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-parameter -Wno-unused-result -Wunused-value @gol
272 -Wunused-variable -Wunused-but-set-parameter -Wunused-but-set-variable @gol
273 -Wvariadic-macros -Wvla -Wvolatile-register-var -Wwrite-strings}
275 @item C and Objective-C-only Warning Options
276 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
277 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
278 -Wold-style-declaration -Wold-style-definition @gol
279 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
280 -Wdeclaration-after-statement -Wpointer-sign}
282 @item Debugging Options
283 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
284 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
285 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
286 -fdisable-ipa-@var{pass_name} @gol
287 -fdisable-rtl-@var{pass_name} @gol
288 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
289 -fdisable-tree-@var{pass_name} @gol
290 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
291 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
292 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
293 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
294 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
296 -fdump-statistics @gol
298 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
299 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
300 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
302 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
303 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
304 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
305 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
306 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
307 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
308 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
309 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
310 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
311 -fdump-tree-nrv -fdump-tree-vect @gol
312 -fdump-tree-sink @gol
313 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
314 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
315 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
316 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
317 -ftree-vectorizer-verbose=@var{n} @gol
318 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
319 -fdump-final-insns=@var{file} @gol
320 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
321 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
322 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
323 -fenable-@var{kind}-@var{pass} @gol
324 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
325 -fdebug-types-section @gol
326 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
327 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
328 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
329 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
330 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
331 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
332 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
333 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
334 -gvms -gxcoff -gxcoff+ @gol
335 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
336 -fdebug-prefix-map=@var{old}=@var{new} @gol
337 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
338 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
339 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
340 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
341 -print-prog-name=@var{program} -print-search-dirs -Q @gol
342 -print-sysroot -print-sysroot-headers-suffix @gol
343 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
345 @item Optimization Options
346 @xref{Optimize Options,,Options that Control Optimization}.
347 @gccoptlist{-falign-functions[=@var{n}] -falign-jumps[=@var{n}] @gol
348 -falign-labels[=@var{n}] -falign-loops[=@var{n}] -fassociative-math @gol
349 -fauto-inc-dec -fbranch-probabilities -fbranch-target-load-optimize @gol
350 -fbranch-target-load-optimize2 -fbtr-bb-exclusive -fcaller-saves @gol
351 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
352 -fcompare-elim -fcprop-registers -fcrossjumping @gol
353 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
354 -fcx-limited-range @gol
355 -fdata-sections -fdce -fdce -fdelayed-branch @gol
356 -fdelete-null-pointer-checks -fdse -fdevirtualize -fdse @gol
357 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffast-math @gol
358 -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
359 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
360 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
361 -fgcse-sm -fif-conversion -fif-conversion2 -findirect-inlining @gol
362 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
363 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-matrix-reorg @gol
364 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference @gol
365 -fira-algorithm=@var{algorithm} @gol
366 -fira-region=@var{region} @gol
367 -fira-loop-pressure -fno-ira-share-save-slots @gol
368 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
369 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
370 -floop-block -floop-flatten -floop-interchange -floop-strip-mine @gol
371 -floop-parallelize-all -flto -flto-compression-level @gol
372 -flto-partition=@var{alg} -flto-report -fmerge-all-constants @gol
373 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
374 -fmove-loop-invariants fmudflap -fmudflapir -fmudflapth -fno-branch-count-reg @gol
375 -fno-default-inline @gol
376 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
377 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
378 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
379 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
380 -fomit-frame-pointer -foptimize-register-move -foptimize-sibling-calls @gol
381 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
382 -fprefetch-loop-arrays @gol
383 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
384 -fprofile-generate=@var{path} @gol
385 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
386 -freciprocal-math -fregmove -frename-registers -freorder-blocks @gol
387 -freorder-blocks-and-partition -freorder-functions @gol
388 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
389 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
390 -fsched-spec-load -fsched-spec-load-dangerous @gol
391 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
392 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
393 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
394 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
395 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
396 -fselective-scheduling -fselective-scheduling2 @gol
397 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
398 -fsignaling-nans -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
399 -fsplit-wide-types -fstack-protector -fstack-protector-all @gol
400 -fstrict-aliasing -fstrict-overflow -fthread-jumps -ftracer @gol
402 -ftree-builtin-call-dce -ftree-ccp -ftree-ch -ftree-copy-prop @gol
403 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
404 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
405 -ftree-loop-if-convert-stores -ftree-loop-im @gol
406 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
407 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
408 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-pta -ftree-reassoc @gol
409 -ftree-sink -ftree-sra -ftree-switch-conversion @gol
410 -ftree-ter -ftree-vect-loop-version -ftree-vectorize -ftree-vrp @gol
411 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
412 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
413 -fvariable-expansion-in-unroller -fvect-cost-model -fvpt -fweb @gol
414 -fwhole-program -fwpa -fuse-linker-plugin @gol
415 --param @var{name}=@var{value}
416 -O -O0 -O1 -O2 -O3 -Os -Ofast}
418 @item Preprocessor Options
419 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
420 @gccoptlist{-A@var{question}=@var{answer} @gol
421 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
422 -C -dD -dI -dM -dN @gol
423 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
424 -idirafter @var{dir} @gol
425 -include @var{file} -imacros @var{file} @gol
426 -iprefix @var{file} -iwithprefix @var{dir} @gol
427 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
428 -imultilib @var{dir} -isysroot @var{dir} @gol
429 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
430 -P -fworking-directory -remap @gol
431 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
432 -Xpreprocessor @var{option}}
434 @item Assembler Option
435 @xref{Assembler Options,,Passing Options to the Assembler}.
436 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
439 @xref{Link Options,,Options for Linking}.
440 @gccoptlist{@var{object-file-name} -l@var{library} @gol
441 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
442 -s -static -static-libgcc -static-libstdc++ -shared @gol
443 -shared-libgcc -symbolic @gol
444 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
447 @item Directory Options
448 @xref{Directory Options,,Options for Directory Search}.
449 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
450 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
453 @item Machine Dependent Options
454 @xref{Submodel Options,,Hardware Models and Configurations}.
455 @c This list is ordered alphanumerically by subsection name.
456 @c Try and put the significant identifier (CPU or system) first,
457 @c so users have a clue at guessing where the ones they want will be.
460 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
461 -mabi=@var{name} @gol
462 -mapcs-stack-check -mno-apcs-stack-check @gol
463 -mapcs-float -mno-apcs-float @gol
464 -mapcs-reentrant -mno-apcs-reentrant @gol
465 -msched-prolog -mno-sched-prolog @gol
466 -mlittle-endian -mbig-endian -mwords-little-endian @gol
467 -mfloat-abi=@var{name} -mfpe @gol
468 -mfp16-format=@var{name}
469 -mthumb-interwork -mno-thumb-interwork @gol
470 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
471 -mstructure-size-boundary=@var{n} @gol
472 -mabort-on-noreturn @gol
473 -mlong-calls -mno-long-calls @gol
474 -msingle-pic-base -mno-single-pic-base @gol
475 -mpic-register=@var{reg} @gol
476 -mnop-fun-dllimport @gol
477 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
478 -mpoke-function-name @gol
480 -mtpcs-frame -mtpcs-leaf-frame @gol
481 -mcaller-super-interworking -mcallee-super-interworking @gol
482 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
483 -mword-relocations @gol
484 -mfix-cortex-m3-ldrd}
487 @gccoptlist{-mmcu=@var{mcu} -mno-interrupts @gol
488 -mcall-prologues -mtiny-stack -mint8}
490 @emph{Blackfin Options}
491 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
492 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
493 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
494 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
495 -mno-id-shared-library -mshared-library-id=@var{n} @gol
496 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
497 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
498 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
502 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
503 -msim -msdata=@var{sdata-type}}
506 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
507 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
508 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
509 -mstack-align -mdata-align -mconst-align @gol
510 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
511 -melf -maout -melinux -mlinux -sim -sim2 @gol
512 -mmul-bug-workaround -mno-mul-bug-workaround}
514 @emph{Darwin Options}
515 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
516 -arch_only -bind_at_load -bundle -bundle_loader @gol
517 -client_name -compatibility_version -current_version @gol
519 -dependency-file -dylib_file -dylinker_install_name @gol
520 -dynamic -dynamiclib -exported_symbols_list @gol
521 -filelist -flat_namespace -force_cpusubtype_ALL @gol
522 -force_flat_namespace -headerpad_max_install_names @gol
524 -image_base -init -install_name -keep_private_externs @gol
525 -multi_module -multiply_defined -multiply_defined_unused @gol
526 -noall_load -no_dead_strip_inits_and_terms @gol
527 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
528 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
529 -private_bundle -read_only_relocs -sectalign @gol
530 -sectobjectsymbols -whyload -seg1addr @gol
531 -sectcreate -sectobjectsymbols -sectorder @gol
532 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
533 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
534 -segprot -segs_read_only_addr -segs_read_write_addr @gol
535 -single_module -static -sub_library -sub_umbrella @gol
536 -twolevel_namespace -umbrella -undefined @gol
537 -unexported_symbols_list -weak_reference_mismatches @gol
538 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
539 -mkernel -mone-byte-bool}
541 @emph{DEC Alpha Options}
542 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
543 -mieee -mieee-with-inexact -mieee-conformant @gol
544 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
545 -mtrap-precision=@var{mode} -mbuild-constants @gol
546 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
547 -mbwx -mmax -mfix -mcix @gol
548 -mfloat-vax -mfloat-ieee @gol
549 -mexplicit-relocs -msmall-data -mlarge-data @gol
550 -msmall-text -mlarge-text @gol
551 -mmemory-latency=@var{time}}
553 @emph{DEC Alpha/VMS Options}
554 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
557 @gccoptlist{-msmall-model -mno-lsim}
560 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
561 -mhard-float -msoft-float @gol
562 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
563 -mdouble -mno-double @gol
564 -mmedia -mno-media -mmuladd -mno-muladd @gol
565 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
566 -mlinked-fp -mlong-calls -malign-labels @gol
567 -mlibrary-pic -macc-4 -macc-8 @gol
568 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
569 -moptimize-membar -mno-optimize-membar @gol
570 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
571 -mvliw-branch -mno-vliw-branch @gol
572 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
573 -mno-nested-cond-exec -mtomcat-stats @gol
577 @emph{GNU/Linux Options}
578 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
579 -tno-android-cc -tno-android-ld}
581 @emph{H8/300 Options}
582 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
585 @gccoptlist{-march=@var{architecture-type} @gol
586 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
587 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
588 -mfixed-range=@var{register-range} @gol
589 -mjump-in-delay -mlinker-opt -mlong-calls @gol
590 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
591 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
592 -mno-jump-in-delay -mno-long-load-store @gol
593 -mno-portable-runtime -mno-soft-float @gol
594 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
595 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
596 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
597 -munix=@var{unix-std} -nolibdld -static -threads}
599 @emph{i386 and x86-64 Options}
600 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
601 -mfpmath=@var{unit} @gol
602 -masm=@var{dialect} -mno-fancy-math-387 @gol
603 -mno-fp-ret-in-387 -msoft-float @gol
604 -mno-wide-multiply -mrtd -malign-double @gol
605 -mpreferred-stack-boundary=@var{num} @gol
606 -mincoming-stack-boundary=@var{num} @gol
607 -mcld -mcx16 -msahf -mmovbe -mcrc32 -mrecip -mvzeroupper @gol
608 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
609 -mavx2 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
610 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
611 -mbmi2 -mlwp -mthreads -mno-align-stringops -minline-all-stringops @gol
612 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
613 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
614 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
615 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
616 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
617 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
618 -mcmodel=@var{code-model} -mabi=@var{name} @gol
619 -m32 -m64 -mx32 -mlarge-data-threshold=@var{num} @gol
620 -msse2avx -mfentry -m8bit-idiv @gol
621 -mavx256-split-unaligned-load -mavx256-split-unaligned-store}
623 @emph{i386 and x86-64 Windows Options}
624 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
625 -mnop-fun-dllimport -mthread @gol
626 -municode -mwin32 -mwindows -fno-set-stack-executable}
629 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
630 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
631 -mconstant-gp -mauto-pic -mfused-madd @gol
632 -minline-float-divide-min-latency @gol
633 -minline-float-divide-max-throughput @gol
634 -mno-inline-float-divide @gol
635 -minline-int-divide-min-latency @gol
636 -minline-int-divide-max-throughput @gol
637 -mno-inline-int-divide @gol
638 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
639 -mno-inline-sqrt @gol
640 -mdwarf2-asm -mearly-stop-bits @gol
641 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
642 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
643 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
644 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
645 -msched-spec-ldc -msched-spec-control-ldc @gol
646 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
647 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
648 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
649 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
651 @emph{IA-64/VMS Options}
652 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64}
655 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
656 -msign-extend-enabled -muser-enabled}
658 @emph{M32R/D Options}
659 @gccoptlist{-m32r2 -m32rx -m32r @gol
661 -malign-loops -mno-align-loops @gol
662 -missue-rate=@var{number} @gol
663 -mbranch-cost=@var{number} @gol
664 -mmodel=@var{code-size-model-type} @gol
665 -msdata=@var{sdata-type} @gol
666 -mno-flush-func -mflush-func=@var{name} @gol
667 -mno-flush-trap -mflush-trap=@var{number} @gol
671 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
673 @emph{M680x0 Options}
674 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune}
675 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
676 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
677 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
678 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
679 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
680 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
681 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
685 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
686 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
687 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
688 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
689 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
692 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
693 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
694 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
695 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
698 @emph{MicroBlaze Options}
699 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
700 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
701 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
702 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
703 -mxl-mode-@var{app-model}}
706 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
707 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 @gol
708 -mips64 -mips64r2 @gol
709 -mips16 -mno-mips16 -mflip-mips16 @gol
710 -minterlink-mips16 -mno-interlink-mips16 @gol
711 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
712 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
713 -mgp32 -mgp64 -mfp32 -mfp64 -mhard-float -msoft-float @gol
714 -msingle-float -mdouble-float -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
715 -mfpu=@var{fpu-type} @gol
716 -msmartmips -mno-smartmips @gol
717 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
718 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
719 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
720 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
721 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
722 -membedded-data -mno-embedded-data @gol
723 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
724 -mcode-readable=@var{setting} @gol
725 -msplit-addresses -mno-split-addresses @gol
726 -mexplicit-relocs -mno-explicit-relocs @gol
727 -mcheck-zero-division -mno-check-zero-division @gol
728 -mdivide-traps -mdivide-breaks @gol
729 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
730 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
731 -mfix-24k -mno-fix-24k @gol
732 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
733 -mfix-r10000 -mno-fix-r10000 -mfix-vr4120 -mno-fix-vr4120 @gol
734 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
735 -mflush-func=@var{func} -mno-flush-func @gol
736 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
737 -mfp-exceptions -mno-fp-exceptions @gol
738 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
739 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
742 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
743 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
744 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
745 -mno-base-addresses -msingle-exit -mno-single-exit}
747 @emph{MN10300 Options}
748 @gccoptlist{-mmult-bug -mno-mult-bug @gol
749 -mno-am33 -mam33 -mam33-2 -mam34 @gol
750 -mtune=@var{cpu-type} @gol
751 -mreturn-pointer-on-d0 @gol
752 -mno-crt0 -mrelax -mliw -msetlb}
754 @emph{PDP-11 Options}
755 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
756 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
757 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
758 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
759 -mbranch-expensive -mbranch-cheap @gol
760 -munix-asm -mdec-asm}
762 @emph{picoChip Options}
763 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
764 -msymbol-as-address -mno-inefficient-warnings}
766 @emph{PowerPC Options}
767 See RS/6000 and PowerPC Options.
769 @emph{RS/6000 and PowerPC Options}
770 @gccoptlist{-mcpu=@var{cpu-type} @gol
771 -mtune=@var{cpu-type} @gol
772 -mcmodel=@var{code-model} @gol
773 -mpower -mno-power -mpower2 -mno-power2 @gol
774 -mpowerpc -mpowerpc64 -mno-powerpc @gol
775 -maltivec -mno-altivec @gol
776 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
777 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
778 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
779 -mfprnd -mno-fprnd @gol
780 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
781 -mnew-mnemonics -mold-mnemonics @gol
782 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
783 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
784 -malign-power -malign-natural @gol
785 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
786 -msingle-float -mdouble-float -msimple-fpu @gol
787 -mstring -mno-string -mupdate -mno-update @gol
788 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
789 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
790 -mstrict-align -mno-strict-align -mrelocatable @gol
791 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
792 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
793 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
794 -mprioritize-restricted-insns=@var{priority} @gol
795 -msched-costly-dep=@var{dependence_type} @gol
796 -minsert-sched-nops=@var{scheme} @gol
797 -mcall-sysv -mcall-netbsd @gol
798 -maix-struct-return -msvr4-struct-return @gol
799 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
800 -mblock-move-inline-limit=@var{num} @gol
801 -misel -mno-isel @gol
802 -misel=yes -misel=no @gol
804 -mspe=yes -mspe=no @gol
806 -mgen-cell-microcode -mwarn-cell-microcode @gol
807 -mvrsave -mno-vrsave @gol
808 -mmulhw -mno-mulhw @gol
809 -mdlmzb -mno-dlmzb @gol
810 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
811 -mprototype -mno-prototype @gol
812 -msim -mmvme -mads -myellowknife -memb -msdata @gol
813 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
814 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
815 -mno-recip-precision @gol
816 -mveclibabi=@var{type} -mfriz -mno-friz @gol
817 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
818 -msave-toc-indirect -mno-save-toc-indirect}
821 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
823 -mbig-endian-data -mlittle-endian-data @gol
826 -mas100-syntax -mno-as100-syntax@gol
828 -mmax-constant-size=@gol
830 -msave-acc-in-interrupts}
832 @emph{S/390 and zSeries Options}
833 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
834 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
835 -mlong-double-64 -mlong-double-128 @gol
836 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
837 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
838 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
839 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
840 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
843 @gccoptlist{-meb -mel @gol
847 -mscore5 -mscore5u -mscore7 -mscore7d}
850 @gccoptlist{-m1 -m2 -m2e @gol
851 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
853 -m4-nofpu -m4-single-only -m4-single -m4 @gol
854 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
855 -m5-64media -m5-64media-nofpu @gol
856 -m5-32media -m5-32media-nofpu @gol
857 -m5-compact -m5-compact-nofpu @gol
858 -mb -ml -mdalign -mrelax @gol
859 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
860 -mieee -mbitops -misize -minline-ic_invalidate -mpadstruct -mspace @gol
861 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
862 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
863 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
864 -maccumulate-outgoing-args -minvalid-symbols}
866 @emph{Solaris 2 Options}
867 @gccoptlist{-mimpure-text -mno-impure-text @gol
871 @gccoptlist{-mcpu=@var{cpu-type} @gol
872 -mtune=@var{cpu-type} @gol
873 -mcmodel=@var{code-model} @gol
874 -m32 -m64 -mapp-regs -mno-app-regs @gol
875 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
876 -mfpu -mno-fpu -mhard-float -msoft-float @gol
877 -mhard-quad-float -msoft-quad-float @gol
879 -mstack-bias -mno-stack-bias @gol
880 -munaligned-doubles -mno-unaligned-doubles @gol
881 -mv8plus -mno-v8plus -mvis -mno-vis}
884 @gccoptlist{-mwarn-reloc -merror-reloc @gol
885 -msafe-dma -munsafe-dma @gol
887 -msmall-mem -mlarge-mem -mstdmain @gol
888 -mfixed-range=@var{register-range} @gol
890 -maddress-space-conversion -mno-address-space-conversion @gol
891 -mcache-size=@var{cache-size} @gol
892 -matomic-updates -mno-atomic-updates}
894 @emph{System V Options}
895 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
898 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
899 -mprolog-function -mno-prolog-function -mspace @gol
900 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
901 -mapp-regs -mno-app-regs @gol
902 -mdisable-callt -mno-disable-callt @gol
905 -mv850e1 -mv850es @gol
910 @gccoptlist{-mg -mgnu -munix}
912 @emph{VxWorks Options}
913 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
914 -Xbind-lazy -Xbind-now}
916 @emph{x86-64 Options}
917 See i386 and x86-64 Options.
919 @emph{Xstormy16 Options}
922 @emph{Xtensa Options}
923 @gccoptlist{-mconst16 -mno-const16 @gol
924 -mfused-madd -mno-fused-madd @gol
926 -mserialize-volatile -mno-serialize-volatile @gol
927 -mtext-section-literals -mno-text-section-literals @gol
928 -mtarget-align -mno-target-align @gol
929 -mlongcalls -mno-longcalls}
931 @emph{zSeries Options}
932 See S/390 and zSeries Options.
934 @item Code Generation Options
935 @xref{Code Gen Options,,Options for Code Generation Conventions}.
936 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
937 -ffixed-@var{reg} -fexceptions @gol
938 -fnon-call-exceptions -funwind-tables @gol
939 -fasynchronous-unwind-tables @gol
940 -finhibit-size-directive -finstrument-functions @gol
941 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
942 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
943 -fno-common -fno-ident @gol
944 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
945 -fno-jump-tables @gol
946 -frecord-gcc-switches @gol
947 -freg-struct-return -fshort-enums @gol
948 -fshort-double -fshort-wchar @gol
949 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
950 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
951 -fno-stack-limit -fsplit-stack @gol
952 -fleading-underscore -ftls-model=@var{model} @gol
953 -ftrapv -fwrapv -fbounds-check @gol
954 -fvisibility -fstrict-volatile-bitfields}
958 * Overall Options:: Controlling the kind of output:
959 an executable, object files, assembler files,
960 or preprocessed source.
961 * C Dialect Options:: Controlling the variant of C language compiled.
962 * C++ Dialect Options:: Variations on C++.
963 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
965 * Language Independent Options:: Controlling how diagnostics should be
967 * Warning Options:: How picky should the compiler be?
968 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
969 * Optimize Options:: How much optimization?
970 * Preprocessor Options:: Controlling header files and macro definitions.
971 Also, getting dependency information for Make.
972 * Assembler Options:: Passing options to the assembler.
973 * Link Options:: Specifying libraries and so on.
974 * Directory Options:: Where to find header files and libraries.
975 Where to find the compiler executable files.
976 * Spec Files:: How to pass switches to sub-processes.
977 * Target Options:: Running a cross-compiler, or an old version of GCC.
980 @node Overall Options
981 @section Options Controlling the Kind of Output
983 Compilation can involve up to four stages: preprocessing, compilation
984 proper, assembly and linking, always in that order. GCC is capable of
985 preprocessing and compiling several files either into several
986 assembler input files, or into one assembler input file; then each
987 assembler input file produces an object file, and linking combines all
988 the object files (those newly compiled, and those specified as input)
989 into an executable file.
991 @cindex file name suffix
992 For any given input file, the file name suffix determines what kind of
997 C source code which must be preprocessed.
1000 C source code which should not be preprocessed.
1003 C++ source code which should not be preprocessed.
1006 Objective-C source code. Note that you must link with the @file{libobjc}
1007 library to make an Objective-C program work.
1010 Objective-C source code which should not be preprocessed.
1014 Objective-C++ source code. Note that you must link with the @file{libobjc}
1015 library to make an Objective-C++ program work. Note that @samp{.M} refers
1016 to a literal capital M@.
1018 @item @var{file}.mii
1019 Objective-C++ source code which should not be preprocessed.
1022 C, C++, Objective-C or Objective-C++ header file to be turned into a
1023 precompiled header (default), or C, C++ header file to be turned into an
1024 Ada spec (via the @option{-fdump-ada-spec} switch).
1027 @itemx @var{file}.cp
1028 @itemx @var{file}.cxx
1029 @itemx @var{file}.cpp
1030 @itemx @var{file}.CPP
1031 @itemx @var{file}.c++
1033 C++ source code which must be preprocessed. Note that in @samp{.cxx},
1034 the last two letters must both be literally @samp{x}. Likewise,
1035 @samp{.C} refers to a literal capital C@.
1039 Objective-C++ source code which must be preprocessed.
1041 @item @var{file}.mii
1042 Objective-C++ source code which should not be preprocessed.
1046 @itemx @var{file}.hp
1047 @itemx @var{file}.hxx
1048 @itemx @var{file}.hpp
1049 @itemx @var{file}.HPP
1050 @itemx @var{file}.h++
1051 @itemx @var{file}.tcc
1052 C++ header file to be turned into a precompiled header or Ada spec.
1055 @itemx @var{file}.for
1056 @itemx @var{file}.ftn
1057 Fixed form Fortran source code which should not be preprocessed.
1060 @itemx @var{file}.FOR
1061 @itemx @var{file}.fpp
1062 @itemx @var{file}.FPP
1063 @itemx @var{file}.FTN
1064 Fixed form Fortran source code which must be preprocessed (with the traditional
1067 @item @var{file}.f90
1068 @itemx @var{file}.f95
1069 @itemx @var{file}.f03
1070 @itemx @var{file}.f08
1071 Free form Fortran source code which should not be preprocessed.
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 must be preprocessed (with the
1078 traditional preprocessor).
1083 @c FIXME: Descriptions of Java file types.
1089 @item @var{file}.ads
1090 Ada source code file which contains a library unit declaration (a
1091 declaration of a package, subprogram, or generic, or a generic
1092 instantiation), or a library unit renaming declaration (a package,
1093 generic, or subprogram renaming declaration). Such files are also
1096 @item @var{file}.adb
1097 Ada source code file containing a library unit body (a subprogram or
1098 package body). Such files are also called @dfn{bodies}.
1100 @c GCC also knows about some suffixes for languages not yet included:
1111 @itemx @var{file}.sx
1112 Assembler code which must be preprocessed.
1115 An object file to be fed straight into linking.
1116 Any file name with no recognized suffix is treated this way.
1120 You can specify the input language explicitly with the @option{-x} option:
1123 @item -x @var{language}
1124 Specify explicitly the @var{language} for the following input files
1125 (rather than letting the compiler choose a default based on the file
1126 name suffix). This option applies to all following input files until
1127 the next @option{-x} option. Possible values for @var{language} are:
1129 c c-header cpp-output
1130 c++ c++-header c++-cpp-output
1131 objective-c objective-c-header objective-c-cpp-output
1132 objective-c++ objective-c++-header objective-c++-cpp-output
1133 assembler assembler-with-cpp
1135 f77 f77-cpp-input f95 f95-cpp-input
1141 Turn off any specification of a language, so that subsequent files are
1142 handled according to their file name suffixes (as they are if @option{-x}
1143 has not been used at all).
1145 @item -pass-exit-codes
1146 @opindex pass-exit-codes
1147 Normally the @command{gcc} program will exit with the code of 1 if any
1148 phase of the compiler returns a non-success return code. If you specify
1149 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
1150 numerically highest error produced by any phase that returned an error
1151 indication. The C, C++, and Fortran frontends return 4, if an internal
1152 compiler error is encountered.
1155 If you only want some of the stages of compilation, you can use
1156 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1157 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1158 @command{gcc} is to stop. Note that some combinations (for example,
1159 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1164 Compile or assemble the source files, but do not link. The linking
1165 stage simply is not done. The ultimate output is in the form of an
1166 object file for each source file.
1168 By default, the object file name for a source file is made by replacing
1169 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1171 Unrecognized input files, not requiring compilation or assembly, are
1176 Stop after the stage of compilation proper; do not assemble. The output
1177 is in the form of an assembler code file for each non-assembler input
1180 By default, the assembler file name for a source file is made by
1181 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1183 Input files that don't require compilation are ignored.
1187 Stop after the preprocessing stage; do not run the compiler proper. The
1188 output is in the form of preprocessed source code, which is sent to the
1191 Input files which don't require preprocessing are ignored.
1193 @cindex output file option
1196 Place output in file @var{file}. This applies regardless to whatever
1197 sort of output is being produced, whether it be an executable file,
1198 an object file, an assembler file or preprocessed C code.
1200 If @option{-o} is not specified, the default is to put an executable
1201 file in @file{a.out}, the object file for
1202 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1203 assembler file in @file{@var{source}.s}, a precompiled header file in
1204 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1209 Print (on standard error output) the commands executed to run the stages
1210 of compilation. Also print the version number of the compiler driver
1211 program and of the preprocessor and the compiler proper.
1215 Like @option{-v} except the commands are not executed and arguments
1216 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1217 This is useful for shell scripts to capture the driver-generated command lines.
1221 Use pipes rather than temporary files for communication between the
1222 various stages of compilation. This fails to work on some systems where
1223 the assembler is unable to read from a pipe; but the GNU assembler has
1228 Print (on the standard output) a description of the command line options
1229 understood by @command{gcc}. If the @option{-v} option is also specified
1230 then @option{--help} will also be passed on to the various processes
1231 invoked by @command{gcc}, so that they can display the command line options
1232 they accept. If the @option{-Wextra} option has also been specified
1233 (prior to the @option{--help} option), then command line options which
1234 have no documentation associated with them will also be displayed.
1237 @opindex target-help
1238 Print (on the standard output) a description of target-specific command
1239 line options for each tool. For some targets extra target-specific
1240 information may also be printed.
1242 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1243 Print (on the standard output) a description of the command line
1244 options understood by the compiler that fit into all specified classes
1245 and qualifiers. These are the supported classes:
1248 @item @samp{optimizers}
1249 This will display all of the optimization options supported by the
1252 @item @samp{warnings}
1253 This will display all of the options controlling warning messages
1254 produced by the compiler.
1257 This will display target-specific options. Unlike the
1258 @option{--target-help} option however, target-specific options of the
1259 linker and assembler will not be displayed. This is because those
1260 tools do not currently support the extended @option{--help=} syntax.
1263 This will display the values recognized by the @option{--param}
1266 @item @var{language}
1267 This will display the options supported for @var{language}, where
1268 @var{language} is the name of one of the languages supported in this
1272 This will display the options that are common to all languages.
1275 These are the supported qualifiers:
1278 @item @samp{undocumented}
1279 Display only those options which are undocumented.
1282 Display options which take an argument that appears after an equal
1283 sign in the same continuous piece of text, such as:
1284 @samp{--help=target}.
1286 @item @samp{separate}
1287 Display options which take an argument that appears as a separate word
1288 following the original option, such as: @samp{-o output-file}.
1291 Thus for example to display all the undocumented target-specific
1292 switches supported by the compiler the following can be used:
1295 --help=target,undocumented
1298 The sense of a qualifier can be inverted by prefixing it with the
1299 @samp{^} character, so for example to display all binary warning
1300 options (i.e., ones that are either on or off and that do not take an
1301 argument), which have a description the following can be used:
1304 --help=warnings,^joined,^undocumented
1307 The argument to @option{--help=} should not consist solely of inverted
1310 Combining several classes is possible, although this usually
1311 restricts the output by so much that there is nothing to display. One
1312 case where it does work however is when one of the classes is
1313 @var{target}. So for example to display all the target-specific
1314 optimization options the following can be used:
1317 --help=target,optimizers
1320 The @option{--help=} option can be repeated on the command line. Each
1321 successive use will display its requested class of options, skipping
1322 those that have already been displayed.
1324 If the @option{-Q} option appears on the command line before the
1325 @option{--help=} option, then the descriptive text displayed by
1326 @option{--help=} is changed. Instead of describing the displayed
1327 options, an indication is given as to whether the option is enabled,
1328 disabled or set to a specific value (assuming that the compiler
1329 knows this at the point where the @option{--help=} option is used).
1331 Here is a truncated example from the ARM port of @command{gcc}:
1334 % gcc -Q -mabi=2 --help=target -c
1335 The following options are target specific:
1337 -mabort-on-noreturn [disabled]
1341 The output is sensitive to the effects of previous command line
1342 options, so for example it is possible to find out which optimizations
1343 are enabled at @option{-O2} by using:
1346 -Q -O2 --help=optimizers
1349 Alternatively you can discover which binary optimizations are enabled
1350 by @option{-O3} by using:
1353 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1354 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1355 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1358 @item -no-canonical-prefixes
1359 @opindex no-canonical-prefixes
1360 Do not expand any symbolic links, resolve references to @samp{/../}
1361 or @samp{/./}, or make the path absolute when generating a relative
1366 Display the version number and copyrights of the invoked GCC@.
1370 Invoke all subcommands under a wrapper program. The name of the
1371 wrapper program and its parameters are passed as a comma separated
1375 gcc -c t.c -wrapper gdb,--args
1378 This will invoke all subprograms of @command{gcc} under
1379 @samp{gdb --args}, thus the invocation of @command{cc1} will be
1380 @samp{gdb --args cc1 @dots{}}.
1382 @item -fplugin=@var{name}.so
1383 Load the plugin code in file @var{name}.so, assumed to be a
1384 shared object to be dlopen'd by the compiler. The base name of
1385 the shared object file is used to identify the plugin for the
1386 purposes of argument parsing (See
1387 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1388 Each plugin should define the callback functions specified in the
1391 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1392 Define an argument called @var{key} with a value of @var{value}
1393 for the plugin called @var{name}.
1395 @item -fdump-ada-spec@r{[}-slim@r{]}
1396 For C and C++ source and include files, generate corresponding Ada
1397 specs. @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1398 GNAT User's Guide}, which provides detailed documentation on this feature.
1400 @item -fdump-go-spec=@var{file}
1401 For input files in any language, generate corresponding Go
1402 declarations in @var{file}. This generates Go @code{const},
1403 @code{type}, @code{var}, and @code{func} declarations which may be a
1404 useful way to start writing a Go interface to code written in some
1407 @include @value{srcdir}/../libiberty/at-file.texi
1411 @section Compiling C++ Programs
1413 @cindex suffixes for C++ source
1414 @cindex C++ source file suffixes
1415 C++ source files conventionally use one of the suffixes @samp{.C},
1416 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1417 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1418 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1419 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1420 files with these names and compiles them as C++ programs even if you
1421 call the compiler the same way as for compiling C programs (usually
1422 with the name @command{gcc}).
1426 However, the use of @command{gcc} does not add the C++ library.
1427 @command{g++} is a program that calls GCC and treats @samp{.c},
1428 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1429 files unless @option{-x} is used, and automatically specifies linking
1430 against the C++ library. This program is also useful when
1431 precompiling a C header file with a @samp{.h} extension for use in C++
1432 compilations. On many systems, @command{g++} is also installed with
1433 the name @command{c++}.
1435 @cindex invoking @command{g++}
1436 When you compile C++ programs, you may specify many of the same
1437 command-line options that you use for compiling programs in any
1438 language; or command-line options meaningful for C and related
1439 languages; or options that are meaningful only for C++ programs.
1440 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1441 explanations of options for languages related to C@.
1442 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1443 explanations of options that are meaningful only for C++ programs.
1445 @node C Dialect Options
1446 @section Options Controlling C Dialect
1447 @cindex dialect options
1448 @cindex language dialect options
1449 @cindex options, dialect
1451 The following options control the dialect of C (or languages derived
1452 from C, such as C++, Objective-C and Objective-C++) that the compiler
1456 @cindex ANSI support
1460 In C mode, this is equivalent to @samp{-std=c90}. In C++ mode, it is
1461 equivalent to @samp{-std=c++98}.
1463 This turns off certain features of GCC that are incompatible with ISO
1464 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1465 such as the @code{asm} and @code{typeof} keywords, and
1466 predefined macros such as @code{unix} and @code{vax} that identify the
1467 type of system you are using. It also enables the undesirable and
1468 rarely used ISO trigraph feature. For the C compiler,
1469 it disables recognition of C++ style @samp{//} comments as well as
1470 the @code{inline} keyword.
1472 The alternate keywords @code{__asm__}, @code{__extension__},
1473 @code{__inline__} and @code{__typeof__} continue to work despite
1474 @option{-ansi}. You would not want to use them in an ISO C program, of
1475 course, but it is useful to put them in header files that might be included
1476 in compilations done with @option{-ansi}. Alternate predefined macros
1477 such as @code{__unix__} and @code{__vax__} are also available, with or
1478 without @option{-ansi}.
1480 The @option{-ansi} option does not cause non-ISO programs to be
1481 rejected gratuitously. For that, @option{-pedantic} is required in
1482 addition to @option{-ansi}. @xref{Warning Options}.
1484 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1485 option is used. Some header files may notice this macro and refrain
1486 from declaring certain functions or defining certain macros that the
1487 ISO standard doesn't call for; this is to avoid interfering with any
1488 programs that might use these names for other things.
1490 Functions that would normally be built in but do not have semantics
1491 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1492 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1493 built-in functions provided by GCC}, for details of the functions
1498 Determine the language standard. @xref{Standards,,Language Standards
1499 Supported by GCC}, for details of these standard versions. This option
1500 is currently only supported when compiling C or C++.
1502 The compiler can accept several base standards, such as @samp{c90} or
1503 @samp{c++98}, and GNU dialects of those standards, such as
1504 @samp{gnu90} or @samp{gnu++98}. By specifying a base standard, the
1505 compiler will accept all programs following that standard and those
1506 using GNU extensions that do not contradict it. For example,
1507 @samp{-std=c90} turns off certain features of GCC that are
1508 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1509 keywords, but not other GNU extensions that do not have a meaning in
1510 ISO C90, such as omitting the middle term of a @code{?:}
1511 expression. On the other hand, by specifying a GNU dialect of a
1512 standard, all features the compiler support are enabled, even when
1513 those features change the meaning of the base standard and some
1514 strict-conforming programs may be rejected. The particular standard
1515 is used by @option{-pedantic} to identify which features are GNU
1516 extensions given that version of the standard. For example
1517 @samp{-std=gnu90 -pedantic} would warn about C++ style @samp{//}
1518 comments, while @samp{-std=gnu99 -pedantic} would not.
1520 A value for this option must be provided; possible values are
1526 Support all ISO C90 programs (certain GNU extensions that conflict
1527 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1529 @item iso9899:199409
1530 ISO C90 as modified in amendment 1.
1536 ISO C99. Note that this standard is not yet fully supported; see
1537 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1538 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1541 ISO C1X, the draft of the next revision of the ISO C standard.
1542 Support is limited and experimental and features enabled by this
1543 option may be changed or removed if changed in or removed from the
1548 GNU dialect of ISO C90 (including some C99 features). This
1549 is the default for C code.
1553 GNU dialect of ISO C99. When ISO C99 is fully implemented in GCC,
1554 this will become the default. The name @samp{gnu9x} is deprecated.
1557 GNU dialect of ISO C1X. Support is limited and experimental and
1558 features enabled by this option may be changed or removed if changed
1559 in or removed from the standard draft.
1562 The 1998 ISO C++ standard plus amendments. Same as @option{-ansi} for
1566 GNU dialect of @option{-std=c++98}. This is the default for
1570 The working draft of the upcoming ISO C++0x standard. This option
1571 enables experimental features that are likely to be included in
1572 C++0x. The working draft is constantly changing, and any feature that is
1573 enabled by this flag may be removed from future versions of GCC if it is
1574 not part of the C++0x standard.
1577 GNU dialect of @option{-std=c++0x}. This option enables
1578 experimental features that may be removed in future versions of GCC.
1581 @item -fgnu89-inline
1582 @opindex fgnu89-inline
1583 The option @option{-fgnu89-inline} tells GCC to use the traditional
1584 GNU semantics for @code{inline} functions when in C99 mode.
1585 @xref{Inline,,An Inline Function is As Fast As a Macro}. This option
1586 is accepted and ignored by GCC versions 4.1.3 up to but not including
1587 4.3. In GCC versions 4.3 and later it changes the behavior of GCC in
1588 C99 mode. Using this option is roughly equivalent to adding the
1589 @code{gnu_inline} function attribute to all inline functions
1590 (@pxref{Function Attributes}).
1592 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1593 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1594 specifies the default behavior). This option was first supported in
1595 GCC 4.3. This option is not supported in @option{-std=c90} or
1596 @option{-std=gnu90} mode.
1598 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1599 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1600 in effect for @code{inline} functions. @xref{Common Predefined
1601 Macros,,,cpp,The C Preprocessor}.
1603 @item -aux-info @var{filename}
1605 Output to the given filename prototyped declarations for all functions
1606 declared and/or defined in a translation unit, including those in header
1607 files. This option is silently ignored in any language other than C@.
1609 Besides declarations, the file indicates, in comments, the origin of
1610 each declaration (source file and line), whether the declaration was
1611 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1612 @samp{O} for old, respectively, in the first character after the line
1613 number and the colon), and whether it came from a declaration or a
1614 definition (@samp{C} or @samp{F}, respectively, in the following
1615 character). In the case of function definitions, a K&R-style list of
1616 arguments followed by their declarations is also provided, inside
1617 comments, after the declaration.
1621 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1622 keyword, so that code can use these words as identifiers. You can use
1623 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1624 instead. @option{-ansi} implies @option{-fno-asm}.
1626 In C++, this switch only affects the @code{typeof} keyword, since
1627 @code{asm} and @code{inline} are standard keywords. You may want to
1628 use the @option{-fno-gnu-keywords} flag instead, which has the same
1629 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1630 switch only affects the @code{asm} and @code{typeof} keywords, since
1631 @code{inline} is a standard keyword in ISO C99.
1634 @itemx -fno-builtin-@var{function}
1635 @opindex fno-builtin
1636 @cindex built-in functions
1637 Don't recognize built-in functions that do not begin with
1638 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1639 functions provided by GCC}, for details of the functions affected,
1640 including those which are not built-in functions when @option{-ansi} or
1641 @option{-std} options for strict ISO C conformance are used because they
1642 do not have an ISO standard meaning.
1644 GCC normally generates special code to handle certain built-in functions
1645 more efficiently; for instance, calls to @code{alloca} may become single
1646 instructions that adjust the stack directly, and calls to @code{memcpy}
1647 may become inline copy loops. The resulting code is often both smaller
1648 and faster, but since the function calls no longer appear as such, you
1649 cannot set a breakpoint on those calls, nor can you change the behavior
1650 of the functions by linking with a different library. In addition,
1651 when a function is recognized as a built-in function, GCC may use
1652 information about that function to warn about problems with calls to
1653 that function, or to generate more efficient code, even if the
1654 resulting code still contains calls to that function. For example,
1655 warnings are given with @option{-Wformat} for bad calls to
1656 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1657 known not to modify global memory.
1659 With the @option{-fno-builtin-@var{function}} option
1660 only the built-in function @var{function} is
1661 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1662 function is named that is not built-in in this version of GCC, this
1663 option is ignored. There is no corresponding
1664 @option{-fbuiltin-@var{function}} option; if you wish to enable
1665 built-in functions selectively when using @option{-fno-builtin} or
1666 @option{-ffreestanding}, you may define macros such as:
1669 #define abs(n) __builtin_abs ((n))
1670 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1675 @cindex hosted environment
1677 Assert that compilation takes place in a hosted environment. This implies
1678 @option{-fbuiltin}. A hosted environment is one in which the
1679 entire standard library is available, and in which @code{main} has a return
1680 type of @code{int}. Examples are nearly everything except a kernel.
1681 This is equivalent to @option{-fno-freestanding}.
1683 @item -ffreestanding
1684 @opindex ffreestanding
1685 @cindex hosted environment
1687 Assert that compilation takes place in a freestanding environment. This
1688 implies @option{-fno-builtin}. A freestanding environment
1689 is one in which the standard library may not exist, and program startup may
1690 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1691 This is equivalent to @option{-fno-hosted}.
1693 @xref{Standards,,Language Standards Supported by GCC}, for details of
1694 freestanding and hosted environments.
1698 @cindex OpenMP parallel
1699 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1700 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1701 compiler generates parallel code according to the OpenMP Application
1702 Program Interface v3.0 @w{@uref{http://www.openmp.org/}}. This option
1703 implies @option{-pthread}, and thus is only supported on targets that
1704 have support for @option{-pthread}.
1706 @item -fms-extensions
1707 @opindex fms-extensions
1708 Accept some non-standard constructs used in Microsoft header files.
1710 In C++ code, this allows member names in structures to be similar
1711 to previous types declarations.
1720 Some cases of unnamed fields in structures and unions are only
1721 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1722 fields within structs/unions}, for details.
1724 @item -fplan9-extensions
1725 Accept some non-standard constructs used in Plan 9 code.
1727 This enables @option{-fms-extensions}, permits passing pointers to
1728 structures with anonymous fields to functions which expect pointers to
1729 elements of the type of the field, and permits referring to anonymous
1730 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1731 struct/union fields within structs/unions}, for details. This is only
1732 supported for C, not C++.
1736 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1737 options for strict ISO C conformance) implies @option{-trigraphs}.
