1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
15 Free Software Foundation, Inc.
17 Permission is granted to copy, distribute and/or modify this document
18 under the terms of the GNU Free Documentation License, Version 1.3 or
19 any later version published by the Free Software Foundation; with the
20 Invariant Sections being ``GNU General Public License'' and ``Funding
21 Free Software'', the Front-Cover texts being (a) (see below), and with
22 the Back-Cover Texts being (b) (see below). A copy of the license is
23 included in the gfdl(7) man page.
25 (a) The FSF's Front-Cover Text is:
29 (b) The FSF's Back-Cover Text is:
31 You have freedom to copy and modify this GNU Manual, like GNU
32 software. Copies published by the Free Software Foundation raise
33 funds for GNU development.
35 @c Set file name and title for the man page.
37 @settitle GNU project C and C++ compiler
39 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
40 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
41 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
42 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
43 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
44 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
45 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
47 Only the most useful options are listed here; see below for the
48 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
51 gpl(7), gfdl(7), fsf-funding(7),
52 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
53 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
54 @file{ld}, @file{binutils} and @file{gdb}.
57 For instructions on reporting bugs, see
61 See the Info entry for @command{gcc}, or
62 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
63 for contributors to GCC@.
68 @chapter GCC Command Options
69 @cindex GCC command options
70 @cindex command options
71 @cindex options, GCC command
73 @c man begin DESCRIPTION
74 When you invoke GCC, it normally does preprocessing, compilation,
75 assembly and linking. The ``overall options'' allow you to stop this
76 process at an intermediate stage. For example, the @option{-c} option
77 says not to run the linker. Then the output consists of object files
78 output by the assembler.
80 Other options are passed on to one stage of processing. Some options
81 control the preprocessor and others the compiler itself. Yet other
82 options control the assembler and linker; most of these are not
83 documented here, since you rarely need to use any of them.
85 @cindex C compilation options
86 Most of the command line options that you can use with GCC are useful
87 for C programs; when an option is only useful with another language
88 (usually C++), the explanation says so explicitly. If the description
89 for a particular option does not mention a source language, you can use
90 that option with all supported languages.
92 @cindex C++ compilation options
93 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
94 options for compiling C++ programs.
96 @cindex grouping options
97 @cindex options, grouping
98 The @command{gcc} program accepts options and file names as operands. Many
99 options have multi-letter names; therefore multiple single-letter options
100 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
103 @cindex order of options
104 @cindex options, order
105 You can mix options and other arguments. For the most part, the order
106 you use doesn't matter. Order does matter when you use several
107 options of the same kind; for example, if you specify @option{-L} more
108 than once, the directories are searched in the order specified. Also,
109 the placement of the @option{-l} option is significant.
111 Many options have long names starting with @samp{-f} or with
112 @samp{-W}---for example,
113 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
114 these have both positive and negative forms; the negative form of
115 @option{-ffoo} would be @option{-fno-foo}. This manual documents
116 only one of these two forms, whichever one is not the default.
120 @xref{Option Index}, for an index to GCC's options.
123 * Option Summary:: Brief list of all options, without explanations.
124 * Overall Options:: Controlling the kind of output:
125 an executable, object files, assembler files,
126 or preprocessed source.
127 * Invoking G++:: Compiling C++ programs.
128 * C Dialect Options:: Controlling the variant of C language compiled.
129 * C++ Dialect Options:: Variations on C++.
130 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
132 * Language Independent Options:: Controlling how diagnostics should be
134 * Warning Options:: How picky should the compiler be?
135 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
136 * Optimize Options:: How much optimization?
137 * Preprocessor Options:: Controlling header files and macro definitions.
138 Also, getting dependency information for Make.
139 * Assembler Options:: Passing options to the assembler.
140 * Link Options:: Specifying libraries and so on.
141 * Directory Options:: Where to find header files and libraries.
142 Where to find the compiler executable files.
143 * Spec Files:: How to pass switches to sub-processes.
144 * Target Options:: Running a cross-compiler, or an old version of GCC.
145 * Submodel Options:: Specifying minor hardware or convention variations,
146 such as 68010 vs 68020.
147 * Code Gen Options:: Specifying conventions for function calls, data layout
149 * Environment Variables:: Env vars that affect GCC.
150 * Precompiled Headers:: Compiling a header once, and using it many times.
156 @section Option Summary
158 Here is a summary of all the options, grouped by type. Explanations are
159 in the following sections.
162 @item Overall Options
163 @xref{Overall Options,,Options Controlling the Kind of Output}.
164 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
165 -pipe -pass-exit-codes @gol
166 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
167 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
168 -fdump-ada-spec@r{[}-slim@r{]} -fdump-go-spec=@var{file}}
170 @item C Language Options
171 @xref{C Dialect Options,,Options Controlling C Dialect}.
172 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
173 -aux-info @var{filename} @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} -Wno-free-nonheap-object -Wjump-misses-init @gol
248 -Wignored-qualifiers @gol
249 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
250 -Winit-self -Winline -Wmaybe-uninitialized @gol
251 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
252 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
253 -Wlogical-op -Wlong-long @gol
254 -Wmain -Wmaybe-uninitialized -Wmissing-braces -Wmissing-field-initializers @gol
255 -Wmissing-format-attribute -Wmissing-include-dirs @gol
257 -Wno-multichar -Wnonnull -Wno-overflow @gol
258 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
259 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
260 -Wpointer-arith -Wno-pointer-to-int-cast @gol
261 -Wredundant-decls @gol
262 -Wreturn-type -Wsequence-point -Wshadow @gol
263 -Wsign-compare -Wsign-conversion -Wstack-protector @gol
264 -Wstack-usage=@var{len} -Wstrict-aliasing -Wstrict-aliasing=n @gol
265 -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
266 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]} @gol
267 -Wswitch -Wswitch-default -Wswitch-enum -Wsync-nand @gol
268 -Wsystem-headers -Wtrampolines -Wtrigraphs -Wtype-limits -Wundef @gol
269 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
270 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
271 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
272 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
273 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
274 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
276 @item C and Objective-C-only Warning Options
277 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
278 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
279 -Wold-style-declaration -Wold-style-definition @gol
280 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
281 -Wdeclaration-after-statement -Wpointer-sign}
283 @item Debugging Options
284 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
285 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
286 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
287 -fdisable-ipa-@var{pass_name} @gol
288 -fdisable-rtl-@var{pass_name} @gol
289 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
290 -fdisable-tree-@var{pass_name} @gol
291 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
292 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
293 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
294 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
295 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
297 -fdump-statistics @gol
299 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
301 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
303 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
312 -fdump-tree-nrv -fdump-tree-vect @gol
313 -fdump-tree-sink @gol
314 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
317 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
318 -ftree-vectorizer-verbose=@var{n} @gol
319 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
320 -fdump-final-insns=@var{file} @gol
321 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
322 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
323 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
324 -fenable-@var{kind}-@var{pass} @gol
325 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
326 -fdebug-types-section @gol
327 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
328 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
329 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
330 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
331 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
332 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
333 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
334 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
335 -gvms -gxcoff -gxcoff+ @gol
336 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
337 -fdebug-prefix-map=@var{old}=@var{new} @gol
338 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
339 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
340 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
341 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
342 -print-prog-name=@var{program} -print-search-dirs -Q @gol
343 -print-sysroot -print-sysroot-headers-suffix @gol
344 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
346 @item Optimization Options
347 @xref{Optimize Options,,Options that Control Optimization}.
348 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
349 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
350 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
351 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
352 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
353 -fcompare-elim -fcprop-registers -fcrossjumping @gol
354 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
355 -fcx-limited-range @gol
356 -fdata-sections -fdce -fdce -fdelayed-branch @gol
357 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
358 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
359 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
360 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
361 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
362 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
363 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
364 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
365 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
366 -fira-algorithm=@var{algorithm} @gol
367 -fira-region=@var{region} @gol
368 -fira-loop-pressure -fno-ira-share-save-slots @gol
369 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
370 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
371 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
372 -floop-parallelize-all -flto -flto-compression-level @gol
373 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
374 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
375 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
376 -fno-default-inline @gol
377 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
378 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
379 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
380 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
381 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
382 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
383 -fprefetch-loop-arrays @gol
384 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
385 -fprofile-generate=@var{path} @gol
386 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
387 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
388 -freorder-blocks-and-partition -freorder-functions @gol
389 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
390 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
391 -fsched-spec-load -fsched-spec-load-dangerous @gol
392 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
393 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
394 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
395 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
396 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
397 -fselective-scheduling -fselective-scheduling2 @gol
398 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
399 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
400 -fsplit-ivs-in-unroller -fsplit-wide-types -fstack-protector @gol
401 -fstack-protector-all -fstrict-aliasing -fstrict-overflow @gol
402 -fthread-jumps -ftracer -ftree-bit-ccp @gol
403 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
404 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
405 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
406 -ftree-loop-if-convert-stores -ftree-loop-im @gol
407 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
408 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
409 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
410 -ftree-sink -ftree-sra -ftree-switch-conversion -ftree-tail-merge @gol
411 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
412 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
413 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
414 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
415 -fwhole-program -fwpa -fuse-linker-plugin @gol
416 --param @var{name}=@var{value}
417 -O -O0 -O1 -O2 -O3 -Os -Ofast}
419 @item Preprocessor Options
420 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
421 @gccoptlist{-A@var{question}=@var{answer} @gol
422 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
423 -C -dD -dI -dM -dN @gol
424 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
425 -idirafter @var{dir} @gol
426 -include @var{file} -imacros @var{file} @gol
427 -iprefix @var{file} -iwithprefix @var{dir} @gol
428 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
429 -imultilib @var{dir} -isysroot @var{dir} @gol
430 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
431 -P -fworking-directory -remap @gol
432 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
433 -Xpreprocessor @var{option}}
435 @item Assembler Option
436 @xref{Assembler Options,,Passing Options to the Assembler}.
437 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
440 @xref{Link Options,,Options for Linking}.
441 @gccoptlist{@var{object-file-name} -l@var{library} @gol
442 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
443 -s -static -static-libgcc -static-libstdc++ -shared @gol
444 -shared-libgcc -symbolic @gol
445 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
448 @item Directory Options
449 @xref{Directory Options,,Options for Directory Search}.
450 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
451 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
454 @item Machine Dependent Options
455 @xref{Submodel Options,,Hardware Models and Configurations}.
456 @c This list is ordered alphanumerically by subsection name.
457 @c Try and put the significant identifier (CPU or system) first,
458 @c so users have a clue at guessing where the ones they want will be.
461 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
462 -mabi=@var{name} @gol
463 -mapcs-stack-check -mno-apcs-stack-check @gol
464 -mapcs-float -mno-apcs-float @gol
465 -mapcs-reentrant -mno-apcs-reentrant @gol
466 -msched-prolog -mno-sched-prolog @gol
467 -mlittle-endian -mbig-endian -mwords-little-endian @gol
468 -mfloat-abi=@var{name} -mfpe @gol
469 -mfp16-format=@var{name}
470 -mthumb-interwork -mno-thumb-interwork @gol
471 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
472 -mstructure-size-boundary=@var{n} @gol
473 -mabort-on-noreturn @gol
474 -mlong-calls -mno-long-calls @gol
475 -msingle-pic-base -mno-single-pic-base @gol
476 -mpic-register=@var{reg} @gol
477 -mnop-fun-dllimport @gol
478 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
479 -mpoke-function-name @gol
481 -mtpcs-frame -mtpcs-leaf-frame @gol
482 -mcaller-super-interworking -mcallee-super-interworking @gol
483 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
484 -mword-relocations @gol
485 -mfix-cortex-m3-ldrd}
488 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
489 -mcall-prologues -mtiny-stack -mint8}
491 @emph{Blackfin Options}
492 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
493 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
494 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
495 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
496 -mno-id-shared-library -mshared-library-id=@var{n} @gol
497 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
498 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
499 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
503 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
504 -msim -msdata=@var{sdata-type}}
507 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
508 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
509 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
510 -mstack-align -mdata-align -mconst-align @gol
511 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
512 -melf -maout -melinux -mlinux -sim -sim2 @gol
513 -mmul-bug-workaround -mno-mul-bug-workaround}
515 @emph{Darwin Options}
516 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
517 -arch_only -bind_at_load -bundle -bundle_loader @gol
518 -client_name -compatibility_version -current_version @gol
520 -dependency-file -dylib_file -dylinker_install_name @gol
521 -dynamic -dynamiclib -exported_symbols_list @gol
522 -filelist -flat_namespace -force_cpusubtype_ALL @gol
523 -force_flat_namespace -headerpad_max_install_names @gol
525 -image_base -init -install_name -keep_private_externs @gol
526 -multi_module -multiply_defined -multiply_defined_unused @gol
527 -noall_load -no_dead_strip_inits_and_terms @gol
528 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
529 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
530 -private_bundle -read_only_relocs -sectalign @gol
531 -sectobjectsymbols -whyload -seg1addr @gol
532 -sectcreate -sectobjectsymbols -sectorder @gol
533 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
534 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
535 -segprot -segs_read_only_addr -segs_read_write_addr @gol
536 -single_module -static -sub_library -sub_umbrella @gol
537 -twolevel_namespace -umbrella -undefined @gol
538 -unexported_symbols_list -weak_reference_mismatches @gol
539 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
540 -mkernel -mone-byte-bool}
542 @emph{DEC Alpha Options}
543 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
544 -mieee -mieee-with-inexact -mieee-conformant @gol
545 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
546 -mtrap-precision=@var{mode} -mbuild-constants @gol
547 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
548 -mbwx -mmax -mfix -mcix @gol
549 -mfloat-vax -mfloat-ieee @gol
550 -mexplicit-relocs -msmall-data -mlarge-data @gol
551 -msmall-text -mlarge-text @gol
552 -mmemory-latency=@var{time}}
554 @emph{DEC Alpha/VMS Options}
555 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
558 @gccoptlist{-msmall-model -mno-lsim}
561 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
562 -mhard-float -msoft-float @gol
563 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
564 -mdouble -mno-double @gol
565 -mmedia -mno-media -mmuladd -mno-muladd @gol
566 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
567 -mlinked-fp -mlong-calls -malign-labels @gol
568 -mlibrary-pic -macc-4 -macc-8 @gol
569 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
570 -moptimize-membar -mno-optimize-membar @gol
571 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
572 -mvliw-branch -mno-vliw-branch @gol
573 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
574 -mno-nested-cond-exec -mtomcat-stats @gol
578 @emph{GNU/Linux Options}
579 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
580 -tno-android-cc -tno-android-ld}
582 @emph{H8/300 Options}
583 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
586 @gccoptlist{-march=@var{architecture-type} @gol
587 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
588 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
589 -mfixed-range=@var{register-range} @gol
590 -mjump-in-delay -mlinker-opt -mlong-calls @gol
591 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
592 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
593 -mno-jump-in-delay -mno-long-load-store @gol
594 -mno-portable-runtime -mno-soft-float @gol
595 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
596 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
597 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
598 -munix=@var{unix-std} -nolibdld -static -threads}
600 @emph{i386 and x86-64 Options}
601 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
602 -mfpmath=@var{unit} @gol
603 -masm=@var{dialect} -mno-fancy-math-387 @gol
604 -mno-fp-ret-in-387 -msoft-float @gol
605 -mno-wide-multiply -mrtd -malign-double @gol
606 -mpreferred-stack-boundary=@var{num} @gol
607 -mincoming-stack-boundary=@var{num} @gol
608 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
609 -mrecip -mrecip=@var{opt} @gol
611 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
612 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
613 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
614 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
615 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
616 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
617 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
618 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
619 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
620 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
621 -mcmodel=@var{code-model} -mabi=@var{name} @gol
622 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
623 -msse2avx -mfentry -m8bit-idiv @gol
624 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
626 @emph{i386 and x86-64 Windows Options}
627 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
628 -mnop-fun-dllimport -mthread @gol
629 -municode -mwin32 -mwindows -fno-set-stack-executable}
632 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
633 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
634 -mconstant-gp -mauto-pic -mfused-madd @gol
635 -minline-float-divide-min-latency @gol
636 -minline-float-divide-max-throughput @gol
637 -mno-inline-float-divide @gol
638 -minline-int-divide-min-latency @gol
639 -minline-int-divide-max-throughput @gol
640 -mno-inline-int-divide @gol
641 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
642 -mno-inline-sqrt @gol
643 -mdwarf2-asm -mearly-stop-bits @gol
644 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
645 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
646 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
647 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
648 -msched-spec-ldc -msched-spec-control-ldc @gol
649 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
650 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
651 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
652 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
654 @emph{IA-64/VMS Options}
655 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
658 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
659 -msign-extend-enabled -muser-enabled}
661 @emph{M32R/D Options}
662 @gccoptlist{-m32r2 -m32rx -m32r @gol
664 -malign-loops -mno-align-loops @gol
665 -missue-rate=@var{number} @gol
666 -mbranch-cost=@var{number} @gol
667 -mmodel=@var{code-size-model-type} @gol
668 -msdata=@var{sdata-type} @gol
669 -mno-flush-func -mflush-func=@var{name} @gol
670 -mno-flush-trap -mflush-trap=@var{number} @gol
674 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
676 @emph{M680x0 Options}
677 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
678 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
679 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
680 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
681 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
682 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
683 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
684 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
688 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
689 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
690 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
691 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
692 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
695 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
696 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
697 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
698 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
701 @emph{MicroBlaze Options}
702 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
703 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
704 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
705 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
706 -mxl-mode-@var{app-model}}
709 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
710 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
711 -mips64 -mips64r2 @gol
712 -mips16 -mno-mips16 -mflip-mips16 @gol
713 -minterlink-mips16 -mno-interlink-mips16 @gol
714 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
715 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
716 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
717 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
718 -mfpu=@var{fpu-type} @gol
719 -msmartmips -mno-smartmips @gol
720 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
721 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
722 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
723 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
724 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
725 -membedded-data -mno-embedded-data @gol
726 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
727 -mcode-readable=@var{setting} @gol
728 -msplit-addresses -mno-split-addresses @gol
729 -mexplicit-relocs -mno-explicit-relocs @gol
730 -mcheck-zero-division -mno-check-zero-division @gol
731 -mdivide-traps -mdivide-breaks @gol
732 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
733 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
734 -mfix-24k -mno-fix-24k @gol
735 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
736 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
737 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
738 -mflush-func=@var{func} -mno-flush-func @gol
739 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
740 -mfp-exceptions -mno-fp-exceptions @gol
741 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
742 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
745 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
746 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
747 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
748 -mno-base-addresses -msingle-exit -mno-single-exit}
750 @emph{MN10300 Options}
751 @gccoptlist{-mmult-bug -mno-mult-bug @gol
752 -mno-am33 -mam33 -mam33-2 -mam34 @gol
753 -mtune=@var{cpu-type} @gol
754 -mreturn-pointer-on-d0 @gol
755 -mno-crt0 -mrelax -mliw -msetlb}
757 @emph{PDP-11 Options}
758 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
759 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
760 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
761 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
762 -mbranch-expensive -mbranch-cheap @gol
763 -munix-asm -mdec-asm}
765 @emph{picoChip Options}
766 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
767 -msymbol-as-address -mno-inefficient-warnings}
769 @emph{PowerPC Options}
770 See RS/6000 and PowerPC Options.
772 @emph{RS/6000 and PowerPC Options}
773 @gccoptlist{-mcpu=@var{cpu-type} @gol
774 -mtune=@var{cpu-type} @gol
775 -mcmodel=@var{code-model} @gol
776 -mpower -mno-power -mpower2 -mno-power2 @gol
777 -mpowerpc -mpowerpc64 -mno-powerpc @gol
778 -maltivec -mno-altivec @gol
779 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
780 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
781 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
782 -mfprnd -mno-fprnd @gol
783 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
784 -mnew-mnemonics -mold-mnemonics @gol
785 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
786 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
787 -malign-power -malign-natural @gol
788 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
789 -msingle-float -mdouble-float -msimple-fpu @gol
790 -mstring -mno-string -mupdate -mno-update @gol
791 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
792 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
793 -mstrict-align -mno-strict-align -mrelocatable @gol
794 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
795 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
796 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
797 -mprioritize-restricted-insns=@var{priority} @gol
798 -msched-costly-dep=@var{dependence_type} @gol
799 -minsert-sched-nops=@var{scheme} @gol
800 -mcall-sysv -mcall-netbsd @gol
801 -maix-struct-return -msvr4-struct-return @gol
802 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
803 -mblock-move-inline-limit=@var{num} @gol
804 -misel -mno-isel @gol
805 -misel=yes -misel=no @gol
807 -mspe=yes -mspe=no @gol
809 -mgen-cell-microcode -mwarn-cell-microcode @gol
810 -mvrsave -mno-vrsave @gol
811 -mmulhw -mno-mulhw @gol
812 -mdlmzb -mno-dlmzb @gol
813 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
814 -mprototype -mno-prototype @gol
815 -msim -mmvme -mads -myellowknife -memb -msdata @gol
816 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
817 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
818 -mno-recip-precision @gol
819 -mveclibabi=@var{type} -mfriz -mno-friz @gol
820 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
821 -msave-toc-indirect -mno-save-toc-indirect}
824 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
826 -mbig-endian-data -mlittle-endian-data @gol
829 -mas100-syntax -mno-as100-syntax@gol
831 -mmax-constant-size=@gol
834 -msave-acc-in-interrupts}
836 @emph{S/390 and zSeries Options}
837 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
838 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
839 -mlong-double-64 -mlong-double-128 @gol
840 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
841 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
842 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
843 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
844 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
847 @gccoptlist{-meb -mel @gol
851 -mscore5 -mscore5u -mscore7 -mscore7d}
854 @gccoptlist{-m1 -m2 -m2e @gol
855 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
857 -m4-nofpu -m4-single-only -m4-single -m4 @gol
858 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
859 -m5-64media -m5-64media-nofpu @gol
860 -m5-32media -m5-32media-nofpu @gol
861 -m5-compact -m5-compact-nofpu @gol
862 -mb -ml -mdalign -mrelax @gol
863 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
864 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
865 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
866 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
867 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
868 -maccumulate-outgoing-args -minvalid-symbols}
870 @emph{Solaris 2 Options}
871 @gccoptlist{-mimpure-text -mno-impure-text @gol
875 @gccoptlist{-mcpu=@var{cpu-type} @gol
876 -mtune=@var{cpu-type} @gol
877 -mcmodel=@var{code-model} @gol
878 -m32 -m64 -mapp-regs -mno-app-regs @gol
879 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
880 -mfpu -mno-fpu -mhard-float -msoft-float @gol
881 -mhard-quad-float -msoft-quad-float @gol
883 -mstack-bias -mno-stack-bias @gol
884 -munaligned-doubles -mno-unaligned-doubles @gol
885 -mv8plus -mno-v8plus -mvis -mno-vis @gol
886 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
887 -mfmaf -mno-fmaf -mpopc -mno-popc}
890 @gccoptlist{-mwarn-reloc -merror-reloc @gol
891 -msafe-dma -munsafe-dma @gol
893 -msmall-mem -mlarge-mem -mstdmain @gol
894 -mfixed-range=@var{register-range} @gol
896 -maddress-space-conversion -mno-address-space-conversion @gol
897 -mcache-size=@var{cache-size} @gol
898 -matomic-updates -mno-atomic-updates}
900 @emph{System V Options}
901 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
904 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
905 -mprolog-function -mno-prolog-function -mspace @gol
906 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
907 -mapp-regs -mno-app-regs @gol
908 -mdisable-callt -mno-disable-callt @gol
911 -mv850e1 -mv850es @gol
916 @gccoptlist{-mg -mgnu -munix}
918 @emph{VxWorks Options}
919 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
920 -Xbind-lazy -Xbind-now}
922 @emph{x86-64 Options}
923 See i386 and x86-64 Options.
925 @emph{Xstormy16 Options}
928 @emph{Xtensa Options}
929 @gccoptlist{-mconst16 -mno-const16 @gol
930 -mfused-madd -mno-fused-madd @gol
932 -mserialize-volatile -mno-serialize-volatile @gol
933 -mtext-section-literals -mno-text-section-literals @gol
934 -mtarget-align -mno-target-align @gol
935 -mlongcalls -mno-longcalls}
937 @emph{zSeries Options}
938 See S/390 and zSeries Options.
940 @item Code Generation Options
941 @xref{Code Gen Options,,Options for Code Generation Conventions}.
942 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
943 -ffixed-@var{reg} -fexceptions @gol
944 -fnon-call-exceptions -funwind-tables @gol
945 -fasynchronous-unwind-tables @gol
946 -finhibit-size-directive -finstrument-functions @gol
947 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
948 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
949 -fno-common -fno-ident @gol
950 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
951 -fno-jump-tables @gol
952 -frecord-gcc-switches @gol
953 -freg-struct-return -fshort-enums @gol
954 -fshort-double -fshort-wchar @gol
955 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
956 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
957 -fno-stack-limit -fsplit-stack @gol
958 -fleading-underscore -ftls-model=@var{model} @gol
959 -ftrapv -fwrapv -fbounds-check @gol
960 -fvisibility -fstrict-volatile-bitfields}
964 * Overall Options:: Controlling the kind of output:
965 an executable, object files, assembler files,
966 or preprocessed source.
967 * C Dialect Options:: Controlling the variant of C language compiled.
968 * C++ Dialect Options:: Variations on C++.
969 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
971 * Language Independent Options:: Controlling how diagnostics should be
973 * Warning Options:: How picky should the compiler be?
974 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
975 * Optimize Options:: How much optimization?
976 * Preprocessor Options:: Controlling header files and macro definitions.
977 Also, getting dependency information for Make.
978 * Assembler Options:: Passing options to the assembler.
979 * Link Options:: Specifying libraries and so on.
980 * Directory Options:: Where to find header files and libraries.
981 Where to find the compiler executable files.
982 * Spec Files:: How to pass switches to sub-processes.
983 * Target Options:: Running a cross-compiler, or an old version of GCC.
986 @node Overall Options
987 @section Options Controlling the Kind of Output
989 Compilation can involve up to four stages: preprocessing, compilation
990 proper, assembly and linking, always in that order. GCC is capable of
991 preprocessing and compiling several files either into several
992 assembler input files, or into one assembler input file; then each
993 assembler input file produces an object file, and linking combines all
994 the object files (those newly compiled, and those specified as input)
995 into an executable file.
997 @cindex file name suffix
998 For any given input file, the file name suffix determines what kind of
1003 C source code which must be preprocessed.
1006 C source code which should not be preprocessed.
1009 C++ source code which should not be preprocessed.
1012 Objective-C source code. Note that you must link with the @file{libobjc}
1013 library to make an Objective-C program work.
1016 Objective-C source code which should not be preprocessed.
1020 Objective-C++ source code. Note that you must link with the @file{libobjc}
1021 library to make an Objective-C++ program work. Note that @samp{.M} refers
1022 to a literal capital M@.
1024 @item @var{file}.mii
1025 Objective-C++ source code which should not be preprocessed.
1028 C, C++, Objective-C or Objective-C++ header file to be turned into a
1029 precompiled header (default), or C, C++ header file to be turned into an
1030 Ada spec (via the @option{-fdump-ada-spec} switch).
1033 @itemx @var{file}.cp
1034 @itemx @var{file}.cxx
1035 @itemx @var{file}.cpp
1036 @itemx @var{file}.CPP
1037 @itemx @var{file}.c++
1039 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1040 the last two letters must both be literally @samp{x}. Likewise,
1041 @samp{.C} refers to a literal capital C@.
1045 Objective-C++ source code which must be preprocessed.
1047 @item @var{file}.mii
1048 Objective-C++ source code which should not be preprocessed.
1052 @itemx @var{file}.hp
1053 @itemx @var{file}.hxx
1054 @itemx @var{file}.hpp
1055 @itemx @var{file}.HPP
1056 @itemx @var{file}.h++
1057 @itemx @var{file}.tcc
1058 C++ header file to be turned into a precompiled header or Ada spec.
1061 @itemx @var{file}.for
1062 @itemx @var{file}.ftn
1063 Fixed form Fortran source code which should not be preprocessed.
1066 @itemx @var{file}.FOR
1067 @itemx @var{file}.fpp
1068 @itemx @var{file}.FPP
1069 @itemx @var{file}.FTN
1070 Fixed form Fortran source code which must be preprocessed (with the traditional
1073 @item @var{file}.f90
1074 @itemx @var{file}.f95
1075 @itemx @var{file}.f03
1076 @itemx @var{file}.f08
1077 Free form Fortran source code which should not be preprocessed.
1079 @item @var{file}.F90
1080 @itemx @var{file}.F95
1081 @itemx @var{file}.F03
1082 @itemx @var{file}.F08
1083 Free form Fortran source code which must be preprocessed (with the
1084 traditional preprocessor).
1089 @c FIXME: Descriptions of Java file types.
1095 @item @var{file}.ads
1096 Ada source code file which contains a library unit declaration (a
1097 declaration of a package, subprogram, or generic, or a generic
1098 instantiation), or a library unit renaming declaration (a package,
1099 generic, or subprogram renaming declaration). Such files are also
1102 @item @var{file}.adb
1103 Ada source code file containing a library unit body (a subprogram or
1104 package body). Such files are also called @dfn{bodies}.
1106 @c GCC also knows about some suffixes for languages not yet included:
1117 @itemx @var{file}.sx
1118 Assembler code which must be preprocessed.
1121 An object file to be fed straight into linking.
1122 Any file name with no recognized suffix is treated this way.
1126 You can specify the input language explicitly with the @option{-x} option:
1129 @item -x @var{language}
1130 Specify explicitly the @var{language} for the following input files
1131 (rather than letting the compiler choose a default based on the file
1132 name suffix). This option applies to all following input files until
1133 the next @option{-x} option. Possible values for @var{language} are:
1135 c c-header cpp-output
1136 c++ c++-header c++-cpp-output
1137 objective-c objective-c-header objective-c-cpp-output
1138 objective-c++ objective-c++-header objective-c++-cpp-output
1139 assembler assembler-with-cpp
1141 f77 f77-cpp-input f95 f95-cpp-input
1147 Turn off any specification of a language, so that subsequent files are
1148 handled according to their file name suffixes (as they are if @option{-x}
1149 has not been used at all).
1151 @item -pass-exit-codes
1152 @opindex pass-exit-codes
1153 Normally the @command{gcc} program will exit with the code of 1 if any
1154 phase of the compiler returns a non-success return code. If you specify
1155 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1156 numerically highest error produced by any phase that returned an error
1157 indication. The C, C++, and Fortran frontends return 4, if an internal
1158 compiler error is encountered.
1161 If you only want some of the stages of compilation, you can use
1162 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1163 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1164 @command{gcc} is to stop. Note that some combinations (for example,
1165 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1170 Compile or assemble the source files, but do not link. The linking
1171 stage simply is not done. The ultimate output is in the form of an
1172 object file for each source file.
1174 By default, the object file name for a source file is made by replacing
1175 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1177 Unrecognized input files, not requiring compilation or assembly, are
1182 Stop after the stage of compilation proper; do not assemble. The output
1183 is in the form of an assembler code file for each non-assembler input
1186 By default, the assembler file name for a source file is made by
1187 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1189 Input files that don't require compilation are ignored.
1193 Stop after the preprocessing stage; do not run the compiler proper. The
1194 output is in the form of preprocessed source code, which is sent to the
1197 Input files which don't require preprocessing are ignored.
1199 @cindex output file option
1202 Place output in file @var{file}. This applies regardless to whatever
1203 sort of output is being produced, whether it be an executable file,
1204 an object file, an assembler file or preprocessed C code.
1206 If @option{-o} is not specified, the default is to put an executable
1207 file in @file{a.out}, the object file for
1208 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1209 assembler file in @file{@var{source}.s}, a precompiled header file in
1210 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1215 Print (on standard error output) the commands executed to run the stages
1216 of compilation. Also print the version number of the compiler driver
1217 program and of the preprocessor and the compiler proper.
1221 Like @option{-v} except the commands are not executed and arguments
1222 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1223 This is useful for shell scripts to capture the driver-generated command lines.
1227 Use pipes rather than temporary files for communication between the
1228 various stages of compilation. This fails to work on some systems where
1229 the assembler is unable to read from a pipe; but the GNU assembler has
1234 Print (on the standard output) a description of the command line options
1235 understood by @command{gcc}. If the @option{-v} option is also specified
1236 then @option{--help} will also be passed on to the various processes
1237 invoked by @command{gcc}, so that they can display the command line options
1238 they accept. If the @option{-Wextra} option has also been specified
1239 (prior to the @option{--help} option), then command line options which
1240 have no documentation associated with them will also be displayed.
1243 @opindex target-help
1244 Print (on the standard output) a description of target-specific command
1245 line options for each tool. For some targets extra target-specific
1246 information may also be printed.
1248 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1249 Print (on the standard output) a description of the command line
1250 options understood by the compiler that fit into all specified classes
1251 and qualifiers. These are the supported classes:
1254 @item @samp{optimizers}
1255 This will display all of the optimization options supported by the
1258 @item @samp{warnings}
1259 This will display all of the options controlling warning messages
1260 produced by the compiler.
1263 This will display target-specific options. Unlike the
1264 @option{--target-help} option however, target-specific options of the
1265 linker and assembler will not be displayed. This is because those
1266 tools do not currently support the extended @option{--help=} syntax.
1269 This will display the values recognized by the @option{--param}
1272 @item @var{language}
1273 This will display the options supported for @var{language}, where
1274 @var{language} is the name of one of the languages supported in this
1278 This will display the options that are common to all languages.
1281 These are the supported qualifiers:
1284 @item @samp{undocumented}
1285 Display only those options which are undocumented.
1288 Display options which take an argument that appears after an equal
1289 sign in the same continuous piece of text, such as:
1290 @samp{--help=target}.
1292 @item @samp{separate}
1293 Display options which take an argument that appears as a separate word
1294 following the original option, such as: @samp{-o output-file}.
1297 Thus for example to display all the undocumented target-specific
1298 switches supported by the compiler the following can be used:
1301 --help=target,undocumented
1304 The sense of a qualifier can be inverted by prefixing it with the
1305 @samp{^} character, so for example to display all binary warning
1306 options (i.e., ones that are either on or off and that do not take an
1307 argument), which have a description the following can be used:
1310 --help=warnings,^joined,^undocumented
1313 The argument to @option{--help=} should not consist solely of inverted
1316 Combining several classes is possible, although this usually
1317 restricts the output by so much that there is nothing to display. One
1318 case where it does work however is when one of the classes is
1319 @var{target}. So for example to display all the target-specific
1320 optimization options the following can be used:
1323 --help=target,optimizers
1326 The @option{--help=} option can be repeated on the command line. Each
1327 successive use will display its requested class of options, skipping
1328 those that have already been displayed.
1330 If the @option{-Q} option appears on the command line before the
1331 @option{--help=} option, then the descriptive text displayed by
1332 @option{--help=} is changed. Instead of describing the displayed
1333 options, an indication is given as to whether the option is enabled,
1334 disabled or set to a specific value (assuming that the compiler
1335 knows this at the point where the @option{--help=} option is used).
1337 Here is a truncated example from the ARM port of @command{gcc}:
1340 % gcc -Q -mabi=2 --help=target -c
1341 The following options are target specific:
1343 -mabort-on-noreturn [disabled]
1347 The output is sensitive to the effects of previous command line
1348 options, so for example it is possible to find out which optimizations
1349 are enabled at @option{-O2} by using:
1352 -Q -O2 --help=optimizers
1355 Alternatively you can discover which binary optimizations are enabled
1356 by @option{-O3} by using:
1359 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1360 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1361 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1364 @item -no-canonical-prefixes
1365 @opindex no-canonical-prefixes
1366 Do not expand any symbolic links, resolve references to @samp{/../}
1367 or @samp{/./}, or make the path absolute when generating a relative
1372 Display the version number and copyrights of the invoked GCC@.
1376 Invoke all subcommands under a wrapper program. The name of the
1377 wrapper program and its parameters are passed as a comma separated
1381 gcc -c t.c -wrapper gdb,--args
1384 This will invoke all subprograms of @command{gcc} under
1385 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1386 @samp{gdb --args cc1 @dots{}}.
1388 @item -fplugin=@var{name}.so
1389 Load the plugin code in file @var{name}.so, assumed to be a
1390 shared object to be dlopen'd by the compiler. The base name of
1391 the shared object file is used to identify the plugin for the
1392 purposes of argument parsing (See
1393 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1394 Each plugin should define the callback functions specified in the
1397 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1398 Define an argument called @var{key} with a value of @var{value}
1399 for the plugin called @var{name}.
1401 @item -fdump-ada-spec@r{[}-slim@r{]}
1402 For C and C++ source and include files, generate corresponding Ada
1403 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1404 GNAT User's Guide}, which provides detailed documentation on this feature.
1406 @item -fdump-go-spec=@var{file}
1407 For input files in any language, generate corresponding Go
1408 declarations in @var{file}. This generates Go @code{const},
1409 @code{type}, @code{var}, and @code{func} declarations which may be a
1410 useful way to start writing a Go interface to code written in some
1413 @include @value{srcdir}/../libiberty/at-file.texi
1417 @section Compiling C++ Programs
1419 @cindex suffixes for C++ source
1420 @cindex C++ source file suffixes
1421 C++ source files conventionally use one of the suffixes @samp{.C},
1422 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1423 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1424 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1425 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1426 files with these names and compiles them as C++ programs even if you
1427 call the compiler the same way as for compiling C programs (usually
1428 with the name @command{gcc}).
1432 However, the use of @command{gcc} does not add the C++ library.
1433 @command{g++} is a program that calls GCC and treats @samp{.c},
1434 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1435 files unless @option{-x} is used, and automatically specifies linking
1436 against the C++ library. This program is also useful when
1437 precompiling a C header file with a @samp{.h} extension for use in C++
1438 compilations. On many systems, @command{g++} is also installed with
1439 the name @command{c++}.
1441 @cindex invoking @command{g++}
1442 When you compile C++ programs, you may specify many of the same
1443 command-line options that you use for compiling programs in any
1444 language; or command-line options meaningful for C and related
1445 languages; or options that are meaningful only for C++ programs.
1446 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1447 explanations of options for languages related to C@.
1448 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1449 explanations of options that are meaningful only for C++ programs.
1451 @node C Dialect Options
1452 @section Options Controlling C Dialect
1453 @cindex dialect options
1454 @cindex language dialect options
1455 @cindex options, dialect
1457 The following options control the dialect of C (or languages derived
1458 from C, such as C++, Objective-C and Objective-C++) that the compiler
1462 @cindex ANSI support
1466 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1467 equivalent to @samp{-std=c++98}.
1469 This turns off certain features of GCC that are incompatible with ISO
1470 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1471 such as the @code{asm} and @code{typeof} keywords, and
1472 predefined macros such as @code{unix} and @code{vax} that identify the
1473 type of system you are using. It also enables the undesirable and
1474 rarely used ISO trigraph feature. For the C compiler,
1475 it disables recognition of C++ style @samp{//} comments as well as
1476 the @code{inline} keyword.
1478 The alternate keywords @code{__asm__}, @code{__extension__},
1479 @code{__inline__} and @code{__typeof__} continue to work despite
1480 @option{-ansi}. You would not want to use them in an ISO C program, of
1481 course, but it is useful to put them in header files that might be included
1482 in compilations done with @option{-ansi}. Alternate predefined macros
1483 such as @code{__unix__} and @code{__vax__} are also available, with or
1484 without @option{-ansi}.
1486 The @option{-ansi} option does not cause non-ISO programs to be
1487 rejected gratuitously. For that, @option{-pedantic} is required in
1488 addition to @option{-ansi}. @xref{Warning Options}.
1490 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1491 option is used. Some header files may notice this macro and refrain
1492 from declaring certain functions or defining certain macros that the
1493 ISO standard doesn't call for; this is to avoid interfering with any
1494 programs that might use these names for other things.
1496 Functions that would normally be built in but do not have semantics
1497 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1498 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1499 built-in functions provided by GCC}, for details of the functions
1504 Determine the language standard. @xref{Standards,,Language Standards
1505 Supported by GCC}, for details of these standard versions. This option
1506 is currently only supported when compiling C or C++.
1508 The compiler can accept several base standards, such as @samp{c90} or
1509 @samp{c++98}, and GNU dialects of those standards, such as
1510 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1511 compiler will accept all programs following that standard and those
1512 using GNU extensions that do not contradict it. For example,
1513 @samp{-std=c90} turns off certain features of GCC that are
1514 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1515 keywords, but not other GNU extensions that do not have a meaning in
1516 ISO C90, such as omitting the middle term of a @code{?:}
1517 expression. On the other hand, by specifying a GNU dialect of a
1518 standard, all features the compiler support are enabled, even when
1519 those features change the meaning of the base standard and some
1520 strict-conforming programs may be rejected. The particular standard
1521 is used by @option{-pedantic} to identify which features are GNU
1522 extensions given that version of the standard. For example
1523 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1524 comments, while @samp{-std=gnu99 -pedantic} would not.
1526 A value for this option must be provided; possible values are
1532 Support all ISO C90 programs (certain GNU extensions that conflict
1533 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1535 @item iso9899:199409
1536 ISO C90 as modified in amendment 1.
1542 ISO C99. Note that this standard is not yet fully supported; see
1543 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1544 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1547 ISO C1X, the draft of the next revision of the ISO C standard.
1548 Support is limited and experimental and features enabled by this
1549 option may be changed or removed if changed in or removed from the
1554 GNU dialect of ISO C90 (including some C99 features). This
1555 is the default for C code.
1559 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1560 this will become the default. The name @samp{gnu9x} is deprecated.
1563 GNU dialect of ISO C1X. Support is limited and experimental and
1564 features enabled by this option may be changed or removed if changed
1565 in or removed from the standard draft.
1568 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1572 GNU dialect of @option{-std=c++98}. This is the default for
1576 The working draft of the upcoming ISO C++0x standard. This option
1577 enables experimental features that are likely to be included in
1578 C++0x. The working draft is constantly changing, and any feature that is
1579 enabled by this flag may be removed from future versions of GCC if it is
1580 not part of the C++0x standard.
1583 GNU dialect of @option{-std=c++0x}. This option enables
1584 experimental features that may be removed in future versions of GCC.
1587 @item -fgnu89-inline
1588 @opindex fgnu89-inline
1589 The option @option{-fgnu89-inline} tells GCC to use the traditional
1590 GNU semantics for @code{inline} functions when in C99 mode.
1591 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1592 is accepted and ignored by GCC versions 4.1.3 up to but not including
1593 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1594 C99 mode. Using this option is roughly equivalent to adding the
1595 @code{gnu_inline} function attribute to all inline functions
1596 (@pxref{Function Attributes}).
1598 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1599 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1600 specifies the default behavior). This option was first supported in
1601 GCC 4.3. This option is not supported in @option{-std=c90} or
1602 @option{-std=gnu90} mode.
1604 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1605 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1606 in effect for @code{inline} functions. @xref{Common Predefined
1607 Macros,,,cpp,The C Preprocessor}.
1609 @item -aux-info @var{filename}
1611 Output to the given filename prototyped declarations for all functions
1612 declared and/or defined in a translation unit, including those in header
1613 files. This option is silently ignored in any language other than C@.
1615 Besides declarations, the file indicates, in comments, the origin of
1616 each declaration (source file and line), whether the declaration was
1617 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1618 @samp{O} for old, respectively, in the first character after the line
1619 number and the colon), and whether it came from a declaration or a
1620 definition (@samp{C} or @samp{F}, respectively, in the following
1621 character). In the case of function definitions, a K&R-style list of
1622 arguments followed by their declarations is also provided, inside
1623 comments, after the declaration.
1627 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1628 keyword, so that code can use these words as identifiers. You can use
1629 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1630 instead. @option{-ansi} implies @option{-fno-asm}.
1632 In C++, this switch only affects the @code{typeof} keyword, since
1633 @code{asm} and @code{inline} are standard keywords. You may want to
1634 use the @option{-fno-gnu-keywords} flag instead, which has the same
1635 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1636 switch only affects the @code{asm} and @code{typeof} keywords, since
1637 @code{inline} is a standard keyword in ISO C99.
1640 @itemx -fno-builtin-@var{function}
1641 @opindex fno-builtin
1642 @cindex built-in functions
1643 Don't recognize built-in functions that do not begin with
1644 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1645 functions provided by GCC}, for details of the functions affected,
1646 including those which are not built-in functions when @option{-ansi} or
1647 @option{-std} options for strict ISO C conformance are used because they
1648 do not have an ISO standard meaning.
1650 GCC normally generates special code to handle certain built-in functions
1651 more efficiently; for instance, calls to @code{alloca} may become single
1652 instructions that adjust the stack directly, and calls to @code{memcpy}
1653 may become inline copy loops. The resulting code is often both smaller
1654 and faster, but since the function calls no longer appear as such, you
1655 cannot set a breakpoint on those calls, nor can you change the behavior
1656 of the functions by linking with a different library. In addition,
1657 when a function is recognized as a built-in function, GCC may use
1658 information about that function to warn about problems with calls to
1659 that function, or to generate more efficient code, even if the
1660 resulting code still contains calls to that function. For example,
1661 warnings are given with @option{-Wformat} for bad calls to
1662 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1663 known not to modify global memory.
1665 With the @option{-fno-builtin-@var{function}} option
1666 only the built-in function @var{function} is
1667 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1668 function is named that is not built-in in this version of GCC, this
1669 option is ignored. There is no corresponding
1670 @option{-fbuiltin-@var{function}} option; if you wish to enable
1671 built-in functions selectively when using @option{-fno-builtin} or
1672 @option{-ffreestanding}, you may define macros such as:
1675 #define abs(n) __builtin_abs ((n))
1676 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1681 @cindex hosted environment
1683 Assert that compilation takes place in a hosted environment. This implies
1684 @option{-fbuiltin}. A hosted environment is one in which the
1685 entire standard library is available, and in which @code{main} has a return
1686 type of @code{int}. Examples are nearly everything except a kernel.
1687 This is equivalent to @option{-fno-freestanding}.
1689 @item -ffreestanding
1690 @opindex ffreestanding
1691 @cindex hosted environment
1693 Assert that compilation takes place in a freestanding environment. This
1694 implies @option{-fno-builtin}. A freestanding environment
1695 is one in which the standard library may not exist, and program startup may
1696 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1697 This is equivalent to @option{-fno-hosted}.
1699 @xref{Standards,,Language Standards Supported by GCC}, for details of
1700 freestanding and hosted environments.
1704 @cindex OpenMP parallel
1705 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1706 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1707 compiler generates parallel code according to the OpenMP Application
1708 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1709 implies @option{-pthread}, and thus is only supported on targets that
1710 have support for @option{-pthread}.
1712 @item -fms-extensions
1713 @opindex fms-extensions
1714 Accept some non-standard constructs used in Microsoft header files.
1716 In C++ code, this allows member names in structures to be similar
1717 to previous types declarations.
1726 Some cases of unnamed fields in structures and unions are only
1727 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1728 fields within structs/unions}, for details.
1730 @item -fplan9-extensions
1731 Accept some non-standard constructs used in Plan 9 code.
1733 This enables @option{-fms-extensions}, permits passing pointers to
1734 structures with anonymous fields to functions which expect pointers to
1735 elements of the type of the field, and permits referring to anonymous
1736 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1737 struct/union fields within structs/unions}, for details. This is only
1738 supported for C, not C++.
1742 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1743 options for strict ISO C conformance) implies @option{-trigraphs}.
1745 @item -no-integrated-cpp
1746 @opindex no-integrated-cpp
1747 Performs a compilation in two passes: preprocessing and compiling. This
1748 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1749 @option{-B} option. The user supplied compilation step can then add in
1750 an additional preprocessing step after normal preprocessing but before
1751 compiling. The default is to use the integrated cpp (internal cpp)
1753 The semantics of this option will change if "cc1", "cc1plus", and
1754 "cc1obj" are merged.
1756 @cindex traditional C language
1757 @cindex C language, traditional
1759 @itemx -traditional-cpp
1760 @opindex traditional-cpp
1761 @opindex traditional
1762 Formerly, these options caused GCC to attempt to emulate a pre-standard
1763 C compiler. They are now only supported with the @option{-E} switch.
1764 The preprocessor continues to support a pre-standard mode. See the GNU
1765 CPP manual for details.
1767 @item -fcond-mismatch
1768 @opindex fcond-mismatch
1769 Allow conditional expressions with mismatched types in the second and
1770 third arguments. The value of such an expression is void. This option
1771 is not supported for C++.
1773 @item -flax-vector-conversions
1774 @opindex flax-vector-conversions
1775 Allow implicit conversions between vectors with differing numbers of
1776 elements and/or incompatible element types. This option should not be
1779 @item -funsigned-char
1780 @opindex funsigned-char
1781 Let the type @code{char} be unsigned, like @code{unsigned char}.
1783 Each kind of machine has a default for what @code{char} should
1784 be. It is either like @code{unsigned char} by default or like
1785 @code{signed char} by default.
1787 Ideally, a portable program should always use @code{signed char} or
1788 @code{unsigned char} when it depends on the signedness of an object.
1789 But many programs have been written to use plain @code{char} and
1790 expect it to be signed, or expect it to be unsigned, depending on the
1791 machines they were written for. This option, and its inverse, let you
1792 make such a program work with the opposite default.
1794 The type @code{char} is always a distinct type from each of
1795 @code{signed char} or @code{unsigned char}, even though its behavior
1796 is always just like one of those two.
1799 @opindex fsigned-char
1800 Let the type @code{char} be signed, like @code{signed char}.
1802 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1803 the negative form of @option{-funsigned-char}. Likewise, the option
1804 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1806 @item -fsigned-bitfields
1807 @itemx -funsigned-bitfields
1808 @itemx -fno-signed-bitfields
1809 @itemx -fno-unsigned-bitfields
1810 @opindex fsigned-bitfields
1811 @opindex funsigned-bitfields
1812 @opindex fno-signed-bitfields
1813 @opindex fno-unsigned-bitfields
1814 These options control whether a bit-field is signed or unsigned, when the
1815 declaration does not use either @code{signed} or @code{unsigned}. By
1816 default, such a bit-field is signed, because this is consistent: the
1817 basic integer types such as @code{int} are signed types.
1820 @node C++ Dialect Options
1821 @section Options Controlling C++ Dialect
1823 @cindex compiler options, C++
1824 @cindex C++ options, command line
1825 @cindex options, C++
1826 This section describes the command-line options that are only meaningful
1827 for C++ programs; but you can also use most of the GNU compiler options
1828 regardless of what language your program is in. For example, you
1829 might compile a file @code{firstClass.C} like this:
1832 g++ -g -frepo -O -c firstClass.C
1836 In this example, only @option{-frepo} is an option meant
1837 only for C++ programs; you can use the other options with any
1838 language supported by GCC@.
1840 Here is a list of options that are @emph{only} for compiling C++ programs:
1844 @item -fabi-version=@var{n}
1845 @opindex fabi-version
1846 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1847 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1848 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1849 the version that conforms most closely to the C++ ABI specification.
1850 Therefore, the ABI obtained using version 0 will change as ABI bugs
1853 The default is version 2.
1855 Version 3 corrects an error in mangling a constant address as a
1858 Version 4 implements a standard mangling for vector types.
1860 Version 5 corrects the mangling of attribute const/volatile on
1861 function pointer types, decltype of a plain decl, and use of a
1862 function parameter in the declaration of another parameter.
1864 Version 6 corrects the promotion behavior of C++0x scoped enums.
1866 See also @option{-Wabi}.
1868 @item -fno-access-control
1869 @opindex fno-access-control
1870 Turn off all access checking. This switch is mainly useful for working
1871 around bugs in the access control code.
1875 Check that the pointer returned by @code{operator new} is non-null
1876 before attempting to modify the storage allocated. This check is
1877 normally unnecessary because the C++ standard specifies that
1878 @code{operator new} will only return @code{0} if it is declared
1879 @samp{throw()}, in which case the compiler will always check the
1880 return value even without this option. In all other cases, when
1881 @code{operator new} has a non-empty exception specification, memory
1882 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1883 @samp{new (nothrow)}.
1885 @item -fconserve-space
1886 @opindex fconserve-space
1887 Put uninitialized or runtime-initialized global variables into the
1888 common segment, as C does. This saves space in the executable at the
1889 cost of not diagnosing duplicate definitions. If you compile with this
1890 flag and your program mysteriously crashes after @code{main()} has
1891 completed, you may have an object that is being destroyed twice because
1892 two definitions were merged.
1894 This option is no longer useful on most targets, now that support has
1895 been added for putting variables into BSS without making them common.
1897 @item -fconstexpr-depth=@var{n}
1898 @opindex fconstexpr-depth
1899 Set the maximum nested evaluation depth for C++0x constexpr functions
1900 to @var{n}. A limit is needed to detect endless recursion during
1901 constant expression evaluation. The minimum specified by the standard
1904 @item -fno-deduce-init-list
1905 @opindex fno-deduce-init-list
1906 Disable deduction of a template type parameter as
1907 std::initializer_list from a brace-enclosed initializer list, i.e.
1910 template <class T> auto forward(T t) -> decltype (realfn (t))
1917 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1921 This option is present because this deduction is an extension to the
1922 current specification in the C++0x working draft, and there was
1923 some concern about potential overload resolution problems.
1925 @item -ffriend-injection
1926 @opindex ffriend-injection
1927 Inject friend functions into the enclosing namespace, so that they are
1928 visible outside the scope of the class in which they are declared.
1929 Friend functions were documented to work this way in the old Annotated
1930 C++ Reference Manual, and versions of G++ before 4.1 always worked
1931 that way. However, in ISO C++ a friend function which is not declared
1932 in an enclosing scope can only be found using argument dependent
1933 lookup. This option causes friends to be injected as they were in
1936 This option is for compatibility, and may be removed in a future
1939 @item -fno-elide-constructors
1940 @opindex fno-elide-constructors
1941 The C++ standard allows an implementation to omit creating a temporary
1942 which is only used to initialize another object of the same type.
1943 Specifying this option disables that optimization, and forces G++ to
1944 call the copy constructor in all cases.
1946 @item -fno-enforce-eh-specs
1947 @opindex fno-enforce-eh-specs
1948 Don't generate code to check for violation of exception specifications
1949 at runtime. This option violates the C++ standard, but may be useful
1950 for reducing code size in production builds, much like defining
1951 @samp{NDEBUG}. This does not give user code permission to throw
1952 exceptions in violation of the exception specifications; the compiler
1953 will still optimize based on the specifications, so throwing an
1954 unexpected exception will result in undefined behavior.
1957 @itemx -fno-for-scope
1959 @opindex fno-for-scope
1960 If @option{-ffor-scope} is specified, the scope of variables declared in
1961 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1962 as specified by the C++ standard.
1963 If @option{-fno-for-scope} is specified, the scope of variables declared in
1964 a @i{for-init-statement} extends to the end of the enclosing scope,
1965 as was the case in old versions of G++, and other (traditional)
1966 implementations of C++.
1968 The default if neither flag is given to follow the standard,
1969 but to allow and give a warning for old-style code that would
1970 otherwise be invalid, or have different behavior.
1972 @item -fno-gnu-keywords
1973 @opindex fno-gnu-keywords
1974 Do not recognize @code{typeof} as a keyword, so that code can use this
1975 word as an identifier. You can use the keyword @code{__typeof__} instead.
1976 @option{-ansi} implies @option{-fno-gnu-keywords}.
1978 @item -fno-implicit-templates
1979 @opindex fno-implicit-templates
1980 Never emit code for non-inline templates which are instantiated
1981 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1982 @xref{Template Instantiation}, for more information.
1984 @item -fno-implicit-inline-templates
1985 @opindex fno-implicit-inline-templates
1986 Don't emit code for implicit instantiations of inline templates, either.
1987 The default is to handle inlines differently so that compiles with and
1988 without optimization will need the same set of explicit instantiations.
1990 @item -fno-implement-inlines
1991 @opindex fno-implement-inlines
1992 To save space, do not emit out-of-line copies of inline functions
1993 controlled by @samp{#pragma implementation}. This will cause linker
1994 errors if these functions are not inlined everywhere they are called.
1996 @item -fms-extensions
1997 @opindex fms-extensions
1998 Disable pedantic warnings about constructs used in MFC, such as implicit
1999 int and getting a pointer to member function via non-standard syntax.
2001 @item -fno-nonansi-builtins
2002 @opindex fno-nonansi-builtins
2003 Disable built-in declarations of functions that are not mandated by
2004 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2005 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2008 @opindex fnothrow-opt
2009 Treat a @code{throw()} exception specification as though it were a
2010 @code{noexcept} specification to reduce or eliminate the text size
2011 overhead relative to a function with no exception specification. If
2012 the function has local variables of types with non-trivial
2013 destructors, the exception specification will actually make the
2014 function smaller because the EH cleanups for those variables can be
2015 optimized away. The semantic effect is that an exception thrown out of
2016 a function with such an exception specification will result in a call
2017 to @code{terminate} rather than @code{unexpected}.
2019 @item -fno-operator-names
2020 @opindex fno-operator-names
2021 Do not treat the operator name keywords @code{and}, @code{bitand},
2022 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2023 synonyms as keywords.
2025 @item -fno-optional-diags
2026 @opindex fno-optional-diags
2027 Disable diagnostics that the standard says a compiler does not need to
2028 issue. Currently, the only such diagnostic issued by G++ is the one for
2029 a name having multiple meanings within a class.
2032 @opindex fpermissive
2033 Downgrade some diagnostics about nonconformant code from errors to
2034 warnings. Thus, using @option{-fpermissive} will allow some
2035 nonconforming code to compile.
2037 @item -fno-pretty-templates
2038 @opindex fno-pretty-templates
2039 When an error message refers to a specialization of a function
2040 template, the compiler will normally print the signature of the
2041 template followed by the template arguments and any typedefs or
2042 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2043 rather than @code{void f(int)}) so that it's clear which template is
2044 involved. When an error message refers to a specialization of a class
2045 template, the compiler will omit any template arguments which match
2046 the default template arguments for that template. If either of these
2047 behaviors make it harder to understand the error message rather than
2048 easier, using @option{-fno-pretty-templates} will disable them.
2052 Enable automatic template instantiation at link time. This option also
2053 implies @option{-fno-implicit-templates}. @xref{Template
2054 Instantiation}, for more information.
2058 Disable generation of information about every class with virtual
2059 functions for use by the C++ runtime type identification features
2060 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2061 of the language, you can save some space by using this flag. Note that
2062 exception handling uses the same information, but it will generate it as
2063 needed. The @samp{dynamic_cast} operator can still be used for casts that
2064 do not require runtime type information, i.e.@: casts to @code{void *} or to
2065 unambiguous base classes.
2069 Emit statistics about front-end processing at the end of the compilation.
2070 This information is generally only useful to the G++ development team.
2072 @item -fstrict-enums
2073 @opindex fstrict-enums
2074 Allow the compiler to optimize using the assumption that a value of
2075 enumeration type can only be one of the values of the enumeration (as
2076 defined in the C++ standard; basically, a value which can be
2077 represented in the minimum number of bits needed to represent all the
2078 enumerators). This assumption may not be valid if the program uses a
2079 cast to convert an arbitrary integer value to the enumeration type.
2081 @item -ftemplate-depth=@var{n}
2082 @opindex ftemplate-depth
2083 Set the maximum instantiation depth for template classes to @var{n}.
2084 A limit on the template instantiation depth is needed to detect
2085 endless recursions during template class instantiation. ANSI/ISO C++
2086 conforming programs must not rely on a maximum depth greater than 17
2087 (changed to 1024 in C++0x). The default value is 900, as the compiler
2088 can run out of stack space before hitting 1024 in some situations.
2090 @item -fno-threadsafe-statics
2091 @opindex fno-threadsafe-statics
2092 Do not emit the extra code to use the routines specified in the C++
2093 ABI for thread-safe initialization of local statics. You can use this
2094 option to reduce code size slightly in code that doesn't need to be
2097 @item -fuse-cxa-atexit
2098 @opindex fuse-cxa-atexit
2099 Register destructors for objects with static storage duration with the
2100 @code{__cxa_atexit} function rather than the @code{atexit} function.
2101 This option is required for fully standards-compliant handling of static
2102 destructors, but will only work if your C library supports
2103 @code{__cxa_atexit}.
2105 @item -fno-use-cxa-get-exception-ptr
2106 @opindex fno-use-cxa-get-exception-ptr
2107 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2108 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2109 if the runtime routine is not available.
2111 @item -fvisibility-inlines-hidden
2112 @opindex fvisibility-inlines-hidden
2113 This switch declares that the user does not attempt to compare
2114 pointers to inline methods where the addresses of the two functions
2115 were taken in different shared objects.
2117 The effect of this is that GCC may, effectively, mark inline methods with
2118 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2119 appear in the export table of a DSO and do not require a PLT indirection
2120 when used within the DSO@. Enabling this option can have a dramatic effect
2121 on load and link times of a DSO as it massively reduces the size of the
2122 dynamic export table when the library makes heavy use of templates.
2124 The behavior of this switch is not quite the same as marking the
2125 methods as hidden directly, because it does not affect static variables
2126 local to the function or cause the compiler to deduce that
2127 the function is defined in only one shared object.
2129 You may mark a method as having a visibility explicitly to negate the
2130 effect of the switch for that method. For example, if you do want to
2131 compare pointers to a particular inline method, you might mark it as
2132 having default visibility. Marking the enclosing class with explicit
2133 visibility will have no effect.
2135 Explicitly instantiated inline methods are unaffected by this option
2136 as their linkage might otherwise cross a shared library boundary.
2137 @xref{Template Instantiation}.
2139 @item -fvisibility-ms-compat
2140 @opindex fvisibility-ms-compat
2141 This flag attempts to use visibility settings to make GCC's C++
2142 linkage model compatible with that of Microsoft Visual Studio.
2144 The flag makes these changes to GCC's linkage model:
2148 It sets the default visibility to @code{hidden}, like
2149 @option{-fvisibility=hidden}.
2152 Types, but not their members, are not hidden by default.
2155 The One Definition Rule is relaxed for types without explicit
2156 visibility specifications which are defined in more than one different
2157 shared object: those declarations are permitted if they would have
2158 been permitted when this option was not used.
2161 In new code it is better to use @option{-fvisibility=hidden} and
2162 export those classes which are intended to be externally visible.
2163 Unfortunately it is possible for code to rely, perhaps accidentally,
2164 on the Visual Studio behavior.
2166 Among the consequences of these changes are that static data members
2167 of the same type with the same name but defined in different shared
2168 objects will be different, so changing one will not change the other;
2169 and that pointers to function members defined in different shared
2170 objects may not compare equal. When this flag is given, it is a
2171 violation of the ODR to define types with the same name differently.
2175 Do not use weak symbol support, even if it is provided by the linker.
2176 By default, G++ will use weak symbols if they are available. This
2177 option exists only for testing, and should not be used by end-users;
2178 it will result in inferior code and has no benefits. This option may
2179 be removed in a future release of G++.
2183 Do not search for header files in the standard directories specific to
2184 C++, but do still search the other standard directories. (This option
2185 is used when building the C++ library.)
2188 In addition, these optimization, warning, and code generation options
2189 have meanings only for C++ programs:
2192 @item -fno-default-inline
2193 @opindex fno-default-inline
2194 Do not assume @samp{inline} for functions defined inside a class scope.
2195 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2196 functions will have linkage like inline functions; they just won't be
2199 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2202 Warn when G++ generates code that is probably not compatible with the
2203 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2204 all such cases, there are probably some cases that are not warned about,
2205 even though G++ is generating incompatible code. There may also be
2206 cases where warnings are emitted even though the code that is generated
2209 You should rewrite your code to avoid these warnings if you are
2210 concerned about the fact that code generated by G++ may not be binary
2211 compatible with code generated by other compilers.
2213 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2218 A template with a non-type template parameter of reference type is
2219 mangled incorrectly:
2222 template <int &> struct S @{@};
2226 This is fixed in @option{-fabi-version=3}.
2229 SIMD vector types declared using @code{__attribute ((vector_size))} are
2230 mangled in a non-standard way that does not allow for overloading of
2231 functions taking vectors of different sizes.
2233 The mangling is changed in @option{-fabi-version=4}.
2236 The known incompatibilities in @option{-fabi-version=1} include:
2241 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2242 pack data into the same byte as a base class. For example:
2245 struct A @{ virtual void f(); int f1 : 1; @};
2246 struct B : public A @{ int f2 : 1; @};
2250 In this case, G++ will place @code{B::f2} into the same byte
2251 as@code{A::f1}; other compilers will not. You can avoid this problem
2252 by explicitly padding @code{A} so that its size is a multiple of the
2253 byte size on your platform; that will cause G++ and other compilers to
2254 layout @code{B} identically.
2257 Incorrect handling of tail-padding for virtual bases. G++ does not use
2258 tail padding when laying out virtual bases. For example:
2261 struct A @{ virtual void f(); char c1; @};
2262 struct B @{ B(); char c2; @};
2263 struct C : public A, public virtual B @{@};
2267 In this case, G++ will not place @code{B} into the tail-padding for
2268 @code{A}; other compilers will. You can avoid this problem by
2269 explicitly padding @code{A} so that its size is a multiple of its
2270 alignment (ignoring virtual base classes); that will cause G++ and other
2271 compilers to layout @code{C} identically.
2274 Incorrect handling of bit-fields with declared widths greater than that
2275 of their underlying types, when the bit-fields appear in a union. For
2279 union U @{ int i : 4096; @};
2283 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2284 union too small by the number of bits in an @code{int}.
2287 Empty classes can be placed at incorrect offsets. For example:
2297 struct C : public B, public A @{@};
2301 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2302 it should be placed at offset zero. G++ mistakenly believes that the
2303 @code{A} data member of @code{B} is already at offset zero.
2306 Names of template functions whose types involve @code{typename} or
2307 template template parameters can be mangled incorrectly.
2310 template <typename Q>
2311 void f(typename Q::X) @{@}
2313 template <template <typename> class Q>
2314 void f(typename Q<int>::X) @{@}
2318 Instantiations of these templates may be mangled incorrectly.
2322 It also warns psABI related changes. The known psABI changes at this
2328 For SYSV/x86-64, when passing union with long double, it is changed to
2329 pass in memory as specified in psABI. For example:
2339 @code{union U} will always be passed in memory.
2343 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2344 @opindex Wctor-dtor-privacy
2345 @opindex Wno-ctor-dtor-privacy
2346 Warn when a class seems unusable because all the constructors or
2347 destructors in that class are private, and it has neither friends nor
2348 public static member functions.
2350 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2351 @opindex Wdelete-non-virtual-dtor
2352 @opindex Wno-delete-non-virtual-dtor
2353 Warn when @samp{delete} is used to destroy an instance of a class which
2354 has virtual functions and non-virtual destructor. It is unsafe to delete
2355 an instance of a derived class through a pointer to a base class if the
2356 base class does not have a virtual destructor. This warning is enabled
2359 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2361 @opindex Wno-narrowing
2362 With -std=c++0x, suppress the diagnostic required by the standard for
2363 narrowing conversions within @samp{@{ @}}, e.g.
2366 int i = @{ 2.2 @}; // error: narrowing from double to int
2369 This flag can be useful for compiling valid C++98 code in C++0x mode.
2371 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2373 @opindex Wno-noexcept
2374 Warn when a noexcept-expression evaluates to false because of a call
2375 to a function that does not have a non-throwing exception
2376 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2377 the compiler to never throw an exception.
2379 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2380 @opindex Wnon-virtual-dtor
2381 @opindex Wno-non-virtual-dtor
2382 Warn when a class has virtual functions and accessible non-virtual
2383 destructor, in which case it would be possible but unsafe to delete
2384 an instance of a derived class through a pointer to the base class.
2385 This warning is also enabled if -Weffc++ is specified.
2387 @item -Wreorder @r{(C++ and Objective-C++ only)}
2389 @opindex Wno-reorder
2390 @cindex reordering, warning
2391 @cindex warning for reordering of member initializers
2392 Warn when the order of member initializers given in the code does not
2393 match the order in which they must be executed. For instance:
2399 A(): j (0), i (1) @{ @}
2403 The compiler will rearrange the member initializers for @samp{i}
2404 and @samp{j} to match the declaration order of the members, emitting
2405 a warning to that effect. This warning is enabled by @option{-Wall}.
2408 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2411 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2414 Warn about violations of the following style guidelines from Scott Meyers'
2415 @cite{Effective C++} book:
2419 Item 11: Define a copy constructor and an assignment operator for classes
2420 with dynamically allocated memory.
2423 Item 12: Prefer initialization to assignment in constructors.
2426 Item 14: Make destructors virtual in base classes.
2429 Item 15: Have @code{operator=} return a reference to @code{*this}.
2432 Item 23: Don't try to return a reference when you must return an object.
2436 Also warn about violations of the following style guidelines from
2437 Scott Meyers' @cite{More Effective C++} book:
2441 Item 6: Distinguish between prefix and postfix forms of increment and
2442 decrement operators.
2445 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2449 When selecting this option, be aware that the standard library
2450 headers do not obey all of these guidelines; use @samp{grep -v}
2451 to filter out those warnings.
2453 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2454 @opindex Wstrict-null-sentinel
2455 @opindex Wno-strict-null-sentinel
2456 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2457 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2458 to @code{__null}. Although it is a null pointer constant not a null pointer,
2459 it is guaranteed to be of the same size as a pointer. But this use is
2460 not portable across different compilers.
2462 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2463 @opindex Wno-non-template-friend
2464 @opindex Wnon-template-friend
2465 Disable warnings when non-templatized friend functions are declared
2466 within a template. Since the advent of explicit template specification
2467 support in G++, if the name of the friend is an unqualified-id (i.e.,
2468 @samp{friend foo(int)}), the C++ language specification demands that the
2469 friend declare or define an ordinary, nontemplate function. (Section
2470 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2471 could be interpreted as a particular specialization of a templatized
2472 function. Because this non-conforming behavior is no longer the default
2473 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2474 check existing code for potential trouble spots and is on by default.
2475 This new compiler behavior can be turned off with
2476 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2477 but disables the helpful warning.
2479 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2480 @opindex Wold-style-cast
2481 @opindex Wno-old-style-cast
2482 Warn if an old-style (C-style) cast to a non-void type is used within
2483 a C++ program. The new-style casts (@samp{dynamic_cast},
2484 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2485 less vulnerable to unintended effects and much easier to search for.
2487 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2488 @opindex Woverloaded-virtual
2489 @opindex Wno-overloaded-virtual
2490 @cindex overloaded virtual function, warning
2491 @cindex warning for overloaded virtual function
2492 Warn when a function declaration hides virtual functions from a
2493 base class. For example, in:
2500 struct B: public A @{
2505 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2513 will fail to compile.
2515 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2516 @opindex Wno-pmf-conversions
2517 @opindex Wpmf-conversions
2518 Disable the diagnostic for converting a bound pointer to member function
2521 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2522 @opindex Wsign-promo
2523 @opindex Wno-sign-promo
2524 Warn when overload resolution chooses a promotion from unsigned or
2525 enumerated type to a signed type, over a conversion to an unsigned type of
2526 the same size. Previous versions of G++ would try to preserve
2527 unsignedness, but the standard mandates the current behavior.
2532 A& operator = (int);
2542 In this example, G++ will synthesize a default @samp{A& operator =
2543 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2546 @node Objective-C and Objective-C++ Dialect Options
2547 @section Options Controlling Objective-C and Objective-C++ Dialects
2549 @cindex compiler options, Objective-C and Objective-C++
2550 @cindex Objective-C and Objective-C++ options, command line
2551 @cindex options, Objective-C and Objective-C++
2552 (NOTE: This manual does not describe the Objective-C and Objective-C++
2553 languages themselves. @xref{Standards,,Language Standards
2554 Supported by GCC}, for references.)
2556 This section describes the command-line options that are only meaningful
2557 for Objective-C and Objective-C++ programs, but you can also use most of
2558 the language-independent GNU compiler options.
2559 For example, you might compile a file @code{some_class.m} like this:
2562 gcc -g -fgnu-runtime -O -c some_class.m
2566 In this example, @option{-fgnu-runtime} is an option meant only for
2567 Objective-C and Objective-C++ programs; you can use the other options with
2568 any language supported by GCC@.
2570 Note that since Objective-C is an extension of the C language, Objective-C
2571 compilations may also use options specific to the C front-end (e.g.,
2572 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2573 C++-specific options (e.g., @option{-Wabi}).
2575 Here is a list of options that are @emph{only} for compiling Objective-C
2576 and Objective-C++ programs:
2579 @item -fconstant-string-class=@var{class-name}
2580 @opindex fconstant-string-class
2581 Use @var{class-name} as the name of the class to instantiate for each
2582 literal string specified with the syntax @code{@@"@dots{}"}. The default
2583 class name is @code{NXConstantString} if the GNU runtime is being used, and
2584 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2585 @option{-fconstant-cfstrings} option, if also present, will override the
2586 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2587 to be laid out as constant CoreFoundation strings.
2590 @opindex fgnu-runtime
2591 Generate object code compatible with the standard GNU Objective-C
2592 runtime. This is the default for most types of systems.
