846 into an argument vector in the usual fashion. The function returns
9847 a string which is processed as if it had appeared literally as part
9848 of the current spec.
9850 The following built-in spec functions are provided:
9854 The @code{getenv} spec function takes two arguments: an environment
9855 variable name and a string. If the environment variable is not
9856 defined, a fatal error is issued. Otherwise, the return value is the
9857 value of the environment variable concatenated with the string. For
9858 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9861 %:getenv(TOPDIR /include)
9864 expands to @file{/path/to/top/include}.
9866 @item @code{if-exists}
9867 The @code{if-exists} spec function takes one argument, an absolute
9868 pathname to a file. If the file exists, @code{if-exists} returns the
9869 pathname. Here is a small example of its usage:
9873 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9876 @item @code{if-exists-else}
9877 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9878 spec function, except that it takes two arguments. The first argument is
9879 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9880 returns the pathname. If it does not exist, it returns the second argument.
9881 This way, @code{if-exists-else} can be used to select one file or another,
9882 based on the existence of the first. Here is a small example of its usage:
9886 crt0%O%s %:if-exists(crti%O%s) \
9887 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9890 @item @code{replace-outfile}
9891 The @code{replace-outfile} spec function takes two arguments. It looks for the
9892 first argument in the outfiles array and replaces it with the second argument. Here
9893 is a small example of its usage:
9896 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9899 @item @code{remove-outfile}
9900 The @code{remove-outfile} spec function takes one argument. It looks for the
9901 first argument in the outfiles array and removes it. Here is a small example
9905 %:remove-outfile(-lm)
9908 @item @code{pass-through-libs}
9909 The @code{pass-through-libs} spec function takes any number of arguments. It
9910 finds any @option{-l} options and any non-options ending in ".a" (which it
9911 assumes are the names of linker input library archive files) and returns a
9912 result containing all the found arguments each prepended by
9913 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9914 intended to be passed to the LTO linker plugin.
9917 %:pass-through-libs(%G %L %G)
9920 @item @code{print-asm-header}
9921 The @code{print-asm-header} function takes no arguments and simply
9922 prints a banner like:
9928 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9931 It is used to separate compiler options from assembler options
9932 in the @option{--target-help} output.
9936 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9937 If that switch was not specified, this substitutes nothing. Note that
9938 the leading dash is omitted when specifying this option, and it is
9939 automatically inserted if the substitution is performed. Thus the spec
9940 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9941 and would output the command line option @option{-foo}.
9943 @item %W@{@code{S}@}
9944 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9947 @item %@{@code{S}*@}
9948 Substitutes all the switches specified to GCC whose names start
9949 with @code{-S}, but which also take an argument. This is used for
9950 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9951 GCC considers @option{-o foo} as being
9952 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9953 text, including the space. Thus two arguments would be generated.
9955 @item %@{@code{S}*&@code{T}*@}
9956 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
9957 (the order of @code{S} and @code{T} in the spec is not significant).
9958 There can be any number of ampersand-separated variables; for each the
9959 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
9961 @item %@{@code{S}:@code{X}@}
9962 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
9964 @item %@{!@code{S}:@code{X}@}
9965 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
9967 @item %@{@code{S}*:@code{X}@}
9968 Substitutes @code{X} if one or more switches whose names start with
9969 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
9970 once, no matter how many such switches appeared. However, if @code{%*}
9971 appears somewhere in @code{X}, then @code{X} will be substituted once
9972 for each matching switch, with the @code{%*} replaced by the part of
9973 that switch that matched the @code{*}.
9975 @item %@{.@code{S}:@code{X}@}
9976 Substitutes @code{X}, if processing a file with suffix @code{S}.
9978 @item %@{!.@code{S}:@code{X}@}
9979 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
9981 @item %@{,@code{S}:@code{X}@}
9982 Substitutes @code{X}, if processing a file for language @code{S}.
9984 @item %@{!,@code{S}:@code{X}@}
9985 Substitutes @code{X}, if not processing a file for language @code{S}.
9987 @item %@{@code{S}|@code{P}:@code{X}@}
9988 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
9989 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
9990 @code{*} sequences as well, although they have a stronger binding than
9991 the @samp{|}. If @code{%*} appears in @code{X}, all of the
9992 alternatives must be starred, and only the first matching alternative
9995 For example, a spec string like this:
9998 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10001 will output the following command-line options from the following input
10002 command-line options:
10007 -d fred.c -foo -baz -boggle
10008 -d jim.d -bar -baz -boggle
10011 @item %@{S:X; T:Y; :D@}
10013 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10014 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10015 be as many clauses as you need. This may be combined with @code{.},
10016 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10021 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10022 construct may contain other nested @samp{%} constructs or spaces, or
10023 even newlines. They are processed as usual, as described above.
10024 Trailing white space in @code{X} is ignored. White space may also
10025 appear anywhere on the left side of the colon in these constructs,
10026 except between @code{.} or @code{*} and the corresponding word.
10028 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10029 handled specifically in these constructs. If another value of
10030 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10031 @option{-W} switch is found later in the command line, the earlier
10032 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10033 just one letter, which passes all matching options.
10035 The character @samp{|} at the beginning of the predicate text is used to
10036 indicate that a command should be piped to the following command, but
10037 only if @option{-pipe} is specified.
10039 It is built into GCC which switches take arguments and which do not.
10040 (You might think it would be useful to generalize this to allow each
10041 compiler's spec to say which switches take arguments. But this cannot
10042 be done in a consistent fashion. GCC cannot even decide which input
10043 files have been specified without knowing which switches take arguments,
10044 and it must know which input files to compile in order to tell which
10047 GCC also knows implicitly that arguments starting in @option{-l} are to be
10048 treated as compiler output files, and passed to the linker in their
10049 proper position among the other output files.
10051 @c man begin OPTIONS
10053 @node Target Options
10054 @section Specifying Target Machine and Compiler Version
10055 @cindex target options
10056 @cindex cross compiling
10057 @cindex specifying machine version
10058 @cindex specifying compiler version and target machine
10059 @cindex compiler version, specifying
10060 @cindex target machine, specifying
10062 The usual way to run GCC is to run the executable called @command{gcc}, or
10063 @command{@var{machine}-gcc} when cross-compiling, or
10064 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10065 one that was installed last.
10067 @node Submodel Options
10068 @section Hardware Models and Configurations
10069 @cindex submodel options
10070 @cindex specifying hardware config
10071 @cindex hardware models and configurations, specifying
10072 @cindex machine dependent options
10074 Each target machine types can have its own
10075 special options, starting with @samp{-m}, to choose among various
10076 hardware models or configurations---for example, 68010 vs 68020,
10077 floating coprocessor or none. A single installed version of the
10078 compiler can compile for any model or configuration, according to the
10081 Some configurations of the compiler also support additional special
10082 options, usually for compatibility with other compilers on the same
10085 @c This list is ordered alphanumerically by subsection name.
10086 @c It should be the same order and spelling as these options are listed
10087 @c in Machine Dependent Options
10092 * Blackfin Options::
10095 * DEC Alpha Options::
10096 * DEC Alpha/VMS Options::
10099 * GNU/Linux Options::
10102 * i386 and x86-64 Options::
10103 * i386 and x86-64 Windows Options::
10105 * IA-64/VMS Options::
10112 * MicroBlaze Options::
10115 * MN10300 Options::
10117 * picoChip Options::
10118 * PowerPC Options::
10119 * RS/6000 and PowerPC Options::
10121 * S/390 and zSeries Options::
10124 * Solaris 2 Options::
10127 * System V Options::
10130 * VxWorks Options::
10132 * Xstormy16 Options::
10134 * zSeries Options::
10138 @subsection ARM Options
10139 @cindex ARM options
10141 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10145 @item -mabi=@var{name}
10147 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10148 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10151 @opindex mapcs-frame
10152 Generate a stack frame that is compliant with the ARM Procedure Call
10153 Standard for all functions, even if this is not strictly necessary for
10154 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10155 with this option will cause the stack frames not to be generated for
10156 leaf functions. The default is @option{-mno-apcs-frame}.
10160 This is a synonym for @option{-mapcs-frame}.
10163 @c not currently implemented
10164 @item -mapcs-stack-check
10165 @opindex mapcs-stack-check
10166 Generate code to check the amount of stack space available upon entry to
10167 every function (that actually uses some stack space). If there is
10168 insufficient space available then either the function
10169 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10170 called, depending upon the amount of stack space required. The run time
10171 system is required to provide these functions. The default is
10172 @option{-mno-apcs-stack-check}, since this produces smaller code.
10174 @c not currently implemented
10176 @opindex mapcs-float
10177 Pass floating point arguments using the float point registers. This is
10178 one of the variants of the APCS@. This option is recommended if the
10179 target hardware has a floating point unit or if a lot of floating point
10180 arithmetic is going to be performed by the code. The default is
10181 @option{-mno-apcs-float}, since integer only code is slightly increased in
10182 size if @option{-mapcs-float} is used.
10184 @c not currently implemented
10185 @item -mapcs-reentrant
10186 @opindex mapcs-reentrant
10187 Generate reentrant, position independent code. The default is
10188 @option{-mno-apcs-reentrant}.
10191 @item -mthumb-interwork
10192 @opindex mthumb-interwork
10193 Generate code which supports calling between the ARM and Thumb
10194 instruction sets. Without this option, on pre-v5 architectures, the
10195 two instruction sets cannot be reliably used inside one program. The
10196 default is @option{-mno-thumb-interwork}, since slightly larger code
10197 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10198 configurations this option is meaningless.
10200 @item -mno-sched-prolog
10201 @opindex mno-sched-prolog
10202 Prevent the reordering of instructions in the function prolog, or the
10203 merging of those instruction with the instructions in the function's
10204 body. This means that all functions will start with a recognizable set
10205 of instructions (or in fact one of a choice from a small set of
10206 different function prologues), and this information can be used to
10207 locate the start if functions inside an executable piece of code. The
10208 default is @option{-msched-prolog}.
10210 @item -mfloat-abi=@var{name}
10211 @opindex mfloat-abi
10212 Specifies which floating-point ABI to use. Permissible values
10213 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10215 Specifying @samp{soft} causes GCC to generate output containing
10216 library calls for floating-point operations.
10217 @samp{softfp} allows the generation of code using hardware floating-point
10218 instructions, but still uses the soft-float calling conventions.
10219 @samp{hard} allows generation of floating-point instructions
10220 and uses FPU-specific calling conventions.
10222 The default depends on the specific target configuration. Note that
10223 the hard-float and soft-float ABIs are not link-compatible; you must
10224 compile your entire program with the same ABI, and link with a
10225 compatible set of libraries.
10227 @item -mlittle-endian
10228 @opindex mlittle-endian
10229 Generate code for a processor running in little-endian mode. This is
10230 the default for all standard configurations.
10233 @opindex mbig-endian
10234 Generate code for a processor running in big-endian mode; the default is
10235 to compile code for a little-endian processor.
10237 @item -mwords-little-endian
10238 @opindex mwords-little-endian
10239 This option only applies when generating code for big-endian processors.
10240 Generate code for a little-endian word order but a big-endian byte
10241 order. That is, a byte order of the form @samp{32107654}. Note: this
10242 option should only be used if you require compatibility with code for
10243 big-endian ARM processors generated by versions of the compiler prior to
10244 2.8. This option is now deprecated.
10246 @item -mcpu=@var{name}
10248 This specifies the name of the target ARM processor. GCC uses this name
10249 to determine what kind of instructions it can emit when generating
10250 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10251 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10252 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10253 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10254 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10256 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10257 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10258 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10259 @samp{strongarm1110},
10260 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10261 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10262 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10263 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10264 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10265 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10266 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10267 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10268 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10269 @samp{cortex-m4}, @samp{cortex-m3},
10272 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10273 @samp{fa526}, @samp{fa626},
10274 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10276 @item -mtune=@var{name}
10278 This option is very similar to the @option{-mcpu=} option, except that
10279 instead of specifying the actual target processor type, and hence
10280 restricting which instructions can be used, it specifies that GCC should
10281 tune the performance of the code as if the target were of the type
10282 specified in this option, but still choosing the instructions that it
10283 will generate based on the CPU specified by a @option{-mcpu=} option.
10284 For some ARM implementations better performance can be obtained by using
10287 @item -march=@var{name}
10289 This specifies the name of the target ARM architecture. GCC uses this
10290 name to determine what kind of instructions it can emit when generating
10291 assembly code. This option can be used in conjunction with or instead
10292 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10293 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10294 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10295 @samp{armv6}, @samp{armv6j},
10296 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10297 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10298 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10300 @item -mfpu=@var{name}
10301 @itemx -mfpe=@var{number}
10302 @itemx -mfp=@var{number}
10306 This specifies what floating point hardware (or hardware emulation) is
10307 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10308 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10309 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10310 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10311 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10312 @option{-mfp} and @option{-mfpe} are synonyms for
10313 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10316 If @option{-msoft-float} is specified this specifies the format of
10317 floating point values.
10319 If the selected floating-point hardware includes the NEON extension
10320 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10321 operations will not be used by GCC's auto-vectorization pass unless
10322 @option{-funsafe-math-optimizations} is also specified. This is
10323 because NEON hardware does not fully implement the IEEE 754 standard for
10324 floating-point arithmetic (in particular denormal values are treated as
10325 zero), so the use of NEON instructions may lead to a loss of precision.
10327 @item -mfp16-format=@var{name}
10328 @opindex mfp16-format
10329 Specify the format of the @code{__fp16} half-precision floating-point type.
10330 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10331 the default is @samp{none}, in which case the @code{__fp16} type is not
10332 defined. @xref{Half-Precision}, for more information.
10334 @item -mstructure-size-boundary=@var{n}
10335 @opindex mstructure-size-boundary
10336 The size of all structures and unions will be rounded up to a multiple
10337 of the number of bits set by this option. Permissible values are 8, 32
10338 and 64. The default value varies for different toolchains. For the COFF
10339 targeted toolchain the default value is 8. A value of 64 is only allowed
10340 if the underlying ABI supports it.
10342 Specifying the larger number can produce faster, more efficient code, but
10343 can also increase the size of the program. Different values are potentially
10344 incompatible. Code compiled with one value cannot necessarily expect to
10345 work with code or libraries compiled with another value, if they exchange
10346 information using structures or unions.
10348 @item -mabort-on-noreturn
10349 @opindex mabort-on-noreturn
10350 Generate a call to the function @code{abort} at the end of a
10351 @code{noreturn} function. It will be executed if the function tries to
10355 @itemx -mno-long-calls
10356 @opindex mlong-calls
10357 @opindex mno-long-calls
10358 Tells the compiler to perform function calls by first loading the
10359 address of the function into a register and then performing a subroutine
10360 call on this register. This switch is needed if the target function
10361 will lie outside of the 64 megabyte addressing range of the offset based
10362 version of subroutine call instruction.
10364 Even if this switch is enabled, not all function calls will be turned
10365 into long calls. The heuristic is that static functions, functions
10366 which have the @samp{short-call} attribute, functions that are inside
10367 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10368 definitions have already been compiled within the current compilation
10369 unit, will not be turned into long calls. The exception to this rule is
10370 that weak function definitions, functions with the @samp{long-call}
10371 attribute or the @samp{section} attribute, and functions that are within
10372 the scope of a @samp{#pragma long_calls} directive, will always be
10373 turned into long calls.
10375 This feature is not enabled by default. Specifying
10376 @option{-mno-long-calls} will restore the default behavior, as will
10377 placing the function calls within the scope of a @samp{#pragma
10378 long_calls_off} directive. Note these switches have no effect on how
10379 the compiler generates code to handle function calls via function
10382 @item -msingle-pic-base
10383 @opindex msingle-pic-base
10384 Treat the register used for PIC addressing as read-only, rather than
10385 loading it in the prologue for each function. The run-time system is
10386 responsible for initializing this register with an appropriate value
10387 before execution begins.
10389 @item -mpic-register=@var{reg}
10390 @opindex mpic-register
10391 Specify the register to be used for PIC addressing. The default is R10
10392 unless stack-checking is enabled, when R9 is used.
10394 @item -mcirrus-fix-invalid-insns
10395 @opindex mcirrus-fix-invalid-insns
10396 @opindex mno-cirrus-fix-invalid-insns
10397 Insert NOPs into the instruction stream to in order to work around
10398 problems with invalid Maverick instruction combinations. This option
10399 is only valid if the @option{-mcpu=ep9312} option has been used to
10400 enable generation of instructions for the Cirrus Maverick floating
10401 point co-processor. This option is not enabled by default, since the
10402 problem is only present in older Maverick implementations. The default
10403 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10406 @item -mpoke-function-name
10407 @opindex mpoke-function-name
10408 Write the name of each function into the text section, directly
10409 preceding the function prologue. The generated code is similar to this:
10413 .ascii "arm_poke_function_name", 0
10416 .word 0xff000000 + (t1 - t0)
10417 arm_poke_function_name
10419 stmfd sp!, @{fp, ip, lr, pc@}
10423 When performing a stack backtrace, code can inspect the value of
10424 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10425 location @code{pc - 12} and the top 8 bits are set, then we know that
10426 there is a function name embedded immediately preceding this location
10427 and has length @code{((pc[-3]) & 0xff000000)}.
10434 Select between generating code that executes in ARM and Thumb
10435 states. The default for most configurations is to generate code
10436 that executes in ARM state, but the default can be changed by
10437 configuring GCC with the @option{--with-mode=}@var{state}
10441 @opindex mtpcs-frame
10442 Generate a stack frame that is compliant with the Thumb Procedure Call
10443 Standard for all non-leaf functions. (A leaf function is one that does
10444 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10446 @item -mtpcs-leaf-frame
10447 @opindex mtpcs-leaf-frame
10448 Generate a stack frame that is compliant with the Thumb Procedure Call
10449 Standard for all leaf functions. (A leaf function is one that does
10450 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10452 @item -mcallee-super-interworking
10453 @opindex mcallee-super-interworking
10454 Gives all externally visible functions in the file being compiled an ARM
10455 instruction set header which switches to Thumb mode before executing the
10456 rest of the function. This allows these functions to be called from
10457 non-interworking code. This option is not valid in AAPCS configurations
10458 because interworking is enabled by default.
10460 @item -mcaller-super-interworking
10461 @opindex mcaller-super-interworking
10462 Allows calls via function pointers (including virtual functions) to
10463 execute correctly regardless of whether the target code has been
10464 compiled for interworking or not. There is a small overhead in the cost
10465 of executing a function pointer if this option is enabled. This option
10466 is not valid in AAPCS configurations because interworking is enabled
10469 @item -mtp=@var{name}
10471 Specify the access model for the thread local storage pointer. The valid
10472 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10473 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10474 (supported in the arm6k architecture), and @option{auto}, which uses the
10475 best available method for the selected processor. The default setting is
10478 @item -mtls-dialect=@var{dialect}
10479 @opindex mtls-dialect
10480 Specify the dialect to use for accessing thread local storage. Two
10481 dialects are supported --- @option{gnu} and @option{gnu2}. The
10482 @option{gnu} dialect selects the original GNU scheme for supporting
10483 local and global dynamic TLS models. The @option{gnu2} dialect
10484 selects the GNU descriptor scheme, which provides better performance
10485 for shared libraries. The GNU descriptor scheme is compatible with
10486 the original scheme, but does require new assembler, linker and
10487 library support. Initial and local exec TLS models are unaffected by
10488 this option and always use the original scheme.
10490 @item -mword-relocations
10491 @opindex mword-relocations
10492 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10493 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10494 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10497 @item -mfix-cortex-m3-ldrd
10498 @opindex mfix-cortex-m3-ldrd
10499 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10500 with overlapping destination and base registers are used. This option avoids
10501 generating these instructions. This option is enabled by default when
10502 @option{-mcpu=cortex-m3} is specified.
10507 @subsection AVR Options
10508 @cindex AVR Options
10510 These options are defined for AVR implementations:
10513 @item -mmcu=@var{mcu}
10515 Specify ATMEL AVR instruction set or MCU type.
10517 Instruction set avr1 is for the minimal AVR core, not supported by the C
10518 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10519 attiny11, attiny12, attiny15, attiny28).
10521 Instruction set avr2 (default) is for the classic AVR core with up to
10522 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10523 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10524 at90c8534, at90s8535).
10526 Instruction set avr3 is for the classic AVR core with up to 128K program
10527 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10529 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10530 memory space (MCU types: atmega8, atmega83, atmega85).
10532 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10533 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10534 atmega64, atmega128, at43usb355, at94k).
10536 @item -mno-interrupts
10537 @opindex mno-interrupts
10538 Generated code is not compatible with hardware interrupts.
10539 Code size will be smaller.
10541 @item -mcall-prologues
10542 @opindex mcall-prologues
10543 Functions prologues/epilogues expanded as call to appropriate
10544 subroutines. Code size will be smaller.
10547 @opindex mtiny-stack
10548 Change only the low 8 bits of the stack pointer.
10552 Assume int to be 8 bit integer. This affects the sizes of all types: A
10553 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10554 and long long will be 4 bytes. Please note that this option does not
10555 comply to the C standards, but it will provide you with smaller code
10559 @node Blackfin Options
10560 @subsection Blackfin Options
10561 @cindex Blackfin Options
10564 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10566 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10567 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10568 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10569 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10570 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10571 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10572 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10573 @samp{bf561}, @samp{bf592}.
10574 The optional @var{sirevision} specifies the silicon revision of the target
10575 Blackfin processor. Any workarounds available for the targeted silicon revision
10576 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10577 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10578 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10579 hexadecimal digits representing the major and minor numbers in the silicon
10580 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10581 is not defined. If @var{sirevision} is @samp{any}, the
10582 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10583 If this optional @var{sirevision} is not used, GCC assumes the latest known
10584 silicon revision of the targeted Blackfin processor.
10586 Support for @samp{bf561} is incomplete. For @samp{bf561},
10587 Only the processor macro is defined.
10588 Without this option, @samp{bf532} is used as the processor by default.
10589 The corresponding predefined processor macros for @var{cpu} is to
10590 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10591 provided by libgloss to be linked in if @option{-msim} is not given.
10595 Specifies that the program will be run on the simulator. This causes
10596 the simulator BSP provided by libgloss to be linked in. This option
10597 has effect only for @samp{bfin-elf} toolchain.
10598 Certain other options, such as @option{-mid-shared-library} and
10599 @option{-mfdpic}, imply @option{-msim}.
10601 @item -momit-leaf-frame-pointer
10602 @opindex momit-leaf-frame-pointer
10603 Don't keep the frame pointer in a register for leaf functions. This
10604 avoids the instructions to save, set up and restore frame pointers and
10605 makes an extra register available in leaf functions. The option
10606 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10607 which might make debugging harder.
10609 @item -mspecld-anomaly
10610 @opindex mspecld-anomaly
10611 When enabled, the compiler will ensure that the generated code does not
10612 contain speculative loads after jump instructions. If this option is used,
10613 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10615 @item -mno-specld-anomaly
10616 @opindex mno-specld-anomaly
10617 Don't generate extra code to prevent speculative loads from occurring.
10619 @item -mcsync-anomaly
10620 @opindex mcsync-anomaly
10621 When enabled, the compiler will ensure that the generated code does not
10622 contain CSYNC or SSYNC instructions too soon after conditional branches.
10623 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10625 @item -mno-csync-anomaly
10626 @opindex mno-csync-anomaly
10627 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10628 occurring too soon after a conditional branch.
10632 When enabled, the compiler is free to take advantage of the knowledge that
10633 the entire program fits into the low 64k of memory.
10636 @opindex mno-low-64k
10637 Assume that the program is arbitrarily large. This is the default.
10639 @item -mstack-check-l1
10640 @opindex mstack-check-l1
10641 Do stack checking using information placed into L1 scratchpad memory by the
10644 @item -mid-shared-library
10645 @opindex mid-shared-library
10646 Generate code that supports shared libraries via the library ID method.
10647 This allows for execute in place and shared libraries in an environment
10648 without virtual memory management. This option implies @option{-fPIC}.
10649 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10651 @item -mno-id-shared-library
10652 @opindex mno-id-shared-library
10653 Generate code that doesn't assume ID based shared libraries are being used.
10654 This is the default.
10656 @item -mleaf-id-shared-library
10657 @opindex mleaf-id-shared-library
10658 Generate code that supports shared libraries via the library ID method,
10659 but assumes that this library or executable won't link against any other
10660 ID shared libraries. That allows the compiler to use faster code for jumps
10663 @item -mno-leaf-id-shared-library
10664 @opindex mno-leaf-id-shared-library
10665 Do not assume that the code being compiled won't link against any ID shared
10666 libraries. Slower code will be generated for jump and call insns.
10668 @item -mshared-library-id=n
10669 @opindex mshared-library-id
10670 Specified the identification number of the ID based shared library being
10671 compiled. Specifying a value of 0 will generate more compact code, specifying
10672 other values will force the allocation of that number to the current
10673 library but is no more space or time efficient than omitting this option.
10677 Generate code that allows the data segment to be located in a different
10678 area of memory from the text segment. This allows for execute in place in
10679 an environment without virtual memory management by eliminating relocations
10680 against the text section.
10682 @item -mno-sep-data
10683 @opindex mno-sep-data
10684 Generate code that assumes that the data segment follows the text segment.
10685 This is the default.
10688 @itemx -mno-long-calls
10689 @opindex mlong-calls
10690 @opindex mno-long-calls
10691 Tells the compiler to perform function calls by first loading the
10692 address of the function into a register and then performing a subroutine
10693 call on this register. This switch is needed if the target function
10694 will lie outside of the 24 bit addressing range of the offset based
10695 version of subroutine call instruction.
10697 This feature is not enabled by default. Specifying
10698 @option{-mno-long-calls} will restore the default behavior. Note these
10699 switches have no effect on how the compiler generates code to handle
10700 function calls via function pointers.
10704 Link with the fast floating-point library. This library relaxes some of
10705 the IEEE floating-point standard's rules for checking inputs against
10706 Not-a-Number (NAN), in the interest of performance.
10709 @opindex minline-plt
10710 Enable inlining of PLT entries in function calls to functions that are
10711 not known to bind locally. It has no effect without @option{-mfdpic}.
10714 @opindex mmulticore
10715 Build standalone application for multicore Blackfin processor. Proper
10716 start files and link scripts will be used to support multicore.
10717 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10718 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10719 @option{-mcorea} or @option{-mcoreb}. If it's used without
10720 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10721 programming model is used. In this model, the main function of Core B
10722 should be named as coreb_main. If it's used with @option{-mcorea} or
10723 @option{-mcoreb}, one application per core programming model is used.
10724 If this option is not used, single core application programming
10729 Build standalone application for Core A of BF561 when using
10730 one application per core programming model. Proper start files
10731 and link scripts will be used to support Core A. This option
10732 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10736 Build standalone application for Core B of BF561 when using
10737 one application per core programming model. Proper start files
10738 and link scripts will be used to support Core B. This option
10739 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10740 should be used instead of main. It must be used with
10741 @option{-mmulticore}.
10745 Build standalone application for SDRAM. Proper start files and
10746 link scripts will be used to put the application into SDRAM.
10747 Loader should initialize SDRAM before loading the application
10748 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10752 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10753 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10754 are enabled; for standalone applications the default is off.
10758 @subsection CRIS Options
10759 @cindex CRIS Options
10761 These options are defined specifically for the CRIS ports.
10764 @item -march=@var{architecture-type}
10765 @itemx -mcpu=@var{architecture-type}
10768 Generate code for the specified architecture. The choices for
10769 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10770 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10771 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10774 @item -mtune=@var{architecture-type}
10776 Tune to @var{architecture-type} everything applicable about the generated
10777 code, except for the ABI and the set of available instructions. The
10778 choices for @var{architecture-type} are the same as for
10779 @option{-march=@var{architecture-type}}.
10781 @item -mmax-stack-frame=@var{n}
10782 @opindex mmax-stack-frame
10783 Warn when the stack frame of a function exceeds @var{n} bytes.
10789 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10790 @option{-march=v3} and @option{-march=v8} respectively.
10792 @item -mmul-bug-workaround
10793 @itemx -mno-mul-bug-workaround
10794 @opindex mmul-bug-workaround
10795 @opindex mno-mul-bug-workaround
10796 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10797 models where it applies. This option is active by default.
10801 Enable CRIS-specific verbose debug-related information in the assembly
10802 code. This option also has the effect to turn off the @samp{#NO_APP}
10803 formatted-code indicator to the assembler at the beginning of the
10808 Do not use condition-code results from previous instruction; always emit
10809 compare and test instructions before use of condition codes.
10811 @item -mno-side-effects
10812 @opindex mno-side-effects
10813 Do not emit instructions with side-effects in addressing modes other than
10816 @item -mstack-align
10817 @itemx -mno-stack-align
10818 @itemx -mdata-align
10819 @itemx -mno-data-align
10820 @itemx -mconst-align
10821 @itemx -mno-const-align
10822 @opindex mstack-align
10823 @opindex mno-stack-align
10824 @opindex mdata-align
10825 @opindex mno-data-align
10826 @opindex mconst-align
10827 @opindex mno-const-align
10828 These options (no-options) arranges (eliminate arrangements) for the
10829 stack-frame, individual data and constants to be aligned for the maximum
10830 single data access size for the chosen CPU model. The default is to
10831 arrange for 32-bit alignment. ABI details such as structure layout are
10832 not affected by these options.
10840 Similar to the stack- data- and const-align options above, these options
10841 arrange for stack-frame, writable data and constants to all be 32-bit,
10842 16-bit or 8-bit aligned. The default is 32-bit alignment.
10844 @item -mno-prologue-epilogue
10845 @itemx -mprologue-epilogue
10846 @opindex mno-prologue-epilogue
10847 @opindex mprologue-epilogue
10848 With @option{-mno-prologue-epilogue}, the normal function prologue and
10849 epilogue that sets up the stack-frame are omitted and no return
10850 instructions or return sequences are generated in the code. Use this
10851 option only together with visual inspection of the compiled code: no
10852 warnings or errors are generated when call-saved registers must be saved,
10853 or storage for local variable needs to be allocated.
10857 @opindex mno-gotplt
10859 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10860 instruction sequences that load addresses for functions from the PLT part
10861 of the GOT rather than (traditional on other architectures) calls to the
10862 PLT@. The default is @option{-mgotplt}.
10866 Legacy no-op option only recognized with the cris-axis-elf and
10867 cris-axis-linux-gnu targets.
10871 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10875 This option, recognized for the cris-axis-elf arranges
10876 to link with input-output functions from a simulator library. Code,
10877 initialized data and zero-initialized data are allocated consecutively.
10881 Like @option{-sim}, but pass linker options to locate initialized data at
10882 0x40000000 and zero-initialized data at 0x80000000.
10885 @node Darwin Options
10886 @subsection Darwin Options
10887 @cindex Darwin options
10889 These options are defined for all architectures running the Darwin operating
10892 FSF GCC on Darwin does not create ``fat'' object files; it will create
10893 an object file for the single architecture that it was built to
10894 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10895 @option{-arch} options are used; it does so by running the compiler or
10896 linker multiple times and joining the results together with
10899 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10900 @samp{i686}) is determined by the flags that specify the ISA
10901 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10902 @option{-force_cpusubtype_ALL} option can be used to override this.
10904 The Darwin tools vary in their behavior when presented with an ISA
10905 mismatch. The assembler, @file{as}, will only permit instructions to
10906 be used that are valid for the subtype of the file it is generating,
10907 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
10908 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
10909 and print an error if asked to create a shared library with a less
10910 restrictive subtype than its input files (for instance, trying to put
10911 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
10912 for executables, @file{ld}, will quietly give the executable the most
10913 restrictive subtype of any of its input files.
10918 Add the framework directory @var{dir} to the head of the list of
10919 directories to be searched for header files. These directories are
10920 interleaved with those specified by @option{-I} options and are
10921 scanned in a left-to-right order.
10923 A framework directory is a directory with frameworks in it. A
10924 framework is a directory with a @samp{"Headers"} and/or
10925 @samp{"PrivateHeaders"} directory contained directly in it that ends
10926 in @samp{".framework"}. The name of a framework is the name of this
10927 directory excluding the @samp{".framework"}. Headers associated with
10928 the framework are found in one of those two directories, with
10929 @samp{"Headers"} being searched first. A subframework is a framework
10930 directory that is in a framework's @samp{"Frameworks"} directory.