1739 @item -no-integrated-cpp
1740 @opindex no-integrated-cpp
1741 Performs a compilation in two passes: preprocessing and compiling. This
1742 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1743 @option{-B} option. The user supplied compilation step can then add in
1744 an additional preprocessing step after normal preprocessing but before
1745 compiling. The default is to use the integrated cpp (internal cpp)
1747 The semantics of this option will change if "cc1", "cc1plus", and
1748 "cc1obj" are merged.
1750 @cindex traditional C language
1751 @cindex C language, traditional
1753 @itemx -traditional-cpp
1754 @opindex traditional-cpp
1755 @opindex traditional
1756 Formerly, these options caused GCC to attempt to emulate a pre-standard
1757 C compiler. They are now only supported with the @option{-E} switch.
1758 The preprocessor continues to support a pre-standard mode. See the GNU
1759 CPP manual for details.
1761 @item -fcond-mismatch
1762 @opindex fcond-mismatch
1763 Allow conditional expressions with mismatched types in the second and
1764 third arguments. The value of such an expression is void. This option
1765 is not supported for C++.
1767 @item -flax-vector-conversions
1768 @opindex flax-vector-conversions
1769 Allow implicit conversions between vectors with differing numbers of
1770 elements and/or incompatible element types. This option should not be
1773 @item -funsigned-char
1774 @opindex funsigned-char
1775 Let the type @code{char} be unsigned, like @code{unsigned char}.
1777 Each kind of machine has a default for what @code{char} should
1778 be. It is either like @code{unsigned char} by default or like
1779 @code{signed char} by default.
1781 Ideally, a portable program should always use @code{signed char} or
1782 @code{unsigned char} when it depends on the signedness of an object.
1783 But many programs have been written to use plain @code{char} and
1784 expect it to be signed, or expect it to be unsigned, depending on the
1785 machines they were written for. This option, and its inverse, let you
1786 make such a program work with the opposite default.
1788 The type @code{char} is always a distinct type from each of
1789 @code{signed char} or @code{unsigned char}, even though its behavior
1790 is always just like one of those two.
1793 @opindex fsigned-char
1794 Let the type @code{char} be signed, like @code{signed char}.
1796 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1797 the negative form of @option{-funsigned-char}. Likewise, the option
1798 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1800 @item -fsigned-bitfields
1801 @itemx -funsigned-bitfields
1802 @itemx -fno-signed-bitfields
1803 @itemx -fno-unsigned-bitfields
1804 @opindex fsigned-bitfields
1805 @opindex funsigned-bitfields
1806 @opindex fno-signed-bitfields
1807 @opindex fno-unsigned-bitfields
1808 These options control whether a bit-field is signed or unsigned, when the
1809 declaration does not use either @code{signed} or @code{unsigned}. By
1810 default, such a bit-field is signed, because this is consistent: the
1811 basic integer types such as @code{int} are signed types.
1814 @node C++ Dialect Options
1815 @section Options Controlling C++ Dialect
1817 @cindex compiler options, C++
1818 @cindex C++ options, command line
1819 @cindex options, C++
1820 This section describes the command-line options that are only meaningful
1821 for C++ programs; but you can also use most of the GNU compiler options
1822 regardless of what language your program is in. For example, you
1823 might compile a file @code{firstClass.C} like this:
1826 g++ -g -frepo -O -c firstClass.C
1830 In this example, only @option{-frepo} is an option meant
1831 only for C++ programs; you can use the other options with any
1832 language supported by GCC@.
1834 Here is a list of options that are @emph{only} for compiling C++ programs:
1838 @item -fabi-version=@var{n}
1839 @opindex fabi-version
1840 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1841 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1842 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1843 the version that conforms most closely to the C++ ABI specification.
1844 Therefore, the ABI obtained using version 0 will change as ABI bugs
1847 The default is version 2.
1849 Version 3 corrects an error in mangling a constant address as a
1852 Version 4 implements a standard mangling for vector types.
1854 Version 5 corrects the mangling of attribute const/volatile on
1855 function pointer types, decltype of a plain decl, and use of a
1856 function parameter in the declaration of another parameter.
1858 Version 6 corrects the promotion behavior of C++0x scoped enums.
1860 See also @option{-Wabi}.
1862 @item -fno-access-control
1863 @opindex fno-access-control
1864 Turn off all access checking. This switch is mainly useful for working
1865 around bugs in the access control code.
1869 Check that the pointer returned by @code{operator new} is non-null
1870 before attempting to modify the storage allocated. This check is
1871 normally unnecessary because the C++ standard specifies that
1872 @code{operator new} will only return @code{0} if it is declared
1873 @samp{throw()}, in which case the compiler will always check the
1874 return value even without this option. In all other cases, when
1875 @code{operator new} has a non-empty exception specification, memory
1876 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1877 @samp{new (nothrow)}.
1879 @item -fconserve-space
1880 @opindex fconserve-space
1881 Put uninitialized or runtime-initialized global variables into the
1882 common segment, as C does. This saves space in the executable at the
1883 cost of not diagnosing duplicate definitions. If you compile with this
1884 flag and your program mysteriously crashes after @code{main()} has
1885 completed, you may have an object that is being destroyed twice because
1886 two definitions were merged.
1888 This option is no longer useful on most targets, now that support has
1889 been added for putting variables into BSS without making them common.
1891 @item -fconstexpr-depth=@var{n}
1892 @opindex fconstexpr-depth
1893 Set the maximum nested evaluation depth for C++0x constexpr functions
1894 to @var{n}. A limit is needed to detect endless recursion during
1895 constant expression evaluation. The minimum specified by the standard
1898 @item -fno-deduce-init-list
1899 @opindex fno-deduce-init-list
1900 Disable deduction of a template type parameter as
1901 std::initializer_list from a brace-enclosed initializer list, i.e.
1904 template <class T> auto forward(T t) -> decltype (realfn (t))
1911 forward(@{1,2@}); // call forward<std::initializer_list<int>>
1915 This option is present because this deduction is an extension to the
1916 current specification in the C++0x working draft, and there was
1917 some concern about potential overload resolution problems.
1919 @item -ffriend-injection
1920 @opindex ffriend-injection
1921 Inject friend functions into the enclosing namespace, so that they are
1922 visible outside the scope of the class in which they are declared.
1923 Friend functions were documented to work this way in the old Annotated
1924 C++ Reference Manual, and versions of G++ before 4.1 always worked
1925 that way. However, in ISO C++ a friend function which is not declared
1926 in an enclosing scope can only be found using argument dependent
1927 lookup. This option causes friends to be injected as they were in
1930 This option is for compatibility, and may be removed in a future
1933 @item -fno-elide-constructors
1934 @opindex fno-elide-constructors
1935 The C++ standard allows an implementation to omit creating a temporary
1936 which is only used to initialize another object of the same type.
1937 Specifying this option disables that optimization, and forces G++ to
1938 call the copy constructor in all cases.
1940 @item -fno-enforce-eh-specs
1941 @opindex fno-enforce-eh-specs
1942 Don't generate code to check for violation of exception specifications
1943 at runtime. This option violates the C++ standard, but may be useful
1944 for reducing code size in production builds, much like defining
1945 @samp{NDEBUG}. This does not give user code permission to throw
1946 exceptions in violation of the exception specifications; the compiler
1947 will still optimize based on the specifications, so throwing an
1948 unexpected exception will result in undefined behavior.
1951 @itemx -fno-for-scope
1953 @opindex fno-for-scope
1954 If @option{-ffor-scope} is specified, the scope of variables declared in
1955 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1956 as specified by the C++ standard.
1957 If @option{-fno-for-scope} is specified, the scope of variables declared in
1958 a @i{for-init-statement} extends to the end of the enclosing scope,
1959 as was the case in old versions of G++, and other (traditional)
1960 implementations of C++.
1962 The default if neither flag is given to follow the standard,
1963 but to allow and give a warning for old-style code that would
1964 otherwise be invalid, or have different behavior.
1966 @item -fno-gnu-keywords
1967 @opindex fno-gnu-keywords
1968 Do not recognize @code{typeof} as a keyword, so that code can use this
1969 word as an identifier. You can use the keyword @code{__typeof__} instead.
1970 @option{-ansi} implies @option{-fno-gnu-keywords}.
1972 @item -fno-implicit-templates
1973 @opindex fno-implicit-templates
1974 Never emit code for non-inline templates which are instantiated
1975 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1976 @xref{Template Instantiation}, for more information.
1978 @item -fno-implicit-inline-templates
1979 @opindex fno-implicit-inline-templates
1980 Don't emit code for implicit instantiations of inline templates, either.
1981 The default is to handle inlines differently so that compiles with and
1982 without optimization will need the same set of explicit instantiations.
1984 @item -fno-implement-inlines
1985 @opindex fno-implement-inlines
1986 To save space, do not emit out-of-line copies of inline functions
1987 controlled by @samp{#pragma implementation}. This will cause linker
1988 errors if these functions are not inlined everywhere they are called.
1990 @item -fms-extensions
1991 @opindex fms-extensions
1992 Disable pedantic warnings about constructs used in MFC, such as implicit
1993 int and getting a pointer to member function via non-standard syntax.
1995 @item -fno-nonansi-builtins
1996 @opindex fno-nonansi-builtins
1997 Disable built-in declarations of functions that are not mandated by
1998 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1999 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2002 @opindex fnothrow-opt
2003 Treat a @code{throw()} exception specification as though it were a
2004 @code{noexcept} specification to reduce or eliminate the text size
2005 overhead relative to a function with no exception specification. If
2006 the function has local variables of types with non-trivial
2007 destructors, the exception specification will actually make the
2008 function smaller because the EH cleanups for those variables can be
2009 optimized away. The semantic effect is that an exception thrown out of
2010 a function with such an exception specification will result in a call
2011 to @code{terminate} rather than @code{unexpected}.
2013 @item -fno-operator-names
2014 @opindex fno-operator-names
2015 Do not treat the operator name keywords @code{and}, @code{bitand},
2016 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2017 synonyms as keywords.
2019 @item -fno-optional-diags
2020 @opindex fno-optional-diags
2021 Disable diagnostics that the standard says a compiler does not need to
2022 issue. Currently, the only such diagnostic issued by G++ is the one for
2023 a name having multiple meanings within a class.
2026 @opindex fpermissive
2027 Downgrade some diagnostics about nonconformant code from errors to
2028 warnings. Thus, using @option{-fpermissive} will allow some
2029 nonconforming code to compile.
2031 @item -fno-pretty-templates
2032 @opindex fno-pretty-templates
2033 When an error message refers to a specialization of a function
2034 template, the compiler will normally print the signature of the
2035 template followed by the template arguments and any typedefs or
2036 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2037 rather than @code{void f(int)}) so that it's clear which template is
2038 involved. When an error message refers to a specialization of a class
2039 template, the compiler will omit any template arguments which match
2040 the default template arguments for that template. If either of these
2041 behaviors make it harder to understand the error message rather than
2042 easier, using @option{-fno-pretty-templates} will disable them.
2046 Enable automatic template instantiation at link time. This option also
2047 implies @option{-fno-implicit-templates}. @xref{Template
2048 Instantiation}, for more information.
2052 Disable generation of information about every class with virtual
2053 functions for use by the C++ runtime type identification features
2054 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
2055 of the language, you can save some space by using this flag. Note that
2056 exception handling uses the same information, but it will generate it as
2057 needed. The @samp{dynamic_cast} operator can still be used for casts that
2058 do not require runtime type information, i.e.@: casts to @code{void *} or to
2059 unambiguous base classes.
2063 Emit statistics about front-end processing at the end of the compilation.
2064 This information is generally only useful to the G++ development team.
2066 @item -fstrict-enums
2067 @opindex fstrict-enums
2068 Allow the compiler to optimize using the assumption that a value of
2069 enumeration type can only be one of the values of the enumeration (as
2070 defined in the C++ standard; basically, a value which can be
2071 represented in the minimum number of bits needed to represent all the
2072 enumerators). This assumption may not be valid if the program uses a
2073 cast to convert an arbitrary integer value to the enumeration type.
2075 @item -ftemplate-depth=@var{n}
2076 @opindex ftemplate-depth
2077 Set the maximum instantiation depth for template classes to @var{n}.
2078 A limit on the template instantiation depth is needed to detect
2079 endless recursions during template class instantiation. ANSI/ISO C++
2080 conforming programs must not rely on a maximum depth greater than 17
2081 (changed to 1024 in C++0x). The default value is 900, as the compiler
2082 can run out of stack space before hitting 1024 in some situations.
2084 @item -fno-threadsafe-statics
2085 @opindex fno-threadsafe-statics
2086 Do not emit the extra code to use the routines specified in the C++
2087 ABI for thread-safe initialization of local statics. You can use this
2088 option to reduce code size slightly in code that doesn't need to be
2091 @item -fuse-cxa-atexit
2092 @opindex fuse-cxa-atexit
2093 Register destructors for objects with static storage duration with the
2094 @code{__cxa_atexit} function rather than the @code{atexit} function.
2095 This option is required for fully standards-compliant handling of static
2096 destructors, but will only work if your C library supports
2097 @code{__cxa_atexit}.
2099 @item -fno-use-cxa-get-exception-ptr
2100 @opindex fno-use-cxa-get-exception-ptr
2101 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2102 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
2103 if the runtime routine is not available.
2105 @item -fvisibility-inlines-hidden
2106 @opindex fvisibility-inlines-hidden
2107 This switch declares that the user does not attempt to compare
2108 pointers to inline methods where the addresses of the two functions
2109 were taken in different shared objects.
2111 The effect of this is that GCC may, effectively, mark inline methods with
2112 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2113 appear in the export table of a DSO and do not require a PLT indirection
2114 when used within the DSO@. Enabling this option can have a dramatic effect
2115 on load and link times of a DSO as it massively reduces the size of the
2116 dynamic export table when the library makes heavy use of templates.
2118 The behavior of this switch is not quite the same as marking the
2119 methods as hidden directly, because it does not affect static variables
2120 local to the function or cause the compiler to deduce that
2121 the function is defined in only one shared object.
2123 You may mark a method as having a visibility explicitly to negate the
2124 effect of the switch for that method. For example, if you do want to
2125 compare pointers to a particular inline method, you might mark it as
2126 having default visibility. Marking the enclosing class with explicit
2127 visibility will have no effect.
2129 Explicitly instantiated inline methods are unaffected by this option
2130 as their linkage might otherwise cross a shared library boundary.
2131 @xref{Template Instantiation}.
2133 @item -fvisibility-ms-compat
2134 @opindex fvisibility-ms-compat
2135 This flag attempts to use visibility settings to make GCC's C++
2136 linkage model compatible with that of Microsoft Visual Studio.
2138 The flag makes these changes to GCC's linkage model:
2142 It sets the default visibility to @code{hidden}, like
2143 @option{-fvisibility=hidden}.
2146 Types, but not their members, are not hidden by default.
2149 The One Definition Rule is relaxed for types without explicit
2150 visibility specifications which are defined in more than one different
2151 shared object: those declarations are permitted if they would have
2152 been permitted when this option was not used.
2155 In new code it is better to use @option{-fvisibility=hidden} and
2156 export those classes which are intended to be externally visible.
2157 Unfortunately it is possible for code to rely, perhaps accidentally,
2158 on the Visual Studio behavior.
2160 Among the consequences of these changes are that static data members
2161 of the same type with the same name but defined in different shared
2162 objects will be different, so changing one will not change the other;
2163 and that pointers to function members defined in different shared
2164 objects may not compare equal. When this flag is given, it is a
2165 violation of the ODR to define types with the same name differently.
2169 Do not use weak symbol support, even if it is provided by the linker.
2170 By default, G++ will use weak symbols if they are available. This
2171 option exists only for testing, and should not be used by end-users;
2172 it will result in inferior code and has no benefits. This option may
2173 be removed in a future release of G++.
2177 Do not search for header files in the standard directories specific to
2178 C++, but do still search the other standard directories. (This option
2179 is used when building the C++ library.)
2182 In addition, these optimization, warning, and code generation options
2183 have meanings only for C++ programs:
2186 @item -fno-default-inline
2187 @opindex fno-default-inline
2188 Do not assume @samp{inline} for functions defined inside a class scope.
2189 @xref{Optimize Options,,Options That Control Optimization}. Note that these
2190 functions will have linkage like inline functions; they just won't be
2193 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2196 Warn when G++ generates code that is probably not compatible with the
2197 vendor-neutral C++ ABI@. Although an effort has been made to warn about
2198 all such cases, there are probably some cases that are not warned about,
2199 even though G++ is generating incompatible code. There may also be
2200 cases where warnings are emitted even though the code that is generated
2203 You should rewrite your code to avoid these warnings if you are
2204 concerned about the fact that code generated by G++ may not be binary
2205 compatible with code generated by other compilers.
2207 The known incompatibilities in @option{-fabi-version=2} (the default) include:
2212 A template with a non-type template parameter of reference type is
2213 mangled incorrectly:
2216 template <int &> struct S @{@};
2220 This is fixed in @option{-fabi-version=3}.
2223 SIMD vector types declared using @code{__attribute ((vector_size))} are
2224 mangled in a non-standard way that does not allow for overloading of
2225 functions taking vectors of different sizes.
2227 The mangling is changed in @option{-fabi-version=4}.
2230 The known incompatibilities in @option{-fabi-version=1} include:
2235 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
2236 pack data into the same byte as a base class. For example:
2239 struct A @{ virtual void f(); int f1 : 1; @};
2240 struct B : public A @{ int f2 : 1; @};
2244 In this case, G++ will place @code{B::f2} into the same byte
2245 as@code{A::f1}; other compilers will not. You can avoid this problem
2246 by explicitly padding @code{A} so that its size is a multiple of the
2247 byte size on your platform; that will cause G++ and other compilers to
2248 layout @code{B} identically.
2251 Incorrect handling of tail-padding for virtual bases. G++ does not use
2252 tail padding when laying out virtual bases. For example:
2255 struct A @{ virtual void f(); char c1; @};
2256 struct B @{ B(); char c2; @};
2257 struct C : public A, public virtual B @{@};
2261 In this case, G++ will not place @code{B} into the tail-padding for
2262 @code{A}; other compilers will. You can avoid this problem by
2263 explicitly padding @code{A} so that its size is a multiple of its
2264 alignment (ignoring virtual base classes); that will cause G++ and other
2265 compilers to layout @code{C} identically.
2268 Incorrect handling of bit-fields with declared widths greater than that
2269 of their underlying types, when the bit-fields appear in a union. For
2273 union U @{ int i : 4096; @};
2277 Assuming that an @code{int} does not have 4096 bits, G++ will make the
2278 union too small by the number of bits in an @code{int}.
2281 Empty classes can be placed at incorrect offsets. For example:
2291 struct C : public B, public A @{@};
2295 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
2296 it should be placed at offset zero. G++ mistakenly believes that the
2297 @code{A} data member of @code{B} is already at offset zero.
2300 Names of template functions whose types involve @code{typename} or
2301 template template parameters can be mangled incorrectly.
2304 template <typename Q>
2305 void f(typename Q::X) @{@}
2307 template <template <typename> class Q>
2308 void f(typename Q<int>::X) @{@}
2312 Instantiations of these templates may be mangled incorrectly.
2316 It also warns psABI related changes. The known psABI changes at this
2322 For SYSV/x86-64, when passing union with long double, it is changed to
2323 pass in memory as specified in psABI. For example:
2333 @code{union U} will always be passed in memory.
2337 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2338 @opindex Wctor-dtor-privacy
2339 @opindex Wno-ctor-dtor-privacy
2340 Warn when a class seems unusable because all the constructors or
2341 destructors in that class are private, and it has neither friends nor
2342 public static member functions.
2344 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2345 @opindex Wdelete-non-virtual-dtor
2346 @opindex Wno-delete-non-virtual-dtor
2347 Warn when @samp{delete} is used to destroy an instance of a class which
2348 has virtual functions and non-virtual destructor. It is unsafe to delete
2349 an instance of a derived class through a pointer to a base class if the
2350 base class does not have a virtual destructor. This warning is enabled
2353 @item -Wno-narrowing @r{(C++ and Objective-C++ only)}
2355 @opindex Wno-narrowing
2356 With -std=c++0x, suppress the diagnostic required by the standard for
2357 narrowing conversions within @samp{@{ @}}, e.g.
2360 int i = @{ 2.2 @}; // error: narrowing from double to int
2363 This flag can be useful for compiling valid C++98 code in C++0x mode.
2365 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2367 @opindex Wno-noexcept
2368 Warn when a noexcept-expression evaluates to false because of a call
2369 to a function that does not have a non-throwing exception
2370 specification (i.e. @samp{throw()} or @samp{noexcept}) but is known by
2371 the compiler to never throw an exception.
2373 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2374 @opindex Wnon-virtual-dtor
2375 @opindex Wno-non-virtual-dtor
2376 Warn when a class has virtual functions and accessible non-virtual
2377 destructor, in which case it would be possible but unsafe to delete
2378 an instance of a derived class through a pointer to the base class.
2379 This warning is also enabled if -Weffc++ is specified.
2381 @item -Wreorder @r{(C++ and Objective-C++ only)}
2383 @opindex Wno-reorder
2384 @cindex reordering, warning
2385 @cindex warning for reordering of member initializers
2386 Warn when the order of member initializers given in the code does not
2387 match the order in which they must be executed. For instance:
2393 A(): j (0), i (1) @{ @}
2397 The compiler will rearrange the member initializers for @samp{i}
2398 and @samp{j} to match the declaration order of the members, emitting
2399 a warning to that effect. This warning is enabled by @option{-Wall}.
2402 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2405 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2408 Warn about violations of the following style guidelines from Scott Meyers'
2409 @cite{Effective C++} book:
2413 Item 11: Define a copy constructor and an assignment operator for classes
2414 with dynamically allocated memory.
2417 Item 12: Prefer initialization to assignment in constructors.
2420 Item 14: Make destructors virtual in base classes.
2423 Item 15: Have @code{operator=} return a reference to @code{*this}.
2426 Item 23: Don't try to return a reference when you must return an object.
2430 Also warn about violations of the following style guidelines from
2431 Scott Meyers' @cite{More Effective C++} book:
2435 Item 6: Distinguish between prefix and postfix forms of increment and
2436 decrement operators.
2439 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
2443 When selecting this option, be aware that the standard library
2444 headers do not obey all of these guidelines; use @samp{grep -v}
2445 to filter out those warnings.
2447 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2448 @opindex Wstrict-null-sentinel
2449 @opindex Wno-strict-null-sentinel
2450 Warn also about the use of an uncasted @code{NULL} as sentinel. When
2451 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2452 to @code{__null}. Although it is a null pointer constant not a null pointer,
2453 it is guaranteed to be of the same size as a pointer. But this use is
2454 not portable across different compilers.
2456 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2457 @opindex Wno-non-template-friend
2458 @opindex Wnon-template-friend
2459 Disable warnings when non-templatized friend functions are declared
2460 within a template. Since the advent of explicit template specification
2461 support in G++, if the name of the friend is an unqualified-id (i.e.,
2462 @samp{friend foo(int)}), the C++ language specification demands that the
2463 friend declare or define an ordinary, nontemplate function. (Section
2464 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2465 could be interpreted as a particular specialization of a templatized
2466 function. Because this non-conforming behavior is no longer the default
2467 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2468 check existing code for potential trouble spots and is on by default.
2469 This new compiler behavior can be turned off with
2470 @option{-Wno-non-template-friend} which keeps the conformant compiler code
2471 but disables the helpful warning.
2473 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2474 @opindex Wold-style-cast
2475 @opindex Wno-old-style-cast
2476 Warn if an old-style (C-style) cast to a non-void type is used within
2477 a C++ program. The new-style casts (@samp{dynamic_cast},
2478 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
2479 less vulnerable to unintended effects and much easier to search for.
2481 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2482 @opindex Woverloaded-virtual
2483 @opindex Wno-overloaded-virtual
2484 @cindex overloaded virtual function, warning
2485 @cindex warning for overloaded virtual function
2486 Warn when a function declaration hides virtual functions from a
2487 base class. For example, in:
2494 struct B: public A @{
2499 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2507 will fail to compile.
2509 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2510 @opindex Wno-pmf-conversions
2511 @opindex Wpmf-conversions
2512 Disable the diagnostic for converting a bound pointer to member function
2515 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2516 @opindex Wsign-promo
2517 @opindex Wno-sign-promo
2518 Warn when overload resolution chooses a promotion from unsigned or
2519 enumerated type to a signed type, over a conversion to an unsigned type of
2520 the same size. Previous versions of G++ would try to preserve
2521 unsignedness, but the standard mandates the current behavior.
2526 A& operator = (int);
2536 In this example, G++ will synthesize a default @samp{A& operator =
2537 (const A&);}, while cfront will use the user-defined @samp{operator =}.
2540 @node Objective-C and Objective-C++ Dialect Options
2541 @section Options Controlling Objective-C and Objective-C++ Dialects
2543 @cindex compiler options, Objective-C and Objective-C++
2544 @cindex Objective-C and Objective-C++ options, command line
2545 @cindex options, Objective-C and Objective-C++
2546 (NOTE: This manual does not describe the Objective-C and Objective-C++
2547 languages themselves. @xref{Standards,,Language Standards
2548 Supported by GCC}, for references.)
2550 This section describes the command-line options that are only meaningful
2551 for Objective-C and Objective-C++ programs, but you can also use most of
2552 the language-independent GNU compiler options.
2553 For example, you might compile a file @code{some_class.m} like this:
2556 gcc -g -fgnu-runtime -O -c some_class.m
2560 In this example, @option{-fgnu-runtime} is an option meant only for
2561 Objective-C and Objective-C++ programs; you can use the other options with
2562 any language supported by GCC@.
2564 Note that since Objective-C is an extension of the C language, Objective-C
2565 compilations may also use options specific to the C front-end (e.g.,
2566 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2567 C++-specific options (e.g., @option{-Wabi}).
2569 Here is a list of options that are @emph{only} for compiling Objective-C
2570 and Objective-C++ programs:
2573 @item -fconstant-string-class=@var{class-name}
2574 @opindex fconstant-string-class
2575 Use @var{class-name} as the name of the class to instantiate for each
2576 literal string specified with the syntax @code{@@"@dots{}"}. The default
2577 class name is @code{NXConstantString} if the GNU runtime is being used, and
2578 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2579 @option{-fconstant-cfstrings} option, if also present, will override the
2580 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2581 to be laid out as constant CoreFoundation strings.
2584 @opindex fgnu-runtime
2585 Generate object code compatible with the standard GNU Objective-C
2586 runtime. This is the default for most types of systems.
2588 @item -fnext-runtime
2589 @opindex fnext-runtime
2590 Generate output compatible with the NeXT runtime. This is the default
2591 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2592 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2595 @item -fno-nil-receivers
2596 @opindex fno-nil-receivers
2597 Assume that all Objective-C message dispatches (@code{[receiver
2598 message:arg]}) in this translation unit ensure that the receiver is
2599 not @code{nil}. This allows for more efficient entry points in the
2600 runtime to be used. This option is only available in conjunction with
2601 the NeXT runtime and ABI version 0 or 1.
2603 @item -fobjc-abi-version=@var{n}
2604 @opindex fobjc-abi-version
2605 Use version @var{n} of the Objective-C ABI for the selected runtime.
2606 This option is currently supported only for the NeXT runtime. In that
2607 case, Version 0 is the traditional (32-bit) ABI without support for
2608 properties and other Objective-C 2.0 additions. Version 1 is the
2609 traditional (32-bit) ABI with support for properties and other
2610 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2611 nothing is specified, the default is Version 0 on 32-bit target
2612 machines, and Version 2 on 64-bit target machines.
2614 @item -fobjc-call-cxx-cdtors
2615 @opindex fobjc-call-cxx-cdtors
2616 For each Objective-C class, check if any of its instance variables is a
2617 C++ object with a non-trivial default constructor. If so, synthesize a
2618 special @code{- (id) .cxx_construct} instance method that will run
2619 non-trivial default constructors on any such instance variables, in order,
2620 and then return @code{self}. Similarly, check if any instance variable
2621 is a C++ object with a non-trivial destructor, and if so, synthesize a
2622 special @code{- (void) .cxx_destruct} method that will run
2623 all such default destructors, in reverse order.
2625 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2626 methods thusly generated will only operate on instance variables
2627 declared in the current Objective-C class, and not those inherited
2628 from superclasses. It is the responsibility of the Objective-C
2629 runtime to invoke all such methods in an object's inheritance
2630 hierarchy. The @code{- (id) .cxx_construct} methods will be invoked
2631 by the runtime immediately after a new object instance is allocated;
2632 the @code{- (void) .cxx_destruct} methods will be invoked immediately
2633 before the runtime deallocates an object instance.
2635 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2636 support for invoking the @code{- (id) .cxx_construct} and
2637 @code{- (void) .cxx_destruct} methods.
2639 @item -fobjc-direct-dispatch
2640 @opindex fobjc-direct-dispatch
2641 Allow fast jumps to the message dispatcher. On Darwin this is
2642 accomplished via the comm page.
2644 @item -fobjc-exceptions
2645 @opindex fobjc-exceptions
2646 Enable syntactic support for structured exception handling in
2647 Objective-C, similar to what is offered by C++ and Java. This option
2648 is required to use the Objective-C keywords @code{@@try},
2649 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2650 @code{@@synchronized}. This option is available with both the GNU
2651 runtime and the NeXT runtime (but not available in conjunction with
2652 the NeXT runtime on Mac OS X 10.2 and earlier).
2656 Enable garbage collection (GC) in Objective-C and Objective-C++
2657 programs. This option is only available with the NeXT runtime; the
2658 GNU runtime has a different garbage collection implementation that
2659 does not require special compiler flags.
2661 @item -fobjc-nilcheck
2662 @opindex fobjc-nilcheck
2663 For the NeXT runtime with version 2 of the ABI, check for a nil
2664 receiver in method invocations before doing the actual method call.
2665 This is the default and can be disabled using
2666 @option{-fno-objc-nilcheck}. Class methods and super calls are never
2667 checked for nil in this way no matter what this flag is set to.
2668 Currently this flag does nothing when the GNU runtime, or an older
2669 version of the NeXT runtime ABI, is used.
2671 @item -fobjc-std=objc1
2673 Conform to the language syntax of Objective-C 1.0, the language
2674 recognized by GCC 4.0. This only affects the Objective-C additions to
2675 the C/C++ language; it does not affect conformance to C/C++ standards,
2676 which is controlled by the separate C/C++ dialect option flags. When
2677 this option is used with the Objective-C or Objective-C++ compiler,
2678 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
2679 This is useful if you need to make sure that your Objective-C code can
2680 be compiled with older versions of GCC.
2682 @item -freplace-objc-classes
2683 @opindex freplace-objc-classes
2684 Emit a special marker instructing @command{ld(1)} not to statically link in
2685 the resulting object file, and allow @command{dyld(1)} to load it in at
2686 run time instead. This is used in conjunction with the Fix-and-Continue
2687 debugging mode, where the object file in question may be recompiled and
2688 dynamically reloaded in the course of program execution, without the need
2689 to restart the program itself. Currently, Fix-and-Continue functionality
2690 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2695 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2696 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2697 compile time) with static class references that get initialized at load time,
2698 which improves run-time performance. Specifying the @option{-fzero-link} flag
2699 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2700 to be retained. This is useful in Zero-Link debugging mode, since it allows
2701 for individual class implementations to be modified during program execution.
2702 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
2703 regardless of command line options.
2707 Dump interface declarations for all classes seen in the source file to a
2708 file named @file{@var{sourcename}.decl}.
2710 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
2711 @opindex Wassign-intercept
2712 @opindex Wno-assign-intercept
2713 Warn whenever an Objective-C assignment is being intercepted by the
2716 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
2717 @opindex Wno-protocol
2719 If a class is declared to implement a protocol, a warning is issued for
2720 every method in the protocol that is not implemented by the class. The
2721 default behavior is to issue a warning for every method not explicitly
2722 implemented in the class, even if a method implementation is inherited
2723 from the superclass. If you use the @option{-Wno-protocol} option, then
2724 methods inherited from the superclass are considered to be implemented,
2725 and no warning is issued for them.
2727 @item -Wselector @r{(Objective-C and Objective-C++ only)}
2729 @opindex Wno-selector
2730 Warn if multiple methods of different types for the same selector are
2731 found during compilation. The check is performed on the list of methods
2732 in the final stage of compilation. Additionally, a check is performed
2733 for each selector appearing in a @code{@@selector(@dots{})}
2734 expression, and a corresponding method for that selector has been found
2735 during compilation. Because these checks scan the method table only at
2736 the end of compilation, these warnings are not produced if the final
2737 stage of compilation is not reached, for example because an error is
2738 found during compilation, or because the @option{-fsyntax-only} option is
2741 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
2742 @opindex Wstrict-selector-match
2743 @opindex Wno-strict-selector-match
2744 Warn if multiple methods with differing argument and/or return types are
2745 found for a given selector when attempting to send a message using this
2746 selector to a receiver of type @code{id} or @code{Class}. When this flag
2747 is off (which is the default behavior), the compiler will omit such warnings
2748 if any differences found are confined to types which share the same size
2751 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
2752 @opindex Wundeclared-selector
2753 @opindex Wno-undeclared-selector
2754 Warn if a @code{@@selector(@dots{})} expression referring to an
2755 undeclared selector is found. A selector is considered undeclared if no
2756 method with that name has been declared before the
2757 @code{@@selector(@dots{})} expression, either explicitly in an
2758 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2759 an @code{@@implementation} section. This option always performs its
2760 checks as soon as a @code{@@selector(@dots{})} expression is found,
2761 while @option{-Wselector} only performs its checks in the final stage of
2762 compilation. This also enforces the coding style convention
2763 that methods and selectors must be declared before being used.
2765 @item -print-objc-runtime-info
2766 @opindex print-objc-runtime-info
2767 Generate C header describing the largest structure that is passed by
2772 @node Language Independent Options
2773 @section Options to Control Diagnostic Messages Formatting
2774 @cindex options to control diagnostics formatting
2775 @cindex diagnostic messages
2776 @cindex message formatting
2778 Traditionally, diagnostic messages have been formatted irrespective of
2779 the output device's aspect (e.g.@: its width, @dots{}). The options described
2780 below can be used to control the diagnostic messages formatting
2781 algorithm, e.g.@: how many characters per line, how often source location
2782 information should be reported. Right now, only the C++ front end can
2783 honor these options. However it is expected, in the near future, that
2784 the remaining front ends would be able to digest them correctly.
2787 @item -fmessage-length=@var{n}
2788 @opindex fmessage-length
2789 Try to format error messages so that they fit on lines of about @var{n}
2790 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2791 the front ends supported by GCC@. If @var{n} is zero, then no
2792 line-wrapping will be done; each error message will appear on a single
2795 @opindex fdiagnostics-show-location
2796 @item -fdiagnostics-show-location=once
2797 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2798 reporter to emit @emph{once} source location information; that is, in
2799 case the message is too long to fit on a single physical line and has to
2800 be wrapped, the source location won't be emitted (as prefix) again,
2801 over and over, in subsequent continuation lines. This is the default
2804 @item -fdiagnostics-show-location=every-line
2805 Only meaningful in line-wrapping mode. Instructs the diagnostic
2806 messages reporter to emit the same source location information (as
2807 prefix) for physical lines that result from the process of breaking
2808 a message which is too long to fit on a single line.
2810 @item -fno-diagnostics-show-option
2811 @opindex fno-diagnostics-show-option
2812 @opindex fdiagnostics-show-option
2813 By default, each diagnostic emitted includes text which indicates the
2814 command line option that directly controls the diagnostic (if such an
2815 option is known to the diagnostic machinery). Specifying the
2816 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
2818 @item -Wcoverage-mismatch
2819 @opindex Wcoverage-mismatch
2820 Warn if feedback profiles do not match when using the
2821 @option{-fprofile-use} option.
2822 If a source file was changed between @option{-fprofile-gen} and
2823 @option{-fprofile-use}, the files with the profile feedback can fail
2824 to match the source file and GCC can not use the profile feedback
2825 information. By default, this warning is enabled and is treated as an
2826 error. @option{-Wno-coverage-mismatch} can be used to disable the
2827 warning or @option{-Wno-error=coverage-mismatch} can be used to
2828 disable the error. Disable the error for this warning can result in
2829 poorly optimized code, so disabling the error is useful only in the
2830 case of very minor changes such as bug fixes to an existing code-base.
2831 Completely disabling the warning is not recommended.
2835 @node Warning Options
2836 @section Options to Request or Suppress Warnings
2837 @cindex options to control warnings
2838 @cindex warning messages
2839 @cindex messages, warning
2840 @cindex suppressing warnings
2842 Warnings are diagnostic messages that report constructions which
2843 are not inherently erroneous but which are risky or suggest there
2844 may have been an error.
2846 The following language-independent options do not enable specific
2847 warnings but control the kinds of diagnostics produced by GCC.
2850 @cindex syntax checking
2852 @opindex fsyntax-only
2853 Check the code for syntax errors, but don't do anything beyond that.
2855 @item -fmax-errors=@var{n}
2856 @opindex fmax-errors
2857 Limits the maximum number of error messages to @var{n}, at which point
2858 GCC bails out rather than attempting to continue processing the source
2859 code. If @var{n} is 0 (the default), there is no limit on the number
2860 of error messages produced. If @option{-Wfatal-errors} is also
2861 specified, then @option{-Wfatal-errors} takes precedence over this
2866 Inhibit all warning messages.
2871 Make all warnings into errors.
2876 Make the specified warning into an error. The specifier for a warning
2877 is appended, for example @option{-Werror=switch} turns the warnings
2878 controlled by @option{-Wswitch} into errors. This switch takes a
2879 negative form, to be used to negate @option{-Werror} for specific
2880 warnings, for example @option{-Wno-error=switch} makes
2881 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
2884 The warning message for each controllable warning includes the
2885 option which controls the warning. That option can then be used with
2886 @option{-Werror=} and @option{-Wno-error=} as described above.
2887 (Printing of the option in the warning message can be disabled using the
2888 @option{-fno-diagnostics-show-option} flag.)
2890 Note that specifying @option{-Werror=}@var{foo} automatically implies
2891 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
2894 @item -Wfatal-errors
2895 @opindex Wfatal-errors
2896 @opindex Wno-fatal-errors
2897 This option causes the compiler to abort compilation on the first error
2898 occurred rather than trying to keep going and printing further error
2903 You can request many specific warnings with options beginning
2904 @samp{-W}, for example @option{-Wimplicit} to request warnings on
2905 implicit declarations. Each of these specific warning options also
2906 has a negative form beginning @samp{-Wno-} to turn off warnings; for
2907 example, @option{-Wno-implicit}. This manual lists only one of the
2908 two forms, whichever is not the default. For further,
2909 language-specific options also refer to @ref{C++ Dialect Options} and
2910 @ref{Objective-C and Objective-C++ Dialect Options}.
2912 When an unrecognized warning option is requested (e.g.,
2913 @option{-Wunknown-warning}), GCC will emit a diagnostic stating
2914 that the option is not recognized. However, if the @option{-Wno-} form
2915 is used, the behavior is slightly different: No diagnostic will be
2916 produced for @option{-Wno-unknown-warning} unless other diagnostics
2917 are being produced. This allows the use of new @option{-Wno-} options
2918 with old compilers, but if something goes wrong, the compiler will
2919 warn that an unrecognized option was used.
2924 Issue all the warnings demanded by strict ISO C and ISO C++;
2925 reject all programs that use forbidden extensions, and some other
2926 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2927 version of the ISO C standard specified by any @option{-std} option used.
2929 Valid ISO C and ISO C++ programs should compile properly with or without
2930 this option (though a rare few will require @option{-ansi} or a
2931 @option{-std} option specifying the required version of ISO C)@. However,
2932 without this option, certain GNU extensions and traditional C and C++
2933 features are supported as well. With this option, they are rejected.
2935 @option{-pedantic} does not cause warning messages for use of the
2936 alternate keywords whose names begin and end with @samp{__}. Pedantic
2937 warnings are also disabled in the expression that follows
2938 @code{__extension__}. However, only system header files should use
2939 these escape routes; application programs should avoid them.
2940 @xref{Alternate Keywords}.
2942 Some users try to use @option{-pedantic} to check programs for strict ISO
2943 C conformance. They soon find that it does not do quite what they want:
2944 it finds some non-ISO practices, but not all---only those for which
2945 ISO C @emph{requires} a diagnostic, and some others for which
2946 diagnostics have been added.
2948 A feature to report any failure to conform to ISO C might be useful in
2949 some instances, but would require considerable additional work and would
2950 be quite different from @option{-pedantic}. We don't have plans to
2951 support such a feature in the near future.
2953 Where the standard specified with @option{-std} represents a GNU
2954 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
2955 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2956 extended dialect is based. Warnings from @option{-pedantic} are given
2957 where they are required by the base standard. (It would not make sense
2958 for such warnings to be given only for features not in the specified GNU
2959 C dialect, since by definition the GNU dialects of C include all
2960 features the compiler supports with the given option, and there would be
2961 nothing to warn about.)
2963 @item -pedantic-errors
2964 @opindex pedantic-errors
2965 Like @option{-pedantic}, except that errors are produced rather than
2971 This enables all the warnings about constructions that some users
2972 consider questionable, and that are easy to avoid (or modify to
2973 prevent the warning), even in conjunction with macros. This also
2974 enables some language-specific warnings described in @ref{C++ Dialect
2975 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
2977 @option{-Wall} turns on the following warning flags:
2979 @gccoptlist{-Waddress @gol
2980 -Warray-bounds @r{(only with} @option{-O2}@r{)} @gol
2982 -Wchar-subscripts @gol
2983 -Wenum-compare @r{(in C/Objc; this is on by default in C++)} @gol
2984 -Wimplicit-int @r{(C and Objective-C only)} @gol
2985 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
2988 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
2989 -Wmaybe-uninitialized @gol
2990 -Wmissing-braces @gol
2996 -Wsequence-point @gol
2997 -Wsign-compare @r{(only in C++)} @gol
2998 -Wstrict-aliasing @gol
2999 -Wstrict-overflow=1 @gol
3002 -Wuninitialized @gol
3003 -Wunknown-pragmas @gol
3004 -Wunused-function @gol
3007 -Wunused-variable @gol
3008 -Wvolatile-register-var @gol
3011 Note that some warning flags are not implied by @option{-Wall}. Some of
3012 them warn about constructions that users generally do not consider
3013 questionable, but which occasionally you might wish to check for;
3014 others warn about constructions that are necessary or hard to avoid in
3015 some cases, and there is no simple way to modify the code to suppress
3016 the warning. Some of them are enabled by @option{-Wextra} but many of
3017 them must be enabled individually.
3023 This enables some extra warning flags that are not enabled by
3024 @option{-Wall}. (This option used to be called @option{-W}. The older
3025 name is still supported, but the newer name is more descriptive.)
3027 @gccoptlist{-Wclobbered @gol
3029 -Wignored-qualifiers @gol
3030 -Wmissing-field-initializers @gol
3031 -Wmissing-parameter-type @r{(C only)} @gol
3032 -Wold-style-declaration @r{(C only)} @gol
3033 -Woverride-init @gol
3036 -Wuninitialized @gol
3037 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3038 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3041 The option @option{-Wextra} also prints warning messages for the
3047 A pointer is compared against integer zero with @samp{<}, @samp{<=},
3048 @samp{>}, or @samp{>=}.
3051 (C++ only) An enumerator and a non-enumerator both appear in a
3052 conditional expression.
3055 (C++ only) Ambiguous virtual bases.
3058 (C++ only) Subscripting an array which has been declared @samp{register}.
3061 (C++ only) Taking the address of a variable which has been declared
3065 (C++ only) A base class is not initialized in a derived class' copy
3070 @item -Wchar-subscripts
3071 @opindex Wchar-subscripts
3072 @opindex Wno-char-subscripts
3073 Warn if an array subscript has type @code{char}. This is a common cause
3074 of error, as programmers often forget that this type is signed on some
3076 This warning is enabled by @option{-Wall}.
3080 @opindex Wno-comment
3081 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3082 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3083 This warning is enabled by @option{-Wall}.
3086 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3088 Suppress warning messages emitted by @code{#warning} directives.
3090 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3091 @opindex Wdouble-promotion
3092 @opindex Wno-double-promotion
3093 Give a warning when a value of type @code{float} is implicitly
3094 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3095 floating-point unit implement @code{float} in hardware, but emulate
3096 @code{double} in software. On such a machine, doing computations
3097 using @code{double} values is much more expensive because of the
3098 overhead required for software emulation.