2594 @item -fnext-runtime
2595 @opindex fnext-runtime
2596 Generate output compatible with the NeXT runtime. This is the default
2597 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2598 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2601 @item -fno-nil-receivers
2602 @opindex fno-nil-receivers
2603 Assume that all Objective-C message dispatches (@code{[receiver
2604 message:arg]}) in this translation unit ensure that the receiver is
2605 not @code{nil}. This allows for more efficient entry points in the
2606 runtime to be used. This option is only available in conjunction with
2607 the NeXT runtime and ABI version 0 or 1.
2609 @item -fobjc-abi-version=@var{n}
2610 @opindex fobjc-abi-version
2611 Use version @var{n} of the Objective-C ABI for the selected runtime.
2612 This option is currently supported only for the NeXT runtime. In that
2613 case, Version 0 is the traditional (32-bit) ABI without support for
2614 properties and other Objective-C 2.0 additions. Version 1 is the
2615 traditional (32-bit) ABI with support for properties and other
2616 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2617 nothing is specified, the default is Version 0 on 32-bit target
2618 machines, and Version 2 on 64-bit target machines.
2620 @item -fobjc-call-cxx-cdtors
2621 @opindex fobjc-call-cxx-cdtors
2622 For each Objective-C class, check if any of its instance variables is a
2623 C++ object with a non-trivial default constructor. If so, synthesize a
2624 special @code{- (id) .cxx_construct} instance method that will run
2625 non-trivial default constructors on any such instance variables, in order,
2626 and then return @code{self}. Similarly, check if any instance variable
2627 is a C++ object with a non-trivial destructor, and if so, synthesize a
2628 special @code{- (void) .cxx_destruct} method that will run
2629 all such default destructors, in reverse order.
2631 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2632 methods thusly generated will only operate on instance variables
2633 declared in the current Objective-C class, and not those inherited
2634 from superclasses. It is the responsibility of the Objective-C
2635 runtime to invoke all such methods in an object's inheritance
2636 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2637 by the runtime immediately after a new object instance is allocated;
2638 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2639 before the runtime deallocates an object instance.
2641 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2642 support for invoking the @code{- (id) .cxx_construct} and
2643 @code{- (void) .cxx_destruct} methods.
2645 @item -fobjc-direct-dispatch
2646 @opindex fobjc-direct-dispatch
2647 Allow fast jumps to the message dispatcher. On Darwin this is
2648 accomplished via the comm page.
2650 @item -fobjc-exceptions
2651 @opindex fobjc-exceptions
2652 Enable syntactic support for structured exception handling in
2653 Objective-C, similar to what is offered by C++ and Java. This option
2654 is required to use the Objective-C keywords @code{@@try},
2655 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2656 @code{@@synchronized}. This option is available with both the GNU
2657 runtime and the NeXT runtime (but not available in conjunction with
2658 the NeXT runtime on Mac OS X 10.2 and earlier).
2662 Enable garbage collection (GC) in Objective-C and Objective-C++
2663 programs. This option is only available with the NeXT runtime; the
2664 GNU runtime has a different garbage collection implementation that
2665 does not require special compiler flags.
2667 @item -fobjc-nilcheck
2668 @opindex fobjc-nilcheck
2669 For the NeXT runtime with version 2 of the ABI, check for a nil
2670 receiver in method invocations before doing the actual method call.
2671 This is the default and can be disabled using
2672 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2673 checked for nil in this way no matter what this flag is set to.
2674 Currently this flag does nothing when the GNU runtime, or an older
2675 version of the NeXT runtime ABI, is used.
2677 @item -fobjc-std=objc1
2679 Conform to the language syntax of Objective-C 1.0, the language
2680 recognized by GCC 4.0. This only affects the Objective-C additions to
2681 the C/C++ language; it does not affect conformance to C/C++ standards,
2682 which is controlled by the separate C/C++ dialect option flags. When
2683 this option is used with the Objective-C or Objective-C++ compiler,
2684 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2685 This is useful if you need to make sure that your Objective-C code can
2686 be compiled with older versions of GCC.
2688 @item -freplace-objc-classes
2689 @opindex freplace-objc-classes
2690 Emit a special marker instructing @command{ld(1)} not to statically link in
2691 the resulting object file, and allow @command{dyld(1)} to load it in at
2692 run time instead. This is used in conjunction with the Fix-and-Continue
2693 debugging mode, where the object file in question may be recompiled and
2694 dynamically reloaded in the course of program execution, without the need
2695 to restart the program itself. Currently, Fix-and-Continue functionality
2696 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2701 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2702 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2703 compile time) with static class references that get initialized at load time,
2704 which improves run-time performance. Specifying the @option{-fzero-link} flag
2705 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2706 to be retained. This is useful in Zero-Link debugging mode, since it allows
2707 for individual class implementations to be modified during program execution.
2708 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2709 regardless of command line options.
2713 Dump interface declarations for all classes seen in the source file to a
2714 file named @file{@var{sourcename}.decl}.
2716 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2717 @opindex Wassign-intercept
2718 @opindex Wno-assign-intercept
2719 Warn whenever an Objective-C assignment is being intercepted by the
2722 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2723 @opindex Wno-protocol
2725 If a class is declared to implement a protocol, a warning is issued for
2726 every method in the protocol that is not implemented by the class. The
2727 default behavior is to issue a warning for every method not explicitly
2728 implemented in the class, even if a method implementation is inherited
2729 from the superclass. If you use the @option{-Wno-protocol} option, then
2730 methods inherited from the superclass are considered to be implemented,
2731 and no warning is issued for them.
2733 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2735 @opindex Wno-selector
2736 Warn if multiple methods of different types for the same selector are
2737 found during compilation. The check is performed on the list of methods
2738 in the final stage of compilation. Additionally, a check is performed
2739 for each selector appearing in a @code{@@selector(@dots{})}
2740 expression, and a corresponding method for that selector has been found
2741 during compilation. Because these checks scan the method table only at
2742 the end of compilation, these warnings are not produced if the final
2743 stage of compilation is not reached, for example because an error is
2744 found during compilation, or because the @option{-fsyntax-only} option is
2747 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2748 @opindex Wstrict-selector-match
2749 @opindex Wno-strict-selector-match
2750 Warn if multiple methods with differing argument and/or return types are
2751 found for a given selector when attempting to send a message using this
2752 selector to a receiver of type @code{id} or @code{Class}. When this flag
2753 is off (which is the default behavior), the compiler will omit such warnings
2754 if any differences found are confined to types which share the same size
2757 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2758 @opindex Wundeclared-selector
2759 @opindex Wno-undeclared-selector
2760 Warn if a @code{@@selector(@dots{})} expression referring to an
2761 undeclared selector is found. A selector is considered undeclared if no
2762 method with that name has been declared before the
2763 @code{@@selector(@dots{})} expression, either explicitly in an
2764 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2765 an @code{@@implementation} section. This option always performs its
2766 checks as soon as a @code{@@selector(@dots{})} expression is found,
2767 while @option{-Wselector} only performs its checks in the final stage of
2768 compilation. This also enforces the coding style convention
2769 that methods and selectors must be declared before being used.
2771 @item -print-objc-runtime-info
2772 @opindex print-objc-runtime-info
2773 Generate C header describing the largest structure that is passed by
2778 @node Language Independent Options
2779 @section Options to Control Diagnostic Messages Formatting
2780 @cindex options to control diagnostics formatting
2781 @cindex diagnostic messages
2782 @cindex message formatting
2784 Traditionally, diagnostic messages have been formatted irrespective of
2785 the output device's aspect (e.g.@: its width, @dots{}). The options described
2786 below can be used to control the diagnostic messages formatting
2787 algorithm, e.g.@: how many characters per line, how often source location
2788 information should be reported. Right now, only the C++ front end can
2789 honor these options. However it is expected, in the near future, that
2790 the remaining front ends would be able to digest them correctly.
2793 @item -fmessage-length=@var{n}
2794 @opindex fmessage-length
2795 Try to format error messages so that they fit on lines of about @var{n}
2796 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2797 the front ends supported by GCC@. If @var{n} is zero, then no
2798 line-wrapping will be done; each error message will appear on a single
2801 @opindex fdiagnostics-show-location
2802 @item -fdiagnostics-show-location=once
2803 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2804 reporter to emit @emph{once} source location information; that is, in
2805 case the message is too long to fit on a single physical line and has to
2806 be wrapped, the source location won't be emitted (as prefix) again,
2807 over and over, in subsequent continuation lines. This is the default
2810 @item -fdiagnostics-show-location=every-line
2811 Only meaningful in line-wrapping mode. Instructs the diagnostic
2812 messages reporter to emit the same source location information (as
2813 prefix) for physical lines that result from the process of breaking
2814 a message which is too long to fit on a single line.
2816 @item -fno-diagnostics-show-option
2817 @opindex fno-diagnostics-show-option
2818 @opindex fdiagnostics-show-option
2819 By default, each diagnostic emitted includes text which indicates the
2820 command line option that directly controls the diagnostic (if such an
2821 option is known to the diagnostic machinery). Specifying the
2822 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2824 @item -Wcoverage-mismatch
2825 @opindex Wcoverage-mismatch
2826 Warn if feedback profiles do not match when using the
2827 @option{-fprofile-use} option.
2828 If a source file was changed between @option{-fprofile-gen} and
2829 @option{-fprofile-use}, the files with the profile feedback can fail
2830 to match the source file and GCC can not use the profile feedback
2831 information. By default, this warning is enabled and is treated as an
2832 error. @option{-Wno-coverage-mismatch} can be used to disable the
2833 warning or @option{-Wno-error=coverage-mismatch} can be used to
2834 disable the error. Disable the error for this warning can result in
2835 poorly optimized code, so disabling the error is useful only in the
2836 case of very minor changes such as bug fixes to an existing code-base.
2837 Completely disabling the warning is not recommended.
2841 @node Warning Options
2842 @section Options to Request or Suppress Warnings
2843 @cindex options to control warnings
2844 @cindex warning messages
2845 @cindex messages, warning
2846 @cindex suppressing warnings
2848 Warnings are diagnostic messages that report constructions which
2849 are not inherently erroneous but which are risky or suggest there
2850 may have been an error.
2852 The following language-independent options do not enable specific
2853 warnings but control the kinds of diagnostics produced by GCC.
2856 @cindex syntax checking
2858 @opindex fsyntax-only
2859 Check the code for syntax errors, but don't do anything beyond that.
2861 @item -fmax-errors=@var{n}
2862 @opindex fmax-errors
2863 Limits the maximum number of error messages to @var{n}, at which point
2864 GCC bails out rather than attempting to continue processing the source
2865 code. If @var{n} is 0 (the default), there is no limit on the number
2866 of error messages produced. If @option{-Wfatal-errors} is also
2867 specified, then @option{-Wfatal-errors} takes precedence over this
2872 Inhibit all warning messages.
2877 Make all warnings into errors.
2882 Make the specified warning into an error. The specifier for a warning
2883 is appended, for example @option{-Werror=switch} turns the warnings
2884 controlled by @option{-Wswitch} into errors. This switch takes a
2885 negative form, to be used to negate @option{-Werror} for specific
2886 warnings, for example @option{-Wno-error=switch} makes
2887 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2890 The warning message for each controllable warning includes the
2891 option which controls the warning. That option can then be used with
2892 @option{-Werror=} and @option{-Wno-error=} as described above.
2893 (Printing of the option in the warning message can be disabled using the
2894 @option{-fno-diagnostics-show-option} flag.)
2896 Note that specifying @option{-Werror=}@var{foo} automatically implies
2897 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2900 @item -Wfatal-errors
2901 @opindex Wfatal-errors
2902 @opindex Wno-fatal-errors
2903 This option causes the compiler to abort compilation on the first error
2904 occurred rather than trying to keep going and printing further error
2909 You can request many specific warnings with options beginning
2910 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2911 implicit declarations. Each of these specific warning options also
2912 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2913 example, @option{-Wno-implicit}. This manual lists only one of the
2914 two forms, whichever is not the default. For further,
2915 language-specific options also refer to @ref{C++ Dialect Options} and
2916 @ref{Objective-C and Objective-C++ Dialect Options}.
2918 When an unrecognized warning option is requested (e.g.,
2919 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2920 that the option is not recognized. However, if the @option{-Wno-} form
2921 is used, the behavior is slightly different: No diagnostic will be
2922 produced for @option{-Wno-unknown-warning} unless other diagnostics
2923 are being produced. This allows the use of new @option{-Wno-} options
2924 with old compilers, but if something goes wrong, the compiler will
2925 warn that an unrecognized option was used.
2930 Issue all the warnings demanded by strict ISO C and ISO C++;
2931 reject all programs that use forbidden extensions, and some other
2932 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2933 version of the ISO C standard specified by any @option{-std} option used.
2935 Valid ISO C and ISO C++ programs should compile properly with or without
2936 this option (though a rare few will require @option{-ansi} or a
2937 @option{-std} option specifying the required version of ISO C)@. However,
2938 without this option, certain GNU extensions and traditional C and C++
2939 features are supported as well. With this option, they are rejected.
2941 @option{-pedantic} does not cause warning messages for use of the
2942 alternate keywords whose names begin and end with @samp{__}. Pedantic
2943 warnings are also disabled in the expression that follows
2944 @code{__extension__}. However, only system header files should use
2945 these escape routes; application programs should avoid them.
2946 @xref{Alternate Keywords}.
2948 Some users try to use @option{-pedantic} to check programs for strict ISO
2949 C conformance. They soon find that it does not do quite what they want:
2950 it finds some non-ISO practices, but not all---only those for which
2951 ISO C @emph{requires} a diagnostic, and some others for which
2952 diagnostics have been added.
2954 A feature to report any failure to conform to ISO C might be useful in
2955 some instances, but would require considerable additional work and would
2956 be quite different from @option{-pedantic}. We don't have plans to
2957 support such a feature in the near future.
2959 Where the standard specified with @option{-std} represents a GNU
2960 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2961 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2962 extended dialect is based. Warnings from @option{-pedantic} are given
2963 where they are required by the base standard. (It would not make sense
2964 for such warnings to be given only for features not in the specified GNU
2965 C dialect, since by definition the GNU dialects of C include all
2966 features the compiler supports with the given option, and there would be
2967 nothing to warn about.)
2969 @item -pedantic-errors
2970 @opindex pedantic-errors
2971 Like @option{-pedantic}, except that errors are produced rather than
2977 This enables all the warnings about constructions that some users
2978 consider questionable, and that are easy to avoid (or modify to
2979 prevent the warning), even in conjunction with macros. This also
2980 enables some language-specific warnings described in @ref{C++ Dialect
2981 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2983 @option{-Wall} turns on the following warning flags:
2985 @gccoptlist{-Waddress @gol
2986 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2988 -Wchar-subscripts @gol
2989 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2990 -Wimplicit-int @r{(C and Objective-C only)} @gol
2991 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2994 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2995 -Wmaybe-uninitialized @gol
2996 -Wmissing-braces @gol
3002 -Wsequence-point @gol
3003 -Wsign-compare @r{(only in C++)} @gol
3004 -Wstrict-aliasing @gol
3005 -Wstrict-overflow=1 @gol
3008 -Wuninitialized @gol
3009 -Wunknown-pragmas @gol
3010 -Wunused-function @gol
3013 -Wunused-variable @gol
3014 -Wvolatile-register-var @gol
3017 Note that some warning flags are not implied by @option{-Wall}. Some of
3018 them warn about constructions that users generally do not consider
3019 questionable, but which occasionally you might wish to check for;
3020 others warn about constructions that are necessary or hard to avoid in
3021 some cases, and there is no simple way to modify the code to suppress
3022 the warning. Some of them are enabled by @option{-Wextra} but many of
3023 them must be enabled individually.
3029 This enables some extra warning flags that are not enabled by
3030 @option{-Wall}. (This option used to be called @option{-W}. The older
3031 name is still supported, but the newer name is more descriptive.)
3033 @gccoptlist{-Wclobbered @gol
3035 -Wignored-qualifiers @gol
3036 -Wmissing-field-initializers @gol
3037 -Wmissing-parameter-type @r{(C only)} @gol
3038 -Wold-style-declaration @r{(C only)} @gol
3039 -Woverride-init @gol
3042 -Wuninitialized @gol
3043 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3044 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3047 The option @option{-Wextra} also prints warning messages for the
3053 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3054 @samp{>}, or @samp{>=}.
3057 (C++ only) An enumerator and a non-enumerator both appear in a
3058 conditional expression.
3061 (C++ only) Ambiguous virtual bases.
3064 (C++ only) Subscripting an array which has been declared @samp{register}.
3067 (C++ only) Taking the address of a variable which has been declared
3071 (C++ only) A base class is not initialized in a derived class' copy
3076 @item -Wchar-subscripts
3077 @opindex Wchar-subscripts
3078 @opindex Wno-char-subscripts
3079 Warn if an array subscript has type @code{char}. This is a common cause
3080 of error, as programmers often forget that this type is signed on some
3082 This warning is enabled by @option{-Wall}.
3086 @opindex Wno-comment
3087 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3088 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3089 This warning is enabled by @option{-Wall}.
3092 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3094 Suppress warning messages emitted by @code{#warning} directives.
3096 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3097 @opindex Wdouble-promotion
3098 @opindex Wno-double-promotion
3099 Give a warning when a value of type @code{float} is implicitly
3100 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3101 floating-point unit implement @code{float} in hardware, but emulate
3102 @code{double} in software. On such a machine, doing computations
3103 using @code{double} values is much more expensive because of the
3104 overhead required for software emulation.
3106 It is easy to accidentally do computations with @code{double} because
3107 floating-point literals are implicitly of type @code{double}. For
3111 float area(float radius)
3113 return 3.14159 * radius * radius;
3117 the compiler will perform the entire computation with @code{double}
3118 because the floating-point literal is a @code{double}.
3123 @opindex ffreestanding
3124 @opindex fno-builtin
3125 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3126 the arguments supplied have types appropriate to the format string
3127 specified, and that the conversions specified in the format string make
3128 sense. This includes standard functions, and others specified by format
3129 attributes (@pxref{Function Attributes}), in the @code{printf},
3130 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3131 not in the C standard) families (or other target-specific families).
3132 Which functions are checked without format attributes having been
3133 specified depends on the standard version selected, and such checks of
3134 functions without the attribute specified are disabled by
3135 @option{-ffreestanding} or @option{-fno-builtin}.
3137 The formats are checked against the format features supported by GNU
3138 libc version 2.2. These include all ISO C90 and C99 features, as well
3139 as features from the Single Unix Specification and some BSD and GNU
3140 extensions. Other library implementations may not support all these
3141 features; GCC does not support warning about features that go beyond a
3142 particular library's limitations. However, if @option{-pedantic} is used
3143 with @option{-Wformat}, warnings will be given about format features not
3144 in the selected standard version (but not for @code{strfmon} formats,
3145 since those are not in any version of the C standard). @xref{C Dialect
3146 Options,,Options Controlling C Dialect}.
3148 Since @option{-Wformat} also checks for null format arguments for
3149 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3151 @option{-Wformat} is included in @option{-Wall}. For more control over some
3152 aspects of format checking, the options @option{-Wformat-y2k},
3153 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3154 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3155 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3158 @opindex Wformat-y2k
3159 @opindex Wno-format-y2k
3160 If @option{-Wformat} is specified, also warn about @code{strftime}
3161 formats which may yield only a two-digit year.
3163 @item -Wno-format-contains-nul
3164 @opindex Wno-format-contains-nul
3165 @opindex Wformat-contains-nul
3166 If @option{-Wformat} is specified, do not warn about format strings that
3169 @item -Wno-format-extra-args
3170 @opindex Wno-format-extra-args
3171 @opindex Wformat-extra-args
3172 If @option{-Wformat} is specified, do not warn about excess arguments to a
3173 @code{printf} or @code{scanf} format function. The C standard specifies
3174 that such arguments are ignored.
3176 Where the unused arguments lie between used arguments that are
3177 specified with @samp{$} operand number specifications, normally
3178 warnings are still given, since the implementation could not know what
3179 type to pass to @code{va_arg} to skip the unused arguments. However,
3180 in the case of @code{scanf} formats, this option will suppress the
3181 warning if the unused arguments are all pointers, since the Single
3182 Unix Specification says that such unused arguments are allowed.
3184 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3185 @opindex Wno-format-zero-length
3186 @opindex Wformat-zero-length
3187 If @option{-Wformat} is specified, do not warn about zero-length formats.
3188 The C standard specifies that zero-length formats are allowed.
3190 @item -Wformat-nonliteral
3191 @opindex Wformat-nonliteral
3192 @opindex Wno-format-nonliteral
3193 If @option{-Wformat} is specified, also warn if the format string is not a
3194 string literal and so cannot be checked, unless the format function
3195 takes its format arguments as a @code{va_list}.
3197 @item -Wformat-security
3198 @opindex Wformat-security
3199 @opindex Wno-format-security
3200 If @option{-Wformat} is specified, also warn about uses of format
3201 functions that represent possible security problems. At present, this
3202 warns about calls to @code{printf} and @code{scanf} functions where the
3203 format string is not a string literal and there are no format arguments,
3204 as in @code{printf (foo);}. This may be a security hole if the format
3205 string came from untrusted input and contains @samp{%n}. (This is
3206 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3207 in future warnings may be added to @option{-Wformat-security} that are not
3208 included in @option{-Wformat-nonliteral}.)
3212 @opindex Wno-format=2
3213 Enable @option{-Wformat} plus format checks not included in
3214 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3215 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3217 @item -Wnonnull @r{(C and Objective-C only)}
3219 @opindex Wno-nonnull
3220 Warn about passing a null pointer for arguments marked as
3221 requiring a non-null value by the @code{nonnull} function attribute.
3223 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3224 can be disabled with the @option{-Wno-nonnull} option.
3226 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3228 @opindex Wno-init-self
3229 Warn about uninitialized variables which are initialized with themselves.
3230 Note this option can only be used with the @option{-Wuninitialized} option.
3232 For example, GCC will warn about @code{i} being uninitialized in the
3233 following snippet only when @option{-Winit-self} has been specified:
3244 @item -Wimplicit-int @r{(C and Objective-C only)}
3245 @opindex Wimplicit-int
3246 @opindex Wno-implicit-int
3247 Warn when a declaration does not specify a type.
3248 This warning is enabled by @option{-Wall}.
3250 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3251 @opindex Wimplicit-function-declaration
3252 @opindex Wno-implicit-function-declaration
3253 Give a warning whenever a function is used before being declared. In
3254 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3255 enabled by default and it is made into an error by
3256 @option{-pedantic-errors}. This warning is also enabled by
3259 @item -Wimplicit @r{(C and Objective-C only)}
3261 @opindex Wno-implicit
3262 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3263 This warning is enabled by @option{-Wall}.
3265 @item -Wignored-qualifiers @r{(C and C++ only)}
3266 @opindex Wignored-qualifiers
3267 @opindex Wno-ignored-qualifiers
3268 Warn if the return type of a function has a type qualifier
3269 such as @code{const}. For ISO C such a type qualifier has no effect,
3270 since the value returned by a function is not an lvalue.
3271 For C++, the warning is only emitted for scalar types or @code{void}.
3272 ISO C prohibits qualified @code{void} return types on function
3273 definitions, so such return types always receive a warning
3274 even without this option.
3276 This warning is also enabled by @option{-Wextra}.
3281 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3282 a function with external linkage, returning int, taking either zero
3283 arguments, two, or three arguments of appropriate types. This warning
3284 is enabled by default in C++ and is enabled by either @option{-Wall}
3285 or @option{-pedantic}.
3287 @item -Wmissing-braces
3288 @opindex Wmissing-braces
3289 @opindex Wno-missing-braces
3290 Warn if an aggregate or union initializer is not fully bracketed. In
3291 the following example, the initializer for @samp{a} is not fully
3292 bracketed, but that for @samp{b} is fully bracketed.
3295 int a[2][2] = @{ 0, 1, 2, 3 @};
3296 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3299 This warning is enabled by @option{-Wall}.
3301 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3302 @opindex Wmissing-include-dirs
3303 @opindex Wno-missing-include-dirs
3304 Warn if a user-supplied include directory does not exist.
3307 @opindex Wparentheses
3308 @opindex Wno-parentheses
3309 Warn if parentheses are omitted in certain contexts, such
3310 as when there is an assignment in a context where a truth value
3311 is expected, or when operators are nested whose precedence people
3312 often get confused about.
3314 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3315 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3316 interpretation from that of ordinary mathematical notation.
3318 Also warn about constructions where there may be confusion to which
3319 @code{if} statement an @code{else} branch belongs. Here is an example of
3334 In C/C++, every @code{else} branch belongs to the innermost possible
3335 @code{if} statement, which in this example is @code{if (b)}. This is
3336 often not what the programmer expected, as illustrated in the above
3337 example by indentation the programmer chose. When there is the
3338 potential for this confusion, GCC will issue a warning when this flag
3339 is specified. To eliminate the warning, add explicit braces around
3340 the innermost @code{if} statement so there is no way the @code{else}
3341 could belong to the enclosing @code{if}. The resulting code would
3358 Also warn for dangerous uses of the
3359 ?: with omitted middle operand GNU extension. When the condition
3360 in the ?: operator is a boolean expression the omitted value will
3361 be always 1. Often the user expects it to be a value computed
3362 inside the conditional expression instead.
3364 This warning is enabled by @option{-Wall}.
3366 @item -Wsequence-point
3367 @opindex Wsequence-point
3368 @opindex Wno-sequence-point
3369 Warn about code that may have undefined semantics because of violations
3370 of sequence point rules in the C and C++ standards.
3372 The C and C++ standards defines the order in which expressions in a C/C++
3373 program are evaluated in terms of @dfn{sequence points}, which represent
3374 a partial ordering between the execution of parts of the program: those
3375 executed before the sequence point, and those executed after it. These
3376 occur after the evaluation of a full expression (one which is not part
3377 of a larger expression), after the evaluation of the first operand of a
3378 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3379 function is called (but after the evaluation of its arguments and the
3380 expression denoting the called function), and in certain other places.
3381 Other than as expressed by the sequence point rules, the order of
3382 evaluation of subexpressions of an expression is not specified. All
3383 these rules describe only a partial order rather than a total order,
3384 since, for example, if two functions are called within one expression
3385 with no sequence point between them, the order in which the functions
3386 are called is not specified. However, the standards committee have
3387 ruled that function calls do not overlap.
3389 It is not specified when between sequence points modifications to the
3390 values of objects take effect. Programs whose behavior depends on this
3391 have undefined behavior; the C and C++ standards specify that ``Between
3392 the previous and next sequence point an object shall have its stored
3393 value modified at most once by the evaluation of an expression.
3394 Furthermore, the prior value shall be read only to determine the value
3395 to be stored.''. If a program breaks these rules, the results on any
3396 particular implementation are entirely unpredictable.
3398 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3399 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3400 diagnosed by this option, and it may give an occasional false positive
3401 result, but in general it has been found fairly effective at detecting
3402 this sort of problem in programs.
3404 The standard is worded confusingly, therefore there is some debate
3405 over the precise meaning of the sequence point rules in subtle cases.
3406 Links to discussions of the problem, including proposed formal
3407 definitions, may be found on the GCC readings page, at
3408 @uref{http://gcc.gnu.org/@/readings.html}.
3410 This warning is enabled by @option{-Wall} for C and C++.
3413 @opindex Wreturn-type
3414 @opindex Wno-return-type
3415 Warn whenever a function is defined with a return-type that defaults
3416 to @code{int}. Also warn about any @code{return} statement with no
3417 return-value in a function whose return-type is not @code{void}
3418 (falling off the end of the function body is considered returning
3419 without a value), and about a @code{return} statement with an
3420 expression in a function whose return-type is @code{void}.
3422 For C++, a function without return type always produces a diagnostic
3423 message, even when @option{-Wno-return-type} is specified. The only
3424 exceptions are @samp{main} and functions defined in system headers.
3426 This warning is enabled by @option{-Wall}.
3431 Warn whenever a @code{switch} statement has an index of enumerated type
3432 and lacks a @code{case} for one or more of the named codes of that
3433 enumeration. (The presence of a @code{default} label prevents this
3434 warning.) @code{case} labels outside the enumeration range also
3435 provoke warnings when this option is used (even if there is a
3436 @code{default} label).
3437 This warning is enabled by @option{-Wall}.
3439 @item -Wswitch-default
3440 @opindex Wswitch-default
3441 @opindex Wno-switch-default
3442 Warn whenever a @code{switch} statement does not have a @code{default}
3446 @opindex Wswitch-enum
3447 @opindex Wno-switch-enum
3448 Warn whenever a @code{switch} statement has an index of enumerated type
3449 and lacks a @code{case} for one or more of the named codes of that
3450 enumeration. @code{case} labels outside the enumeration range also
3451 provoke warnings when this option is used. The only difference
3452 between @option{-Wswitch} and this option is that this option gives a
3453 warning about an omitted enumeration code even if there is a
3454 @code{default} label.
3456 @item -Wsync-nand @r{(C and C++ only)}
3458 @opindex Wno-sync-nand
3459 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3460 built-in functions are used. These functions changed semantics in GCC 4.4.
3464 @opindex Wno-trigraphs
3465 Warn if any trigraphs are encountered that might change the meaning of
3466 the program (trigraphs within comments are not warned about).
3467 This warning is enabled by @option{-Wall}.
3469 @item -Wunused-but-set-parameter
3470 @opindex Wunused-but-set-parameter
3471 @opindex Wno-unused-but-set-parameter
3472 Warn whenever a function parameter is assigned to, but otherwise unused
3473 (aside from its declaration).
3475 To suppress this warning use the @samp{unused} attribute
3476 (@pxref{Variable Attributes}).
3478 This warning is also enabled by @option{-Wunused} together with
3481 @item -Wunused-but-set-variable
3482 @opindex Wunused-but-set-variable
3483 @opindex Wno-unused-but-set-variable
3484 Warn whenever a local variable is assigned to, but otherwise unused
3485 (aside from its declaration).
3486 This warning is enabled by @option{-Wall}.
3488 To suppress this warning use the @samp{unused} attribute
3489 (@pxref{Variable Attributes}).
3491 This warning is also enabled by @option{-Wunused}, which is enabled
3494 @item -Wunused-function
3495 @opindex Wunused-function
3496 @opindex Wno-unused-function
3497 Warn whenever a static function is declared but not defined or a
3498 non-inline static function is unused.
3499 This warning is enabled by @option{-Wall}.
3501 @item -Wunused-label
3502 @opindex Wunused-label
3503 @opindex Wno-unused-label
3504 Warn whenever a label is declared but not used.
3505 This warning is enabled by @option{-Wall}.
3507 To suppress this warning use the @samp{unused} attribute
3508 (@pxref{Variable Attributes}).
3510 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
3511 @opindex Wunused-local-typedefs
3512 Warn when a typedef locally defined in a function is not used.
3514 @item -Wunused-parameter
3515 @opindex Wunused-parameter
3516 @opindex Wno-unused-parameter
3517 Warn whenever a function parameter is unused aside from its declaration.
3519 To suppress this warning use the @samp{unused} attribute
3520 (@pxref{Variable Attributes}).
3522 @item -Wno-unused-result
3523 @opindex Wunused-result
3524 @opindex Wno-unused-result
3525 Do not warn if a caller of a function marked with attribute
3526 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3527 its return value. The default is @option{-Wunused-result}.
3529 @item -Wunused-variable
3530 @opindex Wunused-variable
3531 @opindex Wno-unused-variable
3532 Warn whenever a local variable or non-constant static variable is unused
3533 aside from its declaration.
3534 This warning is enabled by @option{-Wall}.
3536 To suppress this warning use the @samp{unused} attribute
3537 (@pxref{Variable Attributes}).
3539 @item -Wunused-value
3540 @opindex Wunused-value
3541 @opindex Wno-unused-value
3542 Warn whenever a statement computes a result that is explicitly not
3543 used. To suppress this warning cast the unused expression to
3544 @samp{void}. This includes an expression-statement or the left-hand
3545 side of a comma expression that contains no side effects. For example,
3546 an expression such as @samp{x[i,j]} will cause a warning, while
3547 @samp{x[(void)i,j]} will not.
3549 This warning is enabled by @option{-Wall}.
3554 All the above @option{-Wunused} options combined.
3556 In order to get a warning about an unused function parameter, you must
3557 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3558 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3560 @item -Wuninitialized
3561 @opindex Wuninitialized
3562 @opindex Wno-uninitialized
3563 Warn if an automatic variable is used without first being initialized
3564 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3565 warn if a non-static reference or non-static @samp{const} member
3566 appears in a class without constructors.
3568 If you want to warn about code which uses the uninitialized value of the
3569 variable in its own initializer, use the @option{-Winit-self} option.
3571 These warnings occur for individual uninitialized or clobbered
3572 elements of structure, union or array variables as well as for
3573 variables which are uninitialized or clobbered as a whole. They do
3574 not occur for variables or elements declared @code{volatile}. Because
3575 these warnings depend on optimization, the exact variables or elements
3576 for which there are warnings will depend on the precise optimization
3577 options and version of GCC used.
3579 Note that there may be no warning about a variable that is used only
3580 to compute a value that itself is never used, because such
3581 computations may be deleted by data flow analysis before the warnings
3584 @item -Wmaybe-uninitialized
3585 @opindex Wmaybe-uninitialized
3586 @opindex Wno-maybe-uninitialized
3587 For an automatic variable, if there exists a path from the function
3588 entry to a use of the variable that is initialized, but there exist
3589 some other paths the variable is not initialized, the compiler will
3590 emit a warning if it can not prove the uninitialized paths do not
3591 happen at runtime. These warnings are made optional because GCC is
3592 not smart enough to see all the reasons why the code might be correct
3593 despite appearing to have an error. Here is one example of how
3614 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3615 always initialized, but GCC doesn't know this. To suppress the
3616 warning, the user needs to provide a default case with assert(0) or
3619 @cindex @code{longjmp} warnings
3620 This option also warns when a non-volatile automatic variable might be
3621 changed by a call to @code{longjmp}. These warnings as well are possible
3622 only in optimizing compilation.
3624 The compiler sees only the calls to @code{setjmp}. It cannot know
3625 where @code{longjmp} will be called; in fact, a signal handler could
3626 call it at any point in the code. As a result, you may get a warning
3627 even when there is in fact no problem because @code{longjmp} cannot
3628 in fact be called at the place which would cause a problem.
3630 Some spurious warnings can be avoided if you declare all the functions
3631 you use that never return as @code{noreturn}. @xref{Function
3634 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3636 @item -Wunknown-pragmas
3637 @opindex Wunknown-pragmas
3638 @opindex Wno-unknown-pragmas
3639 @cindex warning for unknown pragmas
3640 @cindex unknown pragmas, warning
3641 @cindex pragmas, warning of unknown
3642 Warn when a #pragma directive is encountered which is not understood by
3643 GCC@. If this command line option is used, warnings will even be issued
3644 for unknown pragmas in system header files. This is not the case if
3645 the warnings were only enabled by the @option{-Wall} command line option.
3648 @opindex Wno-pragmas
3650 Do not warn about misuses of pragmas, such as incorrect parameters,
3651 invalid syntax, or conflicts between pragmas. See also
3652 @samp{-Wunknown-pragmas}.
3654 @item -Wstrict-aliasing
3655 @opindex Wstrict-aliasing
3656 @opindex Wno-strict-aliasing
3657 This option is only active when @option{-fstrict-aliasing} is active.
3658 It warns about code which might break the strict aliasing rules that the
3659 compiler is using for optimization. The warning does not catch all
3660 cases, but does attempt to catch the more common pitfalls. It is
3661 included in @option{-Wall}.
3662 It is equivalent to @option{-Wstrict-aliasing=3}
3664 @item -Wstrict-aliasing=n
3665 @opindex Wstrict-aliasing=n
3666 @opindex Wno-strict-aliasing=n
3667 This option is only active when @option{-fstrict-aliasing} is active.
3668 It warns about code which might break the strict aliasing rules that the
3669 compiler is using for optimization.
3670 Higher levels correspond to higher accuracy (fewer false positives).
3671 Higher levels also correspond to more effort, similar to the way -O works.
3672 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3675 Level 1: Most aggressive, quick, least accurate.
3676 Possibly useful when higher levels
3677 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3678 false negatives. However, it has many false positives.
3679 Warns for all pointer conversions between possibly incompatible types,
3680 even if never dereferenced. Runs in the frontend only.
3682 Level 2: Aggressive, quick, not too precise.
3683 May still have many false positives (not as many as level 1 though),
3684 and few false negatives (but possibly more than level 1).
3685 Unlike level 1, it only warns when an address is taken. Warns about
3686 incomplete types. Runs in the frontend only.
3688 Level 3 (default for @option{-Wstrict-aliasing}):
3689 Should have very few false positives and few false
3690 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3691 Takes care of the common pun+dereference pattern in the frontend:
3692 @code{*(int*)&some_float}.
3693 If optimization is enabled, it also runs in the backend, where it deals
3694 with multiple statement cases using flow-sensitive points-to information.
3695 Only warns when the converted pointer is dereferenced.
3696 Does not warn about incomplete types.
3698 @item -Wstrict-overflow
3699 @itemx -Wstrict-overflow=@var{n}
3700 @opindex Wstrict-overflow
3701 @opindex Wno-strict-overflow
3702 This option is only active when @option{-fstrict-overflow} is active.
3703 It warns about cases where the compiler optimizes based on the
3704 assumption that signed overflow does not occur. Note that it does not
3705 warn about all cases where the code might overflow: it only warns
3706 about cases where the compiler implements some optimization. Thus
3707 this warning depends on the optimization level.
3709 An optimization which assumes that signed overflow does not occur is
3710 perfectly safe if the values of the variables involved are such that
3711 overflow never does, in fact, occur. Therefore this warning can
3712 easily give a false positive: a warning about code which is not
3713 actually a problem. To help focus on important issues, several
3714 warning levels are defined. No warnings are issued for the use of
3715 undefined signed overflow when estimating how many iterations a loop
3716 will require, in particular when determining whether a loop will be
3720 @item -Wstrict-overflow=1
3721 Warn about cases which are both questionable and easy to avoid. For
3722 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3723 compiler will simplify this to @code{1}. This level of
3724 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3725 are not, and must be explicitly requested.
3727 @item -Wstrict-overflow=2
3728 Also warn about other cases where a comparison is simplified to a
3729 constant. For example: @code{abs (x) >= 0}. This can only be
3730 simplified when @option{-fstrict-overflow} is in effect, because
3731 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3732 zero. @option{-Wstrict-overflow} (with no level) is the same as
3733 @option{-Wstrict-overflow=2}.
3735 @item -Wstrict-overflow=3
3736 Also warn about other cases where a comparison is simplified. For
3737 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3739 @item -Wstrict-overflow=4
3740 Also warn about other simplifications not covered by the above cases.
3741 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3743 @item -Wstrict-overflow=5
3744 Also warn about cases where the compiler reduces the magnitude of a
3745 constant involved in a comparison. For example: @code{x + 2 > y} will
3746 be simplified to @code{x + 1 >= y}. This is reported only at the
3747 highest warning level because this simplification applies to many
3748 comparisons, so this warning level will give a very large number of
3752 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3753 @opindex Wsuggest-attribute=
3754 @opindex Wno-suggest-attribute=
3755 Warn for cases where adding an attribute may be beneficial. The
3756 attributes currently supported are listed below.
3759 @item -Wsuggest-attribute=pure
3760 @itemx -Wsuggest-attribute=const
3761 @itemx -Wsuggest-attribute=noreturn
3762 @opindex Wsuggest-attribute=pure
3763 @opindex Wno-suggest-attribute=pure
3764 @opindex Wsuggest-attribute=const
3765 @opindex Wno-suggest-attribute=const
3766 @opindex Wsuggest-attribute=noreturn
3767 @opindex Wno-suggest-attribute=noreturn
3769 Warn about functions which might be candidates for attributes
3770 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3771 functions visible in other compilation units or (in the case of @code{pure} and
3772 @code{const}) if it cannot prove that the function returns normally. A function
3773 returns normally if it doesn't contain an infinite loop nor returns abnormally
3774 by throwing, calling @code{abort()} or trapping. This analysis requires option
3775 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3776 higher. Higher optimization levels improve the accuracy of the analysis.
3779 @item -Warray-bounds
3780 @opindex Wno-array-bounds
3781 @opindex Warray-bounds
3782 This option is only active when @option{-ftree-vrp} is active
3783 (default for @option{-O2} and above). It warns about subscripts to arrays
3784 that are always out of bounds. This warning is enabled by @option{-Wall}.
3786 @item -Wno-div-by-zero
3787 @opindex Wno-div-by-zero
3788 @opindex Wdiv-by-zero
3789 Do not warn about compile-time integer division by zero. Floating point
3790 division by zero is not warned about, as it can be a legitimate way of
3791 obtaining infinities and NaNs.
3793 @item -Wsystem-headers
3794 @opindex Wsystem-headers
3795 @opindex Wno-system-headers
3796 @cindex warnings from system headers
3797 @cindex system headers, warnings from
3798 Print warning messages for constructs found in system header files.
3799 Warnings from system headers are normally suppressed, on the assumption
3800 that they usually do not indicate real problems and would only make the
3801 compiler output harder to read. Using this command line option tells
3802 GCC to emit warnings from system headers as if they occurred in user
3803 code. However, note that using @option{-Wall} in conjunction with this
3804 option will @emph{not} warn about unknown pragmas in system
3805 headers---for that, @option{-Wunknown-pragmas} must also be used.
3808 @opindex Wtrampolines
3809 @opindex Wno-trampolines
3810 Warn about trampolines generated for pointers to nested functions.
3812 A trampoline is a small piece of data or code that is created at run
3813 time on the stack when the address of a nested function is taken, and
3814 is used to call the nested function indirectly. For some targets, it
3815 is made up of data only and thus requires no special treatment. But,
3816 for most targets, it is made up of code and thus requires the stack
3817 to be made executable in order for the program to work properly.
3820 @opindex Wfloat-equal
3821 @opindex Wno-float-equal
3822 Warn if floating point values are used in equality comparisons.
3824 The idea behind this is that sometimes it is convenient (for the
3825 programmer) to consider floating-point values as approximations to
3826 infinitely precise real numbers. If you are doing this, then you need
3827 to compute (by analyzing the code, or in some other way) the maximum or
3828 likely maximum error that the computation introduces, and allow for it
3829 when performing comparisons (and when producing output, but that's a
3830 different problem). In particular, instead of testing for equality, you
3831 would check to see whether the two values have ranges that overlap; and
3832 this is done with the relational operators, so equality comparisons are
3835 @item -Wtraditional @r{(C and Objective-C only)}
3836 @opindex Wtraditional
3837 @opindex Wno-traditional
3838 Warn about certain constructs that behave differently in traditional and
3839 ISO C@. Also warn about ISO C constructs that have no traditional C
3840 equivalent, and/or problematic constructs which should be avoided.
3844 Macro parameters that appear within string literals in the macro body.
3845 In traditional C macro replacement takes place within string literals,
3846 but does not in ISO C@.
3849 In traditional C, some preprocessor directives did not exist.
3850 Traditional preprocessors would only consider a line to be a directive
3851 if the @samp{#} appeared in column 1 on the line. Therefore
3852 @option{-Wtraditional} warns about directives that traditional C
3853 understands but would ignore because the @samp{#} does not appear as the
3854 first character on the line. It also suggests you hide directives like
3855 @samp{#pragma} not understood by traditional C by indenting them. Some
3856 traditional implementations would not recognize @samp{#elif}, so it
3857 suggests avoiding it altogether.
3860 A function-like macro that appears without arguments.
3863 The unary plus operator.
3866 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3867 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3868 constants.) Note, these suffixes appear in macros defined in the system
3869 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3870 Use of these macros in user code might normally lead to spurious
3871 warnings, however GCC's integrated preprocessor has enough context to
3872 avoid warning in these cases.
3875 A function declared external in one block and then used after the end of
3879 A @code{switch} statement has an operand of type @code{long}.
3882 A non-@code{static} function declaration follows a @code{static} one.
3883 This construct is not accepted by some traditional C compilers.
3886 The ISO type of an integer constant has a different width or
3887 signedness from its traditional type. This warning is only issued if
3888 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3889 typically represent bit patterns, are not warned about.
3892 Usage of ISO string concatenation is detected.
3895 Initialization of automatic aggregates.
3898 Identifier conflicts with labels. Traditional C lacks a separate
3899 namespace for labels.
3902 Initialization of unions. If the initializer is zero, the warning is
3903 omitted. This is done under the assumption that the zero initializer in
3904 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3905 initializer warnings and relies on default initialization to zero in the
3909 Conversions by prototypes between fixed/floating point values and vice
3910 versa. The absence of these prototypes when compiling with traditional
3911 C would cause serious problems. This is a subset of the possible
3912 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3915 Use of ISO C style function definitions. This warning intentionally is
3916 @emph{not} issued for prototype declarations or variadic functions
3917 because these ISO C features will appear in your code when using
3918 libiberty's traditional C compatibility macros, @code{PARAMS} and
3919 @code{VPARAMS}. This warning is also bypassed for nested functions
3920 because that feature is already a GCC extension and thus not relevant to
3921 traditional C compatibility.
3924 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3925 @opindex Wtraditional-conversion
3926 @opindex Wno-traditional-conversion
3927 Warn if a prototype causes a type conversion that is different from what
3928 would happen to the same argument in the absence of a prototype. This
3929 includes conversions of fixed point to floating and vice versa, and
3930 conversions changing the width or signedness of a fixed point argument
3931 except when the same as the default promotion.
3933 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3934 @opindex Wdeclaration-after-statement
3935 @opindex Wno-declaration-after-statement
3936 Warn when a declaration is found after a statement in a block. This
3937 construct, known from C++, was introduced with ISO C99 and is by default
3938 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3939 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3944 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3946 @item -Wno-endif-labels
3947 @opindex Wno-endif-labels
3948 @opindex Wendif-labels
3949 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3954 Warn whenever a local variable or type declaration shadows another variable,
3955 parameter, type, or class member (in C++), or whenever a built-in function
3956 is shadowed. Note that in C++, the compiler will not warn if a local variable
3957 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3959 @item -Wlarger-than=@var{len}
3960 @opindex Wlarger-than=@var{len}
3961 @opindex Wlarger-than-@var{len}
3962 Warn whenever an object of larger than @var{len} bytes is defined.
3964 @item -Wframe-larger-than=@var{len}
3965 @opindex Wframe-larger-than
3966 Warn if the size of a function frame is larger than @var{len} bytes.
3967 The computation done to determine the stack frame size is approximate
3968 and not conservative.
3969 The actual requirements may be somewhat greater than @var{len}
3970 even if you do not get a warning. In addition, any space allocated
3971 via @code{alloca}, variable-length arrays, or related constructs
3972 is not included by the compiler when determining
3973 whether or not to issue a warning.
3975 @item -Wno-free-nonheap-object
3976 @opindex Wno-free-nonheap-object
3977 @opindex Wfree-nonheap-object
3978 Do not warn when attempting to free an object which was not allocated
3981 @item -Wstack-usage=@var{len}
3982 @opindex Wstack-usage
3983 Warn if the stack usage of a function might be larger than @var{len} bytes.
3984 The computation done to determine the stack usage is conservative.
3985 Any space allocated via @code{alloca}, variable-length arrays, or related
3986 constructs is included by the compiler when determining whether or not to
3989 The message is in keeping with the output of @option{-fstack-usage}.
3993 If the stack usage is fully static but exceeds the specified amount, it's:
3996 Â warning: stack usage is 1120 bytes
3999 If the stack usage is (partly) dynamic but bounded, it's:
4002 Â warning: stack usage might be 1648 bytes
4005 If the stack usage is (partly) dynamic and not bounded, it's:
4008 Â warning: stack usage might be unbounded
4012 @item -Wunsafe-loop-optimizations
4013 @opindex Wunsafe-loop-optimizations
4014 @opindex Wno-unsafe-loop-optimizations
4015 Warn if the loop cannot be optimized because the compiler could not
4016 assume anything on the bounds of the loop indices. With
4017 @option{-funsafe-loop-optimizations} warn if the compiler made
4020 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4021 @opindex Wno-pedantic-ms-format
4022 @opindex Wpedantic-ms-format
4023 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4024 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4025 depending on the MS runtime, when you are using the options @option{-Wformat}
4026 and @option{-pedantic} without gnu-extensions.
4028 @item -Wpointer-arith
4029 @opindex Wpointer-arith
4030 @opindex Wno-pointer-arith
4031 Warn about anything that depends on the ``size of'' a function type or
4032 of @code{void}. GNU C assigns these types a size of 1, for
4033 convenience in calculations with @code{void *} pointers and pointers
4034 to functions. In C++, warn also when an arithmetic operation involves
4035 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4038 @opindex Wtype-limits
4039 @opindex Wno-type-limits
4040 Warn if a comparison is always true or always false due to the limited
4041 range of the data type, but do not warn for constant expressions. For
4042 example, warn if an unsigned variable is compared against zero with
4043 @samp{<} or @samp{>=}. This warning is also enabled by
4046 @item -Wbad-function-cast @r{(C and Objective-C only)}
4047 @opindex Wbad-function-cast
4048 @opindex Wno-bad-function-cast
4049 Warn whenever a function call is cast to a non-matching type.
4050 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4052 @item -Wc++-compat @r{(C and Objective-C only)}
4053 Warn about ISO C constructs that are outside of the common subset of
4054 ISO C and ISO C++, e.g.@: request for implicit conversion from
4055 @code{void *} to a pointer to non-@code{void} type.
4057 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
4058 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
4059 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
4060 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
4064 @opindex Wno-cast-qual
4065 Warn whenever a pointer is cast so as to remove a type qualifier from
4066 the target type. For example, warn if a @code{const char *} is cast
4067 to an ordinary @code{char *}.
4069 Also warn when making a cast which introduces a type qualifier in an
4070 unsafe way. For example, casting @code{char **} to @code{const char **}
4071 is unsafe, as in this example:
4074 /* p is char ** value. */
4075 const char **q = (const char **) p;
4076 /* Assignment of readonly string to const char * is OK. */
4078 /* Now char** pointer points to read-only memory. */
4083 @opindex Wcast-align
4084 @opindex Wno-cast-align
4085 Warn whenever a pointer is cast such that the required alignment of the
4086 target is increased. For example, warn if a @code{char *} is cast to
4087 an @code{int *} on machines where integers can only be accessed at
4088 two- or four-byte boundaries.
4090 @item -Wwrite-strings
4091 @opindex Wwrite-strings
4092 @opindex Wno-write-strings
4093 When compiling C, give string constants the type @code{const
4094 char[@var{length}]} so that copying the address of one into a
4095 non-@code{const} @code{char *} pointer will get a warning. These
4096 warnings will help you find at compile time code that can try to write
4097 into a string constant, but only if you have been very careful about
4098 using @code{const} in declarations and prototypes. Otherwise, it will
4099 just be a nuisance. This is why we did not make @option{-Wall} request
4102 When compiling C++, warn about the deprecated conversion from string
4103 literals to @code{char *}. This warning is enabled by default for C++
4108 @opindex Wno-clobbered
4109 Warn for variables that might be changed by @samp{longjmp} or
4110 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4113 @opindex Wconversion
4114 @opindex Wno-conversion
4115 Warn for implicit conversions that may alter a value. This includes
4116 conversions between real and integer, like @code{abs (x)} when
4117 @code{x} is @code{double}; conversions between signed and unsigned,
4118 like @code{unsigned ui = -1}; and conversions to smaller types, like
4119 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4120 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4121 changed by the conversion like in @code{abs (2.0)}. Warnings about
4122 conversions between signed and unsigned integers can be disabled by
4123 using @option{-Wno-sign-conversion}.
4125 For C++, also warn for confusing overload resolution for user-defined
4126 conversions; and conversions that will never use a type conversion
4127 operator: conversions to @code{void}, the same type, a base class or a
4128 reference to them. Warnings about conversions between signed and
4129 unsigned integers are disabled by default in C++ unless
4130 @option{-Wsign-conversion} is explicitly enabled.
4132 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4133 @opindex Wconversion-null
4134 @opindex Wno-conversion-null
4135 Do not warn for conversions between @code{NULL} and non-pointer
4136 types. @option{-Wconversion-null} is enabled by default.
4139 @opindex Wempty-body
4140 @opindex Wno-empty-body
4141 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4142 while} statement. This warning is also enabled by @option{-Wextra}.
4144 @item -Wenum-compare
4145 @opindex Wenum-compare
4146 @opindex Wno-enum-compare
4147 Warn about a comparison between values of different enum types. In C++
4148 this warning is enabled by default. In C this warning is enabled by
4151 @item -Wjump-misses-init @r{(C, Objective-C only)}
4152 @opindex Wjump-misses-init
4153 @opindex Wno-jump-misses-init
4154 Warn if a @code{goto} statement or a @code{switch} statement jumps
4155 forward across the initialization of a variable, or jumps backward to a
4156 label after the variable has been initialized. This only warns about
4157 variables which are initialized when they are declared. This warning is
4158 only supported for C and Objective C; in C++ this sort of branch is an
4161 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4162 can be disabled with the @option{-Wno-jump-misses-init} option.
4164 @item -Wsign-compare
4165 @opindex Wsign-compare
4166 @opindex Wno-sign-compare
4167 @cindex warning for comparison of signed and unsigned values
4168 @cindex comparison of signed and unsigned values, warning
4169 @cindex signed and unsigned values, comparison warning
4170 Warn when a comparison between signed and unsigned values could produce
4171 an incorrect result when the signed value is converted to unsigned.
4172 This warning is also enabled by @option{-Wextra}; to get the other warnings
4173 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4175 @item -Wsign-conversion
4176 @opindex Wsign-conversion
4177 @opindex Wno-sign-conversion
4178 Warn for implicit conversions that may change the sign of an integer
4179 value, like assigning a signed integer expression to an unsigned
4180 integer variable. An explicit cast silences the warning. In C, this
4181 option is enabled also by @option{-Wconversion}.
4185 @opindex Wno-address
4186 Warn about suspicious uses of memory addresses. These include using
4187 the address of a function in a conditional expression, such as
4188 @code{void func(void); if (func)}, and comparisons against the memory
4189 address of a string literal, such as @code{if (x == "abc")}. Such
4190 uses typically indicate a programmer error: the address of a function
4191 always evaluates to true, so their use in a conditional usually
4192 indicate that the programmer forgot the parentheses in a function
4193 call; and comparisons against string literals result in unspecified
4194 behavior and are not portable in C, so they usually indicate that the
4195 programmer intended to use @code{strcmp}. This warning is enabled by
4199 @opindex Wlogical-op
4200 @opindex Wno-logical-op
4201 Warn about suspicious uses of logical operators in expressions.
4202 This includes using logical operators in contexts where a
4203 bit-wise operator is likely to be expected.
4205 @item -Waggregate-return
4206 @opindex Waggregate-return
4207 @opindex Wno-aggregate-return
4208 Warn if any functions that return structures or unions are defined or
4209 called. (In languages where you can return an array, this also elicits
4212 @item -Wno-attributes
4213 @opindex Wno-attributes
4214 @opindex Wattributes
4215 Do not warn if an unexpected @code{__attribute__} is used, such as
4216 unrecognized attributes, function attributes applied to variables,
4217 etc. This will not stop errors for incorrect use of supported
4220 @item -Wno-builtin-macro-redefined
4221 @opindex Wno-builtin-macro-redefined
4222 @opindex Wbuiltin-macro-redefined
4223 Do not warn if certain built-in macros are redefined. This suppresses
4224 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4225 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4227 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4228 @opindex Wstrict-prototypes
4229 @opindex Wno-strict-prototypes
4230 Warn if a function is declared or defined without specifying the
4231 argument types. (An old-style function definition is permitted without
4232 a warning if preceded by a declaration which specifies the argument
4235 @item -Wold-style-declaration @r{(C and Objective-C only)}
4236 @opindex Wold-style-declaration
4237 @opindex Wno-old-style-declaration
4238 Warn for obsolescent usages, according to the C Standard, in a
4239 declaration. For example, warn if storage-class specifiers like
4240 @code{static} are not the first things in a declaration. This warning
4241 is also enabled by @option{-Wextra}.
4243 @item -Wold-style-definition @r{(C and Objective-C only)}
4244 @opindex Wold-style-definition
4245 @opindex Wno-old-style-definition
4246 Warn if an old-style function definition is used. A warning is given
4247 even if there is a previous prototype.
4249 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4250 @opindex Wmissing-parameter-type
4251 @opindex Wno-missing-parameter-type
4252 A function parameter is declared without a type specifier in K&R-style
4259 This warning is also enabled by @option{-Wextra}.
4261 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4262 @opindex Wmissing-prototypes
4263 @opindex Wno-missing-prototypes
4264 Warn if a global function is defined without a previous prototype
4265 declaration. This warning is issued even if the definition itself
4266 provides a prototype. The aim is to detect global functions that fail
4267 to be declared in header files.
4269 @item -Wmissing-declarations
4270 @opindex Wmissing-declarations
4271 @opindex Wno-missing-declarations
4272 Warn if a global function is defined without a previous declaration.
4273 Do so even if the definition itself provides a prototype.
4274 Use this option to detect global functions that are not declared in
4275 header files. In C++, no warnings are issued for function templates,
4276 or for inline functions, or for functions in anonymous namespaces.
4278 @item -Wmissing-field-initializers
4279 @opindex Wmissing-field-initializers
4280 @opindex Wno-missing-field-initializers
4284 Warn if a structure's initializer has some fields missing. For
4285 example, the following code would cause such a warning, because
4286 @code{x.h} is implicitly zero:
4289 struct s @{ int f, g, h; @};
4290 struct s x = @{ 3, 4 @};
4293 This option does not warn about designated initializers, so the following
4294 modification would not trigger a warning:
4297 struct s @{ int f, g, h; @};
4298 struct s x = @{ .f = 3, .g = 4 @};
4301 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4302 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4304 @item -Wmissing-format-attribute
4305 @opindex Wmissing-format-attribute
4306 @opindex Wno-missing-format-attribute
4309 Warn about function pointers which might be candidates for @code{format}
4310 attributes. Note these are only possible candidates, not absolute ones.
4311 GCC will guess that function pointers with @code{format} attributes that
4312 are used in assignment, initialization, parameter passing or return
4313 statements should have a corresponding @code{format} attribute in the
4314 resulting type. I.e.@: the left-hand side of the assignment or
4315 initialization, the type of the parameter variable, or the return type
4316 of the containing function respectively should also have a @code{format}
4317 attribute to avoid the warning.
4319 GCC will also warn about function definitions which might be
4320 candidates for @code{format} attributes. Again, these are only
4321 possible candidates. GCC will guess that @code{format} attributes
4322 might be appropriate for any function that calls a function like
4323 @code{vprintf} or @code{vscanf}, but this might not always be the
4324 case, and some functions for which @code{format} attributes are
4325 appropriate may not be detected.
4327 @item -Wno-multichar
4328 @opindex Wno-multichar
4330 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4331 Usually they indicate a typo in the user's code, as they have
4332 implementation-defined values, and should not be used in portable code.
4334 @item -Wnormalized=<none|id|nfc|nfkc>
4335 @opindex Wnormalized=
4338 @cindex character set, input normalization
4339 In ISO C and ISO C++, two identifiers are different if they are
4340 different sequences of characters. However, sometimes when characters
4341 outside the basic ASCII character set are used, you can have two
4342 different character sequences that look the same. To avoid confusion,
4343 the ISO 10646 standard sets out some @dfn{normalization rules} which
4344 when applied ensure that two sequences that look the same are turned into
4345 the same sequence. GCC can warn you if you are using identifiers which
4346 have not been normalized; this option controls that warning.
4348 There are four levels of warning that GCC supports. The default is
4349 @option{-Wnormalized=nfc}, which warns about any identifier which is
4350 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4351 recommended form for most uses.
4353 Unfortunately, there are some characters which ISO C and ISO C++ allow
4354 in identifiers that when turned into NFC aren't allowable as
4355 identifiers. That is, there's no way to use these symbols in portable
4356 ISO C or C++ and have all your identifiers in NFC@.
4357 @option{-Wnormalized=id} suppresses the warning for these characters.
4358 It is hoped that future versions of the standards involved will correct
4359 this, which is why this option is not the default.
4361 You can switch the warning off for all characters by writing
4362 @option{-Wnormalized=none}. You would only want to do this if you
4363 were using some other normalization scheme (like ``D''), because
4364 otherwise you can easily create bugs that are literally impossible to see.
4366 Some characters in ISO 10646 have distinct meanings but look identical
4367 in some fonts or display methodologies, especially once formatting has
4368 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4369 LETTER N'', will display just like a regular @code{n} which has been
4370 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4371 normalization scheme to convert all these into a standard form as
4372 well, and GCC will warn if your code is not in NFKC if you use
4373 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4374 about every identifier that contains the letter O because it might be
4375 confused with the digit 0, and so is not the default, but may be
4376 useful as a local coding convention if the programming environment is
4377 unable to be fixed to display these characters distinctly.
4379 @item -Wno-deprecated
4380 @opindex Wno-deprecated
4381 @opindex Wdeprecated
4382 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4384 @item -Wno-deprecated-declarations
4385 @opindex Wno-deprecated-declarations
4386 @opindex Wdeprecated-declarations
4387 Do not warn about uses of functions (@pxref{Function Attributes}),
4388 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4389 Attributes}) marked as deprecated by using the @code{deprecated}
4393 @opindex Wno-overflow
4395 Do not warn about compile-time overflow in constant expressions.
4397 @item -Woverride-init @r{(C and Objective-C only)}
4398 @opindex Woverride-init
4399 @opindex Wno-override-init
4403 Warn if an initialized field without side effects is overridden when
4404 using designated initializers (@pxref{Designated Inits, , Designated
4407 This warning is included in @option{-Wextra}. To get other
4408 @option{-Wextra} warnings without this one, use @samp{-Wextra
4409 -Wno-override-init}.
4414 Warn if a structure is given the packed attribute, but the packed
4415 attribute has no effect on the layout or size of the structure.
4416 Such structures may be mis-aligned for little benefit. For
4417 instance, in this code, the variable @code{f.x} in @code{struct bar}
4418 will be misaligned even though @code{struct bar} does not itself
4419 have the packed attribute:
4426 @} __attribute__((packed));
4434 @item -Wpacked-bitfield-compat
4435 @opindex Wpacked-bitfield-compat
4436 @opindex Wno-packed-bitfield-compat
4437 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4438 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4439 the change can lead to differences in the structure layout. GCC
4440 informs you when the offset of such a field has changed in GCC 4.4.
4441 For example there is no longer a 4-bit padding between field @code{a}
4442 and @code{b} in this structure:
4449 @} __attribute__ ((packed));
4452 This warning is enabled by default. Use
4453 @option{-Wno-packed-bitfield-compat} to disable this warning.
4458 Warn if padding is included in a structure, either to align an element
4459 of the structure or to align the whole structure. Sometimes when this
4460 happens it is possible to rearrange the fields of the structure to
4461 reduce the padding and so make the structure smaller.
4463 @item -Wredundant-decls
4464 @opindex Wredundant-decls
4465 @opindex Wno-redundant-decls
4466 Warn if anything is declared more than once in the same scope, even in
4467 cases where multiple declaration is valid and changes nothing.
4469 @item -Wnested-externs @r{(C and Objective-C only)}
4470 @opindex Wnested-externs
4471 @opindex Wno-nested-externs
4472 Warn if an @code{extern} declaration is encountered within a function.
4477 Warn if a function can not be inlined and it was declared as inline.
4478 Even with this option, the compiler will not warn about failures to
4479 inline functions declared in system headers.
4481 The compiler uses a variety of heuristics to determine whether or not
4482 to inline a function. For example, the compiler takes into account
4483 the size of the function being inlined and the amount of inlining
4484 that has already been done in the current function. Therefore,
4485 seemingly insignificant changes in the source program can cause the
4486 warnings produced by @option{-Winline} to appear or disappear.
4488 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4489 @opindex Wno-invalid-offsetof
4490 @opindex Winvalid-offsetof
4491 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4492 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4493 to a non-POD type is undefined. In existing C++ implementations,
4494 however, @samp{offsetof} typically gives meaningful results even when
4495 applied to certain kinds of non-POD types. (Such as a simple
4496 @samp{struct} that fails to be a POD type only by virtue of having a
4497 constructor.) This flag is for users who are aware that they are
4498 writing nonportable code and who have deliberately chosen to ignore the
4501 The restrictions on @samp{offsetof} may be relaxed in a future version
4502 of the C++ standard.
4504 @item -Wno-int-to-pointer-cast
4505 @opindex Wno-int-to-pointer-cast
4506 @opindex Wint-to-pointer-cast
4507 Suppress warnings from casts to pointer type of an integer of a
4508 different size. In C++, casting to a pointer type of smaller size is
4509 an error. @option{Wint-to-pointer-cast} is enabled by default.
4512 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4513 @opindex Wno-pointer-to-int-cast
4514 @opindex Wpointer-to-int-cast
4515 Suppress warnings from casts from a pointer to an integer type of a
4519 @opindex Winvalid-pch
4520 @opindex Wno-invalid-pch
4521 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4522 the search path but can't be used.
4526 @opindex Wno-long-long
4527 Warn if @samp{long long} type is used. This is enabled by either
4528 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4529 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4531 @item -Wvariadic-macros
4532 @opindex Wvariadic-macros
4533 @opindex Wno-variadic-macros
4534 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4535 alternate syntax when in pedantic ISO C99 mode. This is default.
4536 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4541 Warn if variable length array is used in the code.
4542 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4543 the variable length array.
4545 @item -Wvolatile-register-var
4546 @opindex Wvolatile-register-var
4547 @opindex Wno-volatile-register-var
4548 Warn if a register variable is declared volatile. The volatile
4549 modifier does not inhibit all optimizations that may eliminate reads
4550 and/or writes to register variables. This warning is enabled by
4553 @item -Wdisabled-optimization
4554 @opindex Wdisabled-optimization
4555 @opindex Wno-disabled-optimization
4556 Warn if a requested optimization pass is disabled. This warning does
4557 not generally indicate that there is anything wrong with your code; it
4558 merely indicates that GCC's optimizers were unable to handle the code
4559 effectively. Often, the problem is that your code is too big or too
4560 complex; GCC will refuse to optimize programs when the optimization
4561 itself is likely to take inordinate amounts of time.
4563 @item -Wpointer-sign @r{(C and Objective-C only)}
4564 @opindex Wpointer-sign
4565 @opindex Wno-pointer-sign
4566 Warn for pointer argument passing or assignment with different signedness.
4567 This option is only supported for C and Objective-C@. It is implied by
4568 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4569 @option{-Wno-pointer-sign}.
4571 @item -Wstack-protector
4572 @opindex Wstack-protector
4573 @opindex Wno-stack-protector
4574 This option is only active when @option{-fstack-protector} is active. It
4575 warns about functions that will not be protected against stack smashing.
4578 @opindex Wno-mudflap
4579 Suppress warnings about constructs that cannot be instrumented by
4582 @item -Woverlength-strings
4583 @opindex Woverlength-strings
4584 @opindex Wno-overlength-strings
4585 Warn about string constants which are longer than the ``minimum
4586 maximum'' length specified in the C standard. Modern compilers
4587 generally allow string constants which are much longer than the
4588 standard's minimum limit, but very portable programs should avoid
4589 using longer strings.
4591 The limit applies @emph{after} string constant concatenation, and does
4592 not count the trailing NUL@. In C90, the limit was 509 characters; in
4593 C99, it was raised to 4095. C++98 does not specify a normative
4594 minimum maximum, so we do not diagnose overlength strings in C++@.
4596 This option is implied by @option{-pedantic}, and can be disabled with
4597 @option{-Wno-overlength-strings}.
4599 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4600 @opindex Wunsuffixed-float-constants
4602 GCC will issue a warning for any floating constant that does not have
4603 a suffix. When used together with @option{-Wsystem-headers} it will
4604 warn about such constants in system header files. This can be useful
4605 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4606 from the decimal floating-point extension to C99.
4609 @node Debugging Options
4610 @section Options for Debugging Your Program or GCC
4611 @cindex options, debugging
4612 @cindex debugging information options
4614 GCC has various special options that are used for debugging
4615 either your program or GCC:
4620 Produce debugging information in the operating system's native format
4621 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4624 On most systems that use stabs format, @option{-g} enables use of extra
4625 debugging information that only GDB can use; this extra information
4626 makes debugging work better in GDB but will probably make other debuggers
4628 refuse to read the program. If you want to control for certain whether
4629 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4630 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4632 GCC allows you to use @option{-g} with
4633 @option{-O}. The shortcuts taken by optimized code may occasionally
4634 produce surprising results: some variables you declared may not exist
4635 at all; flow of control may briefly move where you did not expect it;
4636 some statements may not be executed because they compute constant
4637 results or their values were already at hand; some statements may
4638 execute in different places because they were moved out of loops.
4640 Nevertheless it proves possible to debug optimized output. This makes
4641 it reasonable to use the optimizer for programs that might have bugs.
4643 The following options are useful when GCC is generated with the
4644 capability for more than one debugging format.
4648 Produce debugging information for use by GDB@. This means to use the
4649 most expressive format available (DWARF 2, stabs, or the native format
4650 if neither of those are supported), including GDB extensions if at all
4655 Produce debugging information in stabs format (if that is supported),
4656 without GDB extensions. This is the format used by DBX on most BSD
4657 systems. On MIPS, Alpha and System V Release 4 systems this option
4658 produces stabs debugging output which is not understood by DBX or SDB@.
4659 On System V Release 4 systems this option requires the GNU assembler.
4661 @item -feliminate-unused-debug-symbols
4662 @opindex feliminate-unused-debug-symbols
4663 Produce debugging information in stabs format (if that is supported),
4664 for only symbols that are actually used.
4666 @item -femit-class-debug-always
4667 Instead of emitting debugging information for a C++ class in only one
4668 object file, emit it in all object files using the class. This option
4669 should be used only with debuggers that are unable to handle the way GCC
4670 normally emits debugging information for classes because using this
4671 option will increase the size of debugging information by as much as a
4674 @item -fno-debug-types-section
4675 @opindex fno-debug-types-section
4676 @opindex fdebug-types-section
4677 By default when using DWARF v4 or higher type DIEs will be put into
4678 their own .debug_types section instead of making them part of the
4679 .debug_info section. It is more efficient to put them in a separate
4680 comdat sections since the linker will then be able to remove duplicates.
4681 But not all DWARF consumers support .debug_types sections yet.
4685 Produce debugging information in stabs format (if that is supported),
4686 using GNU extensions understood only by the GNU debugger (GDB)@. The
4687 use of these extensions is likely to make other debuggers crash or
4688 refuse to read the program.
4692 Produce debugging information in COFF format (if that is supported).
4693 This is the format used by SDB on most System V systems prior to
4698 Produce debugging information in XCOFF format (if that is supported).
4699 This is the format used by the DBX debugger on IBM RS/6000 systems.
4703 Produce debugging information in XCOFF format (if that is supported),
4704 using GNU extensions understood only by the GNU debugger (GDB)@. The
4705 use of these extensions is likely to make other debuggers crash or
4706 refuse to read the program, and may cause assemblers other than the GNU
4707 assembler (GAS) to fail with an error.
4709 @item -gdwarf-@var{version}
4710 @opindex gdwarf-@var{version}
4711 Produce debugging information in DWARF format (if that is
4712 supported). This is the format used by DBX on IRIX 6. The value
4713 of @var{version} may be either 2, 3 or 4; the default version is 2.
4715 Note that with DWARF version 2 some ports require, and will always
4716 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4718 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4719 for maximum benefit.
4721 @item -grecord-gcc-switches
4722 @opindex grecord-gcc-switches
4723 This switch causes the command line options, that were used to invoke the
4724 compiler and may affect code generation, to be appended to the
4725 DW_AT_producer attribute in DWARF debugging information. The options
4726 are concatenated with spaces separating them from each other and from
4727 the compiler version. See also @option{-frecord-gcc-switches} for another
4728 way of storing compiler options into the object file.
4730 @item -gno-record-gcc-switches
4731 @opindex gno-record-gcc-switches
4732 Disallow appending command line options to the DW_AT_producer attribute
4733 in DWARF debugging information. This is the default.
4735 @item -gstrict-dwarf
4736 @opindex gstrict-dwarf
4737 Disallow using extensions of later DWARF standard version than selected
4738 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4739 DWARF extensions from later standard versions is allowed.
4741 @item -gno-strict-dwarf
4742 @opindex gno-strict-dwarf
4743 Allow using extensions of later DWARF standard version than selected with
4744 @option{-gdwarf-@var{version}}.
4748 Produce debugging information in VMS debug format (if that is
4749 supported). This is the format used by DEBUG on VMS systems.
4752 @itemx -ggdb@var{level}
4753 @itemx -gstabs@var{level}
4754 @itemx -gcoff@var{level}
4755 @itemx -gxcoff@var{level}
4756 @itemx -gvms@var{level}
4757 Request debugging information and also use @var{level} to specify how
4758 much information. The default level is 2.