10931 Includes of subframework headers can only appear in a header of a
10932 framework that contains the subframework, or in a sibling subframework
10933 header. Two subframeworks are siblings if they occur in the same
10934 framework. A subframework should not have the same name as a
10935 framework, a warning will be issued if this is violated. Currently a
10936 subframework cannot have subframeworks, in the future, the mechanism
10937 may be extended to support this. The standard frameworks can be found
10938 in @samp{"/System/Library/Frameworks"} and
10939 @samp{"/Library/Frameworks"}. An example include looks like
10940 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
10941 the name of the framework and header.h is found in the
10942 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
10944 @item -iframework@var{dir}
10945 @opindex iframework
10946 Like @option{-F} except the directory is a treated as a system
10947 directory. The main difference between this @option{-iframework} and
10948 @option{-F} is that with @option{-iframework} the compiler does not
10949 warn about constructs contained within header files found via
10950 @var{dir}. This option is valid only for the C family of languages.
10954 Emit debugging information for symbols that are used. For STABS
10955 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
10956 This is by default ON@.
10960 Emit debugging information for all symbols and types.
10962 @item -mmacosx-version-min=@var{version}
10963 The earliest version of MacOS X that this executable will run on
10964 is @var{version}. Typical values of @var{version} include @code{10.1},
10965 @code{10.2}, and @code{10.3.9}.
10967 If the compiler was built to use the system's headers by default,
10968 then the default for this option is the system version on which the
10969 compiler is running, otherwise the default is to make choices which
10970 are compatible with as many systems and code bases as possible.
10974 Enable kernel development mode. The @option{-mkernel} option sets
10975 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
10976 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
10977 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
10978 applicable. This mode also sets @option{-mno-altivec},
10979 @option{-msoft-float}, @option{-fno-builtin} and
10980 @option{-mlong-branch} for PowerPC targets.
10982 @item -mone-byte-bool
10983 @opindex mone-byte-bool
10984 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
10985 By default @samp{sizeof(bool)} is @samp{4} when compiling for
10986 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
10987 option has no effect on x86.
10989 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
10990 to generate code that is not binary compatible with code generated
10991 without that switch. Using this switch may require recompiling all
10992 other modules in a program, including system libraries. Use this
10993 switch to conform to a non-default data model.
10995 @item -mfix-and-continue
10996 @itemx -ffix-and-continue
10997 @itemx -findirect-data
10998 @opindex mfix-and-continue
10999 @opindex ffix-and-continue
11000 @opindex findirect-data
11001 Generate code suitable for fast turn around development. Needed to
11002 enable gdb to dynamically load @code{.o} files into already running
11003 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11004 are provided for backwards compatibility.
11008 Loads all members of static archive libraries.
11009 See man ld(1) for more information.
11011 @item -arch_errors_fatal
11012 @opindex arch_errors_fatal
11013 Cause the errors having to do with files that have the wrong architecture
11016 @item -bind_at_load
11017 @opindex bind_at_load
11018 Causes the output file to be marked such that the dynamic linker will
11019 bind all undefined references when the file is loaded or launched.
11023 Produce a Mach-o bundle format file.
11024 See man ld(1) for more information.
11026 @item -bundle_loader @var{executable}
11027 @opindex bundle_loader
11028 This option specifies the @var{executable} that will be loading the build
11029 output file being linked. See man ld(1) for more information.
11032 @opindex dynamiclib
11033 When passed this option, GCC will produce a dynamic library instead of
11034 an executable when linking, using the Darwin @file{libtool} command.
11036 @item -force_cpusubtype_ALL
11037 @opindex force_cpusubtype_ALL
11038 This causes GCC's output file to have the @var{ALL} subtype, instead of
11039 one controlled by the @option{-mcpu} or @option{-march} option.
11041 @item -allowable_client @var{client_name}
11042 @itemx -client_name
11043 @itemx -compatibility_version
11044 @itemx -current_version
11046 @itemx -dependency-file
11048 @itemx -dylinker_install_name
11050 @itemx -exported_symbols_list
11053 @itemx -flat_namespace
11054 @itemx -force_flat_namespace
11055 @itemx -headerpad_max_install_names
11058 @itemx -install_name
11059 @itemx -keep_private_externs
11060 @itemx -multi_module
11061 @itemx -multiply_defined
11062 @itemx -multiply_defined_unused
11065 @itemx -no_dead_strip_inits_and_terms
11066 @itemx -nofixprebinding
11067 @itemx -nomultidefs
11069 @itemx -noseglinkedit
11070 @itemx -pagezero_size
11072 @itemx -prebind_all_twolevel_modules
11073 @itemx -private_bundle
11075 @itemx -read_only_relocs
11077 @itemx -sectobjectsymbols
11081 @itemx -sectobjectsymbols
11084 @itemx -segs_read_only_addr
11086 @itemx -segs_read_write_addr
11087 @itemx -seg_addr_table
11088 @itemx -seg_addr_table_filename
11089 @itemx -seglinkedit
11091 @itemx -segs_read_only_addr
11092 @itemx -segs_read_write_addr
11093 @itemx -single_module
11095 @itemx -sub_library
11097 @itemx -sub_umbrella
11098 @itemx -twolevel_namespace
11101 @itemx -unexported_symbols_list
11102 @itemx -weak_reference_mismatches
11103 @itemx -whatsloaded
11104 @opindex allowable_client
11105 @opindex client_name
11106 @opindex compatibility_version
11107 @opindex current_version
11108 @opindex dead_strip
11109 @opindex dependency-file
11110 @opindex dylib_file
11111 @opindex dylinker_install_name
11113 @opindex exported_symbols_list
11115 @opindex flat_namespace
11116 @opindex force_flat_namespace
11117 @opindex headerpad_max_install_names
11118 @opindex image_base
11120 @opindex install_name
11121 @opindex keep_private_externs
11122 @opindex multi_module
11123 @opindex multiply_defined
11124 @opindex multiply_defined_unused
11125 @opindex noall_load
11126 @opindex no_dead_strip_inits_and_terms
11127 @opindex nofixprebinding
11128 @opindex nomultidefs
11130 @opindex noseglinkedit
11131 @opindex pagezero_size
11133 @opindex prebind_all_twolevel_modules
11134 @opindex private_bundle
11135 @opindex read_only_relocs
11137 @opindex sectobjectsymbols
11140 @opindex sectcreate
11141 @opindex sectobjectsymbols
11144 @opindex segs_read_only_addr
11145 @opindex segs_read_write_addr
11146 @opindex seg_addr_table
11147 @opindex seg_addr_table_filename
11148 @opindex seglinkedit
11150 @opindex segs_read_only_addr
11151 @opindex segs_read_write_addr
11152 @opindex single_module
11154 @opindex sub_library
11155 @opindex sub_umbrella
11156 @opindex twolevel_namespace
11159 @opindex unexported_symbols_list
11160 @opindex weak_reference_mismatches
11161 @opindex whatsloaded
11162 These options are passed to the Darwin linker. The Darwin linker man page
11163 describes them in detail.
11166 @node DEC Alpha Options
11167 @subsection DEC Alpha Options
11169 These @samp{-m} options are defined for the DEC Alpha implementations:
11172 @item -mno-soft-float
11173 @itemx -msoft-float
11174 @opindex mno-soft-float
11175 @opindex msoft-float
11176 Use (do not use) the hardware floating-point instructions for
11177 floating-point operations. When @option{-msoft-float} is specified,
11178 functions in @file{libgcc.a} will be used to perform floating-point
11179 operations. Unless they are replaced by routines that emulate the
11180 floating-point operations, or compiled in such a way as to call such
11181 emulations routines, these routines will issue floating-point
11182 operations. If you are compiling for an Alpha without floating-point
11183 operations, you must ensure that the library is built so as not to call
11186 Note that Alpha implementations without floating-point operations are
11187 required to have floating-point registers.
11190 @itemx -mno-fp-regs
11192 @opindex mno-fp-regs
11193 Generate code that uses (does not use) the floating-point register set.
11194 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11195 register set is not used, floating point operands are passed in integer
11196 registers as if they were integers and floating-point results are passed
11197 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11198 so any function with a floating-point argument or return value called by code
11199 compiled with @option{-mno-fp-regs} must also be compiled with that
11202 A typical use of this option is building a kernel that does not use,
11203 and hence need not save and restore, any floating-point registers.
11207 The Alpha architecture implements floating-point hardware optimized for
11208 maximum performance. It is mostly compliant with the IEEE floating
11209 point standard. However, for full compliance, software assistance is
11210 required. This option generates code fully IEEE compliant code
11211 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11212 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11213 defined during compilation. The resulting code is less efficient but is
11214 able to correctly support denormalized numbers and exceptional IEEE
11215 values such as not-a-number and plus/minus infinity. Other Alpha
11216 compilers call this option @option{-ieee_with_no_inexact}.
11218 @item -mieee-with-inexact
11219 @opindex mieee-with-inexact
11220 This is like @option{-mieee} except the generated code also maintains
11221 the IEEE @var{inexact-flag}. Turning on this option causes the
11222 generated code to implement fully-compliant IEEE math. In addition to
11223 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11224 macro. On some Alpha implementations the resulting code may execute
11225 significantly slower than the code generated by default. Since there is
11226 very little code that depends on the @var{inexact-flag}, you should
11227 normally not specify this option. Other Alpha compilers call this
11228 option @option{-ieee_with_inexact}.
11230 @item -mfp-trap-mode=@var{trap-mode}
11231 @opindex mfp-trap-mode
11232 This option controls what floating-point related traps are enabled.
11233 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11234 The trap mode can be set to one of four values:
11238 This is the default (normal) setting. The only traps that are enabled
11239 are the ones that cannot be disabled in software (e.g., division by zero
11243 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11247 Like @samp{u}, but the instructions are marked to be safe for software
11248 completion (see Alpha architecture manual for details).
11251 Like @samp{su}, but inexact traps are enabled as well.
11254 @item -mfp-rounding-mode=@var{rounding-mode}
11255 @opindex mfp-rounding-mode
11256 Selects the IEEE rounding mode. Other Alpha compilers call this option
11257 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11262 Normal IEEE rounding mode. Floating point numbers are rounded towards
11263 the nearest machine number or towards the even machine number in case
11267 Round towards minus infinity.
11270 Chopped rounding mode. Floating point numbers are rounded towards zero.
11273 Dynamic rounding mode. A field in the floating point control register
11274 (@var{fpcr}, see Alpha architecture reference manual) controls the
11275 rounding mode in effect. The C library initializes this register for
11276 rounding towards plus infinity. Thus, unless your program modifies the
11277 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11280 @item -mtrap-precision=@var{trap-precision}
11281 @opindex mtrap-precision
11282 In the Alpha architecture, floating point traps are imprecise. This
11283 means without software assistance it is impossible to recover from a
11284 floating trap and program execution normally needs to be terminated.
11285 GCC can generate code that can assist operating system trap handlers
11286 in determining the exact location that caused a floating point trap.
11287 Depending on the requirements of an application, different levels of
11288 precisions can be selected:
11292 Program precision. This option is the default and means a trap handler
11293 can only identify which program caused a floating point exception.
11296 Function precision. The trap handler can determine the function that
11297 caused a floating point exception.
11300 Instruction precision. The trap handler can determine the exact
11301 instruction that caused a floating point exception.
11304 Other Alpha compilers provide the equivalent options called
11305 @option{-scope_safe} and @option{-resumption_safe}.
11307 @item -mieee-conformant
11308 @opindex mieee-conformant
11309 This option marks the generated code as IEEE conformant. You must not
11310 use this option unless you also specify @option{-mtrap-precision=i} and either
11311 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11312 is to emit the line @samp{.eflag 48} in the function prologue of the
11313 generated assembly file. Under DEC Unix, this has the effect that
11314 IEEE-conformant math library routines will be linked in.
11316 @item -mbuild-constants
11317 @opindex mbuild-constants
11318 Normally GCC examines a 32- or 64-bit integer constant to
11319 see if it can construct it from smaller constants in two or three
11320 instructions. If it cannot, it will output the constant as a literal and
11321 generate code to load it from the data segment at runtime.
11323 Use this option to require GCC to construct @emph{all} integer constants
11324 using code, even if it takes more instructions (the maximum is six).
11326 You would typically use this option to build a shared library dynamic
11327 loader. Itself a shared library, it must relocate itself in memory
11328 before it can find the variables and constants in its own data segment.
11334 Select whether to generate code to be assembled by the vendor-supplied
11335 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11353 Indicate whether GCC should generate code to use the optional BWX,
11354 CIX, FIX and MAX instruction sets. The default is to use the instruction
11355 sets supported by the CPU type specified via @option{-mcpu=} option or that
11356 of the CPU on which GCC was built if none was specified.
11359 @itemx -mfloat-ieee
11360 @opindex mfloat-vax
11361 @opindex mfloat-ieee
11362 Generate code that uses (does not use) VAX F and G floating point
11363 arithmetic instead of IEEE single and double precision.
11365 @item -mexplicit-relocs
11366 @itemx -mno-explicit-relocs
11367 @opindex mexplicit-relocs
11368 @opindex mno-explicit-relocs
11369 Older Alpha assemblers provided no way to generate symbol relocations
11370 except via assembler macros. Use of these macros does not allow
11371 optimal instruction scheduling. GNU binutils as of version 2.12
11372 supports a new syntax that allows the compiler to explicitly mark
11373 which relocations should apply to which instructions. This option
11374 is mostly useful for debugging, as GCC detects the capabilities of
11375 the assembler when it is built and sets the default accordingly.
11378 @itemx -mlarge-data
11379 @opindex msmall-data
11380 @opindex mlarge-data
11381 When @option{-mexplicit-relocs} is in effect, static data is
11382 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11383 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11384 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11385 16-bit relocations off of the @code{$gp} register. This limits the
11386 size of the small data area to 64KB, but allows the variables to be
11387 directly accessed via a single instruction.
11389 The default is @option{-mlarge-data}. With this option the data area
11390 is limited to just below 2GB@. Programs that require more than 2GB of
11391 data must use @code{malloc} or @code{mmap} to allocate the data in the
11392 heap instead of in the program's data segment.
11394 When generating code for shared libraries, @option{-fpic} implies
11395 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11398 @itemx -mlarge-text
11399 @opindex msmall-text
11400 @opindex mlarge-text
11401 When @option{-msmall-text} is used, the compiler assumes that the
11402 code of the entire program (or shared library) fits in 4MB, and is
11403 thus reachable with a branch instruction. When @option{-msmall-data}
11404 is used, the compiler can assume that all local symbols share the
11405 same @code{$gp} value, and thus reduce the number of instructions
11406 required for a function call from 4 to 1.
11408 The default is @option{-mlarge-text}.
11410 @item -mcpu=@var{cpu_type}
11412 Set the instruction set and instruction scheduling parameters for
11413 machine type @var{cpu_type}. You can specify either the @samp{EV}
11414 style name or the corresponding chip number. GCC supports scheduling
11415 parameters for the EV4, EV5 and EV6 family of processors and will
11416 choose the default values for the instruction set from the processor
11417 you specify. If you do not specify a processor type, GCC will default
11418 to the processor on which the compiler was built.
11420 Supported values for @var{cpu_type} are
11426 Schedules as an EV4 and has no instruction set extensions.
11430 Schedules as an EV5 and has no instruction set extensions.
11434 Schedules as an EV5 and supports the BWX extension.
11439 Schedules as an EV5 and supports the BWX and MAX extensions.
11443 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11447 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11450 Native Linux/GNU toolchains also support the value @samp{native},
11451 which selects the best architecture option for the host processor.
11452 @option{-mcpu=native} has no effect if GCC does not recognize
11455 @item -mtune=@var{cpu_type}
11457 Set only the instruction scheduling parameters for machine type
11458 @var{cpu_type}. The instruction set is not changed.
11460 Native Linux/GNU toolchains also support the value @samp{native},
11461 which selects the best architecture option for the host processor.
11462 @option{-mtune=native} has no effect if GCC does not recognize
11465 @item -mmemory-latency=@var{time}
11466 @opindex mmemory-latency
11467 Sets the latency the scheduler should assume for typical memory
11468 references as seen by the application. This number is highly
11469 dependent on the memory access patterns used by the application
11470 and the size of the external cache on the machine.
11472 Valid options for @var{time} are
11476 A decimal number representing clock cycles.
11482 The compiler contains estimates of the number of clock cycles for
11483 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11484 (also called Dcache, Scache, and Bcache), as well as to main memory.
11485 Note that L3 is only valid for EV5.
11490 @node DEC Alpha/VMS Options
11491 @subsection DEC Alpha/VMS Options
11493 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11496 @item -mvms-return-codes
11497 @opindex mvms-return-codes
11498 Return VMS condition codes from main. The default is to return POSIX
11499 style condition (e.g.@: error) codes.
11501 @item -mdebug-main=@var{prefix}
11502 @opindex mdebug-main=@var{prefix}
11503 Flag the first routine whose name starts with @var{prefix} as the main
11504 routine for the debugger.
11508 Default to 64bit memory allocation routines.
11512 @subsection FR30 Options
11513 @cindex FR30 Options
11515 These options are defined specifically for the FR30 port.
11519 @item -msmall-model
11520 @opindex msmall-model
11521 Use the small address space model. This can produce smaller code, but
11522 it does assume that all symbolic values and addresses will fit into a
11527 Assume that run-time support has been provided and so there is no need
11528 to include the simulator library (@file{libsim.a}) on the linker
11534 @subsection FRV Options
11535 @cindex FRV Options
11541 Only use the first 32 general purpose registers.
11546 Use all 64 general purpose registers.
11551 Use only the first 32 floating point registers.
11556 Use all 64 floating point registers
11559 @opindex mhard-float
11561 Use hardware instructions for floating point operations.
11564 @opindex msoft-float
11566 Use library routines for floating point operations.
11571 Dynamically allocate condition code registers.
11576 Do not try to dynamically allocate condition code registers, only
11577 use @code{icc0} and @code{fcc0}.
11582 Change ABI to use double word insns.
11587 Do not use double word instructions.
11592 Use floating point double instructions.
11595 @opindex mno-double
11597 Do not use floating point double instructions.
11602 Use media instructions.
11607 Do not use media instructions.
11612 Use multiply and add/subtract instructions.
11615 @opindex mno-muladd
11617 Do not use multiply and add/subtract instructions.
11622 Select the FDPIC ABI, that uses function descriptors to represent
11623 pointers to functions. Without any PIC/PIE-related options, it
11624 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11625 assumes GOT entries and small data are within a 12-bit range from the
11626 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11627 are computed with 32 bits.
11628 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11631 @opindex minline-plt
11633 Enable inlining of PLT entries in function calls to functions that are
11634 not known to bind locally. It has no effect without @option{-mfdpic}.
11635 It's enabled by default if optimizing for speed and compiling for
11636 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11637 optimization option such as @option{-O3} or above is present in the
11643 Assume a large TLS segment when generating thread-local code.
11648 Do not assume a large TLS segment when generating thread-local code.
11653 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11654 that is known to be in read-only sections. It's enabled by default,
11655 except for @option{-fpic} or @option{-fpie}: even though it may help
11656 make the global offset table smaller, it trades 1 instruction for 4.
11657 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11658 one of which may be shared by multiple symbols, and it avoids the need
11659 for a GOT entry for the referenced symbol, so it's more likely to be a
11660 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11662 @item -multilib-library-pic
11663 @opindex multilib-library-pic
11665 Link with the (library, not FD) pic libraries. It's implied by
11666 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11667 @option{-fpic} without @option{-mfdpic}. You should never have to use
11671 @opindex mlinked-fp
11673 Follow the EABI requirement of always creating a frame pointer whenever
11674 a stack frame is allocated. This option is enabled by default and can
11675 be disabled with @option{-mno-linked-fp}.
11678 @opindex mlong-calls
11680 Use indirect addressing to call functions outside the current
11681 compilation unit. This allows the functions to be placed anywhere
11682 within the 32-bit address space.
11684 @item -malign-labels
11685 @opindex malign-labels
11687 Try to align labels to an 8-byte boundary by inserting nops into the
11688 previous packet. This option only has an effect when VLIW packing
11689 is enabled. It doesn't create new packets; it merely adds nops to
11692 @item -mlibrary-pic
11693 @opindex mlibrary-pic
11695 Generate position-independent EABI code.
11700 Use only the first four media accumulator registers.
11705 Use all eight media accumulator registers.
11710 Pack VLIW instructions.
11715 Do not pack VLIW instructions.
11718 @opindex mno-eflags
11720 Do not mark ABI switches in e_flags.
11723 @opindex mcond-move
11725 Enable the use of conditional-move instructions (default).
11727 This switch is mainly for debugging the compiler and will likely be removed
11728 in a future version.
11730 @item -mno-cond-move
11731 @opindex mno-cond-move
11733 Disable the use of conditional-move instructions.
11735 This switch is mainly for debugging the compiler and will likely be removed
11736 in a future version.
11741 Enable the use of conditional set instructions (default).
11743 This switch is mainly for debugging the compiler and will likely be removed
11744 in a future version.
11749 Disable the use of conditional set instructions.
11751 This switch is mainly for debugging the compiler and will likely be removed
11752 in a future version.
11755 @opindex mcond-exec
11757 Enable the use of conditional execution (default).
11759 This switch is mainly for debugging the compiler and will likely be removed
11760 in a future version.
11762 @item -mno-cond-exec
11763 @opindex mno-cond-exec
11765 Disable the use of conditional execution.
11767 This switch is mainly for debugging the compiler and will likely be removed
11768 in a future version.
11770 @item -mvliw-branch
11771 @opindex mvliw-branch
11773 Run a pass to pack branches into VLIW instructions (default).
11775 This switch is mainly for debugging the compiler and will likely be removed
11776 in a future version.
11778 @item -mno-vliw-branch
11779 @opindex mno-vliw-branch
11781 Do not run a pass to pack branches into VLIW instructions.
11783 This switch is mainly for debugging the compiler and will likely be removed
11784 in a future version.
11786 @item -mmulti-cond-exec
11787 @opindex mmulti-cond-exec
11789 Enable optimization of @code{&&} and @code{||} in conditional execution
11792 This switch is mainly for debugging the compiler and will likely be removed
11793 in a future version.
11795 @item -mno-multi-cond-exec
11796 @opindex mno-multi-cond-exec
11798 Disable optimization of @code{&&} and @code{||} in conditional execution.
11800 This switch is mainly for debugging the compiler and will likely be removed
11801 in a future version.
11803 @item -mnested-cond-exec
11804 @opindex mnested-cond-exec
11806 Enable nested conditional execution optimizations (default).
11808 This switch is mainly for debugging the compiler and will likely be removed
11809 in a future version.
11811 @item -mno-nested-cond-exec
11812 @opindex mno-nested-cond-exec
11814 Disable nested conditional execution optimizations.
11816 This switch is mainly for debugging the compiler and will likely be removed
11817 in a future version.
11819 @item -moptimize-membar
11820 @opindex moptimize-membar
11822 This switch removes redundant @code{membar} instructions from the
11823 compiler generated code. It is enabled by default.
11825 @item -mno-optimize-membar
11826 @opindex mno-optimize-membar
11828 This switch disables the automatic removal of redundant @code{membar}
11829 instructions from the generated code.
11831 @item -mtomcat-stats
11832 @opindex mtomcat-stats
11834 Cause gas to print out tomcat statistics.
11836 @item -mcpu=@var{cpu}
11839 Select the processor type for which to generate code. Possible values are
11840 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11841 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11845 @node GNU/Linux Options
11846 @subsection GNU/Linux Options
11848 These @samp{-m} options are defined for GNU/Linux targets:
11853 Use the GNU C library. This is the default except
11854 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11858 Use uClibc C library. This is the default on
11859 @samp{*-*-linux-*uclibc*} targets.
11863 Use Bionic C library. This is the default on
11864 @samp{*-*-linux-*android*} targets.
11868 Compile code compatible with Android platform. This is the default on
11869 @samp{*-*-linux-*android*} targets.
11871 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11872 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11873 this option makes the GCC driver pass Android-specific options to the linker.
11874 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11877 @item -tno-android-cc
11878 @opindex tno-android-cc
11879 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11880 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11881 @option{-fno-rtti} by default.
11883 @item -tno-android-ld
11884 @opindex tno-android-ld
11885 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11886 linking options to the linker.
11890 @node H8/300 Options
11891 @subsection H8/300 Options
11893 These @samp{-m} options are defined for the H8/300 implementations:
11898 Shorten some address references at link time, when possible; uses the
11899 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11900 ld, Using ld}, for a fuller description.
11904 Generate code for the H8/300H@.
11908 Generate code for the H8S@.
11912 Generate code for the H8S and H8/300H in the normal mode. This switch
11913 must be used either with @option{-mh} or @option{-ms}.
11917 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
11921 Make @code{int} data 32 bits by default.
11924 @opindex malign-300
11925 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
11926 The default for the H8/300H and H8S is to align longs and floats on 4
11928 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
11929 This option has no effect on the H8/300.
11933 @subsection HPPA Options
11934 @cindex HPPA Options
11936 These @samp{-m} options are defined for the HPPA family of computers:
11939 @item -march=@var{architecture-type}
11941 Generate code for the specified architecture. The choices for
11942 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
11943 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
11944 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
11945 architecture option for your machine. Code compiled for lower numbered
11946 architectures will run on higher numbered architectures, but not the
11949 @item -mpa-risc-1-0
11950 @itemx -mpa-risc-1-1
11951 @itemx -mpa-risc-2-0
11952 @opindex mpa-risc-1-0
11953 @opindex mpa-risc-1-1
11954 @opindex mpa-risc-2-0
11955 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
11958 @opindex mbig-switch
11959 Generate code suitable for big switch tables. Use this option only if
11960 the assembler/linker complain about out of range branches within a switch
11963 @item -mjump-in-delay
11964 @opindex mjump-in-delay
11965 Fill delay slots of function calls with unconditional jump instructions
11966 by modifying the return pointer for the function call to be the target
11967 of the conditional jump.
11969 @item -mdisable-fpregs
11970 @opindex mdisable-fpregs
11971 Prevent floating point registers from being used in any manner. This is
11972 necessary for compiling kernels which perform lazy context switching of
11973 floating point registers. If you use this option and attempt to perform
11974 floating point operations, the compiler will abort.
11976 @item -mdisable-indexing
11977 @opindex mdisable-indexing
11978 Prevent the compiler from using indexing address modes. This avoids some
11979 rather obscure problems when compiling MIG generated code under MACH@.
11981 @item -mno-space-regs
11982 @opindex mno-space-regs
11983 Generate code that assumes the target has no space registers. This allows
11984 GCC to generate faster indirect calls and use unscaled index address modes.
11986 Such code is suitable for level 0 PA systems and kernels.
11988 @item -mfast-indirect-calls
11989 @opindex mfast-indirect-calls
11990 Generate code that assumes calls never cross space boundaries. This
11991 allows GCC to emit code which performs faster indirect calls.
11993 This option will not work in the presence of shared libraries or nested
11996 @item -mfixed-range=@var{register-range}
11997 @opindex mfixed-range
11998 Generate code treating the given register range as fixed registers.
11999 A fixed register is one that the register allocator can not use. This is
12000 useful when compiling kernel code. A register range is specified as
12001 two registers separated by a dash. Multiple register ranges can be
12002 specified separated by a comma.
12004 @item -mlong-load-store
12005 @opindex mlong-load-store
12006 Generate 3-instruction load and store sequences as sometimes required by
12007 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12010 @item -mportable-runtime
12011 @opindex mportable-runtime
12012 Use the portable calling conventions proposed by HP for ELF systems.
12016 Enable the use of assembler directives only GAS understands.
12018 @item -mschedule=@var{cpu-type}
12020 Schedule code according to the constraints for the machine type
12021 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12022 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12023 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12024 proper scheduling option for your machine. The default scheduling is
12028 @opindex mlinker-opt
12029 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12030 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12031 linkers in which they give bogus error messages when linking some programs.
12034 @opindex msoft-float
12035 Generate output containing library calls for floating point.
12036 @strong{Warning:} the requisite libraries are not available for all HPPA
12037 targets. Normally the facilities of the machine's usual C compiler are
12038 used, but this cannot be done directly in cross-compilation. You must make
12039 your own arrangements to provide suitable library functions for
12042 @option{-msoft-float} changes the calling convention in the output file;
12043 therefore, it is only useful if you compile @emph{all} of a program with
12044 this option. In particular, you need to compile @file{libgcc.a}, the
12045 library that comes with GCC, with @option{-msoft-float} in order for
12050 Generate the predefine, @code{_SIO}, for server IO@. The default is
12051 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12052 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12053 options are available under HP-UX and HI-UX@.
12057 Use GNU ld specific options. This passes @option{-shared} to ld when
12058 building a shared library. It is the default when GCC is configured,
12059 explicitly or implicitly, with the GNU linker. This option does not
12060 have any affect on which ld is called, it only changes what parameters
12061 are passed to that ld. The ld that is called is determined by the
12062 @option{--with-ld} configure option, GCC's program search path, and
12063 finally by the user's @env{PATH}. The linker used by GCC can be printed
12064 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12065 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12069 Use HP ld specific options. This passes @option{-b} to ld when building
12070 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12071 links. It is the default when GCC is configured, explicitly or
12072 implicitly, with the HP linker. This option does not have any affect on
12073 which ld is called, it only changes what parameters are passed to that
12074 ld. The ld that is called is determined by the @option{--with-ld}
12075 configure option, GCC's program search path, and finally by the user's
12076 @env{PATH}. The linker used by GCC can be printed using @samp{which
12077 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12078 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12081 @opindex mno-long-calls
12082 Generate code that uses long call sequences. This ensures that a call
12083 is always able to reach linker generated stubs. The default is to generate
12084 long calls only when the distance from the call site to the beginning
12085 of the function or translation unit, as the case may be, exceeds a
12086 predefined limit set by the branch type being used. The limits for
12087 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12088 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12091 Distances are measured from the beginning of functions when using the
12092 @option{-ffunction-sections} option, or when using the @option{-mgas}
12093 and @option{-mno-portable-runtime} options together under HP-UX with
12096 It is normally not desirable to use this option as it will degrade
12097 performance. However, it may be useful in large applications,
12098 particularly when partial linking is used to build the application.
12100 The types of long calls used depends on the capabilities of the
12101 assembler and linker, and the type of code being generated. The
12102 impact on systems that support long absolute calls, and long pic
12103 symbol-difference or pc-relative calls should be relatively small.
12104 However, an indirect call is used on 32-bit ELF systems in pic code
12105 and it is quite long.
12107 @item -munix=@var{unix-std}
12109 Generate compiler predefines and select a startfile for the specified
12110 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12111 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12112 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12113 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12114 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12117 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12118 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12119 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12120 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12121 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12122 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12124 It is @emph{important} to note that this option changes the interfaces
12125 for various library routines. It also affects the operational behavior
12126 of the C library. Thus, @emph{extreme} care is needed in using this
12129 Library code that is intended to operate with more than one UNIX
12130 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12131 as appropriate. Most GNU software doesn't provide this capability.
12135 Suppress the generation of link options to search libdld.sl when the
12136 @option{-static} option is specified on HP-UX 10 and later.
12140 The HP-UX implementation of setlocale in libc has a dependency on
12141 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12142 when the @option{-static} option is specified, special link options
12143 are needed to resolve this dependency.
12145 On HP-UX 10 and later, the GCC driver adds the necessary options to
12146 link with libdld.sl when the @option{-static} option is specified.
12147 This causes the resulting binary to be dynamic. On the 64-bit port,
12148 the linkers generate dynamic binaries by default in any case. The
12149 @option{-nolibdld} option can be used to prevent the GCC driver from
12150 adding these link options.
12154 Add support for multithreading with the @dfn{dce thread} library
12155 under HP-UX@. This option sets flags for both the preprocessor and
12159 @node i386 and x86-64 Options
12160 @subsection Intel 386 and AMD x86-64 Options
12161 @cindex i386 Options
12162 @cindex x86-64 Options
12163 @cindex Intel 386 Options
12164 @cindex AMD x86-64 Options
12166 These @samp{-m} options are defined for the i386 and x86-64 family of
12170 @item -mtune=@var{cpu-type}
12172 Tune to @var{cpu-type} everything applicable about the generated code, except
12173 for the ABI and the set of available instructions. The choices for
12174 @var{cpu-type} are:
12177 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12178 If you know the CPU on which your code will run, then you should use
12179 the corresponding @option{-mtune} option instead of
12180 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12181 of your application will have, then you should use this option.