3100 It is easy to accidentally do computations with @code{double} because
3101 floating-point literals are implicitly of type @code{double}. For
3105 float area(float radius)
3107 return 3.14159 * radius * radius;
3111 the compiler will perform the entire computation with @code{double}
3112 because the floating-point literal is a @code{double}.
3117 @opindex ffreestanding
3118 @opindex fno-builtin
3119 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3120 the arguments supplied have types appropriate to the format string
3121 specified, and that the conversions specified in the format string make
3122 sense. This includes standard functions, and others specified by format
3123 attributes (@pxref{Function Attributes}), in the @code{printf},
3124 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3125 not in the C standard) families (or other target-specific families).
3126 Which functions are checked without format attributes having been
3127 specified depends on the standard version selected, and such checks of
3128 functions without the attribute specified are disabled by
3129 @option{-ffreestanding} or @option{-fno-builtin}.
3131 The formats are checked against the format features supported by GNU
3132 libc version 2.2. These include all ISO C90 and C99 features, as well
3133 as features from the Single Unix Specification and some BSD and GNU
3134 extensions. Other library implementations may not support all these
3135 features; GCC does not support warning about features that go beyond a
3136 particular library's limitations. However, if @option{-pedantic} is used
3137 with @option{-Wformat}, warnings will be given about format features not
3138 in the selected standard version (but not for @code{strfmon} formats,
3139 since those are not in any version of the C standard). @xref{C Dialect
3140 Options,,Options Controlling C Dialect}.
3142 Since @option{-Wformat} also checks for null format arguments for
3143 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
3145 @option{-Wformat} is included in @option{-Wall}. For more control over some
3146 aspects of format checking, the options @option{-Wformat-y2k},
3147 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
3148 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
3149 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
3152 @opindex Wformat-y2k
3153 @opindex Wno-format-y2k
3154 If @option{-Wformat} is specified, also warn about @code{strftime}
3155 formats which may yield only a two-digit year.
3157 @item -Wno-format-contains-nul
3158 @opindex Wno-format-contains-nul
3159 @opindex Wformat-contains-nul
3160 If @option{-Wformat} is specified, do not warn about format strings that
3163 @item -Wno-format-extra-args
3164 @opindex Wno-format-extra-args
3165 @opindex Wformat-extra-args
3166 If @option{-Wformat} is specified, do not warn about excess arguments to a
3167 @code{printf} or @code{scanf} format function. The C standard specifies
3168 that such arguments are ignored.
3170 Where the unused arguments lie between used arguments that are
3171 specified with @samp{$} operand number specifications, normally
3172 warnings are still given, since the implementation could not know what
3173 type to pass to @code{va_arg} to skip the unused arguments. However,
3174 in the case of @code{scanf} formats, this option will suppress the
3175 warning if the unused arguments are all pointers, since the Single
3176 Unix Specification says that such unused arguments are allowed.
3178 @item -Wno-format-zero-length @r{(C and Objective-C only)}
3179 @opindex Wno-format-zero-length
3180 @opindex Wformat-zero-length
3181 If @option{-Wformat} is specified, do not warn about zero-length formats.
3182 The C standard specifies that zero-length formats are allowed.
3184 @item -Wformat-nonliteral
3185 @opindex Wformat-nonliteral
3186 @opindex Wno-format-nonliteral
3187 If @option{-Wformat} is specified, also warn if the format string is not a
3188 string literal and so cannot be checked, unless the format function
3189 takes its format arguments as a @code{va_list}.
3191 @item -Wformat-security
3192 @opindex Wformat-security
3193 @opindex Wno-format-security
3194 If @option{-Wformat} is specified, also warn about uses of format
3195 functions that represent possible security problems. At present, this
3196 warns about calls to @code{printf} and @code{scanf} functions where the
3197 format string is not a string literal and there are no format arguments,
3198 as in @code{printf (foo);}. This may be a security hole if the format
3199 string came from untrusted input and contains @samp{%n}. (This is
3200 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3201 in future warnings may be added to @option{-Wformat-security} that are not
3202 included in @option{-Wformat-nonliteral}.)
3206 @opindex Wno-format=2
3207 Enable @option{-Wformat} plus format checks not included in
3208 @option{-Wformat}. Currently equivalent to @samp{-Wformat
3209 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
3211 @item -Wnonnull @r{(C and Objective-C only)}
3213 @opindex Wno-nonnull
3214 Warn about passing a null pointer for arguments marked as
3215 requiring a non-null value by the @code{nonnull} function attribute.
3217 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3218 can be disabled with the @option{-Wno-nonnull} option.
3220 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3222 @opindex Wno-init-self
3223 Warn about uninitialized variables which are initialized with themselves.
3224 Note this option can only be used with the @option{-Wuninitialized} option.
3226 For example, GCC will warn about @code{i} being uninitialized in the
3227 following snippet only when @option{-Winit-self} has been specified:
3238 @item -Wimplicit-int @r{(C and Objective-C only)}
3239 @opindex Wimplicit-int
3240 @opindex Wno-implicit-int
3241 Warn when a declaration does not specify a type.
3242 This warning is enabled by @option{-Wall}.
3244 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3245 @opindex Wimplicit-function-declaration
3246 @opindex Wno-implicit-function-declaration
3247 Give a warning whenever a function is used before being declared. In
3248 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3249 enabled by default and it is made into an error by
3250 @option{-pedantic-errors}. This warning is also enabled by
3253 @item -Wimplicit @r{(C and Objective-C only)}
3255 @opindex Wno-implicit
3256 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3257 This warning is enabled by @option{-Wall}.
3259 @item -Wignored-qualifiers @r{(C and C++ only)}
3260 @opindex Wignored-qualifiers
3261 @opindex Wno-ignored-qualifiers
3262 Warn if the return type of a function has a type qualifier
3263 such as @code{const}. For ISO C such a type qualifier has no effect,
3264 since the value returned by a function is not an lvalue.
3265 For C++, the warning is only emitted for scalar types or @code{void}.
3266 ISO C prohibits qualified @code{void} return types on function
3267 definitions, so such return types always receive a warning
3268 even without this option.
3270 This warning is also enabled by @option{-Wextra}.
3275 Warn if the type of @samp{main} is suspicious. @samp{main} should be
3276 a function with external linkage, returning int, taking either zero
3277 arguments, two, or three arguments of appropriate types. This warning
3278 is enabled by default in C++ and is enabled by either @option{-Wall}
3279 or @option{-pedantic}.
3281 @item -Wmissing-braces
3282 @opindex Wmissing-braces
3283 @opindex Wno-missing-braces
3284 Warn if an aggregate or union initializer is not fully bracketed. In
3285 the following example, the initializer for @samp{a} is not fully
3286 bracketed, but that for @samp{b} is fully bracketed.
3289 int a[2][2] = @{ 0, 1, 2, 3 @};
3290 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3293 This warning is enabled by @option{-Wall}.
3295 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3296 @opindex Wmissing-include-dirs
3297 @opindex Wno-missing-include-dirs
3298 Warn if a user-supplied include directory does not exist.
3301 @opindex Wparentheses
3302 @opindex Wno-parentheses
3303 Warn if parentheses are omitted in certain contexts, such
3304 as when there is an assignment in a context where a truth value
3305 is expected, or when operators are nested whose precedence people
3306 often get confused about.
3308 Also warn if a comparison like @samp{x<=y<=z} appears; this is
3309 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
3310 interpretation from that of ordinary mathematical notation.
3312 Also warn about constructions where there may be confusion to which
3313 @code{if} statement an @code{else} branch belongs. Here is an example of
3328 In C/C++, every @code{else} branch belongs to the innermost possible
3329 @code{if} statement, which in this example is @code{if (b)}. This is
3330 often not what the programmer expected, as illustrated in the above
3331 example by indentation the programmer chose. When there is the
3332 potential for this confusion, GCC will issue a warning when this flag
3333 is specified. To eliminate the warning, add explicit braces around
3334 the innermost @code{if} statement so there is no way the @code{else}
3335 could belong to the enclosing @code{if}. The resulting code would
3352 Also warn for dangerous uses of the
3353 ?: with omitted middle operand GNU extension. When the condition
3354 in the ?: operator is a boolean expression the omitted value will
3355 be always 1. Often the user expects it to be a value computed
3356 inside the conditional expression instead.
3358 This warning is enabled by @option{-Wall}.
3360 @item -Wsequence-point
3361 @opindex Wsequence-point
3362 @opindex Wno-sequence-point
3363 Warn about code that may have undefined semantics because of violations
3364 of sequence point rules in the C and C++ standards.
3366 The C and C++ standards defines the order in which expressions in a C/C++
3367 program are evaluated in terms of @dfn{sequence points}, which represent
3368 a partial ordering between the execution of parts of the program: those
3369 executed before the sequence point, and those executed after it. These
3370 occur after the evaluation of a full expression (one which is not part
3371 of a larger expression), after the evaluation of the first operand of a
3372 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3373 function is called (but after the evaluation of its arguments and the
3374 expression denoting the called function), and in certain other places.
3375 Other than as expressed by the sequence point rules, the order of
3376 evaluation of subexpressions of an expression is not specified. All
3377 these rules describe only a partial order rather than a total order,
3378 since, for example, if two functions are called within one expression
3379 with no sequence point between them, the order in which the functions
3380 are called is not specified. However, the standards committee have
3381 ruled that function calls do not overlap.
3383 It is not specified when between sequence points modifications to the
3384 values of objects take effect. Programs whose behavior depends on this
3385 have undefined behavior; the C and C++ standards specify that ``Between
3386 the previous and next sequence point an object shall have its stored
3387 value modified at most once by the evaluation of an expression.
3388 Furthermore, the prior value shall be read only to determine the value
3389 to be stored.''. If a program breaks these rules, the results on any
3390 particular implementation are entirely unpredictable.
3392 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3393 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3394 diagnosed by this option, and it may give an occasional false positive
3395 result, but in general it has been found fairly effective at detecting
3396 this sort of problem in programs.
3398 The standard is worded confusingly, therefore there is some debate
3399 over the precise meaning of the sequence point rules in subtle cases.
3400 Links to discussions of the problem, including proposed formal
3401 definitions, may be found on the GCC readings page, at
3402 @uref{http://gcc.gnu.org/@/readings.html}.
3404 This warning is enabled by @option{-Wall} for C and C++.
3407 @opindex Wreturn-type
3408 @opindex Wno-return-type
3409 Warn whenever a function is defined with a return-type that defaults
3410 to @code{int}. Also warn about any @code{return} statement with no
3411 return-value in a function whose return-type is not @code{void}
3412 (falling off the end of the function body is considered returning
3413 without a value), and about a @code{return} statement with an
3414 expression in a function whose return-type is @code{void}.
3416 For C++, a function without return type always produces a diagnostic
3417 message, even when @option{-Wno-return-type} is specified. The only
3418 exceptions are @samp{main} and functions defined in system headers.
3420 This warning is enabled by @option{-Wall}.
3425 Warn whenever a @code{switch} statement has an index of enumerated type
3426 and lacks a @code{case} for one or more of the named codes of that
3427 enumeration. (The presence of a @code{default} label prevents this
3428 warning.) @code{case} labels outside the enumeration range also
3429 provoke warnings when this option is used (even if there is a
3430 @code{default} label).
3431 This warning is enabled by @option{-Wall}.
3433 @item -Wswitch-default
3434 @opindex Wswitch-default
3435 @opindex Wno-switch-default
3436 Warn whenever a @code{switch} statement does not have a @code{default}
3440 @opindex Wswitch-enum
3441 @opindex Wno-switch-enum
3442 Warn whenever a @code{switch} statement has an index of enumerated type
3443 and lacks a @code{case} for one or more of the named codes of that
3444 enumeration. @code{case} labels outside the enumeration range also
3445 provoke warnings when this option is used. The only difference
3446 between @option{-Wswitch} and this option is that this option gives a
3447 warning about an omitted enumeration code even if there is a
3448 @code{default} label.
3450 @item -Wsync-nand @r{(C and C++ only)}
3452 @opindex Wno-sync-nand
3453 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3454 built-in functions are used. These functions changed semantics in GCC 4.4.
3458 @opindex Wno-trigraphs
3459 Warn if any trigraphs are encountered that might change the meaning of
3460 the program (trigraphs within comments are not warned about).
3461 This warning is enabled by @option{-Wall}.
3463 @item -Wunused-but-set-parameter
3464 @opindex Wunused-but-set-parameter
3465 @opindex Wno-unused-but-set-parameter
3466 Warn whenever a function parameter is assigned to, but otherwise unused
3467 (aside from its declaration).
3469 To suppress this warning use the @samp{unused} attribute
3470 (@pxref{Variable Attributes}).
3472 This warning is also enabled by @option{-Wunused} together with
3475 @item -Wunused-but-set-variable
3476 @opindex Wunused-but-set-variable
3477 @opindex Wno-unused-but-set-variable
3478 Warn whenever a local variable is assigned to, but otherwise unused
3479 (aside from its declaration).
3480 This warning is enabled by @option{-Wall}.
3482 To suppress this warning use the @samp{unused} attribute
3483 (@pxref{Variable Attributes}).
3485 This warning is also enabled by @option{-Wunused}, which is enabled
3488 @item -Wunused-function
3489 @opindex Wunused-function
3490 @opindex Wno-unused-function
3491 Warn whenever a static function is declared but not defined or a
3492 non-inline static function is unused.
3493 This warning is enabled by @option{-Wall}.
3495 @item -Wunused-label
3496 @opindex Wunused-label
3497 @opindex Wno-unused-label
3498 Warn whenever a label is declared but not used.
3499 This warning is enabled by @option{-Wall}.
3501 To suppress this warning use the @samp{unused} attribute
3502 (@pxref{Variable Attributes}).
3504 @item -Wunused-parameter
3505 @opindex Wunused-parameter
3506 @opindex Wno-unused-parameter
3507 Warn whenever a function parameter is unused aside from its declaration.
3509 To suppress this warning use the @samp{unused} attribute
3510 (@pxref{Variable Attributes}).
3512 @item -Wno-unused-result
3513 @opindex Wunused-result
3514 @opindex Wno-unused-result
3515 Do not warn if a caller of a function marked with attribute
3516 @code{warn_unused_result} (@pxref{Variable Attributes}) does not use
3517 its return value. The default is @option{-Wunused-result}.
3519 @item -Wunused-variable
3520 @opindex Wunused-variable
3521 @opindex Wno-unused-variable
3522 Warn whenever a local variable or non-constant static variable is unused
3523 aside from its declaration.
3524 This warning is enabled by @option{-Wall}.
3526 To suppress this warning use the @samp{unused} attribute
3527 (@pxref{Variable Attributes}).
3529 @item -Wunused-value
3530 @opindex Wunused-value
3531 @opindex Wno-unused-value
3532 Warn whenever a statement computes a result that is explicitly not
3533 used. To suppress this warning cast the unused expression to
3534 @samp{void}. This includes an expression-statement or the left-hand
3535 side of a comma expression that contains no side effects. For example,
3536 an expression such as @samp{x[i,j]} will cause a warning, while
3537 @samp{x[(void)i,j]} will not.
3539 This warning is enabled by @option{-Wall}.
3544 All the above @option{-Wunused} options combined.
3546 In order to get a warning about an unused function parameter, you must
3547 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
3548 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
3550 @item -Wuninitialized
3551 @opindex Wuninitialized
3552 @opindex Wno-uninitialized
3553 Warn if an automatic variable is used without first being initialized
3554 or if a variable may be clobbered by a @code{setjmp} call. In C++,
3555 warn if a non-static reference or non-static @samp{const} member
3556 appears in a class without constructors.
3558 If you want to warn about code which uses the uninitialized value of the
3559 variable in its own initializer, use the @option{-Winit-self} option.
3561 These warnings occur for individual uninitialized or clobbered
3562 elements of structure, union or array variables as well as for
3563 variables which are uninitialized or clobbered as a whole. They do
3564 not occur for variables or elements declared @code{volatile}. Because
3565 these warnings depend on optimization, the exact variables or elements
3566 for which there are warnings will depend on the precise optimization
3567 options and version of GCC used.
3569 Note that there may be no warning about a variable that is used only
3570 to compute a value that itself is never used, because such
3571 computations may be deleted by data flow analysis before the warnings
3574 @item -Wmaybe-uninitialized
3575 @opindex Wmaybe-uninitialized
3576 @opindex Wno-maybe-uninitialized
3577 For an automatic variable, if there exists a path from the function
3578 entry to a use of the variable that is initialized, but there exist
3579 some other paths the variable is not initialized, the compiler will
3580 emit a warning if it can not prove the uninitialized paths do not
3581 happen at runtime. These warnings are made optional because GCC is
3582 not smart enough to see all the reasons why the code might be correct
3583 despite appearing to have an error. Here is one example of how
3604 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
3605 always initialized, but GCC doesn't know this. To suppress the
3606 warning, the user needs to provide a default case with assert(0) or
3609 @cindex @code{longjmp} warnings
3610 This option also warns when a non-volatile automatic variable might be
3611 changed by a call to @code{longjmp}. These warnings as well are possible
3612 only in optimizing compilation.
3614 The compiler sees only the calls to @code{setjmp}. It cannot know
3615 where @code{longjmp} will be called; in fact, a signal handler could
3616 call it at any point in the code. As a result, you may get a warning
3617 even when there is in fact no problem because @code{longjmp} cannot
3618 in fact be called at the place which would cause a problem.
3620 Some spurious warnings can be avoided if you declare all the functions
3621 you use that never return as @code{noreturn}. @xref{Function
3624 This warning is enabled by @option{-Wall} or @option{-Wextra}.
3626 @item -Wunknown-pragmas
3627 @opindex Wunknown-pragmas
3628 @opindex Wno-unknown-pragmas
3629 @cindex warning for unknown pragmas
3630 @cindex unknown pragmas, warning
3631 @cindex pragmas, warning of unknown
3632 Warn when a #pragma directive is encountered which is not understood by
3633 GCC@. If this command line option is used, warnings will even be issued
3634 for unknown pragmas in system header files. This is not the case if
3635 the warnings were only enabled by the @option{-Wall} command line option.
3638 @opindex Wno-pragmas
3640 Do not warn about misuses of pragmas, such as incorrect parameters,
3641 invalid syntax, or conflicts between pragmas. See also
3642 @samp{-Wunknown-pragmas}.
3644 @item -Wstrict-aliasing
3645 @opindex Wstrict-aliasing
3646 @opindex Wno-strict-aliasing
3647 This option is only active when @option{-fstrict-aliasing} is active.
3648 It warns about code which might break the strict aliasing rules that the
3649 compiler is using for optimization. The warning does not catch all
3650 cases, but does attempt to catch the more common pitfalls. It is
3651 included in @option{-Wall}.
3652 It is equivalent to @option{-Wstrict-aliasing=3}
3654 @item -Wstrict-aliasing=n
3655 @opindex Wstrict-aliasing=n
3656 @opindex Wno-strict-aliasing=n
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.
3660 Higher levels correspond to higher accuracy (fewer false positives).
3661 Higher levels also correspond to more effort, similar to the way -O works.
3662 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=n},
3665 Level 1: Most aggressive, quick, least accurate.
3666 Possibly useful when higher levels
3667 do not warn but -fstrict-aliasing still breaks the code, as it has very few
3668 false negatives. However, it has many false positives.
3669 Warns for all pointer conversions between possibly incompatible types,
3670 even if never dereferenced. Runs in the frontend only.
3672 Level 2: Aggressive, quick, not too precise.
3673 May still have many false positives (not as many as level 1 though),
3674 and few false negatives (but possibly more than level 1).
3675 Unlike level 1, it only warns when an address is taken. Warns about
3676 incomplete types. Runs in the frontend only.
3678 Level 3 (default for @option{-Wstrict-aliasing}):
3679 Should have very few false positives and few false
3680 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
3681 Takes care of the common pun+dereference pattern in the frontend:
3682 @code{*(int*)&some_float}.
3683 If optimization is enabled, it also runs in the backend, where it deals
3684 with multiple statement cases using flow-sensitive points-to information.
3685 Only warns when the converted pointer is dereferenced.
3686 Does not warn about incomplete types.
3688 @item -Wstrict-overflow
3689 @itemx -Wstrict-overflow=@var{n}
3690 @opindex Wstrict-overflow
3691 @opindex Wno-strict-overflow
3692 This option is only active when @option{-fstrict-overflow} is active.
3693 It warns about cases where the compiler optimizes based on the
3694 assumption that signed overflow does not occur. Note that it does not
3695 warn about all cases where the code might overflow: it only warns
3696 about cases where the compiler implements some optimization. Thus
3697 this warning depends on the optimization level.
3699 An optimization which assumes that signed overflow does not occur is
3700 perfectly safe if the values of the variables involved are such that
3701 overflow never does, in fact, occur. Therefore this warning can
3702 easily give a false positive: a warning about code which is not
3703 actually a problem. To help focus on important issues, several
3704 warning levels are defined. No warnings are issued for the use of
3705 undefined signed overflow when estimating how many iterations a loop
3706 will require, in particular when determining whether a loop will be
3710 @item -Wstrict-overflow=1
3711 Warn about cases which are both questionable and easy to avoid. For
3712 example: @code{x + 1 > x}; with @option{-fstrict-overflow}, the
3713 compiler will simplify this to @code{1}. This level of
3714 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
3715 are not, and must be explicitly requested.
3717 @item -Wstrict-overflow=2
3718 Also warn about other cases where a comparison is simplified to a
3719 constant. For example: @code{abs (x) >= 0}. This can only be
3720 simplified when @option{-fstrict-overflow} is in effect, because
3721 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
3722 zero. @option{-Wstrict-overflow} (with no level) is the same as
3723 @option{-Wstrict-overflow=2}.
3725 @item -Wstrict-overflow=3
3726 Also warn about other cases where a comparison is simplified. For
3727 example: @code{x + 1 > 1} will be simplified to @code{x > 0}.
3729 @item -Wstrict-overflow=4
3730 Also warn about other simplifications not covered by the above cases.
3731 For example: @code{(x * 10) / 5} will be simplified to @code{x * 2}.
3733 @item -Wstrict-overflow=5
3734 Also warn about cases where the compiler reduces the magnitude of a
3735 constant involved in a comparison. For example: @code{x + 2 > y} will
3736 be simplified to @code{x + 1 >= y}. This is reported only at the
3737 highest warning level because this simplification applies to many
3738 comparisons, so this warning level will give a very large number of
3742 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{]}
3743 @opindex Wsuggest-attribute=
3744 @opindex Wno-suggest-attribute=
3745 Warn for cases where adding an attribute may be beneficial. The
3746 attributes currently supported are listed below.
3749 @item -Wsuggest-attribute=pure
3750 @itemx -Wsuggest-attribute=const
3751 @itemx -Wsuggest-attribute=noreturn
3752 @opindex Wsuggest-attribute=pure
3753 @opindex Wno-suggest-attribute=pure
3754 @opindex Wsuggest-attribute=const
3755 @opindex Wno-suggest-attribute=const
3756 @opindex Wsuggest-attribute=noreturn
3757 @opindex Wno-suggest-attribute=noreturn
3759 Warn about functions which might be candidates for attributes
3760 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
3761 functions visible in other compilation units or (in the case of @code{pure} and
3762 @code{const}) if it cannot prove that the function returns normally. A function
3763 returns normally if it doesn't contain an infinite loop nor returns abnormally
3764 by throwing, calling @code{abort()} or trapping. This analysis requires option
3765 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
3766 higher. Higher optimization levels improve the accuracy of the analysis.
3769 @item -Warray-bounds
3770 @opindex Wno-array-bounds
3771 @opindex Warray-bounds
3772 This option is only active when @option{-ftree-vrp} is active
3773 (default for @option{-O2} and above). It warns about subscripts to arrays
3774 that are always out of bounds. This warning is enabled by @option{-Wall}.
3776 @item -Wno-div-by-zero
3777 @opindex Wno-div-by-zero
3778 @opindex Wdiv-by-zero
3779 Do not warn about compile-time integer division by zero. Floating point
3780 division by zero is not warned about, as it can be a legitimate way of
3781 obtaining infinities and NaNs.
3783 @item -Wsystem-headers
3784 @opindex Wsystem-headers
3785 @opindex Wno-system-headers
3786 @cindex warnings from system headers
3787 @cindex system headers, warnings from
3788 Print warning messages for constructs found in system header files.
3789 Warnings from system headers are normally suppressed, on the assumption
3790 that they usually do not indicate real problems and would only make the
3791 compiler output harder to read. Using this command line option tells
3792 GCC to emit warnings from system headers as if they occurred in user
3793 code. However, note that using @option{-Wall} in conjunction with this
3794 option will @emph{not} warn about unknown pragmas in system
3795 headers---for that, @option{-Wunknown-pragmas} must also be used.
3798 @opindex Wtrampolines
3799 @opindex Wno-trampolines
3800 Warn about trampolines generated for pointers to nested functions.
3802 A trampoline is a small piece of data or code that is created at run
3803 time on the stack when the address of a nested function is taken, and
3804 is used to call the nested function indirectly. For some targets, it
3805 is made up of data only and thus requires no special treatment. But,
3806 for most targets, it is made up of code and thus requires the stack
3807 to be made executable in order for the program to work properly.
3810 @opindex Wfloat-equal
3811 @opindex Wno-float-equal
3812 Warn if floating point values are used in equality comparisons.
3814 The idea behind this is that sometimes it is convenient (for the
3815 programmer) to consider floating-point values as approximations to
3816 infinitely precise real numbers. If you are doing this, then you need
3817 to compute (by analyzing the code, or in some other way) the maximum or
3818 likely maximum error that the computation introduces, and allow for it
3819 when performing comparisons (and when producing output, but that's a
3820 different problem). In particular, instead of testing for equality, you
3821 would check to see whether the two values have ranges that overlap; and
3822 this is done with the relational operators, so equality comparisons are
3825 @item -Wtraditional @r{(C and Objective-C only)}
3826 @opindex Wtraditional
3827 @opindex Wno-traditional
3828 Warn about certain constructs that behave differently in traditional and
3829 ISO C@. Also warn about ISO C constructs that have no traditional C
3830 equivalent, and/or problematic constructs which should be avoided.
3834 Macro parameters that appear within string literals in the macro body.
3835 In traditional C macro replacement takes place within string literals,
3836 but does not in ISO C@.
3839 In traditional C, some preprocessor directives did not exist.
3840 Traditional preprocessors would only consider a line to be a directive
3841 if the @samp{#} appeared in column 1 on the line. Therefore
3842 @option{-Wtraditional} warns about directives that traditional C
3843 understands but would ignore because the @samp{#} does not appear as the
3844 first character on the line. It also suggests you hide directives like
3845 @samp{#pragma} not understood by traditional C by indenting them. Some
3846 traditional implementations would not recognize @samp{#elif}, so it
3847 suggests avoiding it altogether.
3850 A function-like macro that appears without arguments.
3853 The unary plus operator.
3856 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3857 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3858 constants.) Note, these suffixes appear in macros defined in the system
3859 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3860 Use of these macros in user code might normally lead to spurious
3861 warnings, however GCC's integrated preprocessor has enough context to
3862 avoid warning in these cases.
3865 A function declared external in one block and then used after the end of
3869 A @code{switch} statement has an operand of type @code{long}.
3872 A non-@code{static} function declaration follows a @code{static} one.
3873 This construct is not accepted by some traditional C compilers.
3876 The ISO type of an integer constant has a different width or
3877 signedness from its traditional type. This warning is only issued if
3878 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3879 typically represent bit patterns, are not warned about.
3882 Usage of ISO string concatenation is detected.
3885 Initialization of automatic aggregates.
3888 Identifier conflicts with labels. Traditional C lacks a separate
3889 namespace for labels.
3892 Initialization of unions. If the initializer is zero, the warning is
3893 omitted. This is done under the assumption that the zero initializer in
3894 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3895 initializer warnings and relies on default initialization to zero in the
3899 Conversions by prototypes between fixed/floating point values and vice
3900 versa. The absence of these prototypes when compiling with traditional
3901 C would cause serious problems. This is a subset of the possible
3902 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3905 Use of ISO C style function definitions. This warning intentionally is
3906 @emph{not} issued for prototype declarations or variadic functions
3907 because these ISO C features will appear in your code when using
3908 libiberty's traditional C compatibility macros, @code{PARAMS} and
3909 @code{VPARAMS}. This warning is also bypassed for nested functions
3910 because that feature is already a GCC extension and thus not relevant to
3911 traditional C compatibility.
3914 @item -Wtraditional-conversion @r{(C and Objective-C only)}
3915 @opindex Wtraditional-conversion
3916 @opindex Wno-traditional-conversion
3917 Warn if a prototype causes a type conversion that is different from what
3918 would happen to the same argument in the absence of a prototype. This
3919 includes conversions of fixed point to floating and vice versa, and
3920 conversions changing the width or signedness of a fixed point argument
3921 except when the same as the default promotion.
3923 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
3924 @opindex Wdeclaration-after-statement
3925 @opindex Wno-declaration-after-statement
3926 Warn when a declaration is found after a statement in a block. This
3927 construct, known from C++, was introduced with ISO C99 and is by default
3928 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3929 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3934 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3936 @item -Wno-endif-labels
3937 @opindex Wno-endif-labels
3938 @opindex Wendif-labels
3939 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3944 Warn whenever a local variable or type declaration shadows another variable,
3945 parameter, type, or class member (in C++), or whenever a built-in function
3946 is shadowed. Note that in C++, the compiler will not warn if a local variable
3947 shadows a struct/class/enum, but will warn if it shadows an explicit typedef.
3949 @item -Wlarger-than=@var{len}
3950 @opindex Wlarger-than=@var{len}
3951 @opindex Wlarger-than-@var{len}
3952 Warn whenever an object of larger than @var{len} bytes is defined.
3954 @item -Wframe-larger-than=@var{len}
3955 @opindex Wframe-larger-than
3956 Warn if the size of a function frame is larger than @var{len} bytes.
3957 The computation done to determine the stack frame size is approximate
3958 and not conservative.
3959 The actual requirements may be somewhat greater than @var{len}
3960 even if you do not get a warning. In addition, any space allocated
3961 via @code{alloca}, variable-length arrays, or related constructs
3962 is not included by the compiler when determining
3963 whether or not to issue a warning.
3965 @item -Wno-free-nonheap-object
3966 @opindex Wno-free-nonheap-object
3967 @opindex Wfree-nonheap-object
3968 Do not warn when attempting to free an object which was not allocated
3971 @item -Wstack-usage=@var{len}
3972 @opindex Wstack-usage
3973 Warn if the stack usage of a function might be larger than @var{len} bytes.
3974 The computation done to determine the stack usage is conservative.
3975 Any space allocated via @code{alloca}, variable-length arrays, or related
3976 constructs is included by the compiler when determining whether or not to
3979 The message is in keeping with the output of @option{-fstack-usage}.
3983 If the stack usage is fully static but exceeds the specified amount, it's:
3986 Â warning: stack usage is 1120 bytes
3989 If the stack usage is (partly) dynamic but bounded, it's:
3992 Â warning: stack usage might be 1648 bytes
3995 If the stack usage is (partly) dynamic and not bounded, it's:
3998 Â warning: stack usage might be unbounded
4002 @item -Wunsafe-loop-optimizations
4003 @opindex Wunsafe-loop-optimizations
4004 @opindex Wno-unsafe-loop-optimizations
4005 Warn if the loop cannot be optimized because the compiler could not
4006 assume anything on the bounds of the loop indices. With
4007 @option{-funsafe-loop-optimizations} warn if the compiler made
4010 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4011 @opindex Wno-pedantic-ms-format
4012 @opindex Wpedantic-ms-format
4013 Disables the warnings about non-ISO @code{printf} / @code{scanf} format
4014 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets
4015 depending on the MS runtime, when you are using the options @option{-Wformat}
4016 and @option{-pedantic} without gnu-extensions.
4018 @item -Wpointer-arith
4019 @opindex Wpointer-arith
4020 @opindex Wno-pointer-arith
4021 Warn about anything that depends on the ``size of'' a function type or
4022 of @code{void}. GNU C assigns these types a size of 1, for
4023 convenience in calculations with @code{void *} pointers and pointers
4024 to functions. In C++, warn also when an arithmetic operation involves
4025 @code{NULL}. This warning is also enabled by @option{-pedantic}.
4028 @opindex Wtype-limits
4029 @opindex Wno-type-limits
4030 Warn if a comparison is always true or always false due to the limited
4031 range of the data type, but do not warn for constant expressions. For
4032 example, warn if an unsigned variable is compared against zero with
4033 @samp{<} or @samp{>=}. This warning is also enabled by
4036 @item -Wbad-function-cast @r{(C and Objective-C only)}
4037 @opindex Wbad-function-cast
4038 @opindex Wno-bad-function-cast
4039 Warn whenever a function call is cast to a non-matching type.
4040 For example, warn if @code{int malloc()} is cast to @code{anything *}.
4042 @item -Wc++-compat @r{(C and Objective-C only)}
4043 Warn about ISO C constructs that are outside of the common subset of
4044 ISO C and ISO C++, e.g.@: request for implicit conversion from
4045 @code{void *} to a pointer to non-@code{void} type.
4047 @item -Wc++0x-compat @r{(C++ and Objective-C++ only)}
4048 Warn about C++ constructs whose meaning differs between ISO C++ 1998 and
4049 ISO C++ 200x, e.g., identifiers in ISO C++ 1998 that will become keywords
4050 in ISO C++ 200x. This warning is enabled by @option{-Wall}.
4054 @opindex Wno-cast-qual
4055 Warn whenever a pointer is cast so as to remove a type qualifier from
4056 the target type. For example, warn if a @code{const char *} is cast
4057 to an ordinary @code{char *}.
4059 Also warn when making a cast which introduces a type qualifier in an
4060 unsafe way. For example, casting @code{char **} to @code{const char **}
4061 is unsafe, as in this example:
4064 /* p is char ** value. */
4065 const char **q = (const char **) p;
4066 /* Assignment of readonly string to const char * is OK. */
4068 /* Now char** pointer points to read-only memory. */
4073 @opindex Wcast-align
4074 @opindex Wno-cast-align
4075 Warn whenever a pointer is cast such that the required alignment of the
4076 target is increased. For example, warn if a @code{char *} is cast to
4077 an @code{int *} on machines where integers can only be accessed at
4078 two- or four-byte boundaries.
4080 @item -Wwrite-strings
4081 @opindex Wwrite-strings
4082 @opindex Wno-write-strings
4083 When compiling C, give string constants the type @code{const
4084 char[@var{length}]} so that copying the address of one into a
4085 non-@code{const} @code{char *} pointer will get a warning. These
4086 warnings will help you find at compile time code that can try to write
4087 into a string constant, but only if you have been very careful about
4088 using @code{const} in declarations and prototypes. Otherwise, it will
4089 just be a nuisance. This is why we did not make @option{-Wall} request
4092 When compiling C++, warn about the deprecated conversion from string
4093 literals to @code{char *}. This warning is enabled by default for C++
4098 @opindex Wno-clobbered
4099 Warn for variables that might be changed by @samp{longjmp} or
4100 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
4103 @opindex Wconversion
4104 @opindex Wno-conversion
4105 Warn for implicit conversions that may alter a value. This includes
4106 conversions between real and integer, like @code{abs (x)} when
4107 @code{x} is @code{double}; conversions between signed and unsigned,
4108 like @code{unsigned ui = -1}; and conversions to smaller types, like
4109 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4110 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4111 changed by the conversion like in @code{abs (2.0)}. Warnings about
4112 conversions between signed and unsigned integers can be disabled by
4113 using @option{-Wno-sign-conversion}.
4115 For C++, also warn for confusing overload resolution for user-defined
4116 conversions; and conversions that will never use a type conversion
4117 operator: conversions to @code{void}, the same type, a base class or a
4118 reference to them. Warnings about conversions between signed and
4119 unsigned integers are disabled by default in C++ unless
4120 @option{-Wsign-conversion} is explicitly enabled.
4122 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4123 @opindex Wconversion-null
4124 @opindex Wno-conversion-null
4125 Do not warn for conversions between @code{NULL} and non-pointer
4126 types. @option{-Wconversion-null} is enabled by default.
4129 @opindex Wempty-body
4130 @opindex Wno-empty-body
4131 Warn if an empty body occurs in an @samp{if}, @samp{else} or @samp{do
4132 while} statement. This warning is also enabled by @option{-Wextra}.
4134 @item -Wenum-compare
4135 @opindex Wenum-compare
4136 @opindex Wno-enum-compare
4137 Warn about a comparison between values of different enum types. In C++
4138 this warning is enabled by default. In C this warning is enabled by
4141 @item -Wjump-misses-init @r{(C, Objective-C only)}
4142 @opindex Wjump-misses-init
4143 @opindex Wno-jump-misses-init
4144 Warn if a @code{goto} statement or a @code{switch} statement jumps
4145 forward across the initialization of a variable, or jumps backward to a
4146 label after the variable has been initialized. This only warns about
4147 variables which are initialized when they are declared. This warning is
4148 only supported for C and Objective C; in C++ this sort of branch is an
4151 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4152 can be disabled with the @option{-Wno-jump-misses-init} option.
4154 @item -Wsign-compare
4155 @opindex Wsign-compare
4156 @opindex Wno-sign-compare
4157 @cindex warning for comparison of signed and unsigned values
4158 @cindex comparison of signed and unsigned values, warning
4159 @cindex signed and unsigned values, comparison warning
4160 Warn when a comparison between signed and unsigned values could produce
4161 an incorrect result when the signed value is converted to unsigned.
4162 This warning is also enabled by @option{-Wextra}; to get the other warnings
4163 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
4165 @item -Wsign-conversion
4166 @opindex Wsign-conversion
4167 @opindex Wno-sign-conversion
4168 Warn for implicit conversions that may change the sign of an integer
4169 value, like assigning a signed integer expression to an unsigned
4170 integer variable. An explicit cast silences the warning. In C, this
4171 option is enabled also by @option{-Wconversion}.
4175 @opindex Wno-address
4176 Warn about suspicious uses of memory addresses. These include using
4177 the address of a function in a conditional expression, such as
4178 @code{void func(void); if (func)}, and comparisons against the memory
4179 address of a string literal, such as @code{if (x == "abc")}. Such
4180 uses typically indicate a programmer error: the address of a function
4181 always evaluates to true, so their use in a conditional usually
4182 indicate that the programmer forgot the parentheses in a function
4183 call; and comparisons against string literals result in unspecified
4184 behavior and are not portable in C, so they usually indicate that the
4185 programmer intended to use @code{strcmp}. This warning is enabled by
4189 @opindex Wlogical-op
4190 @opindex Wno-logical-op
4191 Warn about suspicious uses of logical operators in expressions.
4192 This includes using logical operators in contexts where a
4193 bit-wise operator is likely to be expected.
4195 @item -Waggregate-return
4196 @opindex Waggregate-return
4197 @opindex Wno-aggregate-return
4198 Warn if any functions that return structures or unions are defined or
4199 called. (In languages where you can return an array, this also elicits
4202 @item -Wno-attributes
4203 @opindex Wno-attributes
4204 @opindex Wattributes
4205 Do not warn if an unexpected @code{__attribute__} is used, such as
4206 unrecognized attributes, function attributes applied to variables,
4207 etc. This will not stop errors for incorrect use of supported
4210 @item -Wno-builtin-macro-redefined
4211 @opindex Wno-builtin-macro-redefined
4212 @opindex Wbuiltin-macro-redefined
4213 Do not warn if certain built-in macros are redefined. This suppresses
4214 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4215 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4217 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4218 @opindex Wstrict-prototypes
4219 @opindex Wno-strict-prototypes
4220 Warn if a function is declared or defined without specifying the
4221 argument types. (An old-style function definition is permitted without
4222 a warning if preceded by a declaration which specifies the argument
4225 @item -Wold-style-declaration @r{(C and Objective-C only)}
4226 @opindex Wold-style-declaration
4227 @opindex Wno-old-style-declaration
4228 Warn for obsolescent usages, according to the C Standard, in a
4229 declaration. For example, warn if storage-class specifiers like
4230 @code{static} are not the first things in a declaration. This warning
4231 is also enabled by @option{-Wextra}.
4233 @item -Wold-style-definition @r{(C and Objective-C only)}
4234 @opindex Wold-style-definition
4235 @opindex Wno-old-style-definition
4236 Warn if an old-style function definition is used. A warning is given
4237 even if there is a previous prototype.
4239 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
4240 @opindex Wmissing-parameter-type
4241 @opindex Wno-missing-parameter-type
4242 A function parameter is declared without a type specifier in K&R-style
4249 This warning is also enabled by @option{-Wextra}.
4251 @item -Wmissing-prototypes @r{(C and Objective-C only)}
4252 @opindex Wmissing-prototypes
4253 @opindex Wno-missing-prototypes
4254 Warn if a global function is defined without a previous prototype
4255 declaration. This warning is issued even if the definition itself
4256 provides a prototype. The aim is to detect global functions that fail
4257 to be declared in header files.
4259 @item -Wmissing-declarations
4260 @opindex Wmissing-declarations
4261 @opindex Wno-missing-declarations
4262 Warn if a global function is defined without a previous declaration.
4263 Do so even if the definition itself provides a prototype.
4264 Use this option to detect global functions that are not declared in
4265 header files. In C++, no warnings are issued for function templates,
4266 or for inline functions, or for functions in anonymous namespaces.
4268 @item -Wmissing-field-initializers
4269 @opindex Wmissing-field-initializers
4270 @opindex Wno-missing-field-initializers
4274 Warn if a structure's initializer has some fields missing. For
4275 example, the following code would cause such a warning, because
4276 @code{x.h} is implicitly zero:
4279 struct s @{ int f, g, h; @};
4280 struct s x = @{ 3, 4 @};
4283 This option does not warn about designated initializers, so the following
4284 modification would not trigger a warning:
4287 struct s @{ int f, g, h; @};
4288 struct s x = @{ .f = 3, .g = 4 @};
4291 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
4292 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
4294 @item -Wmissing-format-attribute
4295 @opindex Wmissing-format-attribute
4296 @opindex Wno-missing-format-attribute
4299 Warn about function pointers which might be candidates for @code{format}
4300 attributes. Note these are only possible candidates, not absolute ones.
4301 GCC will guess that function pointers with @code{format} attributes that
4302 are used in assignment, initialization, parameter passing or return
4303 statements should have a corresponding @code{format} attribute in the
4304 resulting type. I.e.@: the left-hand side of the assignment or
4305 initialization, the type of the parameter variable, or the return type
4306 of the containing function respectively should also have a @code{format}
4307 attribute to avoid the warning.
4309 GCC will also warn about function definitions which might be
4310 candidates for @code{format} attributes. Again, these are only
4311 possible candidates. GCC will guess that @code{format} attributes
4312 might be appropriate for any function that calls a function like
4313 @code{vprintf} or @code{vscanf}, but this might not always be the
4314 case, and some functions for which @code{format} attributes are
4315 appropriate may not be detected.
4317 @item -Wno-multichar
4318 @opindex Wno-multichar
4320 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
4321 Usually they indicate a typo in the user's code, as they have
4322 implementation-defined values, and should not be used in portable code.
4324 @item -Wnormalized=<none|id|nfc|nfkc>
4325 @opindex Wnormalized=
4328 @cindex character set, input normalization
4329 In ISO C and ISO C++, two identifiers are different if they are
4330 different sequences of characters. However, sometimes when characters
4331 outside the basic ASCII character set are used, you can have two
4332 different character sequences that look the same. To avoid confusion,
4333 the ISO 10646 standard sets out some @dfn{normalization rules} which
4334 when applied ensure that two sequences that look the same are turned into
4335 the same sequence. GCC can warn you if you are using identifiers which
4336 have not been normalized; this option controls that warning.
4338 There are four levels of warning that GCC supports. The default is
4339 @option{-Wnormalized=nfc}, which warns about any identifier which is
4340 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
4341 recommended form for most uses.
4343 Unfortunately, there are some characters which ISO C and ISO C++ allow
4344 in identifiers that when turned into NFC aren't allowable as
4345 identifiers. That is, there's no way to use these symbols in portable
4346 ISO C or C++ and have all your identifiers in NFC@.
4347 @option{-Wnormalized=id} suppresses the warning for these characters.
4348 It is hoped that future versions of the standards involved will correct
4349 this, which is why this option is not the default.