4760 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4763 Level 1 produces minimal information, enough for making backtraces in
4764 parts of the program that you don't plan to debug. This includes
4765 descriptions of functions and external variables, but no information
4766 about local variables and no line numbers.
4768 Level 3 includes extra information, such as all the macro definitions
4769 present in the program. Some debuggers support macro expansion when
4770 you use @option{-g3}.
4772 @option{-gdwarf-2} does not accept a concatenated debug level, because
4773 GCC used to support an option @option{-gdwarf} that meant to generate
4774 debug information in version 1 of the DWARF format (which is very
4775 different from version 2), and it would have been too confusing. That
4776 debug format is long obsolete, but the option cannot be changed now.
4777 Instead use an additional @option{-g@var{level}} option to change the
4778 debug level for DWARF.
4782 Turn off generation of debug info, if leaving out this option would have
4783 generated it, or turn it on at level 2 otherwise. The position of this
4784 argument in the command line does not matter, it takes effect after all
4785 other options are processed, and it does so only once, no matter how
4786 many times it is given. This is mainly intended to be used with
4787 @option{-fcompare-debug}.
4789 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4790 @opindex fdump-final-insns
4791 Dump the final internal representation (RTL) to @var{file}. If the
4792 optional argument is omitted (or if @var{file} is @code{.}), the name
4793 of the dump file will be determined by appending @code{.gkd} to the
4794 compilation output file name.
4796 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4797 @opindex fcompare-debug
4798 @opindex fno-compare-debug
4799 If no error occurs during compilation, run the compiler a second time,
4800 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4801 passed to the second compilation. Dump the final internal
4802 representation in both compilations, and print an error if they differ.
4804 If the equal sign is omitted, the default @option{-gtoggle} is used.
4806 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4807 and nonzero, implicitly enables @option{-fcompare-debug}. If
4808 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4809 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4812 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4813 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4814 of the final representation and the second compilation, preventing even
4815 @env{GCC_COMPARE_DEBUG} from taking effect.
4817 To verify full coverage during @option{-fcompare-debug} testing, set
4818 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4819 which GCC will reject as an invalid option in any actual compilation
4820 (rather than preprocessing, assembly or linking). To get just a
4821 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4822 not overridden} will do.
4824 @item -fcompare-debug-second
4825 @opindex fcompare-debug-second
4826 This option is implicitly passed to the compiler for the second
4827 compilation requested by @option{-fcompare-debug}, along with options to
4828 silence warnings, and omitting other options that would cause
4829 side-effect compiler outputs to files or to the standard output. Dump
4830 files and preserved temporary files are renamed so as to contain the
4831 @code{.gk} additional extension during the second compilation, to avoid
4832 overwriting those generated by the first.
4834 When this option is passed to the compiler driver, it causes the
4835 @emph{first} compilation to be skipped, which makes it useful for little
4836 other than debugging the compiler proper.
4838 @item -feliminate-dwarf2-dups
4839 @opindex feliminate-dwarf2-dups
4840 Compress DWARF2 debugging information by eliminating duplicated
4841 information about each symbol. This option only makes sense when
4842 generating DWARF2 debugging information with @option{-gdwarf-2}.
4844 @item -femit-struct-debug-baseonly
4845 Emit debug information for struct-like types
4846 only when the base name of the compilation source file
4847 matches the base name of file in which the struct was defined.
4849 This option substantially reduces the size of debugging information,
4850 but at significant potential loss in type information to the debugger.
4851 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4852 See @option{-femit-struct-debug-detailed} for more detailed control.
4854 This option works only with DWARF 2.
4856 @item -femit-struct-debug-reduced
4857 Emit debug information for struct-like types
4858 only when the base name of the compilation source file
4859 matches the base name of file in which the type was defined,
4860 unless the struct is a template or defined in a system header.
4862 This option significantly reduces the size of debugging information,
4863 with some potential loss in type information to the debugger.
4864 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4865 See @option{-femit-struct-debug-detailed} for more detailed control.
4867 This option works only with DWARF 2.
4869 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4870 Specify the struct-like types
4871 for which the compiler will generate debug information.
4872 The intent is to reduce duplicate struct debug information
4873 between different object files within the same program.
4875 This option is a detailed version of
4876 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4877 which will serve for most needs.
4879 A specification has the syntax@*
4880 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4882 The optional first word limits the specification to
4883 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4884 A struct type is used directly when it is the type of a variable, member.
4885 Indirect uses arise through pointers to structs.
4886 That is, when use of an incomplete struct would be legal, the use is indirect.
4888 @samp{struct one direct; struct two * indirect;}.
4890 The optional second word limits the specification to
4891 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4892 Generic structs are a bit complicated to explain.
4893 For C++, these are non-explicit specializations of template classes,
4894 or non-template classes within the above.
4895 Other programming languages have generics,
4896 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4898 The third word specifies the source files for those
4899 structs for which the compiler will emit debug information.
4900 The values @samp{none} and @samp{any} have the normal meaning.
4901 The value @samp{base} means that
4902 the base of name of the file in which the type declaration appears
4903 must match the base of the name of the main compilation file.
4904 In practice, this means that
4905 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4906 but types declared in other header will not.
4907 The value @samp{sys} means those types satisfying @samp{base}
4908 or declared in system or compiler headers.
4910 You may need to experiment to determine the best settings for your application.
4912 The default is @samp{-femit-struct-debug-detailed=all}.
4914 This option works only with DWARF 2.
4916 @item -fno-merge-debug-strings
4917 @opindex fmerge-debug-strings
4918 @opindex fno-merge-debug-strings
4919 Direct the linker to not merge together strings in the debugging
4920 information which are identical in different object files. Merging is
4921 not supported by all assemblers or linkers. Merging decreases the size
4922 of the debug information in the output file at the cost of increasing
4923 link processing time. Merging is enabled by default.
4925 @item -fdebug-prefix-map=@var{old}=@var{new}
4926 @opindex fdebug-prefix-map
4927 When compiling files in directory @file{@var{old}}, record debugging
4928 information describing them as in @file{@var{new}} instead.
4930 @item -fno-dwarf2-cfi-asm
4931 @opindex fdwarf2-cfi-asm
4932 @opindex fno-dwarf2-cfi-asm
4933 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4934 instead of using GAS @code{.cfi_*} directives.
4936 @cindex @command{prof}
4939 Generate extra code to write profile information suitable for the
4940 analysis program @command{prof}. You must use this option when compiling
4941 the source files you want data about, and you must also use it when
4944 @cindex @command{gprof}
4947 Generate extra code to write profile information suitable for the
4948 analysis program @command{gprof}. You must use this option when compiling
4949 the source files you want data about, and you must also use it when
4954 Makes the compiler print out each function name as it is compiled, and
4955 print some statistics about each pass when it finishes.
4958 @opindex ftime-report
4959 Makes the compiler print some statistics about the time consumed by each
4960 pass when it finishes.
4963 @opindex fmem-report
4964 Makes the compiler print some statistics about permanent memory
4965 allocation when it finishes.
4967 @item -fpre-ipa-mem-report
4968 @opindex fpre-ipa-mem-report
4969 @item -fpost-ipa-mem-report
4970 @opindex fpost-ipa-mem-report
4971 Makes the compiler print some statistics about permanent memory
4972 allocation before or after interprocedural optimization.
4975 @opindex fstack-usage
4976 Makes the compiler output stack usage information for the program, on a
4977 per-function basis. The filename for the dump is made by appending
4978 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4979 the output file, if explicitly specified and it is not an executable,
4980 otherwise it is the basename of the source file. An entry is made up
4985 The name of the function.
4989 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4992 The qualifier @code{static} means that the function manipulates the stack
4993 statically: a fixed number of bytes are allocated for the frame on function
4994 entry and released on function exit; no stack adjustments are otherwise made
4995 in the function. The second field is this fixed number of bytes.
4997 The qualifier @code{dynamic} means that the function manipulates the stack
4998 dynamically: in addition to the static allocation described above, stack
4999 adjustments are made in the body of the function, for example to push/pop
5000 arguments around function calls. If the qualifier @code{bounded} is also
5001 present, the amount of these adjustments is bounded at compile-time and
5002 the second field is an upper bound of the total amount of stack used by
5003 the function. If it is not present, the amount of these adjustments is
5004 not bounded at compile-time and the second field only represents the
5007 @item -fprofile-arcs
5008 @opindex fprofile-arcs
5009 Add code so that program flow @dfn{arcs} are instrumented. During
5010 execution the program records how many times each branch and call is
5011 executed and how many times it is taken or returns. When the compiled
5012 program exits it saves this data to a file called
5013 @file{@var{auxname}.gcda} for each source file. The data may be used for
5014 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5015 test coverage analysis (@option{-ftest-coverage}). Each object file's
5016 @var{auxname} is generated from the name of the output file, if
5017 explicitly specified and it is not the final executable, otherwise it is
5018 the basename of the source file. In both cases any suffix is removed
5019 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5020 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5021 @xref{Cross-profiling}.
5023 @cindex @command{gcov}
5027 This option is used to compile and link code instrumented for coverage
5028 analysis. The option is a synonym for @option{-fprofile-arcs}
5029 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5030 linking). See the documentation for those options for more details.
5035 Compile the source files with @option{-fprofile-arcs} plus optimization
5036 and code generation options. For test coverage analysis, use the
5037 additional @option{-ftest-coverage} option. You do not need to profile
5038 every source file in a program.
5041 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5042 (the latter implies the former).
5045 Run the program on a representative workload to generate the arc profile
5046 information. This may be repeated any number of times. You can run
5047 concurrent instances of your program, and provided that the file system
5048 supports locking, the data files will be correctly updated. Also
5049 @code{fork} calls are detected and correctly handled (double counting
5053 For profile-directed optimizations, compile the source files again with
5054 the same optimization and code generation options plus
5055 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5056 Control Optimization}).
5059 For test coverage analysis, use @command{gcov} to produce human readable
5060 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5061 @command{gcov} documentation for further information.
5065 With @option{-fprofile-arcs}, for each function of your program GCC
5066 creates a program flow graph, then finds a spanning tree for the graph.
5067 Only arcs that are not on the spanning tree have to be instrumented: the
5068 compiler adds code to count the number of times that these arcs are
5069 executed. When an arc is the only exit or only entrance to a block, the
5070 instrumentation code can be added to the block; otherwise, a new basic
5071 block must be created to hold the instrumentation code.
5074 @item -ftest-coverage
5075 @opindex ftest-coverage
5076 Produce a notes file that the @command{gcov} code-coverage utility
5077 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5078 show program coverage. Each source file's note file is called
5079 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5080 above for a description of @var{auxname} and instructions on how to
5081 generate test coverage data. Coverage data will match the source files
5082 more closely, if you do not optimize.
5084 @item -fdbg-cnt-list
5085 @opindex fdbg-cnt-list
5086 Print the name and the counter upper bound for all debug counters.
5089 @item -fdbg-cnt=@var{counter-value-list}
5091 Set the internal debug counter upper bound. @var{counter-value-list}
5092 is a comma-separated list of @var{name}:@var{value} pairs
5093 which sets the upper bound of each debug counter @var{name} to @var{value}.
5094 All debug counters have the initial upper bound of @var{UINT_MAX},
5095 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5096 e.g. With -fdbg-cnt=dce:10,tail_call:0
5097 dbg_cnt(dce) will return true only for first 10 invocations
5099 @itemx -fenable-@var{kind}-@var{pass}
5100 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5104 This is a set of debugging options that are used to explicitly disable/enable
5105 optimization passes. For compiler users, regular options for enabling/disabling
5106 passes should be used instead.
5110 @item -fdisable-ipa-@var{pass}
5111 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5112 statically invoked in the compiler multiple times, the pass name should be
5113 appended with a sequential number starting from 1.
5115 @item -fdisable-rtl-@var{pass}
5116 @item -fdisable-rtl-@var{pass}=@var{range-list}
5117 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5118 statically invoked in the compiler multiple times, the pass name should be
5119 appended with a sequential number starting from 1. @var{range-list} is a comma
5120 seperated list of function ranges or assembler names. Each range is a number
5121 pair seperated by a colon. The range is inclusive in both ends. If the range
5122 is trivial, the number pair can be simplified as a single number. If the
5123 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5124 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5125 function header of a dump file, and the pass names can be dumped by using
5126 option @option{-fdump-passes}.
5128 @item -fdisable-tree-@var{pass}
5129 @item -fdisable-tree-@var{pass}=@var{range-list}
5130 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5133 @item -fenable-ipa-@var{pass}
5134 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5135 statically invoked in the compiler multiple times, the pass name should be
5136 appended with a sequential number starting from 1.
5138 @item -fenable-rtl-@var{pass}
5139 @item -fenable-rtl-@var{pass}=@var{range-list}
5140 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5141 description and examples.
5143 @item -fenable-tree-@var{pass}
5144 @item -fenable-tree-@var{pass}=@var{range-list}
5145 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5146 of option arguments.
5150 # disable ccp1 for all functions
5152 # disable complete unroll for function whose cgraph node uid is 1
5153 -fenable-tree-cunroll=1
5154 # disable gcse2 for functions at the following ranges [1,1],
5155 # [300,400], and [400,1000]
5156 # disable gcse2 for functions foo and foo2
5157 -fdisable-rtl-gcse2=foo,foo2
5158 # disable early inlining
5159 -fdisable-tree-einline
5160 # disable ipa inlining
5161 -fdisable-ipa-inline
5162 # enable tree full unroll
5163 -fenable-tree-unroll
5169 @item -d@var{letters}
5170 @itemx -fdump-rtl-@var{pass}
5172 Says to make debugging dumps during compilation at times specified by
5173 @var{letters}. This is used for debugging the RTL-based passes of the
5174 compiler. The file names for most of the dumps are made by appending
5175 a pass number and a word to the @var{dumpname}, and the files are
5176 created in the directory of the output file. Note that the pass
5177 number is computed statically as passes get registered into the pass
5178 manager. Thus the numbering is not related to the dynamic order of
5179 execution of passes. In particular, a pass installed by a plugin
5180 could have a number over 200 even if it executed quite early.
5181 @var{dumpname} is generated from the name of the output file, if
5182 explicitly specified and it is not an executable, otherwise it is the
5183 basename of the source file. These switches may have different effects
5184 when @option{-E} is used for preprocessing.
5186 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5187 @option{-d} option @var{letters}. Here are the possible
5188 letters for use in @var{pass} and @var{letters}, and their meanings:
5192 @item -fdump-rtl-alignments
5193 @opindex fdump-rtl-alignments
5194 Dump after branch alignments have been computed.
5196 @item -fdump-rtl-asmcons
5197 @opindex fdump-rtl-asmcons
5198 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5200 @item -fdump-rtl-auto_inc_dec
5201 @opindex fdump-rtl-auto_inc_dec
5202 Dump after auto-inc-dec discovery. This pass is only run on
5203 architectures that have auto inc or auto dec instructions.
5205 @item -fdump-rtl-barriers
5206 @opindex fdump-rtl-barriers
5207 Dump after cleaning up the barrier instructions.
5209 @item -fdump-rtl-bbpart
5210 @opindex fdump-rtl-bbpart
5211 Dump after partitioning hot and cold basic blocks.
5213 @item -fdump-rtl-bbro
5214 @opindex fdump-rtl-bbro
5215 Dump after block reordering.
5217 @item -fdump-rtl-btl1
5218 @itemx -fdump-rtl-btl2
5219 @opindex fdump-rtl-btl2
5220 @opindex fdump-rtl-btl2
5221 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5222 after the two branch
5223 target load optimization passes.
5225 @item -fdump-rtl-bypass
5226 @opindex fdump-rtl-bypass
5227 Dump after jump bypassing and control flow optimizations.
5229 @item -fdump-rtl-combine
5230 @opindex fdump-rtl-combine
5231 Dump after the RTL instruction combination pass.
5233 @item -fdump-rtl-compgotos
5234 @opindex fdump-rtl-compgotos
5235 Dump after duplicating the computed gotos.
5237 @item -fdump-rtl-ce1
5238 @itemx -fdump-rtl-ce2
5239 @itemx -fdump-rtl-ce3
5240 @opindex fdump-rtl-ce1
5241 @opindex fdump-rtl-ce2
5242 @opindex fdump-rtl-ce3
5243 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5244 @option{-fdump-rtl-ce3} enable dumping after the three
5245 if conversion passes.
5247 @itemx -fdump-rtl-cprop_hardreg
5248 @opindex fdump-rtl-cprop_hardreg
5249 Dump after hard register copy propagation.
5251 @itemx -fdump-rtl-csa
5252 @opindex fdump-rtl-csa
5253 Dump after combining stack adjustments.
5255 @item -fdump-rtl-cse1
5256 @itemx -fdump-rtl-cse2
5257 @opindex fdump-rtl-cse1
5258 @opindex fdump-rtl-cse2
5259 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5260 the two common sub-expression elimination passes.
5262 @itemx -fdump-rtl-dce
5263 @opindex fdump-rtl-dce
5264 Dump after the standalone dead code elimination passes.
5266 @itemx -fdump-rtl-dbr
5267 @opindex fdump-rtl-dbr
5268 Dump after delayed branch scheduling.
5270 @item -fdump-rtl-dce1
5271 @itemx -fdump-rtl-dce2
5272 @opindex fdump-rtl-dce1
5273 @opindex fdump-rtl-dce2
5274 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5275 the two dead store elimination passes.
5278 @opindex fdump-rtl-eh
5279 Dump after finalization of EH handling code.
5281 @item -fdump-rtl-eh_ranges
5282 @opindex fdump-rtl-eh_ranges
5283 Dump after conversion of EH handling range regions.
5285 @item -fdump-rtl-expand
5286 @opindex fdump-rtl-expand
5287 Dump after RTL generation.
5289 @item -fdump-rtl-fwprop1
5290 @itemx -fdump-rtl-fwprop2
5291 @opindex fdump-rtl-fwprop1
5292 @opindex fdump-rtl-fwprop2
5293 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5294 dumping after the two forward propagation passes.
5296 @item -fdump-rtl-gcse1
5297 @itemx -fdump-rtl-gcse2
5298 @opindex fdump-rtl-gcse1
5299 @opindex fdump-rtl-gcse2
5300 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5301 after global common subexpression elimination.
5303 @item -fdump-rtl-init-regs
5304 @opindex fdump-rtl-init-regs
5305 Dump after the initialization of the registers.
5307 @item -fdump-rtl-initvals
5308 @opindex fdump-rtl-initvals
5309 Dump after the computation of the initial value sets.
5311 @itemx -fdump-rtl-into_cfglayout
5312 @opindex fdump-rtl-into_cfglayout
5313 Dump after converting to cfglayout mode.
5315 @item -fdump-rtl-ira
5316 @opindex fdump-rtl-ira
5317 Dump after iterated register allocation.
5319 @item -fdump-rtl-jump
5320 @opindex fdump-rtl-jump
5321 Dump after the second jump optimization.
5323 @item -fdump-rtl-loop2
5324 @opindex fdump-rtl-loop2
5325 @option{-fdump-rtl-loop2} enables dumping after the rtl
5326 loop optimization passes.
5328 @item -fdump-rtl-mach
5329 @opindex fdump-rtl-mach
5330 Dump after performing the machine dependent reorganization pass, if that
5333 @item -fdump-rtl-mode_sw
5334 @opindex fdump-rtl-mode_sw
5335 Dump after removing redundant mode switches.
5337 @item -fdump-rtl-rnreg
5338 @opindex fdump-rtl-rnreg
5339 Dump after register renumbering.
5341 @itemx -fdump-rtl-outof_cfglayout
5342 @opindex fdump-rtl-outof_cfglayout
5343 Dump after converting from cfglayout mode.
5345 @item -fdump-rtl-peephole2
5346 @opindex fdump-rtl-peephole2
5347 Dump after the peephole pass.
5349 @item -fdump-rtl-postreload
5350 @opindex fdump-rtl-postreload
5351 Dump after post-reload optimizations.
5353 @itemx -fdump-rtl-pro_and_epilogue
5354 @opindex fdump-rtl-pro_and_epilogue
5355 Dump after generating the function pro and epilogues.
5357 @item -fdump-rtl-regmove
5358 @opindex fdump-rtl-regmove
5359 Dump after the register move pass.
5361 @item -fdump-rtl-sched1
5362 @itemx -fdump-rtl-sched2
5363 @opindex fdump-rtl-sched1
5364 @opindex fdump-rtl-sched2
5365 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5366 after the basic block scheduling passes.
5368 @item -fdump-rtl-see
5369 @opindex fdump-rtl-see
5370 Dump after sign extension elimination.
5372 @item -fdump-rtl-seqabstr
5373 @opindex fdump-rtl-seqabstr
5374 Dump after common sequence discovery.
5376 @item -fdump-rtl-shorten
5377 @opindex fdump-rtl-shorten
5378 Dump after shortening branches.
5380 @item -fdump-rtl-sibling
5381 @opindex fdump-rtl-sibling
5382 Dump after sibling call optimizations.
5384 @item -fdump-rtl-split1
5385 @itemx -fdump-rtl-split2
5386 @itemx -fdump-rtl-split3
5387 @itemx -fdump-rtl-split4
5388 @itemx -fdump-rtl-split5
5389 @opindex fdump-rtl-split1
5390 @opindex fdump-rtl-split2
5391 @opindex fdump-rtl-split3
5392 @opindex fdump-rtl-split4
5393 @opindex fdump-rtl-split5
5394 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5395 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5396 @option{-fdump-rtl-split5} enable dumping after five rounds of
5397 instruction splitting.
5399 @item -fdump-rtl-sms
5400 @opindex fdump-rtl-sms
5401 Dump after modulo scheduling. This pass is only run on some
5404 @item -fdump-rtl-stack
5405 @opindex fdump-rtl-stack
5406 Dump after conversion from GCC's "flat register file" registers to the
5407 x87's stack-like registers. This pass is only run on x86 variants.
5409 @item -fdump-rtl-subreg1
5410 @itemx -fdump-rtl-subreg2
5411 @opindex fdump-rtl-subreg1
5412 @opindex fdump-rtl-subreg2
5413 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5414 the two subreg expansion passes.
5416 @item -fdump-rtl-unshare
5417 @opindex fdump-rtl-unshare
5418 Dump after all rtl has been unshared.
5420 @item -fdump-rtl-vartrack
5421 @opindex fdump-rtl-vartrack
5422 Dump after variable tracking.
5424 @item -fdump-rtl-vregs
5425 @opindex fdump-rtl-vregs
5426 Dump after converting virtual registers to hard registers.
5428 @item -fdump-rtl-web
5429 @opindex fdump-rtl-web
5430 Dump after live range splitting.
5432 @item -fdump-rtl-regclass
5433 @itemx -fdump-rtl-subregs_of_mode_init
5434 @itemx -fdump-rtl-subregs_of_mode_finish
5435 @itemx -fdump-rtl-dfinit
5436 @itemx -fdump-rtl-dfinish
5437 @opindex fdump-rtl-regclass
5438 @opindex fdump-rtl-subregs_of_mode_init
5439 @opindex fdump-rtl-subregs_of_mode_finish
5440 @opindex fdump-rtl-dfinit
5441 @opindex fdump-rtl-dfinish
5442 These dumps are defined but always produce empty files.
5444 @item -fdump-rtl-all
5445 @opindex fdump-rtl-all
5446 Produce all the dumps listed above.
5450 Annotate the assembler output with miscellaneous debugging information.
5454 Dump all macro definitions, at the end of preprocessing, in addition to
5459 Produce a core dump whenever an error occurs.
5463 Print statistics on memory usage, at the end of the run, to
5468 Annotate the assembler output with a comment indicating which
5469 pattern and alternative was used. The length of each instruction is
5474 Dump the RTL in the assembler output as a comment before each instruction.
5475 Also turns on @option{-dp} annotation.
5479 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5480 dump a representation of the control flow graph suitable for viewing with VCG
5481 to @file{@var{file}.@var{pass}.vcg}.
5485 Just generate RTL for a function instead of compiling it. Usually used
5486 with @option{-fdump-rtl-expand}.
5490 @opindex fdump-noaddr
5491 When doing debugging dumps, suppress address output. This makes it more
5492 feasible to use diff on debugging dumps for compiler invocations with
5493 different compiler binaries and/or different
5494 text / bss / data / heap / stack / dso start locations.
5496 @item -fdump-unnumbered
5497 @opindex fdump-unnumbered
5498 When doing debugging dumps, suppress instruction numbers and address output.
5499 This makes it more feasible to use diff on debugging dumps for compiler
5500 invocations with different options, in particular with and without
5503 @item -fdump-unnumbered-links
5504 @opindex fdump-unnumbered-links
5505 When doing debugging dumps (see @option{-d} option above), suppress
5506 instruction numbers for the links to the previous and next instructions
5509 @item -fdump-translation-unit @r{(C++ only)}
5510 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5511 @opindex fdump-translation-unit
5512 Dump a representation of the tree structure for the entire translation
5513 unit to a file. The file name is made by appending @file{.tu} to the
5514 source file name, and the file is created in the same directory as the
5515 output file. If the @samp{-@var{options}} form is used, @var{options}
5516 controls the details of the dump as described for the
5517 @option{-fdump-tree} options.
5519 @item -fdump-class-hierarchy @r{(C++ only)}
5520 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5521 @opindex fdump-class-hierarchy
5522 Dump a representation of each class's hierarchy and virtual function
5523 table layout to a file. The file name is made by appending
5524 @file{.class} to the source file name, and the file is created in the
5525 same directory as the output file. If the @samp{-@var{options}} form
5526 is used, @var{options} controls the details of the dump as described
5527 for the @option{-fdump-tree} options.
5529 @item -fdump-ipa-@var{switch}
5531 Control the dumping at various stages of inter-procedural analysis
5532 language tree to a file. The file name is generated by appending a
5533 switch specific suffix to the source file name, and the file is created
5534 in the same directory as the output file. The following dumps are
5539 Enables all inter-procedural analysis dumps.
5542 Dumps information about call-graph optimization, unused function removal,
5543 and inlining decisions.
5546 Dump after function inlining.
5551 @opindex fdump-passes
5552 Dump the list of optimization passes that are turned on and off by
5553 the current command line options.
5555 @item -fdump-statistics-@var{option}
5556 @opindex fdump-statistics
5557 Enable and control dumping of pass statistics in a separate file. The
5558 file name is generated by appending a suffix ending in
5559 @samp{.statistics} to the source file name, and the file is created in
5560 the same directory as the output file. If the @samp{-@var{option}}
5561 form is used, @samp{-stats} will cause counters to be summed over the
5562 whole compilation unit while @samp{-details} will dump every event as
5563 the passes generate them. The default with no option is to sum
5564 counters for each function compiled.
5566 @item -fdump-tree-@var{switch}
5567 @itemx -fdump-tree-@var{switch}-@var{options}
5569 Control the dumping at various stages of processing the intermediate
5570 language tree to a file. The file name is generated by appending a
5571 switch specific suffix to the source file name, and the file is
5572 created in the same directory as the output file. If the
5573 @samp{-@var{options}} form is used, @var{options} is a list of
5574 @samp{-} separated options that control the details of the dump. Not
5575 all options are applicable to all dumps, those which are not
5576 meaningful will be ignored. The following options are available
5580 Print the address of each node. Usually this is not meaningful as it
5581 changes according to the environment and source file. Its primary use
5582 is for tying up a dump file with a debug environment.
5584 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5585 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5586 use working backward from mangled names in the assembly file.
5588 Inhibit dumping of members of a scope or body of a function merely
5589 because that scope has been reached. Only dump such items when they
5590 are directly reachable by some other path. When dumping pretty-printed
5591 trees, this option inhibits dumping the bodies of control structures.
5593 Print a raw representation of the tree. By default, trees are
5594 pretty-printed into a C-like representation.
5596 Enable more detailed dumps (not honored by every dump option).
5598 Enable dumping various statistics about the pass (not honored by every dump
5601 Enable showing basic block boundaries (disabled in raw dumps).
5603 Enable showing virtual operands for every statement.
5605 Enable showing line numbers for statements.
5607 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5609 Enable showing the tree dump for each statement.
5611 Enable showing the EH region number holding each statement.
5613 Enable showing scalar evolution analysis details.
5615 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5616 and @option{lineno}.
5619 The following tree dumps are possible:
5623 @opindex fdump-tree-original
5624 Dump before any tree based optimization, to @file{@var{file}.original}.
5627 @opindex fdump-tree-optimized
5628 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5631 @opindex fdump-tree-gimple
5632 Dump each function before and after the gimplification pass to a file. The
5633 file name is made by appending @file{.gimple} to the source file name.
5636 @opindex fdump-tree-cfg
5637 Dump the control flow graph of each function to a file. The file name is
5638 made by appending @file{.cfg} to the source file name.
5641 @opindex fdump-tree-vcg
5642 Dump the control flow graph of each function to a file in VCG format. The
5643 file name is made by appending @file{.vcg} to the source file name. Note
5644 that if the file contains more than one function, the generated file cannot
5645 be used directly by VCG@. You will need to cut and paste each function's
5646 graph into its own separate file first.
5649 @opindex fdump-tree-ch
5650 Dump each function after copying loop headers. The file name is made by
5651 appending @file{.ch} to the source file name.
5654 @opindex fdump-tree-ssa
5655 Dump SSA related information to a file. The file name is made by appending
5656 @file{.ssa} to the source file name.
5659 @opindex fdump-tree-alias
5660 Dump aliasing information for each function. The file name is made by
5661 appending @file{.alias} to the source file name.
5664 @opindex fdump-tree-ccp
5665 Dump each function after CCP@. The file name is made by appending
5666 @file{.ccp} to the source file name.
5669 @opindex fdump-tree-storeccp
5670 Dump each function after STORE-CCP@. The file name is made by appending
5671 @file{.storeccp} to the source file name.
5674 @opindex fdump-tree-pre
5675 Dump trees after partial redundancy elimination. The file name is made
5676 by appending @file{.pre} to the source file name.
5679 @opindex fdump-tree-fre
5680 Dump trees after full redundancy elimination. The file name is made
5681 by appending @file{.fre} to the source file name.
5684 @opindex fdump-tree-copyprop
5685 Dump trees after copy propagation. The file name is made
5686 by appending @file{.copyprop} to the source file name.
5688 @item store_copyprop
5689 @opindex fdump-tree-store_copyprop
5690 Dump trees after store copy-propagation. The file name is made
5691 by appending @file{.store_copyprop} to the source file name.
5694 @opindex fdump-tree-dce
5695 Dump each function after dead code elimination. The file name is made by
5696 appending @file{.dce} to the source file name.
5699 @opindex fdump-tree-mudflap
5700 Dump each function after adding mudflap instrumentation. The file name is
5701 made by appending @file{.mudflap} to the source file name.
5704 @opindex fdump-tree-sra
5705 Dump each function after performing scalar replacement of aggregates. The
5706 file name is made by appending @file{.sra} to the source file name.
5709 @opindex fdump-tree-sink
5710 Dump each function after performing code sinking. The file name is made
5711 by appending @file{.sink} to the source file name.
5714 @opindex fdump-tree-dom
5715 Dump each function after applying dominator tree optimizations. The file
5716 name is made by appending @file{.dom} to the source file name.
5719 @opindex fdump-tree-dse
5720 Dump each function after applying dead store elimination. The file
5721 name is made by appending @file{.dse} to the source file name.
5724 @opindex fdump-tree-phiopt
5725 Dump each function after optimizing PHI nodes into straightline code. The file
5726 name is made by appending @file{.phiopt} to the source file name.
5729 @opindex fdump-tree-forwprop
5730 Dump each function after forward propagating single use variables. The file
5731 name is made by appending @file{.forwprop} to the source file name.
5734 @opindex fdump-tree-copyrename
5735 Dump each function after applying the copy rename optimization. The file
5736 name is made by appending @file{.copyrename} to the source file name.
5739 @opindex fdump-tree-nrv
5740 Dump each function after applying the named return value optimization on
5741 generic trees. The file name is made by appending @file{.nrv} to the source
5745 @opindex fdump-tree-vect
5746 Dump each function after applying vectorization of loops. The file name is
5747 made by appending @file{.vect} to the source file name.
5750 @opindex fdump-tree-slp
5751 Dump each function after applying vectorization of basic blocks. The file name
5752 is made by appending @file{.slp} to the source file name.
5755 @opindex fdump-tree-vrp
5756 Dump each function after Value Range Propagation (VRP). The file name
5757 is made by appending @file{.vrp} to the source file name.
5760 @opindex fdump-tree-all
5761 Enable all the available tree dumps with the flags provided in this option.
5764 @item -ftree-vectorizer-verbose=@var{n}
5765 @opindex ftree-vectorizer-verbose
5766 This option controls the amount of debugging output the vectorizer prints.
5767 This information is written to standard error, unless
5768 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5769 in which case it is output to the usual dump listing file, @file{.vect}.
5770 For @var{n}=0 no diagnostic information is reported.
5771 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5772 and the total number of loops that got vectorized.
5773 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5774 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5775 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5776 level that @option{-fdump-tree-vect-stats} uses.
5777 Higher verbosity levels mean either more information dumped for each
5778 reported loop, or same amount of information reported for more loops:
5779 if @var{n}=3, vectorizer cost model information is reported.
5780 If @var{n}=4, alignment related information is added to the reports.
5781 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5782 memory access-patterns) is added to the reports.
5783 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5784 that did not pass the first analysis phase (i.e., may not be countable, or
5785 may have complicated control-flow).
5786 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5787 If @var{n}=8, SLP related information is added to the reports.
5788 For @var{n}=9, all the information the vectorizer generates during its
5789 analysis and transformation is reported. This is the same verbosity level
5790 that @option{-fdump-tree-vect-details} uses.
5792 @item -frandom-seed=@var{string}
5793 @opindex frandom-seed
5794 This option provides a seed that GCC uses when it would otherwise use
5795 random numbers. It is used to generate certain symbol names
5796 that have to be different in every compiled file. It is also used to
5797 place unique stamps in coverage data files and the object files that
5798 produce them. You can use the @option{-frandom-seed} option to produce
5799 reproducibly identical object files.
5801 The @var{string} should be different for every file you compile.
5803 @item -fsched-verbose=@var{n}
5804 @opindex fsched-verbose
5805 On targets that use instruction scheduling, this option controls the
5806 amount of debugging output the scheduler prints. This information is
5807 written to standard error, unless @option{-fdump-rtl-sched1} or
5808 @option{-fdump-rtl-sched2} is specified, in which case it is output
5809 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5810 respectively. However for @var{n} greater than nine, the output is
5811 always printed to standard error.
5813 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5814 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5815 For @var{n} greater than one, it also output basic block probabilities,
5816 detailed ready list information and unit/insn info. For @var{n} greater
5817 than two, it includes RTL at abort point, control-flow and regions info.
5818 And for @var{n} over four, @option{-fsched-verbose} also includes
5822 @itemx -save-temps=cwd
5824 Store the usual ``temporary'' intermediate files permanently; place them
5825 in the current directory and name them based on the source file. Thus,
5826 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5827 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5828 preprocessed @file{foo.i} output file even though the compiler now
5829 normally uses an integrated preprocessor.
5831 When used in combination with the @option{-x} command line option,
5832 @option{-save-temps} is sensible enough to avoid over writing an
5833 input source file with the same extension as an intermediate file.
5834 The corresponding intermediate file may be obtained by renaming the
5835 source file before using @option{-save-temps}.
5837 If you invoke GCC in parallel, compiling several different source
5838 files that share a common base name in different subdirectories or the
5839 same source file compiled for multiple output destinations, it is
5840 likely that the different parallel compilers will interfere with each
5841 other, and overwrite the temporary files. For instance:
5844 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5845 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5848 may result in @file{foo.i} and @file{foo.o} being written to
5849 simultaneously by both compilers.
5851 @item -save-temps=obj
5852 @opindex save-temps=obj
5853 Store the usual ``temporary'' intermediate files permanently. If the
5854 @option{-o} option is used, the temporary files are based on the
5855 object file. If the @option{-o} option is not used, the
5856 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5861 gcc -save-temps=obj -c foo.c
5862 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5863 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5866 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5867 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5868 @file{dir2/yfoobar.o}.
5870 @item -time@r{[}=@var{file}@r{]}
5872 Report the CPU time taken by each subprocess in the compilation
5873 sequence. For C source files, this is the compiler proper and assembler
5874 (plus the linker if linking is done).
5876 Without the specification of an output file, the output looks like this:
5883 The first number on each line is the ``user time'', that is time spent
5884 executing the program itself. The second number is ``system time'',
5885 time spent executing operating system routines on behalf of the program.
5886 Both numbers are in seconds.
5888 With the specification of an output file, the output is appended to the
5889 named file, and it looks like this:
5892 0.12 0.01 cc1 @var{options}
5893 0.00 0.01 as @var{options}
5896 The ``user time'' and the ``system time'' are moved before the program
5897 name, and the options passed to the program are displayed, so that one
5898 can later tell what file was being compiled, and with which options.
5900 @item -fvar-tracking
5901 @opindex fvar-tracking
5902 Run variable tracking pass. It computes where variables are stored at each
5903 position in code. Better debugging information is then generated
5904 (if the debugging information format supports this information).
5906 It is enabled by default when compiling with optimization (@option{-Os},
5907 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5908 the debug info format supports it.
5910 @item -fvar-tracking-assignments
5911 @opindex fvar-tracking-assignments
5912 @opindex fno-var-tracking-assignments
5913 Annotate assignments to user variables early in the compilation and
5914 attempt to carry the annotations over throughout the compilation all the
5915 way to the end, in an attempt to improve debug information while
5916 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5918 It can be enabled even if var-tracking is disabled, in which case
5919 annotations will be created and maintained, but discarded at the end.
5921 @item -fvar-tracking-assignments-toggle
5922 @opindex fvar-tracking-assignments-toggle
5923 @opindex fno-var-tracking-assignments-toggle
5924 Toggle @option{-fvar-tracking-assignments}, in the same way that
5925 @option{-gtoggle} toggles @option{-g}.
5927 @item -print-file-name=@var{library}
5928 @opindex print-file-name
5929 Print the full absolute name of the library file @var{library} that
5930 would be used when linking---and don't do anything else. With this
5931 option, GCC does not compile or link anything; it just prints the
5934 @item -print-multi-directory
5935 @opindex print-multi-directory
5936 Print the directory name corresponding to the multilib selected by any
5937 other switches present in the command line. This directory is supposed
5938 to exist in @env{GCC_EXEC_PREFIX}.
5940 @item -print-multi-lib
5941 @opindex print-multi-lib
5942 Print the mapping from multilib directory names to compiler switches
5943 that enable them. The directory name is separated from the switches by
5944 @samp{;}, and each switch starts with an @samp{@@} instead of the
5945 @samp{-}, without spaces between multiple switches. This is supposed to
5946 ease shell-processing.
5948 @item -print-multi-os-directory
5949 @opindex print-multi-os-directory
5950 Print the path to OS libraries for the selected
5951 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5952 present in the @file{lib} subdirectory and no multilibs are used, this is
5953 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5954 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5955 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5956 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5958 @item -print-prog-name=@var{program}
5959 @opindex print-prog-name
5960 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5962 @item -print-libgcc-file-name
5963 @opindex print-libgcc-file-name
5964 Same as @option{-print-file-name=libgcc.a}.
5966 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5967 but you do want to link with @file{libgcc.a}. You can do
5970 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5973 @item -print-search-dirs
5974 @opindex print-search-dirs
5975 Print the name of the configured installation directory and a list of
5976 program and library directories @command{gcc} will search---and don't do anything else.
5978 This is useful when @command{gcc} prints the error message
5979 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5980 To resolve this you either need to put @file{cpp0} and the other compiler
5981 components where @command{gcc} expects to find them, or you can set the environment
5982 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5983 Don't forget the trailing @samp{/}.
5984 @xref{Environment Variables}.
5986 @item -print-sysroot
5987 @opindex print-sysroot
5988 Print the target sysroot directory that will be used during
5989 compilation. This is the target sysroot specified either at configure
5990 time or using the @option{--sysroot} option, possibly with an extra
5991 suffix that depends on compilation options. If no target sysroot is
5992 specified, the option prints nothing.
5994 @item -print-sysroot-headers-suffix
5995 @opindex print-sysroot-headers-suffix
5996 Print the suffix added to the target sysroot when searching for
5997 headers, or give an error if the compiler is not configured with such
5998 a suffix---and don't do anything else.
6001 @opindex dumpmachine
6002 Print the compiler's target machine (for example,
6003 @samp{i686-pc-linux-gnu})---and don't do anything else.
6006 @opindex dumpversion
6007 Print the compiler version (for example, @samp{3.0})---and don't do
6012 Print the compiler's built-in specs---and don't do anything else. (This
6013 is used when GCC itself is being built.) @xref{Spec Files}.
6015 @item -feliminate-unused-debug-types
6016 @opindex feliminate-unused-debug-types
6017 Normally, when producing DWARF2 output, GCC will emit debugging
6018 information for all types declared in a compilation
6019 unit, regardless of whether or not they are actually used
6020 in that compilation unit. Sometimes this is useful, such as
6021 if, in the debugger, you want to cast a value to a type that is
6022 not actually used in your program (but is declared). More often,
6023 however, this results in a significant amount of wasted space.
6024 With this option, GCC will avoid producing debug symbol output
6025 for types that are nowhere used in the source file being compiled.
6028 @node Optimize Options
6029 @section Options That Control Optimization
6030 @cindex optimize options
6031 @cindex options, optimization
6033 These options control various sorts of optimizations.
6035 Without any optimization option, the compiler's goal is to reduce the
6036 cost of compilation and to make debugging produce the expected
6037 results. Statements are independent: if you stop the program with a
6038 breakpoint between statements, you can then assign a new value to any
6039 variable or change the program counter to any other statement in the
6040 function and get exactly the results you would expect from the source
6043 Turning on optimization flags makes the compiler attempt to improve
6044 the performance and/or code size at the expense of compilation time
6045 and possibly the ability to debug the program.
6047 The compiler performs optimization based on the knowledge it has of the
6048 program. Compiling multiple files at once to a single output file mode allows
6049 the compiler to use information gained from all of the files when compiling
6052 Not all optimizations are controlled directly by a flag. Only
6053 optimizations that have a flag are listed in this section.
6055 Most optimizations are only enabled if an @option{-O} level is set on
6056 the command line. Otherwise they are disabled, even if individual
6057 optimization flags are specified.
6059 Depending on the target and how GCC was configured, a slightly different
6060 set of optimizations may be enabled at each @option{-O} level than
6061 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6062 to find out the exact set of optimizations that are enabled at each level.
6063 @xref{Overall Options}, for examples.
6070 Optimize. Optimizing compilation takes somewhat more time, and a lot
6071 more memory for a large function.
6073 With @option{-O}, the compiler tries to reduce code size and execution
6074 time, without performing any optimizations that take a great deal of
6077 @option{-O} turns on the following optimization flags:
6081 -fcprop-registers @gol
6084 -fdelayed-branch @gol
6086 -fguess-branch-probability @gol
6087 -fif-conversion2 @gol
6088 -fif-conversion @gol
6089 -fipa-pure-const @gol
6091 -fipa-reference @gol
6093 -fsplit-wide-types @gol
6095 -ftree-builtin-call-dce @gol
6098 -ftree-copyrename @gol
6100 -ftree-dominator-opts @gol
6102 -ftree-forwprop @gol
6110 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6111 where doing so does not interfere with debugging.
6115 Optimize even more. GCC performs nearly all supported optimizations
6116 that do not involve a space-speed tradeoff.
6117 As compared to @option{-O}, this option increases both compilation time
6118 and the performance of the generated code.
6120 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6121 also turns on the following optimization flags:
6122 @gccoptlist{-fthread-jumps @gol
6123 -falign-functions -falign-jumps @gol
6124 -falign-loops -falign-labels @gol
6127 -fcse-follow-jumps -fcse-skip-blocks @gol
6128 -fdelete-null-pointer-checks @gol
6130 -fexpensive-optimizations @gol
6131 -fgcse -fgcse-lm @gol
6132 -finline-small-functions @gol
6133 -findirect-inlining @gol
6135 -foptimize-sibling-calls @gol
6136 -fpartial-inlining @gol
6139 -freorder-blocks -freorder-functions @gol
6140 -frerun-cse-after-loop @gol
6141 -fsched-interblock -fsched-spec @gol
6142 -fschedule-insns -fschedule-insns2 @gol
6143 -fstrict-aliasing -fstrict-overflow @gol
6144 -ftree-switch-conversion -ftree-tail-merge @gol
6148 Please note the warning under @option{-fgcse} about
6149 invoking @option{-O2} on programs that use computed gotos.
6153 Optimize yet more. @option{-O3} turns on all optimizations specified
6154 by @option{-O2} and also turns on the @option{-finline-functions},
6155 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6156 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6157 @option{-fipa-cp-clone} options.
6161 Reduce compilation time and make debugging produce the expected
6162 results. This is the default.
6166 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6167 do not typically increase code size. It also performs further
6168 optimizations designed to reduce code size.
6170 @option{-Os} disables the following optimization flags:
6171 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6172 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6173 -fprefetch-loop-arrays -ftree-vect-loop-version}
6177 Disregard strict standards compliance. @option{-Ofast} enables all
6178 @option{-O3} optimizations. It also enables optimizations that are not
6179 valid for all standard compliant programs.
6180 It turns on @option{-ffast-math} and the Fortran-specific
6181 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6183 If you use multiple @option{-O} options, with or without level numbers,
6184 the last such option is the one that is effective.
6187 Options of the form @option{-f@var{flag}} specify machine-independent
6188 flags. Most flags have both positive and negative forms; the negative
6189 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6190 below, only one of the forms is listed---the one you typically will
6191 use. You can figure out the other form by either removing @samp{no-}
6194 The following options control specific optimizations. They are either
6195 activated by @option{-O} options or are related to ones that are. You
6196 can use the following flags in the rare cases when ``fine-tuning'' of
6197 optimizations to be performed is desired.
6200 @item -fno-default-inline
6201 @opindex fno-default-inline
6202 Do not make member functions inline by default merely because they are
6203 defined inside the class scope (C++ only). Otherwise, when you specify
6204 @w{@option{-O}}, member functions defined inside class scope are compiled
6205 inline by default; i.e., you don't need to add @samp{inline} in front of
6206 the member function name.
6208 @item -fno-defer-pop
6209 @opindex fno-defer-pop
6210 Always pop the arguments to each function call as soon as that function
6211 returns. For machines which must pop arguments after a function call,
6212 the compiler normally lets arguments accumulate on the stack for several
6213 function calls and pops them all at once.
6215 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6217 @item -fforward-propagate
6218 @opindex fforward-propagate
6219 Perform a forward propagation pass on RTL@. The pass tries to combine two
6220 instructions and checks if the result can be simplified. If loop unrolling
6221 is active, two passes are performed and the second is scheduled after
6224 This option is enabled by default at optimization levels @option{-O},
6225 @option{-O2}, @option{-O3}, @option{-Os}.
6227 @item -ffp-contract=@var{style}
6228 @opindex ffp-contract
6229 @option{-ffp-contract=off} disables floating-point expression contraction.
6230 @option{-ffp-contract=fast} enables floating-point expression contraction
6231 such as forming of fused multiply-add operations if the target has
6232 native support for them.
6233 @option{-ffp-contract=on} enables floating-point expression contraction
6234 if allowed by the language standard. This is currently not implemented
6235 and treated equal to @option{-ffp-contract=off}.
6237 The default is @option{-ffp-contract=fast}.
6239 @item -fomit-frame-pointer
6240 @opindex fomit-frame-pointer
6241 Don't keep the frame pointer in a register for functions that
6242 don't need one. This avoids the instructions to save, set up and
6243 restore frame pointers; it also makes an extra register available
6244 in many functions. @strong{It also makes debugging impossible on
6247 On some machines, such as the VAX, this flag has no effect, because
6248 the standard calling sequence automatically handles the frame pointer
6249 and nothing is saved by pretending it doesn't exist. The
6250 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6251 whether a target machine supports this flag. @xref{Registers,,Register
6252 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6254 Starting with GCC version 4.6, the default setting (when not optimizing for
6255 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6256 @option{-fomit-frame-pointer}. The default can be reverted to
6257 @option{-fno-omit-frame-pointer} by configuring GCC with the
6258 @option{--enable-frame-pointer} configure option.
6260 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6262 @item -foptimize-sibling-calls
6263 @opindex foptimize-sibling-calls
6264 Optimize sibling and tail recursive calls.
6266 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6270 Don't pay attention to the @code{inline} keyword. Normally this option
6271 is used to keep the compiler from expanding any functions inline.
6272 Note that if you are not optimizing, no functions can be expanded inline.
6274 @item -finline-small-functions
6275 @opindex finline-small-functions
6276 Integrate functions into their callers when their body is smaller than expected
6277 function call code (so overall size of program gets smaller). The compiler
6278 heuristically decides which functions are simple enough to be worth integrating
6281 Enabled at level @option{-O2}.
6283 @item -findirect-inlining
6284 @opindex findirect-inlining
6285 Inline also indirect calls that are discovered to be known at compile
6286 time thanks to previous inlining. This option has any effect only
6287 when inlining itself is turned on by the @option{-finline-functions}
6288 or @option{-finline-small-functions} options.
6290 Enabled at level @option{-O2}.
6292 @item -finline-functions
6293 @opindex finline-functions
6294 Integrate all simple functions into their callers. The compiler
6295 heuristically decides which functions are simple enough to be worth
6296 integrating in this way.
6298 If all calls to a given function are integrated, and the function is
6299 declared @code{static}, then the function is normally not output as
6300 assembler code in its own right.
6302 Enabled at level @option{-O3}.
6304 @item -finline-functions-called-once
6305 @opindex finline-functions-called-once
6306 Consider all @code{static} functions called once for inlining into their
6307 caller even if they are not marked @code{inline}. If a call to a given
6308 function is integrated, then the function is not output as assembler code
6311 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6313 @item -fearly-inlining
6314 @opindex fearly-inlining
6315 Inline functions marked by @code{always_inline} and functions whose body seems
6316 smaller than the function call overhead early before doing
6317 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6318 makes profiling significantly cheaper and usually inlining faster on programs
6319 having large chains of nested wrapper functions.
6325 Perform interprocedural scalar replacement of aggregates, removal of
6326 unused parameters and replacement of parameters passed by reference
6327 by parameters passed by value.
6329 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6331 @item -finline-limit=@var{n}
6332 @opindex finline-limit
6333 By default, GCC limits the size of functions that can be inlined. This flag
6334 allows coarse control of this limit. @var{n} is the size of functions that
6335 can be inlined in number of pseudo instructions.
6337 Inlining is actually controlled by a number of parameters, which may be
6338 specified individually by using @option{--param @var{name}=@var{value}}.
6339 The @option{-finline-limit=@var{n}} option sets some of these parameters
6343 @item max-inline-insns-single
6344 is set to @var{n}/2.
6345 @item max-inline-insns-auto
6346 is set to @var{n}/2.
6349 See below for a documentation of the individual
6350 parameters controlling inlining and for the defaults of these parameters.
6352 @emph{Note:} there may be no value to @option{-finline-limit} that results
6353 in default behavior.
6355 @emph{Note:} pseudo instruction represents, in this particular context, an
6356 abstract measurement of function's size. In no way does it represent a count
6357 of assembly instructions and as such its exact meaning might change from one
6358 release to an another.
6360 @item -fno-keep-inline-dllexport
6361 @opindex -fno-keep-inline-dllexport
6362 This is a more fine-grained version of @option{-fkeep-inline-functions},
6363 which applies only to functions that are declared using the @code{dllexport}
6364 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6367 @item -fkeep-inline-functions
6368 @opindex fkeep-inline-functions
6369 In C, emit @code{static} functions that are declared @code{inline}
6370 into the object file, even if the function has been inlined into all
6371 of its callers. This switch does not affect functions using the
6372 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6373 inline functions into the object file.
6375 @item -fkeep-static-consts
6376 @opindex fkeep-static-consts
6377 Emit variables declared @code{static const} when optimization isn't turned
6378 on, even if the variables aren't referenced.
6380 GCC enables this option by default. If you want to force the compiler to
6381 check if the variable was referenced, regardless of whether or not
6382 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6384 @item -fmerge-constants
6385 @opindex fmerge-constants
6386 Attempt to merge identical constants (string constants and floating point
6387 constants) across compilation units.
6389 This option is the default for optimized compilation if the assembler and
6390 linker support it. Use @option{-fno-merge-constants} to inhibit this
6393 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6395 @item -fmerge-all-constants
6396 @opindex fmerge-all-constants
6397 Attempt to merge identical constants and identical variables.
6399 This option implies @option{-fmerge-constants}. In addition to
6400 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6401 arrays or initialized constant variables with integral or floating point
6402 types. Languages like C or C++ require each variable, including multiple
6403 instances of the same variable in recursive calls, to have distinct locations,
6404 so using this option will result in non-conforming
6407 @item -fmodulo-sched
6408 @opindex fmodulo-sched
6409 Perform swing modulo scheduling immediately before the first scheduling
6410 pass. This pass looks at innermost loops and reorders their
6411 instructions by overlapping different iterations.
6413 @item -fmodulo-sched-allow-regmoves
6414 @opindex fmodulo-sched-allow-regmoves
6415 Perform more aggressive SMS based modulo scheduling with register moves
6416 allowed. By setting this flag certain anti-dependences edges will be
6417 deleted which will trigger the generation of reg-moves based on the
6418 life-range analysis. This option is effective only with
6419 @option{-fmodulo-sched} enabled.
6421 @item -fno-branch-count-reg
6422 @opindex fno-branch-count-reg
6423 Do not use ``decrement and branch'' instructions on a count register,
6424 but instead generate a sequence of instructions that decrement a
6425 register, compare it against zero, then branch based upon the result.
6426 This option is only meaningful on architectures that support such
6427 instructions, which include x86, PowerPC, IA-64 and S/390.
6429 The default is @option{-fbranch-count-reg}.
6431 @item -fno-function-cse
6432 @opindex fno-function-cse
6433 Do not put function addresses in registers; make each instruction that
6434 calls a constant function contain the function's address explicitly.
6436 This option results in less efficient code, but some strange hacks
6437 that alter the assembler output may be confused by the optimizations
6438 performed when this option is not used.
6440 The default is @option{-ffunction-cse}
6442 @item -fno-zero-initialized-in-bss
6443 @opindex fno-zero-initialized-in-bss
6444 If the target supports a BSS section, GCC by default puts variables that
6445 are initialized to zero into BSS@. This can save space in the resulting
6448 This option turns off this behavior because some programs explicitly
6449 rely on variables going to the data section. E.g., so that the
6450 resulting executable can find the beginning of that section and/or make
6451 assumptions based on that.
6453 The default is @option{-fzero-initialized-in-bss}.
6455 @item -fmudflap -fmudflapth -fmudflapir
6459 @cindex bounds checking
6461 For front-ends that support it (C and C++), instrument all risky
6462 pointer/array dereferencing operations, some standard library
6463 string/heap functions, and some other associated constructs with
6464 range/validity tests. Modules so instrumented should be immune to
6465 buffer overflows, invalid heap use, and some other classes of C/C++
6466 programming errors. The instrumentation relies on a separate runtime
6467 library (@file{libmudflap}), which will be linked into a program if
6468 @option{-fmudflap} is given at link time. Run-time behavior of the
6469 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6470 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6473 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6474 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6475 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6476 instrumentation should ignore pointer reads. This produces less
6477 instrumentation (and therefore faster execution) and still provides
6478 some protection against outright memory corrupting writes, but allows
6479 erroneously read data to propagate within a program.
6481 @item -fthread-jumps
6482 @opindex fthread-jumps
6483 Perform optimizations where we check to see if a jump branches to a
6484 location where another comparison subsumed by the first is found. If
6485 so, the first branch is redirected to either the destination of the
6486 second branch or a point immediately following it, depending on whether
6487 the condition is known to be true or false.
6489 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6491 @item -fsplit-wide-types
6492 @opindex fsplit-wide-types
6493 When using a type that occupies multiple registers, such as @code{long
6494 long} on a 32-bit system, split the registers apart and allocate them
6495 independently. This normally generates better code for those types,
6496 but may make debugging more difficult.
6498 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6501 @item -fcse-follow-jumps
6502 @opindex fcse-follow-jumps
6503 In common subexpression elimination (CSE), scan through jump instructions
6504 when the target of the jump is not reached by any other path. For
6505 example, when CSE encounters an @code{if} statement with an
6506 @code{else} clause, CSE will follow the jump when the condition
6509 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6511 @item -fcse-skip-blocks
6512 @opindex fcse-skip-blocks
6513 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6514 follow jumps which conditionally skip over blocks. When CSE
6515 encounters a simple @code{if} statement with no else clause,
6516 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6517 body of the @code{if}.
6519 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6521 @item -frerun-cse-after-loop
6522 @opindex frerun-cse-after-loop
6523 Re-run common subexpression elimination after loop optimizations has been
6526 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6530 Perform a global common subexpression elimination pass.
6531 This pass also performs global constant and copy propagation.
6533 @emph{Note:} When compiling a program using computed gotos, a GCC
6534 extension, you may get better runtime performance if you disable
6535 the global common subexpression elimination pass by adding
6536 @option{-fno-gcse} to the command line.
6538 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6542 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6543 attempt to move loads which are only killed by stores into themselves. This
6544 allows a loop containing a load/store sequence to be changed to a load outside
6545 the loop, and a copy/store within the loop.
6547 Enabled by default when gcse is enabled.
6551 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6552 global common subexpression elimination. This pass will attempt to move
6553 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6554 loops containing a load/store sequence can be changed to a load before
6555 the loop and a store after the loop.
6557 Not enabled at any optimization level.
6561 When @option{-fgcse-las} is enabled, the global common subexpression
6562 elimination pass eliminates redundant loads that come after stores to the
6563 same memory location (both partial and full redundancies).
6565 Not enabled at any optimization level.
6567 @item -fgcse-after-reload
6568 @opindex fgcse-after-reload
6569 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6570 pass is performed after reload. The purpose of this pass is to cleanup
6573 @item -funsafe-loop-optimizations
6574 @opindex funsafe-loop-optimizations
6575 If given, the loop optimizer will assume that loop indices do not
6576 overflow, and that the loops with nontrivial exit condition are not
6577 infinite. This enables a wider range of loop optimizations even if
6578 the loop optimizer itself cannot prove that these assumptions are valid.
6579 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6580 if it finds this kind of loop.
6582 @item -fcrossjumping
6583 @opindex fcrossjumping
6584 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6585 resulting code may or may not perform better than without cross-jumping.
6587 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6589 @item -fauto-inc-dec
6590 @opindex fauto-inc-dec
6591 Combine increments or decrements of addresses with memory accesses.
6592 This pass is always skipped on architectures that do not have
6593 instructions to support this. Enabled by default at @option{-O} and
6594 higher on architectures that support this.
6598 Perform dead code elimination (DCE) on RTL@.
6599 Enabled by default at @option{-O} and higher.
6603 Perform dead store elimination (DSE) on RTL@.
6604 Enabled by default at @option{-O} and higher.
6606 @item -fif-conversion
6607 @opindex fif-conversion
6608 Attempt to transform conditional jumps into branch-less equivalents. This
6609 include use of conditional moves, min, max, set flags and abs instructions, and
6610 some tricks doable by standard arithmetics. The use of conditional execution
6611 on chips where it is available is controlled by @code{if-conversion2}.
6613 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6615 @item -fif-conversion2
6616 @opindex fif-conversion2
6617 Use conditional execution (where available) to transform conditional jumps into
6618 branch-less equivalents.
6620 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6622 @item -fdelete-null-pointer-checks
6623 @opindex fdelete-null-pointer-checks
6624 Assume that programs cannot safely dereference null pointers, and that
6625 no code or data element resides there. This enables simple constant
6626 folding optimizations at all optimization levels. In addition, other
6627 optimization passes in GCC use this flag to control global dataflow
6628 analyses that eliminate useless checks for null pointers; these assume
6629 that if a pointer is checked after it has already been dereferenced,
6632 Note however that in some environments this assumption is not true.
6633 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6634 for programs which depend on that behavior.
6636 Some targets, especially embedded ones, disable this option at all levels.
6637 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6638 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6639 are enabled independently at different optimization levels.
6641 @item -fdevirtualize
6642 @opindex fdevirtualize
6643 Attempt to convert calls to virtual functions to direct calls. This
6644 is done both within a procedure and interprocedurally as part of
6645 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6646 propagation (@option{-fipa-cp}).
6647 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6649 @item -fexpensive-optimizations
6650 @opindex fexpensive-optimizations
6651 Perform a number of minor optimizations that are relatively expensive.
6653 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6655 @item -foptimize-register-move
6657 @opindex foptimize-register-move
6659 Attempt to reassign register numbers in move instructions and as
6660 operands of other simple instructions in order to maximize the amount of
6661 register tying. This is especially helpful on machines with two-operand
6664 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6667 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6669 @item -fira-algorithm=@var{algorithm}
6670 Use specified coloring algorithm for the integrated register
6671 allocator. The @var{algorithm} argument should be @code{priority} or
6672 @code{CB}. The first algorithm specifies Chow's priority coloring,
6673 the second one specifies Chaitin-Briggs coloring. The second
6674 algorithm can be unimplemented for some architectures. If it is
6675 implemented, it is the default because Chaitin-Briggs coloring as a
6676 rule generates a better code.
6678 @item -fira-region=@var{region}
6679 Use specified regions for the integrated register allocator. The
6680 @var{region} argument should be one of @code{all}, @code{mixed}, or
6681 @code{one}. The first value means using all loops as register
6682 allocation regions, the second value which is the default means using
6683 all loops except for loops with small register pressure as the
6684 regions, and third one means using all function as a single region.
6685 The first value can give best result for machines with small size and
6686 irregular register set, the third one results in faster and generates
6687 decent code and the smallest size code, and the default value usually
6688 give the best results in most cases and for most architectures.
6690 @item -fira-loop-pressure
6691 @opindex fira-loop-pressure
6692 Use IRA to evaluate register pressure in loops for decision to move
6693 loop invariants. Usage of this option usually results in generation
6694 of faster and smaller code on machines with big register files (>= 32
6695 registers) but it can slow compiler down.
6697 This option is enabled at level @option{-O3} for some targets.
6699 @item -fno-ira-share-save-slots
6700 @opindex fno-ira-share-save-slots
6701 Switch off sharing stack slots used for saving call used hard
6702 registers living through a call. Each hard register will get a
6703 separate stack slot and as a result function stack frame will be
6706 @item -fno-ira-share-spill-slots
6707 @opindex fno-ira-share-spill-slots
6708 Switch off sharing stack slots allocated for pseudo-registers. Each
6709 pseudo-register which did not get a hard register will get a separate
6710 stack slot and as a result function stack frame will be bigger.
6712 @item -fira-verbose=@var{n}
6713 @opindex fira-verbose
6714 Set up how verbose dump file for the integrated register allocator
6715 will be. Default value is 5. If the value is greater or equal to 10,
6716 the dump file will be stderr as if the value were @var{n} minus 10.
6718 @item -fdelayed-branch
6719 @opindex fdelayed-branch
6720 If supported for the target machine, attempt to reorder instructions
6721 to exploit instruction slots available after delayed branch
6724 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6726 @item -fschedule-insns
6727 @opindex fschedule-insns
6728 If supported for the target machine, attempt to reorder instructions to
6729 eliminate execution stalls due to required data being unavailable. This
6730 helps machines that have slow floating point or memory load instructions
6731 by allowing other instructions to be issued until the result of the load
6732 or floating point instruction is required.
6734 Enabled at levels @option{-O2}, @option{-O3}.
6736 @item -fschedule-insns2
6737 @opindex fschedule-insns2
6738 Similar to @option{-fschedule-insns}, but requests an additional pass of
6739 instruction scheduling after register allocation has been done. This is
6740 especially useful on machines with a relatively small number of
6741 registers and where memory load instructions take more than one cycle.
6743 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6745 @item -fno-sched-interblock
6746 @opindex fno-sched-interblock
6747 Don't schedule instructions across basic blocks. This is normally
6748 enabled by default when scheduling before register allocation, i.e.@:
6749 with @option{-fschedule-insns} or at @option{-O2} or higher.
6751 @item -fno-sched-spec
6752 @opindex fno-sched-spec
6753 Don't allow speculative motion of non-load instructions. This is normally
6754 enabled by default when scheduling before register allocation, i.e.@:
6755 with @option{-fschedule-insns} or at @option{-O2} or higher.
6757 @item -fsched-pressure
6758 @opindex fsched-pressure
6759 Enable register pressure sensitive insn scheduling before the register
6760 allocation. This only makes sense when scheduling before register
6761 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6762 @option{-O2} or higher. Usage of this option can improve the
6763 generated code and decrease its size by preventing register pressure
6764 increase above the number of available hard registers and as a
6765 consequence register spills in the register allocation.
6767 @item -fsched-spec-load
6768 @opindex fsched-spec-load
6769 Allow speculative motion of some load instructions. This only makes
6770 sense when scheduling before register allocation, i.e.@: with
6771 @option{-fschedule-insns} or at @option{-O2} or higher.
6773 @item -fsched-spec-load-dangerous
6774 @opindex fsched-spec-load-dangerous
6775 Allow speculative motion of more load instructions. This only makes
6776 sense when scheduling before register allocation, i.e.@: with
6777 @option{-fschedule-insns} or at @option{-O2} or higher.
6779 @item -fsched-stalled-insns
6780 @itemx -fsched-stalled-insns=@var{n}
6781 @opindex fsched-stalled-insns
6782 Define how many insns (if any) can be moved prematurely from the queue
6783 of stalled insns into the ready list, during the second scheduling pass.
6784 @option{-fno-sched-stalled-insns} means that no insns will be moved
6785 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6786 on how many queued insns can be moved prematurely.
6787 @option{-fsched-stalled-insns} without a value is equivalent to
6788 @option{-fsched-stalled-insns=1}.
6790 @item -fsched-stalled-insns-dep
6791 @itemx -fsched-stalled-insns-dep=@var{n}
6792 @opindex fsched-stalled-insns-dep
6793 Define how many insn groups (cycles) will be examined for a dependency
6794 on a stalled insn that is candidate for premature removal from the queue
6795 of stalled insns. This has an effect only during the second scheduling pass,
6796 and only if @option{-fsched-stalled-insns} is used.
6797 @option{-fno-sched-stalled-insns-dep} is equivalent to
6798 @option{-fsched-stalled-insns-dep=0}.
6799 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6800 @option{-fsched-stalled-insns-dep=1}.
6802 @item -fsched2-use-superblocks
6803 @opindex fsched2-use-superblocks
6804 When scheduling after register allocation, do use superblock scheduling
6805 algorithm. Superblock scheduling allows motion across basic block boundaries
6806 resulting on faster schedules. This option is experimental, as not all machine
6807 descriptions used by GCC model the CPU closely enough to avoid unreliable
6808 results from the algorithm.
6810 This only makes sense when scheduling after register allocation, i.e.@: with
6811 @option{-fschedule-insns2} or at @option{-O2} or higher.
6813 @item -fsched-group-heuristic
6814 @opindex fsched-group-heuristic
6815 Enable the group heuristic in the scheduler. This heuristic favors
6816 the instruction that belongs to a schedule group. This is enabled
6817 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6818 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6820 @item -fsched-critical-path-heuristic
6821 @opindex fsched-critical-path-heuristic
6822 Enable the critical-path heuristic in the scheduler. This heuristic favors
6823 instructions on the critical path. This is enabled by default when
6824 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6825 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6827 @item -fsched-spec-insn-heuristic
6828 @opindex fsched-spec-insn-heuristic
6829 Enable the speculative instruction heuristic in the scheduler. This
6830 heuristic favors speculative instructions with greater dependency weakness.
6831 This is enabled by default when scheduling is enabled, i.e.@:
6832 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6833 or at @option{-O2} or higher.
6835 @item -fsched-rank-heuristic
6836 @opindex fsched-rank-heuristic
6837 Enable the rank heuristic in the scheduler. This heuristic favors
6838 the instruction belonging to a basic block with greater size or frequency.
6839 This is enabled by default when scheduling is enabled, i.e.@:
6840 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6841 at @option{-O2} or higher.
6843 @item -fsched-last-insn-heuristic
6844 @opindex fsched-last-insn-heuristic
6845 Enable the last-instruction heuristic in the scheduler. This heuristic
6846 favors the instruction that is less dependent on the last instruction
6847 scheduled. This is enabled by default when scheduling is enabled,
6848 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6849 at @option{-O2} or higher.
6851 @item -fsched-dep-count-heuristic
6852 @opindex fsched-dep-count-heuristic
6853 Enable the dependent-count heuristic in the scheduler. This heuristic
6854 favors the instruction that has more instructions depending on it.
6855 This is enabled by default when scheduling is enabled, i.e.@:
6856 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6857 at @option{-O2} or higher.
6859 @item -freschedule-modulo-scheduled-loops
6860 @opindex freschedule-modulo-scheduled-loops
6861 The modulo scheduling comes before the traditional scheduling, if a loop
6862 was modulo scheduled we may want to prevent the later scheduling passes
6863 from changing its schedule, we use this option to control that.
6865 @item -fselective-scheduling
6866 @opindex fselective-scheduling
6867 Schedule instructions using selective scheduling algorithm. Selective
6868 scheduling runs instead of the first scheduler pass.
6870 @item -fselective-scheduling2
6871 @opindex fselective-scheduling2
6872 Schedule instructions using selective scheduling algorithm. Selective
6873 scheduling runs instead of the second scheduler pass.
6875 @item -fsel-sched-pipelining
6876 @opindex fsel-sched-pipelining
6877 Enable software pipelining of innermost loops during selective scheduling.
6878 This option has no effect until one of @option{-fselective-scheduling} or
6879 @option{-fselective-scheduling2} is turned on.
6881 @item -fsel-sched-pipelining-outer-loops
6882 @opindex fsel-sched-pipelining-outer-loops
6883 When pipelining loops during selective scheduling, also pipeline outer loops.
6884 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6887 @opindex fshrink-wrap
6888 Emit function prologues only before parts of the function that need it,
6889 rather than at the top of the function. This flag is enabled by default at
6890 @option{-O} and higher.
6892 @item -fcaller-saves
6893 @opindex fcaller-saves
6894 Enable values to be allocated in registers that will be clobbered by
6895 function calls, by emitting extra instructions to save and restore the
6896 registers around such calls. Such allocation is done only when it
6897 seems to result in better code than would otherwise be produced.
6899 This option is always enabled by default on certain machines, usually
6900 those which have no call-preserved registers to use instead.
6902 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6904 @item -fcombine-stack-adjustments
6905 @opindex fcombine-stack-adjustments
6906 Tracks stack adjustments (pushes and pops) and stack memory references
6907 and then tries to find ways to combine them.
6909 Enabled by default at @option{-O1} and higher.
6911 @item -fconserve-stack
6912 @opindex fconserve-stack
6913 Attempt to minimize stack usage. The compiler will attempt to use less
6914 stack space, even if that makes the program slower. This option
6915 implies setting the @option{large-stack-frame} parameter to 100
6916 and the @option{large-stack-frame-growth} parameter to 400.
6918 @item -ftree-reassoc
6919 @opindex ftree-reassoc
6920 Perform reassociation on trees. This flag is enabled by default
6921 at @option{-O} and higher.
6925 Perform partial redundancy elimination (PRE) on trees. This flag is
6926 enabled by default at @option{-O2} and @option{-O3}.
6928 @item -ftree-forwprop
6929 @opindex ftree-forwprop
6930 Perform forward propagation on trees. This flag is enabled by default
6931 at @option{-O} and higher.
6935 Perform full redundancy elimination (FRE) on trees. The difference
6936 between FRE and PRE is that FRE only considers expressions
6937 that are computed on all paths leading to the redundant computation.
6938 This analysis is faster than PRE, though it exposes fewer redundancies.
6939 This flag is enabled by default at @option{-O} and higher.
6941 @item -ftree-phiprop
6942 @opindex ftree-phiprop
6943 Perform hoisting of loads from conditional pointers on trees. This
6944 pass is enabled by default at @option{-O} and higher.
6946 @item -ftree-copy-prop
6947 @opindex ftree-copy-prop
6948 Perform copy propagation on trees. This pass eliminates unnecessary
6949 copy operations. This flag is enabled by default at @option{-O} and
6952 @item -fipa-pure-const
6953 @opindex fipa-pure-const
6954 Discover which functions are pure or constant.
6955 Enabled by default at @option{-O} and higher.
6957 @item -fipa-reference
6958 @opindex fipa-reference
6959 Discover which static variables do not escape cannot escape the
6961 Enabled by default at @option{-O} and higher.
6965 Perform interprocedural pointer analysis and interprocedural modification
6966 and reference analysis. This option can cause excessive memory and
6967 compile-time usage on large compilation units. It is not enabled by
6968 default at any optimization level.