12183 As new processors are deployed in the marketplace, the behavior of this
12184 option will change. Therefore, if you upgrade to a newer version of
12185 GCC, the code generated option will change to reflect the processors
12186 that were most common when that version of GCC was released.
12188 There is no @option{-march=generic} option because @option{-march}
12189 indicates the instruction set the compiler can use, and there is no
12190 generic instruction set applicable to all processors. In contrast,
12191 @option{-mtune} indicates the processor (or, in this case, collection of
12192 processors) for which the code is optimized.
12194 This selects the CPU to tune for at compilation time by determining
12195 the processor type of the compiling machine. Using @option{-mtune=native}
12196 will produce code optimized for the local machine under the constraints
12197 of the selected instruction set. Using @option{-march=native} will
12198 enable all instruction subsets supported by the local machine (hence
12199 the result might not run on different machines).
12201 Original Intel's i386 CPU@.
12203 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12204 @item i586, pentium
12205 Intel Pentium CPU with no MMX support.
12207 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12209 Intel PentiumPro CPU@.
12211 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12212 instruction set will be used, so the code will run on all i686 family chips.
12214 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12215 @item pentium3, pentium3m
12216 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12219 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12220 support. Used by Centrino notebooks.
12221 @item pentium4, pentium4m
12222 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12224 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12227 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12228 SSE2 and SSE3 instruction set support.
12230 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12231 instruction set support.
12233 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12234 and SSE4.2 instruction set support.
12236 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12237 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12239 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12240 instruction set support.
12242 AMD K6 CPU with MMX instruction set support.
12244 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12245 @item athlon, athlon-tbird
12246 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12248 @item athlon-4, athlon-xp, athlon-mp
12249 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12250 instruction set support.
12251 @item k8, opteron, athlon64, athlon-fx
12252 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12253 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12254 @item k8-sse3, opteron-sse3, athlon64-sse3
12255 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12256 @item amdfam10, barcelona
12257 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12258 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12259 instruction set extensions.)
12261 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12264 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12265 instruction set support.
12267 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12268 implemented for this chip.)
12270 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12271 implemented for this chip.)
12273 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12276 While picking a specific @var{cpu-type} will schedule things appropriately
12277 for that particular chip, the compiler will not generate any code that
12278 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12281 @item -march=@var{cpu-type}
12283 Generate instructions for the machine type @var{cpu-type}. The choices
12284 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12285 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12287 @item -mcpu=@var{cpu-type}
12289 A deprecated synonym for @option{-mtune}.
12291 @item -mfpmath=@var{unit}
12293 Generate floating point arithmetics for selected unit @var{unit}. The choices
12294 for @var{unit} are:
12298 Use the standard 387 floating point coprocessor present majority of chips and
12299 emulated otherwise. Code compiled with this option will run almost everywhere.
12300 The temporary results are computed in 80bit precision instead of precision
12301 specified by the type resulting in slightly different results compared to most
12302 of other chips. See @option{-ffloat-store} for more detailed description.
12304 This is the default choice for i386 compiler.
12307 Use scalar floating point instructions present in the SSE instruction set.
12308 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12309 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12310 instruction set supports only single precision arithmetics, thus the double and
12311 extended precision arithmetics is still done using 387. Later version, present
12312 only in Pentium4 and the future AMD x86-64 chips supports double precision
12315 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12316 or @option{-msse2} switches to enable SSE extensions and make this option
12317 effective. For the x86-64 compiler, these extensions are enabled by default.
12319 The resulting code should be considerably faster in the majority of cases and avoid
12320 the numerical instability problems of 387 code, but may break some existing
12321 code that expects temporaries to be 80bit.
12323 This is the default choice for the x86-64 compiler.
12328 Attempt to utilize both instruction sets at once. This effectively double the
12329 amount of available registers and on chips with separate execution units for
12330 387 and SSE the execution resources too. Use this option with care, as it is
12331 still experimental, because the GCC register allocator does not model separate
12332 functional units well resulting in instable performance.
12335 @item -masm=@var{dialect}
12336 @opindex masm=@var{dialect}
12337 Output asm instructions using selected @var{dialect}. Supported
12338 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12339 not support @samp{intel}.
12342 @itemx -mno-ieee-fp
12344 @opindex mno-ieee-fp
12345 Control whether or not the compiler uses IEEE floating point
12346 comparisons. These handle correctly the case where the result of a
12347 comparison is unordered.
12350 @opindex msoft-float
12351 Generate output containing library calls for floating point.
12352 @strong{Warning:} the requisite libraries are not part of GCC@.
12353 Normally the facilities of the machine's usual C compiler are used, but
12354 this can't be done directly in cross-compilation. You must make your
12355 own arrangements to provide suitable library functions for
12358 On machines where a function returns floating point results in the 80387
12359 register stack, some floating point opcodes may be emitted even if
12360 @option{-msoft-float} is used.
12362 @item -mno-fp-ret-in-387
12363 @opindex mno-fp-ret-in-387
12364 Do not use the FPU registers for return values of functions.
12366 The usual calling convention has functions return values of types
12367 @code{float} and @code{double} in an FPU register, even if there
12368 is no FPU@. The idea is that the operating system should emulate
12371 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12372 in ordinary CPU registers instead.
12374 @item -mno-fancy-math-387
12375 @opindex mno-fancy-math-387
12376 Some 387 emulators do not support the @code{sin}, @code{cos} and
12377 @code{sqrt} instructions for the 387. Specify this option to avoid
12378 generating those instructions. This option is the default on FreeBSD,
12379 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12380 indicates that the target CPU will always have an FPU and so the
12381 instruction will not need emulation. As of revision 2.6.1, these
12382 instructions are not generated unless you also use the
12383 @option{-funsafe-math-optimizations} switch.
12385 @item -malign-double
12386 @itemx -mno-align-double
12387 @opindex malign-double
12388 @opindex mno-align-double
12389 Control whether GCC aligns @code{double}, @code{long double}, and
12390 @code{long long} variables on a two word boundary or a one word
12391 boundary. Aligning @code{double} variables on a two word boundary will
12392 produce code that runs somewhat faster on a @samp{Pentium} at the
12393 expense of more memory.
12395 On x86-64, @option{-malign-double} is enabled by default.
12397 @strong{Warning:} if you use the @option{-malign-double} switch,
12398 structures containing the above types will be aligned differently than
12399 the published application binary interface specifications for the 386
12400 and will not be binary compatible with structures in code compiled
12401 without that switch.
12403 @item -m96bit-long-double
12404 @itemx -m128bit-long-double
12405 @opindex m96bit-long-double
12406 @opindex m128bit-long-double
12407 These switches control the size of @code{long double} type. The i386
12408 application binary interface specifies the size to be 96 bits,
12409 so @option{-m96bit-long-double} is the default in 32 bit mode.
12411 Modern architectures (Pentium and newer) would prefer @code{long double}
12412 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12413 conforming to the ABI, this would not be possible. So specifying a
12414 @option{-m128bit-long-double} will align @code{long double}
12415 to a 16 byte boundary by padding the @code{long double} with an additional
12418 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12419 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12421 Notice that neither of these options enable any extra precision over the x87
12422 standard of 80 bits for a @code{long double}.
12424 @strong{Warning:} if you override the default value for your target ABI, the
12425 structures and arrays containing @code{long double} variables will change
12426 their size as well as function calling convention for function taking
12427 @code{long double} will be modified. Hence they will not be binary
12428 compatible with arrays or structures in code compiled without that switch.
12430 @item -mlarge-data-threshold=@var{number}
12431 @opindex mlarge-data-threshold=@var{number}
12432 When @option{-mcmodel=medium} is specified, the data greater than
12433 @var{threshold} are placed in large data section. This value must be the
12434 same across all object linked into the binary and defaults to 65535.
12438 Use a different function-calling convention, in which functions that
12439 take a fixed number of arguments return with the @code{ret} @var{num}
12440 instruction, which pops their arguments while returning. This saves one
12441 instruction in the caller since there is no need to pop the arguments
12444 You can specify that an individual function is called with this calling
12445 sequence with the function attribute @samp{stdcall}. You can also
12446 override the @option{-mrtd} option by using the function attribute
12447 @samp{cdecl}. @xref{Function Attributes}.
12449 @strong{Warning:} this calling convention is incompatible with the one
12450 normally used on Unix, so you cannot use it if you need to call
12451 libraries compiled with the Unix compiler.
12453 Also, you must provide function prototypes for all functions that
12454 take variable numbers of arguments (including @code{printf});
12455 otherwise incorrect code will be generated for calls to those
12458 In addition, seriously incorrect code will result if you call a
12459 function with too many arguments. (Normally, extra arguments are
12460 harmlessly ignored.)
12462 @item -mregparm=@var{num}
12464 Control how many registers are used to pass integer arguments. By
12465 default, no registers are used to pass arguments, and at most 3
12466 registers can be used. You can control this behavior for a specific
12467 function by using the function attribute @samp{regparm}.
12468 @xref{Function Attributes}.
12470 @strong{Warning:} if you use this switch, and
12471 @var{num} is nonzero, then you must build all modules with the same
12472 value, including any libraries. This includes the system libraries and
12476 @opindex msseregparm
12477 Use SSE register passing conventions for float and double arguments
12478 and return values. You can control this behavior for a specific
12479 function by using the function attribute @samp{sseregparm}.
12480 @xref{Function Attributes}.
12482 @strong{Warning:} if you use this switch then you must build all
12483 modules with the same value, including any libraries. This includes
12484 the system libraries and startup modules.
12486 @item -mvect8-ret-in-mem
12487 @opindex mvect8-ret-in-mem
12488 Return 8-byte vectors in memory instead of MMX registers. This is the
12489 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12490 Studio compilers until version 12. Later compiler versions (starting
12491 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12492 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12493 you need to remain compatible with existing code produced by those
12494 previous compiler versions or older versions of GCC.
12503 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12504 is specified, the significands of results of floating-point operations are
12505 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12506 significands of results of floating-point operations to 53 bits (double
12507 precision) and @option{-mpc80} rounds the significands of results of
12508 floating-point operations to 64 bits (extended double precision), which is
12509 the default. When this option is used, floating-point operations in higher
12510 precisions are not available to the programmer without setting the FPU
12511 control word explicitly.
12513 Setting the rounding of floating-point operations to less than the default
12514 80 bits can speed some programs by 2% or more. Note that some mathematical
12515 libraries assume that extended precision (80 bit) floating-point operations
12516 are enabled by default; routines in such libraries could suffer significant
12517 loss of accuracy, typically through so-called "catastrophic cancellation",
12518 when this option is used to set the precision to less than extended precision.
12520 @item -mstackrealign
12521 @opindex mstackrealign
12522 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12523 option will generate an alternate prologue and epilogue that realigns the
12524 runtime stack if necessary. This supports mixing legacy codes that keep
12525 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12526 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12527 applicable to individual functions.
12529 @item -mpreferred-stack-boundary=@var{num}
12530 @opindex mpreferred-stack-boundary
12531 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12532 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12533 the default is 4 (16 bytes or 128 bits).
12535 @item -mincoming-stack-boundary=@var{num}
12536 @opindex mincoming-stack-boundary
12537 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12538 boundary. If @option{-mincoming-stack-boundary} is not specified,
12539 the one specified by @option{-mpreferred-stack-boundary} will be used.
12541 On Pentium and PentiumPro, @code{double} and @code{long double} values
12542 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12543 suffer significant run time performance penalties. On Pentium III, the
12544 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12545 properly if it is not 16 byte aligned.
12547 To ensure proper alignment of this values on the stack, the stack boundary
12548 must be as aligned as that required by any value stored on the stack.
12549 Further, every function must be generated such that it keeps the stack
12550 aligned. Thus calling a function compiled with a higher preferred
12551 stack boundary from a function compiled with a lower preferred stack
12552 boundary will most likely misalign the stack. It is recommended that
12553 libraries that use callbacks always use the default setting.
12555 This extra alignment does consume extra stack space, and generally
12556 increases code size. Code that is sensitive to stack space usage, such
12557 as embedded systems and operating system kernels, may want to reduce the
12558 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12585 @itemx -mno-fsgsbase
12617 These switches enable or disable the use of instructions in the MMX, SSE,
12618 SSE2, SSE3, SSSE3, SSE4.1, AVX, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA,
12619 SSE4A, FMA4, XOP, LWP, ABM, BMI, or 3DNow!@: extended instruction sets.
12620 These extensions are also available as built-in functions: see
12621 @ref{X86 Built-in Functions}, for details of the functions enabled and
12622 disabled by these switches.
12624 To have SSE/SSE2 instructions generated automatically from floating-point
12625 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12627 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12628 generates new AVX instructions or AVX equivalence for all SSEx instructions
12631 These options will enable GCC to use these extended instructions in
12632 generated code, even without @option{-mfpmath=sse}. Applications which
12633 perform runtime CPU detection must compile separate files for each
12634 supported architecture, using the appropriate flags. In particular,
12635 the file containing the CPU detection code should be compiled without
12640 This option instructs GCC to emit a @code{cld} instruction in the prologue
12641 of functions that use string instructions. String instructions depend on
12642 the DF flag to select between autoincrement or autodecrement mode. While the
12643 ABI specifies the DF flag to be cleared on function entry, some operating
12644 systems violate this specification by not clearing the DF flag in their
12645 exception dispatchers. The exception handler can be invoked with the DF flag
12646 set which leads to wrong direction mode, when string instructions are used.
12647 This option can be enabled by default on 32-bit x86 targets by configuring
12648 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12649 instructions can be suppressed with the @option{-mno-cld} compiler option
12653 @opindex mvzeroupper
12654 This option instructs GCC to emit a @code{vzeroupper} instruction
12655 before a transfer of control flow out of the function to minimize
12656 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12661 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12662 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12663 data types. This is useful for high resolution counters that could be updated
12664 by multiple processors (or cores). This instruction is generated as part of
12665 atomic built-in functions: see @ref{Atomic Builtins} for details.
12669 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12670 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12671 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12672 SAHF are load and store instructions, respectively, for certain status flags.
12673 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12674 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12678 This option will enable GCC to use movbe instruction to implement
12679 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12683 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12684 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12685 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12689 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12690 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12691 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12692 variants) for single precision floating point arguments. These instructions
12693 are generated only when @option{-funsafe-math-optimizations} is enabled
12694 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12695 Note that while the throughput of the sequence is higher than the throughput
12696 of the non-reciprocal instruction, the precision of the sequence can be
12697 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12699 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12700 already with @option{-ffast-math} (or the above option combination), and
12701 doesn't need @option{-mrecip}.
12703 @item -mveclibabi=@var{type}
12704 @opindex mveclibabi
12705 Specifies the ABI type to use for vectorizing intrinsics using an
12706 external library. Supported types are @code{svml} for the Intel short
12707 vector math library and @code{acml} for the AMD math core library style
12708 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12709 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12710 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12711 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12712 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12713 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12714 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12715 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12716 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12717 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12718 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12719 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12720 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12721 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12722 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12723 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12724 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12725 compatible library will have to be specified at link time.
12727 @item -mabi=@var{name}
12729 Generate code for the specified calling convention. Permissible values
12730 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12731 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12732 ABI when targeting Windows. On all other systems, the default is the
12733 SYSV ABI. You can control this behavior for a specific function by
12734 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12735 @xref{Function Attributes}.
12737 @item -mtls-dialect=@var{type}
12738 @opindex mtls-dialect
12739 Generate code to access thread-local storage using the @samp{gnu} or
12740 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
12741 @samp{gnu2} is more efficient, but it may add compile- and run-time
12742 requirements that cannot be satisfied on all systems.
12745 @itemx -mno-push-args
12746 @opindex mpush-args
12747 @opindex mno-push-args
12748 Use PUSH operations to store outgoing parameters. This method is shorter
12749 and usually equally fast as method using SUB/MOV operations and is enabled
12750 by default. In some cases disabling it may improve performance because of
12751 improved scheduling and reduced dependencies.
12753 @item -maccumulate-outgoing-args
12754 @opindex maccumulate-outgoing-args
12755 If enabled, the maximum amount of space required for outgoing arguments will be
12756 computed in the function prologue. This is faster on most modern CPUs
12757 because of reduced dependencies, improved scheduling and reduced stack usage
12758 when preferred stack boundary is not equal to 2. The drawback is a notable
12759 increase in code size. This switch implies @option{-mno-push-args}.
12763 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12764 on thread-safe exception handling must compile and link all code with the
12765 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12766 @option{-D_MT}; when linking, it links in a special thread helper library
12767 @option{-lmingwthrd} which cleans up per thread exception handling data.
12769 @item -mno-align-stringops
12770 @opindex mno-align-stringops
12771 Do not align destination of inlined string operations. This switch reduces
12772 code size and improves performance in case the destination is already aligned,
12773 but GCC doesn't know about it.
12775 @item -minline-all-stringops
12776 @opindex minline-all-stringops
12777 By default GCC inlines string operations only when destination is known to be
12778 aligned at least to 4 byte boundary. This enables more inlining, increase code
12779 size, but may improve performance of code that depends on fast memcpy, strlen
12780 and memset for short lengths.
12782 @item -minline-stringops-dynamically
12783 @opindex minline-stringops-dynamically
12784 For string operation of unknown size, inline runtime checks so for small
12785 blocks inline code is used, while for large blocks library call is used.
12787 @item -mstringop-strategy=@var{alg}
12788 @opindex mstringop-strategy=@var{alg}
12789 Overwrite internal decision heuristic about particular algorithm to inline
12790 string operation with. The allowed values are @code{rep_byte},
12791 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12792 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12793 expanding inline loop, @code{libcall} for always expanding library call.
12795 @item -momit-leaf-frame-pointer
12796 @opindex momit-leaf-frame-pointer
12797 Don't keep the frame pointer in a register for leaf functions. This
12798 avoids the instructions to save, set up and restore frame pointers and
12799 makes an extra register available in leaf functions. The option
12800 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12801 which might make debugging harder.
12803 @item -mtls-direct-seg-refs
12804 @itemx -mno-tls-direct-seg-refs
12805 @opindex mtls-direct-seg-refs
12806 Controls whether TLS variables may be accessed with offsets from the
12807 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12808 or whether the thread base pointer must be added. Whether or not this
12809 is legal depends on the operating system, and whether it maps the
12810 segment to cover the entire TLS area.
12812 For systems that use GNU libc, the default is on.
12815 @itemx -mno-sse2avx
12817 Specify that the assembler should encode SSE instructions with VEX
12818 prefix. The option @option{-mavx} turns this on by default.
12823 If profiling is active @option{-pg} put the profiling
12824 counter call before prologue.
12825 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12826 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12829 @itemx -mno-8bit-idiv
12831 On some processors, like Intel Atom, 8bit unsigned integer divide is
12832 much faster than 32bit/64bit integer divide. This option will generate a
12833 runt-time check. If both dividend and divisor are within range of 0
12834 to 255, 8bit unsigned integer divide will be used instead of
12835 32bit/64bit integer divide.
12837 @item -mavx256-split-unaligned-load
12838 @item -mavx256-split-unaligned-store
12839 @opindex avx256-split-unaligned-load
12840 @opindex avx256-split-unaligned-store
12841 Split 32-byte AVX unaligned load and store.
12845 These @samp{-m} switches are supported in addition to the above
12846 on AMD x86-64 processors in 64-bit environments.
12855 Generate code for a 32-bit or 64-bit environment.
12856 The @option{-m32} option sets int, long and pointer to 32 bits and
12857 generates code that runs on any i386 system.
12858 The @option{-m64} option sets int to 32 bits and long and pointer
12859 to 64 bits and generates code for AMD's x86-64 architecture.
12860 The @option{-mx32} option sets int, long and pointer to 32 bits and
12861 generates code for AMD's x86-64 architecture.
12862 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
12863 and @option{-mdynamic-no-pic} options.
12865 @item -mno-red-zone
12866 @opindex mno-red-zone
12867 Do not use a so called red zone for x86-64 code. The red zone is mandated
12868 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12869 stack pointer that will not be modified by signal or interrupt handlers
12870 and therefore can be used for temporary data without adjusting the stack
12871 pointer. The flag @option{-mno-red-zone} disables this red zone.
12873 @item -mcmodel=small
12874 @opindex mcmodel=small
12875 Generate code for the small code model: the program and its symbols must
12876 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12877 Programs can be statically or dynamically linked. This is the default
12880 @item -mcmodel=kernel
12881 @opindex mcmodel=kernel
12882 Generate code for the kernel code model. The kernel runs in the
12883 negative 2 GB of the address space.
12884 This model has to be used for Linux kernel code.
12886 @item -mcmodel=medium
12887 @opindex mcmodel=medium
12888 Generate code for the medium model: The program is linked in the lower 2
12889 GB of the address space. Small symbols are also placed there. Symbols
12890 with sizes larger than @option{-mlarge-data-threshold} are put into
12891 large data or bss sections and can be located above 2GB. Programs can
12892 be statically or dynamically linked.
12894 @item -mcmodel=large
12895 @opindex mcmodel=large
12896 Generate code for the large model: This model makes no assumptions
12897 about addresses and sizes of sections.
12900 @node i386 and x86-64 Windows Options
12901 @subsection i386 and x86-64 Windows Options
12902 @cindex i386 and x86-64 Windows Options
12904 These additional options are available for Windows targets:
12909 This option is available for Cygwin and MinGW targets. It
12910 specifies that a console application is to be generated, by
12911 instructing the linker to set the PE header subsystem type
12912 required for console applications.
12913 This is the default behavior for Cygwin and MinGW targets.
12917 This option is available for Cygwin and MinGW targets. It
12918 specifies that a DLL - a dynamic link library - is to be
12919 generated, enabling the selection of the required runtime
12920 startup object and entry point.
12922 @item -mnop-fun-dllimport
12923 @opindex mnop-fun-dllimport
12924 This option is available for Cygwin and MinGW targets. It
12925 specifies that the dllimport attribute should be ignored.
12929 This option is available for MinGW targets. It specifies
12930 that MinGW-specific thread support is to be used.
12934 This option is available for mingw-w64 targets. It specifies
12935 that the UNICODE macro is getting pre-defined and that the
12936 unicode capable runtime startup code is chosen.
12940 This option is available for Cygwin and MinGW targets. It
12941 specifies that the typical Windows pre-defined macros are to
12942 be set in the pre-processor, but does not influence the choice
12943 of runtime library/startup code.
12947 This option is available for Cygwin and MinGW targets. It
12948 specifies that a GUI application is to be generated by
12949 instructing the linker to set the PE header subsystem type
12952 @item -fno-set-stack-executable
12953 @opindex fno-set-stack-executable
12954 This option is available for MinGW targets. It specifies that
12955 the executable flag for stack used by nested functions isn't
12956 set. This is necessary for binaries running in kernel mode of
12957 Windows, as there the user32 API, which is used to set executable
12958 privileges, isn't available.
12960 @item -mpe-aligned-commons
12961 @opindex mpe-aligned-commons
12962 This option is available for Cygwin and MinGW targets. It
12963 specifies that the GNU extension to the PE file format that
12964 permits the correct alignment of COMMON variables should be
12965 used when generating code. It will be enabled by default if
12966 GCC detects that the target assembler found during configuration
12967 supports the feature.
12970 See also under @ref{i386 and x86-64 Options} for standard options.
12972 @node IA-64 Options
12973 @subsection IA-64 Options
12974 @cindex IA-64 Options
12976 These are the @samp{-m} options defined for the Intel IA-64 architecture.
12980 @opindex mbig-endian
12981 Generate code for a big endian target. This is the default for HP-UX@.
12983 @item -mlittle-endian
12984 @opindex mlittle-endian
12985 Generate code for a little endian target. This is the default for AIX5
12991 @opindex mno-gnu-as
12992 Generate (or don't) code for the GNU assembler. This is the default.
12993 @c Also, this is the default if the configure option @option{--with-gnu-as}
12999 @opindex mno-gnu-ld
13000 Generate (or don't) code for the GNU linker. This is the default.
13001 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13006 Generate code that does not use a global pointer register. The result
13007 is not position independent code, and violates the IA-64 ABI@.
13009 @item -mvolatile-asm-stop
13010 @itemx -mno-volatile-asm-stop
13011 @opindex mvolatile-asm-stop
13012 @opindex mno-volatile-asm-stop
13013 Generate (or don't) a stop bit immediately before and after volatile asm
13016 @item -mregister-names
13017 @itemx -mno-register-names
13018 @opindex mregister-names
13019 @opindex mno-register-names
13020 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13021 the stacked registers. This may make assembler output more readable.
13027 Disable (or enable) optimizations that use the small data section. This may
13028 be useful for working around optimizer bugs.
13030 @item -mconstant-gp
13031 @opindex mconstant-gp
13032 Generate code that uses a single constant global pointer value. This is
13033 useful when compiling kernel code.
13037 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13038 This is useful when compiling firmware code.
13040 @item -minline-float-divide-min-latency
13041 @opindex minline-float-divide-min-latency
13042 Generate code for inline divides of floating point values
13043 using the minimum latency algorithm.
13045 @item -minline-float-divide-max-throughput
13046 @opindex minline-float-divide-max-throughput
13047 Generate code for inline divides of floating point values
13048 using the maximum throughput algorithm.
13050 @item -mno-inline-float-divide
13051 @opindex mno-inline-float-divide
13052 Do not generate inline code for divides of floating point values.
13054 @item -minline-int-divide-min-latency
13055 @opindex minline-int-divide-min-latency
13056 Generate code for inline divides of integer values
13057 using the minimum latency algorithm.
13059 @item -minline-int-divide-max-throughput
13060 @opindex minline-int-divide-max-throughput
13061 Generate code for inline divides of integer values
13062 using the maximum throughput algorithm.
13064 @item -mno-inline-int-divide
13065 @opindex mno-inline-int-divide
13066 Do not generate inline code for divides of integer values.
13068 @item -minline-sqrt-min-latency
13069 @opindex minline-sqrt-min-latency
13070 Generate code for inline square roots
13071 using the minimum latency algorithm.
13073 @item -minline-sqrt-max-throughput
13074 @opindex minline-sqrt-max-throughput
13075 Generate code for inline square roots
13076 using the maximum throughput algorithm.
13078 @item -mno-inline-sqrt
13079 @opindex mno-inline-sqrt
13080 Do not generate inline code for sqrt.
13083 @itemx -mno-fused-madd
13084 @opindex mfused-madd
13085 @opindex mno-fused-madd
13086 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13087 instructions. The default is to use these instructions.
13089 @item -mno-dwarf2-asm
13090 @itemx -mdwarf2-asm
13091 @opindex mno-dwarf2-asm
13092 @opindex mdwarf2-asm
13093 Don't (or do) generate assembler code for the DWARF2 line number debugging
13094 info. This may be useful when not using the GNU assembler.
13096 @item -mearly-stop-bits
13097 @itemx -mno-early-stop-bits
13098 @opindex mearly-stop-bits
13099 @opindex mno-early-stop-bits
13100 Allow stop bits to be placed earlier than immediately preceding the
13101 instruction that triggered the stop bit. This can improve instruction
13102 scheduling, but does not always do so.
13104 @item -mfixed-range=@var{register-range}
13105 @opindex mfixed-range
13106 Generate code treating the given register range as fixed registers.
13107 A fixed register is one that the register allocator can not use. This is
13108 useful when compiling kernel code. A register range is specified as
13109 two registers separated by a dash. Multiple register ranges can be
13110 specified separated by a comma.
13112 @item -mtls-size=@var{tls-size}
13114 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13117 @item -mtune=@var{cpu-type}
13119 Tune the instruction scheduling for a particular CPU, Valid values are
13120 itanium, itanium1, merced, itanium2, and mckinley.
13126 Generate code for a 32-bit or 64-bit environment.
13127 The 32-bit environment sets int, long and pointer to 32 bits.
13128 The 64-bit environment sets int to 32 bits and long and pointer
13129 to 64 bits. These are HP-UX specific flags.
13131 @item -mno-sched-br-data-spec
13132 @itemx -msched-br-data-spec
13133 @opindex mno-sched-br-data-spec
13134 @opindex msched-br-data-spec
13135 (Dis/En)able data speculative scheduling before reload.
13136 This will result in generation of the ld.a instructions and
13137 the corresponding check instructions (ld.c / chk.a).
13138 The default is 'disable'.
13140 @item -msched-ar-data-spec
13141 @itemx -mno-sched-ar-data-spec
13142 @opindex msched-ar-data-spec
13143 @opindex mno-sched-ar-data-spec
13144 (En/Dis)able data speculative scheduling after reload.
13145 This will result in generation of the ld.a instructions and
13146 the corresponding check instructions (ld.c / chk.a).
13147 The default is 'enable'.
13149 @item -mno-sched-control-spec
13150 @itemx -msched-control-spec
13151 @opindex mno-sched-control-spec
13152 @opindex msched-control-spec
13153 (Dis/En)able control speculative scheduling. This feature is
13154 available only during region scheduling (i.e.@: before reload).
13155 This will result in generation of the ld.s instructions and
13156 the corresponding check instructions chk.s .
13157 The default is 'disable'.
13159 @item -msched-br-in-data-spec
13160 @itemx -mno-sched-br-in-data-spec
13161 @opindex msched-br-in-data-spec
13162 @opindex mno-sched-br-in-data-spec
13163 (En/Dis)able speculative scheduling of the instructions that
13164 are dependent on the data speculative loads before reload.
13165 This is effective only with @option{-msched-br-data-spec} enabled.
13166 The default is 'enable'.
13168 @item -msched-ar-in-data-spec
13169 @itemx -mno-sched-ar-in-data-spec
13170 @opindex msched-ar-in-data-spec
13171 @opindex mno-sched-ar-in-data-spec
13172 (En/Dis)able speculative scheduling of the instructions that
13173 are dependent on the data speculative loads after reload.
13174 This is effective only with @option{-msched-ar-data-spec} enabled.
13175 The default is 'enable'.
13177 @item -msched-in-control-spec
13178 @itemx -mno-sched-in-control-spec
13179 @opindex msched-in-control-spec
13180 @opindex mno-sched-in-control-spec
13181 (En/Dis)able speculative scheduling of the instructions that
13182 are dependent on the control speculative loads.
13183 This is effective only with @option{-msched-control-spec} enabled.
13184 The default is 'enable'.
13186 @item -mno-sched-prefer-non-data-spec-insns
13187 @itemx -msched-prefer-non-data-spec-insns
13188 @opindex mno-sched-prefer-non-data-spec-insns
13189 @opindex msched-prefer-non-data-spec-insns
13190 If enabled, data speculative instructions will be chosen for schedule
13191 only if there are no other choices at the moment. This will make
13192 the use of the data speculation much more conservative.
13193 The default is 'disable'.
13195 @item -mno-sched-prefer-non-control-spec-insns
13196 @itemx -msched-prefer-non-control-spec-insns
13197 @opindex mno-sched-prefer-non-control-spec-insns
13198 @opindex msched-prefer-non-control-spec-insns
13199 If enabled, control speculative instructions will be chosen for schedule
13200 only if there are no other choices at the moment. This will make
13201 the use of the control speculation much more conservative.
13202 The default is 'disable'.
13204 @item -mno-sched-count-spec-in-critical-path
13205 @itemx -msched-count-spec-in-critical-path
13206 @opindex mno-sched-count-spec-in-critical-path
13207 @opindex msched-count-spec-in-critical-path
13208 If enabled, speculative dependencies will be considered during
13209 computation of the instructions priorities. This will make the use of the
13210 speculation a bit more conservative.
13211 The default is 'disable'.
13213 @item -msched-spec-ldc
13214 @opindex msched-spec-ldc
13215 Use a simple data speculation check. This option is on by default.
13217 @item -msched-control-spec-ldc
13218 @opindex msched-spec-ldc
13219 Use a simple check for control speculation. This option is on by default.
13221 @item -msched-stop-bits-after-every-cycle
13222 @opindex msched-stop-bits-after-every-cycle
13223 Place a stop bit after every cycle when scheduling. This option is on
13226 @item -msched-fp-mem-deps-zero-cost
13227 @opindex msched-fp-mem-deps-zero-cost
13228 Assume that floating-point stores and loads are not likely to cause a conflict
13229 when placed into the same instruction group. This option is disabled by
13232 @item -msel-sched-dont-check-control-spec
13233 @opindex msel-sched-dont-check-control-spec
13234 Generate checks for control speculation in selective scheduling.
13235 This flag is disabled by default.
13237 @item -msched-max-memory-insns=@var{max-insns}
13238 @opindex msched-max-memory-insns
13239 Limit on the number of memory insns per instruction group, giving lower
13240 priority to subsequent memory insns attempting to schedule in the same
13241 instruction group. Frequently useful to prevent cache bank conflicts.