4351 You can switch the warning off for all characters by writing
4352 @option{-Wnormalized=none}. You would only want to do this if you
4353 were using some other normalization scheme (like ``D''), because
4354 otherwise you can easily create bugs that are literally impossible to see.
4356 Some characters in ISO 10646 have distinct meanings but look identical
4357 in some fonts or display methodologies, especially once formatting has
4358 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
4359 LETTER N'', will display just like a regular @code{n} which has been
4360 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
4361 normalization scheme to convert all these into a standard form as
4362 well, and GCC will warn if your code is not in NFKC if you use
4363 @option{-Wnormalized=nfkc}. This warning is comparable to warning
4364 about every identifier that contains the letter O because it might be
4365 confused with the digit 0, and so is not the default, but may be
4366 useful as a local coding convention if the programming environment is
4367 unable to be fixed to display these characters distinctly.
4369 @item -Wno-deprecated
4370 @opindex Wno-deprecated
4371 @opindex Wdeprecated
4372 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
4374 @item -Wno-deprecated-declarations
4375 @opindex Wno-deprecated-declarations
4376 @opindex Wdeprecated-declarations
4377 Do not warn about uses of functions (@pxref{Function Attributes}),
4378 variables (@pxref{Variable Attributes}), and types (@pxref{Type
4379 Attributes}) marked as deprecated by using the @code{deprecated}
4383 @opindex Wno-overflow
4385 Do not warn about compile-time overflow in constant expressions.
4387 @item -Woverride-init @r{(C and Objective-C only)}
4388 @opindex Woverride-init
4389 @opindex Wno-override-init
4393 Warn if an initialized field without side effects is overridden when
4394 using designated initializers (@pxref{Designated Inits, , Designated
4397 This warning is included in @option{-Wextra}. To get other
4398 @option{-Wextra} warnings without this one, use @samp{-Wextra
4399 -Wno-override-init}.
4404 Warn if a structure is given the packed attribute, but the packed
4405 attribute has no effect on the layout or size of the structure.
4406 Such structures may be mis-aligned for little benefit. For
4407 instance, in this code, the variable @code{f.x} in @code{struct bar}
4408 will be misaligned even though @code{struct bar} does not itself
4409 have the packed attribute:
4416 @} __attribute__((packed));
4424 @item -Wpacked-bitfield-compat
4425 @opindex Wpacked-bitfield-compat
4426 @opindex Wno-packed-bitfield-compat
4427 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
4428 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
4429 the change can lead to differences in the structure layout. GCC
4430 informs you when the offset of such a field has changed in GCC 4.4.
4431 For example there is no longer a 4-bit padding between field @code{a}
4432 and @code{b} in this structure:
4439 @} __attribute__ ((packed));
4442 This warning is enabled by default. Use
4443 @option{-Wno-packed-bitfield-compat} to disable this warning.
4448 Warn if padding is included in a structure, either to align an element
4449 of the structure or to align the whole structure. Sometimes when this
4450 happens it is possible to rearrange the fields of the structure to
4451 reduce the padding and so make the structure smaller.
4453 @item -Wredundant-decls
4454 @opindex Wredundant-decls
4455 @opindex Wno-redundant-decls
4456 Warn if anything is declared more than once in the same scope, even in
4457 cases where multiple declaration is valid and changes nothing.
4459 @item -Wnested-externs @r{(C and Objective-C only)}
4460 @opindex Wnested-externs
4461 @opindex Wno-nested-externs
4462 Warn if an @code{extern} declaration is encountered within a function.
4467 Warn if a function can not be inlined and it was declared as inline.
4468 Even with this option, the compiler will not warn about failures to
4469 inline functions declared in system headers.
4471 The compiler uses a variety of heuristics to determine whether or not
4472 to inline a function. For example, the compiler takes into account
4473 the size of the function being inlined and the amount of inlining
4474 that has already been done in the current function. Therefore,
4475 seemingly insignificant changes in the source program can cause the
4476 warnings produced by @option{-Winline} to appear or disappear.
4478 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
4479 @opindex Wno-invalid-offsetof
4480 @opindex Winvalid-offsetof
4481 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
4482 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
4483 to a non-POD type is undefined. In existing C++ implementations,
4484 however, @samp{offsetof} typically gives meaningful results even when
4485 applied to certain kinds of non-POD types. (Such as a simple
4486 @samp{struct} that fails to be a POD type only by virtue of having a
4487 constructor.) This flag is for users who are aware that they are
4488 writing nonportable code and who have deliberately chosen to ignore the
4491 The restrictions on @samp{offsetof} may be relaxed in a future version
4492 of the C++ standard.
4494 @item -Wno-int-to-pointer-cast
4495 @opindex Wno-int-to-pointer-cast
4496 @opindex Wint-to-pointer-cast
4497 Suppress warnings from casts to pointer type of an integer of a
4498 different size. In C++, casting to a pointer type of smaller size is
4499 an error. @option{Wint-to-pointer-cast} is enabled by default.
4502 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
4503 @opindex Wno-pointer-to-int-cast
4504 @opindex Wpointer-to-int-cast
4505 Suppress warnings from casts from a pointer to an integer type of a
4509 @opindex Winvalid-pch
4510 @opindex Wno-invalid-pch
4511 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
4512 the search path but can't be used.
4516 @opindex Wno-long-long
4517 Warn if @samp{long long} type is used. This is enabled by either
4518 @option{-pedantic} or @option{-Wtraditional} in ISO C90 and C++98
4519 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
4521 @item -Wvariadic-macros
4522 @opindex Wvariadic-macros
4523 @opindex Wno-variadic-macros
4524 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
4525 alternate syntax when in pedantic ISO C99 mode. This is default.
4526 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
4531 Warn if variable length array is used in the code.
4532 @option{-Wno-vla} will prevent the @option{-pedantic} warning of
4533 the variable length array.
4535 @item -Wvolatile-register-var
4536 @opindex Wvolatile-register-var
4537 @opindex Wno-volatile-register-var
4538 Warn if a register variable is declared volatile. The volatile
4539 modifier does not inhibit all optimizations that may eliminate reads
4540 and/or writes to register variables. This warning is enabled by
4543 @item -Wdisabled-optimization
4544 @opindex Wdisabled-optimization
4545 @opindex Wno-disabled-optimization
4546 Warn if a requested optimization pass is disabled. This warning does
4547 not generally indicate that there is anything wrong with your code; it
4548 merely indicates that GCC's optimizers were unable to handle the code
4549 effectively. Often, the problem is that your code is too big or too
4550 complex; GCC will refuse to optimize programs when the optimization
4551 itself is likely to take inordinate amounts of time.
4553 @item -Wpointer-sign @r{(C and Objective-C only)}
4554 @opindex Wpointer-sign
4555 @opindex Wno-pointer-sign
4556 Warn for pointer argument passing or assignment with different signedness.
4557 This option is only supported for C and Objective-C@. It is implied by
4558 @option{-Wall} and by @option{-pedantic}, which can be disabled with
4559 @option{-Wno-pointer-sign}.
4561 @item -Wstack-protector
4562 @opindex Wstack-protector
4563 @opindex Wno-stack-protector
4564 This option is only active when @option{-fstack-protector} is active. It
4565 warns about functions that will not be protected against stack smashing.
4568 @opindex Wno-mudflap
4569 Suppress warnings about constructs that cannot be instrumented by
4572 @item -Woverlength-strings
4573 @opindex Woverlength-strings
4574 @opindex Wno-overlength-strings
4575 Warn about string constants which are longer than the ``minimum
4576 maximum'' length specified in the C standard. Modern compilers
4577 generally allow string constants which are much longer than the
4578 standard's minimum limit, but very portable programs should avoid
4579 using longer strings.
4581 The limit applies @emph{after} string constant concatenation, and does
4582 not count the trailing NUL@. In C90, the limit was 509 characters; in
4583 C99, it was raised to 4095. C++98 does not specify a normative
4584 minimum maximum, so we do not diagnose overlength strings in C++@.
4586 This option is implied by @option{-pedantic}, and can be disabled with
4587 @option{-Wno-overlength-strings}.
4589 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
4590 @opindex Wunsuffixed-float-constants
4592 GCC will issue a warning for any floating constant that does not have
4593 a suffix. When used together with @option{-Wsystem-headers} it will
4594 warn about such constants in system header files. This can be useful
4595 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
4596 from the decimal floating-point extension to C99.
4599 @node Debugging Options
4600 @section Options for Debugging Your Program or GCC
4601 @cindex options, debugging
4602 @cindex debugging information options
4604 GCC has various special options that are used for debugging
4605 either your program or GCC:
4610 Produce debugging information in the operating system's native format
4611 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
4614 On most systems that use stabs format, @option{-g} enables use of extra
4615 debugging information that only GDB can use; this extra information
4616 makes debugging work better in GDB but will probably make other debuggers
4618 refuse to read the program. If you want to control for certain whether
4619 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
4620 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
4622 GCC allows you to use @option{-g} with
4623 @option{-O}. The shortcuts taken by optimized code may occasionally
4624 produce surprising results: some variables you declared may not exist
4625 at all; flow of control may briefly move where you did not expect it;
4626 some statements may not be executed because they compute constant
4627 results or their values were already at hand; some statements may
4628 execute in different places because they were moved out of loops.
4630 Nevertheless it proves possible to debug optimized output. This makes
4631 it reasonable to use the optimizer for programs that might have bugs.
4633 The following options are useful when GCC is generated with the
4634 capability for more than one debugging format.
4638 Produce debugging information for use by GDB@. This means to use the
4639 most expressive format available (DWARF 2, stabs, or the native format
4640 if neither of those are supported), including GDB extensions if at all
4645 Produce debugging information in stabs format (if that is supported),
4646 without GDB extensions. This is the format used by DBX on most BSD
4647 systems. On MIPS, Alpha and System V Release 4 systems this option
4648 produces stabs debugging output which is not understood by DBX or SDB@.
4649 On System V Release 4 systems this option requires the GNU assembler.
4651 @item -feliminate-unused-debug-symbols
4652 @opindex feliminate-unused-debug-symbols
4653 Produce debugging information in stabs format (if that is supported),
4654 for only symbols that are actually used.
4656 @item -femit-class-debug-always
4657 Instead of emitting debugging information for a C++ class in only one
4658 object file, emit it in all object files using the class. This option
4659 should be used only with debuggers that are unable to handle the way GCC
4660 normally emits debugging information for classes because using this
4661 option will increase the size of debugging information by as much as a
4664 @item -fno-debug-types-section
4665 @opindex fno-debug-types-section
4666 @opindex fdebug-types-section
4667 By default when using DWARF v4 or higher type DIEs will be put into
4668 their own .debug_types section instead of making them part of the
4669 .debug_info section. It is more efficient to put them in a separate
4670 comdat sections since the linker will then be able to remove duplicates.
4671 But not all DWARF consumers support .debug_types sections yet.
4675 Produce debugging information in stabs format (if that is supported),
4676 using GNU extensions understood only by the GNU debugger (GDB)@. The
4677 use of these extensions is likely to make other debuggers crash or
4678 refuse to read the program.
4682 Produce debugging information in COFF format (if that is supported).
4683 This is the format used by SDB on most System V systems prior to
4688 Produce debugging information in XCOFF format (if that is supported).
4689 This is the format used by the DBX debugger on IBM RS/6000 systems.
4693 Produce debugging information in XCOFF format (if that is supported),
4694 using GNU extensions understood only by the GNU debugger (GDB)@. The
4695 use of these extensions is likely to make other debuggers crash or
4696 refuse to read the program, and may cause assemblers other than the GNU
4697 assembler (GAS) to fail with an error.
4699 @item -gdwarf-@var{version}
4700 @opindex gdwarf-@var{version}
4701 Produce debugging information in DWARF format (if that is
4702 supported). This is the format used by DBX on IRIX 6. The value
4703 of @var{version} may be either 2, 3 or 4; the default version is 2.
4705 Note that with DWARF version 2 some ports require, and will always
4706 use, some non-conflicting DWARF 3 extensions in the unwind tables.
4708 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
4709 for maximum benefit.
4711 @item -grecord-gcc-switches
4712 @opindex grecord-gcc-switches
4713 This switch causes the command line options, that were used to invoke the
4714 compiler and may affect code generation, to be appended to the
4715 DW_AT_producer attribute in DWARF debugging information. The options
4716 are concatenated with spaces separating them from each other and from
4717 the compiler version. See also @option{-frecord-gcc-switches} for another
4718 way of storing compiler options into the object file.
4720 @item -gno-record-gcc-switches
4721 @opindex gno-record-gcc-switches
4722 Disallow appending command line options to the DW_AT_producer attribute
4723 in DWARF debugging information. This is the default.
4725 @item -gstrict-dwarf
4726 @opindex gstrict-dwarf
4727 Disallow using extensions of later DWARF standard version than selected
4728 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
4729 DWARF extensions from later standard versions is allowed.
4731 @item -gno-strict-dwarf
4732 @opindex gno-strict-dwarf
4733 Allow using extensions of later DWARF standard version than selected with
4734 @option{-gdwarf-@var{version}}.
4738 Produce debugging information in VMS debug format (if that is
4739 supported). This is the format used by DEBUG on VMS systems.
4742 @itemx -ggdb@var{level}
4743 @itemx -gstabs@var{level}
4744 @itemx -gcoff@var{level}
4745 @itemx -gxcoff@var{level}
4746 @itemx -gvms@var{level}
4747 Request debugging information and also use @var{level} to specify how
4748 much information. The default level is 2.
4750 Level 0 produces no debug information at all. Thus, @option{-g0} negates
4753 Level 1 produces minimal information, enough for making backtraces in
4754 parts of the program that you don't plan to debug. This includes
4755 descriptions of functions and external variables, but no information
4756 about local variables and no line numbers.
4758 Level 3 includes extra information, such as all the macro definitions
4759 present in the program. Some debuggers support macro expansion when
4760 you use @option{-g3}.
4762 @option{-gdwarf-2} does not accept a concatenated debug level, because
4763 GCC used to support an option @option{-gdwarf} that meant to generate
4764 debug information in version 1 of the DWARF format (which is very
4765 different from version 2), and it would have been too confusing. That
4766 debug format is long obsolete, but the option cannot be changed now.
4767 Instead use an additional @option{-g@var{level}} option to change the
4768 debug level for DWARF.
4772 Turn off generation of debug info, if leaving out this option would have
4773 generated it, or turn it on at level 2 otherwise. The position of this
4774 argument in the command line does not matter, it takes effect after all
4775 other options are processed, and it does so only once, no matter how
4776 many times it is given. This is mainly intended to be used with
4777 @option{-fcompare-debug}.
4779 @item -fdump-final-insns@r{[}=@var{file}@r{]}
4780 @opindex fdump-final-insns
4781 Dump the final internal representation (RTL) to @var{file}. If the
4782 optional argument is omitted (or if @var{file} is @code{.}), the name
4783 of the dump file will be determined by appending @code{.gkd} to the
4784 compilation output file name.
4786 @item -fcompare-debug@r{[}=@var{opts}@r{]}
4787 @opindex fcompare-debug
4788 @opindex fno-compare-debug
4789 If no error occurs during compilation, run the compiler a second time,
4790 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
4791 passed to the second compilation. Dump the final internal
4792 representation in both compilations, and print an error if they differ.
4794 If the equal sign is omitted, the default @option{-gtoggle} is used.
4796 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
4797 and nonzero, implicitly enables @option{-fcompare-debug}. If
4798 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
4799 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
4802 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
4803 is equivalent to @option{-fno-compare-debug}, which disables the dumping
4804 of the final representation and the second compilation, preventing even
4805 @env{GCC_COMPARE_DEBUG} from taking effect.
4807 To verify full coverage during @option{-fcompare-debug} testing, set
4808 @env{GCC_COMPARE_DEBUG} to say @samp{-fcompare-debug-not-overridden},
4809 which GCC will reject as an invalid option in any actual compilation
4810 (rather than preprocessing, assembly or linking). To get just a
4811 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
4812 not overridden} will do.
4814 @item -fcompare-debug-second
4815 @opindex fcompare-debug-second
4816 This option is implicitly passed to the compiler for the second
4817 compilation requested by @option{-fcompare-debug}, along with options to
4818 silence warnings, and omitting other options that would cause
4819 side-effect compiler outputs to files or to the standard output. Dump
4820 files and preserved temporary files are renamed so as to contain the
4821 @code{.gk} additional extension during the second compilation, to avoid
4822 overwriting those generated by the first.
4824 When this option is passed to the compiler driver, it causes the
4825 @emph{first} compilation to be skipped, which makes it useful for little
4826 other than debugging the compiler proper.
4828 @item -feliminate-dwarf2-dups
4829 @opindex feliminate-dwarf2-dups
4830 Compress DWARF2 debugging information by eliminating duplicated
4831 information about each symbol. This option only makes sense when
4832 generating DWARF2 debugging information with @option{-gdwarf-2}.
4834 @item -femit-struct-debug-baseonly
4835 Emit debug information for struct-like types
4836 only when the base name of the compilation source file
4837 matches the base name of file in which the struct was defined.
4839 This option substantially reduces the size of debugging information,
4840 but at significant potential loss in type information to the debugger.
4841 See @option{-femit-struct-debug-reduced} for a less aggressive option.
4842 See @option{-femit-struct-debug-detailed} for more detailed control.
4844 This option works only with DWARF 2.
4846 @item -femit-struct-debug-reduced
4847 Emit debug information for struct-like types
4848 only when the base name of the compilation source file
4849 matches the base name of file in which the type was defined,
4850 unless the struct is a template or defined in a system header.
4852 This option significantly reduces the size of debugging information,
4853 with some potential loss in type information to the debugger.
4854 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
4855 See @option{-femit-struct-debug-detailed} for more detailed control.
4857 This option works only with DWARF 2.
4859 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
4860 Specify the struct-like types
4861 for which the compiler will generate debug information.
4862 The intent is to reduce duplicate struct debug information
4863 between different object files within the same program.
4865 This option is a detailed version of
4866 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
4867 which will serve for most needs.
4869 A specification has the syntax@*
4870 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
4872 The optional first word limits the specification to
4873 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
4874 A struct type is used directly when it is the type of a variable, member.
4875 Indirect uses arise through pointers to structs.
4876 That is, when use of an incomplete struct would be legal, the use is indirect.
4878 @samp{struct one direct; struct two * indirect;}.
4880 The optional second word limits the specification to
4881 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
4882 Generic structs are a bit complicated to explain.
4883 For C++, these are non-explicit specializations of template classes,
4884 or non-template classes within the above.
4885 Other programming languages have generics,
4886 but @samp{-femit-struct-debug-detailed} does not yet implement them.
4888 The third word specifies the source files for those
4889 structs for which the compiler will emit debug information.
4890 The values @samp{none} and @samp{any} have the normal meaning.
4891 The value @samp{base} means that
4892 the base of name of the file in which the type declaration appears
4893 must match the base of the name of the main compilation file.
4894 In practice, this means that
4895 types declared in @file{foo.c} and @file{foo.h} will have debug information,
4896 but types declared in other header will not.
4897 The value @samp{sys} means those types satisfying @samp{base}
4898 or declared in system or compiler headers.
4900 You may need to experiment to determine the best settings for your application.
4902 The default is @samp{-femit-struct-debug-detailed=all}.
4904 This option works only with DWARF 2.
4906 @item -fno-merge-debug-strings
4907 @opindex fmerge-debug-strings
4908 @opindex fno-merge-debug-strings
4909 Direct the linker to not merge together strings in the debugging
4910 information which are identical in different object files. Merging is
4911 not supported by all assemblers or linkers. Merging decreases the size
4912 of the debug information in the output file at the cost of increasing
4913 link processing time. Merging is enabled by default.
4915 @item -fdebug-prefix-map=@var{old}=@var{new}
4916 @opindex fdebug-prefix-map
4917 When compiling files in directory @file{@var{old}}, record debugging
4918 information describing them as in @file{@var{new}} instead.
4920 @item -fno-dwarf2-cfi-asm
4921 @opindex fdwarf2-cfi-asm
4922 @opindex fno-dwarf2-cfi-asm
4923 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
4924 instead of using GAS @code{.cfi_*} directives.
4926 @cindex @command{prof}
4929 Generate extra code to write profile information suitable for the
4930 analysis program @command{prof}. You must use this option when compiling
4931 the source files you want data about, and you must also use it when
4934 @cindex @command{gprof}
4937 Generate extra code to write profile information suitable for the
4938 analysis program @command{gprof}. You must use this option when compiling
4939 the source files you want data about, and you must also use it when
4944 Makes the compiler print out each function name as it is compiled, and
4945 print some statistics about each pass when it finishes.
4948 @opindex ftime-report
4949 Makes the compiler print some statistics about the time consumed by each
4950 pass when it finishes.
4953 @opindex fmem-report
4954 Makes the compiler print some statistics about permanent memory
4955 allocation when it finishes.
4957 @item -fpre-ipa-mem-report
4958 @opindex fpre-ipa-mem-report
4959 @item -fpost-ipa-mem-report
4960 @opindex fpost-ipa-mem-report
4961 Makes the compiler print some statistics about permanent memory
4962 allocation before or after interprocedural optimization.
4965 @opindex fstack-usage
4966 Makes the compiler output stack usage information for the program, on a
4967 per-function basis. The filename for the dump is made by appending
4968 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
4969 the output file, if explicitly specified and it is not an executable,
4970 otherwise it is the basename of the source file. An entry is made up
4975 The name of the function.
4979 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
4982 The qualifier @code{static} means that the function manipulates the stack
4983 statically: a fixed number of bytes are allocated for the frame on function
4984 entry and released on function exit; no stack adjustments are otherwise made
4985 in the function. The second field is this fixed number of bytes.
4987 The qualifier @code{dynamic} means that the function manipulates the stack
4988 dynamically: in addition to the static allocation described above, stack
4989 adjustments are made in the body of the function, for example to push/pop
4990 arguments around function calls. If the qualifier @code{bounded} is also
4991 present, the amount of these adjustments is bounded at compile-time and
4992 the second field is an upper bound of the total amount of stack used by
4993 the function. If it is not present, the amount of these adjustments is
4994 not bounded at compile-time and the second field only represents the
4997 @item -fprofile-arcs
4998 @opindex fprofile-arcs
4999 Add code so that program flow @dfn{arcs} are instrumented. During
5000 execution the program records how many times each branch and call is
5001 executed and how many times it is taken or returns. When the compiled
5002 program exits it saves this data to a file called
5003 @file{@var{auxname}.gcda} for each source file. The data may be used for
5004 profile-directed optimizations (@option{-fbranch-probabilities}), or for
5005 test coverage analysis (@option{-ftest-coverage}). Each object file's
5006 @var{auxname} is generated from the name of the output file, if
5007 explicitly specified and it is not the final executable, otherwise it is
5008 the basename of the source file. In both cases any suffix is removed
5009 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
5010 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
5011 @xref{Cross-profiling}.
5013 @cindex @command{gcov}
5017 This option is used to compile and link code instrumented for coverage
5018 analysis. The option is a synonym for @option{-fprofile-arcs}
5019 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
5020 linking). See the documentation for those options for more details.
5025 Compile the source files with @option{-fprofile-arcs} plus optimization
5026 and code generation options. For test coverage analysis, use the
5027 additional @option{-ftest-coverage} option. You do not need to profile
5028 every source file in a program.
5031 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
5032 (the latter implies the former).
5035 Run the program on a representative workload to generate the arc profile
5036 information. This may be repeated any number of times. You can run
5037 concurrent instances of your program, and provided that the file system
5038 supports locking, the data files will be correctly updated. Also
5039 @code{fork} calls are detected and correctly handled (double counting
5043 For profile-directed optimizations, compile the source files again with
5044 the same optimization and code generation options plus
5045 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
5046 Control Optimization}).
5049 For test coverage analysis, use @command{gcov} to produce human readable
5050 information from the @file{.gcno} and @file{.gcda} files. Refer to the
5051 @command{gcov} documentation for further information.
5055 With @option{-fprofile-arcs}, for each function of your program GCC
5056 creates a program flow graph, then finds a spanning tree for the graph.
5057 Only arcs that are not on the spanning tree have to be instrumented: the
5058 compiler adds code to count the number of times that these arcs are
5059 executed. When an arc is the only exit or only entrance to a block, the
5060 instrumentation code can be added to the block; otherwise, a new basic
5061 block must be created to hold the instrumentation code.
5064 @item -ftest-coverage
5065 @opindex ftest-coverage
5066 Produce a notes file that the @command{gcov} code-coverage utility
5067 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
5068 show program coverage. Each source file's note file is called
5069 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
5070 above for a description of @var{auxname} and instructions on how to
5071 generate test coverage data. Coverage data will match the source files
5072 more closely, if you do not optimize.
5074 @item -fdbg-cnt-list
5075 @opindex fdbg-cnt-list
5076 Print the name and the counter upper bound for all debug counters.
5079 @item -fdbg-cnt=@var{counter-value-list}
5081 Set the internal debug counter upper bound. @var{counter-value-list}
5082 is a comma-separated list of @var{name}:@var{value} pairs
5083 which sets the upper bound of each debug counter @var{name} to @var{value}.
5084 All debug counters have the initial upper bound of @var{UINT_MAX},
5085 thus dbg_cnt() returns true always unless the upper bound is set by this option.
5086 e.g. With -fdbg-cnt=dce:10,tail_call:0
5087 dbg_cnt(dce) will return true only for first 10 invocations
5089 @itemx -fenable-@var{kind}-@var{pass}
5090 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
5094 This is a set of debugging options that are used to explicitly disable/enable
5095 optimization passes. For compiler users, regular options for enabling/disabling
5096 passes should be used instead.
5100 @item -fdisable-ipa-@var{pass}
5101 Disable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5102 statically invoked in the compiler multiple times, the pass name should be
5103 appended with a sequential number starting from 1.
5105 @item -fdisable-rtl-@var{pass}
5106 @item -fdisable-rtl-@var{pass}=@var{range-list}
5107 Disable rtl pass @var{pass}. @var{pass} is the pass name. If the same pass is
5108 statically invoked in the compiler multiple times, the pass name should be
5109 appended with a sequential number starting from 1. @var{range-list} is a comma
5110 seperated list of function ranges or assembler names. Each range is a number
5111 pair seperated by a colon. The range is inclusive in both ends. If the range
5112 is trivial, the number pair can be simplified as a single number. If the
5113 function's cgraph node's @var{uid} is falling within one of the specified ranges,
5114 the @var{pass} is disabled for that function. The @var{uid} is shown in the
5115 function header of a dump file, and the pass names can be dumped by using
5116 option @option{-fdump-passes}.
5118 @item -fdisable-tree-@var{pass}
5119 @item -fdisable-tree-@var{pass}=@var{range-list}
5120 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
5123 @item -fenable-ipa-@var{pass}
5124 Enable ipa pass @var{pass}. @var{pass} is the pass name. If the same pass is
5125 statically invoked in the compiler multiple times, the pass name should be
5126 appended with a sequential number starting from 1.
5128 @item -fenable-rtl-@var{pass}
5129 @item -fenable-rtl-@var{pass}=@var{range-list}
5130 Enable rtl pass @var{pass}. See @option{-fdisable-rtl} for option argument
5131 description and examples.
5133 @item -fenable-tree-@var{pass}
5134 @item -fenable-tree-@var{pass}=@var{range-list}
5135 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
5136 of option arguments.
5140 # disable ccp1 for all functions
5142 # disable complete unroll for function whose cgraph node uid is 1
5143 -fenable-tree-cunroll=1
5144 # disable gcse2 for functions at the following ranges [1,1],
5145 # [300,400], and [400,1000]
5146 # disable gcse2 for functions foo and foo2
5147 -fdisable-rtl-gcse2=foo,foo2
5148 # disable early inlining
5149 -fdisable-tree-einline
5150 # disable ipa inlining
5151 -fdisable-ipa-inline
5152 # enable tree full unroll
5153 -fenable-tree-unroll
5159 @item -d@var{letters}
5160 @itemx -fdump-rtl-@var{pass}
5162 Says to make debugging dumps during compilation at times specified by
5163 @var{letters}. This is used for debugging the RTL-based passes of the
5164 compiler. The file names for most of the dumps are made by appending
5165 a pass number and a word to the @var{dumpname}, and the files are
5166 created in the directory of the output file. Note that the pass
5167 number is computed statically as passes get registered into the pass
5168 manager. Thus the numbering is not related to the dynamic order of
5169 execution of passes. In particular, a pass installed by a plugin
5170 could have a number over 200 even if it executed quite early.
5171 @var{dumpname} is generated from the name of the output file, if
5172 explicitly specified and it is not an executable, otherwise it is the
5173 basename of the source file. These switches may have different effects
5174 when @option{-E} is used for preprocessing.
5176 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
5177 @option{-d} option @var{letters}. Here are the possible
5178 letters for use in @var{pass} and @var{letters}, and their meanings:
5182 @item -fdump-rtl-alignments
5183 @opindex fdump-rtl-alignments
5184 Dump after branch alignments have been computed.
5186 @item -fdump-rtl-asmcons
5187 @opindex fdump-rtl-asmcons
5188 Dump after fixing rtl statements that have unsatisfied in/out constraints.
5190 @item -fdump-rtl-auto_inc_dec
5191 @opindex fdump-rtl-auto_inc_dec
5192 Dump after auto-inc-dec discovery. This pass is only run on
5193 architectures that have auto inc or auto dec instructions.
5195 @item -fdump-rtl-barriers
5196 @opindex fdump-rtl-barriers
5197 Dump after cleaning up the barrier instructions.
5199 @item -fdump-rtl-bbpart
5200 @opindex fdump-rtl-bbpart
5201 Dump after partitioning hot and cold basic blocks.
5203 @item -fdump-rtl-bbro
5204 @opindex fdump-rtl-bbro
5205 Dump after block reordering.
5207 @item -fdump-rtl-btl1
5208 @itemx -fdump-rtl-btl2
5209 @opindex fdump-rtl-btl2
5210 @opindex fdump-rtl-btl2
5211 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
5212 after the two branch
5213 target load optimization passes.
5215 @item -fdump-rtl-bypass
5216 @opindex fdump-rtl-bypass
5217 Dump after jump bypassing and control flow optimizations.
5219 @item -fdump-rtl-combine
5220 @opindex fdump-rtl-combine
5221 Dump after the RTL instruction combination pass.
5223 @item -fdump-rtl-compgotos
5224 @opindex fdump-rtl-compgotos
5225 Dump after duplicating the computed gotos.
5227 @item -fdump-rtl-ce1
5228 @itemx -fdump-rtl-ce2
5229 @itemx -fdump-rtl-ce3
5230 @opindex fdump-rtl-ce1
5231 @opindex fdump-rtl-ce2
5232 @opindex fdump-rtl-ce3
5233 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
5234 @option{-fdump-rtl-ce3} enable dumping after the three
5235 if conversion passes.
5237 @itemx -fdump-rtl-cprop_hardreg
5238 @opindex fdump-rtl-cprop_hardreg
5239 Dump after hard register copy propagation.
5241 @itemx -fdump-rtl-csa
5242 @opindex fdump-rtl-csa
5243 Dump after combining stack adjustments.
5245 @item -fdump-rtl-cse1
5246 @itemx -fdump-rtl-cse2
5247 @opindex fdump-rtl-cse1
5248 @opindex fdump-rtl-cse2
5249 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
5250 the two common sub-expression elimination passes.
5252 @itemx -fdump-rtl-dce
5253 @opindex fdump-rtl-dce
5254 Dump after the standalone dead code elimination passes.
5256 @itemx -fdump-rtl-dbr
5257 @opindex fdump-rtl-dbr
5258 Dump after delayed branch scheduling.
5260 @item -fdump-rtl-dce1
5261 @itemx -fdump-rtl-dce2
5262 @opindex fdump-rtl-dce1
5263 @opindex fdump-rtl-dce2
5264 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
5265 the two dead store elimination passes.
5268 @opindex fdump-rtl-eh
5269 Dump after finalization of EH handling code.
5271 @item -fdump-rtl-eh_ranges
5272 @opindex fdump-rtl-eh_ranges
5273 Dump after conversion of EH handling range regions.
5275 @item -fdump-rtl-expand
5276 @opindex fdump-rtl-expand
5277 Dump after RTL generation.
5279 @item -fdump-rtl-fwprop1
5280 @itemx -fdump-rtl-fwprop2
5281 @opindex fdump-rtl-fwprop1
5282 @opindex fdump-rtl-fwprop2
5283 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
5284 dumping after the two forward propagation passes.
5286 @item -fdump-rtl-gcse1
5287 @itemx -fdump-rtl-gcse2
5288 @opindex fdump-rtl-gcse1
5289 @opindex fdump-rtl-gcse2
5290 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
5291 after global common subexpression elimination.
5293 @item -fdump-rtl-init-regs
5294 @opindex fdump-rtl-init-regs
5295 Dump after the initialization of the registers.
5297 @item -fdump-rtl-initvals
5298 @opindex fdump-rtl-initvals
5299 Dump after the computation of the initial value sets.
5301 @itemx -fdump-rtl-into_cfglayout
5302 @opindex fdump-rtl-into_cfglayout
5303 Dump after converting to cfglayout mode.
5305 @item -fdump-rtl-ira
5306 @opindex fdump-rtl-ira
5307 Dump after iterated register allocation.
5309 @item -fdump-rtl-jump
5310 @opindex fdump-rtl-jump
5311 Dump after the second jump optimization.
5313 @item -fdump-rtl-loop2
5314 @opindex fdump-rtl-loop2
5315 @option{-fdump-rtl-loop2} enables dumping after the rtl
5316 loop optimization passes.
5318 @item -fdump-rtl-mach
5319 @opindex fdump-rtl-mach
5320 Dump after performing the machine dependent reorganization pass, if that
5323 @item -fdump-rtl-mode_sw
5324 @opindex fdump-rtl-mode_sw
5325 Dump after removing redundant mode switches.
5327 @item -fdump-rtl-rnreg
5328 @opindex fdump-rtl-rnreg
5329 Dump after register renumbering.
5331 @itemx -fdump-rtl-outof_cfglayout
5332 @opindex fdump-rtl-outof_cfglayout
5333 Dump after converting from cfglayout mode.
5335 @item -fdump-rtl-peephole2
5336 @opindex fdump-rtl-peephole2
5337 Dump after the peephole pass.
5339 @item -fdump-rtl-postreload
5340 @opindex fdump-rtl-postreload
5341 Dump after post-reload optimizations.
5343 @itemx -fdump-rtl-pro_and_epilogue
5344 @opindex fdump-rtl-pro_and_epilogue
5345 Dump after generating the function pro and epilogues.
5347 @item -fdump-rtl-regmove
5348 @opindex fdump-rtl-regmove
5349 Dump after the register move pass.
5351 @item -fdump-rtl-sched1
5352 @itemx -fdump-rtl-sched2
5353 @opindex fdump-rtl-sched1
5354 @opindex fdump-rtl-sched2
5355 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
5356 after the basic block scheduling passes.
5358 @item -fdump-rtl-see
5359 @opindex fdump-rtl-see
5360 Dump after sign extension elimination.
5362 @item -fdump-rtl-seqabstr
5363 @opindex fdump-rtl-seqabstr
5364 Dump after common sequence discovery.
5366 @item -fdump-rtl-shorten
5367 @opindex fdump-rtl-shorten
5368 Dump after shortening branches.
5370 @item -fdump-rtl-sibling
5371 @opindex fdump-rtl-sibling
5372 Dump after sibling call optimizations.
5374 @item -fdump-rtl-split1
5375 @itemx -fdump-rtl-split2
5376 @itemx -fdump-rtl-split3
5377 @itemx -fdump-rtl-split4
5378 @itemx -fdump-rtl-split5
5379 @opindex fdump-rtl-split1
5380 @opindex fdump-rtl-split2
5381 @opindex fdump-rtl-split3
5382 @opindex fdump-rtl-split4
5383 @opindex fdump-rtl-split5
5384 @option{-fdump-rtl-split1}, @option{-fdump-rtl-split2},
5385 @option{-fdump-rtl-split3}, @option{-fdump-rtl-split4} and
5386 @option{-fdump-rtl-split5} enable dumping after five rounds of
5387 instruction splitting.
5389 @item -fdump-rtl-sms
5390 @opindex fdump-rtl-sms
5391 Dump after modulo scheduling. This pass is only run on some
5394 @item -fdump-rtl-stack
5395 @opindex fdump-rtl-stack
5396 Dump after conversion from GCC's "flat register file" registers to the
5397 x87's stack-like registers. This pass is only run on x86 variants.
5399 @item -fdump-rtl-subreg1
5400 @itemx -fdump-rtl-subreg2
5401 @opindex fdump-rtl-subreg1
5402 @opindex fdump-rtl-subreg2
5403 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
5404 the two subreg expansion passes.
5406 @item -fdump-rtl-unshare
5407 @opindex fdump-rtl-unshare
5408 Dump after all rtl has been unshared.
5410 @item -fdump-rtl-vartrack
5411 @opindex fdump-rtl-vartrack
5412 Dump after variable tracking.
5414 @item -fdump-rtl-vregs
5415 @opindex fdump-rtl-vregs
5416 Dump after converting virtual registers to hard registers.
5418 @item -fdump-rtl-web
5419 @opindex fdump-rtl-web
5420 Dump after live range splitting.
5422 @item -fdump-rtl-regclass
5423 @itemx -fdump-rtl-subregs_of_mode_init
5424 @itemx -fdump-rtl-subregs_of_mode_finish
5425 @itemx -fdump-rtl-dfinit
5426 @itemx -fdump-rtl-dfinish
5427 @opindex fdump-rtl-regclass
5428 @opindex fdump-rtl-subregs_of_mode_init
5429 @opindex fdump-rtl-subregs_of_mode_finish
5430 @opindex fdump-rtl-dfinit
5431 @opindex fdump-rtl-dfinish
5432 These dumps are defined but always produce empty files.
5434 @item -fdump-rtl-all
5435 @opindex fdump-rtl-all
5436 Produce all the dumps listed above.
5440 Annotate the assembler output with miscellaneous debugging information.
5444 Dump all macro definitions, at the end of preprocessing, in addition to
5449 Produce a core dump whenever an error occurs.
5453 Print statistics on memory usage, at the end of the run, to
5458 Annotate the assembler output with a comment indicating which
5459 pattern and alternative was used. The length of each instruction is
5464 Dump the RTL in the assembler output as a comment before each instruction.
5465 Also turns on @option{-dp} annotation.
5469 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
5470 dump a representation of the control flow graph suitable for viewing with VCG
5471 to @file{@var{file}.@var{pass}.vcg}.
5475 Just generate RTL for a function instead of compiling it. Usually used
5476 with @option{-fdump-rtl-expand}.
5480 @opindex fdump-noaddr
5481 When doing debugging dumps, suppress address output. This makes it more
5482 feasible to use diff on debugging dumps for compiler invocations with
5483 different compiler binaries and/or different
5484 text / bss / data / heap / stack / dso start locations.
5486 @item -fdump-unnumbered
5487 @opindex fdump-unnumbered
5488 When doing debugging dumps, suppress instruction numbers and address output.
5489 This makes it more feasible to use diff on debugging dumps for compiler
5490 invocations with different options, in particular with and without
5493 @item -fdump-unnumbered-links
5494 @opindex fdump-unnumbered-links
5495 When doing debugging dumps (see @option{-d} option above), suppress
5496 instruction numbers for the links to the previous and next instructions
5499 @item -fdump-translation-unit @r{(C++ only)}
5500 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
5501 @opindex fdump-translation-unit
5502 Dump a representation of the tree structure for the entire translation
5503 unit to a file. The file name is made by appending @file{.tu} to the
5504 source file name, and the file is created in the same directory as the
5505 output file. If the @samp{-@var{options}} form is used, @var{options}
5506 controls the details of the dump as described for the
5507 @option{-fdump-tree} options.
5509 @item -fdump-class-hierarchy @r{(C++ only)}
5510 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
5511 @opindex fdump-class-hierarchy
5512 Dump a representation of each class's hierarchy and virtual function
5513 table layout to a file. The file name is made by appending
5514 @file{.class} to the source file name, and the file is created in the
5515 same directory as the output file. If the @samp{-@var{options}} form
5516 is used, @var{options} controls the details of the dump as described
5517 for the @option{-fdump-tree} options.
5519 @item -fdump-ipa-@var{switch}
5521 Control the dumping at various stages of inter-procedural analysis
5522 language tree to a file. The file name is generated by appending a
5523 switch specific suffix to the source file name, and the file is created
5524 in the same directory as the output file. The following dumps are
5529 Enables all inter-procedural analysis dumps.
5532 Dumps information about call-graph optimization, unused function removal,
5533 and inlining decisions.
5536 Dump after function inlining.
5541 @opindex fdump-passes
5542 Dump the list of optimization passes that are turned on and off by
5543 the current command line options.
5545 @item -fdump-statistics-@var{option}
5546 @opindex fdump-statistics
5547 Enable and control dumping of pass statistics in a separate file. The
5548 file name is generated by appending a suffix ending in
5549 @samp{.statistics} to the source file name, and the file is created in
5550 the same directory as the output file. If the @samp{-@var{option}}
5551 form is used, @samp{-stats} will cause counters to be summed over the
5552 whole compilation unit while @samp{-details} will dump every event as
5553 the passes generate them. The default with no option is to sum
5554 counters for each function compiled.
5556 @item -fdump-tree-@var{switch}
5557 @itemx -fdump-tree-@var{switch}-@var{options}
5559 Control the dumping at various stages of processing the intermediate
5560 language tree to a file. The file name is generated by appending a
5561 switch specific suffix to the source file name, and the file is
5562 created in the same directory as the output file. If the
5563 @samp{-@var{options}} form is used, @var{options} is a list of
5564 @samp{-} separated options that control the details of the dump. Not
5565 all options are applicable to all dumps, those which are not
5566 meaningful will be ignored. The following options are available
5570 Print the address of each node. Usually this is not meaningful as it
5571 changes according to the environment and source file. Its primary use
5572 is for tying up a dump file with a debug environment.
5574 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
5575 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
5576 use working backward from mangled names in the assembly file.
5578 Inhibit dumping of members of a scope or body of a function merely
5579 because that scope has been reached. Only dump such items when they
5580 are directly reachable by some other path. When dumping pretty-printed
5581 trees, this option inhibits dumping the bodies of control structures.
5583 Print a raw representation of the tree. By default, trees are
5584 pretty-printed into a C-like representation.
5586 Enable more detailed dumps (not honored by every dump option).
5588 Enable dumping various statistics about the pass (not honored by every dump
5591 Enable showing basic block boundaries (disabled in raw dumps).
5593 Enable showing virtual operands for every statement.
5595 Enable showing line numbers for statements.
5597 Enable showing the unique ID (@code{DECL_UID}) for each variable.
5599 Enable showing the tree dump for each statement.
5601 Enable showing the EH region number holding each statement.
5603 Enable showing scalar evolution analysis details.
5605 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
5606 and @option{lineno}.
5609 The following tree dumps are possible:
5613 @opindex fdump-tree-original
5614 Dump before any tree based optimization, to @file{@var{file}.original}.
5617 @opindex fdump-tree-optimized
5618 Dump after all tree based optimization, to @file{@var{file}.optimized}.
5621 @opindex fdump-tree-gimple
5622 Dump each function before and after the gimplification pass to a file. The
5623 file name is made by appending @file{.gimple} to the source file name.
5626 @opindex fdump-tree-cfg
5627 Dump the control flow graph of each function to a file. The file name is
5628 made by appending @file{.cfg} to the source file name.
5631 @opindex fdump-tree-vcg
5632 Dump the control flow graph of each function to a file in VCG format. The
5633 file name is made by appending @file{.vcg} to the source file name. Note
5634 that if the file contains more than one function, the generated file cannot
5635 be used directly by VCG@. You will need to cut and paste each function's
5636 graph into its own separate file first.
5639 @opindex fdump-tree-ch
5640 Dump each function after copying loop headers. The file name is made by
5641 appending @file{.ch} to the source file name.
5644 @opindex fdump-tree-ssa
5645 Dump SSA related information to a file. The file name is made by appending
5646 @file{.ssa} to the source file name.
5649 @opindex fdump-tree-alias
5650 Dump aliasing information for each function. The file name is made by
5651 appending @file{.alias} to the source file name.
5654 @opindex fdump-tree-ccp
5655 Dump each function after CCP@. The file name is made by appending
5656 @file{.ccp} to the source file name.