6971 @opindex fipa-profile
6972 Perform interprocedural profile propagation. The functions called only from
6973 cold functions are marked as cold. Also functions executed once (such as
6974 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6975 functions and loop less parts of functions executed once are then optimized for
6977 Enabled by default at @option{-O} and higher.
6981 Perform interprocedural constant propagation.
6982 This optimization analyzes the program to determine when values passed
6983 to functions are constants and then optimizes accordingly.
6984 This optimization can substantially increase performance
6985 if the application has constants passed to functions.
6986 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6988 @item -fipa-cp-clone
6989 @opindex fipa-cp-clone
6990 Perform function cloning to make interprocedural constant propagation stronger.
6991 When enabled, interprocedural constant propagation will perform function cloning
6992 when externally visible function can be called with constant arguments.
6993 Because this optimization can create multiple copies of functions,
6994 it may significantly increase code size
6995 (see @option{--param ipcp-unit-growth=@var{value}}).
6996 This flag is enabled by default at @option{-O3}.
6998 @item -fipa-matrix-reorg
6999 @opindex fipa-matrix-reorg
7000 Perform matrix flattening and transposing.
7001 Matrix flattening tries to replace an @math{m}-dimensional matrix
7002 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
7003 This reduces the level of indirection needed for accessing the elements
7004 of the matrix. The second optimization is matrix transposing that
7005 attempts to change the order of the matrix's dimensions in order to
7006 improve cache locality.
7007 Both optimizations need the @option{-fwhole-program} flag.
7008 Transposing is enabled only if profiling information is available.
7012 Perform forward store motion on trees. This flag is
7013 enabled by default at @option{-O} and higher.
7015 @item -ftree-bit-ccp
7016 @opindex ftree-bit-ccp
7017 Perform sparse conditional bit constant propagation on trees and propagate
7018 pointer alignment information.
7019 This pass only operates on local scalar variables and is enabled by default
7020 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7024 Perform sparse conditional constant propagation (CCP) on trees. This
7025 pass only operates on local scalar variables and is enabled by default
7026 at @option{-O} and higher.
7028 @item -ftree-switch-conversion
7029 Perform conversion of simple initializations in a switch to
7030 initializations from a scalar array. This flag is enabled by default
7031 at @option{-O2} and higher.
7033 @item -ftree-tail-merge
7034 Look for identical code sequences. When found, replace one with a jump to the
7035 other. This optimization is known as tail merging or cross jumping. This flag
7036 is enabled by default at @option{-O2} and higher. The run time of this pass can
7037 be limited using @option{max-tail-merge-comparisons} parameter and
7038 @option{max-tail-merge-iterations} parameter.
7042 Perform dead code elimination (DCE) on trees. This flag is enabled by
7043 default at @option{-O} and higher.
7045 @item -ftree-builtin-call-dce
7046 @opindex ftree-builtin-call-dce
7047 Perform conditional dead code elimination (DCE) for calls to builtin functions
7048 that may set @code{errno} but are otherwise side-effect free. This flag is
7049 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7052 @item -ftree-dominator-opts
7053 @opindex ftree-dominator-opts
7054 Perform a variety of simple scalar cleanups (constant/copy
7055 propagation, redundancy elimination, range propagation and expression
7056 simplification) based on a dominator tree traversal. This also
7057 performs jump threading (to reduce jumps to jumps). This flag is
7058 enabled by default at @option{-O} and higher.
7062 Perform dead store elimination (DSE) on trees. A dead store is a store into
7063 a memory location which will later be overwritten by another store without
7064 any intervening loads. In this case the earlier store can be deleted. This
7065 flag is enabled by default at @option{-O} and higher.
7069 Perform loop header copying on trees. This is beneficial since it increases
7070 effectiveness of code motion optimizations. It also saves one jump. This flag
7071 is enabled by default at @option{-O} and higher. It is not enabled
7072 for @option{-Os}, since it usually increases code size.
7074 @item -ftree-loop-optimize
7075 @opindex ftree-loop-optimize
7076 Perform loop optimizations on trees. This flag is enabled by default
7077 at @option{-O} and higher.
7079 @item -ftree-loop-linear
7080 @opindex ftree-loop-linear
7081 Perform loop interchange transformations on tree. Same as
7082 @option{-floop-interchange}. To use this code transformation, GCC has
7083 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7084 enable the Graphite loop transformation infrastructure.
7086 @item -floop-interchange
7087 @opindex floop-interchange
7088 Perform loop interchange transformations on loops. Interchanging two
7089 nested loops switches the inner and outer loops. For example, given a
7094 A(J, I) = A(J, I) * C
7098 loop interchange will transform the loop as if the user had written:
7102 A(J, I) = A(J, I) * C
7106 which can be beneficial when @code{N} is larger than the caches,
7107 because in Fortran, the elements of an array are stored in memory
7108 contiguously by column, and the original loop iterates over rows,
7109 potentially creating at each access a cache miss. This optimization
7110 applies to all the languages supported by GCC and is not limited to
7111 Fortran. To use this code transformation, GCC has to be configured
7112 with @option{--with-ppl} and @option{--with-cloog} to enable the
7113 Graphite loop transformation infrastructure.
7115 @item -floop-strip-mine
7116 @opindex floop-strip-mine
7117 Perform loop strip mining transformations on loops. Strip mining
7118 splits a loop into two nested loops. The outer loop has strides
7119 equal to the strip size and the inner loop has strides of the
7120 original loop within a strip. The strip length can be changed
7121 using the @option{loop-block-tile-size} parameter. For example,
7128 loop strip mining will transform the loop as if the user had written:
7131 DO I = II, min (II + 50, N)
7136 This optimization applies to all the languages supported by GCC and is
7137 not limited to Fortran. To use this code transformation, GCC has to
7138 be configured with @option{--with-ppl} and @option{--with-cloog} to
7139 enable the Graphite loop transformation infrastructure.
7142 @opindex floop-block
7143 Perform loop blocking transformations on loops. Blocking strip mines
7144 each loop in the loop nest such that the memory accesses of the
7145 element loops fit inside caches. The strip length can be changed
7146 using the @option{loop-block-tile-size} parameter. For example, given
7151 A(J, I) = B(I) + C(J)
7155 loop blocking will transform the loop as if the user had written:
7159 DO I = II, min (II + 50, N)
7160 DO J = JJ, min (JJ + 50, M)
7161 A(J, I) = B(I) + C(J)
7167 which can be beneficial when @code{M} is larger than the caches,
7168 because the innermost loop will iterate over a smaller amount of data
7169 that can be kept in the caches. This optimization applies to all the
7170 languages supported by GCC and is not limited to Fortran. To use this
7171 code transformation, GCC has to be configured with @option{--with-ppl}
7172 and @option{--with-cloog} to enable the Graphite loop transformation
7175 @item -fgraphite-identity
7176 @opindex fgraphite-identity
7177 Enable the identity transformation for graphite. For every SCoP we generate
7178 the polyhedral representation and transform it back to gimple. Using
7179 @option{-fgraphite-identity} we can check the costs or benefits of the
7180 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7181 are also performed by the code generator CLooG, like index splitting and
7182 dead code elimination in loops.
7184 @item -floop-flatten
7185 @opindex floop-flatten
7186 Removes the loop nesting structure: transforms the loop nest into a
7187 single loop. This transformation can be useful to vectorize all the
7188 levels of the loop nest.
7190 @item -floop-parallelize-all
7191 @opindex floop-parallelize-all
7192 Use the Graphite data dependence analysis to identify loops that can
7193 be parallelized. Parallelize all the loops that can be analyzed to
7194 not contain loop carried dependences without checking that it is
7195 profitable to parallelize the loops.
7197 @item -fcheck-data-deps
7198 @opindex fcheck-data-deps
7199 Compare the results of several data dependence analyzers. This option
7200 is used for debugging the data dependence analyzers.
7202 @item -ftree-loop-if-convert
7203 Attempt to transform conditional jumps in the innermost loops to
7204 branch-less equivalents. The intent is to remove control-flow from
7205 the innermost loops in order to improve the ability of the
7206 vectorization pass to handle these loops. This is enabled by default
7207 if vectorization is enabled.
7209 @item -ftree-loop-if-convert-stores
7210 Attempt to also if-convert conditional jumps containing memory writes.
7211 This transformation can be unsafe for multi-threaded programs as it
7212 transforms conditional memory writes into unconditional memory writes.
7215 for (i = 0; i < N; i++)
7219 would be transformed to
7221 for (i = 0; i < N; i++)
7222 A[i] = cond ? expr : A[i];
7224 potentially producing data races.
7226 @item -ftree-loop-distribution
7227 Perform loop distribution. This flag can improve cache performance on
7228 big loop bodies and allow further loop optimizations, like
7229 parallelization or vectorization, to take place. For example, the loop
7246 @item -ftree-loop-distribute-patterns
7247 Perform loop distribution of patterns that can be code generated with
7248 calls to a library. This flag is enabled by default at @option{-O3}.
7250 This pass distributes the initialization loops and generates a call to
7251 memset zero. For example, the loop
7267 and the initialization loop is transformed into a call to memset zero.
7269 @item -ftree-loop-im
7270 @opindex ftree-loop-im
7271 Perform loop invariant motion on trees. This pass moves only invariants that
7272 would be hard to handle at RTL level (function calls, operations that expand to
7273 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7274 operands of conditions that are invariant out of the loop, so that we can use
7275 just trivial invariantness analysis in loop unswitching. The pass also includes
7278 @item -ftree-loop-ivcanon
7279 @opindex ftree-loop-ivcanon
7280 Create a canonical counter for number of iterations in the loop for that
7281 determining number of iterations requires complicated analysis. Later
7282 optimizations then may determine the number easily. Useful especially
7283 in connection with unrolling.
7287 Perform induction variable optimizations (strength reduction, induction
7288 variable merging and induction variable elimination) on trees.
7290 @item -ftree-parallelize-loops=n
7291 @opindex ftree-parallelize-loops
7292 Parallelize loops, i.e., split their iteration space to run in n threads.
7293 This is only possible for loops whose iterations are independent
7294 and can be arbitrarily reordered. The optimization is only
7295 profitable on multiprocessor machines, for loops that are CPU-intensive,
7296 rather than constrained e.g.@: by memory bandwidth. This option
7297 implies @option{-pthread}, and thus is only supported on targets
7298 that have support for @option{-pthread}.
7302 Perform function-local points-to analysis on trees. This flag is
7303 enabled by default at @option{-O} and higher.
7307 Perform scalar replacement of aggregates. This pass replaces structure
7308 references with scalars to prevent committing structures to memory too
7309 early. This flag is enabled by default at @option{-O} and higher.
7311 @item -ftree-copyrename
7312 @opindex ftree-copyrename
7313 Perform copy renaming on trees. This pass attempts to rename compiler
7314 temporaries to other variables at copy locations, usually resulting in
7315 variable names which more closely resemble the original variables. This flag
7316 is enabled by default at @option{-O} and higher.
7320 Perform temporary expression replacement during the SSA->normal phase. Single
7321 use/single def temporaries are replaced at their use location with their
7322 defining expression. This results in non-GIMPLE code, but gives the expanders
7323 much more complex trees to work on resulting in better RTL generation. This is
7324 enabled by default at @option{-O} and higher.
7326 @item -ftree-vectorize
7327 @opindex ftree-vectorize
7328 Perform loop vectorization on trees. This flag is enabled by default at
7331 @item -ftree-slp-vectorize
7332 @opindex ftree-slp-vectorize
7333 Perform basic block vectorization on trees. This flag is enabled by default at
7334 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7336 @item -ftree-vect-loop-version
7337 @opindex ftree-vect-loop-version
7338 Perform loop versioning when doing loop vectorization on trees. When a loop
7339 appears to be vectorizable except that data alignment or data dependence cannot
7340 be determined at compile time then vectorized and non-vectorized versions of
7341 the loop are generated along with runtime checks for alignment or dependence
7342 to control which version is executed. This option is enabled by default
7343 except at level @option{-Os} where it is disabled.
7345 @item -fvect-cost-model
7346 @opindex fvect-cost-model
7347 Enable cost model for vectorization.
7351 Perform Value Range Propagation on trees. This is similar to the
7352 constant propagation pass, but instead of values, ranges of values are
7353 propagated. This allows the optimizers to remove unnecessary range
7354 checks like array bound checks and null pointer checks. This is
7355 enabled by default at @option{-O2} and higher. Null pointer check
7356 elimination is only done if @option{-fdelete-null-pointer-checks} is
7361 Perform tail duplication to enlarge superblock size. This transformation
7362 simplifies the control flow of the function allowing other optimizations to do
7365 @item -funroll-loops
7366 @opindex funroll-loops
7367 Unroll loops whose number of iterations can be determined at compile
7368 time or upon entry to the loop. @option{-funroll-loops} implies
7369 @option{-frerun-cse-after-loop}. This option makes code larger,
7370 and may or may not make it run faster.
7372 @item -funroll-all-loops
7373 @opindex funroll-all-loops
7374 Unroll all loops, even if their number of iterations is uncertain when
7375 the loop is entered. This usually makes programs run more slowly.
7376 @option{-funroll-all-loops} implies the same options as
7377 @option{-funroll-loops},
7379 @item -fsplit-ivs-in-unroller
7380 @opindex fsplit-ivs-in-unroller
7381 Enables expressing of values of induction variables in later iterations
7382 of the unrolled loop using the value in the first iteration. This breaks
7383 long dependency chains, thus improving efficiency of the scheduling passes.
7385 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7386 same effect. However in cases the loop body is more complicated than
7387 a single basic block, this is not reliable. It also does not work at all
7388 on some of the architectures due to restrictions in the CSE pass.
7390 This optimization is enabled by default.
7392 @item -fvariable-expansion-in-unroller
7393 @opindex fvariable-expansion-in-unroller
7394 With this option, the compiler will create multiple copies of some
7395 local variables when unrolling a loop which can result in superior code.
7397 @item -fpartial-inlining
7398 @opindex fpartial-inlining
7399 Inline parts of functions. This option has any effect only
7400 when inlining itself is turned on by the @option{-finline-functions}
7401 or @option{-finline-small-functions} options.
7403 Enabled at level @option{-O2}.
7405 @item -fpredictive-commoning
7406 @opindex fpredictive-commoning
7407 Perform predictive commoning optimization, i.e., reusing computations
7408 (especially memory loads and stores) performed in previous
7409 iterations of loops.
7411 This option is enabled at level @option{-O3}.
7413 @item -fprefetch-loop-arrays
7414 @opindex fprefetch-loop-arrays
7415 If supported by the target machine, generate instructions to prefetch
7416 memory to improve the performance of loops that access large arrays.
7418 This option may generate better or worse code; results are highly
7419 dependent on the structure of loops within the source code.
7421 Disabled at level @option{-Os}.
7424 @itemx -fno-peephole2
7425 @opindex fno-peephole
7426 @opindex fno-peephole2
7427 Disable any machine-specific peephole optimizations. The difference
7428 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7429 are implemented in the compiler; some targets use one, some use the
7430 other, a few use both.
7432 @option{-fpeephole} is enabled by default.
7433 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7435 @item -fno-guess-branch-probability
7436 @opindex fno-guess-branch-probability
7437 Do not guess branch probabilities using heuristics.
7439 GCC will use heuristics to guess branch probabilities if they are
7440 not provided by profiling feedback (@option{-fprofile-arcs}). These
7441 heuristics are based on the control flow graph. If some branch probabilities
7442 are specified by @samp{__builtin_expect}, then the heuristics will be
7443 used to guess branch probabilities for the rest of the control flow graph,
7444 taking the @samp{__builtin_expect} info into account. The interactions
7445 between the heuristics and @samp{__builtin_expect} can be complex, and in
7446 some cases, it may be useful to disable the heuristics so that the effects
7447 of @samp{__builtin_expect} are easier to understand.
7449 The default is @option{-fguess-branch-probability} at levels
7450 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7452 @item -freorder-blocks
7453 @opindex freorder-blocks
7454 Reorder basic blocks in the compiled function in order to reduce number of
7455 taken branches and improve code locality.
7457 Enabled at levels @option{-O2}, @option{-O3}.
7459 @item -freorder-blocks-and-partition
7460 @opindex freorder-blocks-and-partition
7461 In addition to reordering basic blocks in the compiled function, in order
7462 to reduce number of taken branches, partitions hot and cold basic blocks
7463 into separate sections of the assembly and .o files, to improve
7464 paging and cache locality performance.
7466 This optimization is automatically turned off in the presence of
7467 exception handling, for linkonce sections, for functions with a user-defined
7468 section attribute and on any architecture that does not support named
7471 @item -freorder-functions
7472 @opindex freorder-functions
7473 Reorder functions in the object file in order to
7474 improve code locality. This is implemented by using special
7475 subsections @code{.text.hot} for most frequently executed functions and
7476 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7477 the linker so object file format must support named sections and linker must
7478 place them in a reasonable way.
7480 Also profile feedback must be available in to make this option effective. See
7481 @option{-fprofile-arcs} for details.
7483 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7485 @item -fstrict-aliasing
7486 @opindex fstrict-aliasing
7487 Allow the compiler to assume the strictest aliasing rules applicable to
7488 the language being compiled. For C (and C++), this activates
7489 optimizations based on the type of expressions. In particular, an
7490 object of one type is assumed never to reside at the same address as an
7491 object of a different type, unless the types are almost the same. For
7492 example, an @code{unsigned int} can alias an @code{int}, but not a
7493 @code{void*} or a @code{double}. A character type may alias any other
7496 @anchor{Type-punning}Pay special attention to code like this:
7509 The practice of reading from a different union member than the one most
7510 recently written to (called ``type-punning'') is common. Even with
7511 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7512 is accessed through the union type. So, the code above will work as
7513 expected. @xref{Structures unions enumerations and bit-fields
7514 implementation}. However, this code might not:
7525 Similarly, access by taking the address, casting the resulting pointer
7526 and dereferencing the result has undefined behavior, even if the cast
7527 uses a union type, e.g.:
7531 return ((union a_union *) &d)->i;
7535 The @option{-fstrict-aliasing} option is enabled at levels
7536 @option{-O2}, @option{-O3}, @option{-Os}.
7538 @item -fstrict-overflow
7539 @opindex fstrict-overflow
7540 Allow the compiler to assume strict signed overflow rules, depending
7541 on the language being compiled. For C (and C++) this means that
7542 overflow when doing arithmetic with signed numbers is undefined, which
7543 means that the compiler may assume that it will not happen. This
7544 permits various optimizations. For example, the compiler will assume
7545 that an expression like @code{i + 10 > i} will always be true for
7546 signed @code{i}. This assumption is only valid if signed overflow is
7547 undefined, as the expression is false if @code{i + 10} overflows when
7548 using twos complement arithmetic. When this option is in effect any
7549 attempt to determine whether an operation on signed numbers will
7550 overflow must be written carefully to not actually involve overflow.
7552 This option also allows the compiler to assume strict pointer
7553 semantics: given a pointer to an object, if adding an offset to that
7554 pointer does not produce a pointer to the same object, the addition is
7555 undefined. This permits the compiler to conclude that @code{p + u >
7556 p} is always true for a pointer @code{p} and unsigned integer
7557 @code{u}. This assumption is only valid because pointer wraparound is
7558 undefined, as the expression is false if @code{p + u} overflows using
7559 twos complement arithmetic.
7561 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7562 that integer signed overflow is fully defined: it wraps. When
7563 @option{-fwrapv} is used, there is no difference between
7564 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7565 integers. With @option{-fwrapv} certain types of overflow are
7566 permitted. For example, if the compiler gets an overflow when doing
7567 arithmetic on constants, the overflowed value can still be used with
7568 @option{-fwrapv}, but not otherwise.
7570 The @option{-fstrict-overflow} option is enabled at levels
7571 @option{-O2}, @option{-O3}, @option{-Os}.
7573 @item -falign-functions
7574 @itemx -falign-functions=@var{n}
7575 @opindex falign-functions
7576 Align the start of functions to the next power-of-two greater than
7577 @var{n}, skipping up to @var{n} bytes. For instance,
7578 @option{-falign-functions=32} aligns functions to the next 32-byte
7579 boundary, but @option{-falign-functions=24} would align to the next
7580 32-byte boundary only if this can be done by skipping 23 bytes or less.
7582 @option{-fno-align-functions} and @option{-falign-functions=1} are
7583 equivalent and mean that functions will not be aligned.
7585 Some assemblers only support this flag when @var{n} is a power of two;
7586 in that case, it is rounded up.
7588 If @var{n} is not specified or is zero, use a machine-dependent default.
7590 Enabled at levels @option{-O2}, @option{-O3}.
7592 @item -falign-labels
7593 @itemx -falign-labels=@var{n}
7594 @opindex falign-labels
7595 Align all branch targets to a power-of-two boundary, skipping up to
7596 @var{n} bytes like @option{-falign-functions}. This option can easily
7597 make code slower, because it must insert dummy operations for when the
7598 branch target is reached in the usual flow of the code.
7600 @option{-fno-align-labels} and @option{-falign-labels=1} are
7601 equivalent and mean that labels will not be aligned.
7603 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7604 are greater than this value, then their values are used instead.
7606 If @var{n} is not specified or is zero, use a machine-dependent default
7607 which is very likely to be @samp{1}, meaning no alignment.
7609 Enabled at levels @option{-O2}, @option{-O3}.
7612 @itemx -falign-loops=@var{n}
7613 @opindex falign-loops
7614 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7615 like @option{-falign-functions}. The hope is that the loop will be
7616 executed many times, which will make up for any execution of the dummy
7619 @option{-fno-align-loops} and @option{-falign-loops=1} are
7620 equivalent and mean that loops will not be aligned.
7622 If @var{n} is not specified or is zero, use a machine-dependent default.
7624 Enabled at levels @option{-O2}, @option{-O3}.
7627 @itemx -falign-jumps=@var{n}
7628 @opindex falign-jumps
7629 Align branch targets to a power-of-two boundary, for branch targets
7630 where the targets can only be reached by jumping, skipping up to @var{n}
7631 bytes like @option{-falign-functions}. In this case, no dummy operations
7634 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7635 equivalent and mean that loops will not be aligned.
7637 If @var{n} is not specified or is zero, use a machine-dependent default.
7639 Enabled at levels @option{-O2}, @option{-O3}.
7641 @item -funit-at-a-time
7642 @opindex funit-at-a-time
7643 This option is left for compatibility reasons. @option{-funit-at-a-time}
7644 has no effect, while @option{-fno-unit-at-a-time} implies
7645 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7649 @item -fno-toplevel-reorder
7650 @opindex fno-toplevel-reorder
7651 Do not reorder top-level functions, variables, and @code{asm}
7652 statements. Output them in the same order that they appear in the
7653 input file. When this option is used, unreferenced static variables
7654 will not be removed. This option is intended to support existing code
7655 which relies on a particular ordering. For new code, it is better to
7658 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7659 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7664 Constructs webs as commonly used for register allocation purposes and assign
7665 each web individual pseudo register. This allows the register allocation pass
7666 to operate on pseudos directly, but also strengthens several other optimization
7667 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7668 however, make debugging impossible, since variables will no longer stay in a
7671 Enabled by default with @option{-funroll-loops}.
7673 @item -fwhole-program
7674 @opindex fwhole-program
7675 Assume that the current compilation unit represents the whole program being
7676 compiled. All public functions and variables with the exception of @code{main}
7677 and those merged by attribute @code{externally_visible} become static functions
7678 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.
7679 While this option is equivalent to proper use of the @code{static} keyword for
7680 programs consisting of a single file, in combination with option
7681 @option{-flto} this flag can be used to
7682 compile many smaller scale programs since the functions and variables become
7683 local for the whole combined compilation unit, not for the single source file
7686 This option implies @option{-fwhole-file} for Fortran programs.
7688 @item -flto[=@var{n}]
7690 This option runs the standard link-time optimizer. When invoked
7691 with source code, it generates GIMPLE (one of GCC's internal
7692 representations) and writes it to special ELF sections in the object
7693 file. When the object files are linked together, all the function
7694 bodies are read from these ELF sections and instantiated as if they
7695 had been part of the same translation unit.
7697 To use the link-timer optimizer, @option{-flto} needs to be specified at
7698 compile time and during the final link. For example,
7701 gcc -c -O2 -flto foo.c
7702 gcc -c -O2 -flto bar.c
7703 gcc -o myprog -flto -O2 foo.o bar.o
7706 The first two invocations to GCC will save a bytecode representation
7707 of GIMPLE into special ELF sections inside @file{foo.o} and
7708 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7709 @file{foo.o} and @file{bar.o}, merge the two files into a single
7710 internal image, and compile the result as usual. Since both
7711 @file{foo.o} and @file{bar.o} are merged into a single image, this
7712 causes all the inter-procedural analyses and optimizations in GCC to
7713 work across the two files as if they were a single one. This means,
7714 for example, that the inliner will be able to inline functions in
7715 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7717 Another (simpler) way to enable link-time optimization is,
7720 gcc -o myprog -flto -O2 foo.c bar.c
7723 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7724 merge them together into a single GIMPLE representation and optimize
7725 them as usual to produce @file{myprog}.
7727 The only important thing to keep in mind is that to enable link-time
7728 optimizations the @option{-flto} flag needs to be passed to both the
7729 compile and the link commands.
7731 To make whole program optimization effective, it is necessary to make
7732 certain whole program assumptions. The compiler needs to know
7733 what functions and variables can be accessed by libraries and runtime
7734 outside of the link time optimized unit. When supported by the linker,
7735 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7736 compiler information about used and externally visible symbols. When
7737 the linker plugin is not available, @option{-fwhole-program} should be
7738 used to allow the compiler to make these assumptions, which will lead
7739 to more aggressive optimization decisions.
7741 Note that when a file is compiled with @option{-flto}, the generated
7742 object file will be larger than a regular object file because it will
7743 contain GIMPLE bytecodes and the usual final code. This means that
7744 object files with LTO information can be linked as a normal object
7745 file. So, in the previous example, if the final link is done with
7748 gcc -o myprog foo.o bar.o
7751 The only difference will be that no inter-procedural optimizations
7752 will be applied to produce @file{myprog}. The two object files
7753 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7756 Additionally, the optimization flags used to compile individual files
7757 are not necessarily related to those used at link-time. For instance,
7760 gcc -c -O0 -flto foo.c
7761 gcc -c -O0 -flto bar.c
7762 gcc -o myprog -flto -O3 foo.o bar.o
7765 This will produce individual object files with unoptimized assembler
7766 code, but the resulting binary @file{myprog} will be optimized at
7767 @option{-O3}. Now, if the final binary is generated without
7768 @option{-flto}, then @file{myprog} will not be optimized.
7770 When producing the final binary with @option{-flto}, GCC will only
7771 apply link-time optimizations to those files that contain bytecode.
7772 Therefore, you can mix and match object files and libraries with
7773 GIMPLE bytecodes and final object code. GCC will automatically select
7774 which files to optimize in LTO mode and which files to link without
7777 There are some code generation flags that GCC will preserve when
7778 generating bytecodes, as they need to be used during the final link
7779 stage. Currently, the following options are saved into the GIMPLE
7780 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7781 @option{-m} target flags.
7783 At link time, these options are read-in and reapplied. Note that the
7784 current implementation makes no attempt at recognizing conflicting
7785 values for these options. If two or more files have a conflicting
7786 value (e.g., one file is compiled with @option{-fPIC} and another
7787 isn't), the compiler will simply use the last value read from the
7788 bytecode files. It is recommended, then, that all the files
7789 participating in the same link be compiled with the same options.
7791 Another feature of LTO is that it is possible to apply interprocedural
7792 optimizations on files written in different languages. This requires
7793 some support in the language front end. Currently, the C, C++ and
7794 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7795 something like this should work
7800 gfortran -c -flto baz.f90
7801 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7804 Notice that the final link is done with @command{g++} to get the C++
7805 runtime libraries and @option{-lgfortran} is added to get the Fortran
7806 runtime libraries. In general, when mixing languages in LTO mode, you
7807 should use the same link command used when mixing languages in a
7808 regular (non-LTO) compilation. This means that if your build process
7809 was mixing languages before, all you need to add is @option{-flto} to
7810 all the compile and link commands.
7812 If LTO encounters objects with C linkage declared with incompatible
7813 types in separate translation units to be linked together (undefined
7814 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7815 issued. The behavior is still undefined at runtime.
7817 If object files containing GIMPLE bytecode are stored in a library archive, say
7818 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7819 are using a linker with linker plugin support. To enable this feature, use
7820 the flag @option{-fuse-linker-plugin} at link-time:
7823 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7826 With the linker plugin enabled, the linker will extract the needed
7827 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7828 to make them part of the aggregated GIMPLE image to be optimized.
7830 If you are not using a linker with linker plugin support and/or do not
7831 enable linker plugin then the objects inside @file{libfoo.a}
7832 will be extracted and linked as usual, but they will not participate
7833 in the LTO optimization process.
7835 Link time optimizations do not require the presence of the whole program to
7836 operate. If the program does not require any symbols to be exported, it is
7837 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7838 the interprocedural optimizers to use more aggressive assumptions which may
7839 lead to improved optimization opportunities.
7840 Use of @option{-fwhole-program} is not needed when linker plugin is
7841 active (see @option{-fuse-linker-plugin}).
7843 Regarding portability: the current implementation of LTO makes no
7844 attempt at generating bytecode that can be ported between different
7845 types of hosts. The bytecode files are versioned and there is a
7846 strict version check, so bytecode files generated in one version of
7847 GCC will not work with an older/newer version of GCC.
7849 Link time optimization does not play well with generating debugging
7850 information. Combining @option{-flto} with
7851 @option{-g} is currently experimental and expected to produce wrong
7854 If you specify the optional @var{n}, the optimization and code
7855 generation done at link time is executed in parallel using @var{n}
7856 parallel jobs by utilizing an installed @command{make} program. The
7857 environment variable @env{MAKE} may be used to override the program
7858 used. The default value for @var{n} is 1.
7860 You can also specify @option{-flto=jobserver} to use GNU make's
7861 job server mode to determine the number of parallel jobs. This
7862 is useful when the Makefile calling GCC is already executing in parallel.
7863 The parent Makefile will need a @samp{+} prepended to the command recipe
7864 for this to work. This will likely only work if @env{MAKE} is
7867 This option is disabled by default.
7869 @item -flto-partition=@var{alg}
7870 @opindex flto-partition
7871 Specify the partitioning algorithm used by the link time optimizer.
7872 The value is either @code{1to1} to specify a partitioning mirroring
7873 the original source files or @code{balanced} to specify partitioning
7874 into equally sized chunks (whenever possible). Specifying @code{none}
7875 as an algorithm disables partitioning and streaming completely. The
7876 default value is @code{balanced}.
7878 @item -flto-compression-level=@var{n}
7879 This option specifies the level of compression used for intermediate
7880 language written to LTO object files, and is only meaningful in
7881 conjunction with LTO mode (@option{-flto}). Valid
7882 values are 0 (no compression) to 9 (maximum compression). Values
7883 outside this range are clamped to either 0 or 9. If the option is not
7884 given, a default balanced compression setting is used.
7887 Prints a report with internal details on the workings of the link-time
7888 optimizer. The contents of this report vary from version to version,
7889 it is meant to be useful to GCC developers when processing object
7890 files in LTO mode (via @option{-flto}).
7892 Disabled by default.
7894 @item -fuse-linker-plugin
7895 Enables the use of a linker plugin during link time optimization. This
7896 option relies on plugin support in the linker, which is available in gold
7897 or in GNU ld 2.21 or newer.
7899 This option enables the extraction of object files with GIMPLE bytecode out
7900 of library archives. This improves the quality of optimization by exposing
7901 more code to the link time optimizer. This information specifies what
7902 symbols can be accessed externally (by non-LTO object or during dynamic
7903 linking). Resulting code quality improvements on binaries (and shared
7904 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7905 See @option{-flto} for a description of the effect of this flag and how to
7908 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7909 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7911 @item -ffat-lto-objects
7912 @opindex ffat-lto-objects
7913 Fat LTO objects are object files that contain both the intermediate language
7914 and the object code. This makes them useable for both LTO linking and normal
7915 linking. This option makes effect only with @option{-flto} and is ignored
7918 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
7919 requires the complete toolchain to be aware of LTO. It requires a linker with
7920 linker plugin support for basic functionality. Additionally, nm, ar and ranlib
7921 need to support linker plugins to allow a full-featured build environment
7922 (capable of building static libraries etc).
7924 The default is @option{-ffat-lto-objects} but this default is intended to
7925 change in future releases when linker plugin enabled environments become more
7927 @item -fcompare-elim
7928 @opindex fcompare-elim
7929 After register allocation and post-register allocation instruction splitting,
7930 identify arithmetic instructions that compute processor flags similar to a
7931 comparison operation based on that arithmetic. If possible, eliminate the
7932 explicit comparison operation.
7934 This pass only applies to certain targets that cannot explicitly represent
7935 the comparison operation before register allocation is complete.
7937 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7939 @item -fcprop-registers
7940 @opindex fcprop-registers
7941 After register allocation and post-register allocation instruction splitting,
7942 we perform a copy-propagation pass to try to reduce scheduling dependencies
7943 and occasionally eliminate the copy.
7945 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7947 @item -fprofile-correction
7948 @opindex fprofile-correction
7949 Profiles collected using an instrumented binary for multi-threaded programs may
7950 be inconsistent due to missed counter updates. When this option is specified,
7951 GCC will use heuristics to correct or smooth out such inconsistencies. By
7952 default, GCC will emit an error message when an inconsistent profile is detected.
7954 @item -fprofile-dir=@var{path}
7955 @opindex fprofile-dir
7957 Set the directory to search for the profile data files in to @var{path}.
7958 This option affects only the profile data generated by
7959 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7960 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7961 and its related options. Both absolute and relative paths can be used.
7962 By default, GCC will use the current directory as @var{path}, thus the
7963 profile data file will appear in the same directory as the object file.
7965 @item -fprofile-generate
7966 @itemx -fprofile-generate=@var{path}
7967 @opindex fprofile-generate
7969 Enable options usually used for instrumenting application to produce
7970 profile useful for later recompilation with profile feedback based
7971 optimization. You must use @option{-fprofile-generate} both when
7972 compiling and when linking your program.
7974 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7976 If @var{path} is specified, GCC will look at the @var{path} to find
7977 the profile feedback data files. See @option{-fprofile-dir}.
7980 @itemx -fprofile-use=@var{path}
7981 @opindex fprofile-use
7982 Enable profile feedback directed optimizations, and optimizations
7983 generally profitable only with profile feedback available.
7985 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7986 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7988 By default, GCC emits an error message if the feedback profiles do not
7989 match the source code. This error can be turned into a warning by using
7990 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7993 If @var{path} is specified, GCC will look at the @var{path} to find
7994 the profile feedback data files. See @option{-fprofile-dir}.
7997 The following options control compiler behavior regarding floating
7998 point arithmetic. These options trade off between speed and
7999 correctness. All must be specifically enabled.
8003 @opindex ffloat-store
8004 Do not store floating point variables in registers, and inhibit other
8005 options that might change whether a floating point value is taken from a
8008 @cindex floating point precision
8009 This option prevents undesirable excess precision on machines such as
8010 the 68000 where the floating registers (of the 68881) keep more
8011 precision than a @code{double} is supposed to have. Similarly for the
8012 x86 architecture. For most programs, the excess precision does only
8013 good, but a few programs rely on the precise definition of IEEE floating
8014 point. Use @option{-ffloat-store} for such programs, after modifying
8015 them to store all pertinent intermediate computations into variables.
8017 @item -fexcess-precision=@var{style}
8018 @opindex fexcess-precision
8019 This option allows further control over excess precision on machines
8020 where floating-point registers have more precision than the IEEE
8021 @code{float} and @code{double} types and the processor does not
8022 support operations rounding to those types. By default,
8023 @option{-fexcess-precision=fast} is in effect; this means that
8024 operations are carried out in the precision of the registers and that
8025 it is unpredictable when rounding to the types specified in the source
8026 code takes place. When compiling C, if
8027 @option{-fexcess-precision=standard} is specified then excess
8028 precision will follow the rules specified in ISO C99; in particular,
8029 both casts and assignments cause values to be rounded to their
8030 semantic types (whereas @option{-ffloat-store} only affects
8031 assignments). This option is enabled by default for C if a strict
8032 conformance option such as @option{-std=c99} is used.
8035 @option{-fexcess-precision=standard} is not implemented for languages
8036 other than C, and has no effect if
8037 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8038 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8039 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8040 semantics apply without excess precision, and in the latter, rounding
8045 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8046 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8047 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8049 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8051 This option is not turned on by any @option{-O} option besides
8052 @option{-Ofast} since it can result in incorrect output for programs
8053 which depend on an exact implementation of IEEE or ISO rules/specifications
8054 for math functions. It may, however, yield faster code for programs
8055 that do not require the guarantees of these specifications.
8057 @item -fno-math-errno
8058 @opindex fno-math-errno
8059 Do not set ERRNO after calling math functions that are executed
8060 with a single instruction, e.g., sqrt. A program that relies on
8061 IEEE exceptions for math error handling may want to use this flag
8062 for speed while maintaining IEEE arithmetic compatibility.
8064 This option is not turned on by any @option{-O} option since
8065 it can result in incorrect output for programs which depend on
8066 an exact implementation of IEEE or ISO rules/specifications for
8067 math functions. It may, however, yield faster code for programs
8068 that do not require the guarantees of these specifications.
8070 The default is @option{-fmath-errno}.
8072 On Darwin systems, the math library never sets @code{errno}. There is
8073 therefore no reason for the compiler to consider the possibility that
8074 it might, and @option{-fno-math-errno} is the default.
8076 @item -funsafe-math-optimizations
8077 @opindex funsafe-math-optimizations
8079 Allow optimizations for floating-point arithmetic that (a) assume
8080 that arguments and results are valid and (b) may violate IEEE or
8081 ANSI standards. When used at link-time, it may include libraries
8082 or startup files that change the default FPU control word or other
8083 similar optimizations.
8085 This option is not turned on by any @option{-O} option since
8086 it can result in incorrect output for programs which depend on
8087 an exact implementation of IEEE or ISO rules/specifications for
8088 math functions. It may, however, yield faster code for programs
8089 that do not require the guarantees of these specifications.
8090 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8091 @option{-fassociative-math} and @option{-freciprocal-math}.
8093 The default is @option{-fno-unsafe-math-optimizations}.
8095 @item -fassociative-math
8096 @opindex fassociative-math
8098 Allow re-association of operands in series of floating-point operations.
8099 This violates the ISO C and C++ language standard by possibly changing
8100 computation result. NOTE: re-ordering may change the sign of zero as
8101 well as ignore NaNs and inhibit or create underflow or overflow (and
8102 thus cannot be used on a code which relies on rounding behavior like
8103 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
8104 and thus may not be used when ordered comparisons are required.
8105 This option requires that both @option{-fno-signed-zeros} and
8106 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8107 much sense with @option{-frounding-math}. For Fortran the option
8108 is automatically enabled when both @option{-fno-signed-zeros} and
8109 @option{-fno-trapping-math} are in effect.
8111 The default is @option{-fno-associative-math}.
8113 @item -freciprocal-math
8114 @opindex freciprocal-math
8116 Allow the reciprocal of a value to be used instead of dividing by
8117 the value if this enables optimizations. For example @code{x / y}
8118 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8119 is subject to common subexpression elimination. Note that this loses
8120 precision and increases the number of flops operating on the value.
8122 The default is @option{-fno-reciprocal-math}.
8124 @item -ffinite-math-only
8125 @opindex ffinite-math-only
8126 Allow optimizations for floating-point arithmetic that assume
8127 that arguments and results are not NaNs or +-Infs.
8129 This option is not turned on by any @option{-O} option since
8130 it can result in incorrect output for programs which depend on
8131 an exact implementation of IEEE or ISO rules/specifications for
8132 math functions. It may, however, yield faster code for programs
8133 that do not require the guarantees of these specifications.
8135 The default is @option{-fno-finite-math-only}.
8137 @item -fno-signed-zeros
8138 @opindex fno-signed-zeros
8139 Allow optimizations for floating point arithmetic that ignore the
8140 signedness of zero. IEEE arithmetic specifies the behavior of
8141 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8142 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8143 This option implies that the sign of a zero result isn't significant.
8145 The default is @option{-fsigned-zeros}.
8147 @item -fno-trapping-math
8148 @opindex fno-trapping-math
8149 Compile code assuming that floating-point operations cannot generate
8150 user-visible traps. These traps include division by zero, overflow,
8151 underflow, inexact result and invalid operation. This option requires
8152 that @option{-fno-signaling-nans} be in effect. Setting this option may
8153 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8155 This option should never be turned on by any @option{-O} option since
8156 it can result in incorrect output for programs which depend on
8157 an exact implementation of IEEE or ISO rules/specifications for
8160 The default is @option{-ftrapping-math}.
8162 @item -frounding-math
8163 @opindex frounding-math
8164 Disable transformations and optimizations that assume default floating
8165 point rounding behavior. This is round-to-zero for all floating point
8166 to integer conversions, and round-to-nearest for all other arithmetic
8167 truncations. This option should be specified for programs that change
8168 the FP rounding mode dynamically, or that may be executed with a
8169 non-default rounding mode. This option disables constant folding of
8170 floating point expressions at compile-time (which may be affected by
8171 rounding mode) and arithmetic transformations that are unsafe in the
8172 presence of sign-dependent rounding modes.
8174 The default is @option{-fno-rounding-math}.
8176 This option is experimental and does not currently guarantee to
8177 disable all GCC optimizations that are affected by rounding mode.
8178 Future versions of GCC may provide finer control of this setting
8179 using C99's @code{FENV_ACCESS} pragma. This command line option
8180 will be used to specify the default state for @code{FENV_ACCESS}.
8182 @item -fsignaling-nans
8183 @opindex fsignaling-nans
8184 Compile code assuming that IEEE signaling NaNs may generate user-visible
8185 traps during floating-point operations. Setting this option disables
8186 optimizations that may change the number of exceptions visible with
8187 signaling NaNs. This option implies @option{-ftrapping-math}.
8189 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8192 The default is @option{-fno-signaling-nans}.
8194 This option is experimental and does not currently guarantee to
8195 disable all GCC optimizations that affect signaling NaN behavior.
8197 @item -fsingle-precision-constant
8198 @opindex fsingle-precision-constant
8199 Treat floating point constant as single precision constant instead of
8200 implicitly converting it to double precision constant.
8202 @item -fcx-limited-range
8203 @opindex fcx-limited-range
8204 When enabled, this option states that a range reduction step is not
8205 needed when performing complex division. Also, there is no checking
8206 whether the result of a complex multiplication or division is @code{NaN
8207 + I*NaN}, with an attempt to rescue the situation in that case. The
8208 default is @option{-fno-cx-limited-range}, but is enabled by
8209 @option{-ffast-math}.
8211 This option controls the default setting of the ISO C99
8212 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8215 @item -fcx-fortran-rules
8216 @opindex fcx-fortran-rules
8217 Complex multiplication and division follow Fortran rules. Range
8218 reduction is done as part of complex division, but there is no checking
8219 whether the result of a complex multiplication or division is @code{NaN
8220 + I*NaN}, with an attempt to rescue the situation in that case.
8222 The default is @option{-fno-cx-fortran-rules}.
8226 The following options control optimizations that may improve
8227 performance, but are not enabled by any @option{-O} options. This
8228 section includes experimental options that may produce broken code.
8231 @item -fbranch-probabilities
8232 @opindex fbranch-probabilities
8233 After running a program compiled with @option{-fprofile-arcs}
8234 (@pxref{Debugging Options,, Options for Debugging Your Program or
8235 @command{gcc}}), you can compile it a second time using
8236 @option{-fbranch-probabilities}, to improve optimizations based on
8237 the number of times each branch was taken. When the program
8238 compiled with @option{-fprofile-arcs} exits it saves arc execution
8239 counts to a file called @file{@var{sourcename}.gcda} for each source
8240 file. The information in this data file is very dependent on the
8241 structure of the generated code, so you must use the same source code
8242 and the same optimization options for both compilations.
8244 With @option{-fbranch-probabilities}, GCC puts a
8245 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8246 These can be used to improve optimization. Currently, they are only
8247 used in one place: in @file{reorg.c}, instead of guessing which path a
8248 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8249 exactly determine which path is taken more often.
8251 @item -fprofile-values
8252 @opindex fprofile-values
8253 If combined with @option{-fprofile-arcs}, it adds code so that some
8254 data about values of expressions in the program is gathered.
8256 With @option{-fbranch-probabilities}, it reads back the data gathered
8257 from profiling values of expressions for usage in optimizations.
8259 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8263 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8264 a code to gather information about values of expressions.
8266 With @option{-fbranch-probabilities}, it reads back the data gathered
8267 and actually performs the optimizations based on them.
8268 Currently the optimizations include specialization of division operation
8269 using the knowledge about the value of the denominator.
8271 @item -frename-registers
8272 @opindex frename-registers
8273 Attempt to avoid false dependencies in scheduled code by making use
8274 of registers left over after register allocation. This optimization
8275 will most benefit processors with lots of registers. Depending on the
8276 debug information format adopted by the target, however, it can
8277 make debugging impossible, since variables will no longer stay in
8278 a ``home register''.
8280 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8284 Perform tail duplication to enlarge superblock size. This transformation
8285 simplifies the control flow of the function allowing other optimizations to do
8288 Enabled with @option{-fprofile-use}.
8290 @item -funroll-loops
8291 @opindex funroll-loops
8292 Unroll loops whose number of iterations can be determined at compile time or
8293 upon entry to the loop. @option{-funroll-loops} implies
8294 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8295 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8296 small constant number of iterations). This option makes code larger, and may
8297 or may not make it run faster.
8299 Enabled with @option{-fprofile-use}.
8301 @item -funroll-all-loops
8302 @opindex funroll-all-loops
8303 Unroll all loops, even if their number of iterations is uncertain when
8304 the loop is entered. This usually makes programs run more slowly.
8305 @option{-funroll-all-loops} implies the same options as
8306 @option{-funroll-loops}.
8309 @opindex fpeel-loops
8310 Peels the loops for that there is enough information that they do not
8311 roll much (from profile feedback). It also turns on complete loop peeling
8312 (i.e.@: complete removal of loops with small constant number of iterations).
8314 Enabled with @option{-fprofile-use}.
8316 @item -fmove-loop-invariants
8317 @opindex fmove-loop-invariants
8318 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8319 at level @option{-O1}
8321 @item -funswitch-loops
8322 @opindex funswitch-loops
8323 Move branches with loop invariant conditions out of the loop, with duplicates
8324 of the loop on both branches (modified according to result of the condition).
8326 @item -ffunction-sections
8327 @itemx -fdata-sections
8328 @opindex ffunction-sections
8329 @opindex fdata-sections
8330 Place each function or data item into its own section in the output
8331 file if the target supports arbitrary sections. The name of the
8332 function or the name of the data item determines the section's name
8335 Use these options on systems where the linker can perform optimizations
8336 to improve locality of reference in the instruction space. Most systems
8337 using the ELF object format and SPARC processors running Solaris 2 have
8338 linkers with such optimizations. AIX may have these optimizations in
8341 Only use these options when there are significant benefits from doing
8342 so. When you specify these options, the assembler and linker will
8343 create larger object and executable files and will also be slower.
8344 You will not be able to use @code{gprof} on all systems if you
8345 specify this option and you may have problems with debugging if
8346 you specify both this option and @option{-g}.
8348 @item -fbranch-target-load-optimize
8349 @opindex fbranch-target-load-optimize
8350 Perform branch target register load optimization before prologue / epilogue
8352 The use of target registers can typically be exposed only during reload,
8353 thus hoisting loads out of loops and doing inter-block scheduling needs
8354 a separate optimization pass.
8356 @item -fbranch-target-load-optimize2
8357 @opindex fbranch-target-load-optimize2
8358 Perform branch target register load optimization after prologue / epilogue
8361 @item -fbtr-bb-exclusive
8362 @opindex fbtr-bb-exclusive
8363 When performing branch target register load optimization, don't reuse
8364 branch target registers in within any basic block.
8366 @item -fstack-protector
8367 @opindex fstack-protector
8368 Emit extra code to check for buffer overflows, such as stack smashing
8369 attacks. This is done by adding a guard variable to functions with
8370 vulnerable objects. This includes functions that call alloca, and
8371 functions with buffers larger than 8 bytes. The guards are initialized
8372 when a function is entered and then checked when the function exits.
8373 If a guard check fails, an error message is printed and the program exits.
8375 @item -fstack-protector-all
8376 @opindex fstack-protector-all
8377 Like @option{-fstack-protector} except that all functions are protected.
8379 @item -fsection-anchors
8380 @opindex fsection-anchors
8381 Try to reduce the number of symbolic address calculations by using
8382 shared ``anchor'' symbols to address nearby objects. This transformation
8383 can help to reduce the number of GOT entries and GOT accesses on some
8386 For example, the implementation of the following function @code{foo}:
8390 int foo (void) @{ return a + b + c; @}
8393 would usually calculate the addresses of all three variables, but if you
8394 compile it with @option{-fsection-anchors}, it will access the variables
8395 from a common anchor point instead. The effect is similar to the
8396 following pseudocode (which isn't valid C):
8401 register int *xr = &x;
8402 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8406 Not all targets support this option.
8408 @item --param @var{name}=@var{value}
8410 In some places, GCC uses various constants to control the amount of
8411 optimization that is done. For example, GCC will not inline functions
8412 that contain more that a certain number of instructions. You can
8413 control some of these constants on the command-line using the
8414 @option{--param} option.
8416 The names of specific parameters, and the meaning of the values, are
8417 tied to the internals of the compiler, and are subject to change
8418 without notice in future releases.
8420 In each case, the @var{value} is an integer. The allowable choices for
8421 @var{name} are given in the following table:
8424 @item predictable-branch-outcome
8425 When branch is predicted to be taken with probability lower than this threshold
8426 (in percent), then it is considered well predictable. The default is 10.
8428 @item max-crossjump-edges
8429 The maximum number of incoming edges to consider for crossjumping.
8430 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8431 the number of edges incoming to each block. Increasing values mean
8432 more aggressive optimization, making the compile time increase with
8433 probably small improvement in executable size.
8435 @item min-crossjump-insns
8436 The minimum number of instructions which must be matched at the end
8437 of two blocks before crossjumping will be performed on them. This
8438 value is ignored in the case where all instructions in the block being
8439 crossjumped from are matched. The default value is 5.
8441 @item max-grow-copy-bb-insns
8442 The maximum code size expansion factor when copying basic blocks
8443 instead of jumping. The expansion is relative to a jump instruction.
8444 The default value is 8.
8446 @item max-goto-duplication-insns
8447 The maximum number of instructions to duplicate to a block that jumps
8448 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8449 passes, GCC factors computed gotos early in the compilation process,
8450 and unfactors them as late as possible. Only computed jumps at the
8451 end of a basic blocks with no more than max-goto-duplication-insns are
8452 unfactored. The default value is 8.
8454 @item max-delay-slot-insn-search
8455 The maximum number of instructions to consider when looking for an
8456 instruction to fill a delay slot. If more than this arbitrary number of
8457 instructions is searched, the time savings from filling the delay slot
8458 will be minimal so stop searching. Increasing values mean more
8459 aggressive optimization, making the compile time increase with probably
8460 small improvement in executable run time.
8462 @item max-delay-slot-live-search
8463 When trying to fill delay slots, the maximum number of instructions to
8464 consider when searching for a block with valid live register
8465 information. Increasing this arbitrarily chosen value means more
8466 aggressive optimization, increasing the compile time. This parameter
8467 should be removed when the delay slot code is rewritten to maintain the
8470 @item max-gcse-memory
8471 The approximate maximum amount of memory that will be allocated in
8472 order to perform the global common subexpression elimination
8473 optimization. If more memory than specified is required, the
8474 optimization will not be done.
8476 @item max-gcse-insertion-ratio
8477 If the ratio of expression insertions to deletions is larger than this value
8478 for any expression, then RTL PRE will insert or remove the expression and thus
8479 leave partially redundant computations in the instruction stream. The default value is 20.
8481 @item max-pending-list-length
8482 The maximum number of pending dependencies scheduling will allow
8483 before flushing the current state and starting over. Large functions
8484 with few branches or calls can create excessively large lists which
8485 needlessly consume memory and resources.
8487 @item max-modulo-backtrack-attempts
8488 The maximum number of backtrack attempts the scheduler should make
8489 when modulo scheduling a loop. Larger values can exponentially increase
8492 @item max-inline-insns-single
8493 Several parameters control the tree inliner used in gcc.
8494 This number sets the maximum number of instructions (counted in GCC's
8495 internal representation) in a single function that the tree inliner
8496 will consider for inlining. This only affects functions declared
8497 inline and methods implemented in a class declaration (C++).
8498 The default value is 400.
8500 @item max-inline-insns-auto
8501 When you use @option{-finline-functions} (included in @option{-O3}),
8502 a lot of functions that would otherwise not be considered for inlining
8503 by the compiler will be investigated. To those functions, a different
8504 (more restrictive) limit compared to functions declared inline can
8506 The default value is 40.
8508 @item large-function-insns
8509 The limit specifying really large functions. For functions larger than this
8510 limit after inlining, inlining is constrained by
8511 @option{--param large-function-growth}. This parameter is useful primarily
8512 to avoid extreme compilation time caused by non-linear algorithms used by the
8514 The default value is 2700.
8516 @item large-function-growth
8517 Specifies maximal growth of large function caused by inlining in percents.
8518 The default value is 100 which limits large function growth to 2.0 times
8521 @item large-unit-insns
8522 The limit specifying large translation unit. Growth caused by inlining of
8523 units larger than this limit is limited by @option{--param inline-unit-growth}.
8524 For small units this might be too tight (consider unit consisting of function A
8525 that is inline and B that just calls A three time. If B is small relative to
8526 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8527 large units consisting of small inlineable functions however the overall unit
8528 growth limit is needed to avoid exponential explosion of code size. Thus for
8529 smaller units, the size is increased to @option{--param large-unit-insns}
8530 before applying @option{--param inline-unit-growth}. The default is 10000
8532 @item inline-unit-growth
8533 Specifies maximal overall growth of the compilation unit caused by inlining.
8534 The default value is 30 which limits unit growth to 1.3 times the original
8537 @item ipcp-unit-growth
8538 Specifies maximal overall growth of the compilation unit caused by
8539 interprocedural constant propagation. The default value is 10 which limits
8540 unit growth to 1.1 times the original size.
8542 @item large-stack-frame
8543 The limit specifying large stack frames. While inlining the algorithm is trying
8544 to not grow past this limit too much. Default value is 256 bytes.
8546 @item large-stack-frame-growth
8547 Specifies maximal growth of large stack frames caused by inlining in percents.
8548 The default value is 1000 which limits large stack frame growth to 11 times
8551 @item max-inline-insns-recursive
8552 @itemx max-inline-insns-recursive-auto
8553 Specifies maximum number of instructions out-of-line copy of self recursive inline
8554 function can grow into by performing recursive inlining.
8556 For functions declared inline @option{--param max-inline-insns-recursive} is
8557 taken into account. For function not declared inline, recursive inlining
8558 happens only when @option{-finline-functions} (included in @option{-O3}) is
8559 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8560 default value is 450.
8562 @item max-inline-recursive-depth
8563 @itemx max-inline-recursive-depth-auto
8564 Specifies maximum recursion depth used by the recursive inlining.
8566 For functions declared inline @option{--param max-inline-recursive-depth} is
8567 taken into account. For function not declared inline, recursive inlining
8568 happens only when @option{-finline-functions} (included in @option{-O3}) is
8569 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8572 @item min-inline-recursive-probability
8573 Recursive inlining is profitable only for function having deep recursion
8574 in average and can hurt for function having little recursion depth by
8575 increasing the prologue size or complexity of function body to other
8578 When profile feedback is available (see @option{-fprofile-generate}) the actual
8579 recursion depth can be guessed from probability that function will recurse via
8580 given call expression. This parameter limits inlining only to call expression
8581 whose probability exceeds given threshold (in percents). The default value is
8584 @item early-inlining-insns
8585 Specify growth that early inliner can make. In effect it increases amount of
8586 inlining for code having large abstraction penalty. The default value is 10.
8588 @item max-early-inliner-iterations
8589 @itemx max-early-inliner-iterations
8590 Limit of iterations of early inliner. This basically bounds number of nested
8591 indirect calls early inliner can resolve. Deeper chains are still handled by
8594 @item comdat-sharing-probability
8595 @itemx comdat-sharing-probability
8596 Probability (in percent) that C++ inline function with comdat visibility
8597 will be shared across multiple compilation units. The default value is 20.
8599 @item min-vect-loop-bound
8600 The minimum number of iterations under which a loop will not get vectorized
8601 when @option{-ftree-vectorize} is used. The number of iterations after
8602 vectorization needs to be greater than the value specified by this option
8603 to allow vectorization. The default value is 0.
8605 @item gcse-cost-distance-ratio
8606 Scaling factor in calculation of maximum distance an expression
8607 can be moved by GCSE optimizations. This is currently supported only in the
8608 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8609 will be with simple expressions, i.e., the expressions which have cost
8610 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8611 hoisting of simple expressions. The default value is 10.
8613 @item gcse-unrestricted-cost
8614 Cost, roughly measured as the cost of a single typical machine
8615 instruction, at which GCSE optimizations will not constrain
8616 the distance an expression can travel. This is currently
8617 supported only in the code hoisting pass. The lesser the cost,
8618 the more aggressive code hoisting will be. Specifying 0 will
8619 allow all expressions to travel unrestricted distances.
8620 The default value is 3.
8622 @item max-hoist-depth
8623 The depth of search in the dominator tree for expressions to hoist.
8624 This is used to avoid quadratic behavior in hoisting algorithm.
8625 The value of 0 will avoid limiting the search, but may slow down compilation
8626 of huge functions. The default value is 30.
8628 @item max-tail-merge-comparisons
8629 The maximum amount of similar bbs to compare a bb with. This is used to
8630 avoid quadratic behaviour in tree tail merging. The default value is 10.
8632 @item max-tail-merge-iterations
8633 The maximum amount of iterations of the pass over the function. This is used to
8634 limit run time in tree tail merging. The default value is 2.
8636 @item max-unrolled-insns
8637 The maximum number of instructions that a loop should have if that loop
8638 is unrolled, and if the loop is unrolled, it determines how many times
8639 the loop code is unrolled.
8641 @item max-average-unrolled-insns
8642 The maximum number of instructions biased by probabilities of their execution
8643 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8644 it determines how many times the loop code is unrolled.
8646 @item max-unroll-times
8647 The maximum number of unrollings of a single loop.
8649 @item max-peeled-insns
8650 The maximum number of instructions that a loop should have if that loop
8651 is peeled, and if the loop is peeled, it determines how many times
8652 the loop code is peeled.
8654 @item max-peel-times
8655 The maximum number of peelings of a single loop.
8657 @item max-completely-peeled-insns
8658 The maximum number of insns of a completely peeled loop.
8660 @item max-completely-peel-times
8661 The maximum number of iterations of a loop to be suitable for complete peeling.
8663 @item max-completely-peel-loop-nest-depth
8664 The maximum depth of a loop nest suitable for complete peeling.
8666 @item max-unswitch-insns
8667 The maximum number of insns of an unswitched loop.
8669 @item max-unswitch-level
8670 The maximum number of branches unswitched in a single loop.
8673 The minimum cost of an expensive expression in the loop invariant motion.
8675 @item iv-consider-all-candidates-bound
8676 Bound on number of candidates for induction variables below that
8677 all candidates are considered for each use in induction variable
8678 optimizations. Only the most relevant candidates are considered
8679 if there are more candidates, to avoid quadratic time complexity.
8681 @item iv-max-considered-uses
8682 The induction variable optimizations give up on loops that contain more
8683 induction variable uses.
8685 @item iv-always-prune-cand-set-bound
8686 If number of candidates in the set is smaller than this value,
8687 we always try to remove unnecessary ivs from the set during its
8688 optimization when a new iv is added to the set.
8690 @item scev-max-expr-size
8691 Bound on size of expressions used in the scalar evolutions analyzer.
8692 Large expressions slow the analyzer.
8694 @item scev-max-expr-complexity
8695 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8696 Complex expressions slow the analyzer.
8698 @item omega-max-vars
8699 The maximum number of variables in an Omega constraint system.
8700 The default value is 128.
8702 @item omega-max-geqs
8703 The maximum number of inequalities in an Omega constraint system.
8704 The default value is 256.
8707 The maximum number of equalities in an Omega constraint system.
8708 The default value is 128.
8710 @item omega-max-wild-cards
8711 The maximum number of wildcard variables that the Omega solver will
8712 be able to insert. The default value is 18.
8714 @item omega-hash-table-size
8715 The size of the hash table in the Omega solver. The default value is
8718 @item omega-max-keys
8719 The maximal number of keys used by the Omega solver. The default
8722 @item omega-eliminate-redundant-constraints
8723 When set to 1, use expensive methods to eliminate all redundant
8724 constraints. The default value is 0.
8726 @item vect-max-version-for-alignment-checks
8727 The maximum number of runtime checks that can be performed when
8728 doing loop versioning for alignment in the vectorizer. See option
8729 ftree-vect-loop-version for more information.
8731 @item vect-max-version-for-alias-checks
8732 The maximum number of runtime checks that can be performed when
8733 doing loop versioning for alias in the vectorizer. See option
8734 ftree-vect-loop-version for more information.
8736 @item max-iterations-to-track
8738 The maximum number of iterations of a loop the brute force algorithm
8739 for analysis of # of iterations of the loop tries to evaluate.
8741 @item hot-bb-count-fraction
8742 Select fraction of the maximal count of repetitions of basic block in program
8743 given basic block needs to have to be considered hot.
8745 @item hot-bb-frequency-fraction
8746 Select fraction of the entry block frequency of executions of basic block in
8747 function given basic block needs to have to be considered hot.
8749 @item max-predicted-iterations
8750 The maximum number of loop iterations we predict statically. This is useful
8751 in cases where function contain single loop with known bound and other loop
8752 with unknown. We predict the known number of iterations correctly, while
8753 the unknown number of iterations average to roughly 10. This means that the
8754 loop without bounds would appear artificially cold relative to the other one.
8756 @item align-threshold
8758 Select fraction of the maximal frequency of executions of basic block in
8759 function given basic block will get aligned.
8761 @item align-loop-iterations
8763 A loop expected to iterate at lest the selected number of iterations will get
8766 @item tracer-dynamic-coverage
8767 @itemx tracer-dynamic-coverage-feedback
8769 This value is used to limit superblock formation once the given percentage of
8770 executed instructions is covered. This limits unnecessary code size
8773 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8774 feedback is available. The real profiles (as opposed to statically estimated
8775 ones) are much less balanced allowing the threshold to be larger value.
8777 @item tracer-max-code-growth
8778 Stop tail duplication once code growth has reached given percentage. This is
8779 rather hokey argument, as most of the duplicates will be eliminated later in
8780 cross jumping, so it may be set to much higher values than is the desired code
8783 @item tracer-min-branch-ratio
8785 Stop reverse growth when the reverse probability of best edge is less than this
8786 threshold (in percent).
8788 @item tracer-min-branch-ratio
8789 @itemx tracer-min-branch-ratio-feedback
8791 Stop forward growth if the best edge do have probability lower than this
8794 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8795 compilation for profile feedback and one for compilation without. The value
8796 for compilation with profile feedback needs to be more conservative (higher) in
8797 order to make tracer effective.
8799 @item max-cse-path-length
8801 Maximum number of basic blocks on path that cse considers. The default is 10.
8804 The maximum instructions CSE process before flushing. The default is 1000.
8806 @item ggc-min-expand
8808 GCC uses a garbage collector to manage its own memory allocation. This
8809 parameter specifies the minimum percentage by which the garbage
8810 collector's heap should be allowed to expand between collections.
8811 Tuning this may improve compilation speed; it has no effect on code
8814 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8815 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8816 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8817 GCC is not able to calculate RAM on a particular platform, the lower
8818 bound of 30% is used. Setting this parameter and
8819 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8820 every opportunity. This is extremely slow, but can be useful for
8823 @item ggc-min-heapsize
8825 Minimum size of the garbage collector's heap before it begins bothering
8826 to collect garbage. The first collection occurs after the heap expands
8827 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8828 tuning this may improve compilation speed, and has no effect on code
8831 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8832 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8833 with a lower bound of 4096 (four megabytes) and an upper bound of
8834 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8835 particular platform, the lower bound is used. Setting this parameter
8836 very large effectively disables garbage collection. Setting this
8837 parameter and @option{ggc-min-expand} to zero causes a full collection
8838 to occur at every opportunity.
8840 @item max-reload-search-insns
8841 The maximum number of instruction reload should look backward for equivalent
8842 register. Increasing values mean more aggressive optimization, making the
8843 compile time increase with probably slightly better performance. The default
8846 @item max-cselib-memory-locations
8847 The maximum number of memory locations cselib should take into account.
8848 Increasing values mean more aggressive optimization, making the compile time
8849 increase with probably slightly better performance. The default value is 500.
8851 @item reorder-blocks-duplicate
8852 @itemx reorder-blocks-duplicate-feedback
8854 Used by basic block reordering pass to decide whether to use unconditional
8855 branch or duplicate the code on its destination. Code is duplicated when its
8856 estimated size is smaller than this value multiplied by the estimated size of
8857 unconditional jump in the hot spots of the program.
8859 The @option{reorder-block-duplicate-feedback} is used only when profile
8860 feedback is available and may be set to higher values than
8861 @option{reorder-block-duplicate} since information about the hot spots is more
8864 @item max-sched-ready-insns
8865 The maximum number of instructions ready to be issued the scheduler should
8866 consider at any given time during the first scheduling pass. Increasing
8867 values mean more thorough searches, making the compilation time increase
8868 with probably little benefit. The default value is 100.
8870 @item max-sched-region-blocks
8871 The maximum number of blocks in a region to be considered for
8872 interblock scheduling. The default value is 10.
8874 @item max-pipeline-region-blocks
8875 The maximum number of blocks in a region to be considered for
8876 pipelining in the selective scheduler. The default value is 15.
8878 @item max-sched-region-insns
8879 The maximum number of insns in a region to be considered for
8880 interblock scheduling. The default value is 100.
8882 @item max-pipeline-region-insns
8883 The maximum number of insns in a region to be considered for
8884 pipelining in the selective scheduler. The default value is 200.
8887 The minimum probability (in percents) of reaching a source block
8888 for interblock speculative scheduling. The default value is 40.
8890 @item max-sched-extend-regions-iters
8891 The maximum number of iterations through CFG to extend regions.
8892 0 - disable region extension,
8893 N - do at most N iterations.
8894 The default value is 0.
8896 @item max-sched-insn-conflict-delay
8897 The maximum conflict delay for an insn to be considered for speculative motion.
8898 The default value is 3.
8900 @item sched-spec-prob-cutoff
8901 The minimal probability of speculation success (in percents), so that
8902 speculative insn will be scheduled.
8903 The default value is 40.
8905 @item sched-mem-true-dep-cost
8906 Minimal distance (in CPU cycles) between store and load targeting same
8907 memory locations. The default value is 1.
8909 @item selsched-max-lookahead
8910 The maximum size of the lookahead window of selective scheduling. It is a
8911 depth of search for available instructions.
8912 The default value is 50.
8914 @item selsched-max-sched-times
8915 The maximum number of times that an instruction will be scheduled during
8916 selective scheduling. This is the limit on the number of iterations
8917 through which the instruction may be pipelined. The default value is 2.
8919 @item selsched-max-insns-to-rename
8920 The maximum number of best instructions in the ready list that are considered
8921 for renaming in the selective scheduler. The default value is 2.
8924 The minimum value of stage count that swing modulo scheduler will
8925 generate. The default value is 2.
8927 @item max-last-value-rtl
8928 The maximum size measured as number of RTLs that can be recorded in an expression
8929 in combiner for a pseudo register as last known value of that register. The default
8932 @item integer-share-limit
8933 Small integer constants can use a shared data structure, reducing the
8934 compiler's memory usage and increasing its speed. This sets the maximum
8935 value of a shared integer constant. The default value is 256.
8937 @item min-virtual-mappings
8938 Specifies the minimum number of virtual mappings in the incremental
8939 SSA updater that should be registered to trigger the virtual mappings
8940 heuristic defined by virtual-mappings-ratio. The default value is
8943 @item virtual-mappings-ratio
8944 If the number of virtual mappings is virtual-mappings-ratio bigger
8945 than the number of virtual symbols to be updated, then the incremental
8946 SSA updater switches to a full update for those symbols. The default
8949 @item ssp-buffer-size
8950 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8951 protection when @option{-fstack-protection} is used.
8953 @item max-jump-thread-duplication-stmts
8954 Maximum number of statements allowed in a block that needs to be
8955 duplicated when threading jumps.
8957 @item max-fields-for-field-sensitive
8958 Maximum number of fields in a structure we will treat in
8959 a field sensitive manner during pointer analysis. The default is zero
8960 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8962 @item prefetch-latency
8963 Estimate on average number of instructions that are executed before
8964 prefetch finishes. The distance we prefetch ahead is proportional
8965 to this constant. Increasing this number may also lead to less
8966 streams being prefetched (see @option{simultaneous-prefetches}).
8968 @item simultaneous-prefetches
8969 Maximum number of prefetches that can run at the same time.
8971 @item l1-cache-line-size
8972 The size of cache line in L1 cache, in bytes.
8975 The size of L1 cache, in kilobytes.
8978 The size of L2 cache, in kilobytes.
8980 @item min-insn-to-prefetch-ratio
8981 The minimum ratio between the number of instructions and the
8982 number of prefetches to enable prefetching in a loop.
8984 @item prefetch-min-insn-to-mem-ratio
8985 The minimum ratio between the number of instructions and the
8986 number of memory references to enable prefetching in a loop.
8988 @item use-canonical-types
8989 Whether the compiler should use the ``canonical'' type system. By
8990 default, this should always be 1, which uses a more efficient internal
8991 mechanism for comparing types in C++ and Objective-C++. However, if
8992 bugs in the canonical type system are causing compilation failures,
8993 set this value to 0 to disable canonical types.
8995 @item switch-conversion-max-branch-ratio
8996 Switch initialization conversion will refuse to create arrays that are
8997 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8998 branches in the switch.
9000 @item max-partial-antic-length
9001 Maximum length of the partial antic set computed during the tree
9002 partial redundancy elimination optimization (@option{-ftree-pre}) when
9003 optimizing at @option{-O3} and above. For some sorts of source code
9004 the enhanced partial redundancy elimination optimization can run away,
9005 consuming all of the memory available on the host machine. This
9006 parameter sets a limit on the length of the sets that are computed,
9007 which prevents the runaway behavior. Setting a value of 0 for
9008 this parameter will allow an unlimited set length.
9010 @item sccvn-max-scc-size
9011 Maximum size of a strongly connected component (SCC) during SCCVN
9012 processing. If this limit is hit, SCCVN processing for the whole
9013 function will not be done and optimizations depending on it will
9014 be disabled. The default maximum SCC size is 10000.
9016 @item ira-max-loops-num
9017 IRA uses a regional register allocation by default. If a function
9018 contains loops more than number given by the parameter, only at most
9019 given number of the most frequently executed loops will form regions
9020 for the regional register allocation. The default value of the
9023 @item ira-max-conflict-table-size
9024 Although IRA uses a sophisticated algorithm of compression conflict
9025 table, the table can be still big for huge functions. If the conflict
9026 table for a function could be more than size in MB given by the
9027 parameter, the conflict table is not built and faster, simpler, and
9028 lower quality register allocation algorithm will be used. The
9029 algorithm do not use pseudo-register conflicts. The default value of
9030 the parameter is 2000.
9032 @item ira-loop-reserved-regs
9033 IRA can be used to evaluate more accurate register pressure in loops
9034 for decision to move loop invariants (see @option{-O3}). The number
9035 of available registers reserved for some other purposes is described
9036 by this parameter. The default value of the parameter is 2 which is
9037 minimal number of registers needed for execution of typical
9038 instruction. This value is the best found from numerous experiments.
9040 @item loop-invariant-max-bbs-in-loop
9041 Loop invariant motion can be very expensive, both in compile time and
9042 in amount of needed compile time memory, with very large loops. Loops
9043 with more basic blocks than this parameter won't have loop invariant
9044 motion optimization performed on them. The default value of the
9045 parameter is 1000 for -O1 and 10000 for -O2 and above.
9047 @item max-vartrack-size
9048 Sets a maximum number of hash table slots to use during variable
9049 tracking dataflow analysis of any function. If this limit is exceeded
9050 with variable tracking at assignments enabled, analysis for that
9051 function is retried without it, after removing all debug insns from
9052 the function. If the limit is exceeded even without debug insns, var
9053 tracking analysis is completely disabled for the function. Setting
9054 the parameter to zero makes it unlimited.
9056 @item max-vartrack-expr-depth
9057 Sets a maximum number of recursion levels when attempting to map
9058 variable names or debug temporaries to value expressions. This trades
9059 compile time for more complete debug information. If this is set too
9060 low, value expressions that are available and could be represented in
9061 debug information may end up not being used; setting this higher may
9062 enable the compiler to find more complex debug expressions, but compile
9063 time may grow exponentially, and even then, it may fail to find more
9064 usable expressions. The default is 10.