13242 The default value is 1.
13244 @item -msched-max-memory-insns-hard-limit
13245 @opindex msched-max-memory-insns-hard-limit
13246 Disallow more than `msched-max-memory-insns' in instruction group.
13247 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13248 when limit is reached but may still schedule memory operations.
13252 @node IA-64/VMS Options
13253 @subsection IA-64/VMS Options
13255 These @samp{-m} options are defined for the IA-64/VMS implementations:
13258 @item -mvms-return-codes
13259 @opindex mvms-return-codes
13260 Return VMS condition codes from main. The default is to return POSIX
13261 style condition (e.g.@ error) codes.
13263 @item -mdebug-main=@var{prefix}
13264 @opindex mdebug-main=@var{prefix}
13265 Flag the first routine whose name starts with @var{prefix} as the main
13266 routine for the debugger.
13270 Default to 64bit memory allocation routines.
13274 @subsection LM32 Options
13275 @cindex LM32 options
13277 These @option{-m} options are defined for the Lattice Mico32 architecture:
13280 @item -mbarrel-shift-enabled
13281 @opindex mbarrel-shift-enabled
13282 Enable barrel-shift instructions.
13284 @item -mdivide-enabled
13285 @opindex mdivide-enabled
13286 Enable divide and modulus instructions.
13288 @item -mmultiply-enabled
13289 @opindex multiply-enabled
13290 Enable multiply instructions.
13292 @item -msign-extend-enabled
13293 @opindex msign-extend-enabled
13294 Enable sign extend instructions.
13296 @item -muser-enabled
13297 @opindex muser-enabled
13298 Enable user-defined instructions.
13303 @subsection M32C Options
13304 @cindex M32C options
13307 @item -mcpu=@var{name}
13309 Select the CPU for which code is generated. @var{name} may be one of
13310 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13311 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13312 the M32C/80 series.
13316 Specifies that the program will be run on the simulator. This causes
13317 an alternate runtime library to be linked in which supports, for
13318 example, file I/O@. You must not use this option when generating
13319 programs that will run on real hardware; you must provide your own
13320 runtime library for whatever I/O functions are needed.
13322 @item -memregs=@var{number}
13324 Specifies the number of memory-based pseudo-registers GCC will use
13325 during code generation. These pseudo-registers will be used like real
13326 registers, so there is a tradeoff between GCC's ability to fit the
13327 code into available registers, and the performance penalty of using
13328 memory instead of registers. Note that all modules in a program must
13329 be compiled with the same value for this option. Because of that, you
13330 must not use this option with the default runtime libraries gcc
13335 @node M32R/D Options
13336 @subsection M32R/D Options
13337 @cindex M32R/D options
13339 These @option{-m} options are defined for Renesas M32R/D architectures:
13344 Generate code for the M32R/2@.
13348 Generate code for the M32R/X@.
13352 Generate code for the M32R@. This is the default.
13354 @item -mmodel=small
13355 @opindex mmodel=small
13356 Assume all objects live in the lower 16MB of memory (so that their addresses
13357 can be loaded with the @code{ld24} instruction), and assume all subroutines
13358 are reachable with the @code{bl} instruction.
13359 This is the default.
13361 The addressability of a particular object can be set with the
13362 @code{model} attribute.
13364 @item -mmodel=medium
13365 @opindex mmodel=medium
13366 Assume objects may be anywhere in the 32-bit address space (the compiler
13367 will generate @code{seth/add3} instructions to load their addresses), and
13368 assume all subroutines are reachable with the @code{bl} instruction.
13370 @item -mmodel=large
13371 @opindex mmodel=large
13372 Assume objects may be anywhere in the 32-bit address space (the compiler
13373 will generate @code{seth/add3} instructions to load their addresses), and
13374 assume subroutines may not be reachable with the @code{bl} instruction
13375 (the compiler will generate the much slower @code{seth/add3/jl}
13376 instruction sequence).
13379 @opindex msdata=none
13380 Disable use of the small data area. Variables will be put into
13381 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13382 @code{section} attribute has been specified).
13383 This is the default.
13385 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13386 Objects may be explicitly put in the small data area with the
13387 @code{section} attribute using one of these sections.
13389 @item -msdata=sdata
13390 @opindex msdata=sdata
13391 Put small global and static data in the small data area, but do not
13392 generate special code to reference them.
13395 @opindex msdata=use
13396 Put small global and static data in the small data area, and generate
13397 special instructions to reference them.
13401 @cindex smaller data references
13402 Put global and static objects less than or equal to @var{num} bytes
13403 into the small data or bss sections instead of the normal data or bss
13404 sections. The default value of @var{num} is 8.
13405 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13406 for this option to have any effect.
13408 All modules should be compiled with the same @option{-G @var{num}} value.
13409 Compiling with different values of @var{num} may or may not work; if it
13410 doesn't the linker will give an error message---incorrect code will not be
13415 Makes the M32R specific code in the compiler display some statistics
13416 that might help in debugging programs.
13418 @item -malign-loops
13419 @opindex malign-loops
13420 Align all loops to a 32-byte boundary.
13422 @item -mno-align-loops
13423 @opindex mno-align-loops
13424 Do not enforce a 32-byte alignment for loops. This is the default.
13426 @item -missue-rate=@var{number}
13427 @opindex missue-rate=@var{number}
13428 Issue @var{number} instructions per cycle. @var{number} can only be 1
13431 @item -mbranch-cost=@var{number}
13432 @opindex mbranch-cost=@var{number}
13433 @var{number} can only be 1 or 2. If it is 1 then branches will be
13434 preferred over conditional code, if it is 2, then the opposite will
13437 @item -mflush-trap=@var{number}
13438 @opindex mflush-trap=@var{number}
13439 Specifies the trap number to use to flush the cache. The default is
13440 12. Valid numbers are between 0 and 15 inclusive.
13442 @item -mno-flush-trap
13443 @opindex mno-flush-trap
13444 Specifies that the cache cannot be flushed by using a trap.
13446 @item -mflush-func=@var{name}
13447 @opindex mflush-func=@var{name}
13448 Specifies the name of the operating system function to call to flush
13449 the cache. The default is @emph{_flush_cache}, but a function call
13450 will only be used if a trap is not available.
13452 @item -mno-flush-func
13453 @opindex mno-flush-func
13454 Indicates that there is no OS function for flushing the cache.
13458 @node M680x0 Options
13459 @subsection M680x0 Options
13460 @cindex M680x0 options
13462 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13463 The default settings depend on which architecture was selected when
13464 the compiler was configured; the defaults for the most common choices
13468 @item -march=@var{arch}
13470 Generate code for a specific M680x0 or ColdFire instruction set
13471 architecture. Permissible values of @var{arch} for M680x0
13472 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13473 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13474 architectures are selected according to Freescale's ISA classification
13475 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13476 @samp{isab} and @samp{isac}.
13478 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13479 code for a ColdFire target. The @var{arch} in this macro is one of the
13480 @option{-march} arguments given above.
13482 When used together, @option{-march} and @option{-mtune} select code
13483 that runs on a family of similar processors but that is optimized
13484 for a particular microarchitecture.
13486 @item -mcpu=@var{cpu}
13488 Generate code for a specific M680x0 or ColdFire processor.
13489 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13490 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13491 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13492 below, which also classifies the CPUs into families:
13494 @multitable @columnfractions 0.20 0.80
13495 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13496 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13497 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13498 @item @samp{5206e} @tab @samp{5206e}
13499 @item @samp{5208} @tab @samp{5207} @samp{5208}
13500 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13501 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13502 @item @samp{5216} @tab @samp{5214} @samp{5216}
13503 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13504 @item @samp{5225} @tab @samp{5224} @samp{5225}
13505 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13506 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13507 @item @samp{5249} @tab @samp{5249}
13508 @item @samp{5250} @tab @samp{5250}
13509 @item @samp{5271} @tab @samp{5270} @samp{5271}
13510 @item @samp{5272} @tab @samp{5272}
13511 @item @samp{5275} @tab @samp{5274} @samp{5275}
13512 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13513 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13514 @item @samp{5307} @tab @samp{5307}
13515 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13516 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13517 @item @samp{5407} @tab @samp{5407}
13518 @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}
13521 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13522 @var{arch} is compatible with @var{cpu}. Other combinations of
13523 @option{-mcpu} and @option{-march} are rejected.
13525 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13526 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13527 where the value of @var{family} is given by the table above.
13529 @item -mtune=@var{tune}
13531 Tune the code for a particular microarchitecture, within the
13532 constraints set by @option{-march} and @option{-mcpu}.
13533 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13534 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13535 and @samp{cpu32}. The ColdFire microarchitectures
13536 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13538 You can also use @option{-mtune=68020-40} for code that needs
13539 to run relatively well on 68020, 68030 and 68040 targets.
13540 @option{-mtune=68020-60} is similar but includes 68060 targets
13541 as well. These two options select the same tuning decisions as
13542 @option{-m68020-40} and @option{-m68020-60} respectively.
13544 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13545 when tuning for 680x0 architecture @var{arch}. It also defines
13546 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13547 option is used. If gcc is tuning for a range of architectures,
13548 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13549 it defines the macros for every architecture in the range.
13551 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13552 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13553 of the arguments given above.
13559 Generate output for a 68000. This is the default
13560 when the compiler is configured for 68000-based systems.
13561 It is equivalent to @option{-march=68000}.
13563 Use this option for microcontrollers with a 68000 or EC000 core,
13564 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13568 Generate output for a 68010. This is the default
13569 when the compiler is configured for 68010-based systems.
13570 It is equivalent to @option{-march=68010}.
13576 Generate output for a 68020. This is the default
13577 when the compiler is configured for 68020-based systems.
13578 It is equivalent to @option{-march=68020}.
13582 Generate output for a 68030. This is the default when the compiler is
13583 configured for 68030-based systems. It is equivalent to
13584 @option{-march=68030}.
13588 Generate output for a 68040. This is the default when the compiler is
13589 configured for 68040-based systems. It is equivalent to
13590 @option{-march=68040}.
13592 This option inhibits the use of 68881/68882 instructions that have to be
13593 emulated by software on the 68040. Use this option if your 68040 does not
13594 have code to emulate those instructions.
13598 Generate output for a 68060. This is the default when the compiler is
13599 configured for 68060-based systems. It is equivalent to
13600 @option{-march=68060}.
13602 This option inhibits the use of 68020 and 68881/68882 instructions that
13603 have to be emulated by software on the 68060. Use this option if your 68060
13604 does not have code to emulate those instructions.
13608 Generate output for a CPU32. This is the default
13609 when the compiler is configured for CPU32-based systems.
13610 It is equivalent to @option{-march=cpu32}.
13612 Use this option for microcontrollers with a
13613 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13614 68336, 68340, 68341, 68349 and 68360.
13618 Generate output for a 520X ColdFire CPU@. This is the default
13619 when the compiler is configured for 520X-based systems.
13620 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13621 in favor of that option.
13623 Use this option for microcontroller with a 5200 core, including
13624 the MCF5202, MCF5203, MCF5204 and MCF5206.
13628 Generate output for a 5206e ColdFire CPU@. The option is now
13629 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13633 Generate output for a member of the ColdFire 528X family.
13634 The option is now deprecated in favor of the equivalent
13635 @option{-mcpu=528x}.
13639 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13640 in favor of the equivalent @option{-mcpu=5307}.
13644 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13645 in favor of the equivalent @option{-mcpu=5407}.
13649 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13650 This includes use of hardware floating point instructions.
13651 The option is equivalent to @option{-mcpu=547x}, and is now
13652 deprecated in favor of that option.
13656 Generate output for a 68040, without using any of the new instructions.
13657 This results in code which can run relatively efficiently on either a
13658 68020/68881 or a 68030 or a 68040. The generated code does use the
13659 68881 instructions that are emulated on the 68040.
13661 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13665 Generate output for a 68060, without using any of the new instructions.
13666 This results in code which can run relatively efficiently on either a
13667 68020/68881 or a 68030 or a 68040. The generated code does use the
13668 68881 instructions that are emulated on the 68060.
13670 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13674 @opindex mhard-float
13676 Generate floating-point instructions. This is the default for 68020
13677 and above, and for ColdFire devices that have an FPU@. It defines the
13678 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13679 on ColdFire targets.
13682 @opindex msoft-float
13683 Do not generate floating-point instructions; use library calls instead.
13684 This is the default for 68000, 68010, and 68832 targets. It is also
13685 the default for ColdFire devices that have no FPU.
13691 Generate (do not generate) ColdFire hardware divide and remainder
13692 instructions. If @option{-march} is used without @option{-mcpu},
13693 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13694 architectures. Otherwise, the default is taken from the target CPU
13695 (either the default CPU, or the one specified by @option{-mcpu}). For
13696 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13697 @option{-mcpu=5206e}.
13699 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13703 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13704 Additionally, parameters passed on the stack are also aligned to a
13705 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13709 Do not consider type @code{int} to be 16 bits wide. This is the default.
13712 @itemx -mno-bitfield
13713 @opindex mnobitfield
13714 @opindex mno-bitfield
13715 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13716 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13720 Do use the bit-field instructions. The @option{-m68020} option implies
13721 @option{-mbitfield}. This is the default if you use a configuration
13722 designed for a 68020.
13726 Use a different function-calling convention, in which functions
13727 that take a fixed number of arguments return with the @code{rtd}
13728 instruction, which pops their arguments while returning. This
13729 saves one instruction in the caller since there is no need to pop
13730 the arguments there.
13732 This calling convention is incompatible with the one normally
13733 used on Unix, so you cannot use it if you need to call libraries
13734 compiled with the Unix compiler.
13736 Also, you must provide function prototypes for all functions that
13737 take variable numbers of arguments (including @code{printf});
13738 otherwise incorrect code will be generated for calls to those
13741 In addition, seriously incorrect code will result if you call a
13742 function with too many arguments. (Normally, extra arguments are
13743 harmlessly ignored.)
13745 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13746 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13750 Do not use the calling conventions selected by @option{-mrtd}.
13751 This is the default.
13754 @itemx -mno-align-int
13755 @opindex malign-int
13756 @opindex mno-align-int
13757 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13758 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13759 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13760 Aligning variables on 32-bit boundaries produces code that runs somewhat
13761 faster on processors with 32-bit busses at the expense of more memory.
13763 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13764 align structures containing the above types differently than
13765 most published application binary interface specifications for the m68k.
13769 Use the pc-relative addressing mode of the 68000 directly, instead of
13770 using a global offset table. At present, this option implies @option{-fpic},
13771 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13772 not presently supported with @option{-mpcrel}, though this could be supported for
13773 68020 and higher processors.
13775 @item -mno-strict-align
13776 @itemx -mstrict-align
13777 @opindex mno-strict-align
13778 @opindex mstrict-align
13779 Do not (do) assume that unaligned memory references will be handled by
13783 Generate code that allows the data segment to be located in a different
13784 area of memory from the text segment. This allows for execute in place in
13785 an environment without virtual memory management. This option implies
13788 @item -mno-sep-data
13789 Generate code that assumes that the data segment follows the text segment.
13790 This is the default.
13792 @item -mid-shared-library
13793 Generate code that supports shared libraries via the library ID method.
13794 This allows for execute in place and shared libraries in an environment
13795 without virtual memory management. This option implies @option{-fPIC}.
13797 @item -mno-id-shared-library
13798 Generate code that doesn't assume ID based shared libraries are being used.
13799 This is the default.
13801 @item -mshared-library-id=n
13802 Specified the identification number of the ID based shared library being
13803 compiled. Specifying a value of 0 will generate more compact code, specifying
13804 other values will force the allocation of that number to the current
13805 library but is no more space or time efficient than omitting this option.
13811 When generating position-independent code for ColdFire, generate code
13812 that works if the GOT has more than 8192 entries. This code is
13813 larger and slower than code generated without this option. On M680x0
13814 processors, this option is not needed; @option{-fPIC} suffices.
13816 GCC normally uses a single instruction to load values from the GOT@.
13817 While this is relatively efficient, it only works if the GOT
13818 is smaller than about 64k. Anything larger causes the linker
13819 to report an error such as:
13821 @cindex relocation truncated to fit (ColdFire)
13823 relocation truncated to fit: R_68K_GOT16O foobar
13826 If this happens, you should recompile your code with @option{-mxgot}.
13827 It should then work with very large GOTs. However, code generated with
13828 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13829 the value of a global symbol.
13831 Note that some linkers, including newer versions of the GNU linker,
13832 can create multiple GOTs and sort GOT entries. If you have such a linker,
13833 you should only need to use @option{-mxgot} when compiling a single
13834 object file that accesses more than 8192 GOT entries. Very few do.
13836 These options have no effect unless GCC is generating
13837 position-independent code.
13841 @node MCore Options
13842 @subsection MCore Options
13843 @cindex MCore options
13845 These are the @samp{-m} options defined for the Motorola M*Core
13851 @itemx -mno-hardlit
13853 @opindex mno-hardlit
13854 Inline constants into the code stream if it can be done in two
13855 instructions or less.
13861 Use the divide instruction. (Enabled by default).
13863 @item -mrelax-immediate
13864 @itemx -mno-relax-immediate
13865 @opindex mrelax-immediate
13866 @opindex mno-relax-immediate
13867 Allow arbitrary sized immediates in bit operations.
13869 @item -mwide-bitfields
13870 @itemx -mno-wide-bitfields
13871 @opindex mwide-bitfields
13872 @opindex mno-wide-bitfields
13873 Always treat bit-fields as int-sized.
13875 @item -m4byte-functions
13876 @itemx -mno-4byte-functions
13877 @opindex m4byte-functions
13878 @opindex mno-4byte-functions
13879 Force all functions to be aligned to a four byte boundary.
13881 @item -mcallgraph-data
13882 @itemx -mno-callgraph-data
13883 @opindex mcallgraph-data
13884 @opindex mno-callgraph-data
13885 Emit callgraph information.
13888 @itemx -mno-slow-bytes
13889 @opindex mslow-bytes
13890 @opindex mno-slow-bytes
13891 Prefer word access when reading byte quantities.
13893 @item -mlittle-endian
13894 @itemx -mbig-endian
13895 @opindex mlittle-endian
13896 @opindex mbig-endian
13897 Generate code for a little endian target.
13903 Generate code for the 210 processor.
13907 Assume that run-time support has been provided and so omit the
13908 simulator library (@file{libsim.a)} from the linker command line.
13910 @item -mstack-increment=@var{size}
13911 @opindex mstack-increment
13912 Set the maximum amount for a single stack increment operation. Large
13913 values can increase the speed of programs which contain functions
13914 that need a large amount of stack space, but they can also trigger a
13915 segmentation fault if the stack is extended too much. The default
13921 @subsection MeP Options
13922 @cindex MeP options
13928 Enables the @code{abs} instruction, which is the absolute difference
13929 between two registers.
13933 Enables all the optional instructions - average, multiply, divide, bit
13934 operations, leading zero, absolute difference, min/max, clip, and
13940 Enables the @code{ave} instruction, which computes the average of two
13943 @item -mbased=@var{n}
13945 Variables of size @var{n} bytes or smaller will be placed in the
13946 @code{.based} section by default. Based variables use the @code{$tp}
13947 register as a base register, and there is a 128 byte limit to the
13948 @code{.based} section.
13952 Enables the bit operation instructions - bit test (@code{btstm}), set
13953 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
13954 test-and-set (@code{tas}).
13956 @item -mc=@var{name}
13958 Selects which section constant data will be placed in. @var{name} may
13959 be @code{tiny}, @code{near}, or @code{far}.
13963 Enables the @code{clip} instruction. Note that @code{-mclip} is not
13964 useful unless you also provide @code{-mminmax}.
13966 @item -mconfig=@var{name}
13968 Selects one of the build-in core configurations. Each MeP chip has
13969 one or more modules in it; each module has a core CPU and a variety of
13970 coprocessors, optional instructions, and peripherals. The
13971 @code{MeP-Integrator} tool, not part of GCC, provides these
13972 configurations through this option; using this option is the same as
13973 using all the corresponding command line options. The default
13974 configuration is @code{default}.
13978 Enables the coprocessor instructions. By default, this is a 32-bit
13979 coprocessor. Note that the coprocessor is normally enabled via the
13980 @code{-mconfig=} option.
13984 Enables the 32-bit coprocessor's instructions.
13988 Enables the 64-bit coprocessor's instructions.
13992 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
13996 Causes constant variables to be placed in the @code{.near} section.
14000 Enables the @code{div} and @code{divu} instructions.
14004 Generate big-endian code.
14008 Generate little-endian code.
14010 @item -mio-volatile
14011 @opindex mio-volatile
14012 Tells the compiler that any variable marked with the @code{io}
14013 attribute is to be considered volatile.
14017 Causes variables to be assigned to the @code{.far} section by default.
14021 Enables the @code{leadz} (leading zero) instruction.
14025 Causes variables to be assigned to the @code{.near} section by default.
14029 Enables the @code{min} and @code{max} instructions.
14033 Enables the multiplication and multiply-accumulate instructions.
14037 Disables all the optional instructions enabled by @code{-mall-opts}.
14041 Enables the @code{repeat} and @code{erepeat} instructions, used for
14042 low-overhead looping.
14046 Causes all variables to default to the @code{.tiny} section. Note
14047 that there is a 65536 byte limit to this section. Accesses to these
14048 variables use the @code{%gp} base register.
14052 Enables the saturation instructions. Note that the compiler does not
14053 currently generate these itself, but this option is included for
14054 compatibility with other tools, like @code{as}.
14058 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14062 Link the simulator runtime libraries.
14066 Link the simulator runtime libraries, excluding built-in support
14067 for reset and exception vectors and tables.
14071 Causes all functions to default to the @code{.far} section. Without
14072 this option, functions default to the @code{.near} section.
14074 @item -mtiny=@var{n}
14076 Variables that are @var{n} bytes or smaller will be allocated to the
14077 @code{.tiny} section. These variables use the @code{$gp} base
14078 register. The default for this option is 4, but note that there's a
14079 65536 byte limit to the @code{.tiny} section.
14083 @node MicroBlaze Options
14084 @subsection MicroBlaze Options
14085 @cindex MicroBlaze Options
14090 @opindex msoft-float
14091 Use software emulation for floating point (default).
14094 @opindex mhard-float
14095 Use hardware floating point instructions.
14099 Do not optimize block moves, use @code{memcpy}.
14101 @item -mno-clearbss
14102 @opindex mno-clearbss
14103 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14105 @item -mcpu=@var{cpu-type}
14107 Use features of and schedule code for given CPU.
14108 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14109 where @var{X} is a major version, @var{YY} is the minor version, and
14110 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14111 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14113 @item -mxl-soft-mul
14114 @opindex mxl-soft-mul
14115 Use software multiply emulation (default).
14117 @item -mxl-soft-div
14118 @opindex mxl-soft-div
14119 Use software emulation for divides (default).
14121 @item -mxl-barrel-shift
14122 @opindex mxl-barrel-shift
14123 Use the hardware barrel shifter.
14125 @item -mxl-pattern-compare
14126 @opindex mxl-pattern-compare
14127 Use pattern compare instructions.
14129 @item -msmall-divides
14130 @opindex msmall-divides
14131 Use table lookup optimization for small signed integer divisions.
14133 @item -mxl-stack-check
14134 @opindex mxl-stack-check
14135 This option is deprecated. Use -fstack-check instead.
14138 @opindex mxl-gp-opt
14139 Use GP relative sdata/sbss sections.
14141 @item -mxl-multiply-high
14142 @opindex mxl-multiply-high
14143 Use multiply high instructions for high part of 32x32 multiply.
14145 @item -mxl-float-convert
14146 @opindex mxl-float-convert
14147 Use hardware floating point conversion instructions.
14149 @item -mxl-float-sqrt
14150 @opindex mxl-float-sqrt
14151 Use hardware floating point square root instruction.
14153 @item -mxl-mode-@var{app-model}
14154 Select application model @var{app-model}. Valid models are
14157 normal executable (default), uses startup code @file{crt0.o}.
14160 for use with Xilinx Microprocessor Debugger (XMD) based
14161 software intrusive debug agent called xmdstub. This uses startup file
14162 @file{crt1.o} and sets the start address of the program to be 0x800.
14165 for applications that are loaded using a bootloader.
14166 This model uses startup file @file{crt2.o} which does not contain a processor
14167 reset vector handler. This is suitable for transferring control on a
14168 processor reset to the bootloader rather than the application.
14171 for applications that do not require any of the
14172 MicroBlaze vectors. This option may be useful for applications running
14173 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14176 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14177 @option{-mxl-mode-@var{app-model}}.
14182 @subsection MIPS Options
14183 @cindex MIPS options
14189 Generate big-endian code.
14193 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14196 @item -march=@var{arch}
14198 Generate code that will run on @var{arch}, which can be the name of a
14199 generic MIPS ISA, or the name of a particular processor.
14201 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14202 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14203 The processor names are:
14204 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14205 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14206 @samp{5kc}, @samp{5kf},
14208 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14209 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14210 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14211 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14212 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14213 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14217 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14218 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14219 @samp{rm7000}, @samp{rm9000},
14220 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14223 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14224 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14226 The special value @samp{from-abi} selects the
14227 most compatible architecture for the selected ABI (that is,
14228 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14230 Native Linux/GNU toolchains also support the value @samp{native},
14231 which selects the best architecture option for the host processor.
14232 @option{-march=native} has no effect if GCC does not recognize
14235 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14236 (for example, @samp{-march=r2k}). Prefixes are optional, and
14237 @samp{vr} may be written @samp{r}.
14239 Names of the form @samp{@var{n}f2_1} refer to processors with
14240 FPUs clocked at half the rate of the core, names of the form
14241 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14242 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14243 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14244 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14245 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14246 accepted as synonyms for @samp{@var{n}f1_1}.
14248 GCC defines two macros based on the value of this option. The first
14249 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14250 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14251 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14252 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14253 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14255 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14256 above. In other words, it will have the full prefix and will not
14257 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14258 the macro names the resolved architecture (either @samp{"mips1"} or
14259 @samp{"mips3"}). It names the default architecture when no
14260 @option{-march} option is given.
14262 @item -mtune=@var{arch}
14264 Optimize for @var{arch}. Among other things, this option controls
14265 the way instructions are scheduled, and the perceived cost of arithmetic
14266 operations. The list of @var{arch} values is the same as for
14269 When this option is not used, GCC will optimize for the processor
14270 specified by @option{-march}. By using @option{-march} and
14271 @option{-mtune} together, it is possible to generate code that will
14272 run on a family of processors, but optimize the code for one
14273 particular member of that family.
14275 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14276 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14277 @samp{-march} ones described above.
14281 Equivalent to @samp{-march=mips1}.
14285 Equivalent to @samp{-march=mips2}.
14289 Equivalent to @samp{-march=mips3}.
14293 Equivalent to @samp{-march=mips4}.
14297 Equivalent to @samp{-march=mips32}.
14301 Equivalent to @samp{-march=mips32r2}.
14305 Equivalent to @samp{-march=mips64}.
14309 Equivalent to @samp{-march=mips64r2}.
14314 @opindex mno-mips16
14315 Generate (do not generate) MIPS16 code. If GCC is targetting a
14316 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14318 MIPS16 code generation can also be controlled on a per-function basis
14319 by means of @code{mips16} and @code{nomips16} attributes.
14320 @xref{Function Attributes}, for more information.
14322 @item -mflip-mips16
14323 @opindex mflip-mips16
14324 Generate MIPS16 code on alternating functions. This option is provided
14325 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14326 not intended for ordinary use in compiling user code.
14328 @item -minterlink-mips16
14329 @itemx -mno-interlink-mips16
14330 @opindex minterlink-mips16
14331 @opindex mno-interlink-mips16
14332 Require (do not require) that non-MIPS16 code be link-compatible with
14335 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14336 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14337 therefore disables direct jumps unless GCC knows that the target of the
14338 jump is not MIPS16.
14350 Generate code for the given ABI@.
14352 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14353 generates 64-bit code when you select a 64-bit architecture, but you
14354 can use @option{-mgp32} to get 32-bit code instead.
14356 For information about the O64 ABI, see
14357 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14359 GCC supports a variant of the o32 ABI in which floating-point registers
14360 are 64 rather than 32 bits wide. You can select this combination with
14361 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14362 and @samp{mfhc1} instructions and is therefore only supported for
14363 MIPS32R2 processors.
14365 The register assignments for arguments and return values remain the
14366 same, but each scalar value is passed in a single 64-bit register
14367 rather than a pair of 32-bit registers. For example, scalar
14368 floating-point values are returned in @samp{$f0} only, not a
14369 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14370 remains the same, but all 64 bits are saved.
14373 @itemx -mno-abicalls
14375 @opindex mno-abicalls
14376 Generate (do not generate) code that is suitable for SVR4-style
14377 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14382 Generate (do not generate) code that is fully position-independent,
14383 and that can therefore be linked into shared libraries. This option
14384 only affects @option{-mabicalls}.
14386 All @option{-mabicalls} code has traditionally been position-independent,
14387 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14388 as an extension, the GNU toolchain allows executables to use absolute
14389 accesses for locally-binding symbols. It can also use shorter GP
14390 initialization sequences and generate direct calls to locally-defined
14391 functions. This mode is selected by @option{-mno-shared}.
14393 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14394 objects that can only be linked by the GNU linker. However, the option
14395 does not affect the ABI of the final executable; it only affects the ABI
14396 of relocatable objects. Using @option{-mno-shared} will generally make
14397 executables both smaller and quicker.
14399 @option{-mshared} is the default.
14405 Assume (do not assume) that the static and dynamic linkers
14406 support PLTs and copy relocations. This option only affects
14407 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14408 has no effect without @samp{-msym32}.
14410 You can make @option{-mplt} the default by configuring
14411 GCC with @option{--with-mips-plt}. The default is
14412 @option{-mno-plt} otherwise.
14418 Lift (do not lift) the usual restrictions on the size of the global
14421 GCC normally uses a single instruction to load values from the GOT@.
14422 While this is relatively efficient, it will only work if the GOT
14423 is smaller than about 64k. Anything larger will cause the linker
14424 to report an error such as:
14426 @cindex relocation truncated to fit (MIPS)
14428 relocation truncated to fit: R_MIPS_GOT16 foobar
14431 If this happens, you should recompile your code with @option{-mxgot}.
14432 It should then work with very large GOTs, although it will also be
14433 less efficient, since it will take three instructions to fetch the
14434 value of a global symbol.
14436 Note that some linkers can create multiple GOTs. If you have such a
14437 linker, you should only need to use @option{-mxgot} when a single object
14438 file accesses more than 64k's worth of GOT entries. Very few do.
14440 These options have no effect unless GCC is generating position
14445 Assume that general-purpose registers are 32 bits wide.
14449 Assume that general-purpose registers are 64 bits wide.
14453 Assume that floating-point registers are 32 bits wide.
14457 Assume that floating-point registers are 64 bits wide.
14460 @opindex mhard-float
14461 Use floating-point coprocessor instructions.
14464 @opindex msoft-float
14465 Do not use floating-point coprocessor instructions. Implement
14466 floating-point calculations using library calls instead.
14468 @item -msingle-float
14469 @opindex msingle-float
14470 Assume that the floating-point coprocessor only supports single-precision
14473 @item -mdouble-float
14474 @opindex mdouble-float
14475 Assume that the floating-point coprocessor supports double-precision
14476 operations. This is the default.
14482 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14483 implement atomic memory built-in functions. When neither option is
14484 specified, GCC will use the instructions if the target architecture
14487 @option{-mllsc} is useful if the runtime environment can emulate the
14488 instructions and @option{-mno-llsc} can be useful when compiling for
14489 nonstandard ISAs. You can make either option the default by
14490 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14491 respectively. @option{--with-llsc} is the default for some
14492 configurations; see the installation documentation for details.
14498 Use (do not use) revision 1 of the MIPS DSP ASE@.
14499 @xref{MIPS DSP Built-in Functions}. This option defines the
14500 preprocessor macro @samp{__mips_dsp}. It also defines
14501 @samp{__mips_dsp_rev} to 1.
14507 Use (do not use) revision 2 of the MIPS DSP ASE@.
14508 @xref{MIPS DSP Built-in Functions}. This option defines the
14509 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14510 It also defines @samp{__mips_dsp_rev} to 2.