5659 @opindex fdump-tree-storeccp
5660 Dump each function after STORE-CCP@. The file name is made by appending
5661 @file{.storeccp} to the source file name.
5664 @opindex fdump-tree-pre
5665 Dump trees after partial redundancy elimination. The file name is made
5666 by appending @file{.pre} to the source file name.
5669 @opindex fdump-tree-fre
5670 Dump trees after full redundancy elimination. The file name is made
5671 by appending @file{.fre} to the source file name.
5674 @opindex fdump-tree-copyprop
5675 Dump trees after copy propagation. The file name is made
5676 by appending @file{.copyprop} to the source file name.
5678 @item store_copyprop
5679 @opindex fdump-tree-store_copyprop
5680 Dump trees after store copy-propagation. The file name is made
5681 by appending @file{.store_copyprop} to the source file name.
5684 @opindex fdump-tree-dce
5685 Dump each function after dead code elimination. The file name is made by
5686 appending @file{.dce} to the source file name.
5689 @opindex fdump-tree-mudflap
5690 Dump each function after adding mudflap instrumentation. The file name is
5691 made by appending @file{.mudflap} to the source file name.
5694 @opindex fdump-tree-sra
5695 Dump each function after performing scalar replacement of aggregates. The
5696 file name is made by appending @file{.sra} to the source file name.
5699 @opindex fdump-tree-sink
5700 Dump each function after performing code sinking. The file name is made
5701 by appending @file{.sink} to the source file name.
5704 @opindex fdump-tree-dom
5705 Dump each function after applying dominator tree optimizations. The file
5706 name is made by appending @file{.dom} to the source file name.
5709 @opindex fdump-tree-dse
5710 Dump each function after applying dead store elimination. The file
5711 name is made by appending @file{.dse} to the source file name.
5714 @opindex fdump-tree-phiopt
5715 Dump each function after optimizing PHI nodes into straightline code. The file
5716 name is made by appending @file{.phiopt} to the source file name.
5719 @opindex fdump-tree-forwprop
5720 Dump each function after forward propagating single use variables. The file
5721 name is made by appending @file{.forwprop} to the source file name.
5724 @opindex fdump-tree-copyrename
5725 Dump each function after applying the copy rename optimization. The file
5726 name is made by appending @file{.copyrename} to the source file name.
5729 @opindex fdump-tree-nrv
5730 Dump each function after applying the named return value optimization on
5731 generic trees. The file name is made by appending @file{.nrv} to the source
5735 @opindex fdump-tree-vect
5736 Dump each function after applying vectorization of loops. The file name is
5737 made by appending @file{.vect} to the source file name.
5740 @opindex fdump-tree-slp
5741 Dump each function after applying vectorization of basic blocks. The file name
5742 is made by appending @file{.slp} to the source file name.
5745 @opindex fdump-tree-vrp
5746 Dump each function after Value Range Propagation (VRP). The file name
5747 is made by appending @file{.vrp} to the source file name.
5750 @opindex fdump-tree-all
5751 Enable all the available tree dumps with the flags provided in this option.
5754 @item -ftree-vectorizer-verbose=@var{n}
5755 @opindex ftree-vectorizer-verbose
5756 This option controls the amount of debugging output the vectorizer prints.
5757 This information is written to standard error, unless
5758 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
5759 in which case it is output to the usual dump listing file, @file{.vect}.
5760 For @var{n}=0 no diagnostic information is reported.
5761 If @var{n}=1 the vectorizer reports each loop that got vectorized,
5762 and the total number of loops that got vectorized.
5763 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
5764 the first analysis phase (vect_analyze_loop_form) - i.e.@: countable,
5765 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
5766 level that @option{-fdump-tree-vect-stats} uses.
5767 Higher verbosity levels mean either more information dumped for each
5768 reported loop, or same amount of information reported for more loops:
5769 if @var{n}=3, vectorizer cost model information is reported.
5770 If @var{n}=4, alignment related information is added to the reports.
5771 If @var{n}=5, data-references related information (e.g.@: memory dependences,
5772 memory access-patterns) is added to the reports.
5773 If @var{n}=6, the vectorizer reports also non-vectorized inner-most loops
5774 that did not pass the first analysis phase (i.e., may not be countable, or
5775 may have complicated control-flow).
5776 If @var{n}=7, the vectorizer reports also non-vectorized nested loops.
5777 If @var{n}=8, SLP related information is added to the reports.
5778 For @var{n}=9, all the information the vectorizer generates during its
5779 analysis and transformation is reported. This is the same verbosity level
5780 that @option{-fdump-tree-vect-details} uses.
5782 @item -frandom-seed=@var{string}
5783 @opindex frandom-seed
5784 This option provides a seed that GCC uses when it would otherwise use
5785 random numbers. It is used to generate certain symbol names
5786 that have to be different in every compiled file. It is also used to
5787 place unique stamps in coverage data files and the object files that
5788 produce them. You can use the @option{-frandom-seed} option to produce
5789 reproducibly identical object files.
5791 The @var{string} should be different for every file you compile.
5793 @item -fsched-verbose=@var{n}
5794 @opindex fsched-verbose
5795 On targets that use instruction scheduling, this option controls the
5796 amount of debugging output the scheduler prints. This information is
5797 written to standard error, unless @option{-fdump-rtl-sched1} or
5798 @option{-fdump-rtl-sched2} is specified, in which case it is output
5799 to the usual dump listing file, @file{.sched1} or @file{.sched2}
5800 respectively. However for @var{n} greater than nine, the output is
5801 always printed to standard error.
5803 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
5804 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
5805 For @var{n} greater than one, it also output basic block probabilities,
5806 detailed ready list information and unit/insn info. For @var{n} greater
5807 than two, it includes RTL at abort point, control-flow and regions info.
5808 And for @var{n} over four, @option{-fsched-verbose} also includes
5812 @itemx -save-temps=cwd
5814 Store the usual ``temporary'' intermediate files permanently; place them
5815 in the current directory and name them based on the source file. Thus,
5816 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
5817 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
5818 preprocessed @file{foo.i} output file even though the compiler now
5819 normally uses an integrated preprocessor.
5821 When used in combination with the @option{-x} command line option,
5822 @option{-save-temps} is sensible enough to avoid over writing an
5823 input source file with the same extension as an intermediate file.
5824 The corresponding intermediate file may be obtained by renaming the
5825 source file before using @option{-save-temps}.
5827 If you invoke GCC in parallel, compiling several different source
5828 files that share a common base name in different subdirectories or the
5829 same source file compiled for multiple output destinations, it is
5830 likely that the different parallel compilers will interfere with each
5831 other, and overwrite the temporary files. For instance:
5834 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
5835 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
5838 may result in @file{foo.i} and @file{foo.o} being written to
5839 simultaneously by both compilers.
5841 @item -save-temps=obj
5842 @opindex save-temps=obj
5843 Store the usual ``temporary'' intermediate files permanently. If the
5844 @option{-o} option is used, the temporary files are based on the
5845 object file. If the @option{-o} option is not used, the
5846 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
5851 gcc -save-temps=obj -c foo.c
5852 gcc -save-temps=obj -c bar.c -o dir/xbar.o
5853 gcc -save-temps=obj foobar.c -o dir2/yfoobar
5856 would create @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
5857 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
5858 @file{dir2/yfoobar.o}.
5860 @item -time@r{[}=@var{file}@r{]}
5862 Report the CPU time taken by each subprocess in the compilation
5863 sequence. For C source files, this is the compiler proper and assembler
5864 (plus the linker if linking is done).
5866 Without the specification of an output file, the output looks like this:
5873 The first number on each line is the ``user time'', that is time spent
5874 executing the program itself. The second number is ``system time'',
5875 time spent executing operating system routines on behalf of the program.
5876 Both numbers are in seconds.
5878 With the specification of an output file, the output is appended to the
5879 named file, and it looks like this:
5882 0.12 0.01 cc1 @var{options}
5883 0.00 0.01 as @var{options}
5886 The ``user time'' and the ``system time'' are moved before the program
5887 name, and the options passed to the program are displayed, so that one
5888 can later tell what file was being compiled, and with which options.
5890 @item -fvar-tracking
5891 @opindex fvar-tracking
5892 Run variable tracking pass. It computes where variables are stored at each
5893 position in code. Better debugging information is then generated
5894 (if the debugging information format supports this information).
5896 It is enabled by default when compiling with optimization (@option{-Os},
5897 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
5898 the debug info format supports it.
5900 @item -fvar-tracking-assignments
5901 @opindex fvar-tracking-assignments
5902 @opindex fno-var-tracking-assignments
5903 Annotate assignments to user variables early in the compilation and
5904 attempt to carry the annotations over throughout the compilation all the
5905 way to the end, in an attempt to improve debug information while
5906 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
5908 It can be enabled even if var-tracking is disabled, in which case
5909 annotations will be created and maintained, but discarded at the end.
5911 @item -fvar-tracking-assignments-toggle
5912 @opindex fvar-tracking-assignments-toggle
5913 @opindex fno-var-tracking-assignments-toggle
5914 Toggle @option{-fvar-tracking-assignments}, in the same way that
5915 @option{-gtoggle} toggles @option{-g}.
5917 @item -print-file-name=@var{library}
5918 @opindex print-file-name
5919 Print the full absolute name of the library file @var{library} that
5920 would be used when linking---and don't do anything else. With this
5921 option, GCC does not compile or link anything; it just prints the
5924 @item -print-multi-directory
5925 @opindex print-multi-directory
5926 Print the directory name corresponding to the multilib selected by any
5927 other switches present in the command line. This directory is supposed
5928 to exist in @env{GCC_EXEC_PREFIX}.
5930 @item -print-multi-lib
5931 @opindex print-multi-lib
5932 Print the mapping from multilib directory names to compiler switches
5933 that enable them. The directory name is separated from the switches by
5934 @samp{;}, and each switch starts with an @samp{@@} instead of the
5935 @samp{-}, without spaces between multiple switches. This is supposed to
5936 ease shell-processing.
5938 @item -print-multi-os-directory
5939 @opindex print-multi-os-directory
5940 Print the path to OS libraries for the selected
5941 multilib, relative to some @file{lib} subdirectory. If OS libraries are
5942 present in the @file{lib} subdirectory and no multilibs are used, this is
5943 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
5944 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
5945 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
5946 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
5948 @item -print-prog-name=@var{program}
5949 @opindex print-prog-name
5950 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
5952 @item -print-libgcc-file-name
5953 @opindex print-libgcc-file-name
5954 Same as @option{-print-file-name=libgcc.a}.
5956 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
5957 but you do want to link with @file{libgcc.a}. You can do
5960 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
5963 @item -print-search-dirs
5964 @opindex print-search-dirs
5965 Print the name of the configured installation directory and a list of
5966 program and library directories @command{gcc} will search---and don't do anything else.
5968 This is useful when @command{gcc} prints the error message
5969 @samp{installation problem, cannot exec cpp0: No such file or directory}.
5970 To resolve this you either need to put @file{cpp0} and the other compiler
5971 components where @command{gcc} expects to find them, or you can set the environment
5972 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
5973 Don't forget the trailing @samp{/}.
5974 @xref{Environment Variables}.
5976 @item -print-sysroot
5977 @opindex print-sysroot
5978 Print the target sysroot directory that will be used during
5979 compilation. This is the target sysroot specified either at configure
5980 time or using the @option{--sysroot} option, possibly with an extra
5981 suffix that depends on compilation options. If no target sysroot is
5982 specified, the option prints nothing.
5984 @item -print-sysroot-headers-suffix
5985 @opindex print-sysroot-headers-suffix
5986 Print the suffix added to the target sysroot when searching for
5987 headers, or give an error if the compiler is not configured with such
5988 a suffix---and don't do anything else.
5991 @opindex dumpmachine
5992 Print the compiler's target machine (for example,
5993 @samp{i686-pc-linux-gnu})---and don't do anything else.
5996 @opindex dumpversion
5997 Print the compiler version (for example, @samp{3.0})---and don't do
6002 Print the compiler's built-in specs---and don't do anything else. (This
6003 is used when GCC itself is being built.) @xref{Spec Files}.
6005 @item -feliminate-unused-debug-types
6006 @opindex feliminate-unused-debug-types
6007 Normally, when producing DWARF2 output, GCC will emit debugging
6008 information for all types declared in a compilation
6009 unit, regardless of whether or not they are actually used
6010 in that compilation unit. Sometimes this is useful, such as
6011 if, in the debugger, you want to cast a value to a type that is
6012 not actually used in your program (but is declared). More often,
6013 however, this results in a significant amount of wasted space.
6014 With this option, GCC will avoid producing debug symbol output
6015 for types that are nowhere used in the source file being compiled.
6018 @node Optimize Options
6019 @section Options That Control Optimization
6020 @cindex optimize options
6021 @cindex options, optimization
6023 These options control various sorts of optimizations.
6025 Without any optimization option, the compiler's goal is to reduce the
6026 cost of compilation and to make debugging produce the expected
6027 results. Statements are independent: if you stop the program with a
6028 breakpoint between statements, you can then assign a new value to any
6029 variable or change the program counter to any other statement in the
6030 function and get exactly the results you would expect from the source
6033 Turning on optimization flags makes the compiler attempt to improve
6034 the performance and/or code size at the expense of compilation time
6035 and possibly the ability to debug the program.
6037 The compiler performs optimization based on the knowledge it has of the
6038 program. Compiling multiple files at once to a single output file mode allows
6039 the compiler to use information gained from all of the files when compiling
6042 Not all optimizations are controlled directly by a flag. Only
6043 optimizations that have a flag are listed in this section.
6045 Most optimizations are only enabled if an @option{-O} level is set on
6046 the command line. Otherwise they are disabled, even if individual
6047 optimization flags are specified.
6049 Depending on the target and how GCC was configured, a slightly different
6050 set of optimizations may be enabled at each @option{-O} level than
6051 those listed here. You can invoke GCC with @samp{-Q --help=optimizers}
6052 to find out the exact set of optimizations that are enabled at each level.
6053 @xref{Overall Options}, for examples.
6060 Optimize. Optimizing compilation takes somewhat more time, and a lot
6061 more memory for a large function.
6063 With @option{-O}, the compiler tries to reduce code size and execution
6064 time, without performing any optimizations that take a great deal of
6067 @option{-O} turns on the following optimization flags:
6071 -fcprop-registers @gol
6074 -fdelayed-branch @gol
6076 -fguess-branch-probability @gol
6077 -fif-conversion2 @gol
6078 -fif-conversion @gol
6079 -fipa-pure-const @gol
6081 -fipa-reference @gol
6083 -fsplit-wide-types @gol
6085 -ftree-builtin-call-dce @gol
6088 -ftree-copyrename @gol
6090 -ftree-dominator-opts @gol
6092 -ftree-forwprop @gol
6100 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6101 where doing so does not interfere with debugging.
6105 Optimize even more. GCC performs nearly all supported optimizations
6106 that do not involve a space-speed tradeoff.
6107 As compared to @option{-O}, this option increases both compilation time
6108 and the performance of the generated code.
6110 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6111 also turns on the following optimization flags:
6112 @gccoptlist{-fthread-jumps @gol
6113 -falign-functions -falign-jumps @gol
6114 -falign-loops -falign-labels @gol
6117 -fcse-follow-jumps -fcse-skip-blocks @gol
6118 -fdelete-null-pointer-checks @gol
6120 -fexpensive-optimizations @gol
6121 -fgcse -fgcse-lm @gol
6122 -finline-small-functions @gol
6123 -findirect-inlining @gol
6125 -foptimize-sibling-calls @gol
6126 -fpartial-inlining @gol
6129 -freorder-blocks -freorder-functions @gol
6130 -frerun-cse-after-loop @gol
6131 -fsched-interblock -fsched-spec @gol
6132 -fschedule-insns -fschedule-insns2 @gol
6133 -fstrict-aliasing -fstrict-overflow @gol
6134 -ftree-switch-conversion @gol
6138 Please note the warning under @option{-fgcse} about
6139 invoking @option{-O2} on programs that use computed gotos.
6143 Optimize yet more. @option{-O3} turns on all optimizations specified
6144 by @option{-O2} and also turns on the @option{-finline-functions},
6145 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6146 @option{-fgcse-after-reload}, @option{-ftree-vectorize} and
6147 @option{-fipa-cp-clone} options.
6151 Reduce compilation time and make debugging produce the expected
6152 results. This is the default.
6156 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6157 do not typically increase code size. It also performs further
6158 optimizations designed to reduce code size.
6160 @option{-Os} disables the following optimization flags:
6161 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6162 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
6163 -fprefetch-loop-arrays -ftree-vect-loop-version}
6167 Disregard strict standards compliance. @option{-Ofast} enables all
6168 @option{-O3} optimizations. It also enables optimizations that are not
6169 valid for all standard compliant programs.
6170 It turns on @option{-ffast-math} and the Fortran-specific
6171 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6173 If you use multiple @option{-O} options, with or without level numbers,
6174 the last such option is the one that is effective.
6177 Options of the form @option{-f@var{flag}} specify machine-independent
6178 flags. Most flags have both positive and negative forms; the negative
6179 form of @option{-ffoo} would be @option{-fno-foo}. In the table
6180 below, only one of the forms is listed---the one you typically will
6181 use. You can figure out the other form by either removing @samp{no-}
6184 The following options control specific optimizations. They are either
6185 activated by @option{-O} options or are related to ones that are. You
6186 can use the following flags in the rare cases when ``fine-tuning'' of
6187 optimizations to be performed is desired.
6190 @item -fno-default-inline
6191 @opindex fno-default-inline
6192 Do not make member functions inline by default merely because they are
6193 defined inside the class scope (C++ only). Otherwise, when you specify
6194 @w{@option{-O}}, member functions defined inside class scope are compiled
6195 inline by default; i.e., you don't need to add @samp{inline} in front of
6196 the member function name.
6198 @item -fno-defer-pop
6199 @opindex fno-defer-pop
6200 Always pop the arguments to each function call as soon as that function
6201 returns. For machines which must pop arguments after a function call,
6202 the compiler normally lets arguments accumulate on the stack for several
6203 function calls and pops them all at once.
6205 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6207 @item -fforward-propagate
6208 @opindex fforward-propagate
6209 Perform a forward propagation pass on RTL@. The pass tries to combine two
6210 instructions and checks if the result can be simplified. If loop unrolling
6211 is active, two passes are performed and the second is scheduled after
6214 This option is enabled by default at optimization levels @option{-O},
6215 @option{-O2}, @option{-O3}, @option{-Os}.
6217 @item -ffp-contract=@var{style}
6218 @opindex ffp-contract
6219 @option{-ffp-contract=off} disables floating-point expression contraction.
6220 @option{-ffp-contract=fast} enables floating-point expression contraction
6221 such as forming of fused multiply-add operations if the target has
6222 native support for them.
6223 @option{-ffp-contract=on} enables floating-point expression contraction
6224 if allowed by the language standard. This is currently not implemented
6225 and treated equal to @option{-ffp-contract=off}.
6227 The default is @option{-ffp-contract=fast}.
6229 @item -fomit-frame-pointer
6230 @opindex fomit-frame-pointer
6231 Don't keep the frame pointer in a register for functions that
6232 don't need one. This avoids the instructions to save, set up and
6233 restore frame pointers; it also makes an extra register available
6234 in many functions. @strong{It also makes debugging impossible on
6237 On some machines, such as the VAX, this flag has no effect, because
6238 the standard calling sequence automatically handles the frame pointer
6239 and nothing is saved by pretending it doesn't exist. The
6240 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6241 whether a target machine supports this flag. @xref{Registers,,Register
6242 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6244 Starting with GCC version 4.6, the default setting (when not optimizing for
6245 size) for 32-bit Linux x86 and 32-bit Darwin x86 targets has been changed to
6246 @option{-fomit-frame-pointer}. The default can be reverted to
6247 @option{-fno-omit-frame-pointer} by configuring GCC with the
6248 @option{--enable-frame-pointer} configure option.
6250 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6252 @item -foptimize-sibling-calls
6253 @opindex foptimize-sibling-calls
6254 Optimize sibling and tail recursive calls.
6256 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6260 Don't pay attention to the @code{inline} keyword. Normally this option
6261 is used to keep the compiler from expanding any functions inline.
6262 Note that if you are not optimizing, no functions can be expanded inline.
6264 @item -finline-small-functions
6265 @opindex finline-small-functions
6266 Integrate functions into their callers when their body is smaller than expected
6267 function call code (so overall size of program gets smaller). The compiler
6268 heuristically decides which functions are simple enough to be worth integrating
6271 Enabled at level @option{-O2}.
6273 @item -findirect-inlining
6274 @opindex findirect-inlining
6275 Inline also indirect calls that are discovered to be known at compile
6276 time thanks to previous inlining. This option has any effect only
6277 when inlining itself is turned on by the @option{-finline-functions}
6278 or @option{-finline-small-functions} options.
6280 Enabled at level @option{-O2}.
6282 @item -finline-functions
6283 @opindex finline-functions
6284 Integrate all simple functions into their callers. The compiler
6285 heuristically decides which functions are simple enough to be worth
6286 integrating in this way.
6288 If all calls to a given function are integrated, and the function is
6289 declared @code{static}, then the function is normally not output as
6290 assembler code in its own right.
6292 Enabled at level @option{-O3}.
6294 @item -finline-functions-called-once
6295 @opindex finline-functions-called-once
6296 Consider all @code{static} functions called once for inlining into their
6297 caller even if they are not marked @code{inline}. If a call to a given
6298 function is integrated, then the function is not output as assembler code
6301 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6303 @item -fearly-inlining
6304 @opindex fearly-inlining
6305 Inline functions marked by @code{always_inline} and functions whose body seems
6306 smaller than the function call overhead early before doing
6307 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6308 makes profiling significantly cheaper and usually inlining faster on programs
6309 having large chains of nested wrapper functions.
6315 Perform interprocedural scalar replacement of aggregates, removal of
6316 unused parameters and replacement of parameters passed by reference
6317 by parameters passed by value.
6319 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6321 @item -finline-limit=@var{n}
6322 @opindex finline-limit
6323 By default, GCC limits the size of functions that can be inlined. This flag
6324 allows coarse control of this limit. @var{n} is the size of functions that
6325 can be inlined in number of pseudo instructions.
6327 Inlining is actually controlled by a number of parameters, which may be
6328 specified individually by using @option{--param @var{name}=@var{value}}.
6329 The @option{-finline-limit=@var{n}} option sets some of these parameters
6333 @item max-inline-insns-single
6334 is set to @var{n}/2.
6335 @item max-inline-insns-auto
6336 is set to @var{n}/2.
6339 See below for a documentation of the individual
6340 parameters controlling inlining and for the defaults of these parameters.
6342 @emph{Note:} there may be no value to @option{-finline-limit} that results
6343 in default behavior.
6345 @emph{Note:} pseudo instruction represents, in this particular context, an
6346 abstract measurement of function's size. In no way does it represent a count
6347 of assembly instructions and as such its exact meaning might change from one
6348 release to an another.
6350 @item -fno-keep-inline-dllexport
6351 @opindex -fno-keep-inline-dllexport
6352 This is a more fine-grained version of @option{-fkeep-inline-functions},
6353 which applies only to functions that are declared using the @code{dllexport}
6354 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6357 @item -fkeep-inline-functions
6358 @opindex fkeep-inline-functions
6359 In C, emit @code{static} functions that are declared @code{inline}
6360 into the object file, even if the function has been inlined into all
6361 of its callers. This switch does not affect functions using the
6362 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6363 inline functions into the object file.
6365 @item -fkeep-static-consts
6366 @opindex fkeep-static-consts
6367 Emit variables declared @code{static const} when optimization isn't turned
6368 on, even if the variables aren't referenced.
6370 GCC enables this option by default. If you want to force the compiler to
6371 check if the variable was referenced, regardless of whether or not
6372 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6374 @item -fmerge-constants
6375 @opindex fmerge-constants
6376 Attempt to merge identical constants (string constants and floating point
6377 constants) across compilation units.
6379 This option is the default for optimized compilation if the assembler and
6380 linker support it. Use @option{-fno-merge-constants} to inhibit this
6383 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6385 @item -fmerge-all-constants
6386 @opindex fmerge-all-constants
6387 Attempt to merge identical constants and identical variables.
6389 This option implies @option{-fmerge-constants}. In addition to
6390 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6391 arrays or initialized constant variables with integral or floating point
6392 types. Languages like C or C++ require each variable, including multiple
6393 instances of the same variable in recursive calls, to have distinct locations,
6394 so using this option will result in non-conforming
6397 @item -fmodulo-sched
6398 @opindex fmodulo-sched
6399 Perform swing modulo scheduling immediately before the first scheduling
6400 pass. This pass looks at innermost loops and reorders their
6401 instructions by overlapping different iterations.
6403 @item -fmodulo-sched-allow-regmoves
6404 @opindex fmodulo-sched-allow-regmoves
6405 Perform more aggressive SMS based modulo scheduling with register moves
6406 allowed. By setting this flag certain anti-dependences edges will be
6407 deleted which will trigger the generation of reg-moves based on the
6408 life-range analysis. This option is effective only with
6409 @option{-fmodulo-sched} enabled.
6411 @item -fno-branch-count-reg
6412 @opindex fno-branch-count-reg
6413 Do not use ``decrement and branch'' instructions on a count register,
6414 but instead generate a sequence of instructions that decrement a
6415 register, compare it against zero, then branch based upon the result.
6416 This option is only meaningful on architectures that support such
6417 instructions, which include x86, PowerPC, IA-64 and S/390.
6419 The default is @option{-fbranch-count-reg}.
6421 @item -fno-function-cse
6422 @opindex fno-function-cse
6423 Do not put function addresses in registers; make each instruction that
6424 calls a constant function contain the function's address explicitly.
6426 This option results in less efficient code, but some strange hacks
6427 that alter the assembler output may be confused by the optimizations
6428 performed when this option is not used.
6430 The default is @option{-ffunction-cse}
6432 @item -fno-zero-initialized-in-bss
6433 @opindex fno-zero-initialized-in-bss
6434 If the target supports a BSS section, GCC by default puts variables that
6435 are initialized to zero into BSS@. This can save space in the resulting
6438 This option turns off this behavior because some programs explicitly
6439 rely on variables going to the data section. E.g., so that the
6440 resulting executable can find the beginning of that section and/or make
6441 assumptions based on that.
6443 The default is @option{-fzero-initialized-in-bss}.
6445 @item -fmudflap -fmudflapth -fmudflapir
6449 @cindex bounds checking
6451 For front-ends that support it (C and C++), instrument all risky
6452 pointer/array dereferencing operations, some standard library
6453 string/heap functions, and some other associated constructs with
6454 range/validity tests. Modules so instrumented should be immune to
6455 buffer overflows, invalid heap use, and some other classes of C/C++
6456 programming errors. The instrumentation relies on a separate runtime
6457 library (@file{libmudflap}), which will be linked into a program if
6458 @option{-fmudflap} is given at link time. Run-time behavior of the
6459 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
6460 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
6463 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
6464 link if your program is multi-threaded. Use @option{-fmudflapir}, in
6465 addition to @option{-fmudflap} or @option{-fmudflapth}, if
6466 instrumentation should ignore pointer reads. This produces less
6467 instrumentation (and therefore faster execution) and still provides
6468 some protection against outright memory corrupting writes, but allows
6469 erroneously read data to propagate within a program.
6471 @item -fthread-jumps
6472 @opindex fthread-jumps
6473 Perform optimizations where we check to see if a jump branches to a
6474 location where another comparison subsumed by the first is found. If
6475 so, the first branch is redirected to either the destination of the
6476 second branch or a point immediately following it, depending on whether
6477 the condition is known to be true or false.
6479 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6481 @item -fsplit-wide-types
6482 @opindex fsplit-wide-types
6483 When using a type that occupies multiple registers, such as @code{long
6484 long} on a 32-bit system, split the registers apart and allocate them
6485 independently. This normally generates better code for those types,
6486 but may make debugging more difficult.
6488 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6491 @item -fcse-follow-jumps
6492 @opindex fcse-follow-jumps
6493 In common subexpression elimination (CSE), scan through jump instructions
6494 when the target of the jump is not reached by any other path. For
6495 example, when CSE encounters an @code{if} statement with an
6496 @code{else} clause, CSE will follow the jump when the condition
6499 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6501 @item -fcse-skip-blocks
6502 @opindex fcse-skip-blocks
6503 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6504 follow jumps which conditionally skip over blocks. When CSE
6505 encounters a simple @code{if} statement with no else clause,
6506 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6507 body of the @code{if}.
6509 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6511 @item -frerun-cse-after-loop
6512 @opindex frerun-cse-after-loop
6513 Re-run common subexpression elimination after loop optimizations has been
6516 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6520 Perform a global common subexpression elimination pass.
6521 This pass also performs global constant and copy propagation.
6523 @emph{Note:} When compiling a program using computed gotos, a GCC
6524 extension, you may get better runtime performance if you disable
6525 the global common subexpression elimination pass by adding
6526 @option{-fno-gcse} to the command line.
6528 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6532 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
6533 attempt to move loads which are only killed by stores into themselves. This
6534 allows a loop containing a load/store sequence to be changed to a load outside
6535 the loop, and a copy/store within the loop.
6537 Enabled by default when gcse is enabled.
6541 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6542 global common subexpression elimination. This pass will attempt to move
6543 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6544 loops containing a load/store sequence can be changed to a load before
6545 the loop and a store after the loop.
6547 Not enabled at any optimization level.
6551 When @option{-fgcse-las} is enabled, the global common subexpression
6552 elimination pass eliminates redundant loads that come after stores to the
6553 same memory location (both partial and full redundancies).
6555 Not enabled at any optimization level.
6557 @item -fgcse-after-reload
6558 @opindex fgcse-after-reload
6559 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6560 pass is performed after reload. The purpose of this pass is to cleanup
6563 @item -funsafe-loop-optimizations
6564 @opindex funsafe-loop-optimizations
6565 If given, the loop optimizer will assume that loop indices do not
6566 overflow, and that the loops with nontrivial exit condition are not
6567 infinite. This enables a wider range of loop optimizations even if
6568 the loop optimizer itself cannot prove that these assumptions are valid.
6569 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
6570 if it finds this kind of loop.
6572 @item -fcrossjumping
6573 @opindex fcrossjumping
6574 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
6575 resulting code may or may not perform better than without cross-jumping.
6577 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6579 @item -fauto-inc-dec
6580 @opindex fauto-inc-dec
6581 Combine increments or decrements of addresses with memory accesses.
6582 This pass is always skipped on architectures that do not have
6583 instructions to support this. Enabled by default at @option{-O} and
6584 higher on architectures that support this.
6588 Perform dead code elimination (DCE) on RTL@.
6589 Enabled by default at @option{-O} and higher.
6593 Perform dead store elimination (DSE) on RTL@.
6594 Enabled by default at @option{-O} and higher.
6596 @item -fif-conversion
6597 @opindex fif-conversion
6598 Attempt to transform conditional jumps into branch-less equivalents. This
6599 include use of conditional moves, min, max, set flags and abs instructions, and
6600 some tricks doable by standard arithmetics. The use of conditional execution
6601 on chips where it is available is controlled by @code{if-conversion2}.
6603 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6605 @item -fif-conversion2
6606 @opindex fif-conversion2
6607 Use conditional execution (where available) to transform conditional jumps into
6608 branch-less equivalents.
6610 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6612 @item -fdelete-null-pointer-checks
6613 @opindex fdelete-null-pointer-checks
6614 Assume that programs cannot safely dereference null pointers, and that
6615 no code or data element resides there. This enables simple constant
6616 folding optimizations at all optimization levels. In addition, other
6617 optimization passes in GCC use this flag to control global dataflow
6618 analyses that eliminate useless checks for null pointers; these assume
6619 that if a pointer is checked after it has already been dereferenced,
6622 Note however that in some environments this assumption is not true.
6623 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6624 for programs which depend on that behavior.
6626 Some targets, especially embedded ones, disable this option at all levels.
6627 Otherwise it is enabled at all levels: @option{-O0}, @option{-O1},
6628 @option{-O2}, @option{-O3}, @option{-Os}. Passes that use the information
6629 are enabled independently at different optimization levels.
6631 @item -fdevirtualize
6632 @opindex fdevirtualize
6633 Attempt to convert calls to virtual functions to direct calls. This
6634 is done both within a procedure and interprocedurally as part of
6635 indirect inlining (@code{-findirect-inlining}) and interprocedural constant
6636 propagation (@option{-fipa-cp}).
6637 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6639 @item -fexpensive-optimizations
6640 @opindex fexpensive-optimizations
6641 Perform a number of minor optimizations that are relatively expensive.
6643 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6645 @item -foptimize-register-move
6647 @opindex foptimize-register-move
6649 Attempt to reassign register numbers in move instructions and as
6650 operands of other simple instructions in order to maximize the amount of
6651 register tying. This is especially helpful on machines with two-operand
6654 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
6657 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6659 @item -fira-algorithm=@var{algorithm}
6660 Use specified coloring algorithm for the integrated register
6661 allocator. The @var{algorithm} argument should be @code{priority} or
6662 @code{CB}. The first algorithm specifies Chow's priority coloring,
6663 the second one specifies Chaitin-Briggs coloring. The second
6664 algorithm can be unimplemented for some architectures. If it is
6665 implemented, it is the default because Chaitin-Briggs coloring as a
6666 rule generates a better code.
6668 @item -fira-region=@var{region}
6669 Use specified regions for the integrated register allocator. The
6670 @var{region} argument should be one of @code{all}, @code{mixed}, or
6671 @code{one}. The first value means using all loops as register
6672 allocation regions, the second value which is the default means using
6673 all loops except for loops with small register pressure as the
6674 regions, and third one means using all function as a single region.
6675 The first value can give best result for machines with small size and
6676 irregular register set, the third one results in faster and generates
6677 decent code and the smallest size code, and the default value usually
6678 give the best results in most cases and for most architectures.
6680 @item -fira-loop-pressure
6681 @opindex fira-loop-pressure
6682 Use IRA to evaluate register pressure in loops for decision to move
6683 loop invariants. Usage of this option usually results in generation
6684 of faster and smaller code on machines with big register files (>= 32
6685 registers) but it can slow compiler down.
6687 This option is enabled at level @option{-O3} for some targets.
6689 @item -fno-ira-share-save-slots
6690 @opindex fno-ira-share-save-slots
6691 Switch off sharing stack slots used for saving call used hard
6692 registers living through a call. Each hard register will get a
6693 separate stack slot and as a result function stack frame will be
6696 @item -fno-ira-share-spill-slots
6697 @opindex fno-ira-share-spill-slots
6698 Switch off sharing stack slots allocated for pseudo-registers. Each
6699 pseudo-register which did not get a hard register will get a separate
6700 stack slot and as a result function stack frame will be bigger.
6702 @item -fira-verbose=@var{n}
6703 @opindex fira-verbose
6704 Set up how verbose dump file for the integrated register allocator
6705 will be. Default value is 5. If the value is greater or equal to 10,
6706 the dump file will be stderr as if the value were @var{n} minus 10.
6708 @item -fdelayed-branch
6709 @opindex fdelayed-branch
6710 If supported for the target machine, attempt to reorder instructions
6711 to exploit instruction slots available after delayed branch
6714 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6716 @item -fschedule-insns
6717 @opindex fschedule-insns
6718 If supported for the target machine, attempt to reorder instructions to
6719 eliminate execution stalls due to required data being unavailable. This
6720 helps machines that have slow floating point or memory load instructions
6721 by allowing other instructions to be issued until the result of the load
6722 or floating point instruction is required.
6724 Enabled at levels @option{-O2}, @option{-O3}.
6726 @item -fschedule-insns2
6727 @opindex fschedule-insns2
6728 Similar to @option{-fschedule-insns}, but requests an additional pass of
6729 instruction scheduling after register allocation has been done. This is
6730 especially useful on machines with a relatively small number of
6731 registers and where memory load instructions take more than one cycle.
6733 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6735 @item -fno-sched-interblock
6736 @opindex fno-sched-interblock
6737 Don't schedule instructions across basic blocks. This is normally
6738 enabled by default when scheduling before register allocation, i.e.@:
6739 with @option{-fschedule-insns} or at @option{-O2} or higher.
6741 @item -fno-sched-spec
6742 @opindex fno-sched-spec
6743 Don't allow speculative motion of non-load instructions. This is normally
6744 enabled by default when scheduling before register allocation, i.e.@:
6745 with @option{-fschedule-insns} or at @option{-O2} or higher.
6747 @item -fsched-pressure
6748 @opindex fsched-pressure
6749 Enable register pressure sensitive insn scheduling before the register
6750 allocation. This only makes sense when scheduling before register
6751 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
6752 @option{-O2} or higher. Usage of this option can improve the
6753 generated code and decrease its size by preventing register pressure
6754 increase above the number of available hard registers and as a
6755 consequence register spills in the register allocation.
6757 @item -fsched-spec-load
6758 @opindex fsched-spec-load
6759 Allow speculative motion of some load instructions. This only makes
6760 sense when scheduling before register allocation, i.e.@: with
6761 @option{-fschedule-insns} or at @option{-O2} or higher.
6763 @item -fsched-spec-load-dangerous
6764 @opindex fsched-spec-load-dangerous
6765 Allow speculative motion of more load instructions. This only makes
6766 sense when scheduling before register allocation, i.e.@: with
6767 @option{-fschedule-insns} or at @option{-O2} or higher.
6769 @item -fsched-stalled-insns
6770 @itemx -fsched-stalled-insns=@var{n}
6771 @opindex fsched-stalled-insns
6772 Define how many insns (if any) can be moved prematurely from the queue
6773 of stalled insns into the ready list, during the second scheduling pass.
6774 @option{-fno-sched-stalled-insns} means that no insns will be moved
6775 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
6776 on how many queued insns can be moved prematurely.
6777 @option{-fsched-stalled-insns} without a value is equivalent to
6778 @option{-fsched-stalled-insns=1}.
6780 @item -fsched-stalled-insns-dep
6781 @itemx -fsched-stalled-insns-dep=@var{n}
6782 @opindex fsched-stalled-insns-dep
6783 Define how many insn groups (cycles) will be examined for a dependency
6784 on a stalled insn that is candidate for premature removal from the queue
6785 of stalled insns. This has an effect only during the second scheduling pass,
6786 and only if @option{-fsched-stalled-insns} is used.
6787 @option{-fno-sched-stalled-insns-dep} is equivalent to
6788 @option{-fsched-stalled-insns-dep=0}.
6789 @option{-fsched-stalled-insns-dep} without a value is equivalent to
6790 @option{-fsched-stalled-insns-dep=1}.
6792 @item -fsched2-use-superblocks
6793 @opindex fsched2-use-superblocks
6794 When scheduling after register allocation, do use superblock scheduling
6795 algorithm. Superblock scheduling allows motion across basic block boundaries
6796 resulting on faster schedules. This option is experimental, as not all machine
6797 descriptions used by GCC model the CPU closely enough to avoid unreliable
6798 results from the algorithm.
6800 This only makes sense when scheduling after register allocation, i.e.@: with
6801 @option{-fschedule-insns2} or at @option{-O2} or higher.
6803 @item -fsched-group-heuristic
6804 @opindex fsched-group-heuristic
6805 Enable the group heuristic in the scheduler. This heuristic favors
6806 the instruction that belongs to a schedule group. This is enabled
6807 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6808 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6810 @item -fsched-critical-path-heuristic
6811 @opindex fsched-critical-path-heuristic
6812 Enable the critical-path heuristic in the scheduler. This heuristic favors
6813 instructions on the critical path. This is enabled by default when
6814 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
6815 or @option{-fschedule-insns2} or at @option{-O2} or higher.
6817 @item -fsched-spec-insn-heuristic
6818 @opindex fsched-spec-insn-heuristic
6819 Enable the speculative instruction heuristic in the scheduler. This
6820 heuristic favors speculative instructions with greater dependency weakness.
6821 This is enabled by default when scheduling is enabled, i.e.@:
6822 with @option{-fschedule-insns} or @option{-fschedule-insns2}
6823 or at @option{-O2} or higher.
6825 @item -fsched-rank-heuristic
6826 @opindex fsched-rank-heuristic
6827 Enable the rank heuristic in the scheduler. This heuristic favors
6828 the instruction belonging to a basic block with greater size or frequency.
6829 This is enabled by default when scheduling is enabled, i.e.@:
6830 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6831 at @option{-O2} or higher.
6833 @item -fsched-last-insn-heuristic
6834 @opindex fsched-last-insn-heuristic
6835 Enable the last-instruction heuristic in the scheduler. This heuristic
6836 favors the instruction that is less dependent on the last instruction
6837 scheduled. This is enabled by default when scheduling is enabled,
6838 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6839 at @option{-O2} or higher.
6841 @item -fsched-dep-count-heuristic
6842 @opindex fsched-dep-count-heuristic
6843 Enable the dependent-count heuristic in the scheduler. This heuristic
6844 favors the instruction that has more instructions depending on it.
6845 This is enabled by default when scheduling is enabled, i.e.@:
6846 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
6847 at @option{-O2} or higher.
6849 @item -freschedule-modulo-scheduled-loops
6850 @opindex freschedule-modulo-scheduled-loops
6851 The modulo scheduling comes before the traditional scheduling, if a loop
6852 was modulo scheduled we may want to prevent the later scheduling passes
6853 from changing its schedule, we use this option to control that.
6855 @item -fselective-scheduling
6856 @opindex fselective-scheduling
6857 Schedule instructions using selective scheduling algorithm. Selective
6858 scheduling runs instead of the first scheduler pass.
6860 @item -fselective-scheduling2
6861 @opindex fselective-scheduling2
6862 Schedule instructions using selective scheduling algorithm. Selective
6863 scheduling runs instead of the second scheduler pass.
6865 @item -fsel-sched-pipelining
6866 @opindex fsel-sched-pipelining
6867 Enable software pipelining of innermost loops during selective scheduling.
6868 This option has no effect until one of @option{-fselective-scheduling} or
6869 @option{-fselective-scheduling2} is turned on.
6871 @item -fsel-sched-pipelining-outer-loops
6872 @opindex fsel-sched-pipelining-outer-loops
6873 When pipelining loops during selective scheduling, also pipeline outer loops.
6874 This option has no effect until @option{-fsel-sched-pipelining} is turned on.
6876 @item -fcaller-saves
6877 @opindex fcaller-saves
6878 Enable values to be allocated in registers that will be clobbered by
6879 function calls, by emitting extra instructions to save and restore the
6880 registers around such calls. Such allocation is done only when it
6881 seems to result in better code than would otherwise be produced.
6883 This option is always enabled by default on certain machines, usually
6884 those which have no call-preserved registers to use instead.
6886 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6888 @item -fcombine-stack-adjustments
6889 @opindex fcombine-stack-adjustments
6890 Tracks stack adjustments (pushes and pops) and stack memory references
6891 and then tries to find ways to combine them.
6893 Enabled by default at @option{-O1} and higher.
6895 @item -fconserve-stack
6896 @opindex fconserve-stack
6897 Attempt to minimize stack usage. The compiler will attempt to use less
6898 stack space, even if that makes the program slower. This option
6899 implies setting the @option{large-stack-frame} parameter to 100
6900 and the @option{large-stack-frame-growth} parameter to 400.
6902 @item -ftree-reassoc
6903 @opindex ftree-reassoc
6904 Perform reassociation on trees. This flag is enabled by default
6905 at @option{-O} and higher.
6909 Perform partial redundancy elimination (PRE) on trees. This flag is
6910 enabled by default at @option{-O2} and @option{-O3}.
6912 @item -ftree-forwprop
6913 @opindex ftree-forwprop
6914 Perform forward propagation on trees. This flag is enabled by default
6915 at @option{-O} and higher.
6919 Perform full redundancy elimination (FRE) on trees. The difference
6920 between FRE and PRE is that FRE only considers expressions
6921 that are computed on all paths leading to the redundant computation.
6922 This analysis is faster than PRE, though it exposes fewer redundancies.
6923 This flag is enabled by default at @option{-O} and higher.
6925 @item -ftree-phiprop
6926 @opindex ftree-phiprop
6927 Perform hoisting of loads from conditional pointers on trees. This
6928 pass is enabled by default at @option{-O} and higher.