9066 @item min-nondebug-insn-uid
9067 Use uids starting at this parameter for nondebug insns. The range below
9068 the parameter is reserved exclusively for debug insns created by
9069 @option{-fvar-tracking-assignments}, but debug insns may get
9070 (non-overlapping) uids above it if the reserved range is exhausted.
9072 @item ipa-sra-ptr-growth-factor
9073 IPA-SRA will replace a pointer to an aggregate with one or more new
9074 parameters only when their cumulative size is less or equal to
9075 @option{ipa-sra-ptr-growth-factor} times the size of the original
9078 @item graphite-max-nb-scop-params
9079 To avoid exponential effects in the Graphite loop transforms, the
9080 number of parameters in a Static Control Part (SCoP) is bounded. The
9081 default value is 10 parameters. A variable whose value is unknown at
9082 compile time and defined outside a SCoP is a parameter of the SCoP.
9084 @item graphite-max-bbs-per-function
9085 To avoid exponential effects in the detection of SCoPs, the size of
9086 the functions analyzed by Graphite is bounded. The default value is
9089 @item loop-block-tile-size
9090 Loop blocking or strip mining transforms, enabled with
9091 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9092 loop in the loop nest by a given number of iterations. The strip
9093 length can be changed using the @option{loop-block-tile-size}
9094 parameter. The default value is 51 iterations.
9096 @item ipa-cp-value-list-size
9097 IPA-CP attempts to track all possible values and types passed to a function's
9098 parameter in order to propagate them and perform devirtualization.
9099 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9100 stores per one formal parameter of a function.
9102 @item lto-partitions
9103 Specify desired number of partitions produced during WHOPR compilation.
9104 The number of partitions should exceed the number of CPUs used for compilation.
9105 The default value is 32.
9107 @item lto-minpartition
9108 Size of minimal partition for WHOPR (in estimated instructions).
9109 This prevents expenses of splitting very small programs into too many
9112 @item cxx-max-namespaces-for-diagnostic-help
9113 The maximum number of namespaces to consult for suggestions when C++
9114 name lookup fails for an identifier. The default is 1000.
9116 @item max-stores-to-sink
9117 The maximum number of conditional stores paires that can be sunk. Set to 0
9118 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9119 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9121 @item allow-store-data-races
9122 Allow optimizers to introduce new data races on stores.
9123 Set to 1 to allow, otherwise to 0. This option is enabled by default
9124 unless implicitly set by the @option{-fmemory-model=} option.
9126 @item case-values-threshold
9127 The smallest number of different values for which it is best to use a
9128 jump-table instead of a tree of conditional branches. If the value is
9129 0, use the default for the machine. The default is 0.
9131 @item tree-reassoc-width
9132 Set the maximum number of instructions executed in parallel in
9133 reassociated tree. This parameter overrides target dependent
9134 heuristics used by default if has non zero value.
9139 @node Preprocessor Options
9140 @section Options Controlling the Preprocessor
9141 @cindex preprocessor options
9142 @cindex options, preprocessor
9144 These options control the C preprocessor, which is run on each C source
9145 file before actual compilation.
9147 If you use the @option{-E} option, nothing is done except preprocessing.
9148 Some of these options make sense only together with @option{-E} because
9149 they cause the preprocessor output to be unsuitable for actual
9153 @item -Wp,@var{option}
9155 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9156 and pass @var{option} directly through to the preprocessor. If
9157 @var{option} contains commas, it is split into multiple options at the
9158 commas. However, many options are modified, translated or interpreted
9159 by the compiler driver before being passed to the preprocessor, and
9160 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9161 interface is undocumented and subject to change, so whenever possible
9162 you should avoid using @option{-Wp} and let the driver handle the
9165 @item -Xpreprocessor @var{option}
9166 @opindex Xpreprocessor
9167 Pass @var{option} as an option to the preprocessor. You can use this to
9168 supply system-specific preprocessor options which GCC does not know how to
9171 If you want to pass an option that takes an argument, you must use
9172 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9175 @include cppopts.texi
9177 @node Assembler Options
9178 @section Passing Options to the Assembler
9180 @c prevent bad page break with this line
9181 You can pass options to the assembler.
9184 @item -Wa,@var{option}
9186 Pass @var{option} as an option to the assembler. If @var{option}
9187 contains commas, it is split into multiple options at the commas.
9189 @item -Xassembler @var{option}
9191 Pass @var{option} as an option to the assembler. You can use this to
9192 supply system-specific assembler options which GCC does not know how to
9195 If you want to pass an option that takes an argument, you must use
9196 @option{-Xassembler} twice, once for the option and once for the argument.
9201 @section Options for Linking
9202 @cindex link options
9203 @cindex options, linking
9205 These options come into play when the compiler links object files into
9206 an executable output file. They are meaningless if the compiler is
9207 not doing a link step.
9211 @item @var{object-file-name}
9212 A file name that does not end in a special recognized suffix is
9213 considered to name an object file or library. (Object files are
9214 distinguished from libraries by the linker according to the file
9215 contents.) If linking is done, these object files are used as input
9224 If any of these options is used, then the linker is not run, and
9225 object file names should not be used as arguments. @xref{Overall
9229 @item -l@var{library}
9230 @itemx -l @var{library}
9232 Search the library named @var{library} when linking. (The second
9233 alternative with the library as a separate argument is only for
9234 POSIX compliance and is not recommended.)
9236 It makes a difference where in the command you write this option; the
9237 linker searches and processes libraries and object files in the order they
9238 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9239 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9240 to functions in @samp{z}, those functions may not be loaded.
9242 The linker searches a standard list of directories for the library,
9243 which is actually a file named @file{lib@var{library}.a}. The linker
9244 then uses this file as if it had been specified precisely by name.
9246 The directories searched include several standard system directories
9247 plus any that you specify with @option{-L}.
9249 Normally the files found this way are library files---archive files
9250 whose members are object files. The linker handles an archive file by
9251 scanning through it for members which define symbols that have so far
9252 been referenced but not defined. But if the file that is found is an
9253 ordinary object file, it is linked in the usual fashion. The only
9254 difference between using an @option{-l} option and specifying a file name
9255 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9256 and searches several directories.
9260 You need this special case of the @option{-l} option in order to
9261 link an Objective-C or Objective-C++ program.
9264 @opindex nostartfiles
9265 Do not use the standard system startup files when linking.
9266 The standard system libraries are used normally, unless @option{-nostdlib}
9267 or @option{-nodefaultlibs} is used.
9269 @item -nodefaultlibs
9270 @opindex nodefaultlibs
9271 Do not use the standard system libraries when linking.
9272 Only the libraries you specify will be passed to the linker, options
9273 specifying linkage of the system libraries, such as @code{-static-libgcc}
9274 or @code{-shared-libgcc}, will be ignored.
9275 The standard startup files are used normally, unless @option{-nostartfiles}
9276 is used. The compiler may generate calls to @code{memcmp},
9277 @code{memset}, @code{memcpy} and @code{memmove}.
9278 These entries are usually resolved by entries in
9279 libc. These entry points should be supplied through some other
9280 mechanism when this option is specified.
9284 Do not use the standard system startup files or libraries when linking.
9285 No startup files and only the libraries you specify will be passed to
9286 the linker, options specifying linkage of the system libraries, such as
9287 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9288 The compiler may generate calls to @code{memcmp}, @code{memset},
9289 @code{memcpy} and @code{memmove}.
9290 These entries are usually resolved by entries in
9291 libc. These entry points should be supplied through some other
9292 mechanism when this option is specified.
9294 @cindex @option{-lgcc}, use with @option{-nostdlib}
9295 @cindex @option{-nostdlib} and unresolved references
9296 @cindex unresolved references and @option{-nostdlib}
9297 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9298 @cindex @option{-nodefaultlibs} and unresolved references
9299 @cindex unresolved references and @option{-nodefaultlibs}
9300 One of the standard libraries bypassed by @option{-nostdlib} and
9301 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9302 that GCC uses to overcome shortcomings of particular machines, or special
9303 needs for some languages.
9304 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9305 Collection (GCC) Internals},
9306 for more discussion of @file{libgcc.a}.)
9307 In most cases, you need @file{libgcc.a} even when you want to avoid
9308 other standard libraries. In other words, when you specify @option{-nostdlib}
9309 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9310 This ensures that you have no unresolved references to internal GCC
9311 library subroutines. (For example, @samp{__main}, used to ensure C++
9312 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9313 GNU Compiler Collection (GCC) Internals}.)
9317 Produce a position independent executable on targets which support it.
9318 For predictable results, you must also specify the same set of options
9319 that were used to generate code (@option{-fpie}, @option{-fPIE},
9320 or model suboptions) when you specify this option.
9324 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9325 that support it. This instructs the linker to add all symbols, not
9326 only used ones, to the dynamic symbol table. This option is needed
9327 for some uses of @code{dlopen} or to allow obtaining backtraces
9328 from within a program.
9332 Remove all symbol table and relocation information from the executable.
9336 On systems that support dynamic linking, this prevents linking with the shared
9337 libraries. On other systems, this option has no effect.
9341 Produce a shared object which can then be linked with other objects to
9342 form an executable. Not all systems support this option. For predictable
9343 results, you must also specify the same set of options that were used to
9344 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9345 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9346 needs to build supplementary stub code for constructors to work. On
9347 multi-libbed systems, @samp{gcc -shared} must select the correct support
9348 libraries to link against. Failing to supply the correct flags may lead
9349 to subtle defects. Supplying them in cases where they are not necessary
9352 @item -shared-libgcc
9353 @itemx -static-libgcc
9354 @opindex shared-libgcc
9355 @opindex static-libgcc
9356 On systems that provide @file{libgcc} as a shared library, these options
9357 force the use of either the shared or static version respectively.
9358 If no shared version of @file{libgcc} was built when the compiler was
9359 configured, these options have no effect.
9361 There are several situations in which an application should use the
9362 shared @file{libgcc} instead of the static version. The most common
9363 of these is when the application wishes to throw and catch exceptions
9364 across different shared libraries. In that case, each of the libraries
9365 as well as the application itself should use the shared @file{libgcc}.
9367 Therefore, the G++ and GCJ drivers automatically add
9368 @option{-shared-libgcc} whenever you build a shared library or a main
9369 executable, because C++ and Java programs typically use exceptions, so
9370 this is the right thing to do.
9372 If, instead, you use the GCC driver to create shared libraries, you may
9373 find that they will not always be linked with the shared @file{libgcc}.
9374 If GCC finds, at its configuration time, that you have a non-GNU linker
9375 or a GNU linker that does not support option @option{--eh-frame-hdr},
9376 it will link the shared version of @file{libgcc} into shared libraries
9377 by default. Otherwise, it will take advantage of the linker and optimize
9378 away the linking with the shared version of @file{libgcc}, linking with
9379 the static version of libgcc by default. This allows exceptions to
9380 propagate through such shared libraries, without incurring relocation
9381 costs at library load time.
9383 However, if a library or main executable is supposed to throw or catch
9384 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9385 for the languages used in the program, or using the option
9386 @option{-shared-libgcc}, such that it is linked with the shared
9389 @item -static-libstdc++
9390 When the @command{g++} program is used to link a C++ program, it will
9391 normally automatically link against @option{libstdc++}. If
9392 @file{libstdc++} is available as a shared library, and the
9393 @option{-static} option is not used, then this will link against the
9394 shared version of @file{libstdc++}. That is normally fine. However, it
9395 is sometimes useful to freeze the version of @file{libstdc++} used by
9396 the program without going all the way to a fully static link. The
9397 @option{-static-libstdc++} option directs the @command{g++} driver to
9398 link @file{libstdc++} statically, without necessarily linking other
9399 libraries statically.
9403 Bind references to global symbols when building a shared object. Warn
9404 about any unresolved references (unless overridden by the link editor
9405 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9408 @item -T @var{script}
9410 @cindex linker script
9411 Use @var{script} as the linker script. This option is supported by most
9412 systems using the GNU linker. On some targets, such as bare-board
9413 targets without an operating system, the @option{-T} option may be required
9414 when linking to avoid references to undefined symbols.
9416 @item -Xlinker @var{option}
9418 Pass @var{option} as an option to the linker. You can use this to
9419 supply system-specific linker options which GCC does not know how to
9422 If you want to pass an option that takes a separate argument, you must use
9423 @option{-Xlinker} twice, once for the option and once for the argument.
9424 For example, to pass @option{-assert definitions}, you must write
9425 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9426 @option{-Xlinker "-assert definitions"}, because this passes the entire
9427 string as a single argument, which is not what the linker expects.
9429 When using the GNU linker, it is usually more convenient to pass
9430 arguments to linker options using the @option{@var{option}=@var{value}}
9431 syntax than as separate arguments. For example, you can specify
9432 @samp{-Xlinker -Map=output.map} rather than
9433 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9434 this syntax for command-line options.
9436 @item -Wl,@var{option}
9438 Pass @var{option} as an option to the linker. If @var{option} contains
9439 commas, it is split into multiple options at the commas. You can use this
9440 syntax to pass an argument to the option.
9441 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9442 linker. When using the GNU linker, you can also get the same effect with
9443 @samp{-Wl,-Map=output.map}.
9445 @item -u @var{symbol}
9447 Pretend the symbol @var{symbol} is undefined, to force linking of
9448 library modules to define it. You can use @option{-u} multiple times with
9449 different symbols to force loading of additional library modules.
9452 @node Directory Options
9453 @section Options for Directory Search
9454 @cindex directory options
9455 @cindex options, directory search
9458 These options specify directories to search for header files, for
9459 libraries and for parts of the compiler:
9464 Add the directory @var{dir} to the head of the list of directories to be
9465 searched for header files. This can be used to override a system header
9466 file, substituting your own version, since these directories are
9467 searched before the system header file directories. However, you should
9468 not use this option to add directories that contain vendor-supplied
9469 system header files (use @option{-isystem} for that). If you use more than
9470 one @option{-I} option, the directories are scanned in left-to-right
9471 order; the standard system directories come after.
9473 If a standard system include directory, or a directory specified with
9474 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9475 option will be ignored. The directory will still be searched but as a
9476 system directory at its normal position in the system include chain.
9477 This is to ensure that GCC's procedure to fix buggy system headers and
9478 the ordering for the include_next directive are not inadvertently changed.
9479 If you really need to change the search order for system directories,
9480 use the @option{-nostdinc} and/or @option{-isystem} options.
9482 @item -iplugindir=@var{dir}
9483 Set the directory to search for plugins which are passed
9484 by @option{-fplugin=@var{name}} instead of
9485 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9486 to be used by the user, but only passed by the driver.
9488 @item -iquote@var{dir}
9490 Add the directory @var{dir} to the head of the list of directories to
9491 be searched for header files only for the case of @samp{#include
9492 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9493 otherwise just like @option{-I}.
9497 Add directory @var{dir} to the list of directories to be searched
9500 @item -B@var{prefix}
9502 This option specifies where to find the executables, libraries,
9503 include files, and data files of the compiler itself.
9505 The compiler driver program runs one or more of the subprograms
9506 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9507 @var{prefix} as a prefix for each program it tries to run, both with and
9508 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9510 For each subprogram to be run, the compiler driver first tries the
9511 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9512 was not specified, the driver tries two standard prefixes, which are
9513 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9514 those results in a file name that is found, the unmodified program
9515 name is searched for using the directories specified in your
9516 @env{PATH} environment variable.
9518 The compiler will check to see if the path provided by the @option{-B}
9519 refers to a directory, and if necessary it will add a directory
9520 separator character at the end of the path.
9522 @option{-B} prefixes that effectively specify directory names also apply
9523 to libraries in the linker, because the compiler translates these
9524 options into @option{-L} options for the linker. They also apply to
9525 includes files in the preprocessor, because the compiler translates these
9526 options into @option{-isystem} options for the preprocessor. In this case,
9527 the compiler appends @samp{include} to the prefix.
9529 The run-time support file @file{libgcc.a} can also be searched for using
9530 the @option{-B} prefix, if needed. If it is not found there, the two
9531 standard prefixes above are tried, and that is all. The file is left
9532 out of the link if it is not found by those means.
9534 Another way to specify a prefix much like the @option{-B} prefix is to use
9535 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9538 As a special kludge, if the path provided by @option{-B} is
9539 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9540 9, then it will be replaced by @file{[dir/]include}. This is to help
9541 with boot-strapping the compiler.
9543 @item -specs=@var{file}
9545 Process @var{file} after the compiler reads in the standard @file{specs}
9546 file, in order to override the defaults that the @file{gcc} driver
9547 program uses when determining what switches to pass to @file{cc1},
9548 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9549 @option{-specs=@var{file}} can be specified on the command line, and they
9550 are processed in order, from left to right.
9552 @item --sysroot=@var{dir}
9554 Use @var{dir} as the logical root directory for headers and libraries.
9555 For example, if the compiler would normally search for headers in
9556 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9557 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9559 If you use both this option and the @option{-isysroot} option, then
9560 the @option{--sysroot} option will apply to libraries, but the
9561 @option{-isysroot} option will apply to header files.
9563 The GNU linker (beginning with version 2.16) has the necessary support
9564 for this option. If your linker does not support this option, the
9565 header file aspect of @option{--sysroot} will still work, but the
9566 library aspect will not.
9570 This option has been deprecated. Please use @option{-iquote} instead for
9571 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9572 Any directories you specify with @option{-I} options before the @option{-I-}
9573 option are searched only for the case of @samp{#include "@var{file}"};
9574 they are not searched for @samp{#include <@var{file}>}.
9576 If additional directories are specified with @option{-I} options after
9577 the @option{-I-}, these directories are searched for all @samp{#include}
9578 directives. (Ordinarily @emph{all} @option{-I} directories are used
9581 In addition, the @option{-I-} option inhibits the use of the current
9582 directory (where the current input file came from) as the first search
9583 directory for @samp{#include "@var{file}"}. There is no way to
9584 override this effect of @option{-I-}. With @option{-I.} you can specify
9585 searching the directory which was current when the compiler was
9586 invoked. That is not exactly the same as what the preprocessor does
9587 by default, but it is often satisfactory.
9589 @option{-I-} does not inhibit the use of the standard system directories
9590 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9597 @section Specifying subprocesses and the switches to pass to them
9600 @command{gcc} is a driver program. It performs its job by invoking a
9601 sequence of other programs to do the work of compiling, assembling and
9602 linking. GCC interprets its command-line parameters and uses these to
9603 deduce which programs it should invoke, and which command-line options
9604 it ought to place on their command lines. This behavior is controlled
9605 by @dfn{spec strings}. In most cases there is one spec string for each
9606 program that GCC can invoke, but a few programs have multiple spec
9607 strings to control their behavior. The spec strings built into GCC can
9608 be overridden by using the @option{-specs=} command-line switch to specify
9611 @dfn{Spec files} are plaintext files that are used to construct spec
9612 strings. They consist of a sequence of directives separated by blank
9613 lines. The type of directive is determined by the first non-whitespace
9614 character on the line and it can be one of the following:
9617 @item %@var{command}
9618 Issues a @var{command} to the spec file processor. The commands that can
9622 @item %include <@var{file}>
9623 @cindex @code{%include}
9624 Search for @var{file} and insert its text at the current point in the
9627 @item %include_noerr <@var{file}>
9628 @cindex @code{%include_noerr}
9629 Just like @samp{%include}, but do not generate an error message if the include
9630 file cannot be found.
9632 @item %rename @var{old_name} @var{new_name}
9633 @cindex @code{%rename}
9634 Rename the spec string @var{old_name} to @var{new_name}.
9638 @item *[@var{spec_name}]:
9639 This tells the compiler to create, override or delete the named spec
9640 string. All lines after this directive up to the next directive or
9641 blank line are considered to be the text for the spec string. If this
9642 results in an empty string then the spec will be deleted. (Or, if the
9643 spec did not exist, then nothing will happen.) Otherwise, if the spec
9644 does not currently exist a new spec will be created. If the spec does
9645 exist then its contents will be overridden by the text of this
9646 directive, unless the first character of that text is the @samp{+}
9647 character, in which case the text will be appended to the spec.
9649 @item [@var{suffix}]:
9650 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9651 and up to the next directive or blank line are considered to make up the
9652 spec string for the indicated suffix. When the compiler encounters an
9653 input file with the named suffix, it will processes the spec string in
9654 order to work out how to compile that file. For example:
9661 This says that any input file whose name ends in @samp{.ZZ} should be
9662 passed to the program @samp{z-compile}, which should be invoked with the
9663 command-line switch @option{-input} and with the result of performing the
9664 @samp{%i} substitution. (See below.)
9666 As an alternative to providing a spec string, the text that follows a
9667 suffix directive can be one of the following:
9670 @item @@@var{language}
9671 This says that the suffix is an alias for a known @var{language}. This is
9672 similar to using the @option{-x} command-line switch to GCC to specify a
9673 language explicitly. For example:
9680 Says that .ZZ files are, in fact, C++ source files.
9683 This causes an error messages saying:
9686 @var{name} compiler not installed on this system.
9690 GCC already has an extensive list of suffixes built into it.
9691 This directive will add an entry to the end of the list of suffixes, but
9692 since the list is searched from the end backwards, it is effectively
9693 possible to override earlier entries using this technique.
9697 GCC has the following spec strings built into it. Spec files can
9698 override these strings or create their own. Note that individual
9699 targets can also add their own spec strings to this list.
9702 asm Options to pass to the assembler
9703 asm_final Options to pass to the assembler post-processor
9704 cpp Options to pass to the C preprocessor
9705 cc1 Options to pass to the C compiler
9706 cc1plus Options to pass to the C++ compiler
9707 endfile Object files to include at the end of the link
9708 link Options to pass to the linker
9709 lib Libraries to include on the command line to the linker
9710 libgcc Decides which GCC support library to pass to the linker
9711 linker Sets the name of the linker
9712 predefines Defines to be passed to the C preprocessor
9713 signed_char Defines to pass to CPP to say whether @code{char} is signed
9715 startfile Object files to include at the start of the link
9718 Here is a small example of a spec file:
9724 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9727 This example renames the spec called @samp{lib} to @samp{old_lib} and
9728 then overrides the previous definition of @samp{lib} with a new one.
9729 The new definition adds in some extra command-line options before
9730 including the text of the old definition.
9732 @dfn{Spec strings} are a list of command-line options to be passed to their
9733 corresponding program. In addition, the spec strings can contain
9734 @samp{%}-prefixed sequences to substitute variable text or to
9735 conditionally insert text into the command line. Using these constructs
9736 it is possible to generate quite complex command lines.
9738 Here is a table of all defined @samp{%}-sequences for spec
9739 strings. Note that spaces are not generated automatically around the
9740 results of expanding these sequences. Therefore you can concatenate them
9741 together or combine them with constant text in a single argument.
9745 Substitute one @samp{%} into the program name or argument.
9748 Substitute the name of the input file being processed.
9751 Substitute the basename of the input file being processed.
9752 This is the substring up to (and not including) the last period
9753 and not including the directory.
9756 This is the same as @samp{%b}, but include the file suffix (text after
9760 Marks the argument containing or following the @samp{%d} as a
9761 temporary file name, so that that file will be deleted if GCC exits
9762 successfully. Unlike @samp{%g}, this contributes no text to the
9765 @item %g@var{suffix}
9766 Substitute a file name that has suffix @var{suffix} and is chosen
9767 once per compilation, and mark the argument in the same way as
9768 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9769 name is now chosen in a way that is hard to predict even when previously
9770 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9771 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9772 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9773 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9774 was simply substituted with a file name chosen once per compilation,
9775 without regard to any appended suffix (which was therefore treated
9776 just like ordinary text), making such attacks more likely to succeed.
9778 @item %u@var{suffix}
9779 Like @samp{%g}, but generates a new temporary file name even if
9780 @samp{%u@var{suffix}} was already seen.
9782 @item %U@var{suffix}
9783 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9784 new one if there is no such last file name. In the absence of any
9785 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9786 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9787 would involve the generation of two distinct file names, one
9788 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9789 simply substituted with a file name chosen for the previous @samp{%u},
9790 without regard to any appended suffix.
9792 @item %j@var{suffix}
9793 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9794 writable, and if save-temps is off; otherwise, substitute the name
9795 of a temporary file, just like @samp{%u}. This temporary file is not
9796 meant for communication between processes, but rather as a junk
9799 @item %|@var{suffix}
9800 @itemx %m@var{suffix}
9801 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9802 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9803 all. These are the two most common ways to instruct a program that it
9804 should read from standard input or write to standard output. If you
9805 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9806 construct: see for example @file{f/lang-specs.h}.
9808 @item %.@var{SUFFIX}
9809 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9810 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9811 terminated by the next space or %.
9814 Marks the argument containing or following the @samp{%w} as the
9815 designated output file of this compilation. This puts the argument
9816 into the sequence of arguments that @samp{%o} will substitute later.
9819 Substitutes the names of all the output files, with spaces
9820 automatically placed around them. You should write spaces
9821 around the @samp{%o} as well or the results are undefined.
9822 @samp{%o} is for use in the specs for running the linker.
9823 Input files whose names have no recognized suffix are not compiled
9824 at all, but they are included among the output files, so they will
9828 Substitutes the suffix for object files. Note that this is
9829 handled specially when it immediately follows @samp{%g, %u, or %U},
9830 because of the need for those to form complete file names. The
9831 handling is such that @samp{%O} is treated exactly as if it had already
9832 been substituted, except that @samp{%g, %u, and %U} do not currently
9833 support additional @var{suffix} characters following @samp{%O} as they would
9834 following, for example, @samp{.o}.
9837 Substitutes the standard macro predefinitions for the
9838 current target machine. Use this when running @code{cpp}.
9841 Like @samp{%p}, but puts @samp{__} before and after the name of each
9842 predefined macro, except for macros that start with @samp{__} or with
9843 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9847 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9848 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9849 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9850 and @option{-imultilib} as necessary.
9853 Current argument is the name of a library or startup file of some sort.
9854 Search for that file in a standard list of directories and substitute
9855 the full name found. The current working directory is included in the
9856 list of directories scanned.
9859 Current argument is the name of a linker script. Search for that file
9860 in the current list of directories to scan for libraries. If the file
9861 is located insert a @option{--script} option into the command line
9862 followed by the full path name found. If the file is not found then
9863 generate an error message. Note: the current working directory is not
9867 Print @var{str} as an error message. @var{str} is terminated by a newline.
9868 Use this when inconsistent options are detected.
9871 Substitute the contents of spec string @var{name} at this point.
9873 @item %x@{@var{option}@}
9874 Accumulate an option for @samp{%X}.
9877 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9881 Output the accumulated assembler options specified by @option{-Wa}.
9884 Output the accumulated preprocessor options specified by @option{-Wp}.
9887 Process the @code{asm} spec. This is used to compute the
9888 switches to be passed to the assembler.
9891 Process the @code{asm_final} spec. This is a spec string for
9892 passing switches to an assembler post-processor, if such a program is
9896 Process the @code{link} spec. This is the spec for computing the
9897 command line passed to the linker. Typically it will make use of the
9898 @samp{%L %G %S %D and %E} sequences.
9901 Dump out a @option{-L} option for each directory that GCC believes might
9902 contain startup files. If the target supports multilibs then the
9903 current multilib directory will be prepended to each of these paths.
9906 Process the @code{lib} spec. This is a spec string for deciding which
9907 libraries should be included on the command line to the linker.
9910 Process the @code{libgcc} spec. This is a spec string for deciding
9911 which GCC support library should be included on the command line to the linker.
9914 Process the @code{startfile} spec. This is a spec for deciding which
9915 object files should be the first ones passed to the linker. Typically
9916 this might be a file named @file{crt0.o}.
9919 Process the @code{endfile} spec. This is a spec string that specifies
9920 the last object files that will be passed to the linker.
9923 Process the @code{cpp} spec. This is used to construct the arguments
9924 to be passed to the C preprocessor.
9927 Process the @code{cc1} spec. This is used to construct the options to be
9928 passed to the actual C compiler (@samp{cc1}).
9931 Process the @code{cc1plus} spec. This is used to construct the options to be
9932 passed to the actual C++ compiler (@samp{cc1plus}).
9935 Substitute the variable part of a matched option. See below.
9936 Note that each comma in the substituted string is replaced by
9940 Remove all occurrences of @code{-S} from the command line. Note---this
9941 command is position dependent. @samp{%} commands in the spec string
9942 before this one will see @code{-S}, @samp{%} commands in the spec string
9943 after this one will not.
9945 @item %:@var{function}(@var{args})
9946 Call the named function @var{function}, passing it @var{args}.
9947 @var{args} is first processed as a nested spec string, then split
9948 into an argument vector in the usual fashion. The function returns
9949 a string which is processed as if it had appeared literally as part
9950 of the current spec.
9952 The following built-in spec functions are provided:
9956 The @code{getenv} spec function takes two arguments: an environment
9957 variable name and a string. If the environment variable is not
9958 defined, a fatal error is issued. Otherwise, the return value is the
9959 value of the environment variable concatenated with the string. For
9960 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9963 %:getenv(TOPDIR /include)
9966 expands to @file{/path/to/top/include}.
9968 @item @code{if-exists}
9969 The @code{if-exists} spec function takes one argument, an absolute
9970 pathname to a file. If the file exists, @code{if-exists} returns the
9971 pathname. Here is a small example of its usage:
9975 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9978 @item @code{if-exists-else}
9979 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9980 spec function, except that it takes two arguments. The first argument is
9981 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9982 returns the pathname. If it does not exist, it returns the second argument.
9983 This way, @code{if-exists-else} can be used to select one file or another,
9984 based on the existence of the first. Here is a small example of its usage:
9988 crt0%O%s %:if-exists(crti%O%s) \
9989 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9992 @item @code{replace-outfile}
9993 The @code{replace-outfile} spec function takes two arguments. It looks for the
9994 first argument in the outfiles array and replaces it with the second argument. Here
9995 is a small example of its usage:
9998 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
10001 @item @code{remove-outfile}
10002 The @code{remove-outfile} spec function takes one argument. It looks for the
10003 first argument in the outfiles array and removes it. Here is a small example
10007 %:remove-outfile(-lm)
10010 @item @code{pass-through-libs}
10011 The @code{pass-through-libs} spec function takes any number of arguments. It
10012 finds any @option{-l} options and any non-options ending in ".a" (which it
10013 assumes are the names of linker input library archive files) and returns a
10014 result containing all the found arguments each prepended by
10015 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
10016 intended to be passed to the LTO linker plugin.
10019 %:pass-through-libs(%G %L %G)
10022 @item @code{print-asm-header}
10023 The @code{print-asm-header} function takes no arguments and simply
10024 prints a banner like:
10030 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
10033 It is used to separate compiler options from assembler options
10034 in the @option{--target-help} output.
10037 @item %@{@code{S}@}
10038 Substitutes the @code{-S} switch, if that switch was given to GCC@.
10039 If that switch was not specified, this substitutes nothing. Note that
10040 the leading dash is omitted when specifying this option, and it is
10041 automatically inserted if the substitution is performed. Thus the spec
10042 string @samp{%@{foo@}} would match the command-line option @option{-foo}
10043 and would output the command line option @option{-foo}.
10045 @item %W@{@code{S}@}
10046 Like %@{@code{S}@} but mark last argument supplied within as a file to be
10047 deleted on failure.
10049 @item %@{@code{S}*@}
10050 Substitutes all the switches specified to GCC whose names start
10051 with @code{-S}, but which also take an argument. This is used for
10052 switches like @option{-o}, @option{-D}, @option{-I}, etc.
10053 GCC considers @option{-o foo} as being
10054 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
10055 text, including the space. Thus two arguments would be generated.
10057 @item %@{@code{S}*&@code{T}*@}
10058 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10059 (the order of @code{S} and @code{T} in the spec is not significant).
10060 There can be any number of ampersand-separated variables; for each the
10061 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10063 @item %@{@code{S}:@code{X}@}
10064 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10066 @item %@{!@code{S}:@code{X}@}
10067 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10069 @item %@{@code{S}*:@code{X}@}
10070 Substitutes @code{X} if one or more switches whose names start with
10071 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10072 once, no matter how many such switches appeared. However, if @code{%*}
10073 appears somewhere in @code{X}, then @code{X} will be substituted once
10074 for each matching switch, with the @code{%*} replaced by the part of
10075 that switch that matched the @code{*}.
10077 @item %@{.@code{S}:@code{X}@}
10078 Substitutes @code{X}, if processing a file with suffix @code{S}.
10080 @item %@{!.@code{S}:@code{X}@}
10081 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10083 @item %@{,@code{S}:@code{X}@}
10084 Substitutes @code{X}, if processing a file for language @code{S}.
10086 @item %@{!,@code{S}:@code{X}@}
10087 Substitutes @code{X}, if not processing a file for language @code{S}.
10089 @item %@{@code{S}|@code{P}:@code{X}@}
10090 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10091 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10092 @code{*} sequences as well, although they have a stronger binding than
10093 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10094 alternatives must be starred, and only the first matching alternative
10097 For example, a spec string like this:
10100 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10103 will output the following command-line options from the following input
10104 command-line options:
10109 -d fred.c -foo -baz -boggle
10110 -d jim.d -bar -baz -boggle
10113 @item %@{S:X; T:Y; :D@}
10115 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10116 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10117 be as many clauses as you need. This may be combined with @code{.},
10118 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10123 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10124 construct may contain other nested @samp{%} constructs or spaces, or
10125 even newlines. They are processed as usual, as described above.
10126 Trailing white space in @code{X} is ignored. White space may also
10127 appear anywhere on the left side of the colon in these constructs,
10128 except between @code{.} or @code{*} and the corresponding word.
10130 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10131 handled specifically in these constructs. If another value of
10132 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10133 @option{-W} switch is found later in the command line, the earlier
10134 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10135 just one letter, which passes all matching options.
10137 The character @samp{|} at the beginning of the predicate text is used to
10138 indicate that a command should be piped to the following command, but
10139 only if @option{-pipe} is specified.
10141 It is built into GCC which switches take arguments and which do not.
10142 (You might think it would be useful to generalize this to allow each
10143 compiler's spec to say which switches take arguments. But this cannot
10144 be done in a consistent fashion. GCC cannot even decide which input
10145 files have been specified without knowing which switches take arguments,
10146 and it must know which input files to compile in order to tell which
10149 GCC also knows implicitly that arguments starting in @option{-l} are to be
10150 treated as compiler output files, and passed to the linker in their
10151 proper position among the other output files.
10153 @c man begin OPTIONS
10155 @node Target Options
10156 @section Specifying Target Machine and Compiler Version
10157 @cindex target options
10158 @cindex cross compiling
10159 @cindex specifying machine version
10160 @cindex specifying compiler version and target machine
10161 @cindex compiler version, specifying
10162 @cindex target machine, specifying
10164 The usual way to run GCC is to run the executable called @command{gcc}, or
10165 @command{@var{machine}-gcc} when cross-compiling, or
10166 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10167 one that was installed last.
10169 @node Submodel Options
10170 @section Hardware Models and Configurations
10171 @cindex submodel options
10172 @cindex specifying hardware config
10173 @cindex hardware models and configurations, specifying
10174 @cindex machine dependent options
10176 Each target machine types can have its own
10177 special options, starting with @samp{-m}, to choose among various
10178 hardware models or configurations---for example, 68010 vs 68020,
10179 floating coprocessor or none. A single installed version of the
10180 compiler can compile for any model or configuration, according to the
10183 Some configurations of the compiler also support additional special
10184 options, usually for compatibility with other compilers on the same
10187 @c This list is ordered alphanumerically by subsection name.
10188 @c It should be the same order and spelling as these options are listed
10189 @c in Machine Dependent Options
10194 * Blackfin Options::
10198 * DEC Alpha Options::
10199 * DEC Alpha/VMS Options::
10202 * GNU/Linux Options::
10205 * i386 and x86-64 Options::
10206 * i386 and x86-64 Windows Options::
10208 * IA-64/VMS Options::
10215 * MicroBlaze Options::
10218 * MN10300 Options::
10220 * picoChip Options::
10221 * PowerPC Options::
10222 * RS/6000 and PowerPC Options::
10224 * S/390 and zSeries Options::
10227 * Solaris 2 Options::
10230 * System V Options::
10233 * VxWorks Options::
10235 * Xstormy16 Options::
10237 * zSeries Options::
10241 @subsection ARM Options
10242 @cindex ARM options
10244 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10248 @item -mabi=@var{name}
10250 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10251 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10254 @opindex mapcs-frame
10255 Generate a stack frame that is compliant with the ARM Procedure Call
10256 Standard for all functions, even if this is not strictly necessary for
10257 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10258 with this option will cause the stack frames not to be generated for
10259 leaf functions. The default is @option{-mno-apcs-frame}.
10263 This is a synonym for @option{-mapcs-frame}.
10266 @c not currently implemented
10267 @item -mapcs-stack-check
10268 @opindex mapcs-stack-check
10269 Generate code to check the amount of stack space available upon entry to
10270 every function (that actually uses some stack space). If there is
10271 insufficient space available then either the function
10272 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10273 called, depending upon the amount of stack space required. The run time
10274 system is required to provide these functions. The default is
10275 @option{-mno-apcs-stack-check}, since this produces smaller code.
10277 @c not currently implemented
10279 @opindex mapcs-float
10280 Pass floating point arguments using the float point registers. This is
10281 one of the variants of the APCS@. This option is recommended if the
10282 target hardware has a floating point unit or if a lot of floating point
10283 arithmetic is going to be performed by the code. The default is
10284 @option{-mno-apcs-float}, since integer only code is slightly increased in
10285 size if @option{-mapcs-float} is used.
10287 @c not currently implemented
10288 @item -mapcs-reentrant
10289 @opindex mapcs-reentrant
10290 Generate reentrant, position independent code. The default is
10291 @option{-mno-apcs-reentrant}.
10294 @item -mthumb-interwork
10295 @opindex mthumb-interwork
10296 Generate code which supports calling between the ARM and Thumb
10297 instruction sets. Without this option, on pre-v5 architectures, the
10298 two instruction sets cannot be reliably used inside one program. The
10299 default is @option{-mno-thumb-interwork}, since slightly larger code
10300 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10301 configurations this option is meaningless.
10303 @item -mno-sched-prolog
10304 @opindex mno-sched-prolog
10305 Prevent the reordering of instructions in the function prolog, or the
10306 merging of those instruction with the instructions in the function's
10307 body. This means that all functions will start with a recognizable set
10308 of instructions (or in fact one of a choice from a small set of
10309 different function prologues), and this information can be used to
10310 locate the start if functions inside an executable piece of code. The
10311 default is @option{-msched-prolog}.
10313 @item -mfloat-abi=@var{name}
10314 @opindex mfloat-abi
10315 Specifies which floating-point ABI to use. Permissible values
10316 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10318 Specifying @samp{soft} causes GCC to generate output containing
10319 library calls for floating-point operations.
10320 @samp{softfp} allows the generation of code using hardware floating-point
10321 instructions, but still uses the soft-float calling conventions.
10322 @samp{hard} allows generation of floating-point instructions
10323 and uses FPU-specific calling conventions.
10325 The default depends on the specific target configuration. Note that
10326 the hard-float and soft-float ABIs are not link-compatible; you must
10327 compile your entire program with the same ABI, and link with a
10328 compatible set of libraries.
10330 @item -mlittle-endian
10331 @opindex mlittle-endian
10332 Generate code for a processor running in little-endian mode. This is
10333 the default for all standard configurations.
10336 @opindex mbig-endian
10337 Generate code for a processor running in big-endian mode; the default is
10338 to compile code for a little-endian processor.
10340 @item -mwords-little-endian
10341 @opindex mwords-little-endian
10342 This option only applies when generating code for big-endian processors.
10343 Generate code for a little-endian word order but a big-endian byte
10344 order. That is, a byte order of the form @samp{32107654}. Note: this
10345 option should only be used if you require compatibility with code for
10346 big-endian ARM processors generated by versions of the compiler prior to
10347 2.8. This option is now deprecated.
10349 @item -mcpu=@var{name}
10351 This specifies the name of the target ARM processor. GCC uses this name
10352 to determine what kind of instructions it can emit when generating
10353 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10354 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10355 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10356 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10357 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10359 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10360 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10361 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10362 @samp{strongarm1110},
10363 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10364 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10365 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10366 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10367 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10368 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10369 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10370 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10371 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10372 @samp{cortex-m4}, @samp{cortex-m3},
10375 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10376 @samp{fa526}, @samp{fa626},
10377 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10379 @option{-mcpu=generic-@var{arch}} is also permissible, and is
10380 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
10381 See @option{-mtune} for more information.
10383 @item -mtune=@var{name}
10385 This option is very similar to the @option{-mcpu=} option, except that
10386 instead of specifying the actual target processor type, and hence
10387 restricting which instructions can be used, it specifies that GCC should
10388 tune the performance of the code as if the target were of the type
10389 specified in this option, but still choosing the instructions that it
10390 will generate based on the CPU specified by a @option{-mcpu=} option.
10391 For some ARM implementations better performance can be obtained by using
10394 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
10395 performance for a blend of processors within architecture @var{arch}.
10396 The aim is to generate code that run well on the current most popular
10397 processors, balancing between optimizations that benefit some CPUs in the
10398 range, and avoiding performance pitfalls of other CPUs. The effects of
10399 this option may change in future GCC versions as CPU models come and go.
10401 @item -march=@var{name}
10403 This specifies the name of the target ARM architecture. GCC uses this
10404 name to determine what kind of instructions it can emit when generating
10405 assembly code. This option can be used in conjunction with or instead
10406 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10407 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10408 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10409 @samp{armv6}, @samp{armv6j},
10410 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10411 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10412 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10414 @item -mfpu=@var{name}
10415 @itemx -mfpe=@var{number}
10416 @itemx -mfp=@var{number}
10420 This specifies what floating point hardware (or hardware emulation) is
10421 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10422 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10423 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10424 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10425 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10426 @option{-mfp} and @option{-mfpe} are synonyms for
10427 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10430 If @option{-msoft-float} is specified this specifies the format of
10431 floating point values.
10433 If the selected floating-point hardware includes the NEON extension
10434 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10435 operations will not be used by GCC's auto-vectorization pass unless
10436 @option{-funsafe-math-optimizations} is also specified. This is
10437 because NEON hardware does not fully implement the IEEE 754 standard for
10438 floating-point arithmetic (in particular denormal values are treated as
10439 zero), so the use of NEON instructions may lead to a loss of precision.
10441 @item -mfp16-format=@var{name}
10442 @opindex mfp16-format
10443 Specify the format of the @code{__fp16} half-precision floating-point type.
10444 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10445 the default is @samp{none}, in which case the @code{__fp16} type is not
10446 defined. @xref{Half-Precision}, for more information.
10448 @item -mstructure-size-boundary=@var{n}
10449 @opindex mstructure-size-boundary
10450 The size of all structures and unions will be rounded up to a multiple
10451 of the number of bits set by this option. Permissible values are 8, 32
10452 and 64. The default value varies for different toolchains. For the COFF
10453 targeted toolchain the default value is 8. A value of 64 is only allowed
10454 if the underlying ABI supports it.
10456 Specifying the larger number can produce faster, more efficient code, but
10457 can also increase the size of the program. Different values are potentially
10458 incompatible. Code compiled with one value cannot necessarily expect to
10459 work with code or libraries compiled with another value, if they exchange
10460 information using structures or unions.
10462 @item -mabort-on-noreturn
10463 @opindex mabort-on-noreturn
10464 Generate a call to the function @code{abort} at the end of a
10465 @code{noreturn} function. It will be executed if the function tries to
10469 @itemx -mno-long-calls
10470 @opindex mlong-calls
10471 @opindex mno-long-calls
10472 Tells the compiler to perform function calls by first loading the
10473 address of the function into a register and then performing a subroutine
10474 call on this register. This switch is needed if the target function
10475 will lie outside of the 64 megabyte addressing range of the offset based
10476 version of subroutine call instruction.
10478 Even if this switch is enabled, not all function calls will be turned
10479 into long calls. The heuristic is that static functions, functions
10480 which have the @samp{short-call} attribute, functions that are inside
10481 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10482 definitions have already been compiled within the current compilation
10483 unit, will not be turned into long calls. The exception to this rule is
10484 that weak function definitions, functions with the @samp{long-call}
10485 attribute or the @samp{section} attribute, and functions that are within
10486 the scope of a @samp{#pragma long_calls} directive, will always be
10487 turned into long calls.
10489 This feature is not enabled by default. Specifying
10490 @option{-mno-long-calls} will restore the default behavior, as will
10491 placing the function calls within the scope of a @samp{#pragma
10492 long_calls_off} directive. Note these switches have no effect on how
10493 the compiler generates code to handle function calls via function
10496 @item -msingle-pic-base
10497 @opindex msingle-pic-base
10498 Treat the register used for PIC addressing as read-only, rather than
10499 loading it in the prologue for each function. The run-time system is
10500 responsible for initializing this register with an appropriate value
10501 before execution begins.
10503 @item -mpic-register=@var{reg}
10504 @opindex mpic-register
10505 Specify the register to be used for PIC addressing. The default is R10
10506 unless stack-checking is enabled, when R9 is used.
10508 @item -mcirrus-fix-invalid-insns
10509 @opindex mcirrus-fix-invalid-insns
10510 @opindex mno-cirrus-fix-invalid-insns
10511 Insert NOPs into the instruction stream to in order to work around
10512 problems with invalid Maverick instruction combinations. This option
10513 is only valid if the @option{-mcpu=ep9312} option has been used to
10514 enable generation of instructions for the Cirrus Maverick floating
10515 point co-processor. This option is not enabled by default, since the
10516 problem is only present in older Maverick implementations. The default
10517 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10520 @item -mpoke-function-name
10521 @opindex mpoke-function-name
10522 Write the name of each function into the text section, directly
10523 preceding the function prologue. The generated code is similar to this:
10527 .ascii "arm_poke_function_name", 0
10530 .word 0xff000000 + (t1 - t0)
10531 arm_poke_function_name
10533 stmfd sp!, @{fp, ip, lr, pc@}
10537 When performing a stack backtrace, code can inspect the value of
10538 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10539 location @code{pc - 12} and the top 8 bits are set, then we know that
10540 there is a function name embedded immediately preceding this location
10541 and has length @code{((pc[-3]) & 0xff000000)}.
10548 Select between generating code that executes in ARM and Thumb
10549 states. The default for most configurations is to generate code
10550 that executes in ARM state, but the default can be changed by
10551 configuring GCC with the @option{--with-mode=}@var{state}
10555 @opindex mtpcs-frame
10556 Generate a stack frame that is compliant with the Thumb Procedure Call
10557 Standard for all non-leaf functions. (A leaf function is one that does
10558 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10560 @item -mtpcs-leaf-frame
10561 @opindex mtpcs-leaf-frame
10562 Generate a stack frame that is compliant with the Thumb Procedure Call
10563 Standard for all leaf functions. (A leaf function is one that does
10564 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10566 @item -mcallee-super-interworking
10567 @opindex mcallee-super-interworking
10568 Gives all externally visible functions in the file being compiled an ARM
10569 instruction set header which switches to Thumb mode before executing the
10570 rest of the function. This allows these functions to be called from
10571 non-interworking code. This option is not valid in AAPCS configurations
10572 because interworking is enabled by default.
10574 @item -mcaller-super-interworking
10575 @opindex mcaller-super-interworking
10576 Allows calls via function pointers (including virtual functions) to
10577 execute correctly regardless of whether the target code has been
10578 compiled for interworking or not. There is a small overhead in the cost
10579 of executing a function pointer if this option is enabled. This option
10580 is not valid in AAPCS configurations because interworking is enabled
10583 @item -mtp=@var{name}
10585 Specify the access model for the thread local storage pointer. The valid
10586 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10587 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10588 (supported in the arm6k architecture), and @option{auto}, which uses the
10589 best available method for the selected processor. The default setting is
10592 @item -mtls-dialect=@var{dialect}
10593 @opindex mtls-dialect
10594 Specify the dialect to use for accessing thread local storage. Two
10595 dialects are supported --- @option{gnu} and @option{gnu2}. The
10596 @option{gnu} dialect selects the original GNU scheme for supporting
10597 local and global dynamic TLS models. The @option{gnu2} dialect
10598 selects the GNU descriptor scheme, which provides better performance
10599 for shared libraries. The GNU descriptor scheme is compatible with
10600 the original scheme, but does require new assembler, linker and
10601 library support. Initial and local exec TLS models are unaffected by
10602 this option and always use the original scheme.
10604 @item -mword-relocations
10605 @opindex mword-relocations
10606 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10607 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10608 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10611 @item -mfix-cortex-m3-ldrd
10612 @opindex mfix-cortex-m3-ldrd
10613 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10614 with overlapping destination and base registers are used. This option avoids
10615 generating these instructions. This option is enabled by default when
10616 @option{-mcpu=cortex-m3} is specified.
10621 @subsection AVR Options
10622 @cindex AVR Options
10624 These options are defined for AVR implementations:
10627 @item -mmcu=@var{mcu}
10629 Specify ATMEL AVR instruction set or MCU type.
10631 Instruction set avr1 is for the minimal AVR core, not supported by the C
10632 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10633 attiny11, attiny12, attiny15, attiny28).
10635 Instruction set avr2 (default) is for the classic AVR core with up to
10636 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10637 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10638 at90c8534, at90s8535).
10640 Instruction set avr3 is for the classic AVR core with up to 128K program
10641 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10643 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10644 memory space (MCU types: atmega8, atmega83, atmega85).
10646 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10647 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10648 atmega64, atmega128, at43usb355, at94k).
10650 @item -mno-interrupts
10651 @opindex mno-interrupts
10652 Generated code is not compatible with hardware interrupts.
10653 Code size will be smaller.
10655 @item -mcall-prologues
10656 @opindex mcall-prologues
10657 Functions prologues/epilogues expanded as call to appropriate
10658 subroutines. Code size will be smaller.
10661 @opindex mtiny-stack
10662 Change only the low 8 bits of the stack pointer.
10666 Assume int to be 8 bit integer. This affects the sizes of all types: A
10667 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10668 and long long will be 4 bytes. Please note that this option does not
10669 comply to the C standards, but it will provide you with smaller code
10673 @node Blackfin Options
10674 @subsection Blackfin Options
10675 @cindex Blackfin Options
10678 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10680 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10681 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10682 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10683 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10684 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10685 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10686 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10687 @samp{bf561}, @samp{bf592}.
10688 The optional @var{sirevision} specifies the silicon revision of the target
10689 Blackfin processor. Any workarounds available for the targeted silicon revision
10690 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10691 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10692 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10693 hexadecimal digits representing the major and minor numbers in the silicon
10694 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10695 is not defined. If @var{sirevision} is @samp{any}, the
10696 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10697 If this optional @var{sirevision} is not used, GCC assumes the latest known
10698 silicon revision of the targeted Blackfin processor.
10700 Support for @samp{bf561} is incomplete. For @samp{bf561},
10701 Only the processor macro is defined.
10702 Without this option, @samp{bf532} is used as the processor by default.
10703 The corresponding predefined processor macros for @var{cpu} is to
10704 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10705 provided by libgloss to be linked in if @option{-msim} is not given.
10709 Specifies that the program will be run on the simulator. This causes
10710 the simulator BSP provided by libgloss to be linked in. This option
10711 has effect only for @samp{bfin-elf} toolchain.
10712 Certain other options, such as @option{-mid-shared-library} and
10713 @option{-mfdpic}, imply @option{-msim}.
10715 @item -momit-leaf-frame-pointer
10716 @opindex momit-leaf-frame-pointer
10717 Don't keep the frame pointer in a register for leaf functions. This
10718 avoids the instructions to save, set up and restore frame pointers and
10719 makes an extra register available in leaf functions. The option
10720 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10721 which might make debugging harder.
10723 @item -mspecld-anomaly
10724 @opindex mspecld-anomaly
10725 When enabled, the compiler will ensure that the generated code does not
10726 contain speculative loads after jump instructions. If this option is used,
10727 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10729 @item -mno-specld-anomaly
10730 @opindex mno-specld-anomaly
10731 Don't generate extra code to prevent speculative loads from occurring.
10733 @item -mcsync-anomaly
10734 @opindex mcsync-anomaly
10735 When enabled, the compiler will ensure that the generated code does not
10736 contain CSYNC or SSYNC instructions too soon after conditional branches.
10737 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10739 @item -mno-csync-anomaly
10740 @opindex mno-csync-anomaly
10741 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10742 occurring too soon after a conditional branch.
10746 When enabled, the compiler is free to take advantage of the knowledge that
10747 the entire program fits into the low 64k of memory.
10750 @opindex mno-low-64k
10751 Assume that the program is arbitrarily large. This is the default.
10753 @item -mstack-check-l1
10754 @opindex mstack-check-l1
10755 Do stack checking using information placed into L1 scratchpad memory by the
10758 @item -mid-shared-library
10759 @opindex mid-shared-library
10760 Generate code that supports shared libraries via the library ID method.
10761 This allows for execute in place and shared libraries in an environment
10762 without virtual memory management. This option implies @option{-fPIC}.
10763 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10765 @item -mno-id-shared-library
10766 @opindex mno-id-shared-library
10767 Generate code that doesn't assume ID based shared libraries are being used.
10768 This is the default.
10770 @item -mleaf-id-shared-library
10771 @opindex mleaf-id-shared-library
10772 Generate code that supports shared libraries via the library ID method,
10773 but assumes that this library or executable won't link against any other
10774 ID shared libraries. That allows the compiler to use faster code for jumps
10777 @item -mno-leaf-id-shared-library
10778 @opindex mno-leaf-id-shared-library
10779 Do not assume that the code being compiled won't link against any ID shared
10780 libraries. Slower code will be generated for jump and call insns.
10782 @item -mshared-library-id=n
10783 @opindex mshared-library-id
10784 Specified the identification number of the ID based shared library being
10785 compiled. Specifying a value of 0 will generate more compact code, specifying
10786 other values will force the allocation of that number to the current
10787 library but is no more space or time efficient than omitting this option.
10791 Generate code that allows the data segment to be located in a different
10792 area of memory from the text segment. This allows for execute in place in
10793 an environment without virtual memory management by eliminating relocations
10794 against the text section.
10796 @item -mno-sep-data
10797 @opindex mno-sep-data
10798 Generate code that assumes that the data segment follows the text segment.
10799 This is the default.
10802 @itemx -mno-long-calls
10803 @opindex mlong-calls
10804 @opindex mno-long-calls
10805 Tells the compiler to perform function calls by first loading the
10806 address of the function into a register and then performing a subroutine
10807 call on this register. This switch is needed if the target function
10808 will lie outside of the 24 bit addressing range of the offset based
10809 version of subroutine call instruction.
10811 This feature is not enabled by default. Specifying
10812 @option{-mno-long-calls} will restore the default behavior. Note these
10813 switches have no effect on how the compiler generates code to handle
10814 function calls via function pointers.
10818 Link with the fast floating-point library. This library relaxes some of
10819 the IEEE floating-point standard's rules for checking inputs against
10820 Not-a-Number (NAN), in the interest of performance.
10823 @opindex minline-plt
10824 Enable inlining of PLT entries in function calls to functions that are
10825 not known to bind locally. It has no effect without @option{-mfdpic}.
10828 @opindex mmulticore
10829 Build standalone application for multicore Blackfin processor. Proper
10830 start files and link scripts will be used to support multicore.
10831 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10832 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10833 @option{-mcorea} or @option{-mcoreb}. If it's used without
10834 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10835 programming model is used. In this model, the main function of Core B
10836 should be named as coreb_main. If it's used with @option{-mcorea} or
10837 @option{-mcoreb}, one application per core programming model is used.
10838 If this option is not used, single core application programming
10843 Build standalone application for Core A of BF561 when using
10844 one application per core programming model. Proper start files
10845 and link scripts will be used to support Core A. This option
10846 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10850 Build standalone application for Core B of BF561 when using
10851 one application per core programming model. Proper start files
10852 and link scripts will be used to support Core B. This option
10853 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10854 should be used instead of main. It must be used with
10855 @option{-mmulticore}.
10859 Build standalone application for SDRAM. Proper start files and
10860 link scripts will be used to put the application into SDRAM.
10861 Loader should initialize SDRAM before loading the application
10862 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10866 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10867 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10868 are enabled; for standalone applications the default is off.
10872 @subsection C6X Options
10873 @cindex C6X Options
10876 @item -march=@var{name}
10878 This specifies the name of the target architecture. GCC uses this
10879 name to determine what kind of instructions it can emit when generating
10880 assembly code. Permissible names are: @samp{c62x},
10881 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
10884 @opindex mbig-endian
10885 Generate code for a big endian target.
10887 @item -mlittle-endian
10888 @opindex mlittle-endian
10889 Generate code for a little endian target. This is the default.
10893 Choose startup files and linker script suitable for the simulator.
10895 @item -msdata=default
10896 @opindex msdata=default
10897 Put small global and static data in the @samp{.neardata} section,
10898 which is pointed to by register @code{B14}. Put small uninitialized
10899 global and static data in the @samp{.bss} section, which is adjacent
10900 to the @samp{.neardata} section. Put small read-only data into the
10901 @samp{.rodata} section. The corresponding sections used for large
10902 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
10905 @opindex msdata=all
10906 Put all data, not just small objets, into the sections reserved for
10907 small data, and use addressing relative to the @code{B14} register to
10911 @opindex msdata=none
10912 Make no use of the sections reserved for small data, and use absolute
10913 addresses to access all data. Put all initialized global and static
10914 data in the @samp{.fardata} section, and all uninitialized data in the
10915 @samp{.far} section. Put all constant data into the @samp{.const}
10920 @subsection CRIS Options
10921 @cindex CRIS Options
10923 These options are defined specifically for the CRIS ports.
10926 @item -march=@var{architecture-type}
10927 @itemx -mcpu=@var{architecture-type}
10930 Generate code for the specified architecture. The choices for
10931 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10932 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10933 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10936 @item -mtune=@var{architecture-type}
10938 Tune to @var{architecture-type} everything applicable about the generated
10939 code, except for the ABI and the set of available instructions. The
10940 choices for @var{architecture-type} are the same as for
10941 @option{-march=@var{architecture-type}}.
10943 @item -mmax-stack-frame=@var{n}
10944 @opindex mmax-stack-frame
10945 Warn when the stack frame of a function exceeds @var{n} bytes.
10951 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10952 @option{-march=v3} and @option{-march=v8} respectively.
10954 @item -mmul-bug-workaround
10955 @itemx -mno-mul-bug-workaround
10956 @opindex mmul-bug-workaround
10957 @opindex mno-mul-bug-workaround
10958 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10959 models where it applies. This option is active by default.
10963 Enable CRIS-specific verbose debug-related information in the assembly
10964 code. This option also has the effect to turn off the @samp{#NO_APP}
10965 formatted-code indicator to the assembler at the beginning of the
10970 Do not use condition-code results from previous instruction; always emit
10971 compare and test instructions before use of condition codes.
10973 @item -mno-side-effects
10974 @opindex mno-side-effects
10975 Do not emit instructions with side-effects in addressing modes other than
10978 @item -mstack-align
10979 @itemx -mno-stack-align
10980 @itemx -mdata-align
10981 @itemx -mno-data-align
10982 @itemx -mconst-align
10983 @itemx -mno-const-align
10984 @opindex mstack-align
10985 @opindex mno-stack-align
10986 @opindex mdata-align
10987 @opindex mno-data-align
10988 @opindex mconst-align
10989 @opindex mno-const-align
10990 These options (no-options) arranges (eliminate arrangements) for the
10991 stack-frame, individual data and constants to be aligned for the maximum
10992 single data access size for the chosen CPU model. The default is to
10993 arrange for 32-bit alignment. ABI details such as structure layout are
10994 not affected by these options.
11002 Similar to the stack- data- and const-align options above, these options
11003 arrange for stack-frame, writable data and constants to all be 32-bit,
11004 16-bit or 8-bit aligned. The default is 32-bit alignment.
11006 @item -mno-prologue-epilogue
11007 @itemx -mprologue-epilogue
11008 @opindex mno-prologue-epilogue
11009 @opindex mprologue-epilogue
11010 With @option{-mno-prologue-epilogue}, the normal function prologue and
11011 epilogue that sets up the stack-frame are omitted and no return
11012 instructions or return sequences are generated in the code. Use this
11013 option only together with visual inspection of the compiled code: no
11014 warnings or errors are generated when call-saved registers must be saved,
11015 or storage for local variable needs to be allocated.
11019 @opindex mno-gotplt
11021 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
11022 instruction sequences that load addresses for functions from the PLT part
11023 of the GOT rather than (traditional on other architectures) calls to the
11024 PLT@. The default is @option{-mgotplt}.
11028 Legacy no-op option only recognized with the cris-axis-elf and
11029 cris-axis-linux-gnu targets.
11033 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
11037 This option, recognized for the cris-axis-elf arranges
11038 to link with input-output functions from a simulator library. Code,
11039 initialized data and zero-initialized data are allocated consecutively.
11043 Like @option{-sim}, but pass linker options to locate initialized data at
11044 0x40000000 and zero-initialized data at 0x80000000.
11047 @node Darwin Options
11048 @subsection Darwin Options
11049 @cindex Darwin options
11051 These options are defined for all architectures running the Darwin operating
11054 FSF GCC on Darwin does not create ``fat'' object files; it will create
11055 an object file for the single architecture that it was built to
11056 target. Apple's GCC on Darwin does create ``fat'' files if multiple
11057 @option{-arch} options are used; it does so by running the compiler or
11058 linker multiple times and joining the results together with
11061 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
11062 @samp{i686}) is determined by the flags that specify the ISA
11063 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
11064 @option{-force_cpusubtype_ALL} option can be used to override this.
11066 The Darwin tools vary in their behavior when presented with an ISA
11067 mismatch. The assembler, @file{as}, will only permit instructions to
11068 be used that are valid for the subtype of the file it is generating,
11069 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11070 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11071 and print an error if asked to create a shared library with a less
11072 restrictive subtype than its input files (for instance, trying to put
11073 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11074 for executables, @file{ld}, will quietly give the executable the most
11075 restrictive subtype of any of its input files.
11080 Add the framework directory @var{dir} to the head of the list of
11081 directories to be searched for header files. These directories are
11082 interleaved with those specified by @option{-I} options and are
11083 scanned in a left-to-right order.
11085 A framework directory is a directory with frameworks in it. A
11086 framework is a directory with a @samp{"Headers"} and/or
11087 @samp{"PrivateHeaders"} directory contained directly in it that ends
11088 in @samp{".framework"}. The name of a framework is the name of this
11089 directory excluding the @samp{".framework"}. Headers associated with
11090 the framework are found in one of those two directories, with
11091 @samp{"Headers"} being searched first. A subframework is a framework
11092 directory that is in a framework's @samp{"Frameworks"} directory.
11093 Includes of subframework headers can only appear in a header of a
11094 framework that contains the subframework, or in a sibling subframework
11095 header. Two subframeworks are siblings if they occur in the same
11096 framework. A subframework should not have the same name as a
11097 framework, a warning will be issued if this is violated. Currently a
11098 subframework cannot have subframeworks, in the future, the mechanism
11099 may be extended to support this. The standard frameworks can be found
11100 in @samp{"/System/Library/Frameworks"} and
11101 @samp{"/Library/Frameworks"}. An example include looks like
11102 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11103 the name of the framework and header.h is found in the
11104 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11106 @item -iframework@var{dir}
11107 @opindex iframework
11108 Like @option{-F} except the directory is a treated as a system
11109 directory. The main difference between this @option{-iframework} and
11110 @option{-F} is that with @option{-iframework} the compiler does not
11111 warn about constructs contained within header files found via
11112 @var{dir}. This option is valid only for the C family of languages.
11116 Emit debugging information for symbols that are used. For STABS
11117 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11118 This is by default ON@.
11122 Emit debugging information for all symbols and types.
11124 @item -mmacosx-version-min=@var{version}
11125 The earliest version of MacOS X that this executable will run on
11126 is @var{version}. Typical values of @var{version} include @code{10.1},
11127 @code{10.2}, and @code{10.3.9}.
11129 If the compiler was built to use the system's headers by default,
11130 then the default for this option is the system version on which the
11131 compiler is running, otherwise the default is to make choices which
11132 are compatible with as many systems and code bases as possible.
11136 Enable kernel development mode. The @option{-mkernel} option sets
11137 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11138 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11139 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11140 applicable. This mode also sets @option{-mno-altivec},
11141 @option{-msoft-float}, @option{-fno-builtin} and
11142 @option{-mlong-branch} for PowerPC targets.
11144 @item -mone-byte-bool
11145 @opindex mone-byte-bool
11146 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11147 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11148 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11149 option has no effect on x86.
11151 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11152 to generate code that is not binary compatible with code generated
11153 without that switch. Using this switch may require recompiling all
11154 other modules in a program, including system libraries. Use this
11155 switch to conform to a non-default data model.
11157 @item -mfix-and-continue
11158 @itemx -ffix-and-continue
11159 @itemx -findirect-data
11160 @opindex mfix-and-continue
11161 @opindex ffix-and-continue
11162 @opindex findirect-data
11163 Generate code suitable for fast turn around development. Needed to
11164 enable gdb to dynamically load @code{.o} files into already running
11165 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11166 are provided for backwards compatibility.
11170 Loads all members of static archive libraries.
11171 See man ld(1) for more information.
11173 @item -arch_errors_fatal
11174 @opindex arch_errors_fatal
11175 Cause the errors having to do with files that have the wrong architecture
11178 @item -bind_at_load
11179 @opindex bind_at_load
11180 Causes the output file to be marked such that the dynamic linker will
11181 bind all undefined references when the file is loaded or launched.
11185 Produce a Mach-o bundle format file.
11186 See man ld(1) for more information.
11188 @item -bundle_loader @var{executable}
11189 @opindex bundle_loader
11190 This option specifies the @var{executable} that will be loading the build
11191 output file being linked. See man ld(1) for more information.
11194 @opindex dynamiclib
11195 When passed this option, GCC will produce a dynamic library instead of
11196 an executable when linking, using the Darwin @file{libtool} command.
11198 @item -force_cpusubtype_ALL
11199 @opindex force_cpusubtype_ALL
11200 This causes GCC's output file to have the @var{ALL} subtype, instead of
11201 one controlled by the @option{-mcpu} or @option{-march} option.
11203 @item -allowable_client @var{client_name}
11204 @itemx -client_name
11205 @itemx -compatibility_version
11206 @itemx -current_version
11208 @itemx -dependency-file
11210 @itemx -dylinker_install_name
11212 @itemx -exported_symbols_list
11215 @itemx -flat_namespace
11216 @itemx -force_flat_namespace
11217 @itemx -headerpad_max_install_names
11220 @itemx -install_name
11221 @itemx -keep_private_externs
11222 @itemx -multi_module
11223 @itemx -multiply_defined
11224 @itemx -multiply_defined_unused
11227 @itemx -no_dead_strip_inits_and_terms
11228 @itemx -nofixprebinding
11229 @itemx -nomultidefs
11231 @itemx -noseglinkedit
11232 @itemx -pagezero_size
11234 @itemx -prebind_all_twolevel_modules
11235 @itemx -private_bundle
11237 @itemx -read_only_relocs
11239 @itemx -sectobjectsymbols
11243 @itemx -sectobjectsymbols
11246 @itemx -segs_read_only_addr
11248 @itemx -segs_read_write_addr
11249 @itemx -seg_addr_table
11250 @itemx -seg_addr_table_filename
11251 @itemx -seglinkedit
11253 @itemx -segs_read_only_addr
11254 @itemx -segs_read_write_addr
11255 @itemx -single_module
11257 @itemx -sub_library
11259 @itemx -sub_umbrella
11260 @itemx -twolevel_namespace
11263 @itemx -unexported_symbols_list
11264 @itemx -weak_reference_mismatches
11265 @itemx -whatsloaded
11266 @opindex allowable_client
11267 @opindex client_name
11268 @opindex compatibility_version
11269 @opindex current_version
11270 @opindex dead_strip
11271 @opindex dependency-file
11272 @opindex dylib_file
11273 @opindex dylinker_install_name
11275 @opindex exported_symbols_list
11277 @opindex flat_namespace
11278 @opindex force_flat_namespace
11279 @opindex headerpad_max_install_names
11280 @opindex image_base
11282 @opindex install_name
11283 @opindex keep_private_externs
11284 @opindex multi_module
11285 @opindex multiply_defined
11286 @opindex multiply_defined_unused
11287 @opindex noall_load
11288 @opindex no_dead_strip_inits_and_terms
11289 @opindex nofixprebinding
11290 @opindex nomultidefs
11292 @opindex noseglinkedit
11293 @opindex pagezero_size
11295 @opindex prebind_all_twolevel_modules
11296 @opindex private_bundle
11297 @opindex read_only_relocs
11299 @opindex sectobjectsymbols
11302 @opindex sectcreate
11303 @opindex sectobjectsymbols
11306 @opindex segs_read_only_addr
11307 @opindex segs_read_write_addr
11308 @opindex seg_addr_table
11309 @opindex seg_addr_table_filename
11310 @opindex seglinkedit
11312 @opindex segs_read_only_addr
11313 @opindex segs_read_write_addr
11314 @opindex single_module
11316 @opindex sub_library
11317 @opindex sub_umbrella
11318 @opindex twolevel_namespace
11321 @opindex unexported_symbols_list
11322 @opindex weak_reference_mismatches
11323 @opindex whatsloaded
11324 These options are passed to the Darwin linker. The Darwin linker man page
11325 describes them in detail.
11328 @node DEC Alpha Options
11329 @subsection DEC Alpha Options
11331 These @samp{-m} options are defined for the DEC Alpha implementations:
11334 @item -mno-soft-float
11335 @itemx -msoft-float
11336 @opindex mno-soft-float
11337 @opindex msoft-float
11338 Use (do not use) the hardware floating-point instructions for
11339 floating-point operations. When @option{-msoft-float} is specified,
11340 functions in @file{libgcc.a} will be used to perform floating-point
11341 operations. Unless they are replaced by routines that emulate the
11342 floating-point operations, or compiled in such a way as to call such
11343 emulations routines, these routines will issue floating-point
11344 operations. If you are compiling for an Alpha without floating-point
11345 operations, you must ensure that the library is built so as not to call
11348 Note that Alpha implementations without floating-point operations are
11349 required to have floating-point registers.
11352 @itemx -mno-fp-regs
11354 @opindex mno-fp-regs
11355 Generate code that uses (does not use) the floating-point register set.
11356 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11357 register set is not used, floating point operands are passed in integer
11358 registers as if they were integers and floating-point results are passed
11359 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11360 so any function with a floating-point argument or return value called by code
11361 compiled with @option{-mno-fp-regs} must also be compiled with that
11364 A typical use of this option is building a kernel that does not use,
11365 and hence need not save and restore, any floating-point registers.
11369 The Alpha architecture implements floating-point hardware optimized for
11370 maximum performance. It is mostly compliant with the IEEE floating
11371 point standard. However, for full compliance, software assistance is
11372 required. This option generates code fully IEEE compliant code
11373 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11374 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11375 defined during compilation. The resulting code is less efficient but is
11376 able to correctly support denormalized numbers and exceptional IEEE
11377 values such as not-a-number and plus/minus infinity. Other Alpha
11378 compilers call this option @option{-ieee_with_no_inexact}.
11380 @item -mieee-with-inexact
11381 @opindex mieee-with-inexact
11382 This is like @option{-mieee} except the generated code also maintains
11383 the IEEE @var{inexact-flag}. Turning on this option causes the
11384 generated code to implement fully-compliant IEEE math. In addition to
11385 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11386 macro. On some Alpha implementations the resulting code may execute
11387 significantly slower than the code generated by default. Since there is
11388 very little code that depends on the @var{inexact-flag}, you should
11389 normally not specify this option. Other Alpha compilers call this
11390 option @option{-ieee_with_inexact}.
11392 @item -mfp-trap-mode=@var{trap-mode}
11393 @opindex mfp-trap-mode
11394 This option controls what floating-point related traps are enabled.
11395 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11396 The trap mode can be set to one of four values:
11400 This is the default (normal) setting. The only traps that are enabled
11401 are the ones that cannot be disabled in software (e.g., division by zero
11405 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11409 Like @samp{u}, but the instructions are marked to be safe for software
11410 completion (see Alpha architecture manual for details).
11413 Like @samp{su}, but inexact traps are enabled as well.
11416 @item -mfp-rounding-mode=@var{rounding-mode}
11417 @opindex mfp-rounding-mode
11418 Selects the IEEE rounding mode. Other Alpha compilers call this option
11419 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11424 Normal IEEE rounding mode. Floating point numbers are rounded towards
11425 the nearest machine number or towards the even machine number in case
11429 Round towards minus infinity.
11432 Chopped rounding mode. Floating point numbers are rounded towards zero.
11435 Dynamic rounding mode. A field in the floating point control register
11436 (@var{fpcr}, see Alpha architecture reference manual) controls the
11437 rounding mode in effect. The C library initializes this register for
11438 rounding towards plus infinity. Thus, unless your program modifies the
11439 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11442 @item -mtrap-precision=@var{trap-precision}
11443 @opindex mtrap-precision
11444 In the Alpha architecture, floating point traps are imprecise. This
11445 means without software assistance it is impossible to recover from a
11446 floating trap and program execution normally needs to be terminated.