14513 @itemx -mno-smartmips
14514 @opindex msmartmips
14515 @opindex mno-smartmips
14516 Use (do not use) the MIPS SmartMIPS ASE.
14518 @item -mpaired-single
14519 @itemx -mno-paired-single
14520 @opindex mpaired-single
14521 @opindex mno-paired-single
14522 Use (do not use) paired-single floating-point instructions.
14523 @xref{MIPS Paired-Single Support}. This option requires
14524 hardware floating-point support to be enabled.
14530 Use (do not use) MIPS Digital Media Extension instructions.
14531 This option can only be used when generating 64-bit code and requires
14532 hardware floating-point support to be enabled.
14537 @opindex mno-mips3d
14538 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14539 The option @option{-mips3d} implies @option{-mpaired-single}.
14545 Use (do not use) MT Multithreading instructions.
14549 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14550 an explanation of the default and the way that the pointer size is
14555 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14557 The default size of @code{int}s, @code{long}s and pointers depends on
14558 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14559 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14560 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14561 or the same size as integer registers, whichever is smaller.
14567 Assume (do not assume) that all symbols have 32-bit values, regardless
14568 of the selected ABI@. This option is useful in combination with
14569 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14570 to generate shorter and faster references to symbolic addresses.
14574 Put definitions of externally-visible data in a small data section
14575 if that data is no bigger than @var{num} bytes. GCC can then access
14576 the data more efficiently; see @option{-mgpopt} for details.
14578 The default @option{-G} option depends on the configuration.
14580 @item -mlocal-sdata
14581 @itemx -mno-local-sdata
14582 @opindex mlocal-sdata
14583 @opindex mno-local-sdata
14584 Extend (do not extend) the @option{-G} behavior to local data too,
14585 such as to static variables in C@. @option{-mlocal-sdata} is the
14586 default for all configurations.
14588 If the linker complains that an application is using too much small data,
14589 you might want to try rebuilding the less performance-critical parts with
14590 @option{-mno-local-sdata}. You might also want to build large
14591 libraries with @option{-mno-local-sdata}, so that the libraries leave
14592 more room for the main program.
14594 @item -mextern-sdata
14595 @itemx -mno-extern-sdata
14596 @opindex mextern-sdata
14597 @opindex mno-extern-sdata
14598 Assume (do not assume) that externally-defined data will be in
14599 a small data section if that data is within the @option{-G} limit.
14600 @option{-mextern-sdata} is the default for all configurations.
14602 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14603 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14604 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14605 is placed in a small data section. If @var{Var} is defined by another
14606 module, you must either compile that module with a high-enough
14607 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14608 definition. If @var{Var} is common, you must link the application
14609 with a high-enough @option{-G} setting.
14611 The easiest way of satisfying these restrictions is to compile
14612 and link every module with the same @option{-G} option. However,
14613 you may wish to build a library that supports several different
14614 small data limits. You can do this by compiling the library with
14615 the highest supported @option{-G} setting and additionally using
14616 @option{-mno-extern-sdata} to stop the library from making assumptions
14617 about externally-defined data.
14623 Use (do not use) GP-relative accesses for symbols that are known to be
14624 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14625 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14628 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14629 might not hold the value of @code{_gp}. For example, if the code is
14630 part of a library that might be used in a boot monitor, programs that
14631 call boot monitor routines will pass an unknown value in @code{$gp}.
14632 (In such situations, the boot monitor itself would usually be compiled
14633 with @option{-G0}.)
14635 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14636 @option{-mno-extern-sdata}.
14638 @item -membedded-data
14639 @itemx -mno-embedded-data
14640 @opindex membedded-data
14641 @opindex mno-embedded-data
14642 Allocate variables to the read-only data section first if possible, then
14643 next in the small data section if possible, otherwise in data. This gives
14644 slightly slower code than the default, but reduces the amount of RAM required
14645 when executing, and thus may be preferred for some embedded systems.
14647 @item -muninit-const-in-rodata
14648 @itemx -mno-uninit-const-in-rodata
14649 @opindex muninit-const-in-rodata
14650 @opindex mno-uninit-const-in-rodata
14651 Put uninitialized @code{const} variables in the read-only data section.
14652 This option is only meaningful in conjunction with @option{-membedded-data}.
14654 @item -mcode-readable=@var{setting}
14655 @opindex mcode-readable
14656 Specify whether GCC may generate code that reads from executable sections.
14657 There are three possible settings:
14660 @item -mcode-readable=yes
14661 Instructions may freely access executable sections. This is the
14664 @item -mcode-readable=pcrel
14665 MIPS16 PC-relative load instructions can access executable sections,
14666 but other instructions must not do so. This option is useful on 4KSc
14667 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14668 It is also useful on processors that can be configured to have a dual
14669 instruction/data SRAM interface and that, like the M4K, automatically
14670 redirect PC-relative loads to the instruction RAM.
14672 @item -mcode-readable=no
14673 Instructions must not access executable sections. This option can be
14674 useful on targets that are configured to have a dual instruction/data
14675 SRAM interface but that (unlike the M4K) do not automatically redirect
14676 PC-relative loads to the instruction RAM.
14679 @item -msplit-addresses
14680 @itemx -mno-split-addresses
14681 @opindex msplit-addresses
14682 @opindex mno-split-addresses
14683 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14684 relocation operators. This option has been superseded by
14685 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14687 @item -mexplicit-relocs
14688 @itemx -mno-explicit-relocs
14689 @opindex mexplicit-relocs
14690 @opindex mno-explicit-relocs
14691 Use (do not use) assembler relocation operators when dealing with symbolic
14692 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14693 is to use assembler macros instead.
14695 @option{-mexplicit-relocs} is the default if GCC was configured
14696 to use an assembler that supports relocation operators.
14698 @item -mcheck-zero-division
14699 @itemx -mno-check-zero-division
14700 @opindex mcheck-zero-division
14701 @opindex mno-check-zero-division
14702 Trap (do not trap) on integer division by zero.
14704 The default is @option{-mcheck-zero-division}.
14706 @item -mdivide-traps
14707 @itemx -mdivide-breaks
14708 @opindex mdivide-traps
14709 @opindex mdivide-breaks
14710 MIPS systems check for division by zero by generating either a
14711 conditional trap or a break instruction. Using traps results in
14712 smaller code, but is only supported on MIPS II and later. Also, some
14713 versions of the Linux kernel have a bug that prevents trap from
14714 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14715 allow conditional traps on architectures that support them and
14716 @option{-mdivide-breaks} to force the use of breaks.
14718 The default is usually @option{-mdivide-traps}, but this can be
14719 overridden at configure time using @option{--with-divide=breaks}.
14720 Divide-by-zero checks can be completely disabled using
14721 @option{-mno-check-zero-division}.
14726 @opindex mno-memcpy
14727 Force (do not force) the use of @code{memcpy()} for non-trivial block
14728 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14729 most constant-sized copies.
14732 @itemx -mno-long-calls
14733 @opindex mlong-calls
14734 @opindex mno-long-calls
14735 Disable (do not disable) use of the @code{jal} instruction. Calling
14736 functions using @code{jal} is more efficient but requires the caller
14737 and callee to be in the same 256 megabyte segment.
14739 This option has no effect on abicalls code. The default is
14740 @option{-mno-long-calls}.
14746 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14747 instructions, as provided by the R4650 ISA@.
14750 @itemx -mno-fused-madd
14751 @opindex mfused-madd
14752 @opindex mno-fused-madd
14753 Enable (disable) use of the floating point multiply-accumulate
14754 instructions, when they are available. The default is
14755 @option{-mfused-madd}.
14757 When multiply-accumulate instructions are used, the intermediate
14758 product is calculated to infinite precision and is not subject to
14759 the FCSR Flush to Zero bit. This may be undesirable in some
14764 Tell the MIPS assembler to not run its preprocessor over user
14765 assembler files (with a @samp{.s} suffix) when assembling them.
14770 @opindex mno-fix-24k
14771 Work around the 24K E48 (lost data on stores during refill) errata.
14772 The workarounds are implemented by the assembler rather than by GCC.
14775 @itemx -mno-fix-r4000
14776 @opindex mfix-r4000
14777 @opindex mno-fix-r4000
14778 Work around certain R4000 CPU errata:
14781 A double-word or a variable shift may give an incorrect result if executed
14782 immediately after starting an integer division.
14784 A double-word or a variable shift may give an incorrect result if executed
14785 while an integer multiplication is in progress.
14787 An integer division may give an incorrect result if started in a delay slot
14788 of a taken branch or a jump.
14792 @itemx -mno-fix-r4400
14793 @opindex mfix-r4400
14794 @opindex mno-fix-r4400
14795 Work around certain R4400 CPU errata:
14798 A double-word or a variable shift may give an incorrect result if executed
14799 immediately after starting an integer division.
14803 @itemx -mno-fix-r10000
14804 @opindex mfix-r10000
14805 @opindex mno-fix-r10000
14806 Work around certain R10000 errata:
14809 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14810 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14813 This option can only be used if the target architecture supports
14814 branch-likely instructions. @option{-mfix-r10000} is the default when
14815 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14819 @itemx -mno-fix-vr4120
14820 @opindex mfix-vr4120
14821 Work around certain VR4120 errata:
14824 @code{dmultu} does not always produce the correct result.
14826 @code{div} and @code{ddiv} do not always produce the correct result if one
14827 of the operands is negative.
14829 The workarounds for the division errata rely on special functions in
14830 @file{libgcc.a}. At present, these functions are only provided by
14831 the @code{mips64vr*-elf} configurations.
14833 Other VR4120 errata require a nop to be inserted between certain pairs of
14834 instructions. These errata are handled by the assembler, not by GCC itself.
14837 @opindex mfix-vr4130
14838 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14839 workarounds are implemented by the assembler rather than by GCC,
14840 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14841 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14842 instructions are available instead.
14845 @itemx -mno-fix-sb1
14847 Work around certain SB-1 CPU core errata.
14848 (This flag currently works around the SB-1 revision 2
14849 ``F1'' and ``F2'' floating point errata.)
14851 @item -mr10k-cache-barrier=@var{setting}
14852 @opindex mr10k-cache-barrier
14853 Specify whether GCC should insert cache barriers to avoid the
14854 side-effects of speculation on R10K processors.
14856 In common with many processors, the R10K tries to predict the outcome
14857 of a conditional branch and speculatively executes instructions from
14858 the ``taken'' branch. It later aborts these instructions if the
14859 predicted outcome was wrong. However, on the R10K, even aborted
14860 instructions can have side effects.
14862 This problem only affects kernel stores and, depending on the system,
14863 kernel loads. As an example, a speculatively-executed store may load
14864 the target memory into cache and mark the cache line as dirty, even if
14865 the store itself is later aborted. If a DMA operation writes to the
14866 same area of memory before the ``dirty'' line is flushed, the cached
14867 data will overwrite the DMA-ed data. See the R10K processor manual
14868 for a full description, including other potential problems.
14870 One workaround is to insert cache barrier instructions before every memory
14871 access that might be speculatively executed and that might have side
14872 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14873 controls GCC's implementation of this workaround. It assumes that
14874 aborted accesses to any byte in the following regions will not have
14879 the memory occupied by the current function's stack frame;
14882 the memory occupied by an incoming stack argument;
14885 the memory occupied by an object with a link-time-constant address.
14888 It is the kernel's responsibility to ensure that speculative
14889 accesses to these regions are indeed safe.
14891 If the input program contains a function declaration such as:
14897 then the implementation of @code{foo} must allow @code{j foo} and
14898 @code{jal foo} to be executed speculatively. GCC honors this
14899 restriction for functions it compiles itself. It expects non-GCC
14900 functions (such as hand-written assembly code) to do the same.
14902 The option has three forms:
14905 @item -mr10k-cache-barrier=load-store
14906 Insert a cache barrier before a load or store that might be
14907 speculatively executed and that might have side effects even
14910 @item -mr10k-cache-barrier=store
14911 Insert a cache barrier before a store that might be speculatively
14912 executed and that might have side effects even if aborted.
14914 @item -mr10k-cache-barrier=none
14915 Disable the insertion of cache barriers. This is the default setting.
14918 @item -mflush-func=@var{func}
14919 @itemx -mno-flush-func
14920 @opindex mflush-func
14921 Specifies the function to call to flush the I and D caches, or to not
14922 call any such function. If called, the function must take the same
14923 arguments as the common @code{_flush_func()}, that is, the address of the
14924 memory range for which the cache is being flushed, the size of the
14925 memory range, and the number 3 (to flush both caches). The default
14926 depends on the target GCC was configured for, but commonly is either
14927 @samp{_flush_func} or @samp{__cpu_flush}.
14929 @item mbranch-cost=@var{num}
14930 @opindex mbranch-cost
14931 Set the cost of branches to roughly @var{num} ``simple'' instructions.
14932 This cost is only a heuristic and is not guaranteed to produce
14933 consistent results across releases. A zero cost redundantly selects
14934 the default, which is based on the @option{-mtune} setting.
14936 @item -mbranch-likely
14937 @itemx -mno-branch-likely
14938 @opindex mbranch-likely
14939 @opindex mno-branch-likely
14940 Enable or disable use of Branch Likely instructions, regardless of the
14941 default for the selected architecture. By default, Branch Likely
14942 instructions may be generated if they are supported by the selected
14943 architecture. An exception is for the MIPS32 and MIPS64 architectures
14944 and processors which implement those architectures; for those, Branch
14945 Likely instructions will not be generated by default because the MIPS32
14946 and MIPS64 architectures specifically deprecate their use.
14948 @item -mfp-exceptions
14949 @itemx -mno-fp-exceptions
14950 @opindex mfp-exceptions
14951 Specifies whether FP exceptions are enabled. This affects how we schedule
14952 FP instructions for some processors. The default is that FP exceptions are
14955 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
14956 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
14959 @item -mvr4130-align
14960 @itemx -mno-vr4130-align
14961 @opindex mvr4130-align
14962 The VR4130 pipeline is two-way superscalar, but can only issue two
14963 instructions together if the first one is 8-byte aligned. When this
14964 option is enabled, GCC will align pairs of instructions that it
14965 thinks should execute in parallel.
14967 This option only has an effect when optimizing for the VR4130.
14968 It normally makes code faster, but at the expense of making it bigger.
14969 It is enabled by default at optimization level @option{-O3}.
14974 Enable (disable) generation of @code{synci} instructions on
14975 architectures that support it. The @code{synci} instructions (if
14976 enabled) will be generated when @code{__builtin___clear_cache()} is
14979 This option defaults to @code{-mno-synci}, but the default can be
14980 overridden by configuring with @code{--with-synci}.
14982 When compiling code for single processor systems, it is generally safe
14983 to use @code{synci}. However, on many multi-core (SMP) systems, it
14984 will not invalidate the instruction caches on all cores and may lead
14985 to undefined behavior.
14987 @item -mrelax-pic-calls
14988 @itemx -mno-relax-pic-calls
14989 @opindex mrelax-pic-calls
14990 Try to turn PIC calls that are normally dispatched via register
14991 @code{$25} into direct calls. This is only possible if the linker can
14992 resolve the destination at link-time and if the destination is within
14993 range for a direct call.
14995 @option{-mrelax-pic-calls} is the default if GCC was configured to use
14996 an assembler and a linker that supports the @code{.reloc} assembly
14997 directive and @code{-mexplicit-relocs} is in effect. With
14998 @code{-mno-explicit-relocs}, this optimization can be performed by the
14999 assembler and the linker alone without help from the compiler.
15001 @item -mmcount-ra-address
15002 @itemx -mno-mcount-ra-address
15003 @opindex mmcount-ra-address
15004 @opindex mno-mcount-ra-address
15005 Emit (do not emit) code that allows @code{_mcount} to modify the
15006 calling function's return address. When enabled, this option extends
15007 the usual @code{_mcount} interface with a new @var{ra-address}
15008 parameter, which has type @code{intptr_t *} and is passed in register
15009 @code{$12}. @code{_mcount} can then modify the return address by
15010 doing both of the following:
15013 Returning the new address in register @code{$31}.
15015 Storing the new address in @code{*@var{ra-address}},
15016 if @var{ra-address} is nonnull.
15019 The default is @option{-mno-mcount-ra-address}.
15024 @subsection MMIX Options
15025 @cindex MMIX Options
15027 These options are defined for the MMIX:
15031 @itemx -mno-libfuncs
15033 @opindex mno-libfuncs
15034 Specify that intrinsic library functions are being compiled, passing all
15035 values in registers, no matter the size.
15038 @itemx -mno-epsilon
15040 @opindex mno-epsilon
15041 Generate floating-point comparison instructions that compare with respect
15042 to the @code{rE} epsilon register.
15044 @item -mabi=mmixware
15046 @opindex mabi=mmixware
15048 Generate code that passes function parameters and return values that (in
15049 the called function) are seen as registers @code{$0} and up, as opposed to
15050 the GNU ABI which uses global registers @code{$231} and up.
15052 @item -mzero-extend
15053 @itemx -mno-zero-extend
15054 @opindex mzero-extend
15055 @opindex mno-zero-extend
15056 When reading data from memory in sizes shorter than 64 bits, use (do not
15057 use) zero-extending load instructions by default, rather than
15058 sign-extending ones.
15061 @itemx -mno-knuthdiv
15063 @opindex mno-knuthdiv
15064 Make the result of a division yielding a remainder have the same sign as
15065 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15066 remainder follows the sign of the dividend. Both methods are
15067 arithmetically valid, the latter being almost exclusively used.
15069 @item -mtoplevel-symbols
15070 @itemx -mno-toplevel-symbols
15071 @opindex mtoplevel-symbols
15072 @opindex mno-toplevel-symbols
15073 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15074 code can be used with the @code{PREFIX} assembly directive.
15078 Generate an executable in the ELF format, rather than the default
15079 @samp{mmo} format used by the @command{mmix} simulator.
15081 @item -mbranch-predict
15082 @itemx -mno-branch-predict
15083 @opindex mbranch-predict
15084 @opindex mno-branch-predict
15085 Use (do not use) the probable-branch instructions, when static branch
15086 prediction indicates a probable branch.
15088 @item -mbase-addresses
15089 @itemx -mno-base-addresses
15090 @opindex mbase-addresses
15091 @opindex mno-base-addresses
15092 Generate (do not generate) code that uses @emph{base addresses}. Using a
15093 base address automatically generates a request (handled by the assembler
15094 and the linker) for a constant to be set up in a global register. The
15095 register is used for one or more base address requests within the range 0
15096 to 255 from the value held in the register. The generally leads to short
15097 and fast code, but the number of different data items that can be
15098 addressed is limited. This means that a program that uses lots of static
15099 data may require @option{-mno-base-addresses}.
15101 @item -msingle-exit
15102 @itemx -mno-single-exit
15103 @opindex msingle-exit
15104 @opindex mno-single-exit
15105 Force (do not force) generated code to have a single exit point in each
15109 @node MN10300 Options
15110 @subsection MN10300 Options
15111 @cindex MN10300 options
15113 These @option{-m} options are defined for Matsushita MN10300 architectures:
15118 Generate code to avoid bugs in the multiply instructions for the MN10300
15119 processors. This is the default.
15121 @item -mno-mult-bug
15122 @opindex mno-mult-bug
15123 Do not generate code to avoid bugs in the multiply instructions for the
15124 MN10300 processors.
15128 Generate code which uses features specific to the AM33 processor.
15132 Do not generate code which uses features specific to the AM33 processor. This
15137 Generate code which uses features specific to the AM33/2.0 processor.
15141 Generate code which uses features specific to the AM34 processor.
15143 @item -mtune=@var{cpu-type}
15145 Use the timing characteristics of the indicated CPU type when
15146 scheduling instructions. This does not change the targeted processor
15147 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15148 @samp{am33-2} or @samp{am34}.
15150 @item -mreturn-pointer-on-d0
15151 @opindex mreturn-pointer-on-d0
15152 When generating a function which returns a pointer, return the pointer
15153 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15154 only in a0, and attempts to call such functions without a prototype
15155 would result in errors. Note that this option is on by default; use
15156 @option{-mno-return-pointer-on-d0} to disable it.
15160 Do not link in the C run-time initialization object file.
15164 Indicate to the linker that it should perform a relaxation optimization pass
15165 to shorten branches, calls and absolute memory addresses. This option only
15166 has an effect when used on the command line for the final link step.
15168 This option makes symbolic debugging impossible.
15172 Allow the compiler to generate @emph{Long Instruction Word}
15173 instructions if the target is the @samp{AM33} or later. This is the
15174 default. This option defines the preprocessor macro @samp{__LIW__}.
15178 Do not allow the compiler to generate @emph{Long Instruction Word}
15179 instructions. This option defines the preprocessor macro
15184 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15185 instructions if the target is the @samp{AM33} or later. This is the
15186 default. This option defines the preprocessor macro @samp{__SETLB__}.
15190 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15191 instructions. This option defines the preprocessor macro
15192 @samp{__NO_SETLB__}.
15196 @node PDP-11 Options
15197 @subsection PDP-11 Options
15198 @cindex PDP-11 Options
15200 These options are defined for the PDP-11:
15205 Use hardware FPP floating point. This is the default. (FIS floating
15206 point on the PDP-11/40 is not supported.)
15209 @opindex msoft-float
15210 Do not use hardware floating point.
15214 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15218 Return floating-point results in memory. This is the default.
15222 Generate code for a PDP-11/40.
15226 Generate code for a PDP-11/45. This is the default.
15230 Generate code for a PDP-11/10.
15232 @item -mbcopy-builtin
15233 @opindex mbcopy-builtin
15234 Use inline @code{movmemhi} patterns for copying memory. This is the
15239 Do not use inline @code{movmemhi} patterns for copying memory.
15245 Use 16-bit @code{int}. This is the default.
15251 Use 32-bit @code{int}.
15254 @itemx -mno-float32
15256 @opindex mno-float32
15257 Use 64-bit @code{float}. This is the default.
15260 @itemx -mno-float64
15262 @opindex mno-float64
15263 Use 32-bit @code{float}.
15267 Use @code{abshi2} pattern. This is the default.
15271 Do not use @code{abshi2} pattern.
15273 @item -mbranch-expensive
15274 @opindex mbranch-expensive
15275 Pretend that branches are expensive. This is for experimenting with
15276 code generation only.
15278 @item -mbranch-cheap
15279 @opindex mbranch-cheap
15280 Do not pretend that branches are expensive. This is the default.
15284 Use Unix assembler syntax. This is the default when configured for
15285 @samp{pdp11-*-bsd}.
15289 Use DEC assembler syntax. This is the default when configured for any
15290 PDP-11 target other than @samp{pdp11-*-bsd}.
15293 @node picoChip Options
15294 @subsection picoChip Options
15295 @cindex picoChip options
15297 These @samp{-m} options are defined for picoChip implementations:
15301 @item -mae=@var{ae_type}
15303 Set the instruction set, register set, and instruction scheduling
15304 parameters for array element type @var{ae_type}. Supported values
15305 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15307 @option{-mae=ANY} selects a completely generic AE type. Code
15308 generated with this option will run on any of the other AE types. The
15309 code will not be as efficient as it would be if compiled for a specific
15310 AE type, and some types of operation (e.g., multiplication) will not
15311 work properly on all types of AE.
15313 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15314 for compiled code, and is the default.
15316 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15317 option may suffer from poor performance of byte (char) manipulation,
15318 since the DSP AE does not provide hardware support for byte load/stores.
15320 @item -msymbol-as-address
15321 Enable the compiler to directly use a symbol name as an address in a
15322 load/store instruction, without first loading it into a
15323 register. Typically, the use of this option will generate larger
15324 programs, which run faster than when the option isn't used. However, the
15325 results vary from program to program, so it is left as a user option,
15326 rather than being permanently enabled.
15328 @item -mno-inefficient-warnings
15329 Disables warnings about the generation of inefficient code. These
15330 warnings can be generated, for example, when compiling code which
15331 performs byte-level memory operations on the MAC AE type. The MAC AE has
15332 no hardware support for byte-level memory operations, so all byte
15333 load/stores must be synthesized from word load/store operations. This is
15334 inefficient and a warning will be generated indicating to the programmer
15335 that they should rewrite the code to avoid byte operations, or to target
15336 an AE type which has the necessary hardware support. This option enables
15337 the warning to be turned off.
15341 @node PowerPC Options
15342 @subsection PowerPC Options
15343 @cindex PowerPC options
15345 These are listed under @xref{RS/6000 and PowerPC Options}.
15347 @node RS/6000 and PowerPC Options
15348 @subsection IBM RS/6000 and PowerPC Options
15349 @cindex RS/6000 and PowerPC Options
15350 @cindex IBM RS/6000 and PowerPC Options
15352 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15359 @itemx -mno-powerpc
15360 @itemx -mpowerpc-gpopt
15361 @itemx -mno-powerpc-gpopt
15362 @itemx -mpowerpc-gfxopt
15363 @itemx -mno-powerpc-gfxopt
15366 @itemx -mno-powerpc64
15370 @itemx -mno-popcntb
15372 @itemx -mno-popcntd
15381 @itemx -mno-hard-dfp
15385 @opindex mno-power2
15387 @opindex mno-powerpc
15388 @opindex mpowerpc-gpopt
15389 @opindex mno-powerpc-gpopt
15390 @opindex mpowerpc-gfxopt
15391 @opindex mno-powerpc-gfxopt
15392 @opindex mpowerpc64
15393 @opindex mno-powerpc64
15397 @opindex mno-popcntb
15399 @opindex mno-popcntd
15405 @opindex mno-mfpgpr
15407 @opindex mno-hard-dfp
15408 GCC supports two related instruction set architectures for the
15409 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15410 instructions supported by the @samp{rios} chip set used in the original
15411 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15412 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15413 the IBM 4xx, 6xx, and follow-on microprocessors.
15415 Neither architecture is a subset of the other. However there is a
15416 large common subset of instructions supported by both. An MQ
15417 register is included in processors supporting the POWER architecture.
15419 You use these options to specify which instructions are available on the
15420 processor you are using. The default value of these options is
15421 determined when configuring GCC@. Specifying the
15422 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15423 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15424 rather than the options listed above.
15426 The @option{-mpower} option allows GCC to generate instructions that
15427 are found only in the POWER architecture and to use the MQ register.
15428 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15429 to generate instructions that are present in the POWER2 architecture but
15430 not the original POWER architecture.
15432 The @option{-mpowerpc} option allows GCC to generate instructions that
15433 are found only in the 32-bit subset of the PowerPC architecture.
15434 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15435 GCC to use the optional PowerPC architecture instructions in the
15436 General Purpose group, including floating-point square root. Specifying
15437 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15438 use the optional PowerPC architecture instructions in the Graphics
15439 group, including floating-point select.
15441 The @option{-mmfcrf} option allows GCC to generate the move from
15442 condition register field instruction implemented on the POWER4
15443 processor and other processors that support the PowerPC V2.01
15445 The @option{-mpopcntb} option allows GCC to generate the popcount and
15446 double precision FP reciprocal estimate instruction implemented on the
15447 POWER5 processor and other processors that support the PowerPC V2.02
15449 The @option{-mpopcntd} option allows GCC to generate the popcount
15450 instruction implemented on the POWER7 processor and other processors
15451 that support the PowerPC V2.06 architecture.
15452 The @option{-mfprnd} option allows GCC to generate the FP round to
15453 integer instructions implemented on the POWER5+ processor and other
15454 processors that support the PowerPC V2.03 architecture.
15455 The @option{-mcmpb} option allows GCC to generate the compare bytes
15456 instruction implemented on the POWER6 processor and other processors
15457 that support the PowerPC V2.05 architecture.
15458 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15459 general purpose register instructions implemented on the POWER6X
15460 processor and other processors that support the extended PowerPC V2.05
15462 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15463 point instructions implemented on some POWER processors.
15465 The @option{-mpowerpc64} option allows GCC to generate the additional
15466 64-bit instructions that are found in the full PowerPC64 architecture
15467 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15468 @option{-mno-powerpc64}.
15470 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15471 will use only the instructions in the common subset of both
15472 architectures plus some special AIX common-mode calls, and will not use
15473 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15474 permits GCC to use any instruction from either architecture and to
15475 allow use of the MQ register; specify this for the Motorola MPC601.
15477 @item -mnew-mnemonics
15478 @itemx -mold-mnemonics
15479 @opindex mnew-mnemonics
15480 @opindex mold-mnemonics
15481 Select which mnemonics to use in the generated assembler code. With
15482 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15483 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15484 assembler mnemonics defined for the POWER architecture. Instructions
15485 defined in only one architecture have only one mnemonic; GCC uses that
15486 mnemonic irrespective of which of these options is specified.
15488 GCC defaults to the mnemonics appropriate for the architecture in
15489 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15490 value of these option. Unless you are building a cross-compiler, you
15491 should normally not specify either @option{-mnew-mnemonics} or
15492 @option{-mold-mnemonics}, but should instead accept the default.
15494 @item -mcpu=@var{cpu_type}
15496 Set architecture type, register usage, choice of mnemonics, and
15497 instruction scheduling parameters for machine type @var{cpu_type}.
15498 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15499 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15500 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15501 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15502 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15503 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15504 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15505 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15506 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15507 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15508 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15510 @option{-mcpu=common} selects a completely generic processor. Code
15511 generated under this option will run on any POWER or PowerPC processor.
15512 GCC will use only the instructions in the common subset of both
15513 architectures, and will not use the MQ register. GCC assumes a generic
15514 processor model for scheduling purposes.
15516 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15517 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15518 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15519 types, with an appropriate, generic processor model assumed for
15520 scheduling purposes.
15522 The other options specify a specific processor. Code generated under
15523 those options will run best on that processor, and may not run at all on
15526 The @option{-mcpu} options automatically enable or disable the
15529 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15530 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15531 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15532 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15534 The particular options set for any particular CPU will vary between
15535 compiler versions, depending on what setting seems to produce optimal
15536 code for that CPU; it doesn't necessarily reflect the actual hardware's
15537 capabilities. If you wish to set an individual option to a particular
15538 value, you may specify it after the @option{-mcpu} option, like
15539 @samp{-mcpu=970 -mno-altivec}.
15541 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15542 not enabled or disabled by the @option{-mcpu} option at present because
15543 AIX does not have full support for these options. You may still
15544 enable or disable them individually if you're sure it'll work in your
15547 @item -mtune=@var{cpu_type}
15549 Set the instruction scheduling parameters for machine type
15550 @var{cpu_type}, but do not set the architecture type, register usage, or
15551 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15552 values for @var{cpu_type} are used for @option{-mtune} as for
15553 @option{-mcpu}. If both are specified, the code generated will use the
15554 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15555 scheduling parameters set by @option{-mtune}.
15557 @item -mcmodel=small
15558 @opindex mcmodel=small
15559 Generate PowerPC64 code for the small model: The TOC is limited to
15562 @item -mcmodel=medium
15563 @opindex mcmodel=medium
15564 Generate PowerPC64 code for the medium model: The TOC and other static
15565 data may be up to a total of 4G in size.
15567 @item -mcmodel=large
15568 @opindex mcmodel=large
15569 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15570 in size. Other data and code is only limited by the 64-bit address
15574 @itemx -mno-altivec
15576 @opindex mno-altivec
15577 Generate code that uses (does not use) AltiVec instructions, and also
15578 enable the use of built-in functions that allow more direct access to
15579 the AltiVec instruction set. You may also need to set
15580 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15586 @opindex mno-vrsave
15587 Generate VRSAVE instructions when generating AltiVec code.
15589 @item -mgen-cell-microcode
15590 @opindex mgen-cell-microcode
15591 Generate Cell microcode instructions
15593 @item -mwarn-cell-microcode
15594 @opindex mwarn-cell-microcode
15595 Warning when a Cell microcode instruction is going to emitted. An example
15596 of a Cell microcode instruction is a variable shift.
15599 @opindex msecure-plt
15600 Generate code that allows ld and ld.so to build executables and shared
15601 libraries with non-exec .plt and .got sections. This is a PowerPC
15602 32-bit SYSV ABI option.
15606 Generate code that uses a BSS .plt section that ld.so fills in, and
15607 requires .plt and .got sections that are both writable and executable.
15608 This is a PowerPC 32-bit SYSV ABI option.
15614 This switch enables or disables the generation of ISEL instructions.
15616 @item -misel=@var{yes/no}
15617 This switch has been deprecated. Use @option{-misel} and
15618 @option{-mno-isel} instead.