6930 @item -ftree-copy-prop
6931 @opindex ftree-copy-prop
6932 Perform copy propagation on trees. This pass eliminates unnecessary
6933 copy operations. This flag is enabled by default at @option{-O} and
6936 @item -fipa-pure-const
6937 @opindex fipa-pure-const
6938 Discover which functions are pure or constant.
6939 Enabled by default at @option{-O} and higher.
6941 @item -fipa-reference
6942 @opindex fipa-reference
6943 Discover which static variables do not escape cannot escape the
6945 Enabled by default at @option{-O} and higher.
6949 Perform interprocedural pointer analysis and interprocedural modification
6950 and reference analysis. This option can cause excessive memory and
6951 compile-time usage on large compilation units. It is not enabled by
6952 default at any optimization level.
6955 @opindex fipa-profile
6956 Perform interprocedural profile propagation. The functions called only from
6957 cold functions are marked as cold. Also functions executed once (such as
6958 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
6959 functions and loop less parts of functions executed once are then optimized for
6961 Enabled by default at @option{-O} and higher.
6965 Perform interprocedural constant propagation.
6966 This optimization analyzes the program to determine when values passed
6967 to functions are constants and then optimizes accordingly.
6968 This optimization can substantially increase performance
6969 if the application has constants passed to functions.
6970 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
6972 @item -fipa-cp-clone
6973 @opindex fipa-cp-clone
6974 Perform function cloning to make interprocedural constant propagation stronger.
6975 When enabled, interprocedural constant propagation will perform function cloning
6976 when externally visible function can be called with constant arguments.
6977 Because this optimization can create multiple copies of functions,
6978 it may significantly increase code size
6979 (see @option{--param ipcp-unit-growth=@var{value}}).
6980 This flag is enabled by default at @option{-O3}.
6982 @item -fipa-matrix-reorg
6983 @opindex fipa-matrix-reorg
6984 Perform matrix flattening and transposing.
6985 Matrix flattening tries to replace an @math{m}-dimensional matrix
6986 with its equivalent @math{n}-dimensional matrix, where @math{n < m}.
6987 This reduces the level of indirection needed for accessing the elements
6988 of the matrix. The second optimization is matrix transposing that
6989 attempts to change the order of the matrix's dimensions in order to
6990 improve cache locality.
6991 Both optimizations need the @option{-fwhole-program} flag.
6992 Transposing is enabled only if profiling information is available.
6996 Perform forward store motion on trees. This flag is
6997 enabled by default at @option{-O} and higher.
6999 @item -ftree-bit-ccp
7000 @opindex ftree-bit-ccp
7001 Perform sparse conditional bit constant propagation on trees and propagate
7002 pointer alignment information.
7003 This pass only operates on local scalar variables and is enabled by default
7004 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7008 Perform sparse conditional constant propagation (CCP) on trees. This
7009 pass only operates on local scalar variables and is enabled by default
7010 at @option{-O} and higher.
7012 @item -ftree-switch-conversion
7013 Perform conversion of simple initializations in a switch to
7014 initializations from a scalar array. This flag is enabled by default
7015 at @option{-O2} and higher.
7019 Perform dead code elimination (DCE) on trees. This flag is enabled by
7020 default at @option{-O} and higher.
7022 @item -ftree-builtin-call-dce
7023 @opindex ftree-builtin-call-dce
7024 Perform conditional dead code elimination (DCE) for calls to builtin functions
7025 that may set @code{errno} but are otherwise side-effect free. This flag is
7026 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7029 @item -ftree-dominator-opts
7030 @opindex ftree-dominator-opts
7031 Perform a variety of simple scalar cleanups (constant/copy
7032 propagation, redundancy elimination, range propagation and expression
7033 simplification) based on a dominator tree traversal. This also
7034 performs jump threading (to reduce jumps to jumps). This flag is
7035 enabled by default at @option{-O} and higher.
7039 Perform dead store elimination (DSE) on trees. A dead store is a store into
7040 a memory location which will later be overwritten by another store without
7041 any intervening loads. In this case the earlier store can be deleted. This
7042 flag is enabled by default at @option{-O} and higher.
7046 Perform loop header copying on trees. This is beneficial since it increases
7047 effectiveness of code motion optimizations. It also saves one jump. This flag
7048 is enabled by default at @option{-O} and higher. It is not enabled
7049 for @option{-Os}, since it usually increases code size.
7051 @item -ftree-loop-optimize
7052 @opindex ftree-loop-optimize
7053 Perform loop optimizations on trees. This flag is enabled by default
7054 at @option{-O} and higher.
7056 @item -ftree-loop-linear
7057 @opindex ftree-loop-linear
7058 Perform loop interchange transformations on tree. Same as
7059 @option{-floop-interchange}. To use this code transformation, GCC has
7060 to be configured with @option{--with-ppl} and @option{--with-cloog} to
7061 enable the Graphite loop transformation infrastructure.
7063 @item -floop-interchange
7064 @opindex floop-interchange
7065 Perform loop interchange transformations on loops. Interchanging two
7066 nested loops switches the inner and outer loops. For example, given a
7071 A(J, I) = A(J, I) * C
7075 loop interchange will transform the loop as if the user had written:
7079 A(J, I) = A(J, I) * C
7083 which can be beneficial when @code{N} is larger than the caches,
7084 because in Fortran, the elements of an array are stored in memory
7085 contiguously by column, and the original loop iterates over rows,
7086 potentially creating at each access a cache miss. This optimization
7087 applies to all the languages supported by GCC and is not limited to
7088 Fortran. To use this code transformation, GCC has to be configured
7089 with @option{--with-ppl} and @option{--with-cloog} to enable the
7090 Graphite loop transformation infrastructure.
7092 @item -floop-strip-mine
7093 @opindex floop-strip-mine
7094 Perform loop strip mining transformations on loops. Strip mining
7095 splits a loop into two nested loops. The outer loop has strides
7096 equal to the strip size and the inner loop has strides of the
7097 original loop within a strip. The strip length can be changed
7098 using the @option{loop-block-tile-size} parameter. For example,
7105 loop strip mining will transform the loop as if the user had written:
7108 DO I = II, min (II + 50, N)
7113 This optimization applies to all the languages supported by GCC and is
7114 not limited to Fortran. To use this code transformation, GCC has to
7115 be configured with @option{--with-ppl} and @option{--with-cloog} to
7116 enable the Graphite loop transformation infrastructure.
7119 @opindex floop-block
7120 Perform loop blocking transformations on loops. Blocking strip mines
7121 each loop in the loop nest such that the memory accesses of the
7122 element loops fit inside caches. The strip length can be changed
7123 using the @option{loop-block-tile-size} parameter. For example, given
7128 A(J, I) = B(I) + C(J)
7132 loop blocking will transform the loop as if the user had written:
7136 DO I = II, min (II + 50, N)
7137 DO J = JJ, min (JJ + 50, M)
7138 A(J, I) = B(I) + C(J)
7144 which can be beneficial when @code{M} is larger than the caches,
7145 because the innermost loop will iterate over a smaller amount of data
7146 that can be kept in the caches. This optimization applies to all the
7147 languages supported by GCC and is not limited to Fortran. To use this
7148 code transformation, GCC has to be configured with @option{--with-ppl}
7149 and @option{--with-cloog} to enable the Graphite loop transformation
7152 @item -fgraphite-identity
7153 @opindex fgraphite-identity
7154 Enable the identity transformation for graphite. For every SCoP we generate
7155 the polyhedral representation and transform it back to gimple. Using
7156 @option{-fgraphite-identity} we can check the costs or benefits of the
7157 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7158 are also performed by the code generator CLooG, like index splitting and
7159 dead code elimination in loops.
7161 @item -floop-flatten
7162 @opindex floop-flatten
7163 Removes the loop nesting structure: transforms the loop nest into a
7164 single loop. This transformation can be useful to vectorize all the
7165 levels of the loop nest.
7167 @item -floop-parallelize-all
7168 @opindex floop-parallelize-all
7169 Use the Graphite data dependence analysis to identify loops that can
7170 be parallelized. Parallelize all the loops that can be analyzed to
7171 not contain loop carried dependences without checking that it is
7172 profitable to parallelize the loops.
7174 @item -fcheck-data-deps
7175 @opindex fcheck-data-deps
7176 Compare the results of several data dependence analyzers. This option
7177 is used for debugging the data dependence analyzers.
7179 @item -ftree-loop-if-convert
7180 Attempt to transform conditional jumps in the innermost loops to
7181 branch-less equivalents. The intent is to remove control-flow from
7182 the innermost loops in order to improve the ability of the
7183 vectorization pass to handle these loops. This is enabled by default
7184 if vectorization is enabled.
7186 @item -ftree-loop-if-convert-stores
7187 Attempt to also if-convert conditional jumps containing memory writes.
7188 This transformation can be unsafe for multi-threaded programs as it
7189 transforms conditional memory writes into unconditional memory writes.
7192 for (i = 0; i < N; i++)
7196 would be transformed to
7198 for (i = 0; i < N; i++)
7199 A[i] = cond ? expr : A[i];
7201 potentially producing data races.
7203 @item -ftree-loop-distribution
7204 Perform loop distribution. This flag can improve cache performance on
7205 big loop bodies and allow further loop optimizations, like
7206 parallelization or vectorization, to take place. For example, the loop
7223 @item -ftree-loop-distribute-patterns
7224 Perform loop distribution of patterns that can be code generated with
7225 calls to a library. This flag is enabled by default at @option{-O3}.
7227 This pass distributes the initialization loops and generates a call to
7228 memset zero. For example, the loop
7244 and the initialization loop is transformed into a call to memset zero.
7246 @item -ftree-loop-im
7247 @opindex ftree-loop-im
7248 Perform loop invariant motion on trees. This pass moves only invariants that
7249 would be hard to handle at RTL level (function calls, operations that expand to
7250 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7251 operands of conditions that are invariant out of the loop, so that we can use
7252 just trivial invariantness analysis in loop unswitching. The pass also includes
7255 @item -ftree-loop-ivcanon
7256 @opindex ftree-loop-ivcanon
7257 Create a canonical counter for number of iterations in the loop for that
7258 determining number of iterations requires complicated analysis. Later
7259 optimizations then may determine the number easily. Useful especially
7260 in connection with unrolling.
7264 Perform induction variable optimizations (strength reduction, induction
7265 variable merging and induction variable elimination) on trees.
7267 @item -ftree-parallelize-loops=n
7268 @opindex ftree-parallelize-loops
7269 Parallelize loops, i.e., split their iteration space to run in n threads.
7270 This is only possible for loops whose iterations are independent
7271 and can be arbitrarily reordered. The optimization is only
7272 profitable on multiprocessor machines, for loops that are CPU-intensive,
7273 rather than constrained e.g.@: by memory bandwidth. This option
7274 implies @option{-pthread}, and thus is only supported on targets
7275 that have support for @option{-pthread}.
7279 Perform function-local points-to analysis on trees. This flag is
7280 enabled by default at @option{-O} and higher.
7284 Perform scalar replacement of aggregates. This pass replaces structure
7285 references with scalars to prevent committing structures to memory too
7286 early. This flag is enabled by default at @option{-O} and higher.
7288 @item -ftree-copyrename
7289 @opindex ftree-copyrename
7290 Perform copy renaming on trees. This pass attempts to rename compiler
7291 temporaries to other variables at copy locations, usually resulting in
7292 variable names which more closely resemble the original variables. This flag
7293 is enabled by default at @option{-O} and higher.
7297 Perform temporary expression replacement during the SSA->normal phase. Single
7298 use/single def temporaries are replaced at their use location with their
7299 defining expression. This results in non-GIMPLE code, but gives the expanders
7300 much more complex trees to work on resulting in better RTL generation. This is
7301 enabled by default at @option{-O} and higher.
7303 @item -ftree-vectorize
7304 @opindex ftree-vectorize
7305 Perform loop vectorization on trees. This flag is enabled by default at
7308 @item -ftree-slp-vectorize
7309 @opindex ftree-slp-vectorize
7310 Perform basic block vectorization on trees. This flag is enabled by default at
7311 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7313 @item -ftree-vect-loop-version
7314 @opindex ftree-vect-loop-version
7315 Perform loop versioning when doing loop vectorization on trees. When a loop
7316 appears to be vectorizable except that data alignment or data dependence cannot
7317 be determined at compile time then vectorized and non-vectorized versions of
7318 the loop are generated along with runtime checks for alignment or dependence
7319 to control which version is executed. This option is enabled by default
7320 except at level @option{-Os} where it is disabled.
7322 @item -fvect-cost-model
7323 @opindex fvect-cost-model
7324 Enable cost model for vectorization.
7328 Perform Value Range Propagation on trees. This is similar to the
7329 constant propagation pass, but instead of values, ranges of values are
7330 propagated. This allows the optimizers to remove unnecessary range
7331 checks like array bound checks and null pointer checks. This is
7332 enabled by default at @option{-O2} and higher. Null pointer check
7333 elimination is only done if @option{-fdelete-null-pointer-checks} is
7338 Perform tail duplication to enlarge superblock size. This transformation
7339 simplifies the control flow of the function allowing other optimizations to do
7342 @item -funroll-loops
7343 @opindex funroll-loops
7344 Unroll loops whose number of iterations can be determined at compile
7345 time or upon entry to the loop. @option{-funroll-loops} implies
7346 @option{-frerun-cse-after-loop}. This option makes code larger,
7347 and may or may not make it run faster.
7349 @item -funroll-all-loops
7350 @opindex funroll-all-loops
7351 Unroll all loops, even if their number of iterations is uncertain when
7352 the loop is entered. This usually makes programs run more slowly.
7353 @option{-funroll-all-loops} implies the same options as
7354 @option{-funroll-loops},
7356 @item -fsplit-ivs-in-unroller
7357 @opindex fsplit-ivs-in-unroller
7358 Enables expressing of values of induction variables in later iterations
7359 of the unrolled loop using the value in the first iteration. This breaks
7360 long dependency chains, thus improving efficiency of the scheduling passes.
7362 Combination of @option{-fweb} and CSE is often sufficient to obtain the
7363 same effect. However in cases the loop body is more complicated than
7364 a single basic block, this is not reliable. It also does not work at all
7365 on some of the architectures due to restrictions in the CSE pass.
7367 This optimization is enabled by default.
7369 @item -fvariable-expansion-in-unroller
7370 @opindex fvariable-expansion-in-unroller
7371 With this option, the compiler will create multiple copies of some
7372 local variables when unrolling a loop which can result in superior code.
7374 @item -fpartial-inlining
7375 @opindex fpartial-inlining
7376 Inline parts of functions. This option has any effect only
7377 when inlining itself is turned on by the @option{-finline-functions}
7378 or @option{-finline-small-functions} options.
7380 Enabled at level @option{-O2}.
7382 @item -fpredictive-commoning
7383 @opindex fpredictive-commoning
7384 Perform predictive commoning optimization, i.e., reusing computations
7385 (especially memory loads and stores) performed in previous
7386 iterations of loops.
7388 This option is enabled at level @option{-O3}.
7390 @item -fprefetch-loop-arrays
7391 @opindex fprefetch-loop-arrays
7392 If supported by the target machine, generate instructions to prefetch
7393 memory to improve the performance of loops that access large arrays.
7395 This option may generate better or worse code; results are highly
7396 dependent on the structure of loops within the source code.
7398 Disabled at level @option{-Os}.
7401 @itemx -fno-peephole2
7402 @opindex fno-peephole
7403 @opindex fno-peephole2
7404 Disable any machine-specific peephole optimizations. The difference
7405 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7406 are implemented in the compiler; some targets use one, some use the
7407 other, a few use both.
7409 @option{-fpeephole} is enabled by default.
7410 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7412 @item -fno-guess-branch-probability
7413 @opindex fno-guess-branch-probability
7414 Do not guess branch probabilities using heuristics.
7416 GCC will use heuristics to guess branch probabilities if they are
7417 not provided by profiling feedback (@option{-fprofile-arcs}). These
7418 heuristics are based on the control flow graph. If some branch probabilities
7419 are specified by @samp{__builtin_expect}, then the heuristics will be
7420 used to guess branch probabilities for the rest of the control flow graph,
7421 taking the @samp{__builtin_expect} info into account. The interactions
7422 between the heuristics and @samp{__builtin_expect} can be complex, and in
7423 some cases, it may be useful to disable the heuristics so that the effects
7424 of @samp{__builtin_expect} are easier to understand.
7426 The default is @option{-fguess-branch-probability} at levels
7427 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7429 @item -freorder-blocks
7430 @opindex freorder-blocks
7431 Reorder basic blocks in the compiled function in order to reduce number of
7432 taken branches and improve code locality.
7434 Enabled at levels @option{-O2}, @option{-O3}.
7436 @item -freorder-blocks-and-partition
7437 @opindex freorder-blocks-and-partition
7438 In addition to reordering basic blocks in the compiled function, in order
7439 to reduce number of taken branches, partitions hot and cold basic blocks
7440 into separate sections of the assembly and .o files, to improve
7441 paging and cache locality performance.
7443 This optimization is automatically turned off in the presence of
7444 exception handling, for linkonce sections, for functions with a user-defined
7445 section attribute and on any architecture that does not support named
7448 @item -freorder-functions
7449 @opindex freorder-functions
7450 Reorder functions in the object file in order to
7451 improve code locality. This is implemented by using special
7452 subsections @code{.text.hot} for most frequently executed functions and
7453 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7454 the linker so object file format must support named sections and linker must
7455 place them in a reasonable way.
7457 Also profile feedback must be available in to make this option effective. See
7458 @option{-fprofile-arcs} for details.
7460 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7462 @item -fstrict-aliasing
7463 @opindex fstrict-aliasing
7464 Allow the compiler to assume the strictest aliasing rules applicable to
7465 the language being compiled. For C (and C++), this activates
7466 optimizations based on the type of expressions. In particular, an
7467 object of one type is assumed never to reside at the same address as an
7468 object of a different type, unless the types are almost the same. For
7469 example, an @code{unsigned int} can alias an @code{int}, but not a
7470 @code{void*} or a @code{double}. A character type may alias any other
7473 @anchor{Type-punning}Pay special attention to code like this:
7486 The practice of reading from a different union member than the one most
7487 recently written to (called ``type-punning'') is common. Even with
7488 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7489 is accessed through the union type. So, the code above will work as
7490 expected. @xref{Structures unions enumerations and bit-fields
7491 implementation}. However, this code might not:
7502 Similarly, access by taking the address, casting the resulting pointer
7503 and dereferencing the result has undefined behavior, even if the cast
7504 uses a union type, e.g.:
7508 return ((union a_union *) &d)->i;
7512 The @option{-fstrict-aliasing} option is enabled at levels
7513 @option{-O2}, @option{-O3}, @option{-Os}.
7515 @item -fstrict-overflow
7516 @opindex fstrict-overflow
7517 Allow the compiler to assume strict signed overflow rules, depending
7518 on the language being compiled. For C (and C++) this means that
7519 overflow when doing arithmetic with signed numbers is undefined, which
7520 means that the compiler may assume that it will not happen. This
7521 permits various optimizations. For example, the compiler will assume
7522 that an expression like @code{i + 10 > i} will always be true for
7523 signed @code{i}. This assumption is only valid if signed overflow is
7524 undefined, as the expression is false if @code{i + 10} overflows when
7525 using twos complement arithmetic. When this option is in effect any
7526 attempt to determine whether an operation on signed numbers will
7527 overflow must be written carefully to not actually involve overflow.
7529 This option also allows the compiler to assume strict pointer
7530 semantics: given a pointer to an object, if adding an offset to that
7531 pointer does not produce a pointer to the same object, the addition is
7532 undefined. This permits the compiler to conclude that @code{p + u >
7533 p} is always true for a pointer @code{p} and unsigned integer
7534 @code{u}. This assumption is only valid because pointer wraparound is
7535 undefined, as the expression is false if @code{p + u} overflows using
7536 twos complement arithmetic.
7538 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7539 that integer signed overflow is fully defined: it wraps. When
7540 @option{-fwrapv} is used, there is no difference between
7541 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7542 integers. With @option{-fwrapv} certain types of overflow are
7543 permitted. For example, if the compiler gets an overflow when doing
7544 arithmetic on constants, the overflowed value can still be used with
7545 @option{-fwrapv}, but not otherwise.
7547 The @option{-fstrict-overflow} option is enabled at levels
7548 @option{-O2}, @option{-O3}, @option{-Os}.
7550 @item -falign-functions
7551 @itemx -falign-functions=@var{n}
7552 @opindex falign-functions
7553 Align the start of functions to the next power-of-two greater than
7554 @var{n}, skipping up to @var{n} bytes. For instance,
7555 @option{-falign-functions=32} aligns functions to the next 32-byte
7556 boundary, but @option{-falign-functions=24} would align to the next
7557 32-byte boundary only if this can be done by skipping 23 bytes or less.
7559 @option{-fno-align-functions} and @option{-falign-functions=1} are
7560 equivalent and mean that functions will not be aligned.
7562 Some assemblers only support this flag when @var{n} is a power of two;
7563 in that case, it is rounded up.
7565 If @var{n} is not specified or is zero, use a machine-dependent default.
7567 Enabled at levels @option{-O2}, @option{-O3}.
7569 @item -falign-labels
7570 @itemx -falign-labels=@var{n}
7571 @opindex falign-labels
7572 Align all branch targets to a power-of-two boundary, skipping up to
7573 @var{n} bytes like @option{-falign-functions}. This option can easily
7574 make code slower, because it must insert dummy operations for when the
7575 branch target is reached in the usual flow of the code.
7577 @option{-fno-align-labels} and @option{-falign-labels=1} are
7578 equivalent and mean that labels will not be aligned.
7580 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7581 are greater than this value, then their values are used instead.
7583 If @var{n} is not specified or is zero, use a machine-dependent default
7584 which is very likely to be @samp{1}, meaning no alignment.
7586 Enabled at levels @option{-O2}, @option{-O3}.
7589 @itemx -falign-loops=@var{n}
7590 @opindex falign-loops
7591 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7592 like @option{-falign-functions}. The hope is that the loop will be
7593 executed many times, which will make up for any execution of the dummy
7596 @option{-fno-align-loops} and @option{-falign-loops=1} are
7597 equivalent and mean that loops will not be aligned.
7599 If @var{n} is not specified or is zero, use a machine-dependent default.
7601 Enabled at levels @option{-O2}, @option{-O3}.
7604 @itemx -falign-jumps=@var{n}
7605 @opindex falign-jumps
7606 Align branch targets to a power-of-two boundary, for branch targets
7607 where the targets can only be reached by jumping, skipping up to @var{n}
7608 bytes like @option{-falign-functions}. In this case, no dummy operations
7611 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7612 equivalent and mean that loops will not be aligned.
7614 If @var{n} is not specified or is zero, use a machine-dependent default.
7616 Enabled at levels @option{-O2}, @option{-O3}.
7618 @item -funit-at-a-time
7619 @opindex funit-at-a-time
7620 This option is left for compatibility reasons. @option{-funit-at-a-time}
7621 has no effect, while @option{-fno-unit-at-a-time} implies
7622 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7626 @item -fno-toplevel-reorder
7627 @opindex fno-toplevel-reorder
7628 Do not reorder top-level functions, variables, and @code{asm}
7629 statements. Output them in the same order that they appear in the
7630 input file. When this option is used, unreferenced static variables
7631 will not be removed. This option is intended to support existing code
7632 which relies on a particular ordering. For new code, it is better to
7635 Enabled at level @option{-O0}. When disabled explicitly, it also imply
7636 @option{-fno-section-anchors} that is otherwise enabled at @option{-O0} on some
7641 Constructs webs as commonly used for register allocation purposes and assign
7642 each web individual pseudo register. This allows the register allocation pass
7643 to operate on pseudos directly, but also strengthens several other optimization
7644 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
7645 however, make debugging impossible, since variables will no longer stay in a
7648 Enabled by default with @option{-funroll-loops}.
7650 @item -fwhole-program
7651 @opindex fwhole-program
7652 Assume that the current compilation unit represents the whole program being
7653 compiled. All public functions and variables with the exception of @code{main}
7654 and those merged by attribute @code{externally_visible} become static functions
7655 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.
7656 While this option is equivalent to proper use of the @code{static} keyword for
7657 programs consisting of a single file, in combination with option
7658 @option{-flto} this flag can be used to
7659 compile many smaller scale programs since the functions and variables become
7660 local for the whole combined compilation unit, not for the single source file
7663 This option implies @option{-fwhole-file} for Fortran programs.
7665 @item -flto[=@var{n}]
7667 This option runs the standard link-time optimizer. When invoked
7668 with source code, it generates GIMPLE (one of GCC's internal
7669 representations) and writes it to special ELF sections in the object
7670 file. When the object files are linked together, all the function
7671 bodies are read from these ELF sections and instantiated as if they
7672 had been part of the same translation unit.
7674 To use the link-timer optimizer, @option{-flto} needs to be specified at
7675 compile time and during the final link. For example,
7678 gcc -c -O2 -flto foo.c
7679 gcc -c -O2 -flto bar.c
7680 gcc -o myprog -flto -O2 foo.o bar.o
7683 The first two invocations to GCC will save a bytecode representation
7684 of GIMPLE into special ELF sections inside @file{foo.o} and
7685 @file{bar.o}. The final invocation will read the GIMPLE bytecode from
7686 @file{foo.o} and @file{bar.o}, merge the two files into a single
7687 internal image, and compile the result as usual. Since both
7688 @file{foo.o} and @file{bar.o} are merged into a single image, this
7689 causes all the inter-procedural analyses and optimizations in GCC to
7690 work across the two files as if they were a single one. This means,
7691 for example, that the inliner will be able to inline functions in
7692 @file{bar.o} into functions in @file{foo.o} and vice-versa.
7694 Another (simpler) way to enable link-time optimization is,
7697 gcc -o myprog -flto -O2 foo.c bar.c
7700 The above will generate bytecode for @file{foo.c} and @file{bar.c},
7701 merge them together into a single GIMPLE representation and optimize
7702 them as usual to produce @file{myprog}.
7704 The only important thing to keep in mind is that to enable link-time
7705 optimizations the @option{-flto} flag needs to be passed to both the
7706 compile and the link commands.
7708 To make whole program optimization effective, it is necessary to make
7709 certain whole program assumptions. The compiler needs to know
7710 what functions and variables can be accessed by libraries and runtime
7711 outside of the link time optimized unit. When supported by the linker,
7712 the linker plugin (see @option{-fuse-linker-plugin}) passes to the
7713 compiler information about used and externally visible symbols. When
7714 the linker plugin is not available, @option{-fwhole-program} should be
7715 used to allow the compiler to make these assumptions, which will lead
7716 to more aggressive optimization decisions.
7718 Note that when a file is compiled with @option{-flto}, the generated
7719 object file will be larger than a regular object file because it will
7720 contain GIMPLE bytecodes and the usual final code. This means that
7721 object files with LTO information can be linked as a normal object
7722 file. So, in the previous example, if the final link is done with
7725 gcc -o myprog foo.o bar.o
7728 The only difference will be that no inter-procedural optimizations
7729 will be applied to produce @file{myprog}. The two object files
7730 @file{foo.o} and @file{bar.o} will be simply sent to the regular
7733 Additionally, the optimization flags used to compile individual files
7734 are not necessarily related to those used at link-time. For instance,
7737 gcc -c -O0 -flto foo.c
7738 gcc -c -O0 -flto bar.c
7739 gcc -o myprog -flto -O3 foo.o bar.o
7742 This will produce individual object files with unoptimized assembler
7743 code, but the resulting binary @file{myprog} will be optimized at
7744 @option{-O3}. Now, if the final binary is generated without
7745 @option{-flto}, then @file{myprog} will not be optimized.
7747 When producing the final binary with @option{-flto}, GCC will only
7748 apply link-time optimizations to those files that contain bytecode.
7749 Therefore, you can mix and match object files and libraries with
7750 GIMPLE bytecodes and final object code. GCC will automatically select
7751 which files to optimize in LTO mode and which files to link without
7754 There are some code generation flags that GCC will preserve when
7755 generating bytecodes, as they need to be used during the final link
7756 stage. Currently, the following options are saved into the GIMPLE
7757 bytecode files: @option{-fPIC}, @option{-fcommon} and all the
7758 @option{-m} target flags.
7760 At link time, these options are read-in and reapplied. Note that the
7761 current implementation makes no attempt at recognizing conflicting
7762 values for these options. If two or more files have a conflicting
7763 value (e.g., one file is compiled with @option{-fPIC} and another
7764 isn't), the compiler will simply use the last value read from the
7765 bytecode files. It is recommended, then, that all the files
7766 participating in the same link be compiled with the same options.
7768 Another feature of LTO is that it is possible to apply interprocedural
7769 optimizations on files written in different languages. This requires
7770 some support in the language front end. Currently, the C, C++ and
7771 Fortran front ends are capable of emitting GIMPLE bytecodes, so
7772 something like this should work
7777 gfortran -c -flto baz.f90
7778 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
7781 Notice that the final link is done with @command{g++} to get the C++
7782 runtime libraries and @option{-lgfortran} is added to get the Fortran
7783 runtime libraries. In general, when mixing languages in LTO mode, you
7784 should use the same link command used when mixing languages in a
7785 regular (non-LTO) compilation. This means that if your build process
7786 was mixing languages before, all you need to add is @option{-flto} to
7787 all the compile and link commands.
7789 If LTO encounters objects with C linkage declared with incompatible
7790 types in separate translation units to be linked together (undefined
7791 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
7792 issued. The behavior is still undefined at runtime.
7794 If object files containing GIMPLE bytecode are stored in a library archive, say
7795 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
7796 are using a linker with linker plugin support. To enable this feature, use
7797 the flag @option{-fuse-linker-plugin} at link-time:
7800 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
7803 With the linker plugin enabled, the linker will extract the needed
7804 GIMPLE files from @file{libfoo.a} and pass them on to the running GCC
7805 to make them part of the aggregated GIMPLE image to be optimized.
7807 If you are not using a linker with linker plugin support and/or do not
7808 enable linker plugin then the objects inside @file{libfoo.a}
7809 will be extracted and linked as usual, but they will not participate
7810 in the LTO optimization process.
7812 Link time optimizations do not require the presence of the whole program to
7813 operate. If the program does not require any symbols to be exported, it is
7814 possible to combine @option{-flto} and with @option{-fwhole-program} to allow
7815 the interprocedural optimizers to use more aggressive assumptions which may
7816 lead to improved optimization opportunities.
7817 Use of @option{-fwhole-program} is not needed when linker plugin is
7818 active (see @option{-fuse-linker-plugin}).
7820 Regarding portability: the current implementation of LTO makes no
7821 attempt at generating bytecode that can be ported between different
7822 types of hosts. The bytecode files are versioned and there is a
7823 strict version check, so bytecode files generated in one version of
7824 GCC will not work with an older/newer version of GCC.
7826 Link time optimization does not play well with generating debugging
7827 information. Combining @option{-flto} with
7828 @option{-g} is currently experimental and expected to produce wrong
7831 If you specify the optional @var{n}, the optimization and code
7832 generation done at link time is executed in parallel using @var{n}
7833 parallel jobs by utilizing an installed @command{make} program. The
7834 environment variable @env{MAKE} may be used to override the program
7835 used. The default value for @var{n} is 1.
7837 You can also specify @option{-flto=jobserver} to use GNU make's
7838 job server mode to determine the number of parallel jobs. This
7839 is useful when the Makefile calling GCC is already executing in parallel.
7840 The parent Makefile will need a @samp{+} prepended to the command recipe
7841 for this to work. This will likely only work if @env{MAKE} is
7844 This option is disabled by default.
7846 @item -flto-partition=@var{alg}
7847 @opindex flto-partition
7848 Specify the partitioning algorithm used by the link time optimizer.
7849 The value is either @code{1to1} to specify a partitioning mirroring
7850 the original source files or @code{balanced} to specify partitioning
7851 into equally sized chunks (whenever possible). Specifying @code{none}
7852 as an algorithm disables partitioning and streaming completely. The
7853 default value is @code{balanced}.
7855 @item -flto-compression-level=@var{n}
7856 This option specifies the level of compression used for intermediate
7857 language written to LTO object files, and is only meaningful in
7858 conjunction with LTO mode (@option{-flto}). Valid
7859 values are 0 (no compression) to 9 (maximum compression). Values
7860 outside this range are clamped to either 0 or 9. If the option is not
7861 given, a default balanced compression setting is used.
7864 Prints a report with internal details on the workings of the link-time
7865 optimizer. The contents of this report vary from version to version,
7866 it is meant to be useful to GCC developers when processing object
7867 files in LTO mode (via @option{-flto}).
7869 Disabled by default.
7871 @item -fuse-linker-plugin
7872 Enables the use of a linker plugin during link time optimization. This
7873 option relies on plugin support in the linker, which is available in gold
7874 or in GNU ld 2.21 or newer.
7876 This option enables the extraction of object files with GIMPLE bytecode out
7877 of library archives. This improves the quality of optimization by exposing
7878 more code to the link time optimizer. This information specifies what
7879 symbols can be accessed externally (by non-LTO object or during dynamic
7880 linking). Resulting code quality improvements on binaries (and shared
7881 libraries that use hidden visibility) are similar to @code{-fwhole-program}.
7882 See @option{-flto} for a description of the effect of this flag and how to
7885 Enabled by default when LTO support in GCC is enabled and GCC was compiled
7886 with a linker supporting plugins (GNU ld 2.21 or newer or gold).
7888 @item -fcompare-elim
7889 @opindex fcompare-elim
7890 After register allocation and post-register allocation instruction splitting,
7891 identify arithmetic instructions that compute processor flags similar to a
7892 comparison operation based on that arithmetic. If possible, eliminate the
7893 explicit comparison operation.
7895 This pass only applies to certain targets that cannot explicitly represent
7896 the comparison operation before register allocation is complete.
7898 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7900 @item -fcprop-registers
7901 @opindex fcprop-registers
7902 After register allocation and post-register allocation instruction splitting,
7903 we perform a copy-propagation pass to try to reduce scheduling dependencies
7904 and occasionally eliminate the copy.
7906 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7908 @item -fprofile-correction
7909 @opindex fprofile-correction
7910 Profiles collected using an instrumented binary for multi-threaded programs may
7911 be inconsistent due to missed counter updates. When this option is specified,
7912 GCC will use heuristics to correct or smooth out such inconsistencies. By
7913 default, GCC will emit an error message when an inconsistent profile is detected.
7915 @item -fprofile-dir=@var{path}
7916 @opindex fprofile-dir
7918 Set the directory to search for the profile data files in to @var{path}.
7919 This option affects only the profile data generated by
7920 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
7921 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
7922 and its related options. Both absolute and relative paths can be used.
7923 By default, GCC will use the current directory as @var{path}, thus the
7924 profile data file will appear in the same directory as the object file.
7926 @item -fprofile-generate
7927 @itemx -fprofile-generate=@var{path}
7928 @opindex fprofile-generate
7930 Enable options usually used for instrumenting application to produce
7931 profile useful for later recompilation with profile feedback based
7932 optimization. You must use @option{-fprofile-generate} both when
7933 compiling and when linking your program.
7935 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
7937 If @var{path} is specified, GCC will look at the @var{path} to find
7938 the profile feedback data files. See @option{-fprofile-dir}.
7941 @itemx -fprofile-use=@var{path}
7942 @opindex fprofile-use
7943 Enable profile feedback directed optimizations, and optimizations
7944 generally profitable only with profile feedback available.
7946 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
7947 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
7949 By default, GCC emits an error message if the feedback profiles do not
7950 match the source code. This error can be turned into a warning by using
7951 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
7954 If @var{path} is specified, GCC will look at the @var{path} to find
7955 the profile feedback data files. See @option{-fprofile-dir}.
7958 The following options control compiler behavior regarding floating
7959 point arithmetic. These options trade off between speed and
7960 correctness. All must be specifically enabled.
7964 @opindex ffloat-store
7965 Do not store floating point variables in registers, and inhibit other
7966 options that might change whether a floating point value is taken from a
7969 @cindex floating point precision
7970 This option prevents undesirable excess precision on machines such as
7971 the 68000 where the floating registers (of the 68881) keep more
7972 precision than a @code{double} is supposed to have. Similarly for the
7973 x86 architecture. For most programs, the excess precision does only
7974 good, but a few programs rely on the precise definition of IEEE floating
7975 point. Use @option{-ffloat-store} for such programs, after modifying
7976 them to store all pertinent intermediate computations into variables.
7978 @item -fexcess-precision=@var{style}
7979 @opindex fexcess-precision
7980 This option allows further control over excess precision on machines
7981 where floating-point registers have more precision than the IEEE
7982 @code{float} and @code{double} types and the processor does not
7983 support operations rounding to those types. By default,
7984 @option{-fexcess-precision=fast} is in effect; this means that
7985 operations are carried out in the precision of the registers and that
7986 it is unpredictable when rounding to the types specified in the source
7987 code takes place. When compiling C, if
7988 @option{-fexcess-precision=standard} is specified then excess
7989 precision will follow the rules specified in ISO C99; in particular,
7990 both casts and assignments cause values to be rounded to their
7991 semantic types (whereas @option{-ffloat-store} only affects
7992 assignments). This option is enabled by default for C if a strict
7993 conformance option such as @option{-std=c99} is used.
7996 @option{-fexcess-precision=standard} is not implemented for languages
7997 other than C, and has no effect if
7998 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
7999 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8000 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8001 semantics apply without excess precision, and in the latter, rounding
8006 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8007 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8008 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8010 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8012 This option is not turned on by any @option{-O} option besides
8013 @option{-Ofast} since it can result in incorrect output for programs
8014 which depend on an exact implementation of IEEE or ISO rules/specifications
8015 for math functions. It may, however, yield faster code for programs
8016 that do not require the guarantees of these specifications.
8018 @item -fno-math-errno
8019 @opindex fno-math-errno
8020 Do not set ERRNO after calling math functions that are executed
8021 with a single instruction, e.g., sqrt. A program that relies on
8022 IEEE exceptions for math error handling may want to use this flag
8023 for speed while maintaining IEEE arithmetic compatibility.
8025 This option is not turned on by any @option{-O} option since
8026 it can result in incorrect output for programs which depend on
8027 an exact implementation of IEEE or ISO rules/specifications for
8028 math functions. It may, however, yield faster code for programs
8029 that do not require the guarantees of these specifications.
8031 The default is @option{-fmath-errno}.
8033 On Darwin systems, the math library never sets @code{errno}. There is
8034 therefore no reason for the compiler to consider the possibility that
8035 it might, and @option{-fno-math-errno} is the default.
8037 @item -funsafe-math-optimizations
8038 @opindex funsafe-math-optimizations
8040 Allow optimizations for floating-point arithmetic that (a) assume
8041 that arguments and results are valid and (b) may violate IEEE or
8042 ANSI standards. When used at link-time, it may include libraries
8043 or startup files that change the default FPU control word or other
8044 similar optimizations.
8046 This option is not turned on by any @option{-O} option since
8047 it can result in incorrect output for programs which depend on
8048 an exact implementation of IEEE or ISO rules/specifications for
8049 math functions. It may, however, yield faster code for programs
8050 that do not require the guarantees of these specifications.
8051 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8052 @option{-fassociative-math} and @option{-freciprocal-math}.
8054 The default is @option{-fno-unsafe-math-optimizations}.
8056 @item -fassociative-math
8057 @opindex fassociative-math
8059 Allow re-association of operands in series of floating-point operations.
8060 This violates the ISO C and C++ language standard by possibly changing
8061 computation result. NOTE: re-ordering may change the sign of zero as
8062 well as ignore NaNs and inhibit or create underflow or overflow (and
8063 thus cannot be used on a code which relies on rounding behavior like
8064 @code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
8065 and thus may not be used when ordered comparisons are required.
8066 This option requires that both @option{-fno-signed-zeros} and
8067 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8068 much sense with @option{-frounding-math}. For Fortran the option
8069 is automatically enabled when both @option{-fno-signed-zeros} and
8070 @option{-fno-trapping-math} are in effect.
8072 The default is @option{-fno-associative-math}.
8074 @item -freciprocal-math
8075 @opindex freciprocal-math
8077 Allow the reciprocal of a value to be used instead of dividing by
8078 the value if this enables optimizations. For example @code{x / y}
8079 can be replaced with @code{x * (1/y)} which is useful if @code{(1/y)}
8080 is subject to common subexpression elimination. Note that this loses
8081 precision and increases the number of flops operating on the value.
8083 The default is @option{-fno-reciprocal-math}.
8085 @item -ffinite-math-only
8086 @opindex ffinite-math-only
8087 Allow optimizations for floating-point arithmetic that assume
8088 that arguments and results are not NaNs or +-Infs.
8090 This option is not turned on by any @option{-O} option since
8091 it can result in incorrect output for programs which depend on
8092 an exact implementation of IEEE or ISO rules/specifications for
8093 math functions. It may, however, yield faster code for programs
8094 that do not require the guarantees of these specifications.
8096 The default is @option{-fno-finite-math-only}.
8098 @item -fno-signed-zeros
8099 @opindex fno-signed-zeros
8100 Allow optimizations for floating point arithmetic that ignore the
8101 signedness of zero. IEEE arithmetic specifies the behavior of
8102 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8103 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8104 This option implies that the sign of a zero result isn't significant.
8106 The default is @option{-fsigned-zeros}.
8108 @item -fno-trapping-math
8109 @opindex fno-trapping-math
8110 Compile code assuming that floating-point operations cannot generate
8111 user-visible traps. These traps include division by zero, overflow,
8112 underflow, inexact result and invalid operation. This option requires
8113 that @option{-fno-signaling-nans} be in effect. Setting this option may
8114 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8116 This option should never be turned on by any @option{-O} option since
8117 it can result in incorrect output for programs which depend on
8118 an exact implementation of IEEE or ISO rules/specifications for
8121 The default is @option{-ftrapping-math}.
8123 @item -frounding-math
8124 @opindex frounding-math
8125 Disable transformations and optimizations that assume default floating
8126 point rounding behavior. This is round-to-zero for all floating point
8127 to integer conversions, and round-to-nearest for all other arithmetic
8128 truncations. This option should be specified for programs that change
8129 the FP rounding mode dynamically, or that may be executed with a
8130 non-default rounding mode. This option disables constant folding of
8131 floating point expressions at compile-time (which may be affected by
8132 rounding mode) and arithmetic transformations that are unsafe in the
8133 presence of sign-dependent rounding modes.
8135 The default is @option{-fno-rounding-math}.
8137 This option is experimental and does not currently guarantee to
8138 disable all GCC optimizations that are affected by rounding mode.
8139 Future versions of GCC may provide finer control of this setting
8140 using C99's @code{FENV_ACCESS} pragma. This command line option
8141 will be used to specify the default state for @code{FENV_ACCESS}.
8143 @item -fsignaling-nans
8144 @opindex fsignaling-nans
8145 Compile code assuming that IEEE signaling NaNs may generate user-visible
8146 traps during floating-point operations. Setting this option disables
8147 optimizations that may change the number of exceptions visible with
8148 signaling NaNs. This option implies @option{-ftrapping-math}.
8150 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8153 The default is @option{-fno-signaling-nans}.
8155 This option is experimental and does not currently guarantee to
8156 disable all GCC optimizations that affect signaling NaN behavior.
8158 @item -fsingle-precision-constant
8159 @opindex fsingle-precision-constant
8160 Treat floating point constant as single precision constant instead of
8161 implicitly converting it to double precision constant.
8163 @item -fcx-limited-range
8164 @opindex fcx-limited-range
8165 When enabled, this option states that a range reduction step is not
8166 needed when performing complex division. Also, there is no checking
8167 whether the result of a complex multiplication or division is @code{NaN
8168 + I*NaN}, with an attempt to rescue the situation in that case. The
8169 default is @option{-fno-cx-limited-range}, but is enabled by
8170 @option{-ffast-math}.
8172 This option controls the default setting of the ISO C99
8173 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8176 @item -fcx-fortran-rules
8177 @opindex fcx-fortran-rules
8178 Complex multiplication and division follow Fortran rules. Range
8179 reduction is done as part of complex division, but there is no checking
8180 whether the result of a complex multiplication or division is @code{NaN
8181 + I*NaN}, with an attempt to rescue the situation in that case.
8183 The default is @option{-fno-cx-fortran-rules}.
8187 The following options control optimizations that may improve
8188 performance, but are not enabled by any @option{-O} options. This
8189 section includes experimental options that may produce broken code.