11447 GCC can generate code that can assist operating system trap handlers
11448 in determining the exact location that caused a floating point trap.
11449 Depending on the requirements of an application, different levels of
11450 precisions can be selected:
11454 Program precision. This option is the default and means a trap handler
11455 can only identify which program caused a floating point exception.
11458 Function precision. The trap handler can determine the function that
11459 caused a floating point exception.
11462 Instruction precision. The trap handler can determine the exact
11463 instruction that caused a floating point exception.
11466 Other Alpha compilers provide the equivalent options called
11467 @option{-scope_safe} and @option{-resumption_safe}.
11469 @item -mieee-conformant
11470 @opindex mieee-conformant
11471 This option marks the generated code as IEEE conformant. You must not
11472 use this option unless you also specify @option{-mtrap-precision=i} and either
11473 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11474 is to emit the line @samp{.eflag 48} in the function prologue of the
11475 generated assembly file. Under DEC Unix, this has the effect that
11476 IEEE-conformant math library routines will be linked in.
11478 @item -mbuild-constants
11479 @opindex mbuild-constants
11480 Normally GCC examines a 32- or 64-bit integer constant to
11481 see if it can construct it from smaller constants in two or three
11482 instructions. If it cannot, it will output the constant as a literal and
11483 generate code to load it from the data segment at runtime.
11485 Use this option to require GCC to construct @emph{all} integer constants
11486 using code, even if it takes more instructions (the maximum is six).
11488 You would typically use this option to build a shared library dynamic
11489 loader. Itself a shared library, it must relocate itself in memory
11490 before it can find the variables and constants in its own data segment.
11496 Select whether to generate code to be assembled by the vendor-supplied
11497 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11515 Indicate whether GCC should generate code to use the optional BWX,
11516 CIX, FIX and MAX instruction sets. The default is to use the instruction
11517 sets supported by the CPU type specified via @option{-mcpu=} option or that
11518 of the CPU on which GCC was built if none was specified.
11521 @itemx -mfloat-ieee
11522 @opindex mfloat-vax
11523 @opindex mfloat-ieee
11524 Generate code that uses (does not use) VAX F and G floating point
11525 arithmetic instead of IEEE single and double precision.
11527 @item -mexplicit-relocs
11528 @itemx -mno-explicit-relocs
11529 @opindex mexplicit-relocs
11530 @opindex mno-explicit-relocs
11531 Older Alpha assemblers provided no way to generate symbol relocations
11532 except via assembler macros. Use of these macros does not allow
11533 optimal instruction scheduling. GNU binutils as of version 2.12
11534 supports a new syntax that allows the compiler to explicitly mark
11535 which relocations should apply to which instructions. This option
11536 is mostly useful for debugging, as GCC detects the capabilities of
11537 the assembler when it is built and sets the default accordingly.
11540 @itemx -mlarge-data
11541 @opindex msmall-data
11542 @opindex mlarge-data
11543 When @option{-mexplicit-relocs} is in effect, static data is
11544 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11545 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11546 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11547 16-bit relocations off of the @code{$gp} register. This limits the
11548 size of the small data area to 64KB, but allows the variables to be
11549 directly accessed via a single instruction.
11551 The default is @option{-mlarge-data}. With this option the data area
11552 is limited to just below 2GB@. Programs that require more than 2GB of
11553 data must use @code{malloc} or @code{mmap} to allocate the data in the
11554 heap instead of in the program's data segment.
11556 When generating code for shared libraries, @option{-fpic} implies
11557 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11560 @itemx -mlarge-text
11561 @opindex msmall-text
11562 @opindex mlarge-text
11563 When @option{-msmall-text} is used, the compiler assumes that the
11564 code of the entire program (or shared library) fits in 4MB, and is
11565 thus reachable with a branch instruction. When @option{-msmall-data}
11566 is used, the compiler can assume that all local symbols share the
11567 same @code{$gp} value, and thus reduce the number of instructions
11568 required for a function call from 4 to 1.
11570 The default is @option{-mlarge-text}.
11572 @item -mcpu=@var{cpu_type}
11574 Set the instruction set and instruction scheduling parameters for
11575 machine type @var{cpu_type}. You can specify either the @samp{EV}
11576 style name or the corresponding chip number. GCC supports scheduling
11577 parameters for the EV4, EV5 and EV6 family of processors and will
11578 choose the default values for the instruction set from the processor
11579 you specify. If you do not specify a processor type, GCC will default
11580 to the processor on which the compiler was built.
11582 Supported values for @var{cpu_type} are
11588 Schedules as an EV4 and has no instruction set extensions.
11592 Schedules as an EV5 and has no instruction set extensions.
11596 Schedules as an EV5 and supports the BWX extension.
11601 Schedules as an EV5 and supports the BWX and MAX extensions.
11605 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11609 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11612 Native toolchains also support the value @samp{native},
11613 which selects the best architecture option for the host processor.
11614 @option{-mcpu=native} has no effect if GCC does not recognize
11617 @item -mtune=@var{cpu_type}
11619 Set only the instruction scheduling parameters for machine type
11620 @var{cpu_type}. The instruction set is not changed.
11622 Native toolchains also support the value @samp{native},
11623 which selects the best architecture option for the host processor.
11624 @option{-mtune=native} has no effect if GCC does not recognize
11627 @item -mmemory-latency=@var{time}
11628 @opindex mmemory-latency
11629 Sets the latency the scheduler should assume for typical memory
11630 references as seen by the application. This number is highly
11631 dependent on the memory access patterns used by the application
11632 and the size of the external cache on the machine.
11634 Valid options for @var{time} are
11638 A decimal number representing clock cycles.
11644 The compiler contains estimates of the number of clock cycles for
11645 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11646 (also called Dcache, Scache, and Bcache), as well as to main memory.
11647 Note that L3 is only valid for EV5.
11652 @node DEC Alpha/VMS Options
11653 @subsection DEC Alpha/VMS Options
11655 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11658 @item -mvms-return-codes
11659 @opindex mvms-return-codes
11660 Return VMS condition codes from main. The default is to return POSIX
11661 style condition (e.g.@: error) codes.
11663 @item -mdebug-main=@var{prefix}
11664 @opindex mdebug-main=@var{prefix}
11665 Flag the first routine whose name starts with @var{prefix} as the main
11666 routine for the debugger.
11670 Default to 64bit memory allocation routines.
11674 @subsection FR30 Options
11675 @cindex FR30 Options
11677 These options are defined specifically for the FR30 port.
11681 @item -msmall-model
11682 @opindex msmall-model
11683 Use the small address space model. This can produce smaller code, but
11684 it does assume that all symbolic values and addresses will fit into a
11689 Assume that run-time support has been provided and so there is no need
11690 to include the simulator library (@file{libsim.a}) on the linker
11696 @subsection FRV Options
11697 @cindex FRV Options
11703 Only use the first 32 general purpose registers.
11708 Use all 64 general purpose registers.
11713 Use only the first 32 floating point registers.
11718 Use all 64 floating point registers
11721 @opindex mhard-float
11723 Use hardware instructions for floating point operations.
11726 @opindex msoft-float
11728 Use library routines for floating point operations.
11733 Dynamically allocate condition code registers.
11738 Do not try to dynamically allocate condition code registers, only
11739 use @code{icc0} and @code{fcc0}.
11744 Change ABI to use double word insns.
11749 Do not use double word instructions.
11754 Use floating point double instructions.
11757 @opindex mno-double
11759 Do not use floating point double instructions.
11764 Use media instructions.
11769 Do not use media instructions.
11774 Use multiply and add/subtract instructions.
11777 @opindex mno-muladd
11779 Do not use multiply and add/subtract instructions.
11784 Select the FDPIC ABI, that uses function descriptors to represent
11785 pointers to functions. Without any PIC/PIE-related options, it
11786 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11787 assumes GOT entries and small data are within a 12-bit range from the
11788 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11789 are computed with 32 bits.
11790 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11793 @opindex minline-plt
11795 Enable inlining of PLT entries in function calls to functions that are
11796 not known to bind locally. It has no effect without @option{-mfdpic}.
11797 It's enabled by default if optimizing for speed and compiling for
11798 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11799 optimization option such as @option{-O3} or above is present in the
11805 Assume a large TLS segment when generating thread-local code.
11810 Do not assume a large TLS segment when generating thread-local code.
11815 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11816 that is known to be in read-only sections. It's enabled by default,
11817 except for @option{-fpic} or @option{-fpie}: even though it may help
11818 make the global offset table smaller, it trades 1 instruction for 4.
11819 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11820 one of which may be shared by multiple symbols, and it avoids the need
11821 for a GOT entry for the referenced symbol, so it's more likely to be a
11822 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11824 @item -multilib-library-pic
11825 @opindex multilib-library-pic
11827 Link with the (library, not FD) pic libraries. It's implied by
11828 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11829 @option{-fpic} without @option{-mfdpic}. You should never have to use
11833 @opindex mlinked-fp
11835 Follow the EABI requirement of always creating a frame pointer whenever
11836 a stack frame is allocated. This option is enabled by default and can
11837 be disabled with @option{-mno-linked-fp}.
11840 @opindex mlong-calls
11842 Use indirect addressing to call functions outside the current
11843 compilation unit. This allows the functions to be placed anywhere
11844 within the 32-bit address space.
11846 @item -malign-labels
11847 @opindex malign-labels
11849 Try to align labels to an 8-byte boundary by inserting nops into the
11850 previous packet. This option only has an effect when VLIW packing
11851 is enabled. It doesn't create new packets; it merely adds nops to
11854 @item -mlibrary-pic
11855 @opindex mlibrary-pic
11857 Generate position-independent EABI code.
11862 Use only the first four media accumulator registers.
11867 Use all eight media accumulator registers.
11872 Pack VLIW instructions.
11877 Do not pack VLIW instructions.
11880 @opindex mno-eflags
11882 Do not mark ABI switches in e_flags.
11885 @opindex mcond-move
11887 Enable the use of conditional-move instructions (default).
11889 This switch is mainly for debugging the compiler and will likely be removed
11890 in a future version.
11892 @item -mno-cond-move
11893 @opindex mno-cond-move
11895 Disable the use of conditional-move instructions.
11897 This switch is mainly for debugging the compiler and will likely be removed
11898 in a future version.
11903 Enable the use of conditional set instructions (default).
11905 This switch is mainly for debugging the compiler and will likely be removed
11906 in a future version.
11911 Disable the use of conditional set instructions.
11913 This switch is mainly for debugging the compiler and will likely be removed
11914 in a future version.
11917 @opindex mcond-exec
11919 Enable the use of conditional execution (default).
11921 This switch is mainly for debugging the compiler and will likely be removed
11922 in a future version.
11924 @item -mno-cond-exec
11925 @opindex mno-cond-exec
11927 Disable the use of conditional execution.
11929 This switch is mainly for debugging the compiler and will likely be removed
11930 in a future version.
11932 @item -mvliw-branch
11933 @opindex mvliw-branch
11935 Run a pass to pack branches into VLIW instructions (default).
11937 This switch is mainly for debugging the compiler and will likely be removed
11938 in a future version.
11940 @item -mno-vliw-branch
11941 @opindex mno-vliw-branch
11943 Do not run a pass to pack branches into VLIW instructions.
11945 This switch is mainly for debugging the compiler and will likely be removed
11946 in a future version.
11948 @item -mmulti-cond-exec
11949 @opindex mmulti-cond-exec
11951 Enable optimization of @code{&&} and @code{||} in conditional execution
11954 This switch is mainly for debugging the compiler and will likely be removed
11955 in a future version.
11957 @item -mno-multi-cond-exec
11958 @opindex mno-multi-cond-exec
11960 Disable optimization of @code{&&} and @code{||} in conditional execution.
11962 This switch is mainly for debugging the compiler and will likely be removed
11963 in a future version.
11965 @item -mnested-cond-exec
11966 @opindex mnested-cond-exec
11968 Enable nested conditional execution optimizations (default).
11970 This switch is mainly for debugging the compiler and will likely be removed
11971 in a future version.
11973 @item -mno-nested-cond-exec
11974 @opindex mno-nested-cond-exec
11976 Disable nested conditional execution optimizations.
11978 This switch is mainly for debugging the compiler and will likely be removed
11979 in a future version.
11981 @item -moptimize-membar
11982 @opindex moptimize-membar
11984 This switch removes redundant @code{membar} instructions from the
11985 compiler generated code. It is enabled by default.
11987 @item -mno-optimize-membar
11988 @opindex mno-optimize-membar
11990 This switch disables the automatic removal of redundant @code{membar}
11991 instructions from the generated code.
11993 @item -mtomcat-stats
11994 @opindex mtomcat-stats
11996 Cause gas to print out tomcat statistics.
11998 @item -mcpu=@var{cpu}
12001 Select the processor type for which to generate code. Possible values are
12002 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
12003 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
12007 @node GNU/Linux Options
12008 @subsection GNU/Linux Options
12010 These @samp{-m} options are defined for GNU/Linux targets:
12015 Use the GNU C library. This is the default except
12016 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
12020 Use uClibc C library. This is the default on
12021 @samp{*-*-linux-*uclibc*} targets.
12025 Use Bionic C library. This is the default on
12026 @samp{*-*-linux-*android*} targets.
12030 Compile code compatible with Android platform. This is the default on
12031 @samp{*-*-linux-*android*} targets.
12033 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
12034 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
12035 this option makes the GCC driver pass Android-specific options to the linker.
12036 Finally, this option causes the preprocessor macro @code{__ANDROID__}
12039 @item -tno-android-cc
12040 @opindex tno-android-cc
12041 Disable compilation effects of @option{-mandroid}, i.e., do not enable
12042 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
12043 @option{-fno-rtti} by default.
12045 @item -tno-android-ld
12046 @opindex tno-android-ld
12047 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
12048 linking options to the linker.
12052 @node H8/300 Options
12053 @subsection H8/300 Options
12055 These @samp{-m} options are defined for the H8/300 implementations:
12060 Shorten some address references at link time, when possible; uses the
12061 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
12062 ld, Using ld}, for a fuller description.
12066 Generate code for the H8/300H@.
12070 Generate code for the H8S@.
12074 Generate code for the H8S and H8/300H in the normal mode. This switch
12075 must be used either with @option{-mh} or @option{-ms}.
12079 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12083 Make @code{int} data 32 bits by default.
12086 @opindex malign-300
12087 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12088 The default for the H8/300H and H8S is to align longs and floats on 4
12090 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12091 This option has no effect on the H8/300.
12095 @subsection HPPA Options
12096 @cindex HPPA Options
12098 These @samp{-m} options are defined for the HPPA family of computers:
12101 @item -march=@var{architecture-type}
12103 Generate code for the specified architecture. The choices for
12104 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12105 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12106 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12107 architecture option for your machine. Code compiled for lower numbered
12108 architectures will run on higher numbered architectures, but not the
12111 @item -mpa-risc-1-0
12112 @itemx -mpa-risc-1-1
12113 @itemx -mpa-risc-2-0
12114 @opindex mpa-risc-1-0
12115 @opindex mpa-risc-1-1
12116 @opindex mpa-risc-2-0
12117 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12120 @opindex mbig-switch
12121 Generate code suitable for big switch tables. Use this option only if
12122 the assembler/linker complain about out of range branches within a switch
12125 @item -mjump-in-delay
12126 @opindex mjump-in-delay
12127 Fill delay slots of function calls with unconditional jump instructions
12128 by modifying the return pointer for the function call to be the target
12129 of the conditional jump.
12131 @item -mdisable-fpregs
12132 @opindex mdisable-fpregs
12133 Prevent floating point registers from being used in any manner. This is
12134 necessary for compiling kernels which perform lazy context switching of
12135 floating point registers. If you use this option and attempt to perform
12136 floating point operations, the compiler will abort.
12138 @item -mdisable-indexing
12139 @opindex mdisable-indexing
12140 Prevent the compiler from using indexing address modes. This avoids some
12141 rather obscure problems when compiling MIG generated code under MACH@.
12143 @item -mno-space-regs
12144 @opindex mno-space-regs
12145 Generate code that assumes the target has no space registers. This allows
12146 GCC to generate faster indirect calls and use unscaled index address modes.
12148 Such code is suitable for level 0 PA systems and kernels.
12150 @item -mfast-indirect-calls
12151 @opindex mfast-indirect-calls
12152 Generate code that assumes calls never cross space boundaries. This
12153 allows GCC to emit code which performs faster indirect calls.
12155 This option will not work in the presence of shared libraries or nested
12158 @item -mfixed-range=@var{register-range}
12159 @opindex mfixed-range
12160 Generate code treating the given register range as fixed registers.
12161 A fixed register is one that the register allocator can not use. This is
12162 useful when compiling kernel code. A register range is specified as
12163 two registers separated by a dash. Multiple register ranges can be
12164 specified separated by a comma.
12166 @item -mlong-load-store
12167 @opindex mlong-load-store
12168 Generate 3-instruction load and store sequences as sometimes required by
12169 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12172 @item -mportable-runtime
12173 @opindex mportable-runtime
12174 Use the portable calling conventions proposed by HP for ELF systems.
12178 Enable the use of assembler directives only GAS understands.
12180 @item -mschedule=@var{cpu-type}
12182 Schedule code according to the constraints for the machine type
12183 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12184 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12185 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12186 proper scheduling option for your machine. The default scheduling is
12190 @opindex mlinker-opt
12191 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12192 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12193 linkers in which they give bogus error messages when linking some programs.
12196 @opindex msoft-float
12197 Generate output containing library calls for floating point.
12198 @strong{Warning:} the requisite libraries are not available for all HPPA
12199 targets. Normally the facilities of the machine's usual C compiler are
12200 used, but this cannot be done directly in cross-compilation. You must make
12201 your own arrangements to provide suitable library functions for
12204 @option{-msoft-float} changes the calling convention in the output file;
12205 therefore, it is only useful if you compile @emph{all} of a program with
12206 this option. In particular, you need to compile @file{libgcc.a}, the
12207 library that comes with GCC, with @option{-msoft-float} in order for
12212 Generate the predefine, @code{_SIO}, for server IO@. The default is
12213 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12214 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12215 options are available under HP-UX and HI-UX@.
12219 Use GNU ld specific options. This passes @option{-shared} to ld when
12220 building a shared library. It is the default when GCC is configured,
12221 explicitly or implicitly, with the GNU linker. This option does not
12222 have any affect on which ld is called, it only changes what parameters
12223 are passed to that ld. The ld that is called is determined by the
12224 @option{--with-ld} configure option, GCC's program search path, and
12225 finally by the user's @env{PATH}. The linker used by GCC can be printed
12226 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12227 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12231 Use HP ld specific options. This passes @option{-b} to ld when building
12232 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12233 links. It is the default when GCC is configured, explicitly or
12234 implicitly, with the HP linker. This option does not have any affect on
12235 which ld is called, it only changes what parameters are passed to that
12236 ld. The ld that is called is determined by the @option{--with-ld}
12237 configure option, GCC's program search path, and finally by the user's
12238 @env{PATH}. The linker used by GCC can be printed using @samp{which
12239 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12240 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12243 @opindex mno-long-calls
12244 Generate code that uses long call sequences. This ensures that a call
12245 is always able to reach linker generated stubs. The default is to generate
12246 long calls only when the distance from the call site to the beginning
12247 of the function or translation unit, as the case may be, exceeds a
12248 predefined limit set by the branch type being used. The limits for
12249 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12250 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12253 Distances are measured from the beginning of functions when using the
12254 @option{-ffunction-sections} option, or when using the @option{-mgas}
12255 and @option{-mno-portable-runtime} options together under HP-UX with
12258 It is normally not desirable to use this option as it will degrade
12259 performance. However, it may be useful in large applications,
12260 particularly when partial linking is used to build the application.
12262 The types of long calls used depends on the capabilities of the
12263 assembler and linker, and the type of code being generated. The
12264 impact on systems that support long absolute calls, and long pic
12265 symbol-difference or pc-relative calls should be relatively small.
12266 However, an indirect call is used on 32-bit ELF systems in pic code
12267 and it is quite long.
12269 @item -munix=@var{unix-std}
12271 Generate compiler predefines and select a startfile for the specified
12272 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12273 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12274 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12275 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12276 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12279 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12280 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12281 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12282 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12283 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12284 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12286 It is @emph{important} to note that this option changes the interfaces
12287 for various library routines. It also affects the operational behavior
12288 of the C library. Thus, @emph{extreme} care is needed in using this
12291 Library code that is intended to operate with more than one UNIX
12292 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12293 as appropriate. Most GNU software doesn't provide this capability.
12297 Suppress the generation of link options to search libdld.sl when the
12298 @option{-static} option is specified on HP-UX 10 and later.
12302 The HP-UX implementation of setlocale in libc has a dependency on
12303 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12304 when the @option{-static} option is specified, special link options
12305 are needed to resolve this dependency.
12307 On HP-UX 10 and later, the GCC driver adds the necessary options to
12308 link with libdld.sl when the @option{-static} option is specified.
12309 This causes the resulting binary to be dynamic. On the 64-bit port,
12310 the linkers generate dynamic binaries by default in any case. The
12311 @option{-nolibdld} option can be used to prevent the GCC driver from
12312 adding these link options.
12316 Add support for multithreading with the @dfn{dce thread} library
12317 under HP-UX@. This option sets flags for both the preprocessor and
12321 @node i386 and x86-64 Options
12322 @subsection Intel 386 and AMD x86-64 Options
12323 @cindex i386 Options
12324 @cindex x86-64 Options
12325 @cindex Intel 386 Options
12326 @cindex AMD x86-64 Options
12328 These @samp{-m} options are defined for the i386 and x86-64 family of
12332 @item -mtune=@var{cpu-type}
12334 Tune to @var{cpu-type} everything applicable about the generated code, except
12335 for the ABI and the set of available instructions. The choices for
12336 @var{cpu-type} are:
12339 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12340 If you know the CPU on which your code will run, then you should use
12341 the corresponding @option{-mtune} option instead of
12342 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12343 of your application will have, then you should use this option.
12345 As new processors are deployed in the marketplace, the behavior of this
12346 option will change. Therefore, if you upgrade to a newer version of
12347 GCC, the code generated option will change to reflect the processors
12348 that were most common when that version of GCC was released.
12350 There is no @option{-march=generic} option because @option{-march}
12351 indicates the instruction set the compiler can use, and there is no
12352 generic instruction set applicable to all processors. In contrast,
12353 @option{-mtune} indicates the processor (or, in this case, collection of
12354 processors) for which the code is optimized.
12356 This selects the CPU to tune for at compilation time by determining
12357 the processor type of the compiling machine. Using @option{-mtune=native}
12358 will produce code optimized for the local machine under the constraints
12359 of the selected instruction set. Using @option{-march=native} will
12360 enable all instruction subsets supported by the local machine (hence
12361 the result might not run on different machines).
12363 Original Intel's i386 CPU@.
12365 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12366 @item i586, pentium
12367 Intel Pentium CPU with no MMX support.
12369 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12371 Intel PentiumPro CPU@.
12373 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12374 instruction set will be used, so the code will run on all i686 family chips.
12376 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12377 @item pentium3, pentium3m
12378 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12381 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12382 support. Used by Centrino notebooks.
12383 @item pentium4, pentium4m
12384 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12386 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12389 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12390 SSE2 and SSE3 instruction set support.
12392 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12393 instruction set support.
12395 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12396 and SSE4.2 instruction set support.
12398 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12399 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12401 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12402 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
12405 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12406 instruction set support.
12408 AMD K6 CPU with MMX instruction set support.
12410 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12411 @item athlon, athlon-tbird
12412 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12414 @item athlon-4, athlon-xp, athlon-mp
12415 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12416 instruction set support.
12417 @item k8, opteron, athlon64, athlon-fx
12418 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12419 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12420 @item k8-sse3, opteron-sse3, athlon64-sse3
12421 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12422 @item amdfam10, barcelona
12423 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12424 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12425 instruction set extensions.)
12427 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12430 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12431 instruction set support.
12433 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12434 implemented for this chip.)
12436 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12437 implemented for this chip.)
12439 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12442 While picking a specific @var{cpu-type} will schedule things appropriately
12443 for that particular chip, the compiler will not generate any code that
12444 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12447 @item -march=@var{cpu-type}
12449 Generate instructions for the machine type @var{cpu-type}. The choices
12450 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12451 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12453 @item -mcpu=@var{cpu-type}
12455 A deprecated synonym for @option{-mtune}.
12457 @item -mfpmath=@var{unit}
12459 Generate floating point arithmetics for selected unit @var{unit}. The choices
12460 for @var{unit} are:
12464 Use the standard 387 floating point coprocessor present majority of chips and
12465 emulated otherwise. Code compiled with this option will run almost everywhere.
12466 The temporary results are computed in 80bit precision instead of precision
12467 specified by the type resulting in slightly different results compared to most
12468 of other chips. See @option{-ffloat-store} for more detailed description.
12470 This is the default choice for i386 compiler.
12473 Use scalar floating point instructions present in the SSE instruction set.
12474 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12475 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12476 instruction set supports only single precision arithmetics, thus the double and
12477 extended precision arithmetics is still done using 387. Later version, present
12478 only in Pentium4 and the future AMD x86-64 chips supports double precision
12481 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12482 or @option{-msse2} switches to enable SSE extensions and make this option
12483 effective. For the x86-64 compiler, these extensions are enabled by default.
12485 The resulting code should be considerably faster in the majority of cases and avoid
12486 the numerical instability problems of 387 code, but may break some existing
12487 code that expects temporaries to be 80bit.
12489 This is the default choice for the x86-64 compiler.
12494 Attempt to utilize both instruction sets at once. This effectively double the
12495 amount of available registers and on chips with separate execution units for
12496 387 and SSE the execution resources too. Use this option with care, as it is
12497 still experimental, because the GCC register allocator does not model separate
12498 functional units well resulting in instable performance.
12501 @item -masm=@var{dialect}
12502 @opindex masm=@var{dialect}
12503 Output asm instructions using selected @var{dialect}. Supported
12504 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12505 not support @samp{intel}.
12508 @itemx -mno-ieee-fp
12510 @opindex mno-ieee-fp
12511 Control whether or not the compiler uses IEEE floating point
12512 comparisons. These handle correctly the case where the result of a
12513 comparison is unordered.
12516 @opindex msoft-float
12517 Generate output containing library calls for floating point.
12518 @strong{Warning:} the requisite libraries are not part of GCC@.
12519 Normally the facilities of the machine's usual C compiler are used, but
12520 this can't be done directly in cross-compilation. You must make your
12521 own arrangements to provide suitable library functions for
12524 On machines where a function returns floating point results in the 80387
12525 register stack, some floating point opcodes may be emitted even if
12526 @option{-msoft-float} is used.
12528 @item -mno-fp-ret-in-387
12529 @opindex mno-fp-ret-in-387
12530 Do not use the FPU registers for return values of functions.
12532 The usual calling convention has functions return values of types
12533 @code{float} and @code{double} in an FPU register, even if there
12534 is no FPU@. The idea is that the operating system should emulate
12537 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12538 in ordinary CPU registers instead.
12540 @item -mno-fancy-math-387
12541 @opindex mno-fancy-math-387
12542 Some 387 emulators do not support the @code{sin}, @code{cos} and
12543 @code{sqrt} instructions for the 387. Specify this option to avoid
12544 generating those instructions. This option is the default on FreeBSD,
12545 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12546 indicates that the target CPU will always have an FPU and so the
12547 instruction will not need emulation. As of revision 2.6.1, these
12548 instructions are not generated unless you also use the
12549 @option{-funsafe-math-optimizations} switch.
12551 @item -malign-double
12552 @itemx -mno-align-double
12553 @opindex malign-double
12554 @opindex mno-align-double
12555 Control whether GCC aligns @code{double}, @code{long double}, and
12556 @code{long long} variables on a two word boundary or a one word
12557 boundary. Aligning @code{double} variables on a two word boundary will
12558 produce code that runs somewhat faster on a @samp{Pentium} at the
12559 expense of more memory.
12561 On x86-64, @option{-malign-double} is enabled by default.
12563 @strong{Warning:} if you use the @option{-malign-double} switch,
12564 structures containing the above types will be aligned differently than
12565 the published application binary interface specifications for the 386
12566 and will not be binary compatible with structures in code compiled
12567 without that switch.
12569 @item -m96bit-long-double
12570 @itemx -m128bit-long-double
12571 @opindex m96bit-long-double
12572 @opindex m128bit-long-double
12573 These switches control the size of @code{long double} type. The i386
12574 application binary interface specifies the size to be 96 bits,
12575 so @option{-m96bit-long-double} is the default in 32 bit mode.
12577 Modern architectures (Pentium and newer) would prefer @code{long double}
12578 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12579 conforming to the ABI, this would not be possible. So specifying a
12580 @option{-m128bit-long-double} will align @code{long double}
12581 to a 16 byte boundary by padding the @code{long double} with an additional
12584 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12585 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12587 Notice that neither of these options enable any extra precision over the x87
12588 standard of 80 bits for a @code{long double}.
12590 @strong{Warning:} if you override the default value for your target ABI, the
12591 structures and arrays containing @code{long double} variables will change
12592 their size as well as function calling convention for function taking
12593 @code{long double} will be modified. Hence they will not be binary
12594 compatible with arrays or structures in code compiled without that switch.
12596 @item -mlarge-data-threshold=@var{number}
12597 @opindex mlarge-data-threshold=@var{number}
12598 When @option{-mcmodel=medium} is specified, the data greater than
12599 @var{threshold} are placed in large data section. This value must be the
12600 same across all object linked into the binary and defaults to 65535.
12604 Use a different function-calling convention, in which functions that
12605 take a fixed number of arguments return with the @code{ret} @var{num}
12606 instruction, which pops their arguments while returning. This saves one
12607 instruction in the caller since there is no need to pop the arguments
12610 You can specify that an individual function is called with this calling
12611 sequence with the function attribute @samp{stdcall}. You can also
12612 override the @option{-mrtd} option by using the function attribute
12613 @samp{cdecl}. @xref{Function Attributes}.
12615 @strong{Warning:} this calling convention is incompatible with the one
12616 normally used on Unix, so you cannot use it if you need to call
12617 libraries compiled with the Unix compiler.
12619 Also, you must provide function prototypes for all functions that
12620 take variable numbers of arguments (including @code{printf});
12621 otherwise incorrect code will be generated for calls to those
12624 In addition, seriously incorrect code will result if you call a
12625 function with too many arguments. (Normally, extra arguments are
12626 harmlessly ignored.)
12628 @item -mregparm=@var{num}
12630 Control how many registers are used to pass integer arguments. By
12631 default, no registers are used to pass arguments, and at most 3
12632 registers can be used. You can control this behavior for a specific
12633 function by using the function attribute @samp{regparm}.
12634 @xref{Function Attributes}.
12636 @strong{Warning:} if you use this switch, and
12637 @var{num} is nonzero, then you must build all modules with the same
12638 value, including any libraries. This includes the system libraries and
12642 @opindex msseregparm
12643 Use SSE register passing conventions for float and double arguments
12644 and return values. You can control this behavior for a specific
12645 function by using the function attribute @samp{sseregparm}.
12646 @xref{Function Attributes}.
12648 @strong{Warning:} if you use this switch then you must build all
12649 modules with the same value, including any libraries. This includes
12650 the system libraries and startup modules.
12652 @item -mvect8-ret-in-mem
12653 @opindex mvect8-ret-in-mem
12654 Return 8-byte vectors in memory instead of MMX registers. This is the
12655 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12656 Studio compilers until version 12. Later compiler versions (starting
12657 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12658 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12659 you need to remain compatible with existing code produced by those
12660 previous compiler versions or older versions of GCC.
12669 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12670 is specified, the significands of results of floating-point operations are
12671 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12672 significands of results of floating-point operations to 53 bits (double
12673 precision) and @option{-mpc80} rounds the significands of results of
12674 floating-point operations to 64 bits (extended double precision), which is
12675 the default. When this option is used, floating-point operations in higher
12676 precisions are not available to the programmer without setting the FPU
12677 control word explicitly.
12679 Setting the rounding of floating-point operations to less than the default
12680 80 bits can speed some programs by 2% or more. Note that some mathematical
12681 libraries assume that extended precision (80 bit) floating-point operations
12682 are enabled by default; routines in such libraries could suffer significant
12683 loss of accuracy, typically through so-called "catastrophic cancellation",
12684 when this option is used to set the precision to less than extended precision.
12686 @item -mstackrealign
12687 @opindex mstackrealign
12688 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12689 option will generate an alternate prologue and epilogue that realigns the
12690 runtime stack if necessary. This supports mixing legacy codes that keep
12691 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12692 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12693 applicable to individual functions.
12695 @item -mpreferred-stack-boundary=@var{num}
12696 @opindex mpreferred-stack-boundary
12697 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12698 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12699 the default is 4 (16 bytes or 128 bits).
12701 @item -mincoming-stack-boundary=@var{num}
12702 @opindex mincoming-stack-boundary
12703 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12704 boundary. If @option{-mincoming-stack-boundary} is not specified,
12705 the one specified by @option{-mpreferred-stack-boundary} will be used.
12707 On Pentium and PentiumPro, @code{double} and @code{long double} values
12708 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12709 suffer significant run time performance penalties. On Pentium III, the
12710 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12711 properly if it is not 16 byte aligned.
12713 To ensure proper alignment of this values on the stack, the stack boundary
12714 must be as aligned as that required by any value stored on the stack.
12715 Further, every function must be generated such that it keeps the stack
12716 aligned. Thus calling a function compiled with a higher preferred
12717 stack boundary from a function compiled with a lower preferred stack
12718 boundary will most likely misalign the stack. It is recommended that
12719 libraries that use callbacks always use the default setting.
12721 This extra alignment does consume extra stack space, and generally
12722 increases code size. Code that is sensitive to stack space usage, such
12723 as embedded systems and operating system kernels, may want to reduce the
12724 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12753 @itemx -mno-fsgsbase
12789 These switches enable or disable the use of instructions in the MMX, SSE,
12790 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
12791 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
12792 @: extended instruction sets.
12793 These extensions are also available as built-in functions: see
12794 @ref{X86 Built-in Functions}, for details of the functions enabled and
12795 disabled by these switches.
12797 To have SSE/SSE2 instructions generated automatically from floating-point
12798 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12800 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12801 generates new AVX instructions or AVX equivalence for all SSEx instructions
12804 These options will enable GCC to use these extended instructions in
12805 generated code, even without @option{-mfpmath=sse}. Applications which
12806 perform runtime CPU detection must compile separate files for each
12807 supported architecture, using the appropriate flags. In particular,
12808 the file containing the CPU detection code should be compiled without
12813 This option instructs GCC to emit a @code{cld} instruction in the prologue
12814 of functions that use string instructions. String instructions depend on
12815 the DF flag to select between autoincrement or autodecrement mode. While the
12816 ABI specifies the DF flag to be cleared on function entry, some operating
12817 systems violate this specification by not clearing the DF flag in their
12818 exception dispatchers. The exception handler can be invoked with the DF flag
12819 set which leads to wrong direction mode, when string instructions are used.
12820 This option can be enabled by default on 32-bit x86 targets by configuring
12821 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12822 instructions can be suppressed with the @option{-mno-cld} compiler option
12826 @opindex mvzeroupper
12827 This option instructs GCC to emit a @code{vzeroupper} instruction
12828 before a transfer of control flow out of the function to minimize
12829 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12834 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12835 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12836 data types. This is useful for high resolution counters that could be updated
12837 by multiple processors (or cores). This instruction is generated as part of
12838 atomic built-in functions: see @ref{Atomic Builtins} for details.
12842 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12843 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12844 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12845 SAHF are load and store instructions, respectively, for certain status flags.
12846 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12847 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12851 This option will enable GCC to use movbe instruction to implement
12852 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12856 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12857 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12858 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12862 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12863 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12864 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12865 variants) for single precision floating point arguments. These instructions
12866 are generated only when @option{-funsafe-math-optimizations} is enabled
12867 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12868 Note that while the throughput of the sequence is higher than the throughput
12869 of the non-reciprocal instruction, the precision of the sequence can be
12870 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12872 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12873 already with @option{-ffast-math} (or the above option combination), and
12874 doesn't need @option{-mrecip}.
12876 @item -mrecip=@var{opt}
12877 @opindex mrecip=opt
12878 This option allows to control which reciprocal estimate instructions
12879 may be used. @var{opt} is a comma separated list of options, that may
12880 be preceded by a @code{!} to invert the option:
12881 @code{all}: enable all estimate instructions,
12882 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
12883 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip},
12884 @code{div}: enable the approximation for scalar division,
12885 @code{vec-div}: enable the approximation for vectorized division,
12886 @code{sqrt}: enable the approximation for scalar square root,
12887 @code{vec-sqrt}: enable the approximation for vectorized square root.
12889 So for example, @option{-mrecip=all,!sqrt} would enable
12890 all of the reciprocal approximations, except for square root.
12892 @item -mveclibabi=@var{type}
12893 @opindex mveclibabi
12894 Specifies the ABI type to use for vectorizing intrinsics using an
12895 external library. Supported types are @code{svml} for the Intel short
12896 vector math library and @code{acml} for the AMD math core library style
12897 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12898 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12899 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12900 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12901 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12902 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12903 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12904 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12905 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12906 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12907 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12908 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12909 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12910 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12911 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12912 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12913 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12914 compatible library will have to be specified at link time.
12916 @item -mabi=@var{name}
12918 Generate code for the specified calling convention. Permissible values
12919 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12920 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12921 ABI when targeting Windows. On all other systems, the default is the
12922 SYSV ABI. You can control this behavior for a specific function by
12923 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12924 @xref{Function Attributes}.
12926 @item -mtls-dialect=@var{type}
12927 @opindex mtls-dialect
12928 Generate code to access thread-local storage using the @samp{gnu} or
12929 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
12930 @samp{gnu2} is more efficient, but it may add compile- and run-time
12931 requirements that cannot be satisfied on all systems.
12934 @itemx -mno-push-args
12935 @opindex mpush-args
12936 @opindex mno-push-args
12937 Use PUSH operations to store outgoing parameters. This method is shorter
12938 and usually equally fast as method using SUB/MOV operations and is enabled
12939 by default. In some cases disabling it may improve performance because of
12940 improved scheduling and reduced dependencies.
12942 @item -maccumulate-outgoing-args
12943 @opindex maccumulate-outgoing-args
12944 If enabled, the maximum amount of space required for outgoing arguments will be
12945 computed in the function prologue. This is faster on most modern CPUs
12946 because of reduced dependencies, improved scheduling and reduced stack usage
12947 when preferred stack boundary is not equal to 2. The drawback is a notable
12948 increase in code size. This switch implies @option{-mno-push-args}.
12952 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12953 on thread-safe exception handling must compile and link all code with the
12954 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12955 @option{-D_MT}; when linking, it links in a special thread helper library
12956 @option{-lmingwthrd} which cleans up per thread exception handling data.
12958 @item -mno-align-stringops
12959 @opindex mno-align-stringops
12960 Do not align destination of inlined string operations. This switch reduces
12961 code size and improves performance in case the destination is already aligned,
12962 but GCC doesn't know about it.
12964 @item -minline-all-stringops
12965 @opindex minline-all-stringops
12966 By default GCC inlines string operations only when destination is known to be
12967 aligned at least to 4 byte boundary. This enables more inlining, increase code
12968 size, but may improve performance of code that depends on fast memcpy, strlen
12969 and memset for short lengths.
12971 @item -minline-stringops-dynamically
12972 @opindex minline-stringops-dynamically
12973 For string operation of unknown size, inline runtime checks so for small
12974 blocks inline code is used, while for large blocks library call is used.
12976 @item -mstringop-strategy=@var{alg}
12977 @opindex mstringop-strategy=@var{alg}
12978 Overwrite internal decision heuristic about particular algorithm to inline
12979 string operation with. The allowed values are @code{rep_byte},
12980 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12981 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12982 expanding inline loop, @code{libcall} for always expanding library call.
12984 @item -momit-leaf-frame-pointer
12985 @opindex momit-leaf-frame-pointer
12986 Don't keep the frame pointer in a register for leaf functions. This
12987 avoids the instructions to save, set up and restore frame pointers and
12988 makes an extra register available in leaf functions. The option
12989 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12990 which might make debugging harder.
12992 @item -mtls-direct-seg-refs
12993 @itemx -mno-tls-direct-seg-refs
12994 @opindex mtls-direct-seg-refs
12995 Controls whether TLS variables may be accessed with offsets from the
12996 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12997 or whether the thread base pointer must be added. Whether or not this
12998 is legal depends on the operating system, and whether it maps the
12999 segment to cover the entire TLS area.
13001 For systems that use GNU libc, the default is on.
13004 @itemx -mno-sse2avx
13006 Specify that the assembler should encode SSE instructions with VEX
13007 prefix. The option @option{-mavx} turns this on by default.
13012 If profiling is active @option{-pg} put the profiling
13013 counter call before prologue.
13014 Note: On x86 architectures the attribute @code{ms_hook_prologue}
13015 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
13018 @itemx -mno-8bit-idiv
13020 On some processors, like Intel Atom, 8bit unsigned integer divide is
13021 much faster than 32bit/64bit integer divide. This option will generate a
13022 runt-time check. If both dividend and divisor are within range of 0
13023 to 255, 8bit unsigned integer divide will be used instead of
13024 32bit/64bit integer divide.
13026 @item -mavx256-split-unaligned-load
13027 @item -mavx256-split-unaligned-store
13028 @opindex avx256-split-unaligned-load
13029 @opindex avx256-split-unaligned-store
13030 Split 32-byte AVX unaligned load and store.
13034 These @samp{-m} switches are supported in addition to the above
13035 on AMD x86-64 processors in 64-bit environments.
13044 Generate code for a 32-bit or 64-bit environment.
13045 The @option{-m32} option sets int, long and pointer to 32 bits and
13046 generates code that runs on any i386 system.
13047 The @option{-m64} option sets int to 32 bits and long and pointer
13048 to 64 bits and generates code for AMD's x86-64 architecture.
13049 The @option{-mx32} option sets int, long and pointer to 32 bits and
13050 generates code for AMD's x86-64 architecture.
13051 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
13052 and @option{-mdynamic-no-pic} options.
13054 @item -mno-red-zone
13055 @opindex mno-red-zone
13056 Do not use a so called red zone for x86-64 code. The red zone is mandated
13057 by the x86-64 ABI, it is a 128-byte area beyond the location of the
13058 stack pointer that will not be modified by signal or interrupt handlers
13059 and therefore can be used for temporary data without adjusting the stack
13060 pointer. The flag @option{-mno-red-zone} disables this red zone.
13062 @item -mcmodel=small
13063 @opindex mcmodel=small
13064 Generate code for the small code model: the program and its symbols must
13065 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
13066 Programs can be statically or dynamically linked. This is the default
13069 @item -mcmodel=kernel
13070 @opindex mcmodel=kernel
13071 Generate code for the kernel code model. The kernel runs in the
13072 negative 2 GB of the address space.
13073 This model has to be used for Linux kernel code.
13075 @item -mcmodel=medium
13076 @opindex mcmodel=medium
13077 Generate code for the medium model: The program is linked in the lower 2
13078 GB of the address space. Small symbols are also placed there. Symbols
13079 with sizes larger than @option{-mlarge-data-threshold} are put into
13080 large data or bss sections and can be located above 2GB. Programs can
13081 be statically or dynamically linked.
13083 @item -mcmodel=large
13084 @opindex mcmodel=large
13085 Generate code for the large model: This model makes no assumptions
13086 about addresses and sizes of sections.
13089 @node i386 and x86-64 Windows Options
13090 @subsection i386 and x86-64 Windows Options
13091 @cindex i386 and x86-64 Windows Options
13093 These additional options are available for Windows targets:
13098 This option is available for Cygwin and MinGW targets. It
13099 specifies that a console application is to be generated, by
13100 instructing the linker to set the PE header subsystem type
13101 required for console applications.
13102 This is the default behavior for Cygwin and MinGW targets.
13106 This option is available for Cygwin and MinGW targets. It
13107 specifies that a DLL - a dynamic link library - is to be
13108 generated, enabling the selection of the required runtime
13109 startup object and entry point.
13111 @item -mnop-fun-dllimport
13112 @opindex mnop-fun-dllimport
13113 This option is available for Cygwin and MinGW targets. It
13114 specifies that the dllimport attribute should be ignored.
13118 This option is available for MinGW targets. It specifies
13119 that MinGW-specific thread support is to be used.
13123 This option is available for mingw-w64 targets. It specifies
13124 that the UNICODE macro is getting pre-defined and that the
13125 unicode capable runtime startup code is chosen.
13129 This option is available for Cygwin and MinGW targets. It
13130 specifies that the typical Windows pre-defined macros are to
13131 be set in the pre-processor, but does not influence the choice
13132 of runtime library/startup code.
13136 This option is available for Cygwin and MinGW targets. It
13137 specifies that a GUI application is to be generated by
13138 instructing the linker to set the PE header subsystem type
13141 @item -fno-set-stack-executable
13142 @opindex fno-set-stack-executable
13143 This option is available for MinGW targets. It specifies that
13144 the executable flag for stack used by nested functions isn't
13145 set. This is necessary for binaries running in kernel mode of
13146 Windows, as there the user32 API, which is used to set executable
13147 privileges, isn't available.
13149 @item -mpe-aligned-commons
13150 @opindex mpe-aligned-commons
13151 This option is available for Cygwin and MinGW targets. It
13152 specifies that the GNU extension to the PE file format that
13153 permits the correct alignment of COMMON variables should be
13154 used when generating code. It will be enabled by default if
13155 GCC detects that the target assembler found during configuration
13156 supports the feature.
13159 See also under @ref{i386 and x86-64 Options} for standard options.
13161 @node IA-64 Options
13162 @subsection IA-64 Options
13163 @cindex IA-64 Options
13165 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13169 @opindex mbig-endian
13170 Generate code for a big endian target. This is the default for HP-UX@.
13172 @item -mlittle-endian
13173 @opindex mlittle-endian
13174 Generate code for a little endian target. This is the default for AIX5
13180 @opindex mno-gnu-as
13181 Generate (or don't) code for the GNU assembler. This is the default.
13182 @c Also, this is the default if the configure option @option{--with-gnu-as}
13188 @opindex mno-gnu-ld
13189 Generate (or don't) code for the GNU linker. This is the default.
13190 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13195 Generate code that does not use a global pointer register. The result
13196 is not position independent code, and violates the IA-64 ABI@.
13198 @item -mvolatile-asm-stop
13199 @itemx -mno-volatile-asm-stop
13200 @opindex mvolatile-asm-stop
13201 @opindex mno-volatile-asm-stop
13202 Generate (or don't) a stop bit immediately before and after volatile asm
13205 @item -mregister-names
13206 @itemx -mno-register-names
13207 @opindex mregister-names
13208 @opindex mno-register-names
13209 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13210 the stacked registers. This may make assembler output more readable.
13216 Disable (or enable) optimizations that use the small data section. This may
13217 be useful for working around optimizer bugs.
13219 @item -mconstant-gp
13220 @opindex mconstant-gp
13221 Generate code that uses a single constant global pointer value. This is
13222 useful when compiling kernel code.
13226 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13227 This is useful when compiling firmware code.
13229 @item -minline-float-divide-min-latency
13230 @opindex minline-float-divide-min-latency
13231 Generate code for inline divides of floating point values
13232 using the minimum latency algorithm.
13234 @item -minline-float-divide-max-throughput
13235 @opindex minline-float-divide-max-throughput
13236 Generate code for inline divides of floating point values
13237 using the maximum throughput algorithm.
13239 @item -mno-inline-float-divide
13240 @opindex mno-inline-float-divide
13241 Do not generate inline code for divides of floating point values.
13243 @item -minline-int-divide-min-latency
13244 @opindex minline-int-divide-min-latency
13245 Generate code for inline divides of integer values
13246 using the minimum latency algorithm.
13248 @item -minline-int-divide-max-throughput
13249 @opindex minline-int-divide-max-throughput
13250 Generate code for inline divides of integer values
13251 using the maximum throughput algorithm.
13253 @item -mno-inline-int-divide
13254 @opindex mno-inline-int-divide
13255 Do not generate inline code for divides of integer values.
13257 @item -minline-sqrt-min-latency
13258 @opindex minline-sqrt-min-latency
13259 Generate code for inline square roots
13260 using the minimum latency algorithm.
13262 @item -minline-sqrt-max-throughput
13263 @opindex minline-sqrt-max-throughput
13264 Generate code for inline square roots
13265 using the maximum throughput algorithm.
13267 @item -mno-inline-sqrt
13268 @opindex mno-inline-sqrt
13269 Do not generate inline code for sqrt.
13272 @itemx -mno-fused-madd
13273 @opindex mfused-madd
13274 @opindex mno-fused-madd
13275 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13276 instructions. The default is to use these instructions.
13278 @item -mno-dwarf2-asm
13279 @itemx -mdwarf2-asm
13280 @opindex mno-dwarf2-asm
13281 @opindex mdwarf2-asm
13282 Don't (or do) generate assembler code for the DWARF2 line number debugging
13283 info. This may be useful when not using the GNU assembler.
13285 @item -mearly-stop-bits
13286 @itemx -mno-early-stop-bits
13287 @opindex mearly-stop-bits
13288 @opindex mno-early-stop-bits
13289 Allow stop bits to be placed earlier than immediately preceding the
13290 instruction that triggered the stop bit. This can improve instruction
13291 scheduling, but does not always do so.
13293 @item -mfixed-range=@var{register-range}
13294 @opindex mfixed-range
13295 Generate code treating the given register range as fixed registers.
13296 A fixed register is one that the register allocator can not use. This is
13297 useful when compiling kernel code. A register range is specified as
13298 two registers separated by a dash. Multiple register ranges can be
13299 specified separated by a comma.
13301 @item -mtls-size=@var{tls-size}
13303 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13306 @item -mtune=@var{cpu-type}
13308 Tune the instruction scheduling for a particular CPU, Valid values are
13309 itanium, itanium1, merced, itanium2, and mckinley.
13315 Generate code for a 32-bit or 64-bit environment.
13316 The 32-bit environment sets int, long and pointer to 32 bits.
13317 The 64-bit environment sets int to 32 bits and long and pointer
13318 to 64 bits. These are HP-UX specific flags.
13320 @item -mno-sched-br-data-spec
13321 @itemx -msched-br-data-spec
13322 @opindex mno-sched-br-data-spec
13323 @opindex msched-br-data-spec
13324 (Dis/En)able data speculative scheduling before reload.
13325 This will result in generation of the ld.a instructions and
13326 the corresponding check instructions (ld.c / chk.a).
13327 The default is 'disable'.
13329 @item -msched-ar-data-spec
13330 @itemx -mno-sched-ar-data-spec
13331 @opindex msched-ar-data-spec
13332 @opindex mno-sched-ar-data-spec
13333 (En/Dis)able data speculative scheduling after reload.
13334 This will result in generation of the ld.a instructions and
13335 the corresponding check instructions (ld.c / chk.a).
13336 The default is 'enable'.
13338 @item -mno-sched-control-spec
13339 @itemx -msched-control-spec
13340 @opindex mno-sched-control-spec
13341 @opindex msched-control-spec
13342 (Dis/En)able control speculative scheduling. This feature is
13343 available only during region scheduling (i.e.@: before reload).
13344 This will result in generation of the ld.s instructions and
13345 the corresponding check instructions chk.s .
13346 The default is 'disable'.
13348 @item -msched-br-in-data-spec
13349 @itemx -mno-sched-br-in-data-spec
13350 @opindex msched-br-in-data-spec
13351 @opindex mno-sched-br-in-data-spec
13352 (En/Dis)able speculative scheduling of the instructions that
13353 are dependent on the data speculative loads before reload.
13354 This is effective only with @option{-msched-br-data-spec} enabled.
13355 The default is 'enable'.
13357 @item -msched-ar-in-data-spec
13358 @itemx -mno-sched-ar-in-data-spec
13359 @opindex msched-ar-in-data-spec
13360 @opindex mno-sched-ar-in-data-spec
13361 (En/Dis)able speculative scheduling of the instructions that
13362 are dependent on the data speculative loads after reload.
13363 This is effective only with @option{-msched-ar-data-spec} enabled.
13364 The default is 'enable'.
13366 @item -msched-in-control-spec
13367 @itemx -mno-sched-in-control-spec
13368 @opindex msched-in-control-spec
13369 @opindex mno-sched-in-control-spec
13370 (En/Dis)able speculative scheduling of the instructions that
13371 are dependent on the control speculative loads.
13372 This is effective only with @option{-msched-control-spec} enabled.
13373 The default is 'enable'.
13375 @item -mno-sched-prefer-non-data-spec-insns
13376 @itemx -msched-prefer-non-data-spec-insns
13377 @opindex mno-sched-prefer-non-data-spec-insns
13378 @opindex msched-prefer-non-data-spec-insns
13379 If enabled, data speculative instructions will be chosen for schedule
13380 only if there are no other choices at the moment. This will make
13381 the use of the data speculation much more conservative.
13382 The default is 'disable'.
13384 @item -mno-sched-prefer-non-control-spec-insns
13385 @itemx -msched-prefer-non-control-spec-insns
13386 @opindex mno-sched-prefer-non-control-spec-insns
13387 @opindex msched-prefer-non-control-spec-insns
13388 If enabled, control speculative instructions will be chosen for schedule
13389 only if there are no other choices at the moment. This will make
13390 the use of the control speculation much more conservative.
13391 The default is 'disable'.
13393 @item -mno-sched-count-spec-in-critical-path
13394 @itemx -msched-count-spec-in-critical-path
13395 @opindex mno-sched-count-spec-in-critical-path
13396 @opindex msched-count-spec-in-critical-path
13397 If enabled, speculative dependencies will be considered during
13398 computation of the instructions priorities. This will make the use of the
13399 speculation a bit more conservative.
13400 The default is 'disable'.
13402 @item -msched-spec-ldc
13403 @opindex msched-spec-ldc
13404 Use a simple data speculation check. This option is on by default.
13406 @item -msched-control-spec-ldc
13407 @opindex msched-spec-ldc
13408 Use a simple check for control speculation. This option is on by default.
13410 @item -msched-stop-bits-after-every-cycle
13411 @opindex msched-stop-bits-after-every-cycle
13412 Place a stop bit after every cycle when scheduling. This option is on
13415 @item -msched-fp-mem-deps-zero-cost
13416 @opindex msched-fp-mem-deps-zero-cost
13417 Assume that floating-point stores and loads are not likely to cause a conflict
13418 when placed into the same instruction group. This option is disabled by
13421 @item -msel-sched-dont-check-control-spec
13422 @opindex msel-sched-dont-check-control-spec
13423 Generate checks for control speculation in selective scheduling.
13424 This flag is disabled by default.
13426 @item -msched-max-memory-insns=@var{max-insns}
13427 @opindex msched-max-memory-insns
13428 Limit on the number of memory insns per instruction group, giving lower
13429 priority to subsequent memory insns attempting to schedule in the same
13430 instruction group. Frequently useful to prevent cache bank conflicts.
13431 The default value is 1.
13433 @item -msched-max-memory-insns-hard-limit
13434 @opindex msched-max-memory-insns-hard-limit
13435 Disallow more than `msched-max-memory-insns' in instruction group.
13436 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13437 when limit is reached but may still schedule memory operations.
13441 @node IA-64/VMS Options
13442 @subsection IA-64/VMS Options
13444 These @samp{-m} options are defined for the IA-64/VMS implementations:
13447 @item -mvms-return-codes
13448 @opindex mvms-return-codes
13449 Return VMS condition codes from main. The default is to return POSIX
13450 style condition (e.g.@ error) codes.
13452 @item -mdebug-main=@var{prefix}
13453 @opindex mdebug-main=@var{prefix}
13454 Flag the first routine whose name starts with @var{prefix} as the main
13455 routine for the debugger.
13459 Default to 64bit memory allocation routines.
13463 @subsection LM32 Options
13464 @cindex LM32 options
13466 These @option{-m} options are defined for the Lattice Mico32 architecture:
13469 @item -mbarrel-shift-enabled
13470 @opindex mbarrel-shift-enabled
13471 Enable barrel-shift instructions.
13473 @item -mdivide-enabled
13474 @opindex mdivide-enabled
13475 Enable divide and modulus instructions.
13477 @item -mmultiply-enabled
13478 @opindex multiply-enabled
13479 Enable multiply instructions.
13481 @item -msign-extend-enabled
13482 @opindex msign-extend-enabled
13483 Enable sign extend instructions.
13485 @item -muser-enabled
13486 @opindex muser-enabled
13487 Enable user-defined instructions.
13492 @subsection M32C Options
13493 @cindex M32C options
13496 @item -mcpu=@var{name}
13498 Select the CPU for which code is generated. @var{name} may be one of
13499 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13500 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13501 the M32C/80 series.
13505 Specifies that the program will be run on the simulator. This causes
13506 an alternate runtime library to be linked in which supports, for
13507 example, file I/O@. You must not use this option when generating
13508 programs that will run on real hardware; you must provide your own
13509 runtime library for whatever I/O functions are needed.
13511 @item -memregs=@var{number}
13513 Specifies the number of memory-based pseudo-registers GCC will use
13514 during code generation. These pseudo-registers will be used like real
13515 registers, so there is a tradeoff between GCC's ability to fit the
13516 code into available registers, and the performance penalty of using
13517 memory instead of registers. Note that all modules in a program must
13518 be compiled with the same value for this option. Because of that, you
13519 must not use this option with the default runtime libraries gcc
13524 @node M32R/D Options
13525 @subsection M32R/D Options
13526 @cindex M32R/D options
13528 These @option{-m} options are defined for Renesas M32R/D architectures:
13533 Generate code for the M32R/2@.
13537 Generate code for the M32R/X@.
13541 Generate code for the M32R@. This is the default.
13543 @item -mmodel=small
13544 @opindex mmodel=small
13545 Assume all objects live in the lower 16MB of memory (so that their addresses
13546 can be loaded with the @code{ld24} instruction), and assume all subroutines
13547 are reachable with the @code{bl} instruction.
13548 This is the default.
13550 The addressability of a particular object can be set with the
13551 @code{model} attribute.
13553 @item -mmodel=medium
13554 @opindex mmodel=medium
13555 Assume objects may be anywhere in the 32-bit address space (the compiler
13556 will generate @code{seth/add3} instructions to load their addresses), and
13557 assume all subroutines are reachable with the @code{bl} instruction.
13559 @item -mmodel=large
13560 @opindex mmodel=large
13561 Assume objects may be anywhere in the 32-bit address space (the compiler
13562 will generate @code{seth/add3} instructions to load their addresses), and
13563 assume subroutines may not be reachable with the @code{bl} instruction
13564 (the compiler will generate the much slower @code{seth/add3/jl}
13565 instruction sequence).
13568 @opindex msdata=none
13569 Disable use of the small data area. Variables will be put into
13570 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13571 @code{section} attribute has been specified).
13572 This is the default.
13574 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13575 Objects may be explicitly put in the small data area with the
13576 @code{section} attribute using one of these sections.
13578 @item -msdata=sdata
13579 @opindex msdata=sdata
13580 Put small global and static data in the small data area, but do not
13581 generate special code to reference them.
13584 @opindex msdata=use
13585 Put small global and static data in the small data area, and generate
13586 special instructions to reference them.
13590 @cindex smaller data references
13591 Put global and static objects less than or equal to @var{num} bytes
13592 into the small data or bss sections instead of the normal data or bss
13593 sections. The default value of @var{num} is 8.
13594 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13595 for this option to have any effect.
13597 All modules should be compiled with the same @option{-G @var{num}} value.
13598 Compiling with different values of @var{num} may or may not work; if it
13599 doesn't the linker will give an error message---incorrect code will not be
13604 Makes the M32R specific code in the compiler display some statistics
13605 that might help in debugging programs.
13607 @item -malign-loops
13608 @opindex malign-loops
13609 Align all loops to a 32-byte boundary.
13611 @item -mno-align-loops
13612 @opindex mno-align-loops
13613 Do not enforce a 32-byte alignment for loops. This is the default.
13615 @item -missue-rate=@var{number}
13616 @opindex missue-rate=@var{number}
13617 Issue @var{number} instructions per cycle. @var{number} can only be 1
13620 @item -mbranch-cost=@var{number}
13621 @opindex mbranch-cost=@var{number}
13622 @var{number} can only be 1 or 2. If it is 1 then branches will be
13623 preferred over conditional code, if it is 2, then the opposite will
13626 @item -mflush-trap=@var{number}
13627 @opindex mflush-trap=@var{number}
13628 Specifies the trap number to use to flush the cache. The default is
13629 12. Valid numbers are between 0 and 15 inclusive.
13631 @item -mno-flush-trap
13632 @opindex mno-flush-trap
13633 Specifies that the cache cannot be flushed by using a trap.
13635 @item -mflush-func=@var{name}
13636 @opindex mflush-func=@var{name}
13637 Specifies the name of the operating system function to call to flush
13638 the cache. The default is @emph{_flush_cache}, but a function call
13639 will only be used if a trap is not available.
13641 @item -mno-flush-func
13642 @opindex mno-flush-func
13643 Indicates that there is no OS function for flushing the cache.
13647 @node M680x0 Options
13648 @subsection M680x0 Options
13649 @cindex M680x0 options
13651 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13652 The default settings depend on which architecture was selected when
13653 the compiler was configured; the defaults for the most common choices
13657 @item -march=@var{arch}
13659 Generate code for a specific M680x0 or ColdFire instruction set
13660 architecture. Permissible values of @var{arch} for M680x0
13661 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13662 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13663 architectures are selected according to Freescale's ISA classification
13664 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13665 @samp{isab} and @samp{isac}.
13667 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13668 code for a ColdFire target. The @var{arch} in this macro is one of the
13669 @option{-march} arguments given above.
13671 When used together, @option{-march} and @option{-mtune} select code
13672 that runs on a family of similar processors but that is optimized
13673 for a particular microarchitecture.
13675 @item -mcpu=@var{cpu}
13677 Generate code for a specific M680x0 or ColdFire processor.
13678 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13679 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13680 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13681 below, which also classifies the CPUs into families:
13683 @multitable @columnfractions 0.20 0.80
13684 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13685 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13686 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13687 @item @samp{5206e} @tab @samp{5206e}
13688 @item @samp{5208} @tab @samp{5207} @samp{5208}
13689 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13690 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13691 @item @samp{5216} @tab @samp{5214} @samp{5216}
13692 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13693 @item @samp{5225} @tab @samp{5224} @samp{5225}
13694 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13695 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13696 @item @samp{5249} @tab @samp{5249}
13697 @item @samp{5250} @tab @samp{5250}
13698 @item @samp{5271} @tab @samp{5270} @samp{5271}
13699 @item @samp{5272} @tab @samp{5272}
13700 @item @samp{5275} @tab @samp{5274} @samp{5275}
13701 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13702 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13703 @item @samp{5307} @tab @samp{5307}
13704 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13705 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13706 @item @samp{5407} @tab @samp{5407}
13707 @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}
13710 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13711 @var{arch} is compatible with @var{cpu}. Other combinations of
13712 @option{-mcpu} and @option{-march} are rejected.
13714 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13715 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13716 where the value of @var{family} is given by the table above.
13718 @item -mtune=@var{tune}
13720 Tune the code for a particular microarchitecture, within the
13721 constraints set by @option{-march} and @option{-mcpu}.
13722 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13723 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13724 and @samp{cpu32}. The ColdFire microarchitectures
13725 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13727 You can also use @option{-mtune=68020-40} for code that needs
13728 to run relatively well on 68020, 68030 and 68040 targets.
13729 @option{-mtune=68020-60} is similar but includes 68060 targets
13730 as well. These two options select the same tuning decisions as
13731 @option{-m68020-40} and @option{-m68020-60} respectively.
13733 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13734 when tuning for 680x0 architecture @var{arch}. It also defines
13735 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13736 option is used. If gcc is tuning for a range of architectures,
13737 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13738 it defines the macros for every architecture in the range.
13740 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13741 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13742 of the arguments given above.
13748 Generate output for a 68000. This is the default
13749 when the compiler is configured for 68000-based systems.
13750 It is equivalent to @option{-march=68000}.
13752 Use this option for microcontrollers with a 68000 or EC000 core,
13753 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13757 Generate output for a 68010. This is the default
13758 when the compiler is configured for 68010-based systems.
13759 It is equivalent to @option{-march=68010}.
13765 Generate output for a 68020. This is the default
13766 when the compiler is configured for 68020-based systems.
13767 It is equivalent to @option{-march=68020}.
13771 Generate output for a 68030. This is the default when the compiler is
13772 configured for 68030-based systems. It is equivalent to
13773 @option{-march=68030}.
13777 Generate output for a 68040. This is the default when the compiler is
13778 configured for 68040-based systems. It is equivalent to
13779 @option{-march=68040}.
13781 This option inhibits the use of 68881/68882 instructions that have to be
13782 emulated by software on the 68040. Use this option if your 68040 does not
13783 have code to emulate those instructions.
13787 Generate output for a 68060. This is the default when the compiler is
13788 configured for 68060-based systems. It is equivalent to
13789 @option{-march=68060}.
13791 This option inhibits the use of 68020 and 68881/68882 instructions that
13792 have to be emulated by software on the 68060. Use this option if your 68060
13793 does not have code to emulate those instructions.
13797 Generate output for a CPU32. This is the default
13798 when the compiler is configured for CPU32-based systems.
13799 It is equivalent to @option{-march=cpu32}.
13801 Use this option for microcontrollers with a
13802 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13803 68336, 68340, 68341, 68349 and 68360.
13807 Generate output for a 520X ColdFire CPU@. This is the default
13808 when the compiler is configured for 520X-based systems.
13809 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13810 in favor of that option.
13812 Use this option for microcontroller with a 5200 core, including
13813 the MCF5202, MCF5203, MCF5204 and MCF5206.
13817 Generate output for a 5206e ColdFire CPU@. The option is now
13818 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13822 Generate output for a member of the ColdFire 528X family.
13823 The option is now deprecated in favor of the equivalent
13824 @option{-mcpu=528x}.
13828 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13829 in favor of the equivalent @option{-mcpu=5307}.
13833 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13834 in favor of the equivalent @option{-mcpu=5407}.
13838 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13839 This includes use of hardware floating point instructions.
13840 The option is equivalent to @option{-mcpu=547x}, and is now
13841 deprecated in favor of that option.
13845 Generate output for a 68040, without using any of the new instructions.
13846 This results in code which can run relatively efficiently on either a
13847 68020/68881 or a 68030 or a 68040. The generated code does use the
13848 68881 instructions that are emulated on the 68040.
13850 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13854 Generate output for a 68060, without using any of the new instructions.
13855 This results in code which can run relatively efficiently on either a
13856 68020/68881 or a 68030 or a 68040. The generated code does use the
13857 68881 instructions that are emulated on the 68060.
13859 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13863 @opindex mhard-float
13865 Generate floating-point instructions. This is the default for 68020
13866 and above, and for ColdFire devices that have an FPU@. It defines the
13867 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13868 on ColdFire targets.
13871 @opindex msoft-float
13872 Do not generate floating-point instructions; use library calls instead.
13873 This is the default for 68000, 68010, and 68832 targets. It is also
13874 the default for ColdFire devices that have no FPU.
13880 Generate (do not generate) ColdFire hardware divide and remainder
13881 instructions. If @option{-march} is used without @option{-mcpu},
13882 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13883 architectures. Otherwise, the default is taken from the target CPU
13884 (either the default CPU, or the one specified by @option{-mcpu}). For
13885 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13886 @option{-mcpu=5206e}.
13888 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13892 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13893 Additionally, parameters passed on the stack are also aligned to a
13894 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13898 Do not consider type @code{int} to be 16 bits wide. This is the default.
13901 @itemx -mno-bitfield
13902 @opindex mnobitfield
13903 @opindex mno-bitfield
13904 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13905 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13909 Do use the bit-field instructions. The @option{-m68020} option implies
13910 @option{-mbitfield}. This is the default if you use a configuration
13911 designed for a 68020.
13915 Use a different function-calling convention, in which functions
13916 that take a fixed number of arguments return with the @code{rtd}
13917 instruction, which pops their arguments while returning. This
13918 saves one instruction in the caller since there is no need to pop
13919 the arguments there.
13921 This calling convention is incompatible with the one normally
13922 used on Unix, so you cannot use it if you need to call libraries
13923 compiled with the Unix compiler.
13925 Also, you must provide function prototypes for all functions that
13926 take variable numbers of arguments (including @code{printf});
13927 otherwise incorrect code will be generated for calls to those
13930 In addition, seriously incorrect code will result if you call a
13931 function with too many arguments. (Normally, extra arguments are
13932 harmlessly ignored.)
13934 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13935 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13939 Do not use the calling conventions selected by @option{-mrtd}.
13940 This is the default.
13943 @itemx -mno-align-int
13944 @opindex malign-int
13945 @opindex mno-align-int
13946 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13947 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13948 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13949 Aligning variables on 32-bit boundaries produces code that runs somewhat
13950 faster on processors with 32-bit busses at the expense of more memory.
13952 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13953 align structures containing the above types differently than
13954 most published application binary interface specifications for the m68k.
13958 Use the pc-relative addressing mode of the 68000 directly, instead of
13959 using a global offset table. At present, this option implies @option{-fpic},
13960 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13961 not presently supported with @option{-mpcrel}, though this could be supported for
13962 68020 and higher processors.
13964 @item -mno-strict-align
13965 @itemx -mstrict-align
13966 @opindex mno-strict-align
13967 @opindex mstrict-align
13968 Do not (do) assume that unaligned memory references will be handled by
13972 Generate code that allows the data segment to be located in a different
13973 area of memory from the text segment. This allows for execute in place in
13974 an environment without virtual memory management. This option implies
13977 @item -mno-sep-data
13978 Generate code that assumes that the data segment follows the text segment.
13979 This is the default.
13981 @item -mid-shared-library
13982 Generate code that supports shared libraries via the library ID method.
13983 This allows for execute in place and shared libraries in an environment
13984 without virtual memory management. This option implies @option{-fPIC}.
13986 @item -mno-id-shared-library
13987 Generate code that doesn't assume ID based shared libraries are being used.
13988 This is the default.
13990 @item -mshared-library-id=n
13991 Specified the identification number of the ID based shared library being
13992 compiled. Specifying a value of 0 will generate more compact code, specifying
13993 other values will force the allocation of that number to the current
13994 library but is no more space or time efficient than omitting this option.
14000 When generating position-independent code for ColdFire, generate code
14001 that works if the GOT has more than 8192 entries. This code is
14002 larger and slower than code generated without this option. On M680x0
14003 processors, this option is not needed; @option{-fPIC} suffices.
14005 GCC normally uses a single instruction to load values from the GOT@.
14006 While this is relatively efficient, it only works if the GOT
14007 is smaller than about 64k. Anything larger causes the linker
14008 to report an error such as:
14010 @cindex relocation truncated to fit (ColdFire)
14012 relocation truncated to fit: R_68K_GOT16O foobar
14015 If this happens, you should recompile your code with @option{-mxgot}.
14016 It should then work with very large GOTs. However, code generated with
14017 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
14018 the value of a global symbol.
14020 Note that some linkers, including newer versions of the GNU linker,
14021 can create multiple GOTs and sort GOT entries. If you have such a linker,
14022 you should only need to use @option{-mxgot} when compiling a single
14023 object file that accesses more than 8192 GOT entries. Very few do.
14025 These options have no effect unless GCC is generating
14026 position-independent code.
14030 @node MCore Options
14031 @subsection MCore Options
14032 @cindex MCore options
14034 These are the @samp{-m} options defined for the Motorola M*Core
14040 @itemx -mno-hardlit
14042 @opindex mno-hardlit
14043 Inline constants into the code stream if it can be done in two
14044 instructions or less.
14050 Use the divide instruction. (Enabled by default).
14052 @item -mrelax-immediate
14053 @itemx -mno-relax-immediate
14054 @opindex mrelax-immediate
14055 @opindex mno-relax-immediate
14056 Allow arbitrary sized immediates in bit operations.
14058 @item -mwide-bitfields
14059 @itemx -mno-wide-bitfields
14060 @opindex mwide-bitfields
14061 @opindex mno-wide-bitfields
14062 Always treat bit-fields as int-sized.
14064 @item -m4byte-functions
14065 @itemx -mno-4byte-functions
14066 @opindex m4byte-functions
14067 @opindex mno-4byte-functions
14068 Force all functions to be aligned to a four byte boundary.
14070 @item -mcallgraph-data
14071 @itemx -mno-callgraph-data
14072 @opindex mcallgraph-data
14073 @opindex mno-callgraph-data
14074 Emit callgraph information.
14077 @itemx -mno-slow-bytes
14078 @opindex mslow-bytes
14079 @opindex mno-slow-bytes
14080 Prefer word access when reading byte quantities.
14082 @item -mlittle-endian
14083 @itemx -mbig-endian
14084 @opindex mlittle-endian
14085 @opindex mbig-endian
14086 Generate code for a little endian target.
14092 Generate code for the 210 processor.
14096 Assume that run-time support has been provided and so omit the
14097 simulator library (@file{libsim.a)} from the linker command line.