15624 This switch enables or disables the generation of SPE simd
15630 @opindex mno-paired
15631 This switch enables or disables the generation of PAIRED simd
15634 @item -mspe=@var{yes/no}
15635 This option has been deprecated. Use @option{-mspe} and
15636 @option{-mno-spe} instead.
15642 Generate code that uses (does not use) vector/scalar (VSX)
15643 instructions, and also enable the use of built-in functions that allow
15644 more direct access to the VSX instruction set.
15646 @item -mfloat-gprs=@var{yes/single/double/no}
15647 @itemx -mfloat-gprs
15648 @opindex mfloat-gprs
15649 This switch enables or disables the generation of floating point
15650 operations on the general purpose registers for architectures that
15653 The argument @var{yes} or @var{single} enables the use of
15654 single-precision floating point operations.
15656 The argument @var{double} enables the use of single and
15657 double-precision floating point operations.
15659 The argument @var{no} disables floating point operations on the
15660 general purpose registers.
15662 This option is currently only available on the MPC854x.
15668 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15669 targets (including GNU/Linux). The 32-bit environment sets int, long
15670 and pointer to 32 bits and generates code that runs on any PowerPC
15671 variant. The 64-bit environment sets int to 32 bits and long and
15672 pointer to 64 bits, and generates code for PowerPC64, as for
15673 @option{-mpowerpc64}.
15676 @itemx -mno-fp-in-toc
15677 @itemx -mno-sum-in-toc
15678 @itemx -mminimal-toc
15680 @opindex mno-fp-in-toc
15681 @opindex mno-sum-in-toc
15682 @opindex mminimal-toc
15683 Modify generation of the TOC (Table Of Contents), which is created for
15684 every executable file. The @option{-mfull-toc} option is selected by
15685 default. In that case, GCC will allocate at least one TOC entry for
15686 each unique non-automatic variable reference in your program. GCC
15687 will also place floating-point constants in the TOC@. However, only
15688 16,384 entries are available in the TOC@.
15690 If you receive a linker error message that saying you have overflowed
15691 the available TOC space, you can reduce the amount of TOC space used
15692 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15693 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15694 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15695 generate code to calculate the sum of an address and a constant at
15696 run-time instead of putting that sum into the TOC@. You may specify one
15697 or both of these options. Each causes GCC to produce very slightly
15698 slower and larger code at the expense of conserving TOC space.
15700 If you still run out of space in the TOC even when you specify both of
15701 these options, specify @option{-mminimal-toc} instead. This option causes
15702 GCC to make only one TOC entry for every file. When you specify this
15703 option, GCC will produce code that is slower and larger but which
15704 uses extremely little TOC space. You may wish to use this option
15705 only on files that contain less frequently executed code.
15711 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15712 @code{long} type, and the infrastructure needed to support them.
15713 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15714 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15715 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15718 @itemx -mno-xl-compat
15719 @opindex mxl-compat
15720 @opindex mno-xl-compat
15721 Produce code that conforms more closely to IBM XL compiler semantics
15722 when using AIX-compatible ABI@. Pass floating-point arguments to
15723 prototyped functions beyond the register save area (RSA) on the stack
15724 in addition to argument FPRs. Do not assume that most significant
15725 double in 128-bit long double value is properly rounded when comparing
15726 values and converting to double. Use XL symbol names for long double
15729 The AIX calling convention was extended but not initially documented to
15730 handle an obscure K&R C case of calling a function that takes the
15731 address of its arguments with fewer arguments than declared. IBM XL
15732 compilers access floating point arguments which do not fit in the
15733 RSA from the stack when a subroutine is compiled without
15734 optimization. Because always storing floating-point arguments on the
15735 stack is inefficient and rarely needed, this option is not enabled by
15736 default and only is necessary when calling subroutines compiled by IBM
15737 XL compilers without optimization.
15741 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15742 application written to use message passing with special startup code to
15743 enable the application to run. The system must have PE installed in the
15744 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15745 must be overridden with the @option{-specs=} option to specify the
15746 appropriate directory location. The Parallel Environment does not
15747 support threads, so the @option{-mpe} option and the @option{-pthread}
15748 option are incompatible.
15750 @item -malign-natural
15751 @itemx -malign-power
15752 @opindex malign-natural
15753 @opindex malign-power
15754 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15755 @option{-malign-natural} overrides the ABI-defined alignment of larger
15756 types, such as floating-point doubles, on their natural size-based boundary.
15757 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15758 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15760 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15764 @itemx -mhard-float
15765 @opindex msoft-float
15766 @opindex mhard-float
15767 Generate code that does not use (uses) the floating-point register set.
15768 Software floating point emulation is provided if you use the
15769 @option{-msoft-float} option, and pass the option to GCC when linking.
15771 @item -msingle-float
15772 @itemx -mdouble-float
15773 @opindex msingle-float
15774 @opindex mdouble-float
15775 Generate code for single or double-precision floating point operations.
15776 @option{-mdouble-float} implies @option{-msingle-float}.
15779 @opindex msimple-fpu
15780 Do not generate sqrt and div instructions for hardware floating point unit.
15784 Specify type of floating point unit. Valid values are @var{sp_lite}
15785 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15786 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15787 and @var{dp_full} (equivalent to -mdouble-float).
15790 @opindex mxilinx-fpu
15791 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15794 @itemx -mno-multiple
15796 @opindex mno-multiple
15797 Generate code that uses (does not use) the load multiple word
15798 instructions and the store multiple word instructions. These
15799 instructions are generated by default on POWER systems, and not
15800 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15801 endian PowerPC systems, since those instructions do not work when the
15802 processor is in little endian mode. The exceptions are PPC740 and
15803 PPC750 which permit the instructions usage in little endian mode.
15808 @opindex mno-string
15809 Generate code that uses (does not use) the load string instructions
15810 and the store string word instructions to save multiple registers and
15811 do small block moves. These instructions are generated by default on
15812 POWER systems, and not generated on PowerPC systems. Do not use
15813 @option{-mstring} on little endian PowerPC systems, since those
15814 instructions do not work when the processor is in little endian mode.
15815 The exceptions are PPC740 and PPC750 which permit the instructions
15816 usage in little endian mode.
15821 @opindex mno-update
15822 Generate code that uses (does not use) the load or store instructions
15823 that update the base register to the address of the calculated memory
15824 location. These instructions are generated by default. If you use
15825 @option{-mno-update}, there is a small window between the time that the
15826 stack pointer is updated and the address of the previous frame is
15827 stored, which means code that walks the stack frame across interrupts or
15828 signals may get corrupted data.
15830 @item -mavoid-indexed-addresses
15831 @itemx -mno-avoid-indexed-addresses
15832 @opindex mavoid-indexed-addresses
15833 @opindex mno-avoid-indexed-addresses
15834 Generate code that tries to avoid (not avoid) the use of indexed load
15835 or store instructions. These instructions can incur a performance
15836 penalty on Power6 processors in certain situations, such as when
15837 stepping through large arrays that cross a 16M boundary. This option
15838 is enabled by default when targetting Power6 and disabled otherwise.
15841 @itemx -mno-fused-madd
15842 @opindex mfused-madd
15843 @opindex mno-fused-madd
15844 Generate code that uses (does not use) the floating point multiply and
15845 accumulate instructions. These instructions are generated by default
15846 if hardware floating point is used. The machine dependent
15847 @option{-mfused-madd} option is now mapped to the machine independent
15848 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15849 mapped to @option{-ffp-contract=off}.
15855 Generate code that uses (does not use) the half-word multiply and
15856 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15857 These instructions are generated by default when targetting those
15864 Generate code that uses (does not use) the string-search @samp{dlmzb}
15865 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15866 generated by default when targetting those processors.
15868 @item -mno-bit-align
15870 @opindex mno-bit-align
15871 @opindex mbit-align
15872 On System V.4 and embedded PowerPC systems do not (do) force structures
15873 and unions that contain bit-fields to be aligned to the base type of the
15876 For example, by default a structure containing nothing but 8
15877 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15878 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15879 the structure would be aligned to a 1 byte boundary and be one byte in
15882 @item -mno-strict-align
15883 @itemx -mstrict-align
15884 @opindex mno-strict-align
15885 @opindex mstrict-align
15886 On System V.4 and embedded PowerPC systems do not (do) assume that
15887 unaligned memory references will be handled by the system.
15889 @item -mrelocatable
15890 @itemx -mno-relocatable
15891 @opindex mrelocatable
15892 @opindex mno-relocatable
15893 Generate code that allows (does not allow) a static executable to be
15894 relocated to a different address at runtime. A simple embedded
15895 PowerPC system loader should relocate the entire contents of
15896 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
15897 a table of 32-bit addresses generated by this option. For this to
15898 work, all objects linked together must be compiled with
15899 @option{-mrelocatable} or @option{-mrelocatable-lib}.
15900 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
15902 @item -mrelocatable-lib
15903 @itemx -mno-relocatable-lib
15904 @opindex mrelocatable-lib
15905 @opindex mno-relocatable-lib
15906 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
15907 @code{.fixup} section to allow static executables to be relocated at
15908 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
15909 alignment of @option{-mrelocatable}. Objects compiled with
15910 @option{-mrelocatable-lib} may be linked with objects compiled with
15911 any combination of the @option{-mrelocatable} options.
15917 On System V.4 and embedded PowerPC systems do not (do) assume that
15918 register 2 contains a pointer to a global area pointing to the addresses
15919 used in the program.
15922 @itemx -mlittle-endian
15924 @opindex mlittle-endian
15925 On System V.4 and embedded PowerPC systems compile code for the
15926 processor in little endian mode. The @option{-mlittle-endian} option is
15927 the same as @option{-mlittle}.
15930 @itemx -mbig-endian
15932 @opindex mbig-endian
15933 On System V.4 and embedded PowerPC systems compile code for the
15934 processor in big endian mode. The @option{-mbig-endian} option is
15935 the same as @option{-mbig}.
15937 @item -mdynamic-no-pic
15938 @opindex mdynamic-no-pic
15939 On Darwin and Mac OS X systems, compile code so that it is not
15940 relocatable, but that its external references are relocatable. The
15941 resulting code is suitable for applications, but not shared
15944 @item -msingle-pic-base
15945 @opindex msingle-pic-base
15946 Treat the register used for PIC addressing as read-only, rather than
15947 loading it in the prologue for each function. The run-time system is
15948 responsible for initializing this register with an appropriate value
15949 before execution begins.
15951 @item -mprioritize-restricted-insns=@var{priority}
15952 @opindex mprioritize-restricted-insns
15953 This option controls the priority that is assigned to
15954 dispatch-slot restricted instructions during the second scheduling
15955 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
15956 @var{no/highest/second-highest} priority to dispatch slot restricted
15959 @item -msched-costly-dep=@var{dependence_type}
15960 @opindex msched-costly-dep
15961 This option controls which dependences are considered costly
15962 by the target during instruction scheduling. The argument
15963 @var{dependence_type} takes one of the following values:
15964 @var{no}: no dependence is costly,
15965 @var{all}: all dependences are costly,
15966 @var{true_store_to_load}: a true dependence from store to load is costly,
15967 @var{store_to_load}: any dependence from store to load is costly,
15968 @var{number}: any dependence which latency >= @var{number} is costly.
15970 @item -minsert-sched-nops=@var{scheme}
15971 @opindex minsert-sched-nops
15972 This option controls which nop insertion scheme will be used during
15973 the second scheduling pass. The argument @var{scheme} takes one of the
15975 @var{no}: Don't insert nops.
15976 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
15977 according to the scheduler's grouping.
15978 @var{regroup_exact}: Insert nops to force costly dependent insns into
15979 separate groups. Insert exactly as many nops as needed to force an insn
15980 to a new group, according to the estimated processor grouping.
15981 @var{number}: Insert nops to force costly dependent insns into
15982 separate groups. Insert @var{number} nops to force an insn to a new group.
15985 @opindex mcall-sysv
15986 On System V.4 and embedded PowerPC systems compile code using calling
15987 conventions that adheres to the March 1995 draft of the System V
15988 Application Binary Interface, PowerPC processor supplement. This is the
15989 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
15991 @item -mcall-sysv-eabi
15993 @opindex mcall-sysv-eabi
15994 @opindex mcall-eabi
15995 Specify both @option{-mcall-sysv} and @option{-meabi} options.
15997 @item -mcall-sysv-noeabi
15998 @opindex mcall-sysv-noeabi
15999 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16001 @item -mcall-aixdesc
16003 On System V.4 and embedded PowerPC systems compile code for the AIX
16007 @opindex mcall-linux
16008 On System V.4 and embedded PowerPC systems compile code for the
16009 Linux-based GNU system.
16011 @item -mcall-freebsd
16012 @opindex mcall-freebsd
16013 On System V.4 and embedded PowerPC systems compile code for the
16014 FreeBSD operating system.
16016 @item -mcall-netbsd
16017 @opindex mcall-netbsd
16018 On System V.4 and embedded PowerPC systems compile code for the
16019 NetBSD operating system.
16021 @item -mcall-openbsd
16022 @opindex mcall-netbsd
16023 On System V.4 and embedded PowerPC systems compile code for the
16024 OpenBSD operating system.
16026 @item -maix-struct-return
16027 @opindex maix-struct-return
16028 Return all structures in memory (as specified by the AIX ABI)@.
16030 @item -msvr4-struct-return
16031 @opindex msvr4-struct-return
16032 Return structures smaller than 8 bytes in registers (as specified by the
16035 @item -mabi=@var{abi-type}
16037 Extend the current ABI with a particular extension, or remove such extension.
16038 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16039 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16043 Extend the current ABI with SPE ABI extensions. This does not change
16044 the default ABI, instead it adds the SPE ABI extensions to the current
16048 @opindex mabi=no-spe
16049 Disable Booke SPE ABI extensions for the current ABI@.
16051 @item -mabi=ibmlongdouble
16052 @opindex mabi=ibmlongdouble
16053 Change the current ABI to use IBM extended precision long double.
16054 This is a PowerPC 32-bit SYSV ABI option.
16056 @item -mabi=ieeelongdouble
16057 @opindex mabi=ieeelongdouble
16058 Change the current ABI to use IEEE extended precision long double.
16059 This is a PowerPC 32-bit Linux ABI option.
16062 @itemx -mno-prototype
16063 @opindex mprototype
16064 @opindex mno-prototype
16065 On System V.4 and embedded PowerPC systems assume that all calls to
16066 variable argument functions are properly prototyped. Otherwise, the
16067 compiler must insert an instruction before every non prototyped call to
16068 set or clear bit 6 of the condition code register (@var{CR}) to
16069 indicate whether floating point values were passed in the floating point
16070 registers in case the function takes a variable arguments. With
16071 @option{-mprototype}, only calls to prototyped variable argument functions
16072 will set or clear the bit.
16076 On embedded PowerPC systems, assume that the startup module is called
16077 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16078 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16083 On embedded PowerPC systems, assume that the startup module is called
16084 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16089 On embedded PowerPC systems, assume that the startup module is called
16090 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16093 @item -myellowknife
16094 @opindex myellowknife
16095 On embedded PowerPC systems, assume that the startup module is called
16096 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16101 On System V.4 and embedded PowerPC systems, specify that you are
16102 compiling for a VxWorks system.
16106 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16107 header to indicate that @samp{eabi} extended relocations are used.
16113 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16114 Embedded Applications Binary Interface (eabi) which is a set of
16115 modifications to the System V.4 specifications. Selecting @option{-meabi}
16116 means that the stack is aligned to an 8 byte boundary, a function
16117 @code{__eabi} is called to from @code{main} to set up the eabi
16118 environment, and the @option{-msdata} option can use both @code{r2} and
16119 @code{r13} to point to two separate small data areas. Selecting
16120 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16121 do not call an initialization function from @code{main}, and the
16122 @option{-msdata} option will only use @code{r13} to point to a single
16123 small data area. The @option{-meabi} option is on by default if you
16124 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16127 @opindex msdata=eabi
16128 On System V.4 and embedded PowerPC systems, put small initialized
16129 @code{const} global and static data in the @samp{.sdata2} section, which
16130 is pointed to by register @code{r2}. Put small initialized
16131 non-@code{const} global and static data in the @samp{.sdata} section,
16132 which is pointed to by register @code{r13}. Put small uninitialized
16133 global and static data in the @samp{.sbss} section, which is adjacent to
16134 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16135 incompatible with the @option{-mrelocatable} option. The
16136 @option{-msdata=eabi} option also sets the @option{-memb} option.
16139 @opindex msdata=sysv
16140 On System V.4 and embedded PowerPC systems, put small global and static
16141 data in the @samp{.sdata} section, which is pointed to by register
16142 @code{r13}. Put small uninitialized global and static data in the
16143 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16144 The @option{-msdata=sysv} option is incompatible with the
16145 @option{-mrelocatable} option.
16147 @item -msdata=default
16149 @opindex msdata=default
16151 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16152 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16153 same as @option{-msdata=sysv}.
16156 @opindex msdata=data
16157 On System V.4 and embedded PowerPC systems, put small global
16158 data in the @samp{.sdata} section. Put small uninitialized global
16159 data in the @samp{.sbss} section. Do not use register @code{r13}
16160 to address small data however. This is the default behavior unless
16161 other @option{-msdata} options are used.
16165 @opindex msdata=none
16167 On embedded PowerPC systems, put all initialized global and static data
16168 in the @samp{.data} section, and all uninitialized data in the
16169 @samp{.bss} section.
16171 @item -mblock-move-inline-limit=@var{num}
16172 @opindex mblock-move-inline-limit
16173 Inline all block moves (such as calls to @code{memcpy} or structure
16174 copies) less than or equal to @var{num} bytes. The minimum value for
16175 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16176 targets. The default value is target-specific.
16180 @cindex smaller data references (PowerPC)
16181 @cindex .sdata/.sdata2 references (PowerPC)
16182 On embedded PowerPC systems, put global and static items less than or
16183 equal to @var{num} bytes into the small data or bss sections instead of
16184 the normal data or bss section. By default, @var{num} is 8. The
16185 @option{-G @var{num}} switch is also passed to the linker.
16186 All modules should be compiled with the same @option{-G @var{num}} value.
16189 @itemx -mno-regnames
16191 @opindex mno-regnames
16192 On System V.4 and embedded PowerPC systems do (do not) emit register
16193 names in the assembly language output using symbolic forms.
16196 @itemx -mno-longcall
16198 @opindex mno-longcall
16199 By default assume that all calls are far away so that a longer more
16200 expensive calling sequence is required. This is required for calls
16201 further than 32 megabytes (33,554,432 bytes) from the current location.
16202 A short call will be generated if the compiler knows
16203 the call cannot be that far away. This setting can be overridden by
16204 the @code{shortcall} function attribute, or by @code{#pragma
16207 Some linkers are capable of detecting out-of-range calls and generating
16208 glue code on the fly. On these systems, long calls are unnecessary and
16209 generate slower code. As of this writing, the AIX linker can do this,
16210 as can the GNU linker for PowerPC/64. It is planned to add this feature
16211 to the GNU linker for 32-bit PowerPC systems as well.
16213 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16214 callee, L42'', plus a ``branch island'' (glue code). The two target
16215 addresses represent the callee and the ``branch island''. The
16216 Darwin/PPC linker will prefer the first address and generate a ``bl
16217 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16218 otherwise, the linker will generate ``bl L42'' to call the ``branch
16219 island''. The ``branch island'' is appended to the body of the
16220 calling function; it computes the full 32-bit address of the callee
16223 On Mach-O (Darwin) systems, this option directs the compiler emit to
16224 the glue for every direct call, and the Darwin linker decides whether
16225 to use or discard it.
16227 In the future, we may cause GCC to ignore all longcall specifications
16228 when the linker is known to generate glue.
16230 @item -mtls-markers
16231 @itemx -mno-tls-markers
16232 @opindex mtls-markers
16233 @opindex mno-tls-markers
16234 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16235 specifying the function argument. The relocation allows ld to
16236 reliably associate function call with argument setup instructions for
16237 TLS optimization, which in turn allows gcc to better schedule the
16242 Adds support for multithreading with the @dfn{pthreads} library.
16243 This option sets flags for both the preprocessor and linker.
16248 This option will enable GCC to use the reciprocal estimate and
16249 reciprocal square root estimate instructions with additional
16250 Newton-Raphson steps to increase precision instead of doing a divide or
16251 square root and divide for floating point arguments. You should use
16252 the @option{-ffast-math} option when using @option{-mrecip} (or at
16253 least @option{-funsafe-math-optimizations},
16254 @option{-finite-math-only}, @option{-freciprocal-math} and
16255 @option{-fno-trapping-math}). Note that while the throughput of the
16256 sequence is generally higher than the throughput of the non-reciprocal
16257 instruction, the precision of the sequence can be decreased by up to 2
16258 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16261 @item -mrecip=@var{opt}
16262 @opindex mrecip=opt
16263 This option allows to control which reciprocal estimate instructions
16264 may be used. @var{opt} is a comma separated list of options, that may
16265 be preceded by a @code{!} to invert the option:
16266 @code{all}: enable all estimate instructions,
16267 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16268 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16269 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16270 @code{divf}: enable the single precision reciprocal approximation instructions;
16271 @code{divd}: enable the double precision reciprocal approximation instructions;
16272 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16273 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16274 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16276 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16277 all of the reciprocal estimate instructions, except for the
16278 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16279 which handle the double precision reciprocal square root calculations.
16281 @item -mrecip-precision
16282 @itemx -mno-recip-precision
16283 @opindex mrecip-precision
16284 Assume (do not assume) that the reciprocal estimate instructions
16285 provide higher precision estimates than is mandated by the powerpc
16286 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16287 automatically selects @option{-mrecip-precision}. The double
16288 precision square root estimate instructions are not generated by
16289 default on low precision machines, since they do not provide an
16290 estimate that converges after three steps.
16292 @item -mveclibabi=@var{type}
16293 @opindex mveclibabi
16294 Specifies the ABI type to use for vectorizing intrinsics using an
16295 external library. The only type supported at present is @code{mass},
16296 which specifies to use IBM's Mathematical Acceleration Subsystem
16297 (MASS) libraries for vectorizing intrinsics using external libraries.
16298 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16299 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16300 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16301 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16302 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16303 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16304 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16305 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16306 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16307 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16308 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16309 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16310 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16311 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16312 for power7. Both @option{-ftree-vectorize} and
16313 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16314 libraries will have to be specified at link time.
16319 Generate (do not generate) the @code{friz} instruction when the
16320 @option{-funsafe-math-optimizations} option is used to optimize
16321 rounding a floating point value to 64-bit integer and back to floating
16322 point. The @code{friz} instruction does not return the same value if
16323 the floating point number is too large to fit in an integer.
16328 Generate (do not generate) code to load up the static chain register
16329 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16330 systems where a function pointer points to a 3 word descriptor giving
16331 the function address, TOC value to be loaded in register @var{r2}, and
16332 static chain value to be loaded in register @var{r11}. The
16333 @option{-mr11} is on by default. You will not be able to call through
16334 pointers to nested functions or pointers to functions compiled in
16335 other languages that use the static chain if you use the
16340 @subsection RX Options
16343 These command line options are defined for RX targets:
16346 @item -m64bit-doubles
16347 @itemx -m32bit-doubles
16348 @opindex m64bit-doubles
16349 @opindex m32bit-doubles
16350 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16351 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16352 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16353 works on 32-bit values, which is why the default is
16354 @option{-m32bit-doubles}.
16360 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16361 floating point hardware. The default is enabled for the @var{RX600}
16362 series and disabled for the @var{RX200} series.
16364 Floating point instructions will only be generated for 32-bit floating
16365 point values however, so if the @option{-m64bit-doubles} option is in
16366 use then the FPU hardware will not be used for doubles.
16368 @emph{Note} If the @option{-fpu} option is enabled then
16369 @option{-funsafe-math-optimizations} is also enabled automatically.
16370 This is because the RX FPU instructions are themselves unsafe.
16372 @item -mcpu=@var{name}
16374 Selects the type of RX CPU to be targeted. Currently three types are
16375 supported, the generic @var{RX600} and @var{RX200} series hardware and
16376 the specific @var{RX610} CPU. The default is @var{RX600}.
16378 The only difference between @var{RX600} and @var{RX610} is that the
16379 @var{RX610} does not support the @code{MVTIPL} instruction.
16381 The @var{RX200} series does not have a hardware floating point unit
16382 and so @option{-nofpu} is enabled by default when this type is
16385 @item -mbig-endian-data
16386 @itemx -mlittle-endian-data
16387 @opindex mbig-endian-data
16388 @opindex mlittle-endian-data
16389 Store data (but not code) in the big-endian format. The default is
16390 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16393 @item -msmall-data-limit=@var{N}
16394 @opindex msmall-data-limit
16395 Specifies the maximum size in bytes of global and static variables
16396 which can be placed into the small data area. Using the small data
16397 area can lead to smaller and faster code, but the size of area is
16398 limited and it is up to the programmer to ensure that the area does
16399 not overflow. Also when the small data area is used one of the RX's
16400 registers (@code{r13}) is reserved for use pointing to this area, so
16401 it is no longer available for use by the compiler. This could result
16402 in slower and/or larger code if variables which once could have been
16403 held in @code{r13} are now pushed onto the stack.
16405 Note, common variables (variables which have not been initialised) and
16406 constants are not placed into the small data area as they are assigned
16407 to other sections in the output executable.
16409 The default value is zero, which disables this feature. Note, this
16410 feature is not enabled by default with higher optimization levels
16411 (@option{-O2} etc) because of the potentially detrimental effects of
16412 reserving register @code{r13}. It is up to the programmer to
16413 experiment and discover whether this feature is of benefit to their
16420 Use the simulator runtime. The default is to use the libgloss board
16423 @item -mas100-syntax
16424 @itemx -mno-as100-syntax
16425 @opindex mas100-syntax
16426 @opindex mno-as100-syntax
16427 When generating assembler output use a syntax that is compatible with
16428 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16429 assembler but it has some restrictions so generating it is not the
16432 @item -mmax-constant-size=@var{N}
16433 @opindex mmax-constant-size
16434 Specifies the maximum size, in bytes, of a constant that can be used as
16435 an operand in a RX instruction. Although the RX instruction set does
16436 allow constants of up to 4 bytes in length to be used in instructions,
16437 a longer value equates to a longer instruction. Thus in some
16438 circumstances it can be beneficial to restrict the size of constants
16439 that are used in instructions. Constants that are too big are instead
16440 placed into a constant pool and referenced via register indirection.
16442 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16443 or 4 means that constants of any size are allowed.
16447 Enable linker relaxation. Linker relaxation is a process whereby the
16448 linker will attempt to reduce the size of a program by finding shorter
16449 versions of various instructions. Disabled by default.
16451 @item -mint-register=@var{N}
16452 @opindex mint-register
16453 Specify the number of registers to reserve for fast interrupt handler
16454 functions. The value @var{N} can be between 0 and 4. A value of 1
16455 means that register @code{r13} will be reserved for the exclusive use
16456 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16457 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16458 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16459 A value of 0, the default, does not reserve any registers.
16461 @item -msave-acc-in-interrupts
16462 @opindex msave-acc-in-interrupts
16463 Specifies that interrupt handler functions should preserve the
16464 accumulator register. This is only necessary if normal code might use
16465 the accumulator register, for example because it performs 64-bit
16466 multiplications. The default is to ignore the accumulator as this
16467 makes the interrupt handlers faster.
16471 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16472 has special significance to the RX port when used with the
16473 @code{interrupt} function attribute. This attribute indicates a
16474 function intended to process fast interrupts. GCC will will ensure
16475 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16476 and/or @code{r13} and only provided that the normal use of the
16477 corresponding registers have been restricted via the
16478 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16481 @node S/390 and zSeries Options
16482 @subsection S/390 and zSeries Options
16483 @cindex S/390 and zSeries Options
16485 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16489 @itemx -msoft-float
16490 @opindex mhard-float
16491 @opindex msoft-float
16492 Use (do not use) the hardware floating-point instructions and registers
16493 for floating-point operations. When @option{-msoft-float} is specified,
16494 functions in @file{libgcc.a} will be used to perform floating-point
16495 operations. When @option{-mhard-float} is specified, the compiler
16496 generates IEEE floating-point instructions. This is the default.
16499 @itemx -mno-hard-dfp
16501 @opindex mno-hard-dfp
16502 Use (do not use) the hardware decimal-floating-point instructions for
16503 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16504 specified, functions in @file{libgcc.a} will be used to perform
16505 decimal-floating-point operations. When @option{-mhard-dfp} is
16506 specified, the compiler generates decimal-floating-point hardware
16507 instructions. This is the default for @option{-march=z9-ec} or higher.
16509 @item -mlong-double-64
16510 @itemx -mlong-double-128
16511 @opindex mlong-double-64
16512 @opindex mlong-double-128
16513 These switches control the size of @code{long double} type. A size
16514 of 64bit makes the @code{long double} type equivalent to the @code{double}
16515 type. This is the default.
16518 @itemx -mno-backchain
16519 @opindex mbackchain
16520 @opindex mno-backchain
16521 Store (do not store) the address of the caller's frame as backchain pointer
16522 into the callee's stack frame.
16523 A backchain may be needed to allow debugging using tools that do not understand
16524 DWARF-2 call frame information.
16525 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16526 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16527 the backchain is placed into the topmost word of the 96/160 byte register
16530 In general, code compiled with @option{-mbackchain} is call-compatible with
16531 code compiled with @option{-mmo-backchain}; however, use of the backchain
16532 for debugging purposes usually requires that the whole binary is built with
16533 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16534 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16535 to build a linux kernel use @option{-msoft-float}.
16537 The default is to not maintain the backchain.
16539 @item -mpacked-stack
16540 @itemx -mno-packed-stack
16541 @opindex mpacked-stack
16542 @opindex mno-packed-stack
16543 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16544 specified, the compiler uses the all fields of the 96/160 byte register save
16545 area only for their default purpose; unused fields still take up stack space.
16546 When @option{-mpacked-stack} is specified, register save slots are densely
16547 packed at the top of the register save area; unused space is reused for other
16548 purposes, allowing for more efficient use of the available stack space.
16549 However, when @option{-mbackchain} is also in effect, the topmost word of
16550 the save area is always used to store the backchain, and the return address
16551 register is always saved two words below the backchain.
16553 As long as the stack frame backchain is not used, code generated with
16554 @option{-mpacked-stack} is call-compatible with code generated with
16555 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16556 S/390 or zSeries generated code that uses the stack frame backchain at run
16557 time, not just for debugging purposes. Such code is not call-compatible
16558 with code compiled with @option{-mpacked-stack}. Also, note that the
16559 combination of @option{-mbackchain},
16560 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16561 to build a linux kernel use @option{-msoft-float}.
16563 The default is to not use the packed stack layout.
16566 @itemx -mno-small-exec
16567 @opindex msmall-exec
16568 @opindex mno-small-exec
16569 Generate (or do not generate) code using the @code{bras} instruction
16570 to do subroutine calls.
16571 This only works reliably if the total executable size does not
16572 exceed 64k. The default is to use the @code{basr} instruction instead,
16573 which does not have this limitation.
16579 When @option{-m31} is specified, generate code compliant to the
16580 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16581 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16582 particular to generate 64-bit instructions. For the @samp{s390}
16583 targets, the default is @option{-m31}, while the @samp{s390x}
16584 targets default to @option{-m64}.
16590 When @option{-mzarch} is specified, generate code using the
16591 instructions available on z/Architecture.
16592 When @option{-mesa} is specified, generate code using the
16593 instructions available on ESA/390. Note that @option{-mesa} is
16594 not possible with @option{-m64}.
16595 When generating code compliant to the GNU/Linux for S/390 ABI,
16596 the default is @option{-mesa}. When generating code compliant
16597 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16603 Generate (or do not generate) code using the @code{mvcle} instruction
16604 to perform block moves. When @option{-mno-mvcle} is specified,
16605 use a @code{mvc} loop instead. This is the default unless optimizing for
16612 Print (or do not print) additional debug information when compiling.
16613 The default is to not print debug information.
16615 @item -march=@var{cpu-type}
16617 Generate code that will run on @var{cpu-type}, which is the name of a system
16618 representing a certain processor type. Possible values for
16619 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16620 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16621 When generating code using the instructions available on z/Architecture,
16622 the default is @option{-march=z900}. Otherwise, the default is
16623 @option{-march=g5}.
16625 @item -mtune=@var{cpu-type}
16627 Tune to @var{cpu-type} everything applicable about the generated code,
16628 except for the ABI and the set of available instructions.