8192 @item -fbranch-probabilities
8193 @opindex fbranch-probabilities
8194 After running a program compiled with @option{-fprofile-arcs}
8195 (@pxref{Debugging Options,, Options for Debugging Your Program or
8196 @command{gcc}}), you can compile it a second time using
8197 @option{-fbranch-probabilities}, to improve optimizations based on
8198 the number of times each branch was taken. When the program
8199 compiled with @option{-fprofile-arcs} exits it saves arc execution
8200 counts to a file called @file{@var{sourcename}.gcda} for each source
8201 file. The information in this data file is very dependent on the
8202 structure of the generated code, so you must use the same source code
8203 and the same optimization options for both compilations.
8205 With @option{-fbranch-probabilities}, GCC puts a
8206 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8207 These can be used to improve optimization. Currently, they are only
8208 used in one place: in @file{reorg.c}, instead of guessing which path a
8209 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8210 exactly determine which path is taken more often.
8212 @item -fprofile-values
8213 @opindex fprofile-values
8214 If combined with @option{-fprofile-arcs}, it adds code so that some
8215 data about values of expressions in the program is gathered.
8217 With @option{-fbranch-probabilities}, it reads back the data gathered
8218 from profiling values of expressions for usage in optimizations.
8220 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8224 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
8225 a code to gather information about values of expressions.
8227 With @option{-fbranch-probabilities}, it reads back the data gathered
8228 and actually performs the optimizations based on them.
8229 Currently the optimizations include specialization of division operation
8230 using the knowledge about the value of the denominator.
8232 @item -frename-registers
8233 @opindex frename-registers
8234 Attempt to avoid false dependencies in scheduled code by making use
8235 of registers left over after register allocation. This optimization
8236 will most benefit processors with lots of registers. Depending on the
8237 debug information format adopted by the target, however, it can
8238 make debugging impossible, since variables will no longer stay in
8239 a ``home register''.
8241 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
8245 Perform tail duplication to enlarge superblock size. This transformation
8246 simplifies the control flow of the function allowing other optimizations to do
8249 Enabled with @option{-fprofile-use}.
8251 @item -funroll-loops
8252 @opindex funroll-loops
8253 Unroll loops whose number of iterations can be determined at compile time or
8254 upon entry to the loop. @option{-funroll-loops} implies
8255 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8256 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8257 small constant number of iterations). This option makes code larger, and may
8258 or may not make it run faster.
8260 Enabled with @option{-fprofile-use}.
8262 @item -funroll-all-loops
8263 @opindex funroll-all-loops
8264 Unroll all loops, even if their number of iterations is uncertain when
8265 the loop is entered. This usually makes programs run more slowly.
8266 @option{-funroll-all-loops} implies the same options as
8267 @option{-funroll-loops}.
8270 @opindex fpeel-loops
8271 Peels the loops for that there is enough information that they do not
8272 roll much (from profile feedback). It also turns on complete loop peeling
8273 (i.e.@: complete removal of loops with small constant number of iterations).
8275 Enabled with @option{-fprofile-use}.
8277 @item -fmove-loop-invariants
8278 @opindex fmove-loop-invariants
8279 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8280 at level @option{-O1}
8282 @item -funswitch-loops
8283 @opindex funswitch-loops
8284 Move branches with loop invariant conditions out of the loop, with duplicates
8285 of the loop on both branches (modified according to result of the condition).
8287 @item -ffunction-sections
8288 @itemx -fdata-sections
8289 @opindex ffunction-sections
8290 @opindex fdata-sections
8291 Place each function or data item into its own section in the output
8292 file if the target supports arbitrary sections. The name of the
8293 function or the name of the data item determines the section's name
8296 Use these options on systems where the linker can perform optimizations
8297 to improve locality of reference in the instruction space. Most systems
8298 using the ELF object format and SPARC processors running Solaris 2 have
8299 linkers with such optimizations. AIX may have these optimizations in
8302 Only use these options when there are significant benefits from doing
8303 so. When you specify these options, the assembler and linker will
8304 create larger object and executable files and will also be slower.
8305 You will not be able to use @code{gprof} on all systems if you
8306 specify this option and you may have problems with debugging if
8307 you specify both this option and @option{-g}.
8309 @item -fbranch-target-load-optimize
8310 @opindex fbranch-target-load-optimize
8311 Perform branch target register load optimization before prologue / epilogue
8313 The use of target registers can typically be exposed only during reload,
8314 thus hoisting loads out of loops and doing inter-block scheduling needs
8315 a separate optimization pass.
8317 @item -fbranch-target-load-optimize2
8318 @opindex fbranch-target-load-optimize2
8319 Perform branch target register load optimization after prologue / epilogue
8322 @item -fbtr-bb-exclusive
8323 @opindex fbtr-bb-exclusive
8324 When performing branch target register load optimization, don't reuse
8325 branch target registers in within any basic block.
8327 @item -fstack-protector
8328 @opindex fstack-protector
8329 Emit extra code to check for buffer overflows, such as stack smashing
8330 attacks. This is done by adding a guard variable to functions with
8331 vulnerable objects. This includes functions that call alloca, and
8332 functions with buffers larger than 8 bytes. The guards are initialized
8333 when a function is entered and then checked when the function exits.
8334 If a guard check fails, an error message is printed and the program exits.
8336 @item -fstack-protector-all
8337 @opindex fstack-protector-all
8338 Like @option{-fstack-protector} except that all functions are protected.
8340 @item -fsection-anchors
8341 @opindex fsection-anchors
8342 Try to reduce the number of symbolic address calculations by using
8343 shared ``anchor'' symbols to address nearby objects. This transformation
8344 can help to reduce the number of GOT entries and GOT accesses on some
8347 For example, the implementation of the following function @code{foo}:
8351 int foo (void) @{ return a + b + c; @}
8354 would usually calculate the addresses of all three variables, but if you
8355 compile it with @option{-fsection-anchors}, it will access the variables
8356 from a common anchor point instead. The effect is similar to the
8357 following pseudocode (which isn't valid C):
8362 register int *xr = &x;
8363 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8367 Not all targets support this option.
8369 @item --param @var{name}=@var{value}
8371 In some places, GCC uses various constants to control the amount of
8372 optimization that is done. For example, GCC will not inline functions
8373 that contain more that a certain number of instructions. You can
8374 control some of these constants on the command-line using the
8375 @option{--param} option.
8377 The names of specific parameters, and the meaning of the values, are
8378 tied to the internals of the compiler, and are subject to change
8379 without notice in future releases.
8381 In each case, the @var{value} is an integer. The allowable choices for
8382 @var{name} are given in the following table:
8385 @item predictable-branch-outcome
8386 When branch is predicted to be taken with probability lower than this threshold
8387 (in percent), then it is considered well predictable. The default is 10.
8389 @item max-crossjump-edges
8390 The maximum number of incoming edges to consider for crossjumping.
8391 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8392 the number of edges incoming to each block. Increasing values mean
8393 more aggressive optimization, making the compile time increase with
8394 probably small improvement in executable size.
8396 @item min-crossjump-insns
8397 The minimum number of instructions which must be matched at the end
8398 of two blocks before crossjumping will be performed on them. This
8399 value is ignored in the case where all instructions in the block being
8400 crossjumped from are matched. The default value is 5.
8402 @item max-grow-copy-bb-insns
8403 The maximum code size expansion factor when copying basic blocks
8404 instead of jumping. The expansion is relative to a jump instruction.
8405 The default value is 8.
8407 @item max-goto-duplication-insns
8408 The maximum number of instructions to duplicate to a block that jumps
8409 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8410 passes, GCC factors computed gotos early in the compilation process,
8411 and unfactors them as late as possible. Only computed jumps at the
8412 end of a basic blocks with no more than max-goto-duplication-insns are
8413 unfactored. The default value is 8.
8415 @item max-delay-slot-insn-search
8416 The maximum number of instructions to consider when looking for an
8417 instruction to fill a delay slot. If more than this arbitrary number of
8418 instructions is searched, the time savings from filling the delay slot
8419 will be minimal so stop searching. Increasing values mean more
8420 aggressive optimization, making the compile time increase with probably
8421 small improvement in executable run time.
8423 @item max-delay-slot-live-search
8424 When trying to fill delay slots, the maximum number of instructions to
8425 consider when searching for a block with valid live register
8426 information. Increasing this arbitrarily chosen value means more
8427 aggressive optimization, increasing the compile time. This parameter
8428 should be removed when the delay slot code is rewritten to maintain the
8431 @item max-gcse-memory
8432 The approximate maximum amount of memory that will be allocated in
8433 order to perform the global common subexpression elimination
8434 optimization. If more memory than specified is required, the
8435 optimization will not be done.
8437 @item max-gcse-insertion-ratio
8438 If the ratio of expression insertions to deletions is larger than this value
8439 for any expression, then RTL PRE will insert or remove the expression and thus
8440 leave partially redundant computations in the instruction stream. The default value is 20.
8442 @item max-pending-list-length
8443 The maximum number of pending dependencies scheduling will allow
8444 before flushing the current state and starting over. Large functions
8445 with few branches or calls can create excessively large lists which
8446 needlessly consume memory and resources.
8448 @item max-inline-insns-single
8449 Several parameters control the tree inliner used in gcc.
8450 This number sets the maximum number of instructions (counted in GCC's
8451 internal representation) in a single function that the tree inliner
8452 will consider for inlining. This only affects functions declared
8453 inline and methods implemented in a class declaration (C++).
8454 The default value is 400.
8456 @item max-inline-insns-auto
8457 When you use @option{-finline-functions} (included in @option{-O3}),
8458 a lot of functions that would otherwise not be considered for inlining
8459 by the compiler will be investigated. To those functions, a different
8460 (more restrictive) limit compared to functions declared inline can
8462 The default value is 40.
8464 @item large-function-insns
8465 The limit specifying really large functions. For functions larger than this
8466 limit after inlining, inlining is constrained by
8467 @option{--param large-function-growth}. This parameter is useful primarily
8468 to avoid extreme compilation time caused by non-linear algorithms used by the
8470 The default value is 2700.
8472 @item large-function-growth
8473 Specifies maximal growth of large function caused by inlining in percents.
8474 The default value is 100 which limits large function growth to 2.0 times
8477 @item large-unit-insns
8478 The limit specifying large translation unit. Growth caused by inlining of
8479 units larger than this limit is limited by @option{--param inline-unit-growth}.
8480 For small units this might be too tight (consider unit consisting of function A
8481 that is inline and B that just calls A three time. If B is small relative to
8482 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8483 large units consisting of small inlineable functions however the overall unit
8484 growth limit is needed to avoid exponential explosion of code size. Thus for
8485 smaller units, the size is increased to @option{--param large-unit-insns}
8486 before applying @option{--param inline-unit-growth}. The default is 10000
8488 @item inline-unit-growth
8489 Specifies maximal overall growth of the compilation unit caused by inlining.
8490 The default value is 30 which limits unit growth to 1.3 times the original
8493 @item ipcp-unit-growth
8494 Specifies maximal overall growth of the compilation unit caused by
8495 interprocedural constant propagation. The default value is 10 which limits
8496 unit growth to 1.1 times the original size.
8498 @item large-stack-frame
8499 The limit specifying large stack frames. While inlining the algorithm is trying
8500 to not grow past this limit too much. Default value is 256 bytes.
8502 @item large-stack-frame-growth
8503 Specifies maximal growth of large stack frames caused by inlining in percents.
8504 The default value is 1000 which limits large stack frame growth to 11 times
8507 @item max-inline-insns-recursive
8508 @itemx max-inline-insns-recursive-auto
8509 Specifies maximum number of instructions out-of-line copy of self recursive inline
8510 function can grow into by performing recursive inlining.
8512 For functions declared inline @option{--param max-inline-insns-recursive} is
8513 taken into account. For function not declared inline, recursive inlining
8514 happens only when @option{-finline-functions} (included in @option{-O3}) is
8515 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
8516 default value is 450.
8518 @item max-inline-recursive-depth
8519 @itemx max-inline-recursive-depth-auto
8520 Specifies maximum recursion depth used by the recursive inlining.
8522 For functions declared inline @option{--param max-inline-recursive-depth} is
8523 taken into account. For function not declared inline, recursive inlining
8524 happens only when @option{-finline-functions} (included in @option{-O3}) is
8525 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
8528 @item min-inline-recursive-probability
8529 Recursive inlining is profitable only for function having deep recursion
8530 in average and can hurt for function having little recursion depth by
8531 increasing the prologue size or complexity of function body to other
8534 When profile feedback is available (see @option{-fprofile-generate}) the actual
8535 recursion depth can be guessed from probability that function will recurse via
8536 given call expression. This parameter limits inlining only to call expression
8537 whose probability exceeds given threshold (in percents). The default value is
8540 @item early-inlining-insns
8541 Specify growth that early inliner can make. In effect it increases amount of
8542 inlining for code having large abstraction penalty. The default value is 10.
8544 @item max-early-inliner-iterations
8545 @itemx max-early-inliner-iterations
8546 Limit of iterations of early inliner. This basically bounds number of nested
8547 indirect calls early inliner can resolve. Deeper chains are still handled by
8550 @item comdat-sharing-probability
8551 @itemx comdat-sharing-probability
8552 Probability (in percent) that C++ inline function with comdat visibility
8553 will be shared across multiple compilation units. The default value is 20.
8555 @item min-vect-loop-bound
8556 The minimum number of iterations under which a loop will not get vectorized
8557 when @option{-ftree-vectorize} is used. The number of iterations after
8558 vectorization needs to be greater than the value specified by this option
8559 to allow vectorization. The default value is 0.
8561 @item gcse-cost-distance-ratio
8562 Scaling factor in calculation of maximum distance an expression
8563 can be moved by GCSE optimizations. This is currently supported only in the
8564 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
8565 will be with simple expressions, i.e., the expressions which have cost
8566 less than @option{gcse-unrestricted-cost}. Specifying 0 will disable
8567 hoisting of simple expressions. The default value is 10.
8569 @item gcse-unrestricted-cost
8570 Cost, roughly measured as the cost of a single typical machine
8571 instruction, at which GCSE optimizations will not constrain
8572 the distance an expression can travel. This is currently
8573 supported only in the code hoisting pass. The lesser the cost,
8574 the more aggressive code hoisting will be. Specifying 0 will
8575 allow all expressions to travel unrestricted distances.
8576 The default value is 3.
8578 @item max-hoist-depth
8579 The depth of search in the dominator tree for expressions to hoist.
8580 This is used to avoid quadratic behavior in hoisting algorithm.
8581 The value of 0 will avoid limiting the search, but may slow down compilation
8582 of huge functions. The default value is 30.
8584 @item max-unrolled-insns
8585 The maximum number of instructions that a loop should have if that loop
8586 is unrolled, and if the loop is unrolled, it determines how many times
8587 the loop code is unrolled.
8589 @item max-average-unrolled-insns
8590 The maximum number of instructions biased by probabilities of their execution
8591 that a loop should have if that loop is unrolled, and if the loop is unrolled,
8592 it determines how many times the loop code is unrolled.
8594 @item max-unroll-times
8595 The maximum number of unrollings of a single loop.
8597 @item max-peeled-insns
8598 The maximum number of instructions that a loop should have if that loop
8599 is peeled, and if the loop is peeled, it determines how many times
8600 the loop code is peeled.
8602 @item max-peel-times
8603 The maximum number of peelings of a single loop.
8605 @item max-completely-peeled-insns
8606 The maximum number of insns of a completely peeled loop.
8608 @item max-completely-peel-times
8609 The maximum number of iterations of a loop to be suitable for complete peeling.
8611 @item max-completely-peel-loop-nest-depth
8612 The maximum depth of a loop nest suitable for complete peeling.
8614 @item max-unswitch-insns
8615 The maximum number of insns of an unswitched loop.
8617 @item max-unswitch-level
8618 The maximum number of branches unswitched in a single loop.
8621 The minimum cost of an expensive expression in the loop invariant motion.
8623 @item iv-consider-all-candidates-bound
8624 Bound on number of candidates for induction variables below that
8625 all candidates are considered for each use in induction variable
8626 optimizations. Only the most relevant candidates are considered
8627 if there are more candidates, to avoid quadratic time complexity.
8629 @item iv-max-considered-uses
8630 The induction variable optimizations give up on loops that contain more
8631 induction variable uses.
8633 @item iv-always-prune-cand-set-bound
8634 If number of candidates in the set is smaller than this value,
8635 we always try to remove unnecessary ivs from the set during its
8636 optimization when a new iv is added to the set.
8638 @item scev-max-expr-size
8639 Bound on size of expressions used in the scalar evolutions analyzer.
8640 Large expressions slow the analyzer.
8642 @item scev-max-expr-complexity
8643 Bound on the complexity of the expressions in the scalar evolutions analyzer.
8644 Complex expressions slow the analyzer.
8646 @item omega-max-vars
8647 The maximum number of variables in an Omega constraint system.
8648 The default value is 128.
8650 @item omega-max-geqs
8651 The maximum number of inequalities in an Omega constraint system.
8652 The default value is 256.
8655 The maximum number of equalities in an Omega constraint system.
8656 The default value is 128.
8658 @item omega-max-wild-cards
8659 The maximum number of wildcard variables that the Omega solver will
8660 be able to insert. The default value is 18.
8662 @item omega-hash-table-size
8663 The size of the hash table in the Omega solver. The default value is
8666 @item omega-max-keys
8667 The maximal number of keys used by the Omega solver. The default
8670 @item omega-eliminate-redundant-constraints
8671 When set to 1, use expensive methods to eliminate all redundant
8672 constraints. The default value is 0.
8674 @item vect-max-version-for-alignment-checks
8675 The maximum number of runtime checks that can be performed when
8676 doing loop versioning for alignment in the vectorizer. See option
8677 ftree-vect-loop-version for more information.
8679 @item vect-max-version-for-alias-checks
8680 The maximum number of runtime checks that can be performed when
8681 doing loop versioning for alias in the vectorizer. See option
8682 ftree-vect-loop-version for more information.
8684 @item max-iterations-to-track
8686 The maximum number of iterations of a loop the brute force algorithm
8687 for analysis of # of iterations of the loop tries to evaluate.
8689 @item hot-bb-count-fraction
8690 Select fraction of the maximal count of repetitions of basic block in program
8691 given basic block needs to have to be considered hot.
8693 @item hot-bb-frequency-fraction
8694 Select fraction of the entry block frequency of executions of basic block in
8695 function given basic block needs to have to be considered hot.
8697 @item max-predicted-iterations
8698 The maximum number of loop iterations we predict statically. This is useful
8699 in cases where function contain single loop with known bound and other loop
8700 with unknown. We predict the known number of iterations correctly, while
8701 the unknown number of iterations average to roughly 10. This means that the
8702 loop without bounds would appear artificially cold relative to the other one.
8704 @item align-threshold
8706 Select fraction of the maximal frequency of executions of basic block in
8707 function given basic block will get aligned.
8709 @item align-loop-iterations
8711 A loop expected to iterate at lest the selected number of iterations will get
8714 @item tracer-dynamic-coverage
8715 @itemx tracer-dynamic-coverage-feedback
8717 This value is used to limit superblock formation once the given percentage of
8718 executed instructions is covered. This limits unnecessary code size
8721 The @option{tracer-dynamic-coverage-feedback} is used only when profile
8722 feedback is available. The real profiles (as opposed to statically estimated
8723 ones) are much less balanced allowing the threshold to be larger value.
8725 @item tracer-max-code-growth
8726 Stop tail duplication once code growth has reached given percentage. This is
8727 rather hokey argument, as most of the duplicates will be eliminated later in
8728 cross jumping, so it may be set to much higher values than is the desired code
8731 @item tracer-min-branch-ratio
8733 Stop reverse growth when the reverse probability of best edge is less than this
8734 threshold (in percent).
8736 @item tracer-min-branch-ratio
8737 @itemx tracer-min-branch-ratio-feedback
8739 Stop forward growth if the best edge do have probability lower than this
8742 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
8743 compilation for profile feedback and one for compilation without. The value
8744 for compilation with profile feedback needs to be more conservative (higher) in
8745 order to make tracer effective.
8747 @item max-cse-path-length
8749 Maximum number of basic blocks on path that cse considers. The default is 10.
8752 The maximum instructions CSE process before flushing. The default is 1000.
8754 @item ggc-min-expand
8756 GCC uses a garbage collector to manage its own memory allocation. This
8757 parameter specifies the minimum percentage by which the garbage
8758 collector's heap should be allowed to expand between collections.
8759 Tuning this may improve compilation speed; it has no effect on code
8762 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
8763 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
8764 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
8765 GCC is not able to calculate RAM on a particular platform, the lower
8766 bound of 30% is used. Setting this parameter and
8767 @option{ggc-min-heapsize} to zero causes a full collection to occur at
8768 every opportunity. This is extremely slow, but can be useful for
8771 @item ggc-min-heapsize
8773 Minimum size of the garbage collector's heap before it begins bothering
8774 to collect garbage. The first collection occurs after the heap expands
8775 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
8776 tuning this may improve compilation speed, and has no effect on code
8779 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
8780 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
8781 with a lower bound of 4096 (four megabytes) and an upper bound of
8782 131072 (128 megabytes). If GCC is not able to calculate RAM on a
8783 particular platform, the lower bound is used. Setting this parameter
8784 very large effectively disables garbage collection. Setting this
8785 parameter and @option{ggc-min-expand} to zero causes a full collection
8786 to occur at every opportunity.
8788 @item max-reload-search-insns
8789 The maximum number of instruction reload should look backward for equivalent
8790 register. Increasing values mean more aggressive optimization, making the
8791 compile time increase with probably slightly better performance. The default
8794 @item max-cselib-memory-locations
8795 The maximum number of memory locations cselib should take into account.
8796 Increasing values mean more aggressive optimization, making the compile time
8797 increase with probably slightly better performance. The default value is 500.
8799 @item reorder-blocks-duplicate
8800 @itemx reorder-blocks-duplicate-feedback
8802 Used by basic block reordering pass to decide whether to use unconditional
8803 branch or duplicate the code on its destination. Code is duplicated when its
8804 estimated size is smaller than this value multiplied by the estimated size of
8805 unconditional jump in the hot spots of the program.
8807 The @option{reorder-block-duplicate-feedback} is used only when profile
8808 feedback is available and may be set to higher values than
8809 @option{reorder-block-duplicate} since information about the hot spots is more
8812 @item max-sched-ready-insns
8813 The maximum number of instructions ready to be issued the scheduler should
8814 consider at any given time during the first scheduling pass. Increasing
8815 values mean more thorough searches, making the compilation time increase
8816 with probably little benefit. The default value is 100.
8818 @item max-sched-region-blocks
8819 The maximum number of blocks in a region to be considered for
8820 interblock scheduling. The default value is 10.
8822 @item max-pipeline-region-blocks
8823 The maximum number of blocks in a region to be considered for
8824 pipelining in the selective scheduler. The default value is 15.
8826 @item max-sched-region-insns
8827 The maximum number of insns in a region to be considered for
8828 interblock scheduling. The default value is 100.
8830 @item max-pipeline-region-insns
8831 The maximum number of insns in a region to be considered for
8832 pipelining in the selective scheduler. The default value is 200.
8835 The minimum probability (in percents) of reaching a source block
8836 for interblock speculative scheduling. The default value is 40.
8838 @item max-sched-extend-regions-iters
8839 The maximum number of iterations through CFG to extend regions.
8840 0 - disable region extension,
8841 N - do at most N iterations.
8842 The default value is 0.
8844 @item max-sched-insn-conflict-delay
8845 The maximum conflict delay for an insn to be considered for speculative motion.
8846 The default value is 3.
8848 @item sched-spec-prob-cutoff
8849 The minimal probability of speculation success (in percents), so that
8850 speculative insn will be scheduled.
8851 The default value is 40.
8853 @item sched-mem-true-dep-cost
8854 Minimal distance (in CPU cycles) between store and load targeting same
8855 memory locations. The default value is 1.
8857 @item selsched-max-lookahead
8858 The maximum size of the lookahead window of selective scheduling. It is a
8859 depth of search for available instructions.
8860 The default value is 50.
8862 @item selsched-max-sched-times
8863 The maximum number of times that an instruction will be scheduled during
8864 selective scheduling. This is the limit on the number of iterations
8865 through which the instruction may be pipelined. The default value is 2.
8867 @item selsched-max-insns-to-rename
8868 The maximum number of best instructions in the ready list that are considered
8869 for renaming in the selective scheduler. The default value is 2.
8872 The minimum value of stage count that swing modulo scheduler will
8873 generate. The default value is 2.
8875 @item max-last-value-rtl
8876 The maximum size measured as number of RTLs that can be recorded in an expression
8877 in combiner for a pseudo register as last known value of that register. The default
8880 @item integer-share-limit
8881 Small integer constants can use a shared data structure, reducing the
8882 compiler's memory usage and increasing its speed. This sets the maximum
8883 value of a shared integer constant. The default value is 256.
8885 @item min-virtual-mappings
8886 Specifies the minimum number of virtual mappings in the incremental
8887 SSA updater that should be registered to trigger the virtual mappings
8888 heuristic defined by virtual-mappings-ratio. The default value is
8891 @item virtual-mappings-ratio
8892 If the number of virtual mappings is virtual-mappings-ratio bigger
8893 than the number of virtual symbols to be updated, then the incremental
8894 SSA updater switches to a full update for those symbols. The default
8897 @item ssp-buffer-size
8898 The minimum size of buffers (i.e.@: arrays) that will receive stack smashing
8899 protection when @option{-fstack-protection} is used.
8901 @item max-jump-thread-duplication-stmts
8902 Maximum number of statements allowed in a block that needs to be
8903 duplicated when threading jumps.
8905 @item max-fields-for-field-sensitive
8906 Maximum number of fields in a structure we will treat in
8907 a field sensitive manner during pointer analysis. The default is zero
8908 for -O0, and -O1 and 100 for -Os, -O2, and -O3.
8910 @item prefetch-latency
8911 Estimate on average number of instructions that are executed before
8912 prefetch finishes. The distance we prefetch ahead is proportional
8913 to this constant. Increasing this number may also lead to less
8914 streams being prefetched (see @option{simultaneous-prefetches}).
8916 @item simultaneous-prefetches
8917 Maximum number of prefetches that can run at the same time.
8919 @item l1-cache-line-size
8920 The size of cache line in L1 cache, in bytes.
8923 The size of L1 cache, in kilobytes.
8926 The size of L2 cache, in kilobytes.
8928 @item min-insn-to-prefetch-ratio
8929 The minimum ratio between the number of instructions and the
8930 number of prefetches to enable prefetching in a loop.
8932 @item prefetch-min-insn-to-mem-ratio
8933 The minimum ratio between the number of instructions and the
8934 number of memory references to enable prefetching in a loop.
8936 @item use-canonical-types
8937 Whether the compiler should use the ``canonical'' type system. By
8938 default, this should always be 1, which uses a more efficient internal
8939 mechanism for comparing types in C++ and Objective-C++. However, if
8940 bugs in the canonical type system are causing compilation failures,
8941 set this value to 0 to disable canonical types.
8943 @item switch-conversion-max-branch-ratio
8944 Switch initialization conversion will refuse to create arrays that are
8945 bigger than @option{switch-conversion-max-branch-ratio} times the number of
8946 branches in the switch.
8948 @item max-partial-antic-length
8949 Maximum length of the partial antic set computed during the tree
8950 partial redundancy elimination optimization (@option{-ftree-pre}) when
8951 optimizing at @option{-O3} and above. For some sorts of source code
8952 the enhanced partial redundancy elimination optimization can run away,
8953 consuming all of the memory available on the host machine. This
8954 parameter sets a limit on the length of the sets that are computed,
8955 which prevents the runaway behavior. Setting a value of 0 for
8956 this parameter will allow an unlimited set length.
8958 @item sccvn-max-scc-size
8959 Maximum size of a strongly connected component (SCC) during SCCVN
8960 processing. If this limit is hit, SCCVN processing for the whole
8961 function will not be done and optimizations depending on it will
8962 be disabled. The default maximum SCC size is 10000.
8964 @item ira-max-loops-num
8965 IRA uses a regional register allocation by default. If a function
8966 contains loops more than number given by the parameter, only at most
8967 given number of the most frequently executed loops will form regions
8968 for the regional register allocation. The default value of the
8971 @item ira-max-conflict-table-size
8972 Although IRA uses a sophisticated algorithm of compression conflict
8973 table, the table can be still big for huge functions. If the conflict
8974 table for a function could be more than size in MB given by the
8975 parameter, the conflict table is not built and faster, simpler, and
8976 lower quality register allocation algorithm will be used. The
8977 algorithm do not use pseudo-register conflicts. The default value of
8978 the parameter is 2000.
8980 @item ira-loop-reserved-regs
8981 IRA can be used to evaluate more accurate register pressure in loops
8982 for decision to move loop invariants (see @option{-O3}). The number
8983 of available registers reserved for some other purposes is described
8984 by this parameter. The default value of the parameter is 2 which is
8985 minimal number of registers needed for execution of typical
8986 instruction. This value is the best found from numerous experiments.
8988 @item loop-invariant-max-bbs-in-loop
8989 Loop invariant motion can be very expensive, both in compile time and
8990 in amount of needed compile time memory, with very large loops. Loops
8991 with more basic blocks than this parameter won't have loop invariant
8992 motion optimization performed on them. The default value of the
8993 parameter is 1000 for -O1 and 10000 for -O2 and above.
8995 @item max-vartrack-size
8996 Sets a maximum number of hash table slots to use during variable
8997 tracking dataflow analysis of any function. If this limit is exceeded
8998 with variable tracking at assignments enabled, analysis for that
8999 function is retried without it, after removing all debug insns from
9000 the function. If the limit is exceeded even without debug insns, var
9001 tracking analysis is completely disabled for the function. Setting
9002 the parameter to zero makes it unlimited.
9004 @item max-vartrack-expr-depth
9005 Sets a maximum number of recursion levels when attempting to map
9006 variable names or debug temporaries to value expressions. This trades
9007 compile time for more complete debug information. If this is set too
9008 low, value expressions that are available and could be represented in
9009 debug information may end up not being used; setting this higher may
9010 enable the compiler to find more complex debug expressions, but compile
9011 time may grow exponentially, and even then, it may fail to find more
9012 usable expressions. The default is 10.
9014 @item min-nondebug-insn-uid
9015 Use uids starting at this parameter for nondebug insns. The range below
9016 the parameter is reserved exclusively for debug insns created by
9017 @option{-fvar-tracking-assignments}, but debug insns may get
9018 (non-overlapping) uids above it if the reserved range is exhausted.
9020 @item ipa-sra-ptr-growth-factor
9021 IPA-SRA will replace a pointer to an aggregate with one or more new
9022 parameters only when their cumulative size is less or equal to
9023 @option{ipa-sra-ptr-growth-factor} times the size of the original
9026 @item graphite-max-nb-scop-params
9027 To avoid exponential effects in the Graphite loop transforms, the
9028 number of parameters in a Static Control Part (SCoP) is bounded. The
9029 default value is 10 parameters. A variable whose value is unknown at
9030 compile time and defined outside a SCoP is a parameter of the SCoP.
9032 @item graphite-max-bbs-per-function
9033 To avoid exponential effects in the detection of SCoPs, the size of
9034 the functions analyzed by Graphite is bounded. The default value is
9037 @item loop-block-tile-size
9038 Loop blocking or strip mining transforms, enabled with
9039 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9040 loop in the loop nest by a given number of iterations. The strip
9041 length can be changed using the @option{loop-block-tile-size}
9042 parameter. The default value is 51 iterations.
9044 @item ipa-cp-value-list-size
9045 IPA-CP attempts to track all possible values and types passed to a function's
9046 parameter in order to propagate them and perform devirtualization.
9047 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9048 stores per one formal parameter of a function.
9050 @item lto-partitions
9051 Specify desired number of partitions produced during WHOPR compilation.
9052 The number of partitions should exceed the number of CPUs used for compilation.
9053 The default value is 32.
9055 @item lto-minpartition
9056 Size of minimal partition for WHOPR (in estimated instructions).
9057 This prevents expenses of splitting very small programs into too many
9060 @item cxx-max-namespaces-for-diagnostic-help
9061 The maximum number of namespaces to consult for suggestions when C++
9062 name lookup fails for an identifier. The default is 1000.
9064 @item max-stores-to-sink
9065 The maximum number of conditional stores paires that can be sunk. Set to 0
9066 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9067 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9069 @item allow-store-data-races
9070 Allow optimizers to introduce new data races on stores.
9071 Set to 1 to allow, otherwise to 0. This option is enabled by default
9072 unless implicitly set by the @option{-fmemory-model=} option.
9074 @item case-values-threshold
9075 The smallest number of different values for which it is best to use a
9076 jump-table instead of a tree of conditional branches. If the value is
9077 0, use the default for the machine. The default is 0.
9082 @node Preprocessor Options
9083 @section Options Controlling the Preprocessor
9084 @cindex preprocessor options
9085 @cindex options, preprocessor
9087 These options control the C preprocessor, which is run on each C source
9088 file before actual compilation.
9090 If you use the @option{-E} option, nothing is done except preprocessing.
9091 Some of these options make sense only together with @option{-E} because
9092 they cause the preprocessor output to be unsuitable for actual
9096 @item -Wp,@var{option}
9098 You can use @option{-Wp,@var{option}} to bypass the compiler driver
9099 and pass @var{option} directly through to the preprocessor. If
9100 @var{option} contains commas, it is split into multiple options at the
9101 commas. However, many options are modified, translated or interpreted
9102 by the compiler driver before being passed to the preprocessor, and
9103 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
9104 interface is undocumented and subject to change, so whenever possible
9105 you should avoid using @option{-Wp} and let the driver handle the
9108 @item -Xpreprocessor @var{option}
9109 @opindex Xpreprocessor
9110 Pass @var{option} as an option to the preprocessor. You can use this to
9111 supply system-specific preprocessor options which GCC does not know how to
9114 If you want to pass an option that takes an argument, you must use
9115 @option{-Xpreprocessor} twice, once for the option and once for the argument.
9118 @include cppopts.texi
9120 @node Assembler Options
9121 @section Passing Options to the Assembler
9123 @c prevent bad page break with this line
9124 You can pass options to the assembler.
9127 @item -Wa,@var{option}
9129 Pass @var{option} as an option to the assembler. If @var{option}
9130 contains commas, it is split into multiple options at the commas.
9132 @item -Xassembler @var{option}
9134 Pass @var{option} as an option to the assembler. You can use this to
9135 supply system-specific assembler options which GCC does not know how to
9138 If you want to pass an option that takes an argument, you must use
9139 @option{-Xassembler} twice, once for the option and once for the argument.
9144 @section Options for Linking
9145 @cindex link options
9146 @cindex options, linking
9148 These options come into play when the compiler links object files into
9149 an executable output file. They are meaningless if the compiler is
9150 not doing a link step.
9154 @item @var{object-file-name}
9155 A file name that does not end in a special recognized suffix is
9156 considered to name an object file or library. (Object files are
9157 distinguished from libraries by the linker according to the file
9158 contents.) If linking is done, these object files are used as input
9167 If any of these options is used, then the linker is not run, and
9168 object file names should not be used as arguments. @xref{Overall
9172 @item -l@var{library}
9173 @itemx -l @var{library}
9175 Search the library named @var{library} when linking. (The second
9176 alternative with the library as a separate argument is only for
9177 POSIX compliance and is not recommended.)
9179 It makes a difference where in the command you write this option; the
9180 linker searches and processes libraries and object files in the order they
9181 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
9182 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
9183 to functions in @samp{z}, those functions may not be loaded.
9185 The linker searches a standard list of directories for the library,
9186 which is actually a file named @file{lib@var{library}.a}. The linker
9187 then uses this file as if it had been specified precisely by name.
9189 The directories searched include several standard system directories
9190 plus any that you specify with @option{-L}.
9192 Normally the files found this way are library files---archive files
9193 whose members are object files. The linker handles an archive file by
9194 scanning through it for members which define symbols that have so far
9195 been referenced but not defined. But if the file that is found is an
9196 ordinary object file, it is linked in the usual fashion. The only
9197 difference between using an @option{-l} option and specifying a file name
9198 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
9199 and searches several directories.
9203 You need this special case of the @option{-l} option in order to
9204 link an Objective-C or Objective-C++ program.
9207 @opindex nostartfiles
9208 Do not use the standard system startup files when linking.
9209 The standard system libraries are used normally, unless @option{-nostdlib}
9210 or @option{-nodefaultlibs} is used.
9212 @item -nodefaultlibs
9213 @opindex nodefaultlibs
9214 Do not use the standard system libraries when linking.
9215 Only the libraries you specify will be passed to the linker, options
9216 specifying linkage of the system libraries, such as @code{-static-libgcc}
9217 or @code{-shared-libgcc}, will be ignored.
9218 The standard startup files are used normally, unless @option{-nostartfiles}
9219 is used. The compiler may generate calls to @code{memcmp},
9220 @code{memset}, @code{memcpy} and @code{memmove}.
9221 These entries are usually resolved by entries in
9222 libc. These entry points should be supplied through some other
9223 mechanism when this option is specified.
9227 Do not use the standard system startup files or libraries when linking.
9228 No startup files and only the libraries you specify will be passed to
9229 the linker, options specifying linkage of the system libraries, such as
9230 @code{-static-libgcc} or @code{-shared-libgcc}, will be ignored.
9231 The compiler may generate calls to @code{memcmp}, @code{memset},
9232 @code{memcpy} and @code{memmove}.
9233 These entries are usually resolved by entries in
9234 libc. These entry points should be supplied through some other
9235 mechanism when this option is specified.
9237 @cindex @option{-lgcc}, use with @option{-nostdlib}
9238 @cindex @option{-nostdlib} and unresolved references
9239 @cindex unresolved references and @option{-nostdlib}
9240 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
9241 @cindex @option{-nodefaultlibs} and unresolved references
9242 @cindex unresolved references and @option{-nodefaultlibs}
9243 One of the standard libraries bypassed by @option{-nostdlib} and
9244 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
9245 that GCC uses to overcome shortcomings of particular machines, or special
9246 needs for some languages.
9247 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
9248 Collection (GCC) Internals},
9249 for more discussion of @file{libgcc.a}.)
9250 In most cases, you need @file{libgcc.a} even when you want to avoid
9251 other standard libraries. In other words, when you specify @option{-nostdlib}
9252 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
9253 This ensures that you have no unresolved references to internal GCC
9254 library subroutines. (For example, @samp{__main}, used to ensure C++
9255 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
9256 GNU Compiler Collection (GCC) Internals}.)
9260 Produce a position independent executable on targets which support it.
9261 For predictable results, you must also specify the same set of options
9262 that were used to generate code (@option{-fpie}, @option{-fPIE},
9263 or model suboptions) when you specify this option.
9267 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
9268 that support it. This instructs the linker to add all symbols, not
9269 only used ones, to the dynamic symbol table. This option is needed
9270 for some uses of @code{dlopen} or to allow obtaining backtraces
9271 from within a program.
9275 Remove all symbol table and relocation information from the executable.
9279 On systems that support dynamic linking, this prevents linking with the shared
9280 libraries. On other systems, this option has no effect.
9284 Produce a shared object which can then be linked with other objects to
9285 form an executable. Not all systems support this option. For predictable
9286 results, you must also specify the same set of options that were used to
9287 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
9288 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
9289 needs to build supplementary stub code for constructors to work. On
9290 multi-libbed systems, @samp{gcc -shared} must select the correct support
9291 libraries to link against. Failing to supply the correct flags may lead
9292 to subtle defects. Supplying them in cases where they are not necessary
9295 @item -shared-libgcc
9296 @itemx -static-libgcc
9297 @opindex shared-libgcc
9298 @opindex static-libgcc
9299 On systems that provide @file{libgcc} as a shared library, these options
9300 force the use of either the shared or static version respectively.
9301 If no shared version of @file{libgcc} was built when the compiler was
9302 configured, these options have no effect.
9304 There are several situations in which an application should use the
9305 shared @file{libgcc} instead of the static version. The most common
9306 of these is when the application wishes to throw and catch exceptions
9307 across different shared libraries. In that case, each of the libraries
9308 as well as the application itself should use the shared @file{libgcc}.
9310 Therefore, the G++ and GCJ drivers automatically add
9311 @option{-shared-libgcc} whenever you build a shared library or a main
9312 executable, because C++ and Java programs typically use exceptions, so
9313 this is the right thing to do.
9315 If, instead, you use the GCC driver to create shared libraries, you may
9316 find that they will not always be linked with the shared @file{libgcc}.
9317 If GCC finds, at its configuration time, that you have a non-GNU linker
9318 or a GNU linker that does not support option @option{--eh-frame-hdr},
9319 it will link the shared version of @file{libgcc} into shared libraries
9320 by default. Otherwise, it will take advantage of the linker and optimize
9321 away the linking with the shared version of @file{libgcc}, linking with
9322 the static version of libgcc by default. This allows exceptions to
9323 propagate through such shared libraries, without incurring relocation
9324 costs at library load time.
9326 However, if a library or main executable is supposed to throw or catch
9327 exceptions, you must link it using the G++ or GCJ driver, as appropriate
9328 for the languages used in the program, or using the option
9329 @option{-shared-libgcc}, such that it is linked with the shared
9332 @item -static-libstdc++
9333 When the @command{g++} program is used to link a C++ program, it will
9334 normally automatically link against @option{libstdc++}. If
9335 @file{libstdc++} is available as a shared library, and the
9336 @option{-static} option is not used, then this will link against the
9337 shared version of @file{libstdc++}. That is normally fine. However, it
9338 is sometimes useful to freeze the version of @file{libstdc++} used by
9339 the program without going all the way to a fully static link. The
9340 @option{-static-libstdc++} option directs the @command{g++} driver to
9341 link @file{libstdc++} statically, without necessarily linking other
9342 libraries statically.
9346 Bind references to global symbols when building a shared object. Warn
9347 about any unresolved references (unless overridden by the link editor
9348 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
9351 @item -T @var{script}
9353 @cindex linker script
9354 Use @var{script} as the linker script. This option is supported by most
9355 systems using the GNU linker. On some targets, such as bare-board
9356 targets without an operating system, the @option{-T} option may be required
9357 when linking to avoid references to undefined symbols.
9359 @item -Xlinker @var{option}
9361 Pass @var{option} as an option to the linker. You can use this to
9362 supply system-specific linker options which GCC does not know how to
9365 If you want to pass an option that takes a separate argument, you must use
9366 @option{-Xlinker} twice, once for the option and once for the argument.
9367 For example, to pass @option{-assert definitions}, you must write
9368 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
9369 @option{-Xlinker "-assert definitions"}, because this passes the entire
9370 string as a single argument, which is not what the linker expects.
9372 When using the GNU linker, it is usually more convenient to pass
9373 arguments to linker options using the @option{@var{option}=@var{value}}
9374 syntax than as separate arguments. For example, you can specify
9375 @samp{-Xlinker -Map=output.map} rather than
9376 @samp{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
9377 this syntax for command-line options.
9379 @item -Wl,@var{option}
9381 Pass @var{option} as an option to the linker. If @var{option} contains
9382 commas, it is split into multiple options at the commas. You can use this
9383 syntax to pass an argument to the option.
9384 For example, @samp{-Wl,-Map,output.map} passes @samp{-Map output.map} to the
9385 linker. When using the GNU linker, you can also get the same effect with
9386 @samp{-Wl,-Map=output.map}.
9388 @item -u @var{symbol}
9390 Pretend the symbol @var{symbol} is undefined, to force linking of
9391 library modules to define it. You can use @option{-u} multiple times with
9392 different symbols to force loading of additional library modules.
9395 @node Directory Options
9396 @section Options for Directory Search
9397 @cindex directory options
9398 @cindex options, directory search
9401 These options specify directories to search for header files, for
9402 libraries and for parts of the compiler:
9407 Add the directory @var{dir} to the head of the list of directories to be
9408 searched for header files. This can be used to override a system header
9409 file, substituting your own version, since these directories are
9410 searched before the system header file directories. However, you should
9411 not use this option to add directories that contain vendor-supplied
9412 system header files (use @option{-isystem} for that). If you use more than
9413 one @option{-I} option, the directories are scanned in left-to-right
9414 order; the standard system directories come after.
9416 If a standard system include directory, or a directory specified with
9417 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
9418 option will be ignored. The directory will still be searched but as a
9419 system directory at its normal position in the system include chain.
9420 This is to ensure that GCC's procedure to fix buggy system headers and
9421 the ordering for the include_next directive are not inadvertently changed.
9422 If you really need to change the search order for system directories,
9423 use the @option{-nostdinc} and/or @option{-isystem} options.