14099 @item -mstack-increment=@var{size}
14100 @opindex mstack-increment
14101 Set the maximum amount for a single stack increment operation. Large
14102 values can increase the speed of programs which contain functions
14103 that need a large amount of stack space, but they can also trigger a
14104 segmentation fault if the stack is extended too much. The default
14110 @subsection MeP Options
14111 @cindex MeP options
14117 Enables the @code{abs} instruction, which is the absolute difference
14118 between two registers.
14122 Enables all the optional instructions - average, multiply, divide, bit
14123 operations, leading zero, absolute difference, min/max, clip, and
14129 Enables the @code{ave} instruction, which computes the average of two
14132 @item -mbased=@var{n}
14134 Variables of size @var{n} bytes or smaller will be placed in the
14135 @code{.based} section by default. Based variables use the @code{$tp}
14136 register as a base register, and there is a 128 byte limit to the
14137 @code{.based} section.
14141 Enables the bit operation instructions - bit test (@code{btstm}), set
14142 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14143 test-and-set (@code{tas}).
14145 @item -mc=@var{name}
14147 Selects which section constant data will be placed in. @var{name} may
14148 be @code{tiny}, @code{near}, or @code{far}.
14152 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14153 useful unless you also provide @code{-mminmax}.
14155 @item -mconfig=@var{name}
14157 Selects one of the build-in core configurations. Each MeP chip has
14158 one or more modules in it; each module has a core CPU and a variety of
14159 coprocessors, optional instructions, and peripherals. The
14160 @code{MeP-Integrator} tool, not part of GCC, provides these
14161 configurations through this option; using this option is the same as
14162 using all the corresponding command line options. The default
14163 configuration is @code{default}.
14167 Enables the coprocessor instructions. By default, this is a 32-bit
14168 coprocessor. Note that the coprocessor is normally enabled via the
14169 @code{-mconfig=} option.
14173 Enables the 32-bit coprocessor's instructions.
14177 Enables the 64-bit coprocessor's instructions.
14181 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14185 Causes constant variables to be placed in the @code{.near} section.
14189 Enables the @code{div} and @code{divu} instructions.
14193 Generate big-endian code.
14197 Generate little-endian code.
14199 @item -mio-volatile
14200 @opindex mio-volatile
14201 Tells the compiler that any variable marked with the @code{io}
14202 attribute is to be considered volatile.
14206 Causes variables to be assigned to the @code{.far} section by default.
14210 Enables the @code{leadz} (leading zero) instruction.
14214 Causes variables to be assigned to the @code{.near} section by default.
14218 Enables the @code{min} and @code{max} instructions.
14222 Enables the multiplication and multiply-accumulate instructions.
14226 Disables all the optional instructions enabled by @code{-mall-opts}.
14230 Enables the @code{repeat} and @code{erepeat} instructions, used for
14231 low-overhead looping.
14235 Causes all variables to default to the @code{.tiny} section. Note
14236 that there is a 65536 byte limit to this section. Accesses to these
14237 variables use the @code{%gp} base register.
14241 Enables the saturation instructions. Note that the compiler does not
14242 currently generate these itself, but this option is included for
14243 compatibility with other tools, like @code{as}.
14247 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14251 Link the simulator runtime libraries.
14255 Link the simulator runtime libraries, excluding built-in support
14256 for reset and exception vectors and tables.
14260 Causes all functions to default to the @code{.far} section. Without
14261 this option, functions default to the @code{.near} section.
14263 @item -mtiny=@var{n}
14265 Variables that are @var{n} bytes or smaller will be allocated to the
14266 @code{.tiny} section. These variables use the @code{$gp} base
14267 register. The default for this option is 4, but note that there's a
14268 65536 byte limit to the @code{.tiny} section.
14272 @node MicroBlaze Options
14273 @subsection MicroBlaze Options
14274 @cindex MicroBlaze Options
14279 @opindex msoft-float
14280 Use software emulation for floating point (default).
14283 @opindex mhard-float
14284 Use hardware floating point instructions.
14288 Do not optimize block moves, use @code{memcpy}.
14290 @item -mno-clearbss
14291 @opindex mno-clearbss
14292 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14294 @item -mcpu=@var{cpu-type}
14296 Use features of and schedule code for given CPU.
14297 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14298 where @var{X} is a major version, @var{YY} is the minor version, and
14299 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14300 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14302 @item -mxl-soft-mul
14303 @opindex mxl-soft-mul
14304 Use software multiply emulation (default).
14306 @item -mxl-soft-div
14307 @opindex mxl-soft-div
14308 Use software emulation for divides (default).
14310 @item -mxl-barrel-shift
14311 @opindex mxl-barrel-shift
14312 Use the hardware barrel shifter.
14314 @item -mxl-pattern-compare
14315 @opindex mxl-pattern-compare
14316 Use pattern compare instructions.
14318 @item -msmall-divides
14319 @opindex msmall-divides
14320 Use table lookup optimization for small signed integer divisions.
14322 @item -mxl-stack-check
14323 @opindex mxl-stack-check
14324 This option is deprecated. Use -fstack-check instead.
14327 @opindex mxl-gp-opt
14328 Use GP relative sdata/sbss sections.
14330 @item -mxl-multiply-high
14331 @opindex mxl-multiply-high
14332 Use multiply high instructions for high part of 32x32 multiply.
14334 @item -mxl-float-convert
14335 @opindex mxl-float-convert
14336 Use hardware floating point conversion instructions.
14338 @item -mxl-float-sqrt
14339 @opindex mxl-float-sqrt
14340 Use hardware floating point square root instruction.
14342 @item -mxl-mode-@var{app-model}
14343 Select application model @var{app-model}. Valid models are
14346 normal executable (default), uses startup code @file{crt0.o}.
14349 for use with Xilinx Microprocessor Debugger (XMD) based
14350 software intrusive debug agent called xmdstub. This uses startup file
14351 @file{crt1.o} and sets the start address of the program to be 0x800.
14354 for applications that are loaded using a bootloader.
14355 This model uses startup file @file{crt2.o} which does not contain a processor
14356 reset vector handler. This is suitable for transferring control on a
14357 processor reset to the bootloader rather than the application.
14360 for applications that do not require any of the
14361 MicroBlaze vectors. This option may be useful for applications running
14362 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14365 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14366 @option{-mxl-mode-@var{app-model}}.
14371 @subsection MIPS Options
14372 @cindex MIPS options
14378 Generate big-endian code.
14382 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14385 @item -march=@var{arch}
14387 Generate code that will run on @var{arch}, which can be the name of a
14388 generic MIPS ISA, or the name of a particular processor.
14390 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14391 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14392 The processor names are:
14393 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14394 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14395 @samp{5kc}, @samp{5kf},
14397 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14398 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14399 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14400 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14401 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14402 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14406 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14407 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14408 @samp{rm7000}, @samp{rm9000},
14409 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14412 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14413 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14415 The special value @samp{from-abi} selects the
14416 most compatible architecture for the selected ABI (that is,
14417 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14419 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
14420 which selects the best architecture option for the host processor.
14421 @option{-march=native} has no effect if GCC does not recognize
14424 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14425 (for example, @samp{-march=r2k}). Prefixes are optional, and
14426 @samp{vr} may be written @samp{r}.
14428 Names of the form @samp{@var{n}f2_1} refer to processors with
14429 FPUs clocked at half the rate of the core, names of the form
14430 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14431 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14432 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14433 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14434 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14435 accepted as synonyms for @samp{@var{n}f1_1}.
14437 GCC defines two macros based on the value of this option. The first
14438 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14439 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14440 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14441 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14442 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14444 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14445 above. In other words, it will have the full prefix and will not
14446 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14447 the macro names the resolved architecture (either @samp{"mips1"} or
14448 @samp{"mips3"}). It names the default architecture when no
14449 @option{-march} option is given.
14451 @item -mtune=@var{arch}
14453 Optimize for @var{arch}. Among other things, this option controls
14454 the way instructions are scheduled, and the perceived cost of arithmetic
14455 operations. The list of @var{arch} values is the same as for
14458 When this option is not used, GCC will optimize for the processor
14459 specified by @option{-march}. By using @option{-march} and
14460 @option{-mtune} together, it is possible to generate code that will
14461 run on a family of processors, but optimize the code for one
14462 particular member of that family.
14464 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14465 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14466 @samp{-march} ones described above.
14470 Equivalent to @samp{-march=mips1}.
14474 Equivalent to @samp{-march=mips2}.
14478 Equivalent to @samp{-march=mips3}.
14482 Equivalent to @samp{-march=mips4}.
14486 Equivalent to @samp{-march=mips32}.
14490 Equivalent to @samp{-march=mips32r2}.
14494 Equivalent to @samp{-march=mips64}.
14498 Equivalent to @samp{-march=mips64r2}.
14503 @opindex mno-mips16
14504 Generate (do not generate) MIPS16 code. If GCC is targetting a
14505 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14507 MIPS16 code generation can also be controlled on a per-function basis
14508 by means of @code{mips16} and @code{nomips16} attributes.
14509 @xref{Function Attributes}, for more information.
14511 @item -mflip-mips16
14512 @opindex mflip-mips16
14513 Generate MIPS16 code on alternating functions. This option is provided
14514 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14515 not intended for ordinary use in compiling user code.
14517 @item -minterlink-mips16
14518 @itemx -mno-interlink-mips16
14519 @opindex minterlink-mips16
14520 @opindex mno-interlink-mips16
14521 Require (do not require) that non-MIPS16 code be link-compatible with
14524 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14525 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14526 therefore disables direct jumps unless GCC knows that the target of the
14527 jump is not MIPS16.
14539 Generate code for the given ABI@.
14541 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14542 generates 64-bit code when you select a 64-bit architecture, but you
14543 can use @option{-mgp32} to get 32-bit code instead.
14545 For information about the O64 ABI, see
14546 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14548 GCC supports a variant of the o32 ABI in which floating-point registers
14549 are 64 rather than 32 bits wide. You can select this combination with
14550 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14551 and @samp{mfhc1} instructions and is therefore only supported for
14552 MIPS32R2 processors.
14554 The register assignments for arguments and return values remain the
14555 same, but each scalar value is passed in a single 64-bit register
14556 rather than a pair of 32-bit registers. For example, scalar
14557 floating-point values are returned in @samp{$f0} only, not a
14558 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14559 remains the same, but all 64 bits are saved.
14562 @itemx -mno-abicalls
14564 @opindex mno-abicalls
14565 Generate (do not generate) code that is suitable for SVR4-style
14566 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14571 Generate (do not generate) code that is fully position-independent,
14572 and that can therefore be linked into shared libraries. This option
14573 only affects @option{-mabicalls}.
14575 All @option{-mabicalls} code has traditionally been position-independent,
14576 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14577 as an extension, the GNU toolchain allows executables to use absolute
14578 accesses for locally-binding symbols. It can also use shorter GP
14579 initialization sequences and generate direct calls to locally-defined
14580 functions. This mode is selected by @option{-mno-shared}.
14582 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14583 objects that can only be linked by the GNU linker. However, the option
14584 does not affect the ABI of the final executable; it only affects the ABI
14585 of relocatable objects. Using @option{-mno-shared} will generally make
14586 executables both smaller and quicker.
14588 @option{-mshared} is the default.
14594 Assume (do not assume) that the static and dynamic linkers
14595 support PLTs and copy relocations. This option only affects
14596 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14597 has no effect without @samp{-msym32}.
14599 You can make @option{-mplt} the default by configuring
14600 GCC with @option{--with-mips-plt}. The default is
14601 @option{-mno-plt} otherwise.
14607 Lift (do not lift) the usual restrictions on the size of the global
14610 GCC normally uses a single instruction to load values from the GOT@.
14611 While this is relatively efficient, it will only work if the GOT
14612 is smaller than about 64k. Anything larger will cause the linker
14613 to report an error such as:
14615 @cindex relocation truncated to fit (MIPS)
14617 relocation truncated to fit: R_MIPS_GOT16 foobar
14620 If this happens, you should recompile your code with @option{-mxgot}.
14621 It should then work with very large GOTs, although it will also be
14622 less efficient, since it will take three instructions to fetch the
14623 value of a global symbol.
14625 Note that some linkers can create multiple GOTs. If you have such a
14626 linker, you should only need to use @option{-mxgot} when a single object
14627 file accesses more than 64k's worth of GOT entries. Very few do.
14629 These options have no effect unless GCC is generating position
14634 Assume that general-purpose registers are 32 bits wide.
14638 Assume that general-purpose registers are 64 bits wide.
14642 Assume that floating-point registers are 32 bits wide.
14646 Assume that floating-point registers are 64 bits wide.
14649 @opindex mhard-float
14650 Use floating-point coprocessor instructions.
14653 @opindex msoft-float
14654 Do not use floating-point coprocessor instructions. Implement
14655 floating-point calculations using library calls instead.
14657 @item -msingle-float
14658 @opindex msingle-float
14659 Assume that the floating-point coprocessor only supports single-precision
14662 @item -mdouble-float
14663 @opindex mdouble-float
14664 Assume that the floating-point coprocessor supports double-precision
14665 operations. This is the default.
14671 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14672 implement atomic memory built-in functions. When neither option is
14673 specified, GCC will use the instructions if the target architecture
14676 @option{-mllsc} is useful if the runtime environment can emulate the
14677 instructions and @option{-mno-llsc} can be useful when compiling for
14678 nonstandard ISAs. You can make either option the default by
14679 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14680 respectively. @option{--with-llsc} is the default for some
14681 configurations; see the installation documentation for details.
14687 Use (do not use) revision 1 of the MIPS DSP ASE@.
14688 @xref{MIPS DSP Built-in Functions}. This option defines the
14689 preprocessor macro @samp{__mips_dsp}. It also defines
14690 @samp{__mips_dsp_rev} to 1.
14696 Use (do not use) revision 2 of the MIPS DSP ASE@.
14697 @xref{MIPS DSP Built-in Functions}. This option defines the
14698 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14699 It also defines @samp{__mips_dsp_rev} to 2.
14702 @itemx -mno-smartmips
14703 @opindex msmartmips
14704 @opindex mno-smartmips
14705 Use (do not use) the MIPS SmartMIPS ASE.
14707 @item -mpaired-single
14708 @itemx -mno-paired-single
14709 @opindex mpaired-single
14710 @opindex mno-paired-single
14711 Use (do not use) paired-single floating-point instructions.
14712 @xref{MIPS Paired-Single Support}. This option requires
14713 hardware floating-point support to be enabled.
14719 Use (do not use) MIPS Digital Media Extension instructions.
14720 This option can only be used when generating 64-bit code and requires
14721 hardware floating-point support to be enabled.
14726 @opindex mno-mips3d
14727 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14728 The option @option{-mips3d} implies @option{-mpaired-single}.
14734 Use (do not use) MT Multithreading instructions.
14738 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14739 an explanation of the default and the way that the pointer size is
14744 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14746 The default size of @code{int}s, @code{long}s and pointers depends on
14747 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14748 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14749 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14750 or the same size as integer registers, whichever is smaller.
14756 Assume (do not assume) that all symbols have 32-bit values, regardless
14757 of the selected ABI@. This option is useful in combination with
14758 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14759 to generate shorter and faster references to symbolic addresses.
14763 Put definitions of externally-visible data in a small data section
14764 if that data is no bigger than @var{num} bytes. GCC can then access
14765 the data more efficiently; see @option{-mgpopt} for details.
14767 The default @option{-G} option depends on the configuration.
14769 @item -mlocal-sdata
14770 @itemx -mno-local-sdata
14771 @opindex mlocal-sdata
14772 @opindex mno-local-sdata
14773 Extend (do not extend) the @option{-G} behavior to local data too,
14774 such as to static variables in C@. @option{-mlocal-sdata} is the
14775 default for all configurations.
14777 If the linker complains that an application is using too much small data,
14778 you might want to try rebuilding the less performance-critical parts with
14779 @option{-mno-local-sdata}. You might also want to build large
14780 libraries with @option{-mno-local-sdata}, so that the libraries leave
14781 more room for the main program.
14783 @item -mextern-sdata
14784 @itemx -mno-extern-sdata
14785 @opindex mextern-sdata
14786 @opindex mno-extern-sdata
14787 Assume (do not assume) that externally-defined data will be in
14788 a small data section if that data is within the @option{-G} limit.
14789 @option{-mextern-sdata} is the default for all configurations.
14791 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14792 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14793 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14794 is placed in a small data section. If @var{Var} is defined by another
14795 module, you must either compile that module with a high-enough
14796 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14797 definition. If @var{Var} is common, you must link the application
14798 with a high-enough @option{-G} setting.
14800 The easiest way of satisfying these restrictions is to compile
14801 and link every module with the same @option{-G} option. However,
14802 you may wish to build a library that supports several different
14803 small data limits. You can do this by compiling the library with
14804 the highest supported @option{-G} setting and additionally using
14805 @option{-mno-extern-sdata} to stop the library from making assumptions
14806 about externally-defined data.
14812 Use (do not use) GP-relative accesses for symbols that are known to be
14813 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14814 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14817 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14818 might not hold the value of @code{_gp}. For example, if the code is
14819 part of a library that might be used in a boot monitor, programs that
14820 call boot monitor routines will pass an unknown value in @code{$gp}.
14821 (In such situations, the boot monitor itself would usually be compiled
14822 with @option{-G0}.)
14824 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14825 @option{-mno-extern-sdata}.
14827 @item -membedded-data
14828 @itemx -mno-embedded-data
14829 @opindex membedded-data
14830 @opindex mno-embedded-data
14831 Allocate variables to the read-only data section first if possible, then
14832 next in the small data section if possible, otherwise in data. This gives
14833 slightly slower code than the default, but reduces the amount of RAM required
14834 when executing, and thus may be preferred for some embedded systems.
14836 @item -muninit-const-in-rodata
14837 @itemx -mno-uninit-const-in-rodata
14838 @opindex muninit-const-in-rodata
14839 @opindex mno-uninit-const-in-rodata
14840 Put uninitialized @code{const} variables in the read-only data section.
14841 This option is only meaningful in conjunction with @option{-membedded-data}.
14843 @item -mcode-readable=@var{setting}
14844 @opindex mcode-readable
14845 Specify whether GCC may generate code that reads from executable sections.
14846 There are three possible settings:
14849 @item -mcode-readable=yes
14850 Instructions may freely access executable sections. This is the
14853 @item -mcode-readable=pcrel
14854 MIPS16 PC-relative load instructions can access executable sections,
14855 but other instructions must not do so. This option is useful on 4KSc
14856 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14857 It is also useful on processors that can be configured to have a dual
14858 instruction/data SRAM interface and that, like the M4K, automatically
14859 redirect PC-relative loads to the instruction RAM.
14861 @item -mcode-readable=no
14862 Instructions must not access executable sections. This option can be
14863 useful on targets that are configured to have a dual instruction/data
14864 SRAM interface but that (unlike the M4K) do not automatically redirect
14865 PC-relative loads to the instruction RAM.
14868 @item -msplit-addresses
14869 @itemx -mno-split-addresses
14870 @opindex msplit-addresses
14871 @opindex mno-split-addresses
14872 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14873 relocation operators. This option has been superseded by
14874 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14876 @item -mexplicit-relocs
14877 @itemx -mno-explicit-relocs
14878 @opindex mexplicit-relocs
14879 @opindex mno-explicit-relocs
14880 Use (do not use) assembler relocation operators when dealing with symbolic
14881 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14882 is to use assembler macros instead.
14884 @option{-mexplicit-relocs} is the default if GCC was configured
14885 to use an assembler that supports relocation operators.
14887 @item -mcheck-zero-division
14888 @itemx -mno-check-zero-division
14889 @opindex mcheck-zero-division
14890 @opindex mno-check-zero-division
14891 Trap (do not trap) on integer division by zero.
14893 The default is @option{-mcheck-zero-division}.
14895 @item -mdivide-traps
14896 @itemx -mdivide-breaks
14897 @opindex mdivide-traps
14898 @opindex mdivide-breaks
14899 MIPS systems check for division by zero by generating either a
14900 conditional trap or a break instruction. Using traps results in
14901 smaller code, but is only supported on MIPS II and later. Also, some
14902 versions of the Linux kernel have a bug that prevents trap from
14903 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14904 allow conditional traps on architectures that support them and
14905 @option{-mdivide-breaks} to force the use of breaks.
14907 The default is usually @option{-mdivide-traps}, but this can be
14908 overridden at configure time using @option{--with-divide=breaks}.
14909 Divide-by-zero checks can be completely disabled using
14910 @option{-mno-check-zero-division}.
14915 @opindex mno-memcpy
14916 Force (do not force) the use of @code{memcpy()} for non-trivial block
14917 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14918 most constant-sized copies.
14921 @itemx -mno-long-calls
14922 @opindex mlong-calls
14923 @opindex mno-long-calls
14924 Disable (do not disable) use of the @code{jal} instruction. Calling
14925 functions using @code{jal} is more efficient but requires the caller
14926 and callee to be in the same 256 megabyte segment.
14928 This option has no effect on abicalls code. The default is
14929 @option{-mno-long-calls}.
14935 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14936 instructions, as provided by the R4650 ISA@.
14939 @itemx -mno-fused-madd
14940 @opindex mfused-madd
14941 @opindex mno-fused-madd
14942 Enable (disable) use of the floating point multiply-accumulate
14943 instructions, when they are available. The default is
14944 @option{-mfused-madd}.
14946 When multiply-accumulate instructions are used, the intermediate
14947 product is calculated to infinite precision and is not subject to
14948 the FCSR Flush to Zero bit. This may be undesirable in some
14953 Tell the MIPS assembler to not run its preprocessor over user
14954 assembler files (with a @samp{.s} suffix) when assembling them.
14959 @opindex mno-fix-24k
14960 Work around the 24K E48 (lost data on stores during refill) errata.
14961 The workarounds are implemented by the assembler rather than by GCC.
14964 @itemx -mno-fix-r4000
14965 @opindex mfix-r4000
14966 @opindex mno-fix-r4000
14967 Work around certain R4000 CPU errata:
14970 A double-word or a variable shift may give an incorrect result if executed
14971 immediately after starting an integer division.
14973 A double-word or a variable shift may give an incorrect result if executed
14974 while an integer multiplication is in progress.
14976 An integer division may give an incorrect result if started in a delay slot
14977 of a taken branch or a jump.
14981 @itemx -mno-fix-r4400
14982 @opindex mfix-r4400
14983 @opindex mno-fix-r4400
14984 Work around certain R4400 CPU errata:
14987 A double-word or a variable shift may give an incorrect result if executed
14988 immediately after starting an integer division.
14992 @itemx -mno-fix-r10000
14993 @opindex mfix-r10000
14994 @opindex mno-fix-r10000
14995 Work around certain R10000 errata:
14998 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14999 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
15002 This option can only be used if the target architecture supports
15003 branch-likely instructions. @option{-mfix-r10000} is the default when
15004 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
15008 @itemx -mno-fix-vr4120
15009 @opindex mfix-vr4120
15010 Work around certain VR4120 errata:
15013 @code{dmultu} does not always produce the correct result.
15015 @code{div} and @code{ddiv} do not always produce the correct result if one
15016 of the operands is negative.
15018 The workarounds for the division errata rely on special functions in
15019 @file{libgcc.a}. At present, these functions are only provided by
15020 the @code{mips64vr*-elf} configurations.
15022 Other VR4120 errata require a nop to be inserted between certain pairs of
15023 instructions. These errata are handled by the assembler, not by GCC itself.
15026 @opindex mfix-vr4130
15027 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
15028 workarounds are implemented by the assembler rather than by GCC,
15029 although GCC will avoid using @code{mflo} and @code{mfhi} if the
15030 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
15031 instructions are available instead.
15034 @itemx -mno-fix-sb1
15036 Work around certain SB-1 CPU core errata.
15037 (This flag currently works around the SB-1 revision 2
15038 ``F1'' and ``F2'' floating point errata.)
15040 @item -mr10k-cache-barrier=@var{setting}
15041 @opindex mr10k-cache-barrier
15042 Specify whether GCC should insert cache barriers to avoid the
15043 side-effects of speculation on R10K processors.
15045 In common with many processors, the R10K tries to predict the outcome
15046 of a conditional branch and speculatively executes instructions from
15047 the ``taken'' branch. It later aborts these instructions if the
15048 predicted outcome was wrong. However, on the R10K, even aborted
15049 instructions can have side effects.
15051 This problem only affects kernel stores and, depending on the system,
15052 kernel loads. As an example, a speculatively-executed store may load
15053 the target memory into cache and mark the cache line as dirty, even if
15054 the store itself is later aborted. If a DMA operation writes to the
15055 same area of memory before the ``dirty'' line is flushed, the cached
15056 data will overwrite the DMA-ed data. See the R10K processor manual
15057 for a full description, including other potential problems.
15059 One workaround is to insert cache barrier instructions before every memory
15060 access that might be speculatively executed and that might have side
15061 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
15062 controls GCC's implementation of this workaround. It assumes that
15063 aborted accesses to any byte in the following regions will not have
15068 the memory occupied by the current function's stack frame;
15071 the memory occupied by an incoming stack argument;
15074 the memory occupied by an object with a link-time-constant address.
15077 It is the kernel's responsibility to ensure that speculative
15078 accesses to these regions are indeed safe.
15080 If the input program contains a function declaration such as:
15086 then the implementation of @code{foo} must allow @code{j foo} and
15087 @code{jal foo} to be executed speculatively. GCC honors this
15088 restriction for functions it compiles itself. It expects non-GCC
15089 functions (such as hand-written assembly code) to do the same.
15091 The option has three forms:
15094 @item -mr10k-cache-barrier=load-store
15095 Insert a cache barrier before a load or store that might be
15096 speculatively executed and that might have side effects even
15099 @item -mr10k-cache-barrier=store
15100 Insert a cache barrier before a store that might be speculatively
15101 executed and that might have side effects even if aborted.
15103 @item -mr10k-cache-barrier=none
15104 Disable the insertion of cache barriers. This is the default setting.
15107 @item -mflush-func=@var{func}
15108 @itemx -mno-flush-func
15109 @opindex mflush-func
15110 Specifies the function to call to flush the I and D caches, or to not
15111 call any such function. If called, the function must take the same
15112 arguments as the common @code{_flush_func()}, that is, the address of the
15113 memory range for which the cache is being flushed, the size of the
15114 memory range, and the number 3 (to flush both caches). The default
15115 depends on the target GCC was configured for, but commonly is either
15116 @samp{_flush_func} or @samp{__cpu_flush}.
15118 @item mbranch-cost=@var{num}
15119 @opindex mbranch-cost
15120 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15121 This cost is only a heuristic and is not guaranteed to produce
15122 consistent results across releases. A zero cost redundantly selects
15123 the default, which is based on the @option{-mtune} setting.
15125 @item -mbranch-likely
15126 @itemx -mno-branch-likely
15127 @opindex mbranch-likely
15128 @opindex mno-branch-likely
15129 Enable or disable use of Branch Likely instructions, regardless of the
15130 default for the selected architecture. By default, Branch Likely
15131 instructions may be generated if they are supported by the selected
15132 architecture. An exception is for the MIPS32 and MIPS64 architectures
15133 and processors which implement those architectures; for those, Branch
15134 Likely instructions will not be generated by default because the MIPS32
15135 and MIPS64 architectures specifically deprecate their use.
15137 @item -mfp-exceptions
15138 @itemx -mno-fp-exceptions
15139 @opindex mfp-exceptions
15140 Specifies whether FP exceptions are enabled. This affects how we schedule
15141 FP instructions for some processors. The default is that FP exceptions are
15144 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15145 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15148 @item -mvr4130-align
15149 @itemx -mno-vr4130-align
15150 @opindex mvr4130-align
15151 The VR4130 pipeline is two-way superscalar, but can only issue two
15152 instructions together if the first one is 8-byte aligned. When this
15153 option is enabled, GCC will align pairs of instructions that it
15154 thinks should execute in parallel.
15156 This option only has an effect when optimizing for the VR4130.
15157 It normally makes code faster, but at the expense of making it bigger.
15158 It is enabled by default at optimization level @option{-O3}.
15163 Enable (disable) generation of @code{synci} instructions on
15164 architectures that support it. The @code{synci} instructions (if
15165 enabled) will be generated when @code{__builtin___clear_cache()} is
15168 This option defaults to @code{-mno-synci}, but the default can be
15169 overridden by configuring with @code{--with-synci}.
15171 When compiling code for single processor systems, it is generally safe
15172 to use @code{synci}. However, on many multi-core (SMP) systems, it
15173 will not invalidate the instruction caches on all cores and may lead
15174 to undefined behavior.
15176 @item -mrelax-pic-calls
15177 @itemx -mno-relax-pic-calls
15178 @opindex mrelax-pic-calls
15179 Try to turn PIC calls that are normally dispatched via register
15180 @code{$25} into direct calls. This is only possible if the linker can
15181 resolve the destination at link-time and if the destination is within
15182 range for a direct call.
15184 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15185 an assembler and a linker that supports the @code{.reloc} assembly
15186 directive and @code{-mexplicit-relocs} is in effect. With
15187 @code{-mno-explicit-relocs}, this optimization can be performed by the
15188 assembler and the linker alone without help from the compiler.
15190 @item -mmcount-ra-address
15191 @itemx -mno-mcount-ra-address
15192 @opindex mmcount-ra-address
15193 @opindex mno-mcount-ra-address
15194 Emit (do not emit) code that allows @code{_mcount} to modify the
15195 calling function's return address. When enabled, this option extends
15196 the usual @code{_mcount} interface with a new @var{ra-address}
15197 parameter, which has type @code{intptr_t *} and is passed in register
15198 @code{$12}. @code{_mcount} can then modify the return address by
15199 doing both of the following:
15202 Returning the new address in register @code{$31}.
15204 Storing the new address in @code{*@var{ra-address}},
15205 if @var{ra-address} is nonnull.
15208 The default is @option{-mno-mcount-ra-address}.
15213 @subsection MMIX Options
15214 @cindex MMIX Options
15216 These options are defined for the MMIX:
15220 @itemx -mno-libfuncs
15222 @opindex mno-libfuncs
15223 Specify that intrinsic library functions are being compiled, passing all
15224 values in registers, no matter the size.
15227 @itemx -mno-epsilon
15229 @opindex mno-epsilon
15230 Generate floating-point comparison instructions that compare with respect
15231 to the @code{rE} epsilon register.
15233 @item -mabi=mmixware
15235 @opindex mabi=mmixware
15237 Generate code that passes function parameters and return values that (in
15238 the called function) are seen as registers @code{$0} and up, as opposed to
15239 the GNU ABI which uses global registers @code{$231} and up.
15241 @item -mzero-extend
15242 @itemx -mno-zero-extend
15243 @opindex mzero-extend
15244 @opindex mno-zero-extend
15245 When reading data from memory in sizes shorter than 64 bits, use (do not
15246 use) zero-extending load instructions by default, rather than
15247 sign-extending ones.
15250 @itemx -mno-knuthdiv
15252 @opindex mno-knuthdiv
15253 Make the result of a division yielding a remainder have the same sign as
15254 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15255 remainder follows the sign of the dividend. Both methods are
15256 arithmetically valid, the latter being almost exclusively used.
15258 @item -mtoplevel-symbols
15259 @itemx -mno-toplevel-symbols
15260 @opindex mtoplevel-symbols
15261 @opindex mno-toplevel-symbols
15262 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15263 code can be used with the @code{PREFIX} assembly directive.
15267 Generate an executable in the ELF format, rather than the default
15268 @samp{mmo} format used by the @command{mmix} simulator.
15270 @item -mbranch-predict
15271 @itemx -mno-branch-predict
15272 @opindex mbranch-predict
15273 @opindex mno-branch-predict
15274 Use (do not use) the probable-branch instructions, when static branch
15275 prediction indicates a probable branch.
15277 @item -mbase-addresses
15278 @itemx -mno-base-addresses
15279 @opindex mbase-addresses
15280 @opindex mno-base-addresses
15281 Generate (do not generate) code that uses @emph{base addresses}. Using a
15282 base address automatically generates a request (handled by the assembler
15283 and the linker) for a constant to be set up in a global register. The
15284 register is used for one or more base address requests within the range 0
15285 to 255 from the value held in the register. The generally leads to short
15286 and fast code, but the number of different data items that can be
15287 addressed is limited. This means that a program that uses lots of static
15288 data may require @option{-mno-base-addresses}.
15290 @item -msingle-exit
15291 @itemx -mno-single-exit
15292 @opindex msingle-exit
15293 @opindex mno-single-exit
15294 Force (do not force) generated code to have a single exit point in each
15298 @node MN10300 Options
15299 @subsection MN10300 Options
15300 @cindex MN10300 options
15302 These @option{-m} options are defined for Matsushita MN10300 architectures:
15307 Generate code to avoid bugs in the multiply instructions for the MN10300
15308 processors. This is the default.
15310 @item -mno-mult-bug
15311 @opindex mno-mult-bug
15312 Do not generate code to avoid bugs in the multiply instructions for the
15313 MN10300 processors.
15317 Generate code which uses features specific to the AM33 processor.
15321 Do not generate code which uses features specific to the AM33 processor. This
15326 Generate code which uses features specific to the AM33/2.0 processor.
15330 Generate code which uses features specific to the AM34 processor.
15332 @item -mtune=@var{cpu-type}
15334 Use the timing characteristics of the indicated CPU type when
15335 scheduling instructions. This does not change the targeted processor
15336 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15337 @samp{am33-2} or @samp{am34}.
15339 @item -mreturn-pointer-on-d0
15340 @opindex mreturn-pointer-on-d0
15341 When generating a function which returns a pointer, return the pointer
15342 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15343 only in a0, and attempts to call such functions without a prototype
15344 would result in errors. Note that this option is on by default; use
15345 @option{-mno-return-pointer-on-d0} to disable it.
15349 Do not link in the C run-time initialization object file.
15353 Indicate to the linker that it should perform a relaxation optimization pass
15354 to shorten branches, calls and absolute memory addresses. This option only
15355 has an effect when used on the command line for the final link step.
15357 This option makes symbolic debugging impossible.
15361 Allow the compiler to generate @emph{Long Instruction Word}
15362 instructions if the target is the @samp{AM33} or later. This is the
15363 default. This option defines the preprocessor macro @samp{__LIW__}.
15367 Do not allow the compiler to generate @emph{Long Instruction Word}
15368 instructions. This option defines the preprocessor macro
15373 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15374 instructions if the target is the @samp{AM33} or later. This is the
15375 default. This option defines the preprocessor macro @samp{__SETLB__}.
15379 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15380 instructions. This option defines the preprocessor macro
15381 @samp{__NO_SETLB__}.
15385 @node PDP-11 Options
15386 @subsection PDP-11 Options
15387 @cindex PDP-11 Options
15389 These options are defined for the PDP-11:
15394 Use hardware FPP floating point. This is the default. (FIS floating
15395 point on the PDP-11/40 is not supported.)
15398 @opindex msoft-float
15399 Do not use hardware floating point.
15403 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15407 Return floating-point results in memory. This is the default.
15411 Generate code for a PDP-11/40.
15415 Generate code for a PDP-11/45. This is the default.
15419 Generate code for a PDP-11/10.
15421 @item -mbcopy-builtin
15422 @opindex mbcopy-builtin
15423 Use inline @code{movmemhi} patterns for copying memory. This is the
15428 Do not use inline @code{movmemhi} patterns for copying memory.
15434 Use 16-bit @code{int}. This is the default.
15440 Use 32-bit @code{int}.
15443 @itemx -mno-float32
15445 @opindex mno-float32
15446 Use 64-bit @code{float}. This is the default.
15449 @itemx -mno-float64
15451 @opindex mno-float64
15452 Use 32-bit @code{float}.
15456 Use @code{abshi2} pattern. This is the default.
15460 Do not use @code{abshi2} pattern.
15462 @item -mbranch-expensive
15463 @opindex mbranch-expensive
15464 Pretend that branches are expensive. This is for experimenting with
15465 code generation only.
15467 @item -mbranch-cheap
15468 @opindex mbranch-cheap
15469 Do not pretend that branches are expensive. This is the default.
15473 Use Unix assembler syntax. This is the default when configured for
15474 @samp{pdp11-*-bsd}.
15478 Use DEC assembler syntax. This is the default when configured for any
15479 PDP-11 target other than @samp{pdp11-*-bsd}.
15482 @node picoChip Options
15483 @subsection picoChip Options
15484 @cindex picoChip options
15486 These @samp{-m} options are defined for picoChip implementations:
15490 @item -mae=@var{ae_type}
15492 Set the instruction set, register set, and instruction scheduling
15493 parameters for array element type @var{ae_type}. Supported values
15494 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15496 @option{-mae=ANY} selects a completely generic AE type. Code
15497 generated with this option will run on any of the other AE types. The
15498 code will not be as efficient as it would be if compiled for a specific
15499 AE type, and some types of operation (e.g., multiplication) will not
15500 work properly on all types of AE.
15502 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15503 for compiled code, and is the default.
15505 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15506 option may suffer from poor performance of byte (char) manipulation,
15507 since the DSP AE does not provide hardware support for byte load/stores.
15509 @item -msymbol-as-address
15510 Enable the compiler to directly use a symbol name as an address in a
15511 load/store instruction, without first loading it into a
15512 register. Typically, the use of this option will generate larger
15513 programs, which run faster than when the option isn't used. However, the
15514 results vary from program to program, so it is left as a user option,
15515 rather than being permanently enabled.
15517 @item -mno-inefficient-warnings
15518 Disables warnings about the generation of inefficient code. These
15519 warnings can be generated, for example, when compiling code which
15520 performs byte-level memory operations on the MAC AE type. The MAC AE has
15521 no hardware support for byte-level memory operations, so all byte
15522 load/stores must be synthesized from word load/store operations. This is
15523 inefficient and a warning will be generated indicating to the programmer
15524 that they should rewrite the code to avoid byte operations, or to target
15525 an AE type which has the necessary hardware support. This option enables
15526 the warning to be turned off.
15530 @node PowerPC Options
15531 @subsection PowerPC Options
15532 @cindex PowerPC options
15534 These are listed under @xref{RS/6000 and PowerPC Options}.
15536 @node RS/6000 and PowerPC Options
15537 @subsection IBM RS/6000 and PowerPC Options
15538 @cindex RS/6000 and PowerPC Options
15539 @cindex IBM RS/6000 and PowerPC Options
15541 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15548 @itemx -mno-powerpc
15549 @itemx -mpowerpc-gpopt
15550 @itemx -mno-powerpc-gpopt
15551 @itemx -mpowerpc-gfxopt
15552 @itemx -mno-powerpc-gfxopt
15555 @itemx -mno-powerpc64
15559 @itemx -mno-popcntb
15561 @itemx -mno-popcntd
15570 @itemx -mno-hard-dfp
15574 @opindex mno-power2
15576 @opindex mno-powerpc
15577 @opindex mpowerpc-gpopt
15578 @opindex mno-powerpc-gpopt
15579 @opindex mpowerpc-gfxopt
15580 @opindex mno-powerpc-gfxopt
15581 @opindex mpowerpc64
15582 @opindex mno-powerpc64
15586 @opindex mno-popcntb
15588 @opindex mno-popcntd
15594 @opindex mno-mfpgpr
15596 @opindex mno-hard-dfp
15597 GCC supports two related instruction set architectures for the
15598 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15599 instructions supported by the @samp{rios} chip set used in the original
15600 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15601 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15602 the IBM 4xx, 6xx, and follow-on microprocessors.
15604 Neither architecture is a subset of the other. However there is a
15605 large common subset of instructions supported by both. An MQ
15606 register is included in processors supporting the POWER architecture.
15608 You use these options to specify which instructions are available on the
15609 processor you are using. The default value of these options is
15610 determined when configuring GCC@. Specifying the
15611 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15612 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15613 rather than the options listed above.
15615 The @option{-mpower} option allows GCC to generate instructions that
15616 are found only in the POWER architecture and to use the MQ register.
15617 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15618 to generate instructions that are present in the POWER2 architecture but
15619 not the original POWER architecture.
15621 The @option{-mpowerpc} option allows GCC to generate instructions that
15622 are found only in the 32-bit subset of the PowerPC architecture.
15623 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15624 GCC to use the optional PowerPC architecture instructions in the
15625 General Purpose group, including floating-point square root. Specifying
15626 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15627 use the optional PowerPC architecture instructions in the Graphics
15628 group, including floating-point select.
15630 The @option{-mmfcrf} option allows GCC to generate the move from
15631 condition register field instruction implemented on the POWER4
15632 processor and other processors that support the PowerPC V2.01
15634 The @option{-mpopcntb} option allows GCC to generate the popcount and
15635 double precision FP reciprocal estimate instruction implemented on the
15636 POWER5 processor and other processors that support the PowerPC V2.02
15638 The @option{-mpopcntd} option allows GCC to generate the popcount
15639 instruction implemented on the POWER7 processor and other processors
15640 that support the PowerPC V2.06 architecture.
15641 The @option{-mfprnd} option allows GCC to generate the FP round to
15642 integer instructions implemented on the POWER5+ processor and other
15643 processors that support the PowerPC V2.03 architecture.
15644 The @option{-mcmpb} option allows GCC to generate the compare bytes
15645 instruction implemented on the POWER6 processor and other processors
15646 that support the PowerPC V2.05 architecture.
15647 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15648 general purpose register instructions implemented on the POWER6X
15649 processor and other processors that support the extended PowerPC V2.05
15651 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15652 point instructions implemented on some POWER processors.
15654 The @option{-mpowerpc64} option allows GCC to generate the additional
15655 64-bit instructions that are found in the full PowerPC64 architecture
15656 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15657 @option{-mno-powerpc64}.
15659 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15660 will use only the instructions in the common subset of both
15661 architectures plus some special AIX common-mode calls, and will not use
15662 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15663 permits GCC to use any instruction from either architecture and to
15664 allow use of the MQ register; specify this for the Motorola MPC601.
15666 @item -mnew-mnemonics
15667 @itemx -mold-mnemonics
15668 @opindex mnew-mnemonics
15669 @opindex mold-mnemonics
15670 Select which mnemonics to use in the generated assembler code. With
15671 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15672 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15673 assembler mnemonics defined for the POWER architecture. Instructions
15674 defined in only one architecture have only one mnemonic; GCC uses that
15675 mnemonic irrespective of which of these options is specified.
15677 GCC defaults to the mnemonics appropriate for the architecture in
15678 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15679 value of these option. Unless you are building a cross-compiler, you
15680 should normally not specify either @option{-mnew-mnemonics} or
15681 @option{-mold-mnemonics}, but should instead accept the default.
15683 @item -mcpu=@var{cpu_type}
15685 Set architecture type, register usage, choice of mnemonics, and
15686 instruction scheduling parameters for machine type @var{cpu_type}.
15687 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15688 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15689 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15690 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15691 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15692 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15693 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15694 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15695 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15696 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15697 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15699 @option{-mcpu=common} selects a completely generic processor. Code
15700 generated under this option will run on any POWER or PowerPC processor.
15701 GCC will use only the instructions in the common subset of both
15702 architectures, and will not use the MQ register. GCC assumes a generic
15703 processor model for scheduling purposes.
15705 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15706 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15707 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15708 types, with an appropriate, generic processor model assumed for
15709 scheduling purposes.
15711 The other options specify a specific processor. Code generated under
15712 those options will run best on that processor, and may not run at all on
15715 The @option{-mcpu} options automatically enable or disable the
15718 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15719 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15720 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15721 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15723 The particular options set for any particular CPU will vary between
15724 compiler versions, depending on what setting seems to produce optimal
15725 code for that CPU; it doesn't necessarily reflect the actual hardware's
15726 capabilities. If you wish to set an individual option to a particular
15727 value, you may specify it after the @option{-mcpu} option, like
15728 @samp{-mcpu=970 -mno-altivec}.
15730 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15731 not enabled or disabled by the @option{-mcpu} option at present because
15732 AIX does not have full support for these options. You may still
15733 enable or disable them individually if you're sure it'll work in your
15736 @item -mtune=@var{cpu_type}
15738 Set the instruction scheduling parameters for machine type
15739 @var{cpu_type}, but do not set the architecture type, register usage, or
15740 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15741 values for @var{cpu_type} are used for @option{-mtune} as for
15742 @option{-mcpu}. If both are specified, the code generated will use the
15743 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15744 scheduling parameters set by @option{-mtune}.
15746 @item -mcmodel=small
15747 @opindex mcmodel=small
15748 Generate PowerPC64 code for the small model: The TOC is limited to
15751 @item -mcmodel=medium
15752 @opindex mcmodel=medium
15753 Generate PowerPC64 code for the medium model: The TOC and other static
15754 data may be up to a total of 4G in size.
15756 @item -mcmodel=large
15757 @opindex mcmodel=large
15758 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15759 in size. Other data and code is only limited by the 64-bit address
15763 @itemx -mno-altivec
15765 @opindex mno-altivec
15766 Generate code that uses (does not use) AltiVec instructions, and also
15767 enable the use of built-in functions that allow more direct access to
15768 the AltiVec instruction set. You may also need to set
15769 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15775 @opindex mno-vrsave
15776 Generate VRSAVE instructions when generating AltiVec code.
15778 @item -mgen-cell-microcode
15779 @opindex mgen-cell-microcode
15780 Generate Cell microcode instructions
15782 @item -mwarn-cell-microcode
15783 @opindex mwarn-cell-microcode
15784 Warning when a Cell microcode instruction is going to emitted. An example
15785 of a Cell microcode instruction is a variable shift.
15788 @opindex msecure-plt
15789 Generate code that allows ld and ld.so to build executables and shared
15790 libraries with non-exec .plt and .got sections. This is a PowerPC
15791 32-bit SYSV ABI option.
15795 Generate code that uses a BSS .plt section that ld.so fills in, and
15796 requires .plt and .got sections that are both writable and executable.
15797 This is a PowerPC 32-bit SYSV ABI option.
15803 This switch enables or disables the generation of ISEL instructions.
15805 @item -misel=@var{yes/no}
15806 This switch has been deprecated. Use @option{-misel} and
15807 @option{-mno-isel} instead.
15813 This switch enables or disables the generation of SPE simd
15819 @opindex mno-paired
15820 This switch enables or disables the generation of PAIRED simd
15823 @item -mspe=@var{yes/no}
15824 This option has been deprecated. Use @option{-mspe} and
15825 @option{-mno-spe} instead.
15831 Generate code that uses (does not use) vector/scalar (VSX)
15832 instructions, and also enable the use of built-in functions that allow
15833 more direct access to the VSX instruction set.
15835 @item -mfloat-gprs=@var{yes/single/double/no}
15836 @itemx -mfloat-gprs
15837 @opindex mfloat-gprs
15838 This switch enables or disables the generation of floating point
15839 operations on the general purpose registers for architectures that
15842 The argument @var{yes} or @var{single} enables the use of
15843 single-precision floating point operations.
15845 The argument @var{double} enables the use of single and
15846 double-precision floating point operations.
15848 The argument @var{no} disables floating point operations on the
15849 general purpose registers.
15851 This option is currently only available on the MPC854x.
15857 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15858 targets (including GNU/Linux). The 32-bit environment sets int, long
15859 and pointer to 32 bits and generates code that runs on any PowerPC
15860 variant. The 64-bit environment sets int to 32 bits and long and
15861 pointer to 64 bits, and generates code for PowerPC64, as for
15862 @option{-mpowerpc64}.
15865 @itemx -mno-fp-in-toc
15866 @itemx -mno-sum-in-toc
15867 @itemx -mminimal-toc
15869 @opindex mno-fp-in-toc
15870 @opindex mno-sum-in-toc
15871 @opindex mminimal-toc
15872 Modify generation of the TOC (Table Of Contents), which is created for
15873 every executable file. The @option{-mfull-toc} option is selected by
15874 default. In that case, GCC will allocate at least one TOC entry for
15875 each unique non-automatic variable reference in your program. GCC
15876 will also place floating-point constants in the TOC@. However, only
15877 16,384 entries are available in the TOC@.
15879 If you receive a linker error message that saying you have overflowed
15880 the available TOC space, you can reduce the amount of TOC space used
15881 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15882 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15883 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15884 generate code to calculate the sum of an address and a constant at
15885 run-time instead of putting that sum into the TOC@. You may specify one
15886 or both of these options. Each causes GCC to produce very slightly
15887 slower and larger code at the expense of conserving TOC space.
15889 If you still run out of space in the TOC even when you specify both of
15890 these options, specify @option{-mminimal-toc} instead. This option causes
15891 GCC to make only one TOC entry for every file. When you specify this
15892 option, GCC will produce code that is slower and larger but which
15893 uses extremely little TOC space. You may wish to use this option
15894 only on files that contain less frequently executed code.
15900 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15901 @code{long} type, and the infrastructure needed to support them.
15902 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15903 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15904 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15907 @itemx -mno-xl-compat
15908 @opindex mxl-compat
15909 @opindex mno-xl-compat
15910 Produce code that conforms more closely to IBM XL compiler semantics
15911 when using AIX-compatible ABI@. Pass floating-point arguments to
15912 prototyped functions beyond the register save area (RSA) on the stack
15913 in addition to argument FPRs. Do not assume that most significant
15914 double in 128-bit long double value is properly rounded when comparing
15915 values and converting to double. Use XL symbol names for long double
15918 The AIX calling convention was extended but not initially documented to
15919 handle an obscure K&R C case of calling a function that takes the
15920 address of its arguments with fewer arguments than declared. IBM XL
15921 compilers access floating point arguments which do not fit in the
15922 RSA from the stack when a subroutine is compiled without
15923 optimization. Because always storing floating-point arguments on the
15924 stack is inefficient and rarely needed, this option is not enabled by
15925 default and only is necessary when calling subroutines compiled by IBM
15926 XL compilers without optimization.
15930 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15931 application written to use message passing with special startup code to
15932 enable the application to run. The system must have PE installed in the
15933 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15934 must be overridden with the @option{-specs=} option to specify the
15935 appropriate directory location. The Parallel Environment does not
15936 support threads, so the @option{-mpe} option and the @option{-pthread}
15937 option are incompatible.
15939 @item -malign-natural
15940 @itemx -malign-power
15941 @opindex malign-natural
15942 @opindex malign-power
15943 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15944 @option{-malign-natural} overrides the ABI-defined alignment of larger
15945 types, such as floating-point doubles, on their natural size-based boundary.
15946 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15947 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15949 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15953 @itemx -mhard-float
15954 @opindex msoft-float
15955 @opindex mhard-float
15956 Generate code that does not use (uses) the floating-point register set.
15957 Software floating point emulation is provided if you use the
15958 @option{-msoft-float} option, and pass the option to GCC when linking.
15960 @item -msingle-float
15961 @itemx -mdouble-float
15962 @opindex msingle-float
15963 @opindex mdouble-float
15964 Generate code for single or double-precision floating point operations.
15965 @option{-mdouble-float} implies @option{-msingle-float}.
15968 @opindex msimple-fpu
15969 Do not generate sqrt and div instructions for hardware floating point unit.
15973 Specify type of floating point unit. Valid values are @var{sp_lite}
15974 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15975 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15976 and @var{dp_full} (equivalent to -mdouble-float).
15979 @opindex mxilinx-fpu
15980 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15983 @itemx -mno-multiple
15985 @opindex mno-multiple
15986 Generate code that uses (does not use) the load multiple word
15987 instructions and the store multiple word instructions. These
15988 instructions are generated by default on POWER systems, and not
15989 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15990 endian PowerPC systems, since those instructions do not work when the
15991 processor is in little endian mode. The exceptions are PPC740 and
15992 PPC750 which permit the instructions usage in little endian mode.
15997 @opindex mno-string
15998 Generate code that uses (does not use) the load string instructions
15999 and the store string word instructions to save multiple registers and
16000 do small block moves. These instructions are generated by default on
16001 POWER systems, and not generated on PowerPC systems. Do not use
16002 @option{-mstring} on little endian PowerPC systems, since those
16003 instructions do not work when the processor is in little endian mode.
16004 The exceptions are PPC740 and PPC750 which permit the instructions
16005 usage in little endian mode.
16010 @opindex mno-update
16011 Generate code that uses (does not use) the load or store instructions
16012 that update the base register to the address of the calculated memory
16013 location. These instructions are generated by default. If you use
16014 @option{-mno-update}, there is a small window between the time that the
16015 stack pointer is updated and the address of the previous frame is
16016 stored, which means code that walks the stack frame across interrupts or
16017 signals may get corrupted data.
16019 @item -mavoid-indexed-addresses
16020 @itemx -mno-avoid-indexed-addresses
16021 @opindex mavoid-indexed-addresses
16022 @opindex mno-avoid-indexed-addresses
16023 Generate code that tries to avoid (not avoid) the use of indexed load
16024 or store instructions. These instructions can incur a performance
16025 penalty on Power6 processors in certain situations, such as when
16026 stepping through large arrays that cross a 16M boundary. This option
16027 is enabled by default when targetting Power6 and disabled otherwise.
16030 @itemx -mno-fused-madd
16031 @opindex mfused-madd
16032 @opindex mno-fused-madd
16033 Generate code that uses (does not use) the floating point multiply and
16034 accumulate instructions. These instructions are generated by default
16035 if hardware floating point is used. The machine dependent
16036 @option{-mfused-madd} option is now mapped to the machine independent
16037 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
16038 mapped to @option{-ffp-contract=off}.
16044 Generate code that uses (does not use) the half-word multiply and
16045 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
16046 These instructions are generated by default when targetting those
16053 Generate code that uses (does not use) the string-search @samp{dlmzb}
16054 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
16055 generated by default when targetting those processors.
16057 @item -mno-bit-align
16059 @opindex mno-bit-align
16060 @opindex mbit-align
16061 On System V.4 and embedded PowerPC systems do not (do) force structures
16062 and unions that contain bit-fields to be aligned to the base type of the
16065 For example, by default a structure containing nothing but 8
16066 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
16067 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
16068 the structure would be aligned to a 1 byte boundary and be one byte in
16071 @item -mno-strict-align
16072 @itemx -mstrict-align
16073 @opindex mno-strict-align
16074 @opindex mstrict-align
16075 On System V.4 and embedded PowerPC systems do not (do) assume that
16076 unaligned memory references will be handled by the system.
16078 @item -mrelocatable
16079 @itemx -mno-relocatable
16080 @opindex mrelocatable
16081 @opindex mno-relocatable
16082 Generate code that allows (does not allow) a static executable to be
16083 relocated to a different address at runtime. A simple embedded
16084 PowerPC system loader should relocate the entire contents of
16085 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16086 a table of 32-bit addresses generated by this option. For this to
16087 work, all objects linked together must be compiled with
16088 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16089 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
16091 @item -mrelocatable-lib
16092 @itemx -mno-relocatable-lib
16093 @opindex mrelocatable-lib
16094 @opindex mno-relocatable-lib
16095 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16096 @code{.fixup} section to allow static executables to be relocated at
16097 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
16098 alignment of @option{-mrelocatable}. Objects compiled with
16099 @option{-mrelocatable-lib} may be linked with objects compiled with
16100 any combination of the @option{-mrelocatable} options.
16106 On System V.4 and embedded PowerPC systems do not (do) assume that
16107 register 2 contains a pointer to a global area pointing to the addresses
16108 used in the program.
16111 @itemx -mlittle-endian
16113 @opindex mlittle-endian
16114 On System V.4 and embedded PowerPC systems compile code for the
16115 processor in little endian mode. The @option{-mlittle-endian} option is
16116 the same as @option{-mlittle}.
16119 @itemx -mbig-endian
16121 @opindex mbig-endian
16122 On System V.4 and embedded PowerPC systems compile code for the
16123 processor in big endian mode. The @option{-mbig-endian} option is
16124 the same as @option{-mbig}.
16126 @item -mdynamic-no-pic
16127 @opindex mdynamic-no-pic
16128 On Darwin and Mac OS X systems, compile code so that it is not
16129 relocatable, but that its external references are relocatable. The
16130 resulting code is suitable for applications, but not shared
16133 @item -msingle-pic-base
16134 @opindex msingle-pic-base
16135 Treat the register used for PIC addressing as read-only, rather than
16136 loading it in the prologue for each function. The run-time system is
16137 responsible for initializing this register with an appropriate value
16138 before execution begins.
16140 @item -mprioritize-restricted-insns=@var{priority}
16141 @opindex mprioritize-restricted-insns
16142 This option controls the priority that is assigned to
16143 dispatch-slot restricted instructions during the second scheduling
16144 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16145 @var{no/highest/second-highest} priority to dispatch slot restricted
16148 @item -msched-costly-dep=@var{dependence_type}
16149 @opindex msched-costly-dep
16150 This option controls which dependences are considered costly
16151 by the target during instruction scheduling. The argument
16152 @var{dependence_type} takes one of the following values:
16153 @var{no}: no dependence is costly,
16154 @var{all}: all dependences are costly,
16155 @var{true_store_to_load}: a true dependence from store to load is costly,
16156 @var{store_to_load}: any dependence from store to load is costly,
16157 @var{number}: any dependence which latency >= @var{number} is costly.
16159 @item -minsert-sched-nops=@var{scheme}
16160 @opindex minsert-sched-nops
16161 This option controls which nop insertion scheme will be used during
16162 the second scheduling pass. The argument @var{scheme} takes one of the
16164 @var{no}: Don't insert nops.
16165 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16166 according to the scheduler's grouping.
16167 @var{regroup_exact}: Insert nops to force costly dependent insns into
16168 separate groups. Insert exactly as many nops as needed to force an insn
16169 to a new group, according to the estimated processor grouping.
16170 @var{number}: Insert nops to force costly dependent insns into
16171 separate groups. Insert @var{number} nops to force an insn to a new group.
16174 @opindex mcall-sysv
16175 On System V.4 and embedded PowerPC systems compile code using calling
16176 conventions that adheres to the March 1995 draft of the System V
16177 Application Binary Interface, PowerPC processor supplement. This is the
16178 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16180 @item -mcall-sysv-eabi
16182 @opindex mcall-sysv-eabi
16183 @opindex mcall-eabi
16184 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16186 @item -mcall-sysv-noeabi
16187 @opindex mcall-sysv-noeabi
16188 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16190 @item -mcall-aixdesc
16192 On System V.4 and embedded PowerPC systems compile code for the AIX
16196 @opindex mcall-linux
16197 On System V.4 and embedded PowerPC systems compile code for the
16198 Linux-based GNU system.
16200 @item -mcall-freebsd
16201 @opindex mcall-freebsd
16202 On System V.4 and embedded PowerPC systems compile code for the
16203 FreeBSD operating system.
16205 @item -mcall-netbsd
16206 @opindex mcall-netbsd
16207 On System V.4 and embedded PowerPC systems compile code for the
16208 NetBSD operating system.
16210 @item -mcall-openbsd
16211 @opindex mcall-netbsd
16212 On System V.4 and embedded PowerPC systems compile code for the
16213 OpenBSD operating system.
16215 @item -maix-struct-return
16216 @opindex maix-struct-return
16217 Return all structures in memory (as specified by the AIX ABI)@.
16219 @item -msvr4-struct-return
16220 @opindex msvr4-struct-return
16221 Return structures smaller than 8 bytes in registers (as specified by the
16224 @item -mabi=@var{abi-type}
16226 Extend the current ABI with a particular extension, or remove such extension.
16227 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16228 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16232 Extend the current ABI with SPE ABI extensions. This does not change
16233 the default ABI, instead it adds the SPE ABI extensions to the current
16237 @opindex mabi=no-spe
16238 Disable Booke SPE ABI extensions for the current ABI@.
16240 @item -mabi=ibmlongdouble
16241 @opindex mabi=ibmlongdouble
16242 Change the current ABI to use IBM extended precision long double.
16243 This is a PowerPC 32-bit SYSV ABI option.
16245 @item -mabi=ieeelongdouble
16246 @opindex mabi=ieeelongdouble
16247 Change the current ABI to use IEEE extended precision long double.
16248 This is a PowerPC 32-bit Linux ABI option.
16251 @itemx -mno-prototype
16252 @opindex mprototype
16253 @opindex mno-prototype
16254 On System V.4 and embedded PowerPC systems assume that all calls to
16255 variable argument functions are properly prototyped. Otherwise, the
16256 compiler must insert an instruction before every non prototyped call to
16257 set or clear bit 6 of the condition code register (@var{CR}) to
16258 indicate whether floating point values were passed in the floating point
16259 registers in case the function takes a variable arguments. With
16260 @option{-mprototype}, only calls to prototyped variable argument functions
16261 will set or clear the bit.
16265 On embedded PowerPC systems, assume that the startup module is called
16266 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16267 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16272 On embedded PowerPC systems, assume that the startup module is called
16273 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16278 On embedded PowerPC systems, assume that the startup module is called
16279 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16282 @item -myellowknife
16283 @opindex myellowknife
16284 On embedded PowerPC systems, assume that the startup module is called
16285 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16290 On System V.4 and embedded PowerPC systems, specify that you are
16291 compiling for a VxWorks system.
16295 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16296 header to indicate that @samp{eabi} extended relocations are used.
16302 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16303 Embedded Applications Binary Interface (eabi) which is a set of
16304 modifications to the System V.4 specifications. Selecting @option{-meabi}
16305 means that the stack is aligned to an 8 byte boundary, a function
16306 @code{__eabi} is called to from @code{main} to set up the eabi
16307 environment, and the @option{-msdata} option can use both @code{r2} and
16308 @code{r13} to point to two separate small data areas. Selecting
16309 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16310 do not call an initialization function from @code{main}, and the
16311 @option{-msdata} option will only use @code{r13} to point to a single
16312 small data area. The @option{-meabi} option is on by default if you
16313 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16316 @opindex msdata=eabi
16317 On System V.4 and embedded PowerPC systems, put small initialized
16318 @code{const} global and static data in the @samp{.sdata2} section, which
16319 is pointed to by register @code{r2}. Put small initialized
16320 non-@code{const} global and static data in the @samp{.sdata} section,
16321 which is pointed to by register @code{r13}. Put small uninitialized
16322 global and static data in the @samp{.sbss} section, which is adjacent to
16323 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16324 incompatible with the @option{-mrelocatable} option. The
16325 @option{-msdata=eabi} option also sets the @option{-memb} option.
16328 @opindex msdata=sysv
16329 On System V.4 and embedded PowerPC systems, put small global and static
16330 data in the @samp{.sdata} section, which is pointed to by register
16331 @code{r13}. Put small uninitialized global and static data in the
16332 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16333 The @option{-msdata=sysv} option is incompatible with the
16334 @option{-mrelocatable} option.
16336 @item -msdata=default
16338 @opindex msdata=default
16340 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16341 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16342 same as @option{-msdata=sysv}.
16345 @opindex msdata=data
16346 On System V.4 and embedded PowerPC systems, put small global
16347 data in the @samp{.sdata} section. Put small uninitialized global
16348 data in the @samp{.sbss} section. Do not use register @code{r13}
16349 to address small data however. This is the default behavior unless
16350 other @option{-msdata} options are used.
16354 @opindex msdata=none
16356 On embedded PowerPC systems, put all initialized global and static data
16357 in the @samp{.data} section, and all uninitialized data in the
16358 @samp{.bss} section.
16360 @item -mblock-move-inline-limit=@var{num}
16361 @opindex mblock-move-inline-limit
16362 Inline all block moves (such as calls to @code{memcpy} or structure
16363 copies) less than or equal to @var{num} bytes. The minimum value for
16364 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16365 targets. The default value is target-specific.
16369 @cindex smaller data references (PowerPC)
16370 @cindex .sdata/.sdata2 references (PowerPC)
16371 On embedded PowerPC systems, put global and static items less than or
16372 equal to @var{num} bytes into the small data or bss sections instead of
16373 the normal data or bss section. By default, @var{num} is 8. The
16374 @option{-G @var{num}} switch is also passed to the linker.
16375 All modules should be compiled with the same @option{-G @var{num}} value.
16378 @itemx -mno-regnames
16380 @opindex mno-regnames
16381 On System V.4 and embedded PowerPC systems do (do not) emit register
16382 names in the assembly language output using symbolic forms.
16385 @itemx -mno-longcall
16387 @opindex mno-longcall
16388 By default assume that all calls are far away so that a longer more
16389 expensive calling sequence is required. This is required for calls
16390 further than 32 megabytes (33,554,432 bytes) from the current location.
16391 A short call will be generated if the compiler knows
16392 the call cannot be that far away. This setting can be overridden by
16393 the @code{shortcall} function attribute, or by @code{#pragma
16396 Some linkers are capable of detecting out-of-range calls and generating
16397 glue code on the fly. On these systems, long calls are unnecessary and
16398 generate slower code. As of this writing, the AIX linker can do this,
16399 as can the GNU linker for PowerPC/64. It is planned to add this feature
16400 to the GNU linker for 32-bit PowerPC systems as well.
16402 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16403 callee, L42'', plus a ``branch island'' (glue code). The two target
16404 addresses represent the callee and the ``branch island''. The
16405 Darwin/PPC linker will prefer the first address and generate a ``bl
16406 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16407 otherwise, the linker will generate ``bl L42'' to call the ``branch
16408 island''. The ``branch island'' is appended to the body of the
16409 calling function; it computes the full 32-bit address of the callee
16412 On Mach-O (Darwin) systems, this option directs the compiler emit to
16413 the glue for every direct call, and the Darwin linker decides whether
16414 to use or discard it.
16416 In the future, we may cause GCC to ignore all longcall specifications
16417 when the linker is known to generate glue.
16419 @item -mtls-markers
16420 @itemx -mno-tls-markers
16421 @opindex mtls-markers
16422 @opindex mno-tls-markers
16423 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16424 specifying the function argument. The relocation allows ld to
16425 reliably associate function call with argument setup instructions for
16426 TLS optimization, which in turn allows gcc to better schedule the
16431 Adds support for multithreading with the @dfn{pthreads} library.
16432 This option sets flags for both the preprocessor and linker.
16437 This option will enable GCC to use the reciprocal estimate and
16438 reciprocal square root estimate instructions with additional
16439 Newton-Raphson steps to increase precision instead of doing a divide or
16440 square root and divide for floating point arguments. You should use
16441 the @option{-ffast-math} option when using @option{-mrecip} (or at
16442 least @option{-funsafe-math-optimizations},
16443 @option{-finite-math-only}, @option{-freciprocal-math} and
16444 @option{-fno-trapping-math}). Note that while the throughput of the
16445 sequence is generally higher than the throughput of the non-reciprocal
16446 instruction, the precision of the sequence can be decreased by up to 2
16447 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16450 @item -mrecip=@var{opt}
16451 @opindex mrecip=opt
16452 This option allows to control which reciprocal estimate instructions
16453 may be used. @var{opt} is a comma separated list of options, that may
16454 be preceded by a @code{!} to invert the option:
16455 @code{all}: enable all estimate instructions,
16456 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16457 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16458 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16459 @code{divf}: enable the single precision reciprocal approximation instructions;
16460 @code{divd}: enable the double precision reciprocal approximation instructions;
16461 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16462 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16463 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16465 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16466 all of the reciprocal estimate instructions, except for the
16467 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16468 which handle the double precision reciprocal square root calculations.
16470 @item -mrecip-precision
16471 @itemx -mno-recip-precision
16472 @opindex mrecip-precision
16473 Assume (do not assume) that the reciprocal estimate instructions
16474 provide higher precision estimates than is mandated by the powerpc
16475 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16476 automatically selects @option{-mrecip-precision}. The double
16477 precision square root estimate instructions are not generated by
16478 default on low precision machines, since they do not provide an
16479 estimate that converges after three steps.
16481 @item -mveclibabi=@var{type}
16482 @opindex mveclibabi
16483 Specifies the ABI type to use for vectorizing intrinsics using an
16484 external library. The only type supported at present is @code{mass},
16485 which specifies to use IBM's Mathematical Acceleration Subsystem
16486 (MASS) libraries for vectorizing intrinsics using external libraries.
16487 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16488 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16489 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16490 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16491 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16492 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16493 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16494 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16495 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16496 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16497 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16498 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16499 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16500 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16501 for power7. Both @option{-ftree-vectorize} and
16502 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16503 libraries will have to be specified at link time.
16508 Generate (do not generate) the @code{friz} instruction when the
16509 @option{-funsafe-math-optimizations} option is used to optimize
16510 rounding a floating point value to 64-bit integer and back to floating
16511 point. The @code{friz} instruction does not return the same value if
16512 the floating point number is too large to fit in an integer.
16514 @item -mpointers-to-nested-functions
16515 @itemx -mno-pointers-to-nested-functions
16516 @opindex mpointers-to-nested-functions
16517 Generate (do not generate) code to load up the static chain register
16518 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16519 systems where a function pointer points to a 3 word descriptor giving
16520 the function address, TOC value to be loaded in register @var{r2}, and
16521 static chain value to be loaded in register @var{r11}. The
16522 @option{-mpointers-to-nested-functions} is on by default. You will
16523 not be able to call through pointers to nested functions or pointers
16524 to functions compiled in other languages that use the static chain if
16525 you use the @option{-mno-pointers-to-nested-functions}.
16527 @item -msave-toc-indirect
16528 @itemx -mno-save-toc-indirect
16529 @opindex msave-toc-indirect
16530 Generate (do not generate) code to save the TOC value in the reserved
16531 stack location in the function prologue if the function calls through
16532 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
16533 saved in the prologue, it is saved just before the call through the
16534 pointer. The @option{-mno-save-toc-indirect} option is the default.
16538 @subsection RX Options
16541 These command line options are defined for RX targets:
16544 @item -m64bit-doubles
16545 @itemx -m32bit-doubles
16546 @opindex m64bit-doubles
16547 @opindex m32bit-doubles
16548 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16549 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16550 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16551 works on 32-bit values, which is why the default is
16552 @option{-m32bit-doubles}.
16558 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16559 floating point hardware. The default is enabled for the @var{RX600}
16560 series and disabled for the @var{RX200} series.
16562 Floating point instructions will only be generated for 32-bit floating
16563 point values however, so if the @option{-m64bit-doubles} option is in
16564 use then the FPU hardware will not be used for doubles.
16566 @emph{Note} If the @option{-fpu} option is enabled then
16567 @option{-funsafe-math-optimizations} is also enabled automatically.
16568 This is because the RX FPU instructions are themselves unsafe.
16570 @item -mcpu=@var{name}
16572 Selects the type of RX CPU to be targeted. Currently three types are
16573 supported, the generic @var{RX600} and @var{RX200} series hardware and
16574 the specific @var{RX610} CPU. The default is @var{RX600}.
16576 The only difference between @var{RX600} and @var{RX610} is that the
16577 @var{RX610} does not support the @code{MVTIPL} instruction.
16579 The @var{RX200} series does not have a hardware floating point unit
16580 and so @option{-nofpu} is enabled by default when this type is
16583 @item -mbig-endian-data
16584 @itemx -mlittle-endian-data
16585 @opindex mbig-endian-data
16586 @opindex mlittle-endian-data
16587 Store data (but not code) in the big-endian format. The default is
16588 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16591 @item -msmall-data-limit=@var{N}
16592 @opindex msmall-data-limit
16593 Specifies the maximum size in bytes of global and static variables
16594 which can be placed into the small data area. Using the small data
16595 area can lead to smaller and faster code, but the size of area is
16596 limited and it is up to the programmer to ensure that the area does
16597 not overflow. Also when the small data area is used one of the RX's
16598 registers (usually @code{r13}) is reserved for use pointing to this
16599 area, so it is no longer available for use by the compiler. This
16600 could result in slower and/or larger code if variables which once
16601 could have been held in the reserved register are now pushed onto the
16604 Note, common variables (variables which have not been initialised) and
16605 constants are not placed into the small data area as they are assigned
16606 to other sections in the output executable.
16608 The default value is zero, which disables this feature. Note, this
16609 feature is not enabled by default with higher optimization levels
16610 (@option{-O2} etc) because of the potentially detrimental effects of
16611 reserving a register. It is up to the programmer to experiment and
16612 discover whether this feature is of benefit to their program. See the
16613 description of the @option{-mpid} option for a description of how the
16614 actual register to hold the small data area pointer is chosen.
16620 Use the simulator runtime. The default is to use the libgloss board
16623 @item -mas100-syntax
16624 @itemx -mno-as100-syntax
16625 @opindex mas100-syntax
16626 @opindex mno-as100-syntax
16627 When generating assembler output use a syntax that is compatible with
16628 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16629 assembler but it has some restrictions so generating it is not the
16632 @item -mmax-constant-size=@var{N}
16633 @opindex mmax-constant-size
16634 Specifies the maximum size, in bytes, of a constant that can be used as
16635 an operand in a RX instruction. Although the RX instruction set does
16636 allow constants of up to 4 bytes in length to be used in instructions,
16637 a longer value equates to a longer instruction. Thus in some
16638 circumstances it can be beneficial to restrict the size of constants
16639 that are used in instructions. Constants that are too big are instead
16640 placed into a constant pool and referenced via register indirection.