16629 The list of @var{cpu-type} values is the same as for @option{-march}.
16630 The default is the value used for @option{-march}.
16633 @itemx -mno-tpf-trace
16634 @opindex mtpf-trace
16635 @opindex mno-tpf-trace
16636 Generate code that adds (does not add) in TPF OS specific branches to trace
16637 routines in the operating system. This option is off by default, even
16638 when compiling for the TPF OS@.
16641 @itemx -mno-fused-madd
16642 @opindex mfused-madd
16643 @opindex mno-fused-madd
16644 Generate code that uses (does not use) the floating point multiply and
16645 accumulate instructions. These instructions are generated by default if
16646 hardware floating point is used.
16648 @item -mwarn-framesize=@var{framesize}
16649 @opindex mwarn-framesize
16650 Emit a warning if the current function exceeds the given frame size. Because
16651 this is a compile time check it doesn't need to be a real problem when the program
16652 runs. It is intended to identify functions which most probably cause
16653 a stack overflow. It is useful to be used in an environment with limited stack
16654 size e.g.@: the linux kernel.
16656 @item -mwarn-dynamicstack
16657 @opindex mwarn-dynamicstack
16658 Emit a warning if the function calls alloca or uses dynamically
16659 sized arrays. This is generally a bad idea with a limited stack size.
16661 @item -mstack-guard=@var{stack-guard}
16662 @itemx -mstack-size=@var{stack-size}
16663 @opindex mstack-guard
16664 @opindex mstack-size
16665 If these options are provided the s390 back end emits additional instructions in
16666 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16667 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16668 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16669 the frame size of the compiled function is chosen.
16670 These options are intended to be used to help debugging stack overflow problems.
16671 The additionally emitted code causes only little overhead and hence can also be
16672 used in production like systems without greater performance degradation. The given
16673 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16674 @var{stack-guard} without exceeding 64k.
16675 In order to be efficient the extra code makes the assumption that the stack starts
16676 at an address aligned to the value given by @var{stack-size}.
16677 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16680 @node Score Options
16681 @subsection Score Options
16682 @cindex Score Options
16684 These options are defined for Score implementations:
16689 Compile code for big endian mode. This is the default.
16693 Compile code for little endian mode.
16697 Disable generate bcnz instruction.
16701 Enable generate unaligned load and store instruction.
16705 Enable the use of multiply-accumulate instructions. Disabled by default.
16709 Specify the SCORE5 as the target architecture.
16713 Specify the SCORE5U of the target architecture.
16717 Specify the SCORE7 as the target architecture. This is the default.
16721 Specify the SCORE7D as the target architecture.
16725 @subsection SH Options
16727 These @samp{-m} options are defined for the SH implementations:
16732 Generate code for the SH1.
16736 Generate code for the SH2.
16739 Generate code for the SH2e.
16743 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16744 that the floating-point unit is not used.
16746 @item -m2a-single-only
16747 @opindex m2a-single-only
16748 Generate code for the SH2a-FPU, in such a way that no double-precision
16749 floating point operations are used.
16752 @opindex m2a-single
16753 Generate code for the SH2a-FPU assuming the floating-point unit is in
16754 single-precision mode by default.
16758 Generate code for the SH2a-FPU assuming the floating-point unit is in
16759 double-precision mode by default.
16763 Generate code for the SH3.
16767 Generate code for the SH3e.
16771 Generate code for the SH4 without a floating-point unit.
16773 @item -m4-single-only
16774 @opindex m4-single-only
16775 Generate code for the SH4 with a floating-point unit that only
16776 supports single-precision arithmetic.
16780 Generate code for the SH4 assuming the floating-point unit is in
16781 single-precision mode by default.
16785 Generate code for the SH4.
16789 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16790 floating-point unit is not used.
16792 @item -m4a-single-only
16793 @opindex m4a-single-only
16794 Generate code for the SH4a, in such a way that no double-precision
16795 floating point operations are used.
16798 @opindex m4a-single
16799 Generate code for the SH4a assuming the floating-point unit is in
16800 single-precision mode by default.
16804 Generate code for the SH4a.
16808 Same as @option{-m4a-nofpu}, except that it implicitly passes
16809 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16810 instructions at the moment.
16814 Compile code for the processor in big endian mode.
16818 Compile code for the processor in little endian mode.
16822 Align doubles at 64-bit boundaries. Note that this changes the calling
16823 conventions, and thus some functions from the standard C library will
16824 not work unless you recompile it first with @option{-mdalign}.
16828 Shorten some address references at link time, when possible; uses the
16829 linker option @option{-relax}.
16833 Use 32-bit offsets in @code{switch} tables. The default is to use
16838 Enable the use of bit manipulation instructions on SH2A.
16842 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16843 alignment constraints.
16847 Comply with the calling conventions defined by Renesas.
16851 Comply with the calling conventions defined by Renesas.
16855 Comply with the calling conventions defined for GCC before the Renesas
16856 conventions were available. This option is the default for all
16857 targets of the SH toolchain.
16860 @opindex mnomacsave
16861 Mark the @code{MAC} register as call-clobbered, even if
16862 @option{-mhitachi} is given.
16866 Increase IEEE-compliance of floating-point code.
16867 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16868 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16869 comparisons of NANs / infinities incurs extra overhead in every
16870 floating point comparison, therefore the default is set to
16871 @option{-ffinite-math-only}.
16873 @item -minline-ic_invalidate
16874 @opindex minline-ic_invalidate
16875 Inline code to invalidate instruction cache entries after setting up
16876 nested function trampolines.
16877 This option has no effect if -musermode is in effect and the selected
16878 code generation option (e.g. -m4) does not allow the use of the icbi
16880 If the selected code generation option does not allow the use of the icbi
16881 instruction, and -musermode is not in effect, the inlined code will
16882 manipulate the instruction cache address array directly with an associative
16883 write. This not only requires privileged mode, but it will also
16884 fail if the cache line had been mapped via the TLB and has become unmapped.
16888 Dump instruction size and location in the assembly code.
16891 @opindex mpadstruct
16892 This option is deprecated. It pads structures to multiple of 4 bytes,
16893 which is incompatible with the SH ABI@.
16897 Optimize for space instead of speed. Implied by @option{-Os}.
16900 @opindex mprefergot
16901 When generating position-independent code, emit function calls using
16902 the Global Offset Table instead of the Procedure Linkage Table.
16906 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
16907 if the inlined code would not work in user mode.
16908 This is the default when the target is @code{sh-*-linux*}.
16910 @item -multcost=@var{number}
16911 @opindex multcost=@var{number}
16912 Set the cost to assume for a multiply insn.
16914 @item -mdiv=@var{strategy}
16915 @opindex mdiv=@var{strategy}
16916 Set the division strategy to use for SHmedia code. @var{strategy} must be
16917 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
16918 inv:call2, inv:fp .
16919 "fp" performs the operation in floating point. This has a very high latency,
16920 but needs only a few instructions, so it might be a good choice if
16921 your code has enough easily exploitable ILP to allow the compiler to
16922 schedule the floating point instructions together with other instructions.
16923 Division by zero causes a floating point exception.
16924 "inv" uses integer operations to calculate the inverse of the divisor,
16925 and then multiplies the dividend with the inverse. This strategy allows
16926 cse and hoisting of the inverse calculation. Division by zero calculates
16927 an unspecified result, but does not trap.
16928 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
16929 have been found, or if the entire operation has been hoisted to the same
16930 place, the last stages of the inverse calculation are intertwined with the
16931 final multiply to reduce the overall latency, at the expense of using a few
16932 more instructions, and thus offering fewer scheduling opportunities with
16934 "call" calls a library function that usually implements the inv:minlat
16936 This gives high code density for m5-*media-nofpu compilations.
16937 "call2" uses a different entry point of the same library function, where it
16938 assumes that a pointer to a lookup table has already been set up, which
16939 exposes the pointer load to cse / code hoisting optimizations.
16940 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
16941 code generation, but if the code stays unoptimized, revert to the "call",
16942 "call2", or "fp" strategies, respectively. Note that the
16943 potentially-trapping side effect of division by zero is carried by a
16944 separate instruction, so it is possible that all the integer instructions
16945 are hoisted out, but the marker for the side effect stays where it is.
16946 A recombination to fp operations or a call is not possible in that case.
16947 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
16948 that the inverse calculation was nor separated from the multiply, they speed
16949 up division where the dividend fits into 20 bits (plus sign where applicable),
16950 by inserting a test to skip a number of operations in this case; this test
16951 slows down the case of larger dividends. inv20u assumes the case of a such
16952 a small dividend to be unlikely, and inv20l assumes it to be likely.
16954 @item -maccumulate-outgoing-args
16955 @opindex maccumulate-outgoing-args
16956 Reserve space once for outgoing arguments in the function prologue rather
16957 than around each call. Generally beneficial for performance and size. Also
16958 needed for unwinding to avoid changing the stack frame around conditional code.
16960 @item -mdivsi3_libfunc=@var{name}
16961 @opindex mdivsi3_libfunc=@var{name}
16962 Set the name of the library function used for 32 bit signed division to
16963 @var{name}. This only affect the name used in the call and inv:call
16964 division strategies, and the compiler will still expect the same
16965 sets of input/output/clobbered registers as if this option was not present.
16967 @item -mfixed-range=@var{register-range}
16968 @opindex mfixed-range
16969 Generate code treating the given register range as fixed registers.
16970 A fixed register is one that the register allocator can not use. This is
16971 useful when compiling kernel code. A register range is specified as
16972 two registers separated by a dash. Multiple register ranges can be
16973 specified separated by a comma.
16975 @item -madjust-unroll
16976 @opindex madjust-unroll
16977 Throttle unrolling to avoid thrashing target registers.
16978 This option only has an effect if the gcc code base supports the
16979 TARGET_ADJUST_UNROLL_MAX target hook.
16981 @item -mindexed-addressing
16982 @opindex mindexed-addressing
16983 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
16984 This is only safe if the hardware and/or OS implement 32 bit wrap-around
16985 semantics for the indexed addressing mode. The architecture allows the
16986 implementation of processors with 64 bit MMU, which the OS could use to
16987 get 32 bit addressing, but since no current hardware implementation supports
16988 this or any other way to make the indexed addressing mode safe to use in
16989 the 32 bit ABI, the default is -mno-indexed-addressing.
16991 @item -mgettrcost=@var{number}
16992 @opindex mgettrcost=@var{number}
16993 Set the cost assumed for the gettr instruction to @var{number}.
16994 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
16998 Assume pt* instructions won't trap. This will generally generate better
16999 scheduled code, but is unsafe on current hardware. The current architecture
17000 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17001 This has the unintentional effect of making it unsafe to schedule ptabs /
17002 ptrel before a branch, or hoist it out of a loop. For example,
17003 __do_global_ctors, a part of libgcc that runs constructors at program
17004 startup, calls functions in a list which is delimited by @minus{}1. With the
17005 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17006 That means that all the constructors will be run a bit quicker, but when
17007 the loop comes to the end of the list, the program crashes because ptabs
17008 loads @minus{}1 into a target register. Since this option is unsafe for any
17009 hardware implementing the current architecture specification, the default
17010 is -mno-pt-fixed. Unless the user specifies a specific cost with
17011 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17012 this deters register allocation using target registers for storing
17015 @item -minvalid-symbols
17016 @opindex minvalid-symbols
17017 Assume symbols might be invalid. Ordinary function symbols generated by
17018 the compiler will always be valid to load with movi/shori/ptabs or
17019 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17020 to generate symbols that will cause ptabs / ptrel to trap.
17021 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17022 It will then prevent cross-basic-block cse, hoisting and most scheduling
17023 of symbol loads. The default is @option{-mno-invalid-symbols}.
17026 @node Solaris 2 Options
17027 @subsection Solaris 2 Options
17028 @cindex Solaris 2 options
17030 These @samp{-m} options are supported on Solaris 2:
17033 @item -mimpure-text
17034 @opindex mimpure-text
17035 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17036 the compiler to not pass @option{-z text} to the linker when linking a
17037 shared object. Using this option, you can link position-dependent
17038 code into a shared object.
17040 @option{-mimpure-text} suppresses the ``relocations remain against
17041 allocatable but non-writable sections'' linker error message.
17042 However, the necessary relocations will trigger copy-on-write, and the
17043 shared object is not actually shared across processes. Instead of
17044 using @option{-mimpure-text}, you should compile all source code with
17045 @option{-fpic} or @option{-fPIC}.
17049 These switches are supported in addition to the above on Solaris 2:
17054 Add support for multithreading using the POSIX threads library. This
17055 option sets flags for both the preprocessor and linker. This option does
17056 not affect the thread safety of object code produced by the compiler or
17057 that of libraries supplied with it.
17061 This is a synonym for @option{-pthreads}.
17064 @node SPARC Options
17065 @subsection SPARC Options
17066 @cindex SPARC options
17068 These @samp{-m} options are supported on the SPARC:
17071 @item -mno-app-regs
17073 @opindex mno-app-regs
17075 Specify @option{-mapp-regs} to generate output using the global registers
17076 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17079 To be fully SVR4 ABI compliant at the cost of some performance loss,
17080 specify @option{-mno-app-regs}. You should compile libraries and system
17081 software with this option.
17087 With @option{-mflat}, the compiler does not generate save/restore instructions
17088 and uses a ``flat'' or single register window model. This model is compatible
17089 with the regular register window model. The local registers and the input
17090 registers (0--5) are still treated as ``call-saved'' registers and will be
17091 saved on the stack as needed.
17093 With @option{-mno-flat} (the default), the compiler generates save/restore
17094 instructions (except for leaf functions). This is the normal operating mode.
17097 @itemx -mhard-float
17099 @opindex mhard-float
17100 Generate output containing floating point instructions. This is the
17104 @itemx -msoft-float
17106 @opindex msoft-float
17107 Generate output containing library calls for floating point.
17108 @strong{Warning:} the requisite libraries are not available for all SPARC
17109 targets. Normally the facilities of the machine's usual C compiler are
17110 used, but this cannot be done directly in cross-compilation. You must make
17111 your own arrangements to provide suitable library functions for
17112 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17113 @samp{sparclite-*-*} do provide software floating point support.
17115 @option{-msoft-float} changes the calling convention in the output file;
17116 therefore, it is only useful if you compile @emph{all} of a program with
17117 this option. In particular, you need to compile @file{libgcc.a}, the
17118 library that comes with GCC, with @option{-msoft-float} in order for
17121 @item -mhard-quad-float
17122 @opindex mhard-quad-float
17123 Generate output containing quad-word (long double) floating point
17126 @item -msoft-quad-float
17127 @opindex msoft-quad-float
17128 Generate output containing library calls for quad-word (long double)
17129 floating point instructions. The functions called are those specified
17130 in the SPARC ABI@. This is the default.
17132 As of this writing, there are no SPARC implementations that have hardware
17133 support for the quad-word floating point instructions. They all invoke
17134 a trap handler for one of these instructions, and then the trap handler
17135 emulates the effect of the instruction. Because of the trap handler overhead,
17136 this is much slower than calling the ABI library routines. Thus the
17137 @option{-msoft-quad-float} option is the default.
17139 @item -mno-unaligned-doubles
17140 @itemx -munaligned-doubles
17141 @opindex mno-unaligned-doubles
17142 @opindex munaligned-doubles
17143 Assume that doubles have 8 byte alignment. This is the default.
17145 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17146 alignment only if they are contained in another type, or if they have an
17147 absolute address. Otherwise, it assumes they have 4 byte alignment.
17148 Specifying this option avoids some rare compatibility problems with code
17149 generated by other compilers. It is not the default because it results
17150 in a performance loss, especially for floating point code.
17152 @item -mno-faster-structs
17153 @itemx -mfaster-structs
17154 @opindex mno-faster-structs
17155 @opindex mfaster-structs
17156 With @option{-mfaster-structs}, the compiler assumes that structures
17157 should have 8 byte alignment. This enables the use of pairs of
17158 @code{ldd} and @code{std} instructions for copies in structure
17159 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17160 However, the use of this changed alignment directly violates the SPARC
17161 ABI@. Thus, it's intended only for use on targets where the developer
17162 acknowledges that their resulting code will not be directly in line with
17163 the rules of the ABI@.
17165 @item -mcpu=@var{cpu_type}
17167 Set the instruction set, register set, and instruction scheduling parameters
17168 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17169 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17170 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17171 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17172 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
17174 Default instruction scheduling parameters are used for values that select
17175 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17176 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17178 Here is a list of each supported architecture and their supported
17183 v8: supersparc, hypersparc, leon
17184 sparclite: f930, f934, sparclite86x
17186 v9: ultrasparc, ultrasparc3, niagara, niagara2
17189 By default (unless configured otherwise), GCC generates code for the V7
17190 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17191 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17192 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17193 SPARCStation 1, 2, IPX etc.
17195 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17196 architecture. The only difference from V7 code is that the compiler emits
17197 the integer multiply and integer divide instructions which exist in SPARC-V8
17198 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17199 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17202 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17203 the SPARC architecture. This adds the integer multiply, integer divide step
17204 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17205 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17206 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17207 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17208 MB86934 chip, which is the more recent SPARClite with FPU@.
17210 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17211 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17212 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17213 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17214 optimizes it for the TEMIC SPARClet chip.
17216 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17217 architecture. This adds 64-bit integer and floating-point move instructions,
17218 3 additional floating-point condition code registers and conditional move
17219 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17220 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17221 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17222 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17223 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17224 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17225 additionally optimizes it for Sun UltraSPARC T2 chips.
17227 @item -mtune=@var{cpu_type}
17229 Set the instruction scheduling parameters for machine type
17230 @var{cpu_type}, but do not set the instruction set or register set that the
17231 option @option{-mcpu=@var{cpu_type}} would.
17233 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17234 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17235 that select a particular CPU implementation. Those are @samp{cypress},
17236 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17237 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17238 @samp{niagara}, and @samp{niagara2}.
17243 @opindex mno-v8plus
17244 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17245 difference from the V8 ABI is that the global and out registers are
17246 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17247 mode for all SPARC-V9 processors.
17253 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17254 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17257 These @samp{-m} options are supported in addition to the above
17258 on SPARC-V9 processors in 64-bit environments:
17261 @item -mlittle-endian
17262 @opindex mlittle-endian
17263 Generate code for a processor running in little-endian mode. It is only
17264 available for a few configurations and most notably not on Solaris and Linux.
17270 Generate code for a 32-bit or 64-bit environment.
17271 The 32-bit environment sets int, long and pointer to 32 bits.
17272 The 64-bit environment sets int to 32 bits and long and pointer
17275 @item -mcmodel=medlow
17276 @opindex mcmodel=medlow
17277 Generate code for the Medium/Low code model: 64-bit addresses, programs
17278 must be linked in the low 32 bits of memory. Programs can be statically
17279 or dynamically linked.
17281 @item -mcmodel=medmid
17282 @opindex mcmodel=medmid
17283 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17284 must be linked in the low 44 bits of memory, the text and data segments must
17285 be less than 2GB in size and the data segment must be located within 2GB of
17288 @item -mcmodel=medany
17289 @opindex mcmodel=medany
17290 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17291 may be linked anywhere in memory, the text and data segments must be less
17292 than 2GB in size and the data segment must be located within 2GB of the
17295 @item -mcmodel=embmedany
17296 @opindex mcmodel=embmedany
17297 Generate code for the Medium/Anywhere code model for embedded systems:
17298 64-bit addresses, the text and data segments must be less than 2GB in
17299 size, both starting anywhere in memory (determined at link time). The
17300 global register %g4 points to the base of the data segment. Programs
17301 are statically linked and PIC is not supported.
17304 @itemx -mno-stack-bias
17305 @opindex mstack-bias
17306 @opindex mno-stack-bias
17307 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17308 frame pointer if present, are offset by @minus{}2047 which must be added back
17309 when making stack frame references. This is the default in 64-bit mode.
17310 Otherwise, assume no such offset is present.
17314 @subsection SPU Options
17315 @cindex SPU options
17317 These @samp{-m} options are supported on the SPU:
17321 @itemx -merror-reloc
17322 @opindex mwarn-reloc
17323 @opindex merror-reloc
17325 The loader for SPU does not handle dynamic relocations. By default, GCC
17326 will give an error when it generates code that requires a dynamic
17327 relocation. @option{-mno-error-reloc} disables the error,
17328 @option{-mwarn-reloc} will generate a warning instead.
17331 @itemx -munsafe-dma
17333 @opindex munsafe-dma
17335 Instructions which initiate or test completion of DMA must not be
17336 reordered with respect to loads and stores of the memory which is being
17337 accessed. Users typically address this problem using the volatile
17338 keyword, but that can lead to inefficient code in places where the
17339 memory is known to not change. Rather than mark the memory as volatile
17340 we treat the DMA instructions as potentially effecting all memory. With
17341 @option{-munsafe-dma} users must use the volatile keyword to protect
17344 @item -mbranch-hints
17345 @opindex mbranch-hints
17347 By default, GCC will generate a branch hint instruction to avoid
17348 pipeline stalls for always taken or probably taken branches. A hint
17349 will not be generated closer than 8 instructions away from its branch.
17350 There is little reason to disable them, except for debugging purposes,
17351 or to make an object a little bit smaller.
17355 @opindex msmall-mem
17356 @opindex mlarge-mem
17358 By default, GCC generates code assuming that addresses are never larger
17359 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17360 a full 32 bit address.
17365 By default, GCC links against startup code that assumes the SPU-style
17366 main function interface (which has an unconventional parameter list).
17367 With @option{-mstdmain}, GCC will link your program against startup
17368 code that assumes a C99-style interface to @code{main}, including a
17369 local copy of @code{argv} strings.
17371 @item -mfixed-range=@var{register-range}
17372 @opindex mfixed-range
17373 Generate code treating the given register range as fixed registers.
17374 A fixed register is one that the register allocator can not use. This is
17375 useful when compiling kernel code. A register range is specified as
17376 two registers separated by a dash. Multiple register ranges can be
17377 specified separated by a comma.
17383 Compile code assuming that pointers to the PPU address space accessed
17384 via the @code{__ea} named address space qualifier are either 32 or 64
17385 bits wide. The default is 32 bits. As this is an ABI changing option,
17386 all object code in an executable must be compiled with the same setting.
17388 @item -maddress-space-conversion
17389 @itemx -mno-address-space-conversion
17390 @opindex maddress-space-conversion
17391 @opindex mno-address-space-conversion
17392 Allow/disallow treating the @code{__ea} address space as superset
17393 of the generic address space. This enables explicit type casts
17394 between @code{__ea} and generic pointer as well as implicit
17395 conversions of generic pointers to @code{__ea} pointers. The
17396 default is to allow address space pointer conversions.
17398 @item -mcache-size=@var{cache-size}
17399 @opindex mcache-size
17400 This option controls the version of libgcc that the compiler links to an
17401 executable and selects a software-managed cache for accessing variables
17402 in the @code{__ea} address space with a particular cache size. Possible
17403 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17404 and @samp{128}. The default cache size is 64KB.
17406 @item -matomic-updates
17407 @itemx -mno-atomic-updates
17408 @opindex matomic-updates
17409 @opindex mno-atomic-updates
17410 This option controls the version of libgcc that the compiler links to an
17411 executable and selects whether atomic updates to the software-managed
17412 cache of PPU-side variables are used. If you use atomic updates, changes
17413 to a PPU variable from SPU code using the @code{__ea} named address space
17414 qualifier will not interfere with changes to other PPU variables residing
17415 in the same cache line from PPU code. If you do not use atomic updates,
17416 such interference may occur; however, writing back cache lines will be
17417 more efficient. The default behavior is to use atomic updates.
17420 @itemx -mdual-nops=@var{n}
17421 @opindex mdual-nops
17422 By default, GCC will insert nops to increase dual issue when it expects
17423 it to increase performance. @var{n} can be a value from 0 to 10. A
17424 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17425 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17427 @item -mhint-max-nops=@var{n}
17428 @opindex mhint-max-nops
17429 Maximum number of nops to insert for a branch hint. A branch hint must
17430 be at least 8 instructions away from the branch it is effecting. GCC
17431 will insert up to @var{n} nops to enforce this, otherwise it will not
17432 generate the branch hint.
17434 @item -mhint-max-distance=@var{n}
17435 @opindex mhint-max-distance
17436 The encoding of the branch hint instruction limits the hint to be within
17437 256 instructions of the branch it is effecting. By default, GCC makes
17438 sure it is within 125.
17441 @opindex msafe-hints
17442 Work around a hardware bug which causes the SPU to stall indefinitely.
17443 By default, GCC will insert the @code{hbrp} instruction to make sure
17444 this stall won't happen.
17448 @node System V Options
17449 @subsection Options for System V
17451 These additional options are available on System V Release 4 for
17452 compatibility with other compilers on those systems:
17457 Create a shared object.
17458 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17462 Identify the versions of each tool used by the compiler, in a
17463 @code{.ident} assembler directive in the output.
17467 Refrain from adding @code{.ident} directives to the output file (this is
17470 @item -YP,@var{dirs}
17472 Search the directories @var{dirs}, and no others, for libraries
17473 specified with @option{-l}.
17475 @item -Ym,@var{dir}
17477 Look in the directory @var{dir} to find the M4 preprocessor.
17478 The assembler uses this option.
17479 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17480 @c the generic assembler that comes with Solaris takes just -Ym.
17484 @subsection V850 Options
17485 @cindex V850 Options
17487 These @samp{-m} options are defined for V850 implementations:
17491 @itemx -mno-long-calls
17492 @opindex mlong-calls
17493 @opindex mno-long-calls
17494 Treat all calls as being far away (near). If calls are assumed to be
17495 far away, the compiler will always load the functions address up into a
17496 register, and call indirect through the pointer.
17502 Do not optimize (do optimize) basic blocks that use the same index
17503 pointer 4 or more times to copy pointer into the @code{ep} register, and
17504 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17505 option is on by default if you optimize.
17507 @item -mno-prolog-function
17508 @itemx -mprolog-function
17509 @opindex mno-prolog-function
17510 @opindex mprolog-function
17511 Do not use (do use) external functions to save and restore registers
17512 at the prologue and epilogue of a function. The external functions
17513 are slower, but use less code space if more than one function saves
17514 the same number of registers. The @option{-mprolog-function} option
17515 is on by default if you optimize.
17519 Try to make the code as small as possible. At present, this just turns
17520 on the @option{-mep} and @option{-mprolog-function} options.
17522 @item -mtda=@var{n}
17524 Put static or global variables whose size is @var{n} bytes or less into
17525 the tiny data area that register @code{ep} points to. The tiny data
17526 area can hold up to 256 bytes in total (128 bytes for byte references).
17528 @item -msda=@var{n}
17530 Put static or global variables whose size is @var{n} bytes or less into
17531 the small data area that register @code{gp} points to. The small data
17532 area can hold up to 64 kilobytes.
17534 @item -mzda=@var{n}
17536 Put static or global variables whose size is @var{n} bytes or less into
17537 the first 32 kilobytes of memory.
17541 Specify that the target processor is the V850.
17544 @opindex mbig-switch
17545 Generate code suitable for big switch tables. Use this option only if
17546 the assembler/linker complain about out of range branches within a switch
17551 This option will cause r2 and r5 to be used in the code generated by
17552 the compiler. This setting is the default.
17554 @item -mno-app-regs
17555 @opindex mno-app-regs
17556 This option will cause r2 and r5 to be treated as fixed registers.
17560 Specify that the target processor is the V850E2V3. The preprocessor
17561 constants @samp{__v850e2v3__} will be defined if
17562 this option is used.
17566 Specify that the target processor is the V850E2. The preprocessor
17567 constants @samp{__v850e2__} will be defined if this option is used.
17571 Specify that the target processor is the V850E1. The preprocessor
17572 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17573 this option is used.
17577 Specify that the target processor is the V850ES. This is an alias for
17578 the @option{-mv850e1} option.
17582 Specify that the target processor is the V850E@. The preprocessor
17583 constant @samp{__v850e__} will be defined if this option is used.
17585 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17586 nor @option{-mv850e2} nor @option{-mv850e2v3}
17587 are defined then a default target processor will be chosen and the
17588 relevant @samp{__v850*__} preprocessor constant will be defined.
17590 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17591 defined, regardless of which processor variant is the target.
17593 @item -mdisable-callt
17594 @opindex mdisable-callt
17595 This option will suppress generation of the CALLT instruction for the
17596 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17597 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17602 @subsection VAX Options
17603 @cindex VAX options
17605 These @samp{-m} options are defined for the VAX:
17610 Do not output certain jump instructions (@code{aobleq} and so on)
17611 that the Unix assembler for the VAX cannot handle across long
17616 Do output those jump instructions, on the assumption that you
17617 will assemble with the GNU assembler.
17621 Output code for g-format floating point numbers instead of d-format.
17624 @node VxWorks Options
17625 @subsection VxWorks Options
17626 @cindex VxWorks Options
17628 The options in this section are defined for all VxWorks targets.
17629 Options specific to the target hardware are listed with the other
17630 options for that target.
17635 GCC can generate code for both VxWorks kernels and real time processes
17636 (RTPs). This option switches from the former to the latter. It also
17637 defines the preprocessor macro @code{__RTP__}.
17640 @opindex non-static
17641 Link an RTP executable against shared libraries rather than static
17642 libraries. The options @option{-static} and @option{-shared} can
17643 also be used for RTPs (@pxref{Link Options}); @option{-static}
17650 These options are passed down to the linker. They are defined for
17651 compatibility with Diab.
17654 @opindex Xbind-lazy
17655 Enable lazy binding of function calls. This option is equivalent to
17656 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17660 Disable lazy binding of function calls. This option is the default and
17661 is defined for compatibility with Diab.
17664 @node x86-64 Options
17665 @subsection x86-64 Options
17666 @cindex x86-64 options
17668 These are listed under @xref{i386 and x86-64 Options}.
17670 @node Xstormy16 Options
17671 @subsection Xstormy16 Options
17672 @cindex Xstormy16 Options
17674 These options are defined for Xstormy16:
17679 Choose startup files and linker script suitable for the simulator.
17682 @node Xtensa Options
17683 @subsection Xtensa Options
17684 @cindex Xtensa Options
17686 These options are supported for Xtensa targets:
17690 @itemx -mno-const16
17692 @opindex mno-const16
17693 Enable or disable use of @code{CONST16} instructions for loading
17694 constant values. The @code{CONST16} instruction is currently not a
17695 standard option from Tensilica. When enabled, @code{CONST16}
17696 instructions are always used in place of the standard @code{L32R}
17697 instructions. The use of @code{CONST16} is enabled by default only if
17698 the @code{L32R} instruction is not available.
17701 @itemx -mno-fused-madd
17702 @opindex mfused-madd
17703 @opindex mno-fused-madd
17704 Enable or disable use of fused multiply/add and multiply/subtract
17705 instructions in the floating-point option. This has no effect if the
17706 floating-point option is not also enabled. Disabling fused multiply/add
17707 and multiply/subtract instructions forces the compiler to use separate
17708 instructions for the multiply and add/subtract operations. This may be
17709 desirable in some cases where strict IEEE 754-compliant results are
17710 required: the fused multiply add/subtract instructions do not round the
17711 intermediate result, thereby producing results with @emph{more} bits of
17712 precision than specified by the IEEE standard. Disabling fused multiply
17713 add/subtract instructions also ensures that the program output is not
17714 sensitive to the compiler's ability to combine multiply and add/subtract
17717 @item -mserialize-volatile
17718 @itemx -mno-serialize-volatile
17719 @opindex mserialize-volatile
17720 @opindex mno-serialize-volatile
17721 When this option is enabled, GCC inserts @code{MEMW} instructions before
17722 @code{volatile} memory references to guarantee sequential consistency.
17723 The default is @option{-mserialize-volatile}. Use
17724 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17726 @item -mforce-no-pic
17727 @opindex mforce-no-pic
17728 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17729 position-independent code (PIC), this option disables PIC for compiling
17732 @item -mtext-section-literals
17733 @itemx -mno-text-section-literals
17734 @opindex mtext-section-literals
17735 @opindex mno-text-section-literals
17736 Control the treatment of literal pools. The default is
17737 @option{-mno-text-section-literals}, which places literals in a separate
17738 section in the output file. This allows the literal pool to be placed
17739 in a data RAM/ROM, and it also allows the linker to combine literal
17740 pools from separate object files to remove redundant literals and
17741 improve code size. With @option{-mtext-section-literals}, the literals
17742 are interspersed in the text section in order to keep them as close as
17743 possible to their references. This may be necessary for large assembly
17746 @item -mtarget-align
17747 @itemx -mno-target-align
17748 @opindex mtarget-align
17749 @opindex mno-target-align
17750 When this option is enabled, GCC instructs the assembler to
17751 automatically align instructions to reduce branch penalties at the
17752 expense of some code density. The assembler attempts to widen density
17753 instructions to align branch targets and the instructions following call
17754 instructions. If there are not enough preceding safe density
17755 instructions to align a target, no widening will be performed. The
17756 default is @option{-mtarget-align}. These options do not affect the
17757 treatment of auto-aligned instructions like @code{LOOP}, which the
17758 assembler will always align, either by widening density instructions or
17759 by inserting no-op instructions.
17762 @itemx -mno-longcalls
17763 @opindex mlongcalls
17764 @opindex mno-longcalls
17765 When this option is enabled, GCC instructs the assembler to translate
17766 direct calls to indirect calls unless it can determine that the target
17767 of a direct call is in the range allowed by the call instruction. This
17768 translation typically occurs for calls to functions in other source
17769 files. Specifically, the assembler translates a direct @code{CALL}
17770 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17771 The default is @option{-mno-longcalls}. This option should be used in
17772 programs where the call target can potentially be out of range. This
17773 option is implemented in the assembler, not the compiler, so the
17774 assembly code generated by GCC will still show direct call
17775 instructions---look at the disassembled object code to see the actual
17776 instructions. Note that the assembler will use an indirect call for
17777 every cross-file call, not just those that really will be out of range.
17780 @node zSeries Options
17781 @subsection zSeries Options
17782 @cindex zSeries options
17784 These are listed under @xref{S/390 and zSeries Options}.
17786 @node Code Gen Options
17787 @section Options for Code Generation Conventions
17788 @cindex code generation conventions
17789 @cindex options, code generation
17790 @cindex run-time options
17792 These machine-independent options control the interface conventions
17793 used in code generation.
17795 Most of them have both positive and negative forms; the negative form
17796 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17797 one of the forms is listed---the one which is not the default. You
17798 can figure out the other form by either removing @samp{no-} or adding
17802 @item -fbounds-check
17803 @opindex fbounds-check
17804 For front-ends that support it, generate additional code to check that
17805 indices used to access arrays are within the declared range. This is
17806 currently only supported by the Java and Fortran front-ends, where
17807 this option defaults to true and false respectively.
17811 This option generates traps for signed overflow on addition, subtraction,
17812 multiplication operations.
17816 This option instructs the compiler to assume that signed arithmetic
17817 overflow of addition, subtraction and multiplication wraps around
17818 using twos-complement representation. This flag enables some optimizations
17819 and disables others. This option is enabled by default for the Java
17820 front-end, as required by the Java language specification.
17823 @opindex fexceptions
17824 Enable exception handling. Generates extra code needed to propagate
17825 exceptions. For some targets, this implies GCC will generate frame
17826 unwind information for all functions, which can produce significant data
17827 size overhead, although it does not affect execution. If you do not
17828 specify this option, GCC will enable it by default for languages like
17829 C++ which normally require exception handling, and disable it for
17830 languages like C that do not normally require it. However, you may need
17831 to enable this option when compiling C code that needs to interoperate
17832 properly with exception handlers written in C++. You may also wish to
17833 disable this option if you are compiling older C++ programs that don't
17834 use exception handling.
17836 @item -fnon-call-exceptions
17837 @opindex fnon-call-exceptions
17838 Generate code that allows trapping instructions to throw exceptions.
17839 Note that this requires platform-specific runtime support that does
17840 not exist everywhere. Moreover, it only allows @emph{trapping}
17841 instructions to throw exceptions, i.e.@: memory references or floating
17842 point instructions. It does not allow exceptions to be thrown from
17843 arbitrary signal handlers such as @code{SIGALRM}.
17845 @item -funwind-tables
17846 @opindex funwind-tables
17847 Similar to @option{-fexceptions}, except that it will just generate any needed
17848 static data, but will not affect the generated code in any other way.
17849 You will normally not enable this option; instead, a language processor
17850 that needs this handling would enable it on your behalf.
17852 @item -fasynchronous-unwind-tables
17853 @opindex fasynchronous-unwind-tables
17854 Generate unwind table in dwarf2 format, if supported by target machine. The
17855 table is exact at each instruction boundary, so it can be used for stack
17856 unwinding from asynchronous events (such as debugger or garbage collector).
17858 @item -fpcc-struct-return
17859 @opindex fpcc-struct-return
17860 Return ``short'' @code{struct} and @code{union} values in memory like
17861 longer ones, rather than in registers. This convention is less
17862 efficient, but it has the advantage of allowing intercallability between
17863 GCC-compiled files and files compiled with other compilers, particularly
17864 the Portable C Compiler (pcc).
17866 The precise convention for returning structures in memory depends
17867 on the target configuration macros.
17869 Short structures and unions are those whose size and alignment match
17870 that of some integer type.
17872 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17873 switch is not binary compatible with code compiled with the
17874 @option{-freg-struct-return} switch.
17875 Use it to conform to a non-default application binary interface.
17877 @item -freg-struct-return
17878 @opindex freg-struct-return
17879 Return @code{struct} and @code{union} values in registers when possible.
17880 This is more efficient for small structures than
17881 @option{-fpcc-struct-return}.
17883 If you specify neither @option{-fpcc-struct-return} nor
17884 @option{-freg-struct-return}, GCC defaults to whichever convention is
17885 standard for the target. If there is no standard convention, GCC
17886 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
17887 the principal compiler. In those cases, we can choose the standard, and
17888 we chose the more efficient register return alternative.
17890 @strong{Warning:} code compiled with the @option{-freg-struct-return}
17891 switch is not binary compatible with code compiled with the
17892 @option{-fpcc-struct-return} switch.
17893 Use it to conform to a non-default application binary interface.
17895 @item -fshort-enums
17896 @opindex fshort-enums
17897 Allocate to an @code{enum} type only as many bytes as it needs for the
17898 declared range of possible values. Specifically, the @code{enum} type
17899 will be equivalent to the smallest integer type which has enough room.
17901 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
17902 code that is not binary compatible with code generated without that switch.
17903 Use it to conform to a non-default application binary interface.
17905 @item -fshort-double
17906 @opindex fshort-double
17907 Use the same size for @code{double} as for @code{float}.
17909 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
17910 code that is not binary compatible with code generated without that switch.
17911 Use it to conform to a non-default application binary interface.
17913 @item -fshort-wchar
17914 @opindex fshort-wchar
17915 Override the underlying type for @samp{wchar_t} to be @samp{short
17916 unsigned int} instead of the default for the target. This option is
17917 useful for building programs to run under WINE@.
17919 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
17920 code that is not binary compatible with code generated without that switch.
17921 Use it to conform to a non-default application binary interface.
17924 @opindex fno-common
17925 In C code, controls the placement of uninitialized global variables.
17926 Unix C compilers have traditionally permitted multiple definitions of
17927 such variables in different compilation units by placing the variables
17929 This is the behavior specified by @option{-fcommon}, and is the default
17930 for GCC on most targets.
17931 On the other hand, this behavior is not required by ISO C, and on some
17932 targets may carry a speed or code size penalty on variable references.
17933 The @option{-fno-common} option specifies that the compiler should place
17934 uninitialized global variables in the data section of the object file,
17935 rather than generating them as common blocks.
17936 This has the effect that if the same variable is declared
17937 (without @code{extern}) in two different compilations,
17938 you will get a multiple-definition error when you link them.
17939 In this case, you must compile with @option{-fcommon} instead.
17940 Compiling with @option{-fno-common} is useful on targets for which
17941 it provides better performance, or if you wish to verify that the
17942 program will work on other systems which always treat uninitialized
17943 variable declarations this way.
17947 Ignore the @samp{#ident} directive.
17949 @item -finhibit-size-directive
17950 @opindex finhibit-size-directive
17951 Don't output a @code{.size} assembler directive, or anything else that
17952 would cause trouble if the function is split in the middle, and the
17953 two halves are placed at locations far apart in memory. This option is
17954 used when compiling @file{crtstuff.c}; you should not need to use it
17957 @item -fverbose-asm
17958 @opindex fverbose-asm
17959 Put extra commentary information in the generated assembly code to
17960 make it more readable. This option is generally only of use to those
17961 who actually need to read the generated assembly code (perhaps while
17962 debugging the compiler itself).
17964 @option{-fno-verbose-asm}, the default, causes the
17965 extra information to be omitted and is useful when comparing two assembler
17968 @item -frecord-gcc-switches
17969 @opindex frecord-gcc-switches
17970 This switch causes the command line that was used to invoke the
17971 compiler to be recorded into the object file that is being created.
17972 This switch is only implemented on some targets and the exact format
17973 of the recording is target and binary file format dependent, but it
17974 usually takes the form of a section containing ASCII text. This
17975 switch is related to the @option{-fverbose-asm} switch, but that
17976 switch only records information in the assembler output file as
17977 comments, so it never reaches the object file.
17981 @cindex global offset table
17983 Generate position-independent code (PIC) suitable for use in a shared
17984 library, if supported for the target machine. Such code accesses all
17985 constant addresses through a global offset table (GOT)@. The dynamic
17986 loader resolves the GOT entries when the program starts (the dynamic
17987 loader is not part of GCC; it is part of the operating system). If
17988 the GOT size for the linked executable exceeds a machine-specific
17989 maximum size, you get an error message from the linker indicating that
17990 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
17991 instead. (These maximums are 8k on the SPARC and 32k
17992 on the m68k and RS/6000. The 386 has no such limit.)
17994 Position-independent code requires special support, and therefore works
17995 only on certain machines. For the 386, GCC supports PIC for System V
17996 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
17997 position-independent.
17999 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18004 If supported for the target machine, emit position-independent code,
18005 suitable for dynamic linking and avoiding any limit on the size of the
18006 global offset table. This option makes a difference on the m68k,
18007 PowerPC and SPARC@.
18009 Position-independent code requires special support, and therefore works
18010 only on certain machines.
18012 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18019 These options are similar to @option{-fpic} and @option{-fPIC}, but
18020 generated position independent code can be only linked into executables.
18021 Usually these options are used when @option{-pie} GCC option will be
18022 used during linking.
18024 @option{-fpie} and @option{-fPIE} both define the macros
18025 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18026 for @option{-fpie} and 2 for @option{-fPIE}.
18028 @item -fno-jump-tables
18029 @opindex fno-jump-tables
18030 Do not use jump tables for switch statements even where it would be
18031 more efficient than other code generation strategies. This option is
18032 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18033 building code which forms part of a dynamic linker and cannot
18034 reference the address of a jump table. On some targets, jump tables
18035 do not require a GOT and this option is not needed.
18037 @item -ffixed-@var{reg}
18039 Treat the register named @var{reg} as a fixed register; generated code
18040 should never refer to it (except perhaps as a stack pointer, frame
18041 pointer or in some other fixed role).
18043 @var{reg} must be the name of a register. The register names accepted
18044 are machine-specific and are defined in the @code{REGISTER_NAMES}
18045 macro in the machine description macro file.
18047 This flag does not have a negative form, because it specifies a
18050 @item -fcall-used-@var{reg}
18051 @opindex fcall-used
18052 Treat the register named @var{reg} as an allocable register that is
18053 clobbered by function calls. It may be allocated for temporaries or
18054 variables that do not live across a call. Functions compiled this way
18055 will not save and restore the register @var{reg}.
18057 It is an error to used this flag with the frame pointer or stack pointer.
18058 Use of this flag for other registers that have fixed pervasive roles in
18059 the machine's execution model will produce disastrous results.
18061 This flag does not have a negative form, because it specifies a
18064 @item -fcall-saved-@var{reg}
18065 @opindex fcall-saved
18066 Treat the register named @var{reg} as an allocable register saved by
18067 functions. It may be allocated even for temporaries or variables that
18068 live across a call. Functions compiled this way will save and restore
18069 the register @var{reg} if they use it.
18071 It is an error to used this flag with the frame pointer or stack pointer.
18072 Use of this flag for other registers that have fixed pervasive roles in
18073 the machine's execution model will produce disastrous results.
18075 A different sort of disaster will result from the use of this flag for
18076 a register in which function values may be returned.
18078 This flag does not have a negative form, because it specifies a
18081 @item -fpack-struct[=@var{n}]
18082 @opindex fpack-struct
18083 Without a value specified, pack all structure members together without
18084 holes. When a value is specified (which must be a small power of two), pack
18085 structure members according to this value, representing the maximum
18086 alignment (that is, objects with default alignment requirements larger than
18087 this will be output potentially unaligned at the next fitting location.
18089 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18090 code that is not binary compatible with code generated without that switch.
18091 Additionally, it makes the code suboptimal.
18092 Use it to conform to a non-default application binary interface.
18094 @item -finstrument-functions
18095 @opindex finstrument-functions
18096 Generate instrumentation calls for entry and exit to functions. Just
18097 after function entry and just before function exit, the following
18098 profiling functions will be called with the address of the current
18099 function and its call site. (On some platforms,
18100 @code{__builtin_return_address} does not work beyond the current
18101 function, so the call site information may not be available to the
18102 profiling functions otherwise.)
18105 void __cyg_profile_func_enter (void *this_fn,
18107 void __cyg_profile_func_exit (void *this_fn,
18111 The first argument is the address of the start of the current function,
18112 which may be looked up exactly in the symbol table.
18114 This instrumentation is also done for functions expanded inline in other
18115 functions. The profiling calls will indicate where, conceptually, the
18116 inline function is entered and exited. This means that addressable
18117 versions of such functions must be available. If all your uses of a
18118 function are expanded inline, this may mean an additional expansion of
18119 code size. If you use @samp{extern inline} in your C code, an
18120 addressable version of such functions must be provided. (This is
18121 normally the case anyways, but if you get lucky and the optimizer always
18122 expands the functions inline, you might have gotten away without
18123 providing static copies.)
18125 A function may be given the attribute @code{no_instrument_function}, in
18126 which case this instrumentation will not be done. This can be used, for
18127 example, for the profiling functions listed above, high-priority
18128 interrupt routines, and any functions from which the profiling functions
18129 cannot safely be called (perhaps signal handlers, if the profiling
18130 routines generate output or allocate memory).
18132 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18133 @opindex finstrument-functions-exclude-file-list
18135 Set the list of functions that are excluded from instrumentation (see
18136 the description of @code{-finstrument-functions}). If the file that
18137 contains a function definition matches with one of @var{file}, then
18138 that function is not instrumented. The match is done on substrings:
18139 if the @var{file} parameter is a substring of the file name, it is
18140 considered to be a match.
18145 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18149 will exclude any inline function defined in files whose pathnames
18150 contain @code{/bits/stl} or @code{include/sys}.
18152 If, for some reason, you want to include letter @code{','} in one of
18153 @var{sym}, write @code{'\,'}. For example,
18154 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18155 (note the single quote surrounding the option).
18157 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18158 @opindex finstrument-functions-exclude-function-list
18160 This is similar to @code{-finstrument-functions-exclude-file-list},
18161 but this option sets the list of function names to be excluded from
18162 instrumentation. The function name to be matched is its user-visible
18163 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18164 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18165 match is done on substrings: if the @var{sym} parameter is a substring
18166 of the function name, it is considered to be a match. For C99 and C++
18167 extended identifiers, the function name must be given in UTF-8, not
18168 using universal character names.
18170 @item -fstack-check
18171 @opindex fstack-check
18172 Generate code to verify that you do not go beyond the boundary of the
18173 stack. You should specify this flag if you are running in an
18174 environment with multiple threads, but only rarely need to specify it in
18175 a single-threaded environment since stack overflow is automatically
18176 detected on nearly all systems if there is only one stack.
18178 Note that this switch does not actually cause checking to be done; the
18179 operating system or the language runtime must do that. The switch causes
18180 generation of code to ensure that they see the stack being extended.
18182 You can additionally specify a string parameter: @code{no} means no
18183 checking, @code{generic} means force the use of old-style checking,
18184 @code{specific} means use the best checking method and is equivalent
18185 to bare @option{-fstack-check}.
18187 Old-style checking is a generic mechanism that requires no specific
18188 target support in the compiler but comes with the following drawbacks:
18192 Modified allocation strategy for large objects: they will always be
18193 allocated dynamically if their size exceeds a fixed threshold.
18196 Fixed limit on the size of the static frame of functions: when it is
18197 topped by a particular function, stack checking is not reliable and
18198 a warning is issued by the compiler.
18201 Inefficiency: because of both the modified allocation strategy and the
18202 generic implementation, the performances of the code are hampered.
18205 Note that old-style stack checking is also the fallback method for
18206 @code{specific} if no target support has been added in the compiler.
18208 @item -fstack-limit-register=@var{reg}
18209 @itemx -fstack-limit-symbol=@var{sym}
18210 @itemx -fno-stack-limit
18211 @opindex fstack-limit-register
18212 @opindex fstack-limit-symbol
18213 @opindex fno-stack-limit
18214 Generate code to ensure that the stack does not grow beyond a certain value,
18215 either the value of a register or the address of a symbol. If the stack
18216 would grow beyond the value, a signal is raised. For most targets,
18217 the signal is raised before the stack overruns the boundary, so
18218 it is possible to catch the signal without taking special precautions.
18220 For instance, if the stack starts at absolute address @samp{0x80000000}
18221 and grows downwards, you can use the flags
18222 @option{-fstack-limit-symbol=__stack_limit} and
18223 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18224 of 128KB@. Note that this may only work with the GNU linker.
18226 @item -fsplit-stack
18227 @opindex fsplit-stack
18228 Generate code to automatically split the stack before it overflows.
18229 The resulting program has a discontiguous stack which can only
18230 overflow if the program is unable to allocate any more memory. This
18231 is most useful when running threaded programs, as it is no longer
18232 necessary to calculate a good stack size to use for each thread. This
18233 is currently only implemented for the i386 and x86_64 backends running
18236 When code compiled with @option{-fsplit-stack} calls code compiled
18237 without @option{-fsplit-stack}, there may not be much stack space
18238 available for the latter code to run. If compiling all code,
18239 including library code, with @option{-fsplit-stack} is not an option,
18240 then the linker can fix up these calls so that the code compiled
18241 without @option{-fsplit-stack} always has a large stack. Support for
18242 this is implemented in the gold linker in GNU binutils release 2.21
18245 @item -fleading-underscore
18246 @opindex fleading-underscore
18247 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18248 change the way C symbols are represented in the object file. One use
18249 is to help link with legacy assembly code.
18251 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18252 generate code that is not binary compatible with code generated without that
18253 switch. Use it to conform to a non-default application binary interface.
18254 Not all targets provide complete support for this switch.
18256 @item -ftls-model=@var{model}
18257 @opindex ftls-model
18258 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18259 The @var{model} argument should be one of @code{global-dynamic},
18260 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18262 The default without @option{-fpic} is @code{initial-exec}; with
18263 @option{-fpic} the default is @code{global-dynamic}.
18265 @item -fvisibility=@var{default|internal|hidden|protected}
18266 @opindex fvisibility
18267 Set the default ELF image symbol visibility to the specified option---all
18268 symbols will be marked with this unless overridden within the code.
18269 Using this feature can very substantially improve linking and
18270 load times of shared object libraries, produce more optimized
18271 code, provide near-perfect API export and prevent symbol clashes.
18272 It is @strong{strongly} recommended that you use this in any shared objects
18275 Despite the nomenclature, @code{default} always means public; i.e.,
18276 available to be linked against from outside the shared object.
18277 @code{protected} and @code{internal} are pretty useless in real-world
18278 usage so the only other commonly used option will be @code{hidden}.
18279 The default if @option{-fvisibility} isn't specified is
18280 @code{default}, i.e., make every
18281 symbol public---this causes the same behavior as previous versions of
18284 A good explanation of the benefits offered by ensuring ELF
18285 symbols have the correct visibility is given by ``How To Write
18286 Shared Libraries'' by Ulrich Drepper (which can be found at
18287 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18288 solution made possible by this option to marking things hidden when
18289 the default is public is to make the default hidden and mark things
18290 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18291 and @code{__attribute__ ((visibility("default")))} instead of
18292 @code{__declspec(dllexport)} you get almost identical semantics with
18293 identical syntax. This is a great boon to those working with
18294 cross-platform projects.
18296 For those adding visibility support to existing code, you may find
18297 @samp{#pragma GCC visibility} of use. This works by you enclosing
18298 the declarations you wish to set visibility for with (for example)
18299 @samp{#pragma GCC visibility push(hidden)} and
18300 @samp{#pragma GCC visibility pop}.
18301 Bear in mind that symbol visibility should be viewed @strong{as
18302 part of the API interface contract} and thus all new code should
18303 always specify visibility when it is not the default; i.e., declarations
18304 only for use within the local DSO should @strong{always} be marked explicitly
18305 as hidden as so to avoid PLT indirection overheads---making this
18306 abundantly clear also aids readability and self-documentation of the code.
18307 Note that due to ISO C++ specification requirements, operator new and
18308 operator delete must always be of default visibility.
18310 Be aware that headers from outside your project, in particular system
18311 headers and headers from any other library you use, may not be
18312 expecting to be compiled with visibility other than the default. You
18313 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18314 before including any such headers.
18316 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18317 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18318 no modifications. However, this means that calls to @samp{extern}
18319 functions with no explicit visibility will use the PLT, so it is more
18320 effective to use @samp{__attribute ((visibility))} and/or
18321 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18322 declarations should be treated as hidden.
18324 Note that @samp{-fvisibility} does affect C++ vague linkage
18325 entities. This means that, for instance, an exception class that will
18326 be thrown between DSOs must be explicitly marked with default
18327 visibility so that the @samp{type_info} nodes will be unified between
18330 An overview of these techniques, their benefits and how to use them
18331 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18333 @item -fstrict-volatile-bitfields
18334 @opindex fstrict-volatile-bitfields
18335 This option should be used if accesses to volatile bitfields (or other
18336 structure fields, although the compiler usually honors those types
18337 anyway) should use a single access of the width of the
18338 field's type, aligned to a natural alignment if possible. For
18339 example, targets with memory-mapped peripheral registers might require
18340 all such accesses to be 16 bits wide; with this flag the user could
18341 declare all peripheral bitfields as ``unsigned short'' (assuming short
18342 is 16 bits on these targets) to force GCC to use 16 bit accesses
18343 instead of, perhaps, a more efficient 32 bit access.
18345 If this option is disabled, the compiler will use the most efficient
18346 instruction. In the previous example, that might be a 32-bit load
18347 instruction, even though that will access bytes that do not contain
18348 any portion of the bitfield, or memory-mapped registers unrelated to
18349 the one being updated.
18351 If the target requires strict alignment, and honoring the field
18352 type would require violating this alignment, a warning is issued.
18353 If the field has @code{packed} attribute, the access is done without
18354 honoring the field type. If the field doesn't have @code{packed}
18355 attribute, the access is done honoring the field type. In both cases,
18356 GCC assumes that the user knows something about the target hardware
18357 that it is unaware of.
18359 The default value of this option is determined by the application binary
18360 interface for the target processor.
18366 @node Environment Variables
18367 @section Environment Variables Affecting GCC
18368 @cindex environment variables
18370 @c man begin ENVIRONMENT
18371 This section describes several environment variables that affect how GCC
18372 operates. Some of them work by specifying directories or prefixes to use
18373 when searching for various kinds of files. Some are used to specify other
18374 aspects of the compilation environment.
18376 Note that you can also specify places to search using options such as
18377 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18378 take precedence over places specified using environment variables, which
18379 in turn take precedence over those specified by the configuration of GCC@.
18380 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18381 GNU Compiler Collection (GCC) Internals}.
18386 @c @itemx LC_COLLATE
18388 @c @itemx LC_MONETARY
18389 @c @itemx LC_NUMERIC
18394 @c @findex LC_COLLATE
18395 @findex LC_MESSAGES
18396 @c @findex LC_MONETARY
18397 @c @findex LC_NUMERIC
18401 These environment variables control the way that GCC uses
18402 localization information that allow GCC to work with different
18403 national conventions. GCC inspects the locale categories
18404 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18405 so. These locale categories can be set to any value supported by your
18406 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18407 Kingdom encoded in UTF-8.
18409 The @env{LC_CTYPE} environment variable specifies character
18410 classification. GCC uses it to determine the character boundaries in
18411 a string; this is needed for some multibyte encodings that contain quote
18412 and escape characters that would otherwise be interpreted as a string
18415 The @env{LC_MESSAGES} environment variable specifies the language to
18416 use in diagnostic messages.
18418 If the @env{LC_ALL} environment variable is set, it overrides the value
18419 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18420 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18421 environment variable. If none of these variables are set, GCC
18422 defaults to traditional C English behavior.
18426 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18427 files. GCC uses temporary files to hold the output of one stage of
18428 compilation which is to be used as input to the next stage: for example,
18429 the output of the preprocessor, which is the input to the compiler
18432 @item GCC_COMPARE_DEBUG
18433 @findex GCC_COMPARE_DEBUG
18434 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
18435 @option{-fcompare-debug} to the compiler driver. See the documentation
18436 of this option for more details.
18438 @item GCC_EXEC_PREFIX
18439 @findex GCC_EXEC_PREFIX
18440 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18441 names of the subprograms executed by the compiler. No slash is added
18442 when this prefix is combined with the name of a subprogram, but you can
18443 specify a prefix that ends with a slash if you wish.
18445 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18446 an appropriate prefix to use based on the pathname it was invoked with.
18448 If GCC cannot find the subprogram using the specified prefix, it
18449 tries looking in the usual places for the subprogram.
18451 The default value of @env{GCC_EXEC_PREFIX} is
18452 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18453 the installed compiler. In many cases @var{prefix} is the value
18454 of @code{prefix} when you ran the @file{configure} script.
18456 Other prefixes specified with @option{-B} take precedence over this prefix.
18458 This prefix is also used for finding files such as @file{crt0.o} that are
18461 In addition, the prefix is used in an unusual way in finding the
18462 directories to search for header files. For each of the standard
18463 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18464 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18465 replacing that beginning with the specified prefix to produce an
18466 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18467 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18468 These alternate directories are searched first; the standard directories
18469 come next. If a standard directory begins with the configured
18470 @var{prefix} then the value of @var{prefix} is replaced by
18471 @env{GCC_EXEC_PREFIX} when looking for header files.
18473 @item COMPILER_PATH
18474 @findex COMPILER_PATH
18475 The value of @env{COMPILER_PATH} is a colon-separated list of
18476 directories, much like @env{PATH}. GCC tries the directories thus
18477 specified when searching for subprograms, if it can't find the
18478 subprograms using @env{GCC_EXEC_PREFIX}.
18481 @findex LIBRARY_PATH
18482 The value of @env{LIBRARY_PATH} is a colon-separated list of
18483 directories, much like @env{PATH}. When configured as a native compiler,
18484 GCC tries the directories thus specified when searching for special
18485 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18486 using GCC also uses these directories when searching for ordinary
18487 libraries for the @option{-l} option (but directories specified with
18488 @option{-L} come first).
18492 @cindex locale definition
18493 This variable is used to pass locale information to the compiler. One way in
18494 which this information is used is to determine the character set to be used
18495 when character literals, string literals and comments are parsed in C and C++.
18496 When the compiler is configured to allow multibyte characters,
18497 the following values for @env{LANG} are recognized:
18501 Recognize JIS characters.
18503 Recognize SJIS characters.
18505 Recognize EUCJP characters.
18508 If @env{LANG} is not defined, or if it has some other value, then the
18509 compiler will use mblen and mbtowc as defined by the default locale to
18510 recognize and translate multibyte characters.
18514 Some additional environments variables affect the behavior of the
18517 @include cppenv.texi
18521 @node Precompiled Headers
18522 @section Using Precompiled Headers
18523 @cindex precompiled headers
18524 @cindex speed of compilation
18526 Often large projects have many header files that are included in every
18527 source file. The time the compiler takes to process these header files
18528 over and over again can account for nearly all of the time required to
18529 build the project. To make builds faster, GCC allows users to
18530 `precompile' a header file; then, if builds can use the precompiled
18531 header file they will be much faster.
18533 To create a precompiled header file, simply compile it as you would any
18534 other file, if necessary using the @option{-x} option to make the driver
18535 treat it as a C or C++ header file. You will probably want to use a
18536 tool like @command{make} to keep the precompiled header up-to-date when
18537 the headers it contains change.
18539 A precompiled header file will be searched for when @code{#include} is
18540 seen in the compilation. As it searches for the included file
18541 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18542 compiler looks for a precompiled header in each directory just before it
18543 looks for the include file in that directory. The name searched for is
18544 the name specified in the @code{#include} with @samp{.gch} appended. If
18545 the precompiled header file can't be used, it is ignored.
18547 For instance, if you have @code{#include "all.h"}, and you have
18548 @file{all.h.gch} in the same directory as @file{all.h}, then the
18549 precompiled header file will be used if possible, and the original
18550 header will be used otherwise.
18552 Alternatively, you might decide to put the precompiled header file in a
18553 directory and use @option{-I} to ensure that directory is searched
18554 before (or instead of) the directory containing the original header.
18555 Then, if you want to check that the precompiled header file is always
18556 used, you can put a file of the same name as the original header in this
18557 directory containing an @code{#error} command.
18559 This also works with @option{-include}. So yet another way to use
18560 precompiled headers, good for projects not designed with precompiled
18561 header files in mind, is to simply take most of the header files used by
18562 a project, include them from another header file, precompile that header
18563 file, and @option{-include} the precompiled header. If the header files
18564 have guards against multiple inclusion, they will be skipped because
18565 they've already been included (in the precompiled header).
18567 If you need to precompile the same header file for different
18568 languages, targets, or compiler options, you can instead make a
18569 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18570 header in the directory, perhaps using @option{-o}. It doesn't matter
18571 what you call the files in the directory, every precompiled header in
18572 the directory will be considered. The first precompiled header
18573 encountered in the directory that is valid for this compilation will
18574 be used; they're searched in no particular order.
18576 There are many other possibilities, limited only by your imagination,
18577 good sense, and the constraints of your build system.
18579 A precompiled header file can be used only when these conditions apply:
18583 Only one precompiled header can be used in a particular compilation.
18586 A precompiled header can't be used once the first C token is seen. You
18587 can have preprocessor directives before a precompiled header; you can
18588 even include a precompiled header from inside another header, so long as
18589 there are no C tokens before the @code{#include}.
18592 The precompiled header file must be produced for the same language as
18593 the current compilation. You can't use a C precompiled header for a C++
18597 The precompiled header file must have been produced by the same compiler
18598 binary as the current compilation is using.
18601 Any macros defined before the precompiled header is included must
18602 either be defined in the same way as when the precompiled header was
18603 generated, or must not affect the precompiled header, which usually
18604 means that they don't appear in the precompiled header at all.
18606 The @option{-D} option is one way to define a macro before a
18607 precompiled header is included; using a @code{#define} can also do it.
18608 There are also some options that define macros implicitly, like
18609 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18612 @item If debugging information is output when using the precompiled
18613 header, using @option{-g} or similar, the same kind of debugging information
18614 must have been output when building the precompiled header. However,
18615 a precompiled header built using @option{-g} can be used in a compilation
18616 when no debugging information is being output.
18618 @item The same @option{-m} options must generally be used when building
18619 and using the precompiled header. @xref{Submodel Options},
18620 for any cases where this rule is relaxed.
18622 @item Each of the following options must be the same when building and using
18623 the precompiled header:
18625 @gccoptlist{-fexceptions}
18628 Some other command-line options starting with @option{-f},
18629 @option{-p}, or @option{-O} must be defined in the same way as when
18630 the precompiled header was generated. At present, it's not clear
18631 which options are safe to change and which are not; the safest choice
18632 is to use exactly the same options when generating and using the
18633 precompiled header. The following are known to be safe:
18635 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18636 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18637 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18642 For all of these except the last, the compiler will automatically
18643 ignore the precompiled header if the conditions aren't met. If you
18644 find an option combination that doesn't work and doesn't cause the
18645 precompiled header to be ignored, please consider filing a bug report,
18648 If you do use differing options when generating and using the
18649 precompiled header, the actual behavior will be a mixture of the
18650 behavior for the options. For instance, if you use @option{-g} to
18651 generate the precompiled header but not when using it, you may or may
18652 not get debugging information for routines in the precompiled header.