16642 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16643 or 4 means that constants of any size are allowed.
16647 Enable linker relaxation. Linker relaxation is a process whereby the
16648 linker will attempt to reduce the size of a program by finding shorter
16649 versions of various instructions. Disabled by default.
16651 @item -mint-register=@var{N}
16652 @opindex mint-register
16653 Specify the number of registers to reserve for fast interrupt handler
16654 functions. The value @var{N} can be between 0 and 4. A value of 1
16655 means that register @code{r13} will be reserved for the exclusive use
16656 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16657 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16658 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16659 A value of 0, the default, does not reserve any registers.
16661 @item -msave-acc-in-interrupts
16662 @opindex msave-acc-in-interrupts
16663 Specifies that interrupt handler functions should preserve the
16664 accumulator register. This is only necessary if normal code might use
16665 the accumulator register, for example because it performs 64-bit
16666 multiplications. The default is to ignore the accumulator as this
16667 makes the interrupt handlers faster.
16673 Enables the generation of position independent data. When enabled any
16674 access to constant data will done via an offset from a base address
16675 held in a register. This allows the location of constant data to be
16676 determined at run-time without requiring the executable to be
16677 relocated, which is a benefit to embedded applications with tight
16678 memory constraints. Data that can be modified is not affected by this
16681 Note, using this feature reserves a register, usually @code{r13}, for
16682 the constant data base address. This can result in slower and/or
16683 larger code, especially in complicated functions.
16685 The actual register chosen to hold the constant data base address
16686 depends upon whether the @option{-msmall-data-limit} and/or the
16687 @option{-mint-register} command line options are enabled. Starting
16688 with register @code{r13} and proceeding downwards, registers are
16689 allocated first to satisfy the requirements of @option{-mint-register},
16690 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
16691 is possible for the small data area register to be @code{r8} if both
16692 @option{-mint-register=4} and @option{-mpid} are specified on the
16695 By default this feature is not enabled. The default can be restored
16696 via the @option{-mno-pid} command line option.
16700 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16701 has special significance to the RX port when used with the
16702 @code{interrupt} function attribute. This attribute indicates a
16703 function intended to process fast interrupts. GCC will will ensure
16704 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16705 and/or @code{r13} and only provided that the normal use of the
16706 corresponding registers have been restricted via the
16707 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16710 @node S/390 and zSeries Options
16711 @subsection S/390 and zSeries Options
16712 @cindex S/390 and zSeries Options
16714 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16718 @itemx -msoft-float
16719 @opindex mhard-float
16720 @opindex msoft-float
16721 Use (do not use) the hardware floating-point instructions and registers
16722 for floating-point operations. When @option{-msoft-float} is specified,
16723 functions in @file{libgcc.a} will be used to perform floating-point
16724 operations. When @option{-mhard-float} is specified, the compiler
16725 generates IEEE floating-point instructions. This is the default.
16728 @itemx -mno-hard-dfp
16730 @opindex mno-hard-dfp
16731 Use (do not use) the hardware decimal-floating-point instructions for
16732 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16733 specified, functions in @file{libgcc.a} will be used to perform
16734 decimal-floating-point operations. When @option{-mhard-dfp} is
16735 specified, the compiler generates decimal-floating-point hardware
16736 instructions. This is the default for @option{-march=z9-ec} or higher.
16738 @item -mlong-double-64
16739 @itemx -mlong-double-128
16740 @opindex mlong-double-64
16741 @opindex mlong-double-128
16742 These switches control the size of @code{long double} type. A size
16743 of 64bit makes the @code{long double} type equivalent to the @code{double}
16744 type. This is the default.
16747 @itemx -mno-backchain
16748 @opindex mbackchain
16749 @opindex mno-backchain
16750 Store (do not store) the address of the caller's frame as backchain pointer
16751 into the callee's stack frame.
16752 A backchain may be needed to allow debugging using tools that do not understand
16753 DWARF-2 call frame information.
16754 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16755 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16756 the backchain is placed into the topmost word of the 96/160 byte register
16759 In general, code compiled with @option{-mbackchain} is call-compatible with
16760 code compiled with @option{-mmo-backchain}; however, use of the backchain
16761 for debugging purposes usually requires that the whole binary is built with
16762 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16763 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16764 to build a linux kernel use @option{-msoft-float}.
16766 The default is to not maintain the backchain.
16768 @item -mpacked-stack
16769 @itemx -mno-packed-stack
16770 @opindex mpacked-stack
16771 @opindex mno-packed-stack
16772 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16773 specified, the compiler uses the all fields of the 96/160 byte register save
16774 area only for their default purpose; unused fields still take up stack space.
16775 When @option{-mpacked-stack} is specified, register save slots are densely
16776 packed at the top of the register save area; unused space is reused for other
16777 purposes, allowing for more efficient use of the available stack space.
16778 However, when @option{-mbackchain} is also in effect, the topmost word of
16779 the save area is always used to store the backchain, and the return address
16780 register is always saved two words below the backchain.
16782 As long as the stack frame backchain is not used, code generated with
16783 @option{-mpacked-stack} is call-compatible with code generated with
16784 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16785 S/390 or zSeries generated code that uses the stack frame backchain at run
16786 time, not just for debugging purposes. Such code is not call-compatible
16787 with code compiled with @option{-mpacked-stack}. Also, note that the
16788 combination of @option{-mbackchain},
16789 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16790 to build a linux kernel use @option{-msoft-float}.
16792 The default is to not use the packed stack layout.
16795 @itemx -mno-small-exec
16796 @opindex msmall-exec
16797 @opindex mno-small-exec
16798 Generate (or do not generate) code using the @code{bras} instruction
16799 to do subroutine calls.
16800 This only works reliably if the total executable size does not
16801 exceed 64k. The default is to use the @code{basr} instruction instead,
16802 which does not have this limitation.
16808 When @option{-m31} is specified, generate code compliant to the
16809 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16810 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16811 particular to generate 64-bit instructions. For the @samp{s390}
16812 targets, the default is @option{-m31}, while the @samp{s390x}
16813 targets default to @option{-m64}.
16819 When @option{-mzarch} is specified, generate code using the
16820 instructions available on z/Architecture.
16821 When @option{-mesa} is specified, generate code using the
16822 instructions available on ESA/390. Note that @option{-mesa} is
16823 not possible with @option{-m64}.
16824 When generating code compliant to the GNU/Linux for S/390 ABI,
16825 the default is @option{-mesa}. When generating code compliant
16826 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16832 Generate (or do not generate) code using the @code{mvcle} instruction
16833 to perform block moves. When @option{-mno-mvcle} is specified,
16834 use a @code{mvc} loop instead. This is the default unless optimizing for
16841 Print (or do not print) additional debug information when compiling.
16842 The default is to not print debug information.
16844 @item -march=@var{cpu-type}
16846 Generate code that will run on @var{cpu-type}, which is the name of a system
16847 representing a certain processor type. Possible values for
16848 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16849 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16850 When generating code using the instructions available on z/Architecture,
16851 the default is @option{-march=z900}. Otherwise, the default is
16852 @option{-march=g5}.
16854 @item -mtune=@var{cpu-type}
16856 Tune to @var{cpu-type} everything applicable about the generated code,
16857 except for the ABI and the set of available instructions.
16858 The list of @var{cpu-type} values is the same as for @option{-march}.
16859 The default is the value used for @option{-march}.
16862 @itemx -mno-tpf-trace
16863 @opindex mtpf-trace
16864 @opindex mno-tpf-trace
16865 Generate code that adds (does not add) in TPF OS specific branches to trace
16866 routines in the operating system. This option is off by default, even
16867 when compiling for the TPF OS@.
16870 @itemx -mno-fused-madd
16871 @opindex mfused-madd
16872 @opindex mno-fused-madd
16873 Generate code that uses (does not use) the floating point multiply and
16874 accumulate instructions. These instructions are generated by default if
16875 hardware floating point is used.
16877 @item -mwarn-framesize=@var{framesize}
16878 @opindex mwarn-framesize
16879 Emit a warning if the current function exceeds the given frame size. Because
16880 this is a compile time check it doesn't need to be a real problem when the program
16881 runs. It is intended to identify functions which most probably cause
16882 a stack overflow. It is useful to be used in an environment with limited stack
16883 size e.g.@: the linux kernel.
16885 @item -mwarn-dynamicstack
16886 @opindex mwarn-dynamicstack
16887 Emit a warning if the function calls alloca or uses dynamically
16888 sized arrays. This is generally a bad idea with a limited stack size.
16890 @item -mstack-guard=@var{stack-guard}
16891 @itemx -mstack-size=@var{stack-size}
16892 @opindex mstack-guard
16893 @opindex mstack-size
16894 If these options are provided the s390 back end emits additional instructions in
16895 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16896 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16897 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16898 the frame size of the compiled function is chosen.
16899 These options are intended to be used to help debugging stack overflow problems.
16900 The additionally emitted code causes only little overhead and hence can also be
16901 used in production like systems without greater performance degradation. The given
16902 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16903 @var{stack-guard} without exceeding 64k.
16904 In order to be efficient the extra code makes the assumption that the stack starts
16905 at an address aligned to the value given by @var{stack-size}.
16906 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16909 @node Score Options
16910 @subsection Score Options
16911 @cindex Score Options
16913 These options are defined for Score implementations:
16918 Compile code for big endian mode. This is the default.
16922 Compile code for little endian mode.
16926 Disable generate bcnz instruction.
16930 Enable generate unaligned load and store instruction.
16934 Enable the use of multiply-accumulate instructions. Disabled by default.
16938 Specify the SCORE5 as the target architecture.
16942 Specify the SCORE5U of the target architecture.
16946 Specify the SCORE7 as the target architecture. This is the default.
16950 Specify the SCORE7D as the target architecture.
16954 @subsection SH Options
16956 These @samp{-m} options are defined for the SH implementations:
16961 Generate code for the SH1.
16965 Generate code for the SH2.
16968 Generate code for the SH2e.
16972 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16973 that the floating-point unit is not used.
16975 @item -m2a-single-only
16976 @opindex m2a-single-only
16977 Generate code for the SH2a-FPU, in such a way that no double-precision
16978 floating point operations are used.
16981 @opindex m2a-single
16982 Generate code for the SH2a-FPU assuming the floating-point unit is in
16983 single-precision mode by default.
16987 Generate code for the SH2a-FPU assuming the floating-point unit is in
16988 double-precision mode by default.
16992 Generate code for the SH3.
16996 Generate code for the SH3e.
17000 Generate code for the SH4 without a floating-point unit.
17002 @item -m4-single-only
17003 @opindex m4-single-only
17004 Generate code for the SH4 with a floating-point unit that only
17005 supports single-precision arithmetic.
17009 Generate code for the SH4 assuming the floating-point unit is in
17010 single-precision mode by default.
17014 Generate code for the SH4.
17018 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
17019 floating-point unit is not used.
17021 @item -m4a-single-only
17022 @opindex m4a-single-only
17023 Generate code for the SH4a, in such a way that no double-precision
17024 floating point operations are used.
17027 @opindex m4a-single
17028 Generate code for the SH4a assuming the floating-point unit is in
17029 single-precision mode by default.
17033 Generate code for the SH4a.
17037 Same as @option{-m4a-nofpu}, except that it implicitly passes
17038 @option{-dsp} to the assembler. GCC doesn't generate any DSP
17039 instructions at the moment.
17043 Compile code for the processor in big endian mode.
17047 Compile code for the processor in little endian mode.
17051 Align doubles at 64-bit boundaries. Note that this changes the calling
17052 conventions, and thus some functions from the standard C library will
17053 not work unless you recompile it first with @option{-mdalign}.
17057 Shorten some address references at link time, when possible; uses the
17058 linker option @option{-relax}.
17062 Use 32-bit offsets in @code{switch} tables. The default is to use
17067 Enable the use of bit manipulation instructions on SH2A.
17071 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
17072 alignment constraints.
17076 Comply with the calling conventions defined by Renesas.
17080 Comply with the calling conventions defined by Renesas.
17084 Comply with the calling conventions defined for GCC before the Renesas
17085 conventions were available. This option is the default for all
17086 targets of the SH toolchain.
17089 @opindex mnomacsave
17090 Mark the @code{MAC} register as call-clobbered, even if
17091 @option{-mhitachi} is given.
17095 Increase IEEE-compliance of floating-point code.
17096 At the moment, this is equivalent to @option{-fno-finite-math-only}.
17097 When generating 16 bit SH opcodes, getting IEEE-conforming results for
17098 comparisons of NANs / infinities incurs extra overhead in every
17099 floating point comparison, therefore the default is set to
17100 @option{-ffinite-math-only}.
17102 @item -minline-ic_invalidate
17103 @opindex minline-ic_invalidate
17104 Inline code to invalidate instruction cache entries after setting up
17105 nested function trampolines.
17106 This option has no effect if -musermode is in effect and the selected
17107 code generation option (e.g. -m4) does not allow the use of the icbi
17109 If the selected code generation option does not allow the use of the icbi
17110 instruction, and -musermode is not in effect, the inlined code will
17111 manipulate the instruction cache address array directly with an associative
17112 write. This not only requires privileged mode, but it will also
17113 fail if the cache line had been mapped via the TLB and has become unmapped.
17117 Dump instruction size and location in the assembly code.
17120 @opindex mpadstruct
17121 This option is deprecated. It pads structures to multiple of 4 bytes,
17122 which is incompatible with the SH ABI@.
17126 Optimize for space instead of speed. Implied by @option{-Os}.
17129 @opindex mprefergot
17130 When generating position-independent code, emit function calls using
17131 the Global Offset Table instead of the Procedure Linkage Table.
17135 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17136 if the inlined code would not work in user mode.
17137 This is the default when the target is @code{sh-*-linux*}.
17139 @item -multcost=@var{number}
17140 @opindex multcost=@var{number}
17141 Set the cost to assume for a multiply insn.
17143 @item -mdiv=@var{strategy}
17144 @opindex mdiv=@var{strategy}
17145 Set the division strategy to use for SHmedia code. @var{strategy} must be
17146 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17147 inv:call2, inv:fp .
17148 "fp" performs the operation in floating point. This has a very high latency,
17149 but needs only a few instructions, so it might be a good choice if
17150 your code has enough easily exploitable ILP to allow the compiler to
17151 schedule the floating point instructions together with other instructions.
17152 Division by zero causes a floating point exception.
17153 "inv" uses integer operations to calculate the inverse of the divisor,
17154 and then multiplies the dividend with the inverse. This strategy allows
17155 cse and hoisting of the inverse calculation. Division by zero calculates
17156 an unspecified result, but does not trap.
17157 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17158 have been found, or if the entire operation has been hoisted to the same
17159 place, the last stages of the inverse calculation are intertwined with the
17160 final multiply to reduce the overall latency, at the expense of using a few
17161 more instructions, and thus offering fewer scheduling opportunities with
17163 "call" calls a library function that usually implements the inv:minlat
17165 This gives high code density for m5-*media-nofpu compilations.
17166 "call2" uses a different entry point of the same library function, where it
17167 assumes that a pointer to a lookup table has already been set up, which
17168 exposes the pointer load to cse / code hoisting optimizations.
17169 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17170 code generation, but if the code stays unoptimized, revert to the "call",
17171 "call2", or "fp" strategies, respectively. Note that the
17172 potentially-trapping side effect of division by zero is carried by a
17173 separate instruction, so it is possible that all the integer instructions
17174 are hoisted out, but the marker for the side effect stays where it is.
17175 A recombination to fp operations or a call is not possible in that case.
17176 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17177 that the inverse calculation was nor separated from the multiply, they speed
17178 up division where the dividend fits into 20 bits (plus sign where applicable),
17179 by inserting a test to skip a number of operations in this case; this test
17180 slows down the case of larger dividends. inv20u assumes the case of a such
17181 a small dividend to be unlikely, and inv20l assumes it to be likely.
17183 @item -maccumulate-outgoing-args
17184 @opindex maccumulate-outgoing-args
17185 Reserve space once for outgoing arguments in the function prologue rather
17186 than around each call. Generally beneficial for performance and size. Also
17187 needed for unwinding to avoid changing the stack frame around conditional code.
17189 @item -mdivsi3_libfunc=@var{name}
17190 @opindex mdivsi3_libfunc=@var{name}
17191 Set the name of the library function used for 32 bit signed division to
17192 @var{name}. This only affect the name used in the call and inv:call
17193 division strategies, and the compiler will still expect the same
17194 sets of input/output/clobbered registers as if this option was not present.
17196 @item -mfixed-range=@var{register-range}
17197 @opindex mfixed-range
17198 Generate code treating the given register range as fixed registers.
17199 A fixed register is one that the register allocator can not use. This is
17200 useful when compiling kernel code. A register range is specified as
17201 two registers separated by a dash. Multiple register ranges can be
17202 specified separated by a comma.
17204 @item -madjust-unroll
17205 @opindex madjust-unroll
17206 Throttle unrolling to avoid thrashing target registers.
17207 This option only has an effect if the gcc code base supports the
17208 TARGET_ADJUST_UNROLL_MAX target hook.
17210 @item -mindexed-addressing
17211 @opindex mindexed-addressing
17212 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17213 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17214 semantics for the indexed addressing mode. The architecture allows the
17215 implementation of processors with 64 bit MMU, which the OS could use to
17216 get 32 bit addressing, but since no current hardware implementation supports
17217 this or any other way to make the indexed addressing mode safe to use in
17218 the 32 bit ABI, the default is -mno-indexed-addressing.
17220 @item -mgettrcost=@var{number}
17221 @opindex mgettrcost=@var{number}
17222 Set the cost assumed for the gettr instruction to @var{number}.
17223 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17227 Assume pt* instructions won't trap. This will generally generate better
17228 scheduled code, but is unsafe on current hardware. The current architecture
17229 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17230 This has the unintentional effect of making it unsafe to schedule ptabs /
17231 ptrel before a branch, or hoist it out of a loop. For example,
17232 __do_global_ctors, a part of libgcc that runs constructors at program
17233 startup, calls functions in a list which is delimited by @minus{}1. With the
17234 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17235 That means that all the constructors will be run a bit quicker, but when
17236 the loop comes to the end of the list, the program crashes because ptabs
17237 loads @minus{}1 into a target register. Since this option is unsafe for any
17238 hardware implementing the current architecture specification, the default
17239 is -mno-pt-fixed. Unless the user specifies a specific cost with
17240 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17241 this deters register allocation using target registers for storing
17244 @item -minvalid-symbols
17245 @opindex minvalid-symbols
17246 Assume symbols might be invalid. Ordinary function symbols generated by
17247 the compiler will always be valid to load with movi/shori/ptabs or
17248 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17249 to generate symbols that will cause ptabs / ptrel to trap.
17250 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17251 It will then prevent cross-basic-block cse, hoisting and most scheduling
17252 of symbol loads. The default is @option{-mno-invalid-symbols}.
17255 @node Solaris 2 Options
17256 @subsection Solaris 2 Options
17257 @cindex Solaris 2 options
17259 These @samp{-m} options are supported on Solaris 2:
17262 @item -mimpure-text
17263 @opindex mimpure-text
17264 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17265 the compiler to not pass @option{-z text} to the linker when linking a
17266 shared object. Using this option, you can link position-dependent
17267 code into a shared object.
17269 @option{-mimpure-text} suppresses the ``relocations remain against
17270 allocatable but non-writable sections'' linker error message.
17271 However, the necessary relocations will trigger copy-on-write, and the
17272 shared object is not actually shared across processes. Instead of
17273 using @option{-mimpure-text}, you should compile all source code with
17274 @option{-fpic} or @option{-fPIC}.
17278 These switches are supported in addition to the above on Solaris 2:
17283 Add support for multithreading using the POSIX threads library. This
17284 option sets flags for both the preprocessor and linker. This option does
17285 not affect the thread safety of object code produced by the compiler or
17286 that of libraries supplied with it.
17290 This is a synonym for @option{-pthreads}.
17293 @node SPARC Options
17294 @subsection SPARC Options
17295 @cindex SPARC options
17297 These @samp{-m} options are supported on the SPARC:
17300 @item -mno-app-regs
17302 @opindex mno-app-regs
17304 Specify @option{-mapp-regs} to generate output using the global registers
17305 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17308 To be fully SVR4 ABI compliant at the cost of some performance loss,
17309 specify @option{-mno-app-regs}. You should compile libraries and system
17310 software with this option.
17316 With @option{-mflat}, the compiler does not generate save/restore instructions
17317 and uses a ``flat'' or single register window model. This model is compatible
17318 with the regular register window model. The local registers and the input
17319 registers (0--5) are still treated as ``call-saved'' registers and will be
17320 saved on the stack as needed.
17322 With @option{-mno-flat} (the default), the compiler generates save/restore
17323 instructions (except for leaf functions). This is the normal operating mode.
17326 @itemx -mhard-float
17328 @opindex mhard-float
17329 Generate output containing floating point instructions. This is the
17333 @itemx -msoft-float
17335 @opindex msoft-float
17336 Generate output containing library calls for floating point.
17337 @strong{Warning:} the requisite libraries are not available for all SPARC
17338 targets. Normally the facilities of the machine's usual C compiler are
17339 used, but this cannot be done directly in cross-compilation. You must make
17340 your own arrangements to provide suitable library functions for
17341 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17342 @samp{sparclite-*-*} do provide software floating point support.
17344 @option{-msoft-float} changes the calling convention in the output file;
17345 therefore, it is only useful if you compile @emph{all} of a program with
17346 this option. In particular, you need to compile @file{libgcc.a}, the
17347 library that comes with GCC, with @option{-msoft-float} in order for
17350 @item -mhard-quad-float
17351 @opindex mhard-quad-float
17352 Generate output containing quad-word (long double) floating point
17355 @item -msoft-quad-float
17356 @opindex msoft-quad-float
17357 Generate output containing library calls for quad-word (long double)
17358 floating point instructions. The functions called are those specified
17359 in the SPARC ABI@. This is the default.
17361 As of this writing, there are no SPARC implementations that have hardware
17362 support for the quad-word floating point instructions. They all invoke
17363 a trap handler for one of these instructions, and then the trap handler
17364 emulates the effect of the instruction. Because of the trap handler overhead,
17365 this is much slower than calling the ABI library routines. Thus the
17366 @option{-msoft-quad-float} option is the default.
17368 @item -mno-unaligned-doubles
17369 @itemx -munaligned-doubles
17370 @opindex mno-unaligned-doubles
17371 @opindex munaligned-doubles
17372 Assume that doubles have 8 byte alignment. This is the default.
17374 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17375 alignment only if they are contained in another type, or if they have an
17376 absolute address. Otherwise, it assumes they have 4 byte alignment.
17377 Specifying this option avoids some rare compatibility problems with code
17378 generated by other compilers. It is not the default because it results
17379 in a performance loss, especially for floating point code.
17381 @item -mno-faster-structs
17382 @itemx -mfaster-structs
17383 @opindex mno-faster-structs
17384 @opindex mfaster-structs
17385 With @option{-mfaster-structs}, the compiler assumes that structures
17386 should have 8 byte alignment. This enables the use of pairs of
17387 @code{ldd} and @code{std} instructions for copies in structure
17388 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17389 However, the use of this changed alignment directly violates the SPARC
17390 ABI@. Thus, it's intended only for use on targets where the developer
17391 acknowledges that their resulting code will not be directly in line with
17392 the rules of the ABI@.
17394 @item -mcpu=@var{cpu_type}
17396 Set the instruction set, register set, and instruction scheduling parameters
17397 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17398 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17399 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17400 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17401 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
17402 and @samp{niagara4}.
17404 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
17405 which selects the best architecture option for the host processor.
17406 @option{-mcpu=native} has no effect if GCC does not recognize
17409 Default instruction scheduling parameters are used for values that select
17410 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17411 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17413 Here is a list of each supported architecture and their supported
17418 v8: supersparc, hypersparc, leon
17419 sparclite: f930, f934, sparclite86x
17421 v9: ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
17424 By default (unless configured otherwise), GCC generates code for the V7
17425 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17426 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17427 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17428 SPARCStation 1, 2, IPX etc.
17430 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17431 architecture. The only difference from V7 code is that the compiler emits
17432 the integer multiply and integer divide instructions which exist in SPARC-V8
17433 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17434 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17437 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17438 the SPARC architecture. This adds the integer multiply, integer divide step
17439 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17440 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17441 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17442 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17443 MB86934 chip, which is the more recent SPARClite with FPU@.
17445 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17446 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17447 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17448 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17449 optimizes it for the TEMIC SPARClet chip.
17451 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17452 architecture. This adds 64-bit integer and floating-point move instructions,
17453 3 additional floating-point condition code registers and conditional move
17454 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17455 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17456 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17457 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17458 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17459 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17460 additionally optimizes it for Sun UltraSPARC T2 chips. With
17461 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
17462 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
17463 additionally optimizes it for Sun UltraSPARC T4 chips.
17465 @item -mtune=@var{cpu_type}
17467 Set the instruction scheduling parameters for machine type
17468 @var{cpu_type}, but do not set the instruction set or register set that the
17469 option @option{-mcpu=@var{cpu_type}} would.
17471 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17472 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17473 that select a particular CPU implementation. Those are @samp{cypress},
17474 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17475 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17476 @samp{niagara}, @samp{niagara2}, @samp{niagara3} and @samp{niagara4}. With
17477 native Solaris and GNU/Linux toolchains, @samp{native} can also be used.
17482 @opindex mno-v8plus
17483 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17484 difference from the V8 ABI is that the global and out registers are
17485 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17486 mode for all SPARC-V9 processors.
17492 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17493 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17499 With @option{-mvis2}, GCC generates code that takes advantage of
17500 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
17501 default is @option{-mvis2} when targetting a cpu that supports such
17502 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
17503 also sets @option{-mvis}.
17509 With @option{-mvis3}, GCC generates code that takes advantage of
17510 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
17511 default is @option{-mvis3} when targetting a cpu that supports such
17512 instructions, such as niagara-3 and later. Setting @option{-mvis3}
17513 also sets @option{-mvis2} and @option{-mvis}.
17519 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
17520 population count instruction. The default is @option{-mpopc}
17521 when targetting a cpu that supports such instructions, such as Niagara-2 and
17528 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
17529 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
17530 when targetting a cpu that supports such instructions, such as Niagara-3 and
17534 These @samp{-m} options are supported in addition to the above
17535 on SPARC-V9 processors in 64-bit environments:
17538 @item -mlittle-endian
17539 @opindex mlittle-endian
17540 Generate code for a processor running in little-endian mode. It is only
17541 available for a few configurations and most notably not on Solaris and Linux.
17547 Generate code for a 32-bit or 64-bit environment.
17548 The 32-bit environment sets int, long and pointer to 32 bits.
17549 The 64-bit environment sets int to 32 bits and long and pointer
17552 @item -mcmodel=medlow
17553 @opindex mcmodel=medlow
17554 Generate code for the Medium/Low code model: 64-bit addresses, programs
17555 must be linked in the low 32 bits of memory. Programs can be statically
17556 or dynamically linked.
17558 @item -mcmodel=medmid
17559 @opindex mcmodel=medmid
17560 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17561 must be linked in the low 44 bits of memory, the text and data segments must
17562 be less than 2GB in size and the data segment must be located within 2GB of
17565 @item -mcmodel=medany
17566 @opindex mcmodel=medany
17567 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17568 may be linked anywhere in memory, the text and data segments must be less
17569 than 2GB in size and the data segment must be located within 2GB of the
17572 @item -mcmodel=embmedany
17573 @opindex mcmodel=embmedany
17574 Generate code for the Medium/Anywhere code model for embedded systems:
17575 64-bit addresses, the text and data segments must be less than 2GB in
17576 size, both starting anywhere in memory (determined at link time). The
17577 global register %g4 points to the base of the data segment. Programs
17578 are statically linked and PIC is not supported.
17581 @itemx -mno-stack-bias
17582 @opindex mstack-bias
17583 @opindex mno-stack-bias
17584 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17585 frame pointer if present, are offset by @minus{}2047 which must be added back
17586 when making stack frame references. This is the default in 64-bit mode.
17587 Otherwise, assume no such offset is present.
17591 @subsection SPU Options
17592 @cindex SPU options
17594 These @samp{-m} options are supported on the SPU:
17598 @itemx -merror-reloc
17599 @opindex mwarn-reloc
17600 @opindex merror-reloc
17602 The loader for SPU does not handle dynamic relocations. By default, GCC
17603 will give an error when it generates code that requires a dynamic
17604 relocation. @option{-mno-error-reloc} disables the error,
17605 @option{-mwarn-reloc} will generate a warning instead.
17608 @itemx -munsafe-dma
17610 @opindex munsafe-dma
17612 Instructions which initiate or test completion of DMA must not be
17613 reordered with respect to loads and stores of the memory which is being
17614 accessed. Users typically address this problem using the volatile
17615 keyword, but that can lead to inefficient code in places where the
17616 memory is known to not change. Rather than mark the memory as volatile
17617 we treat the DMA instructions as potentially effecting all memory. With
17618 @option{-munsafe-dma} users must use the volatile keyword to protect
17621 @item -mbranch-hints
17622 @opindex mbranch-hints
17624 By default, GCC will generate a branch hint instruction to avoid
17625 pipeline stalls for always taken or probably taken branches. A hint
17626 will not be generated closer than 8 instructions away from its branch.
17627 There is little reason to disable them, except for debugging purposes,
17628 or to make an object a little bit smaller.
17632 @opindex msmall-mem
17633 @opindex mlarge-mem
17635 By default, GCC generates code assuming that addresses are never larger
17636 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17637 a full 32 bit address.
17642 By default, GCC links against startup code that assumes the SPU-style
17643 main function interface (which has an unconventional parameter list).
17644 With @option{-mstdmain}, GCC will link your program against startup
17645 code that assumes a C99-style interface to @code{main}, including a
17646 local copy of @code{argv} strings.
17648 @item -mfixed-range=@var{register-range}
17649 @opindex mfixed-range
17650 Generate code treating the given register range as fixed registers.
17651 A fixed register is one that the register allocator can not use. This is
17652 useful when compiling kernel code. A register range is specified as
17653 two registers separated by a dash. Multiple register ranges can be
17654 specified separated by a comma.
17660 Compile code assuming that pointers to the PPU address space accessed
17661 via the @code{__ea} named address space qualifier are either 32 or 64
17662 bits wide. The default is 32 bits. As this is an ABI changing option,
17663 all object code in an executable must be compiled with the same setting.
17665 @item -maddress-space-conversion
17666 @itemx -mno-address-space-conversion
17667 @opindex maddress-space-conversion
17668 @opindex mno-address-space-conversion
17669 Allow/disallow treating the @code{__ea} address space as superset
17670 of the generic address space. This enables explicit type casts
17671 between @code{__ea} and generic pointer as well as implicit
17672 conversions of generic pointers to @code{__ea} pointers. The
17673 default is to allow address space pointer conversions.
17675 @item -mcache-size=@var{cache-size}
17676 @opindex mcache-size
17677 This option controls the version of libgcc that the compiler links to an
17678 executable and selects a software-managed cache for accessing variables
17679 in the @code{__ea} address space with a particular cache size. Possible
17680 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17681 and @samp{128}. The default cache size is 64KB.
17683 @item -matomic-updates
17684 @itemx -mno-atomic-updates
17685 @opindex matomic-updates
17686 @opindex mno-atomic-updates
17687 This option controls the version of libgcc that the compiler links to an
17688 executable and selects whether atomic updates to the software-managed
17689 cache of PPU-side variables are used. If you use atomic updates, changes
17690 to a PPU variable from SPU code using the @code{__ea} named address space
17691 qualifier will not interfere with changes to other PPU variables residing
17692 in the same cache line from PPU code. If you do not use atomic updates,
17693 such interference may occur; however, writing back cache lines will be
17694 more efficient. The default behavior is to use atomic updates.
17697 @itemx -mdual-nops=@var{n}
17698 @opindex mdual-nops
17699 By default, GCC will insert nops to increase dual issue when it expects
17700 it to increase performance. @var{n} can be a value from 0 to 10. A
17701 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17702 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17704 @item -mhint-max-nops=@var{n}
17705 @opindex mhint-max-nops
17706 Maximum number of nops to insert for a branch hint. A branch hint must
17707 be at least 8 instructions away from the branch it is effecting. GCC
17708 will insert up to @var{n} nops to enforce this, otherwise it will not
17709 generate the branch hint.
17711 @item -mhint-max-distance=@var{n}
17712 @opindex mhint-max-distance
17713 The encoding of the branch hint instruction limits the hint to be within
17714 256 instructions of the branch it is effecting. By default, GCC makes
17715 sure it is within 125.
17718 @opindex msafe-hints
17719 Work around a hardware bug which causes the SPU to stall indefinitely.
17720 By default, GCC will insert the @code{hbrp} instruction to make sure
17721 this stall won't happen.
17725 @node System V Options
17726 @subsection Options for System V
17728 These additional options are available on System V Release 4 for
17729 compatibility with other compilers on those systems:
17734 Create a shared object.
17735 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17739 Identify the versions of each tool used by the compiler, in a
17740 @code{.ident} assembler directive in the output.
17744 Refrain from adding @code{.ident} directives to the output file (this is
17747 @item -YP,@var{dirs}
17749 Search the directories @var{dirs}, and no others, for libraries
17750 specified with @option{-l}.
17752 @item -Ym,@var{dir}
17754 Look in the directory @var{dir} to find the M4 preprocessor.
17755 The assembler uses this option.
17756 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17757 @c the generic assembler that comes with Solaris takes just -Ym.
17761 @subsection V850 Options
17762 @cindex V850 Options
17764 These @samp{-m} options are defined for V850 implementations:
17768 @itemx -mno-long-calls
17769 @opindex mlong-calls
17770 @opindex mno-long-calls
17771 Treat all calls as being far away (near). If calls are assumed to be
17772 far away, the compiler will always load the functions address up into a
17773 register, and call indirect through the pointer.
17779 Do not optimize (do optimize) basic blocks that use the same index
17780 pointer 4 or more times to copy pointer into the @code{ep} register, and
17781 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17782 option is on by default if you optimize.
17784 @item -mno-prolog-function
17785 @itemx -mprolog-function
17786 @opindex mno-prolog-function
17787 @opindex mprolog-function
17788 Do not use (do use) external functions to save and restore registers
17789 at the prologue and epilogue of a function. The external functions
17790 are slower, but use less code space if more than one function saves
17791 the same number of registers. The @option{-mprolog-function} option
17792 is on by default if you optimize.
17796 Try to make the code as small as possible. At present, this just turns
17797 on the @option{-mep} and @option{-mprolog-function} options.
17799 @item -mtda=@var{n}
17801 Put static or global variables whose size is @var{n} bytes or less into
17802 the tiny data area that register @code{ep} points to. The tiny data
17803 area can hold up to 256 bytes in total (128 bytes for byte references).
17805 @item -msda=@var{n}
17807 Put static or global variables whose size is @var{n} bytes or less into
17808 the small data area that register @code{gp} points to. The small data
17809 area can hold up to 64 kilobytes.
17811 @item -mzda=@var{n}
17813 Put static or global variables whose size is @var{n} bytes or less into
17814 the first 32 kilobytes of memory.
17818 Specify that the target processor is the V850.
17821 @opindex mbig-switch
17822 Generate code suitable for big switch tables. Use this option only if
17823 the assembler/linker complain about out of range branches within a switch
17828 This option will cause r2 and r5 to be used in the code generated by
17829 the compiler. This setting is the default.
17831 @item -mno-app-regs
17832 @opindex mno-app-regs
17833 This option will cause r2 and r5 to be treated as fixed registers.
17837 Specify that the target processor is the V850E2V3. The preprocessor
17838 constants @samp{__v850e2v3__} will be defined if
17839 this option is used.
17843 Specify that the target processor is the V850E2. The preprocessor
17844 constants @samp{__v850e2__} will be defined if this option is used.
17848 Specify that the target processor is the V850E1. The preprocessor
17849 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17850 this option is used.
17854 Specify that the target processor is the V850ES. This is an alias for
17855 the @option{-mv850e1} option.
17859 Specify that the target processor is the V850E@. The preprocessor
17860 constant @samp{__v850e__} will be defined if this option is used.
17862 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17863 nor @option{-mv850e2} nor @option{-mv850e2v3}
17864 are defined then a default target processor will be chosen and the
17865 relevant @samp{__v850*__} preprocessor constant will be defined.
17867 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17868 defined, regardless of which processor variant is the target.
17870 @item -mdisable-callt
17871 @opindex mdisable-callt
17872 This option will suppress generation of the CALLT instruction for the
17873 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17874 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17879 @subsection VAX Options
17880 @cindex VAX options
17882 These @samp{-m} options are defined for the VAX:
17887 Do not output certain jump instructions (@code{aobleq} and so on)
17888 that the Unix assembler for the VAX cannot handle across long
17893 Do output those jump instructions, on the assumption that you
17894 will assemble with the GNU assembler.
17898 Output code for g-format floating point numbers instead of d-format.
17901 @node VxWorks Options
17902 @subsection VxWorks Options
17903 @cindex VxWorks Options
17905 The options in this section are defined for all VxWorks targets.
17906 Options specific to the target hardware are listed with the other
17907 options for that target.
17912 GCC can generate code for both VxWorks kernels and real time processes
17913 (RTPs). This option switches from the former to the latter. It also
17914 defines the preprocessor macro @code{__RTP__}.
17917 @opindex non-static
17918 Link an RTP executable against shared libraries rather than static
17919 libraries. The options @option{-static} and @option{-shared} can
17920 also be used for RTPs (@pxref{Link Options}); @option{-static}
17927 These options are passed down to the linker. They are defined for
17928 compatibility with Diab.
17931 @opindex Xbind-lazy
17932 Enable lazy binding of function calls. This option is equivalent to
17933 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17937 Disable lazy binding of function calls. This option is the default and
17938 is defined for compatibility with Diab.
17941 @node x86-64 Options
17942 @subsection x86-64 Options
17943 @cindex x86-64 options
17945 These are listed under @xref{i386 and x86-64 Options}.
17947 @node Xstormy16 Options
17948 @subsection Xstormy16 Options
17949 @cindex Xstormy16 Options
17951 These options are defined for Xstormy16:
17956 Choose startup files and linker script suitable for the simulator.
17959 @node Xtensa Options
17960 @subsection Xtensa Options
17961 @cindex Xtensa Options
17963 These options are supported for Xtensa targets:
17967 @itemx -mno-const16
17969 @opindex mno-const16
17970 Enable or disable use of @code{CONST16} instructions for loading
17971 constant values. The @code{CONST16} instruction is currently not a
17972 standard option from Tensilica. When enabled, @code{CONST16}
17973 instructions are always used in place of the standard @code{L32R}
17974 instructions. The use of @code{CONST16} is enabled by default only if
17975 the @code{L32R} instruction is not available.
17978 @itemx -mno-fused-madd
17979 @opindex mfused-madd
17980 @opindex mno-fused-madd
17981 Enable or disable use of fused multiply/add and multiply/subtract
17982 instructions in the floating-point option. This has no effect if the
17983 floating-point option is not also enabled. Disabling fused multiply/add
17984 and multiply/subtract instructions forces the compiler to use separate
17985 instructions for the multiply and add/subtract operations. This may be
17986 desirable in some cases where strict IEEE 754-compliant results are
17987 required: the fused multiply add/subtract instructions do not round the
17988 intermediate result, thereby producing results with @emph{more} bits of
17989 precision than specified by the IEEE standard. Disabling fused multiply
17990 add/subtract instructions also ensures that the program output is not
17991 sensitive to the compiler's ability to combine multiply and add/subtract
17994 @item -mserialize-volatile
17995 @itemx -mno-serialize-volatile
17996 @opindex mserialize-volatile
17997 @opindex mno-serialize-volatile
17998 When this option is enabled, GCC inserts @code{MEMW} instructions before
17999 @code{volatile} memory references to guarantee sequential consistency.
18000 The default is @option{-mserialize-volatile}. Use
18001 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
18003 @item -mforce-no-pic
18004 @opindex mforce-no-pic
18005 For targets, like GNU/Linux, where all user-mode Xtensa code must be
18006 position-independent code (PIC), this option disables PIC for compiling
18009 @item -mtext-section-literals
18010 @itemx -mno-text-section-literals
18011 @opindex mtext-section-literals
18012 @opindex mno-text-section-literals
18013 Control the treatment of literal pools. The default is
18014 @option{-mno-text-section-literals}, which places literals in a separate
18015 section in the output file. This allows the literal pool to be placed
18016 in a data RAM/ROM, and it also allows the linker to combine literal
18017 pools from separate object files to remove redundant literals and
18018 improve code size. With @option{-mtext-section-literals}, the literals
18019 are interspersed in the text section in order to keep them as close as
18020 possible to their references. This may be necessary for large assembly
18023 @item -mtarget-align
18024 @itemx -mno-target-align
18025 @opindex mtarget-align
18026 @opindex mno-target-align
18027 When this option is enabled, GCC instructs the assembler to
18028 automatically align instructions to reduce branch penalties at the
18029 expense of some code density. The assembler attempts to widen density
18030 instructions to align branch targets and the instructions following call
18031 instructions. If there are not enough preceding safe density
18032 instructions to align a target, no widening will be performed. The
18033 default is @option{-mtarget-align}. These options do not affect the
18034 treatment of auto-aligned instructions like @code{LOOP}, which the
18035 assembler will always align, either by widening density instructions or
18036 by inserting no-op instructions.
18039 @itemx -mno-longcalls
18040 @opindex mlongcalls
18041 @opindex mno-longcalls
18042 When this option is enabled, GCC instructs the assembler to translate
18043 direct calls to indirect calls unless it can determine that the target
18044 of a direct call is in the range allowed by the call instruction. This
18045 translation typically occurs for calls to functions in other source
18046 files. Specifically, the assembler translates a direct @code{CALL}
18047 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
18048 The default is @option{-mno-longcalls}. This option should be used in
18049 programs where the call target can potentially be out of range. This
18050 option is implemented in the assembler, not the compiler, so the
18051 assembly code generated by GCC will still show direct call
18052 instructions---look at the disassembled object code to see the actual
18053 instructions. Note that the assembler will use an indirect call for
18054 every cross-file call, not just those that really will be out of range.
18057 @node zSeries Options
18058 @subsection zSeries Options
18059 @cindex zSeries options
18061 These are listed under @xref{S/390 and zSeries Options}.
18063 @node Code Gen Options
18064 @section Options for Code Generation Conventions
18065 @cindex code generation conventions
18066 @cindex options, code generation
18067 @cindex run-time options
18069 These machine-independent options control the interface conventions
18070 used in code generation.
18072 Most of them have both positive and negative forms; the negative form
18073 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
18074 one of the forms is listed---the one which is not the default. You
18075 can figure out the other form by either removing @samp{no-} or adding
18079 @item -fbounds-check
18080 @opindex fbounds-check
18081 For front-ends that support it, generate additional code to check that
18082 indices used to access arrays are within the declared range. This is
18083 currently only supported by the Java and Fortran front-ends, where
18084 this option defaults to true and false respectively.
18088 This option generates traps for signed overflow on addition, subtraction,
18089 multiplication operations.
18093 This option instructs the compiler to assume that signed arithmetic
18094 overflow of addition, subtraction and multiplication wraps around
18095 using twos-complement representation. This flag enables some optimizations
18096 and disables others. This option is enabled by default for the Java
18097 front-end, as required by the Java language specification.
18100 @opindex fexceptions
18101 Enable exception handling. Generates extra code needed to propagate
18102 exceptions. For some targets, this implies GCC will generate frame
18103 unwind information for all functions, which can produce significant data
18104 size overhead, although it does not affect execution. If you do not
18105 specify this option, GCC will enable it by default for languages like
18106 C++ which normally require exception handling, and disable it for
18107 languages like C that do not normally require it. However, you may need
18108 to enable this option when compiling C code that needs to interoperate
18109 properly with exception handlers written in C++. You may also wish to
18110 disable this option if you are compiling older C++ programs that don't
18111 use exception handling.
18113 @item -fnon-call-exceptions
18114 @opindex fnon-call-exceptions
18115 Generate code that allows trapping instructions to throw exceptions.
18116 Note that this requires platform-specific runtime support that does
18117 not exist everywhere. Moreover, it only allows @emph{trapping}
18118 instructions to throw exceptions, i.e.@: memory references or floating
18119 point instructions. It does not allow exceptions to be thrown from
18120 arbitrary signal handlers such as @code{SIGALRM}.
18122 @item -funwind-tables
18123 @opindex funwind-tables
18124 Similar to @option{-fexceptions}, except that it will just generate any needed
18125 static data, but will not affect the generated code in any other way.
18126 You will normally not enable this option; instead, a language processor
18127 that needs this handling would enable it on your behalf.
18129 @item -fasynchronous-unwind-tables
18130 @opindex fasynchronous-unwind-tables
18131 Generate unwind table in dwarf2 format, if supported by target machine. The
18132 table is exact at each instruction boundary, so it can be used for stack
18133 unwinding from asynchronous events (such as debugger or garbage collector).
18135 @item -fpcc-struct-return
18136 @opindex fpcc-struct-return
18137 Return ``short'' @code{struct} and @code{union} values in memory like
18138 longer ones, rather than in registers. This convention is less
18139 efficient, but it has the advantage of allowing intercallability between
18140 GCC-compiled files and files compiled with other compilers, particularly
18141 the Portable C Compiler (pcc).
18143 The precise convention for returning structures in memory depends
18144 on the target configuration macros.
18146 Short structures and unions are those whose size and alignment match
18147 that of some integer type.
18149 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
18150 switch is not binary compatible with code compiled with the
18151 @option{-freg-struct-return} switch.
18152 Use it to conform to a non-default application binary interface.
18154 @item -freg-struct-return
18155 @opindex freg-struct-return
18156 Return @code{struct} and @code{union} values in registers when possible.
18157 This is more efficient for small structures than
18158 @option{-fpcc-struct-return}.
18160 If you specify neither @option{-fpcc-struct-return} nor
18161 @option{-freg-struct-return}, GCC defaults to whichever convention is
18162 standard for the target. If there is no standard convention, GCC
18163 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18164 the principal compiler. In those cases, we can choose the standard, and
18165 we chose the more efficient register return alternative.
18167 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18168 switch is not binary compatible with code compiled with the
18169 @option{-fpcc-struct-return} switch.
18170 Use it to conform to a non-default application binary interface.
18172 @item -fshort-enums
18173 @opindex fshort-enums
18174 Allocate to an @code{enum} type only as many bytes as it needs for the
18175 declared range of possible values. Specifically, the @code{enum} type
18176 will be equivalent to the smallest integer type which has enough room.
18178 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18179 code that is not binary compatible with code generated without that switch.
18180 Use it to conform to a non-default application binary interface.
18182 @item -fshort-double
18183 @opindex fshort-double
18184 Use the same size for @code{double} as for @code{float}.
18186 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18187 code that is not binary compatible with code generated without that switch.
18188 Use it to conform to a non-default application binary interface.
18190 @item -fshort-wchar
18191 @opindex fshort-wchar
18192 Override the underlying type for @samp{wchar_t} to be @samp{short
18193 unsigned int} instead of the default for the target. This option is
18194 useful for building programs to run under WINE@.
18196 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18197 code that is not binary compatible with code generated without that switch.
18198 Use it to conform to a non-default application binary interface.
18201 @opindex fno-common
18202 In C code, controls the placement of uninitialized global variables.
18203 Unix C compilers have traditionally permitted multiple definitions of
18204 such variables in different compilation units by placing the variables
18206 This is the behavior specified by @option{-fcommon}, and is the default
18207 for GCC on most targets.
18208 On the other hand, this behavior is not required by ISO C, and on some
18209 targets may carry a speed or code size penalty on variable references.
18210 The @option{-fno-common} option specifies that the compiler should place
18211 uninitialized global variables in the data section of the object file,
18212 rather than generating them as common blocks.
18213 This has the effect that if the same variable is declared
18214 (without @code{extern}) in two different compilations,
18215 you will get a multiple-definition error when you link them.
18216 In this case, you must compile with @option{-fcommon} instead.
18217 Compiling with @option{-fno-common} is useful on targets for which
18218 it provides better performance, or if you wish to verify that the
18219 program will work on other systems which always treat uninitialized
18220 variable declarations this way.
18224 Ignore the @samp{#ident} directive.
18226 @item -finhibit-size-directive
18227 @opindex finhibit-size-directive
18228 Don't output a @code{.size} assembler directive, or anything else that
18229 would cause trouble if the function is split in the middle, and the
18230 two halves are placed at locations far apart in memory. This option is
18231 used when compiling @file{crtstuff.c}; you should not need to use it
18234 @item -fverbose-asm
18235 @opindex fverbose-asm
18236 Put extra commentary information in the generated assembly code to
18237 make it more readable. This option is generally only of use to those
18238 who actually need to read the generated assembly code (perhaps while
18239 debugging the compiler itself).
18241 @option{-fno-verbose-asm}, the default, causes the
18242 extra information to be omitted and is useful when comparing two assembler
18245 @item -frecord-gcc-switches
18246 @opindex frecord-gcc-switches
18247 This switch causes the command line that was used to invoke the
18248 compiler to be recorded into the object file that is being created.
18249 This switch is only implemented on some targets and the exact format
18250 of the recording is target and binary file format dependent, but it
18251 usually takes the form of a section containing ASCII text. This
18252 switch is related to the @option{-fverbose-asm} switch, but that
18253 switch only records information in the assembler output file as
18254 comments, so it never reaches the object file.
18255 See also @option{-grecord-gcc-switches} for another
18256 way of storing compiler options into the object file.
18260 @cindex global offset table
18262 Generate position-independent code (PIC) suitable for use in a shared
18263 library, if supported for the target machine. Such code accesses all
18264 constant addresses through a global offset table (GOT)@. The dynamic
18265 loader resolves the GOT entries when the program starts (the dynamic
18266 loader is not part of GCC; it is part of the operating system). If
18267 the GOT size for the linked executable exceeds a machine-specific
18268 maximum size, you get an error message from the linker indicating that
18269 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18270 instead. (These maximums are 8k on the SPARC and 32k
18271 on the m68k and RS/6000. The 386 has no such limit.)
18273 Position-independent code requires special support, and therefore works
18274 only on certain machines. For the 386, GCC supports PIC for System V
18275 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18276 position-independent.
18278 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18283 If supported for the target machine, emit position-independent code,
18284 suitable for dynamic linking and avoiding any limit on the size of the
18285 global offset table. This option makes a difference on the m68k,
18286 PowerPC and SPARC@.
18288 Position-independent code requires special support, and therefore works
18289 only on certain machines.
18291 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18298 These options are similar to @option{-fpic} and @option{-fPIC}, but
18299 generated position independent code can be only linked into executables.
18300 Usually these options are used when @option{-pie} GCC option will be
18301 used during linking.
18303 @option{-fpie} and @option{-fPIE} both define the macros
18304 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18305 for @option{-fpie} and 2 for @option{-fPIE}.
18307 @item -fno-jump-tables
18308 @opindex fno-jump-tables
18309 Do not use jump tables for switch statements even where it would be
18310 more efficient than other code generation strategies. This option is
18311 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18312 building code which forms part of a dynamic linker and cannot
18313 reference the address of a jump table. On some targets, jump tables
18314 do not require a GOT and this option is not needed.
18316 @item -ffixed-@var{reg}
18318 Treat the register named @var{reg} as a fixed register; generated code
18319 should never refer to it (except perhaps as a stack pointer, frame
18320 pointer or in some other fixed role).
18322 @var{reg} must be the name of a register. The register names accepted
18323 are machine-specific and are defined in the @code{REGISTER_NAMES}
18324 macro in the machine description macro file.
18326 This flag does not have a negative form, because it specifies a
18329 @item -fcall-used-@var{reg}
18330 @opindex fcall-used
18331 Treat the register named @var{reg} as an allocable register that is
18332 clobbered by function calls. It may be allocated for temporaries or
18333 variables that do not live across a call. Functions compiled this way
18334 will not save and restore the register @var{reg}.
18336 It is an error to used this flag with the frame pointer or stack pointer.
18337 Use of this flag for other registers that have fixed pervasive roles in
18338 the machine's execution model will produce disastrous results.
18340 This flag does not have a negative form, because it specifies a
18343 @item -fcall-saved-@var{reg}
18344 @opindex fcall-saved
18345 Treat the register named @var{reg} as an allocable register saved by
18346 functions. It may be allocated even for temporaries or variables that
18347 live across a call. Functions compiled this way will save and restore
18348 the register @var{reg} if they use it.
18350 It is an error to used this flag with the frame pointer or stack pointer.
18351 Use of this flag for other registers that have fixed pervasive roles in
18352 the machine's execution model will produce disastrous results.
18354 A different sort of disaster will result from the use of this flag for
18355 a register in which function values may be returned.
18357 This flag does not have a negative form, because it specifies a
18360 @item -fpack-struct[=@var{n}]
18361 @opindex fpack-struct
18362 Without a value specified, pack all structure members together without
18363 holes. When a value is specified (which must be a small power of two), pack
18364 structure members according to this value, representing the maximum
18365 alignment (that is, objects with default alignment requirements larger than
18366 this will be output potentially unaligned at the next fitting location.
18368 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18369 code that is not binary compatible with code generated without that switch.
18370 Additionally, it makes the code suboptimal.
18371 Use it to conform to a non-default application binary interface.
18373 @item -finstrument-functions
18374 @opindex finstrument-functions
18375 Generate instrumentation calls for entry and exit to functions. Just
18376 after function entry and just before function exit, the following
18377 profiling functions will be called with the address of the current
18378 function and its call site. (On some platforms,
18379 @code{__builtin_return_address} does not work beyond the current
18380 function, so the call site information may not be available to the
18381 profiling functions otherwise.)
18384 void __cyg_profile_func_enter (void *this_fn,
18386 void __cyg_profile_func_exit (void *this_fn,
18390 The first argument is the address of the start of the current function,
18391 which may be looked up exactly in the symbol table.
18393 This instrumentation is also done for functions expanded inline in other
18394 functions. The profiling calls will indicate where, conceptually, the
18395 inline function is entered and exited. This means that addressable
18396 versions of such functions must be available. If all your uses of a
18397 function are expanded inline, this may mean an additional expansion of
18398 code size. If you use @samp{extern inline} in your C code, an
18399 addressable version of such functions must be provided. (This is
18400 normally the case anyways, but if you get lucky and the optimizer always
18401 expands the functions inline, you might have gotten away without
18402 providing static copies.)
18404 A function may be given the attribute @code{no_instrument_function}, in
18405 which case this instrumentation will not be done. This can be used, for
18406 example, for the profiling functions listed above, high-priority
18407 interrupt routines, and any functions from which the profiling functions
18408 cannot safely be called (perhaps signal handlers, if the profiling
18409 routines generate output or allocate memory).
18411 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18412 @opindex finstrument-functions-exclude-file-list
18414 Set the list of functions that are excluded from instrumentation (see
18415 the description of @code{-finstrument-functions}). If the file that
18416 contains a function definition matches with one of @var{file}, then
18417 that function is not instrumented. The match is done on substrings:
18418 if the @var{file} parameter is a substring of the file name, it is
18419 considered to be a match.
18424 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18428 will exclude any inline function defined in files whose pathnames
18429 contain @code{/bits/stl} or @code{include/sys}.
18431 If, for some reason, you want to include letter @code{','} in one of
18432 @var{sym}, write @code{'\,'}. For example,
18433 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18434 (note the single quote surrounding the option).
18436 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18437 @opindex finstrument-functions-exclude-function-list
18439 This is similar to @code{-finstrument-functions-exclude-file-list},
18440 but this option sets the list of function names to be excluded from
18441 instrumentation. The function name to be matched is its user-visible
18442 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18443 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18444 match is done on substrings: if the @var{sym} parameter is a substring
18445 of the function name, it is considered to be a match. For C99 and C++
18446 extended identifiers, the function name must be given in UTF-8, not
18447 using universal character names.
18449 @item -fstack-check
18450 @opindex fstack-check
18451 Generate code to verify that you do not go beyond the boundary of the
18452 stack. You should specify this flag if you are running in an
18453 environment with multiple threads, but only rarely need to specify it in
18454 a single-threaded environment since stack overflow is automatically
18455 detected on nearly all systems if there is only one stack.
18457 Note that this switch does not actually cause checking to be done; the
18458 operating system or the language runtime must do that. The switch causes
18459 generation of code to ensure that they see the stack being extended.
18461 You can additionally specify a string parameter: @code{no} means no
18462 checking, @code{generic} means force the use of old-style checking,
18463 @code{specific} means use the best checking method and is equivalent
18464 to bare @option{-fstack-check}.
18466 Old-style checking is a generic mechanism that requires no specific
18467 target support in the compiler but comes with the following drawbacks:
18471 Modified allocation strategy for large objects: they will always be
18472 allocated dynamically if their size exceeds a fixed threshold.
18475 Fixed limit on the size of the static frame of functions: when it is
18476 topped by a particular function, stack checking is not reliable and
18477 a warning is issued by the compiler.
18480 Inefficiency: because of both the modified allocation strategy and the
18481 generic implementation, the performances of the code are hampered.
18484 Note that old-style stack checking is also the fallback method for
18485 @code{specific} if no target support has been added in the compiler.
18487 @item -fstack-limit-register=@var{reg}
18488 @itemx -fstack-limit-symbol=@var{sym}
18489 @itemx -fno-stack-limit
18490 @opindex fstack-limit-register
18491 @opindex fstack-limit-symbol
18492 @opindex fno-stack-limit
18493 Generate code to ensure that the stack does not grow beyond a certain value,
18494 either the value of a register or the address of a symbol. If the stack
18495 would grow beyond the value, a signal is raised. For most targets,
18496 the signal is raised before the stack overruns the boundary, so
18497 it is possible to catch the signal without taking special precautions.
18499 For instance, if the stack starts at absolute address @samp{0x80000000}
18500 and grows downwards, you can use the flags
18501 @option{-fstack-limit-symbol=__stack_limit} and
18502 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18503 of 128KB@. Note that this may only work with the GNU linker.
18505 @item -fsplit-stack
18506 @opindex fsplit-stack
18507 Generate code to automatically split the stack before it overflows.
18508 The resulting program has a discontiguous stack which can only
18509 overflow if the program is unable to allocate any more memory. This
18510 is most useful when running threaded programs, as it is no longer
18511 necessary to calculate a good stack size to use for each thread. This
18512 is currently only implemented for the i386 and x86_64 backends running
18515 When code compiled with @option{-fsplit-stack} calls code compiled
18516 without @option{-fsplit-stack}, there may not be much stack space
18517 available for the latter code to run. If compiling all code,
18518 including library code, with @option{-fsplit-stack} is not an option,
18519 then the linker can fix up these calls so that the code compiled
18520 without @option{-fsplit-stack} always has a large stack. Support for
18521 this is implemented in the gold linker in GNU binutils release 2.21
18524 @item -fleading-underscore
18525 @opindex fleading-underscore
18526 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18527 change the way C symbols are represented in the object file. One use
18528 is to help link with legacy assembly code.
18530 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18531 generate code that is not binary compatible with code generated without that
18532 switch. Use it to conform to a non-default application binary interface.
18533 Not all targets provide complete support for this switch.
18535 @item -ftls-model=@var{model}
18536 @opindex ftls-model
18537 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18538 The @var{model} argument should be one of @code{global-dynamic},
18539 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18541 The default without @option{-fpic} is @code{initial-exec}; with
18542 @option{-fpic} the default is @code{global-dynamic}.
18544 @item -fvisibility=@var{default|internal|hidden|protected}
18545 @opindex fvisibility
18546 Set the default ELF image symbol visibility to the specified option---all
18547 symbols will be marked with this unless overridden within the code.
18548 Using this feature can very substantially improve linking and
18549 load times of shared object libraries, produce more optimized
18550 code, provide near-perfect API export and prevent symbol clashes.
18551 It is @strong{strongly} recommended that you use this in any shared objects
18554 Despite the nomenclature, @code{default} always means public; i.e.,
18555 available to be linked against from outside the shared object.
18556 @code{protected} and @code{internal} are pretty useless in real-world
18557 usage so the only other commonly used option will be @code{hidden}.
18558 The default if @option{-fvisibility} isn't specified is
18559 @code{default}, i.e., make every
18560 symbol public---this causes the same behavior as previous versions of
18563 A good explanation of the benefits offered by ensuring ELF
18564 symbols have the correct visibility is given by ``How To Write
18565 Shared Libraries'' by Ulrich Drepper (which can be found at
18566 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18567 solution made possible by this option to marking things hidden when
18568 the default is public is to make the default hidden and mark things
18569 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18570 and @code{__attribute__ ((visibility("default")))} instead of
18571 @code{__declspec(dllexport)} you get almost identical semantics with
18572 identical syntax. This is a great boon to those working with
18573 cross-platform projects.
18575 For those adding visibility support to existing code, you may find
18576 @samp{#pragma GCC visibility} of use. This works by you enclosing
18577 the declarations you wish to set visibility for with (for example)
18578 @samp{#pragma GCC visibility push(hidden)} and
18579 @samp{#pragma GCC visibility pop}.
18580 Bear in mind that symbol visibility should be viewed @strong{as
18581 part of the API interface contract} and thus all new code should
18582 always specify visibility when it is not the default; i.e., declarations
18583 only for use within the local DSO should @strong{always} be marked explicitly
18584 as hidden as so to avoid PLT indirection overheads---making this
18585 abundantly clear also aids readability and self-documentation of the code.
18586 Note that due to ISO C++ specification requirements, operator new and
18587 operator delete must always be of default visibility.
18589 Be aware that headers from outside your project, in particular system
18590 headers and headers from any other library you use, may not be
18591 expecting to be compiled with visibility other than the default. You
18592 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18593 before including any such headers.
18595 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18596 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18597 no modifications. However, this means that calls to @samp{extern}
18598 functions with no explicit visibility will use the PLT, so it is more
18599 effective to use @samp{__attribute ((visibility))} and/or
18600 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18601 declarations should be treated as hidden.
18603 Note that @samp{-fvisibility} does affect C++ vague linkage
18604 entities. This means that, for instance, an exception class that will
18605 be thrown between DSOs must be explicitly marked with default
18606 visibility so that the @samp{type_info} nodes will be unified between
18609 An overview of these techniques, their benefits and how to use them
18610 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18612 @item -fstrict-volatile-bitfields
18613 @opindex fstrict-volatile-bitfields
18614 This option should be used if accesses to volatile bitfields (or other
18615 structure fields, although the compiler usually honors those types
18616 anyway) should use a single access of the width of the
18617 field's type, aligned to a natural alignment if possible. For
18618 example, targets with memory-mapped peripheral registers might require
18619 all such accesses to be 16 bits wide; with this flag the user could
18620 declare all peripheral bitfields as ``unsigned short'' (assuming short
18621 is 16 bits on these targets) to force GCC to use 16 bit accesses
18622 instead of, perhaps, a more efficient 32 bit access.
18624 If this option is disabled, the compiler will use the most efficient
18625 instruction. In the previous example, that might be a 32-bit load
18626 instruction, even though that will access bytes that do not contain
18627 any portion of the bitfield, or memory-mapped registers unrelated to
18628 the one being updated.
18630 If the target requires strict alignment, and honoring the field
18631 type would require violating this alignment, a warning is issued.
18632 If the field has @code{packed} attribute, the access is done without
18633 honoring the field type. If the field doesn't have @code{packed}
18634 attribute, the access is done honoring the field type. In both cases,
18635 GCC assumes that the user knows something about the target hardware
18636 that it is unaware of.
18638 The default value of this option is determined by the application binary
18639 interface for the target processor.
18645 @node Environment Variables
18646 @section Environment Variables Affecting GCC
18647 @cindex environment variables
18649 @c man begin ENVIRONMENT
18650 This section describes several environment variables that affect how GCC
18651 operates. Some of them work by specifying directories or prefixes to use
18652 when searching for various kinds of files. Some are used to specify other
18653 aspects of the compilation environment.
18655 Note that you can also specify places to search using options such as
18656 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18657 take precedence over places specified using environment variables, which
18658 in turn take precedence over those specified by the configuration of GCC@.
18659 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18660 GNU Compiler Collection (GCC) Internals}.
18665 @c @itemx LC_COLLATE
18667 @c @itemx LC_MONETARY
18668 @c @itemx LC_NUMERIC
18673 @c @findex LC_COLLATE
18674 @findex LC_MESSAGES
18675 @c @findex LC_MONETARY
18676 @c @findex LC_NUMERIC
18680 These environment variables control the way that GCC uses
18681 localization information that allow GCC to work with different
18682 national conventions. GCC inspects the locale categories
18683 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18684 so. These locale categories can be set to any value supported by your
18685 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18686 Kingdom encoded in UTF-8.
18688 The @env{LC_CTYPE} environment variable specifies character
18689 classification. GCC uses it to determine the character boundaries in
18690 a string; this is needed for some multibyte encodings that contain quote
18691 and escape characters that would otherwise be interpreted as a string
18694 The @env{LC_MESSAGES} environment variable specifies the language to
18695 use in diagnostic messages.
18697 If the @env{LC_ALL} environment variable is set, it overrides the value
18698 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18699 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18700 environment variable. If none of these variables are set, GCC
18701 defaults to traditional C English behavior.
18705 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18706 files. GCC uses temporary files to hold the output of one stage of
18707 compilation which is to be used as input to the next stage: for example,
18708 the output of the preprocessor, which is the input to the compiler
18711 @item GCC_COMPARE_DEBUG
18712 @findex GCC_COMPARE_DEBUG
18713 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
18714 @option{-fcompare-debug} to the compiler driver. See the documentation
18715 of this option for more details.
18717 @item GCC_EXEC_PREFIX
18718 @findex GCC_EXEC_PREFIX
18719 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18720 names of the subprograms executed by the compiler. No slash is added
18721 when this prefix is combined with the name of a subprogram, but you can
18722 specify a prefix that ends with a slash if you wish.
18724 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18725 an appropriate prefix to use based on the pathname it was invoked with.
18727 If GCC cannot find the subprogram using the specified prefix, it
18728 tries looking in the usual places for the subprogram.
18730 The default value of @env{GCC_EXEC_PREFIX} is
18731 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18732 the installed compiler. In many cases @var{prefix} is the value
18733 of @code{prefix} when you ran the @file{configure} script.
18735 Other prefixes specified with @option{-B} take precedence over this prefix.
18737 This prefix is also used for finding files such as @file{crt0.o} that are
18740 In addition, the prefix is used in an unusual way in finding the
18741 directories to search for header files. For each of the standard
18742 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18743 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18744 replacing that beginning with the specified prefix to produce an
18745 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18746 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18747 These alternate directories are searched first; the standard directories
18748 come next. If a standard directory begins with the configured
18749 @var{prefix} then the value of @var{prefix} is replaced by
18750 @env{GCC_EXEC_PREFIX} when looking for header files.
18752 @item COMPILER_PATH
18753 @findex COMPILER_PATH
18754 The value of @env{COMPILER_PATH} is a colon-separated list of
18755 directories, much like @env{PATH}. GCC tries the directories thus
18756 specified when searching for subprograms, if it can't find the
18757 subprograms using @env{GCC_EXEC_PREFIX}.
18760 @findex LIBRARY_PATH
18761 The value of @env{LIBRARY_PATH} is a colon-separated list of
18762 directories, much like @env{PATH}. When configured as a native compiler,
18763 GCC tries the directories thus specified when searching for special
18764 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18765 using GCC also uses these directories when searching for ordinary
18766 libraries for the @option{-l} option (but directories specified with
18767 @option{-L} come first).
18771 @cindex locale definition
18772 This variable is used to pass locale information to the compiler. One way in
18773 which this information is used is to determine the character set to be used
18774 when character literals, string literals and comments are parsed in C and C++.
18775 When the compiler is configured to allow multibyte characters,
18776 the following values for @env{LANG} are recognized:
18780 Recognize JIS characters.
18782 Recognize SJIS characters.
18784 Recognize EUCJP characters.
18787 If @env{LANG} is not defined, or if it has some other value, then the
18788 compiler will use mblen and mbtowc as defined by the default locale to
18789 recognize and translate multibyte characters.
18793 Some additional environments variables affect the behavior of the
18796 @include cppenv.texi
18800 @node Precompiled Headers
18801 @section Using Precompiled Headers
18802 @cindex precompiled headers
18803 @cindex speed of compilation
18805 Often large projects have many header files that are included in every
18806 source file. The time the compiler takes to process these header files
18807 over and over again can account for nearly all of the time required to
18808 build the project. To make builds faster, GCC allows users to
18809 `precompile' a header file; then, if builds can use the precompiled
18810 header file they will be much faster.
18812 To create a precompiled header file, simply compile it as you would any
18813 other file, if necessary using the @option{-x} option to make the driver
18814 treat it as a C or C++ header file. You will probably want to use a
18815 tool like @command{make} to keep the precompiled header up-to-date when
18816 the headers it contains change.
18818 A precompiled header file will be searched for when @code{#include} is
18819 seen in the compilation. As it searches for the included file
18820 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18821 compiler looks for a precompiled header in each directory just before it
18822 looks for the include file in that directory. The name searched for is
18823 the name specified in the @code{#include} with @samp{.gch} appended. If
18824 the precompiled header file can't be used, it is ignored.
18826 For instance, if you have @code{#include "all.h"}, and you have
18827 @file{all.h.gch} in the same directory as @file{all.h}, then the
18828 precompiled header file will be used if possible, and the original
18829 header will be used otherwise.
18831 Alternatively, you might decide to put the precompiled header file in a
18832 directory and use @option{-I} to ensure that directory is searched
18833 before (or instead of) the directory containing the original header.
18834 Then, if you want to check that the precompiled header file is always
18835 used, you can put a file of the same name as the original header in this
18836 directory containing an @code{#error} command.
18838 This also works with @option{-include}. So yet another way to use
18839 precompiled headers, good for projects not designed with precompiled
18840 header files in mind, is to simply take most of the header files used by
18841 a project, include them from another header file, precompile that header
18842 file, and @option{-include} the precompiled header. If the header files
18843 have guards against multiple inclusion, they will be skipped because
18844 they've already been included (in the precompiled header).
18846 If you need to precompile the same header file for different
18847 languages, targets, or compiler options, you can instead make a
18848 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18849 header in the directory, perhaps using @option{-o}. It doesn't matter
18850 what you call the files in the directory, every precompiled header in
18851 the directory will be considered. The first precompiled header
18852 encountered in the directory that is valid for this compilation will
18853 be used; they're searched in no particular order.
18855 There are many other possibilities, limited only by your imagination,
18856 good sense, and the constraints of your build system.
18858 A precompiled header file can be used only when these conditions apply:
18862 Only one precompiled header can be used in a particular compilation.
18865 A precompiled header can't be used once the first C token is seen. You
18866 can have preprocessor directives before a precompiled header; you can
18867 even include a precompiled header from inside another header, so long as
18868 there are no C tokens before the @code{#include}.
18871 The precompiled header file must be produced for the same language as
18872 the current compilation. You can't use a C precompiled header for a C++
18876 The precompiled header file must have been produced by the same compiler
18877 binary as the current compilation is using.
18880 Any macros defined before the precompiled header is included must
18881 either be defined in the same way as when the precompiled header was
18882 generated, or must not affect the precompiled header, which usually
18883 means that they don't appear in the precompiled header at all.
18885 The @option{-D} option is one way to define a macro before a
18886 precompiled header is included; using a @code{#define} can also do it.
18887 There are also some options that define macros implicitly, like
18888 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18891 @item If debugging information is output when using the precompiled
18892 header, using @option{-g} or similar, the same kind of debugging information
18893 must have been output when building the precompiled header. However,
18894 a precompiled header built using @option{-g} can be used in a compilation
18895 when no debugging information is being output.
18897 @item The same @option{-m} options must generally be used when building
18898 and using the precompiled header. @xref{Submodel Options},
18899 for any cases where this rule is relaxed.
18901 @item Each of the following options must be the same when building and using
18902 the precompiled header:
18904 @gccoptlist{-fexceptions}
18907 Some other command-line options starting with @option{-f},
18908 @option{-p}, or @option{-O} must be defined in the same way as when
18909 the precompiled header was generated. At present, it's not clear
18910 which options are safe to change and which are not; the safest choice
18911 is to use exactly the same options when generating and using the
18912 precompiled header. The following are known to be safe:
18914 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18915 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18916 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18921 For all of these except the last, the compiler will automatically
18922 ignore the precompiled header if the conditions aren't met. If you
18923 find an option combination that doesn't work and doesn't cause the
18924 precompiled header to be ignored, please consider filing a bug report,
18927 If you do use differing options when generating and using the
18928 precompiled header, the actual behavior will be a mixture of the
18929 behavior for the options. For instance, if you use @option{-g} to
18930 generate the precompiled header but not when using it, you may or may
18931 not get debugging information for routines in the precompiled header.