9425 @item -iplugindir=@var{dir}
9426 Set the directory to search for plugins which are passed
9427 by @option{-fplugin=@var{name}} instead of
9428 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
9429 to be used by the user, but only passed by the driver.
9431 @item -iquote@var{dir}
9433 Add the directory @var{dir} to the head of the list of directories to
9434 be searched for header files only for the case of @samp{#include
9435 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
9436 otherwise just like @option{-I}.
9440 Add directory @var{dir} to the list of directories to be searched
9443 @item -B@var{prefix}
9445 This option specifies where to find the executables, libraries,
9446 include files, and data files of the compiler itself.
9448 The compiler driver program runs one or more of the subprograms
9449 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
9450 @var{prefix} as a prefix for each program it tries to run, both with and
9451 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
9453 For each subprogram to be run, the compiler driver first tries the
9454 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
9455 was not specified, the driver tries two standard prefixes, which are
9456 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
9457 those results in a file name that is found, the unmodified program
9458 name is searched for using the directories specified in your
9459 @env{PATH} environment variable.
9461 The compiler will check to see if the path provided by the @option{-B}
9462 refers to a directory, and if necessary it will add a directory
9463 separator character at the end of the path.
9465 @option{-B} prefixes that effectively specify directory names also apply
9466 to libraries in the linker, because the compiler translates these
9467 options into @option{-L} options for the linker. They also apply to
9468 includes files in the preprocessor, because the compiler translates these
9469 options into @option{-isystem} options for the preprocessor. In this case,
9470 the compiler appends @samp{include} to the prefix.
9472 The run-time support file @file{libgcc.a} can also be searched for using
9473 the @option{-B} prefix, if needed. If it is not found there, the two
9474 standard prefixes above are tried, and that is all. The file is left
9475 out of the link if it is not found by those means.
9477 Another way to specify a prefix much like the @option{-B} prefix is to use
9478 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
9481 As a special kludge, if the path provided by @option{-B} is
9482 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
9483 9, then it will be replaced by @file{[dir/]include}. This is to help
9484 with boot-strapping the compiler.
9486 @item -specs=@var{file}
9488 Process @var{file} after the compiler reads in the standard @file{specs}
9489 file, in order to override the defaults that the @file{gcc} driver
9490 program uses when determining what switches to pass to @file{cc1},
9491 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
9492 @option{-specs=@var{file}} can be specified on the command line, and they
9493 are processed in order, from left to right.
9495 @item --sysroot=@var{dir}
9497 Use @var{dir} as the logical root directory for headers and libraries.
9498 For example, if the compiler would normally search for headers in
9499 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
9500 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
9502 If you use both this option and the @option{-isysroot} option, then
9503 the @option{--sysroot} option will apply to libraries, but the
9504 @option{-isysroot} option will apply to header files.
9506 The GNU linker (beginning with version 2.16) has the necessary support
9507 for this option. If your linker does not support this option, the
9508 header file aspect of @option{--sysroot} will still work, but the
9509 library aspect will not.
9513 This option has been deprecated. Please use @option{-iquote} instead for
9514 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
9515 Any directories you specify with @option{-I} options before the @option{-I-}
9516 option are searched only for the case of @samp{#include "@var{file}"};
9517 they are not searched for @samp{#include <@var{file}>}.
9519 If additional directories are specified with @option{-I} options after
9520 the @option{-I-}, these directories are searched for all @samp{#include}
9521 directives. (Ordinarily @emph{all} @option{-I} directories are used
9524 In addition, the @option{-I-} option inhibits the use of the current
9525 directory (where the current input file came from) as the first search
9526 directory for @samp{#include "@var{file}"}. There is no way to
9527 override this effect of @option{-I-}. With @option{-I.} you can specify
9528 searching the directory which was current when the compiler was
9529 invoked. That is not exactly the same as what the preprocessor does
9530 by default, but it is often satisfactory.
9532 @option{-I-} does not inhibit the use of the standard system directories
9533 for header files. Thus, @option{-I-} and @option{-nostdinc} are
9540 @section Specifying subprocesses and the switches to pass to them
9543 @command{gcc} is a driver program. It performs its job by invoking a
9544 sequence of other programs to do the work of compiling, assembling and
9545 linking. GCC interprets its command-line parameters and uses these to
9546 deduce which programs it should invoke, and which command-line options
9547 it ought to place on their command lines. This behavior is controlled
9548 by @dfn{spec strings}. In most cases there is one spec string for each
9549 program that GCC can invoke, but a few programs have multiple spec
9550 strings to control their behavior. The spec strings built into GCC can
9551 be overridden by using the @option{-specs=} command-line switch to specify
9554 @dfn{Spec files} are plaintext files that are used to construct spec
9555 strings. They consist of a sequence of directives separated by blank
9556 lines. The type of directive is determined by the first non-whitespace
9557 character on the line and it can be one of the following:
9560 @item %@var{command}
9561 Issues a @var{command} to the spec file processor. The commands that can
9565 @item %include <@var{file}>
9566 @cindex @code{%include}
9567 Search for @var{file} and insert its text at the current point in the
9570 @item %include_noerr <@var{file}>
9571 @cindex @code{%include_noerr}
9572 Just like @samp{%include}, but do not generate an error message if the include
9573 file cannot be found.
9575 @item %rename @var{old_name} @var{new_name}
9576 @cindex @code{%rename}
9577 Rename the spec string @var{old_name} to @var{new_name}.
9581 @item *[@var{spec_name}]:
9582 This tells the compiler to create, override or delete the named spec
9583 string. All lines after this directive up to the next directive or
9584 blank line are considered to be the text for the spec string. If this
9585 results in an empty string then the spec will be deleted. (Or, if the
9586 spec did not exist, then nothing will happen.) Otherwise, if the spec
9587 does not currently exist a new spec will be created. If the spec does
9588 exist then its contents will be overridden by the text of this
9589 directive, unless the first character of that text is the @samp{+}
9590 character, in which case the text will be appended to the spec.
9592 @item [@var{suffix}]:
9593 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
9594 and up to the next directive or blank line are considered to make up the
9595 spec string for the indicated suffix. When the compiler encounters an
9596 input file with the named suffix, it will processes the spec string in
9597 order to work out how to compile that file. For example:
9604 This says that any input file whose name ends in @samp{.ZZ} should be
9605 passed to the program @samp{z-compile}, which should be invoked with the
9606 command-line switch @option{-input} and with the result of performing the
9607 @samp{%i} substitution. (See below.)
9609 As an alternative to providing a spec string, the text that follows a
9610 suffix directive can be one of the following:
9613 @item @@@var{language}
9614 This says that the suffix is an alias for a known @var{language}. This is
9615 similar to using the @option{-x} command-line switch to GCC to specify a
9616 language explicitly. For example:
9623 Says that .ZZ files are, in fact, C++ source files.
9626 This causes an error messages saying:
9629 @var{name} compiler not installed on this system.
9633 GCC already has an extensive list of suffixes built into it.
9634 This directive will add an entry to the end of the list of suffixes, but
9635 since the list is searched from the end backwards, it is effectively
9636 possible to override earlier entries using this technique.
9640 GCC has the following spec strings built into it. Spec files can
9641 override these strings or create their own. Note that individual
9642 targets can also add their own spec strings to this list.
9645 asm Options to pass to the assembler
9646 asm_final Options to pass to the assembler post-processor
9647 cpp Options to pass to the C preprocessor
9648 cc1 Options to pass to the C compiler
9649 cc1plus Options to pass to the C++ compiler
9650 endfile Object files to include at the end of the link
9651 link Options to pass to the linker
9652 lib Libraries to include on the command line to the linker
9653 libgcc Decides which GCC support library to pass to the linker
9654 linker Sets the name of the linker
9655 predefines Defines to be passed to the C preprocessor
9656 signed_char Defines to pass to CPP to say whether @code{char} is signed
9658 startfile Object files to include at the start of the link
9661 Here is a small example of a spec file:
9667 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
9670 This example renames the spec called @samp{lib} to @samp{old_lib} and
9671 then overrides the previous definition of @samp{lib} with a new one.
9672 The new definition adds in some extra command-line options before
9673 including the text of the old definition.
9675 @dfn{Spec strings} are a list of command-line options to be passed to their
9676 corresponding program. In addition, the spec strings can contain
9677 @samp{%}-prefixed sequences to substitute variable text or to
9678 conditionally insert text into the command line. Using these constructs
9679 it is possible to generate quite complex command lines.
9681 Here is a table of all defined @samp{%}-sequences for spec
9682 strings. Note that spaces are not generated automatically around the
9683 results of expanding these sequences. Therefore you can concatenate them
9684 together or combine them with constant text in a single argument.
9688 Substitute one @samp{%} into the program name or argument.
9691 Substitute the name of the input file being processed.
9694 Substitute the basename of the input file being processed.
9695 This is the substring up to (and not including) the last period
9696 and not including the directory.
9699 This is the same as @samp{%b}, but include the file suffix (text after
9703 Marks the argument containing or following the @samp{%d} as a
9704 temporary file name, so that that file will be deleted if GCC exits
9705 successfully. Unlike @samp{%g}, this contributes no text to the
9708 @item %g@var{suffix}
9709 Substitute a file name that has suffix @var{suffix} and is chosen
9710 once per compilation, and mark the argument in the same way as
9711 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
9712 name is now chosen in a way that is hard to predict even when previously
9713 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
9714 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
9715 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
9716 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
9717 was simply substituted with a file name chosen once per compilation,
9718 without regard to any appended suffix (which was therefore treated
9719 just like ordinary text), making such attacks more likely to succeed.
9721 @item %u@var{suffix}
9722 Like @samp{%g}, but generates a new temporary file name even if
9723 @samp{%u@var{suffix}} was already seen.
9725 @item %U@var{suffix}
9726 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
9727 new one if there is no such last file name. In the absence of any
9728 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
9729 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
9730 would involve the generation of two distinct file names, one
9731 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
9732 simply substituted with a file name chosen for the previous @samp{%u},
9733 without regard to any appended suffix.
9735 @item %j@var{suffix}
9736 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
9737 writable, and if save-temps is off; otherwise, substitute the name
9738 of a temporary file, just like @samp{%u}. This temporary file is not
9739 meant for communication between processes, but rather as a junk
9742 @item %|@var{suffix}
9743 @itemx %m@var{suffix}
9744 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
9745 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
9746 all. These are the two most common ways to instruct a program that it
9747 should read from standard input or write to standard output. If you
9748 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
9749 construct: see for example @file{f/lang-specs.h}.
9751 @item %.@var{SUFFIX}
9752 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
9753 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
9754 terminated by the next space or %.
9757 Marks the argument containing or following the @samp{%w} as the
9758 designated output file of this compilation. This puts the argument
9759 into the sequence of arguments that @samp{%o} will substitute later.
9762 Substitutes the names of all the output files, with spaces
9763 automatically placed around them. You should write spaces
9764 around the @samp{%o} as well or the results are undefined.
9765 @samp{%o} is for use in the specs for running the linker.
9766 Input files whose names have no recognized suffix are not compiled
9767 at all, but they are included among the output files, so they will
9771 Substitutes the suffix for object files. Note that this is
9772 handled specially when it immediately follows @samp{%g, %u, or %U},
9773 because of the need for those to form complete file names. The
9774 handling is such that @samp{%O} is treated exactly as if it had already
9775 been substituted, except that @samp{%g, %u, and %U} do not currently
9776 support additional @var{suffix} characters following @samp{%O} as they would
9777 following, for example, @samp{.o}.
9780 Substitutes the standard macro predefinitions for the
9781 current target machine. Use this when running @code{cpp}.
9784 Like @samp{%p}, but puts @samp{__} before and after the name of each
9785 predefined macro, except for macros that start with @samp{__} or with
9786 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
9790 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
9791 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
9792 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
9793 and @option{-imultilib} as necessary.
9796 Current argument is the name of a library or startup file of some sort.
9797 Search for that file in a standard list of directories and substitute
9798 the full name found. The current working directory is included in the
9799 list of directories scanned.
9802 Current argument is the name of a linker script. Search for that file
9803 in the current list of directories to scan for libraries. If the file
9804 is located insert a @option{--script} option into the command line
9805 followed by the full path name found. If the file is not found then
9806 generate an error message. Note: the current working directory is not
9810 Print @var{str} as an error message. @var{str} is terminated by a newline.
9811 Use this when inconsistent options are detected.
9814 Substitute the contents of spec string @var{name} at this point.
9816 @item %x@{@var{option}@}
9817 Accumulate an option for @samp{%X}.
9820 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
9824 Output the accumulated assembler options specified by @option{-Wa}.
9827 Output the accumulated preprocessor options specified by @option{-Wp}.
9830 Process the @code{asm} spec. This is used to compute the
9831 switches to be passed to the assembler.
9834 Process the @code{asm_final} spec. This is a spec string for
9835 passing switches to an assembler post-processor, if such a program is
9839 Process the @code{link} spec. This is the spec for computing the
9840 command line passed to the linker. Typically it will make use of the
9841 @samp{%L %G %S %D and %E} sequences.
9844 Dump out a @option{-L} option for each directory that GCC believes might
9845 contain startup files. If the target supports multilibs then the
9846 current multilib directory will be prepended to each of these paths.
9849 Process the @code{lib} spec. This is a spec string for deciding which
9850 libraries should be included on the command line to the linker.
9853 Process the @code{libgcc} spec. This is a spec string for deciding
9854 which GCC support library should be included on the command line to the linker.
9857 Process the @code{startfile} spec. This is a spec for deciding which
9858 object files should be the first ones passed to the linker. Typically
9859 this might be a file named @file{crt0.o}.
9862 Process the @code{endfile} spec. This is a spec string that specifies
9863 the last object files that will be passed to the linker.
9866 Process the @code{cpp} spec. This is used to construct the arguments
9867 to be passed to the C preprocessor.
9870 Process the @code{cc1} spec. This is used to construct the options to be
9871 passed to the actual C compiler (@samp{cc1}).
9874 Process the @code{cc1plus} spec. This is used to construct the options to be
9875 passed to the actual C++ compiler (@samp{cc1plus}).
9878 Substitute the variable part of a matched option. See below.
9879 Note that each comma in the substituted string is replaced by
9883 Remove all occurrences of @code{-S} from the command line. Note---this
9884 command is position dependent. @samp{%} commands in the spec string
9885 before this one will see @code{-S}, @samp{%} commands in the spec string
9886 after this one will not.
9888 @item %:@var{function}(@var{args})
9889 Call the named function @var{function}, passing it @var{args}.
9890 @var{args} is first processed as a nested spec string, then split
9891 into an argument vector in the usual fashion. The function returns
9892 a string which is processed as if it had appeared literally as part
9893 of the current spec.
9895 The following built-in spec functions are provided:
9899 The @code{getenv} spec function takes two arguments: an environment
9900 variable name and a string. If the environment variable is not
9901 defined, a fatal error is issued. Otherwise, the return value is the
9902 value of the environment variable concatenated with the string. For
9903 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
9906 %:getenv(TOPDIR /include)
9909 expands to @file{/path/to/top/include}.
9911 @item @code{if-exists}
9912 The @code{if-exists} spec function takes one argument, an absolute
9913 pathname to a file. If the file exists, @code{if-exists} returns the
9914 pathname. Here is a small example of its usage:
9918 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
9921 @item @code{if-exists-else}
9922 The @code{if-exists-else} spec function is similar to the @code{if-exists}
9923 spec function, except that it takes two arguments. The first argument is
9924 an absolute pathname to a file. If the file exists, @code{if-exists-else}
9925 returns the pathname. If it does not exist, it returns the second argument.
9926 This way, @code{if-exists-else} can be used to select one file or another,
9927 based on the existence of the first. Here is a small example of its usage:
9931 crt0%O%s %:if-exists(crti%O%s) \
9932 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
9935 @item @code{replace-outfile}
9936 The @code{replace-outfile} spec function takes two arguments. It looks for the
9937 first argument in the outfiles array and replaces it with the second argument. Here
9938 is a small example of its usage:
9941 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
9944 @item @code{remove-outfile}
9945 The @code{remove-outfile} spec function takes one argument. It looks for the
9946 first argument in the outfiles array and removes it. Here is a small example
9950 %:remove-outfile(-lm)
9953 @item @code{pass-through-libs}
9954 The @code{pass-through-libs} spec function takes any number of arguments. It
9955 finds any @option{-l} options and any non-options ending in ".a" (which it
9956 assumes are the names of linker input library archive files) and returns a
9957 result containing all the found arguments each prepended by
9958 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
9959 intended to be passed to the LTO linker plugin.
9962 %:pass-through-libs(%G %L %G)
9965 @item @code{print-asm-header}
9966 The @code{print-asm-header} function takes no arguments and simply
9967 prints a banner like:
9973 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
9976 It is used to separate compiler options from assembler options
9977 in the @option{--target-help} output.
9981 Substitutes the @code{-S} switch, if that switch was given to GCC@.
9982 If that switch was not specified, this substitutes nothing. Note that
9983 the leading dash is omitted when specifying this option, and it is
9984 automatically inserted if the substitution is performed. Thus the spec
9985 string @samp{%@{foo@}} would match the command-line option @option{-foo}
9986 and would output the command line option @option{-foo}.
9988 @item %W@{@code{S}@}
9989 Like %@{@code{S}@} but mark last argument supplied within as a file to be
9992 @item %@{@code{S}*@}
9993 Substitutes all the switches specified to GCC whose names start
9994 with @code{-S}, but which also take an argument. This is used for
9995 switches like @option{-o}, @option{-D}, @option{-I}, etc.
9996 GCC considers @option{-o foo} as being
9997 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
9998 text, including the space. Thus two arguments would be generated.
10000 @item %@{@code{S}*&@code{T}*@}
10001 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
10002 (the order of @code{S} and @code{T} in the spec is not significant).
10003 There can be any number of ampersand-separated variables; for each the
10004 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
10006 @item %@{@code{S}:@code{X}@}
10007 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
10009 @item %@{!@code{S}:@code{X}@}
10010 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
10012 @item %@{@code{S}*:@code{X}@}
10013 Substitutes @code{X} if one or more switches whose names start with
10014 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
10015 once, no matter how many such switches appeared. However, if @code{%*}
10016 appears somewhere in @code{X}, then @code{X} will be substituted once
10017 for each matching switch, with the @code{%*} replaced by the part of
10018 that switch that matched the @code{*}.
10020 @item %@{.@code{S}:@code{X}@}
10021 Substitutes @code{X}, if processing a file with suffix @code{S}.
10023 @item %@{!.@code{S}:@code{X}@}
10024 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
10026 @item %@{,@code{S}:@code{X}@}
10027 Substitutes @code{X}, if processing a file for language @code{S}.
10029 @item %@{!,@code{S}:@code{X}@}
10030 Substitutes @code{X}, if not processing a file for language @code{S}.
10032 @item %@{@code{S}|@code{P}:@code{X}@}
10033 Substitutes @code{X} if either @code{-S} or @code{-P} was given to
10034 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
10035 @code{*} sequences as well, although they have a stronger binding than
10036 the @samp{|}. If @code{%*} appears in @code{X}, all of the
10037 alternatives must be starred, and only the first matching alternative
10040 For example, a spec string like this:
10043 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
10046 will output the following command-line options from the following input
10047 command-line options:
10052 -d fred.c -foo -baz -boggle
10053 -d jim.d -bar -baz -boggle
10056 @item %@{S:X; T:Y; :D@}
10058 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
10059 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
10060 be as many clauses as you need. This may be combined with @code{.},
10061 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
10066 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
10067 construct may contain other nested @samp{%} constructs or spaces, or
10068 even newlines. They are processed as usual, as described above.
10069 Trailing white space in @code{X} is ignored. White space may also
10070 appear anywhere on the left side of the colon in these constructs,
10071 except between @code{.} or @code{*} and the corresponding word.
10073 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
10074 handled specifically in these constructs. If another value of
10075 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
10076 @option{-W} switch is found later in the command line, the earlier
10077 switch value is ignored, except with @{@code{S}*@} where @code{S} is
10078 just one letter, which passes all matching options.
10080 The character @samp{|} at the beginning of the predicate text is used to
10081 indicate that a command should be piped to the following command, but
10082 only if @option{-pipe} is specified.
10084 It is built into GCC which switches take arguments and which do not.
10085 (You might think it would be useful to generalize this to allow each
10086 compiler's spec to say which switches take arguments. But this cannot
10087 be done in a consistent fashion. GCC cannot even decide which input
10088 files have been specified without knowing which switches take arguments,
10089 and it must know which input files to compile in order to tell which
10092 GCC also knows implicitly that arguments starting in @option{-l} are to be
10093 treated as compiler output files, and passed to the linker in their
10094 proper position among the other output files.
10096 @c man begin OPTIONS
10098 @node Target Options
10099 @section Specifying Target Machine and Compiler Version
10100 @cindex target options
10101 @cindex cross compiling
10102 @cindex specifying machine version
10103 @cindex specifying compiler version and target machine
10104 @cindex compiler version, specifying
10105 @cindex target machine, specifying
10107 The usual way to run GCC is to run the executable called @command{gcc}, or
10108 @command{@var{machine}-gcc} when cross-compiling, or
10109 @command{@var{machine}-gcc-@var{version}} to run a version other than the
10110 one that was installed last.
10112 @node Submodel Options
10113 @section Hardware Models and Configurations
10114 @cindex submodel options
10115 @cindex specifying hardware config
10116 @cindex hardware models and configurations, specifying
10117 @cindex machine dependent options
10119 Each target machine types can have its own
10120 special options, starting with @samp{-m}, to choose among various
10121 hardware models or configurations---for example, 68010 vs 68020,
10122 floating coprocessor or none. A single installed version of the
10123 compiler can compile for any model or configuration, according to the
10126 Some configurations of the compiler also support additional special
10127 options, usually for compatibility with other compilers on the same
10130 @c This list is ordered alphanumerically by subsection name.
10131 @c It should be the same order and spelling as these options are listed
10132 @c in Machine Dependent Options
10137 * Blackfin Options::
10141 * DEC Alpha Options::
10142 * DEC Alpha/VMS Options::
10145 * GNU/Linux Options::
10148 * i386 and x86-64 Options::
10149 * i386 and x86-64 Windows Options::
10151 * IA-64/VMS Options::
10158 * MicroBlaze Options::
10161 * MN10300 Options::
10163 * picoChip Options::
10164 * PowerPC Options::
10165 * RS/6000 and PowerPC Options::
10167 * S/390 and zSeries Options::
10170 * Solaris 2 Options::
10173 * System V Options::
10176 * VxWorks Options::
10178 * Xstormy16 Options::
10180 * zSeries Options::
10184 @subsection ARM Options
10185 @cindex ARM options
10187 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
10191 @item -mabi=@var{name}
10193 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
10194 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
10197 @opindex mapcs-frame
10198 Generate a stack frame that is compliant with the ARM Procedure Call
10199 Standard for all functions, even if this is not strictly necessary for
10200 correct execution of the code. Specifying @option{-fomit-frame-pointer}
10201 with this option will cause the stack frames not to be generated for
10202 leaf functions. The default is @option{-mno-apcs-frame}.
10206 This is a synonym for @option{-mapcs-frame}.
10209 @c not currently implemented
10210 @item -mapcs-stack-check
10211 @opindex mapcs-stack-check
10212 Generate code to check the amount of stack space available upon entry to
10213 every function (that actually uses some stack space). If there is
10214 insufficient space available then either the function
10215 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
10216 called, depending upon the amount of stack space required. The run time
10217 system is required to provide these functions. The default is
10218 @option{-mno-apcs-stack-check}, since this produces smaller code.
10220 @c not currently implemented
10222 @opindex mapcs-float
10223 Pass floating point arguments using the float point registers. This is
10224 one of the variants of the APCS@. This option is recommended if the
10225 target hardware has a floating point unit or if a lot of floating point
10226 arithmetic is going to be performed by the code. The default is
10227 @option{-mno-apcs-float}, since integer only code is slightly increased in
10228 size if @option{-mapcs-float} is used.
10230 @c not currently implemented
10231 @item -mapcs-reentrant
10232 @opindex mapcs-reentrant
10233 Generate reentrant, position independent code. The default is
10234 @option{-mno-apcs-reentrant}.
10237 @item -mthumb-interwork
10238 @opindex mthumb-interwork
10239 Generate code which supports calling between the ARM and Thumb
10240 instruction sets. Without this option, on pre-v5 architectures, the
10241 two instruction sets cannot be reliably used inside one program. The
10242 default is @option{-mno-thumb-interwork}, since slightly larger code
10243 is generated when @option{-mthumb-interwork} is specified. In AAPCS
10244 configurations this option is meaningless.
10246 @item -mno-sched-prolog
10247 @opindex mno-sched-prolog
10248 Prevent the reordering of instructions in the function prolog, or the
10249 merging of those instruction with the instructions in the function's
10250 body. This means that all functions will start with a recognizable set
10251 of instructions (or in fact one of a choice from a small set of
10252 different function prologues), and this information can be used to
10253 locate the start if functions inside an executable piece of code. The
10254 default is @option{-msched-prolog}.
10256 @item -mfloat-abi=@var{name}
10257 @opindex mfloat-abi
10258 Specifies which floating-point ABI to use. Permissible values
10259 are: @samp{soft}, @samp{softfp} and @samp{hard}.
10261 Specifying @samp{soft} causes GCC to generate output containing
10262 library calls for floating-point operations.
10263 @samp{softfp} allows the generation of code using hardware floating-point
10264 instructions, but still uses the soft-float calling conventions.
10265 @samp{hard} allows generation of floating-point instructions
10266 and uses FPU-specific calling conventions.
10268 The default depends on the specific target configuration. Note that
10269 the hard-float and soft-float ABIs are not link-compatible; you must
10270 compile your entire program with the same ABI, and link with a
10271 compatible set of libraries.
10273 @item -mlittle-endian
10274 @opindex mlittle-endian
10275 Generate code for a processor running in little-endian mode. This is
10276 the default for all standard configurations.
10279 @opindex mbig-endian
10280 Generate code for a processor running in big-endian mode; the default is
10281 to compile code for a little-endian processor.
10283 @item -mwords-little-endian
10284 @opindex mwords-little-endian
10285 This option only applies when generating code for big-endian processors.
10286 Generate code for a little-endian word order but a big-endian byte
10287 order. That is, a byte order of the form @samp{32107654}. Note: this
10288 option should only be used if you require compatibility with code for
10289 big-endian ARM processors generated by versions of the compiler prior to
10290 2.8. This option is now deprecated.
10292 @item -mcpu=@var{name}
10294 This specifies the name of the target ARM processor. GCC uses this name
10295 to determine what kind of instructions it can emit when generating
10296 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
10297 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
10298 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
10299 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
10300 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
10302 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
10303 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
10304 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
10305 @samp{strongarm1110},
10306 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
10307 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
10308 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
10309 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
10310 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
10311 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
10312 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
10313 @samp{cortex-a5}, @samp{cortex-a8}, @samp{cortex-a9}, @samp{cortex-a15},
10314 @samp{cortex-r4}, @samp{cortex-r4f}, @samp{cortex-r5},
10315 @samp{cortex-m4}, @samp{cortex-m3},
10318 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
10319 @samp{fa526}, @samp{fa626},
10320 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te}.
10322 @item -mtune=@var{name}
10324 This option is very similar to the @option{-mcpu=} option, except that
10325 instead of specifying the actual target processor type, and hence
10326 restricting which instructions can be used, it specifies that GCC should
10327 tune the performance of the code as if the target were of the type
10328 specified in this option, but still choosing the instructions that it
10329 will generate based on the CPU specified by a @option{-mcpu=} option.
10330 For some ARM implementations better performance can be obtained by using
10333 @item -march=@var{name}
10335 This specifies the name of the target ARM architecture. GCC uses this
10336 name to determine what kind of instructions it can emit when generating
10337 assembly code. This option can be used in conjunction with or instead
10338 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
10339 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
10340 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
10341 @samp{armv6}, @samp{armv6j},
10342 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
10343 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m},
10344 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
10346 @item -mfpu=@var{name}
10347 @itemx -mfpe=@var{number}
10348 @itemx -mfp=@var{number}
10352 This specifies what floating point hardware (or hardware emulation) is
10353 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
10354 @samp{fpe3}, @samp{maverick}, @samp{vfp}, @samp{vfpv3}, @samp{vfpv3-fp16},
10355 @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd}, @samp{vfpv3xd-fp16},
10356 @samp{neon}, @samp{neon-fp16}, @samp{vfpv4}, @samp{vfpv4-d16},
10357 @samp{fpv4-sp-d16} and @samp{neon-vfpv4}.
10358 @option{-mfp} and @option{-mfpe} are synonyms for
10359 @option{-mfpu}=@samp{fpe}@var{number}, for compatibility with older versions
10362 If @option{-msoft-float} is specified this specifies the format of
10363 floating point values.
10365 If the selected floating-point hardware includes the NEON extension
10366 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
10367 operations will not be used by GCC's auto-vectorization pass unless
10368 @option{-funsafe-math-optimizations} is also specified. This is
10369 because NEON hardware does not fully implement the IEEE 754 standard for
10370 floating-point arithmetic (in particular denormal values are treated as
10371 zero), so the use of NEON instructions may lead to a loss of precision.
10373 @item -mfp16-format=@var{name}
10374 @opindex mfp16-format
10375 Specify the format of the @code{__fp16} half-precision floating-point type.
10376 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
10377 the default is @samp{none}, in which case the @code{__fp16} type is not
10378 defined. @xref{Half-Precision}, for more information.
10380 @item -mstructure-size-boundary=@var{n}
10381 @opindex mstructure-size-boundary
10382 The size of all structures and unions will be rounded up to a multiple
10383 of the number of bits set by this option. Permissible values are 8, 32
10384 and 64. The default value varies for different toolchains. For the COFF
10385 targeted toolchain the default value is 8. A value of 64 is only allowed
10386 if the underlying ABI supports it.
10388 Specifying the larger number can produce faster, more efficient code, but
10389 can also increase the size of the program. Different values are potentially
10390 incompatible. Code compiled with one value cannot necessarily expect to
10391 work with code or libraries compiled with another value, if they exchange
10392 information using structures or unions.
10394 @item -mabort-on-noreturn
10395 @opindex mabort-on-noreturn
10396 Generate a call to the function @code{abort} at the end of a
10397 @code{noreturn} function. It will be executed if the function tries to
10401 @itemx -mno-long-calls
10402 @opindex mlong-calls
10403 @opindex mno-long-calls
10404 Tells the compiler to perform function calls by first loading the
10405 address of the function into a register and then performing a subroutine
10406 call on this register. This switch is needed if the target function
10407 will lie outside of the 64 megabyte addressing range of the offset based
10408 version of subroutine call instruction.
10410 Even if this switch is enabled, not all function calls will be turned
10411 into long calls. The heuristic is that static functions, functions
10412 which have the @samp{short-call} attribute, functions that are inside
10413 the scope of a @samp{#pragma no_long_calls} directive and functions whose
10414 definitions have already been compiled within the current compilation
10415 unit, will not be turned into long calls. The exception to this rule is
10416 that weak function definitions, functions with the @samp{long-call}
10417 attribute or the @samp{section} attribute, and functions that are within
10418 the scope of a @samp{#pragma long_calls} directive, will always be
10419 turned into long calls.
10421 This feature is not enabled by default. Specifying
10422 @option{-mno-long-calls} will restore the default behavior, as will
10423 placing the function calls within the scope of a @samp{#pragma
10424 long_calls_off} directive. Note these switches have no effect on how
10425 the compiler generates code to handle function calls via function
10428 @item -msingle-pic-base
10429 @opindex msingle-pic-base
10430 Treat the register used for PIC addressing as read-only, rather than
10431 loading it in the prologue for each function. The run-time system is
10432 responsible for initializing this register with an appropriate value
10433 before execution begins.
10435 @item -mpic-register=@var{reg}
10436 @opindex mpic-register
10437 Specify the register to be used for PIC addressing. The default is R10
10438 unless stack-checking is enabled, when R9 is used.
10440 @item -mcirrus-fix-invalid-insns
10441 @opindex mcirrus-fix-invalid-insns
10442 @opindex mno-cirrus-fix-invalid-insns
10443 Insert NOPs into the instruction stream to in order to work around
10444 problems with invalid Maverick instruction combinations. This option
10445 is only valid if the @option{-mcpu=ep9312} option has been used to
10446 enable generation of instructions for the Cirrus Maverick floating
10447 point co-processor. This option is not enabled by default, since the
10448 problem is only present in older Maverick implementations. The default
10449 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
10452 @item -mpoke-function-name
10453 @opindex mpoke-function-name
10454 Write the name of each function into the text section, directly
10455 preceding the function prologue. The generated code is similar to this:
10459 .ascii "arm_poke_function_name", 0
10462 .word 0xff000000 + (t1 - t0)
10463 arm_poke_function_name
10465 stmfd sp!, @{fp, ip, lr, pc@}
10469 When performing a stack backtrace, code can inspect the value of
10470 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
10471 location @code{pc - 12} and the top 8 bits are set, then we know that
10472 there is a function name embedded immediately preceding this location
10473 and has length @code{((pc[-3]) & 0xff000000)}.
10480 Select between generating code that executes in ARM and Thumb
10481 states. The default for most configurations is to generate code
10482 that executes in ARM state, but the default can be changed by
10483 configuring GCC with the @option{--with-mode=}@var{state}
10487 @opindex mtpcs-frame
10488 Generate a stack frame that is compliant with the Thumb Procedure Call
10489 Standard for all non-leaf functions. (A leaf function is one that does
10490 not call any other functions.) The default is @option{-mno-tpcs-frame}.
10492 @item -mtpcs-leaf-frame
10493 @opindex mtpcs-leaf-frame
10494 Generate a stack frame that is compliant with the Thumb Procedure Call
10495 Standard for all leaf functions. (A leaf function is one that does
10496 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
10498 @item -mcallee-super-interworking
10499 @opindex mcallee-super-interworking
10500 Gives all externally visible functions in the file being compiled an ARM
10501 instruction set header which switches to Thumb mode before executing the
10502 rest of the function. This allows these functions to be called from
10503 non-interworking code. This option is not valid in AAPCS configurations
10504 because interworking is enabled by default.
10506 @item -mcaller-super-interworking
10507 @opindex mcaller-super-interworking
10508 Allows calls via function pointers (including virtual functions) to
10509 execute correctly regardless of whether the target code has been
10510 compiled for interworking or not. There is a small overhead in the cost
10511 of executing a function pointer if this option is enabled. This option
10512 is not valid in AAPCS configurations because interworking is enabled
10515 @item -mtp=@var{name}
10517 Specify the access model for the thread local storage pointer. The valid
10518 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
10519 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
10520 (supported in the arm6k architecture), and @option{auto}, which uses the
10521 best available method for the selected processor. The default setting is
10524 @item -mtls-dialect=@var{dialect}
10525 @opindex mtls-dialect
10526 Specify the dialect to use for accessing thread local storage. Two
10527 dialects are supported --- @option{gnu} and @option{gnu2}. The
10528 @option{gnu} dialect selects the original GNU scheme for supporting
10529 local and global dynamic TLS models. The @option{gnu2} dialect
10530 selects the GNU descriptor scheme, which provides better performance
10531 for shared libraries. The GNU descriptor scheme is compatible with
10532 the original scheme, but does require new assembler, linker and
10533 library support. Initial and local exec TLS models are unaffected by
10534 this option and always use the original scheme.
10536 @item -mword-relocations
10537 @opindex mword-relocations
10538 Only generate absolute relocations on word sized values (i.e. R_ARM_ABS32).
10539 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
10540 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
10543 @item -mfix-cortex-m3-ldrd
10544 @opindex mfix-cortex-m3-ldrd
10545 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
10546 with overlapping destination and base registers are used. This option avoids
10547 generating these instructions. This option is enabled by default when
10548 @option{-mcpu=cortex-m3} is specified.
10553 @subsection AVR Options
10554 @cindex AVR Options
10556 These options are defined for AVR implementations:
10559 @item -mmcu=@var{mcu}
10561 Specify ATMEL AVR instruction set or MCU type.
10563 Instruction set avr1 is for the minimal AVR core, not supported by the C
10564 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
10565 attiny11, attiny12, attiny15, attiny28).
10567 Instruction set avr2 (default) is for the classic AVR core with up to
10568 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
10569 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
10570 at90c8534, at90s8535).
10572 Instruction set avr3 is for the classic AVR core with up to 128K program
10573 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
10575 Instruction set avr4 is for the enhanced AVR core with up to 8K program
10576 memory space (MCU types: atmega8, atmega83, atmega85).
10578 Instruction set avr5 is for the enhanced AVR core with up to 128K program
10579 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
10580 atmega64, atmega128, at43usb355, at94k).
10582 @item -mno-interrupts
10583 @opindex mno-interrupts
10584 Generated code is not compatible with hardware interrupts.
10585 Code size will be smaller.
10587 @item -mcall-prologues
10588 @opindex mcall-prologues
10589 Functions prologues/epilogues expanded as call to appropriate
10590 subroutines. Code size will be smaller.
10593 @opindex mtiny-stack
10594 Change only the low 8 bits of the stack pointer.
10598 Assume int to be 8 bit integer. This affects the sizes of all types: A
10599 char will be 1 byte, an int will be 1 byte, a long will be 2 bytes
10600 and long long will be 4 bytes. Please note that this option does not
10601 comply to the C standards, but it will provide you with smaller code
10605 @node Blackfin Options
10606 @subsection Blackfin Options
10607 @cindex Blackfin Options
10610 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
10612 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
10613 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
10614 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
10615 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
10616 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
10617 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
10618 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
10619 @samp{bf561}, @samp{bf592}.
10620 The optional @var{sirevision} specifies the silicon revision of the target
10621 Blackfin processor. Any workarounds available for the targeted silicon revision
10622 will be enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
10623 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
10624 will be enabled. The @code{__SILICON_REVISION__} macro is defined to two
10625 hexadecimal digits representing the major and minor numbers in the silicon
10626 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
10627 is not defined. If @var{sirevision} is @samp{any}, the
10628 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
10629 If this optional @var{sirevision} is not used, GCC assumes the latest known
10630 silicon revision of the targeted Blackfin processor.
10632 Support for @samp{bf561} is incomplete. For @samp{bf561},
10633 Only the processor macro is defined.
10634 Without this option, @samp{bf532} is used as the processor by default.
10635 The corresponding predefined processor macros for @var{cpu} is to
10636 be defined. And for @samp{bfin-elf} toolchain, this causes the hardware BSP
10637 provided by libgloss to be linked in if @option{-msim} is not given.
10641 Specifies that the program will be run on the simulator. This causes
10642 the simulator BSP provided by libgloss to be linked in. This option
10643 has effect only for @samp{bfin-elf} toolchain.
10644 Certain other options, such as @option{-mid-shared-library} and
10645 @option{-mfdpic}, imply @option{-msim}.
10647 @item -momit-leaf-frame-pointer
10648 @opindex momit-leaf-frame-pointer
10649 Don't keep the frame pointer in a register for leaf functions. This
10650 avoids the instructions to save, set up and restore frame pointers and
10651 makes an extra register available in leaf functions. The option
10652 @option{-fomit-frame-pointer} removes the frame pointer for all functions
10653 which might make debugging harder.
10655 @item -mspecld-anomaly
10656 @opindex mspecld-anomaly
10657 When enabled, the compiler will ensure that the generated code does not
10658 contain speculative loads after jump instructions. If this option is used,
10659 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
10661 @item -mno-specld-anomaly
10662 @opindex mno-specld-anomaly
10663 Don't generate extra code to prevent speculative loads from occurring.
10665 @item -mcsync-anomaly
10666 @opindex mcsync-anomaly
10667 When enabled, the compiler will ensure that the generated code does not
10668 contain CSYNC or SSYNC instructions too soon after conditional branches.
10669 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
10671 @item -mno-csync-anomaly
10672 @opindex mno-csync-anomaly
10673 Don't generate extra code to prevent CSYNC or SSYNC instructions from
10674 occurring too soon after a conditional branch.
10678 When enabled, the compiler is free to take advantage of the knowledge that
10679 the entire program fits into the low 64k of memory.
10682 @opindex mno-low-64k
10683 Assume that the program is arbitrarily large. This is the default.
10685 @item -mstack-check-l1
10686 @opindex mstack-check-l1
10687 Do stack checking using information placed into L1 scratchpad memory by the
10690 @item -mid-shared-library
10691 @opindex mid-shared-library
10692 Generate code that supports shared libraries via the library ID method.
10693 This allows for execute in place and shared libraries in an environment
10694 without virtual memory management. This option implies @option{-fPIC}.
10695 With a @samp{bfin-elf} target, this option implies @option{-msim}.
10697 @item -mno-id-shared-library
10698 @opindex mno-id-shared-library
10699 Generate code that doesn't assume ID based shared libraries are being used.
10700 This is the default.
10702 @item -mleaf-id-shared-library
10703 @opindex mleaf-id-shared-library
10704 Generate code that supports shared libraries via the library ID method,
10705 but assumes that this library or executable won't link against any other
10706 ID shared libraries. That allows the compiler to use faster code for jumps
10709 @item -mno-leaf-id-shared-library
10710 @opindex mno-leaf-id-shared-library
10711 Do not assume that the code being compiled won't link against any ID shared
10712 libraries. Slower code will be generated for jump and call insns.
10714 @item -mshared-library-id=n
10715 @opindex mshared-library-id
10716 Specified the identification number of the ID based shared library being
10717 compiled. Specifying a value of 0 will generate more compact code, specifying
10718 other values will force the allocation of that number to the current
10719 library but is no more space or time efficient than omitting this option.
10723 Generate code that allows the data segment to be located in a different
10724 area of memory from the text segment. This allows for execute in place in
10725 an environment without virtual memory management by eliminating relocations
10726 against the text section.
10728 @item -mno-sep-data
10729 @opindex mno-sep-data
10730 Generate code that assumes that the data segment follows the text segment.
10731 This is the default.
10734 @itemx -mno-long-calls
10735 @opindex mlong-calls
10736 @opindex mno-long-calls
10737 Tells the compiler to perform function calls by first loading the
10738 address of the function into a register and then performing a subroutine
10739 call on this register. This switch is needed if the target function
10740 will lie outside of the 24 bit addressing range of the offset based
10741 version of subroutine call instruction.
10743 This feature is not enabled by default. Specifying
10744 @option{-mno-long-calls} will restore the default behavior. Note these
10745 switches have no effect on how the compiler generates code to handle
10746 function calls via function pointers.
10750 Link with the fast floating-point library. This library relaxes some of
10751 the IEEE floating-point standard's rules for checking inputs against
10752 Not-a-Number (NAN), in the interest of performance.
10755 @opindex minline-plt
10756 Enable inlining of PLT entries in function calls to functions that are
10757 not known to bind locally. It has no effect without @option{-mfdpic}.
10760 @opindex mmulticore
10761 Build standalone application for multicore Blackfin processor. Proper
10762 start files and link scripts will be used to support multicore.
10763 This option defines @code{__BFIN_MULTICORE}. It can only be used with
10764 @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}. It can be used with
10765 @option{-mcorea} or @option{-mcoreb}. If it's used without
10766 @option{-mcorea} or @option{-mcoreb}, single application/dual core
10767 programming model is used. In this model, the main function of Core B
10768 should be named as coreb_main. If it's used with @option{-mcorea} or
10769 @option{-mcoreb}, one application per core programming model is used.
10770 If this option is not used, single core application programming
10775 Build standalone application for Core A of BF561 when using
10776 one application per core programming model. Proper start files
10777 and link scripts will be used to support Core A. This option
10778 defines @code{__BFIN_COREA}. It must be used with @option{-mmulticore}.
10782 Build standalone application for Core B of BF561 when using
10783 one application per core programming model. Proper start files
10784 and link scripts will be used to support Core B. This option
10785 defines @code{__BFIN_COREB}. When this option is used, coreb_main
10786 should be used instead of main. It must be used with
10787 @option{-mmulticore}.
10791 Build standalone application for SDRAM. Proper start files and
10792 link scripts will be used to put the application into SDRAM.
10793 Loader should initialize SDRAM before loading the application
10794 into SDRAM. This option defines @code{__BFIN_SDRAM}.
10798 Assume that ICPLBs are enabled at runtime. This has an effect on certain
10799 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
10800 are enabled; for standalone applications the default is off.
10804 @subsection C6X Options
10805 @cindex C6X Options
10808 @item -march=@var{name}
10810 This specifies the name of the target architecture. GCC uses this
10811 name to determine what kind of instructions it can emit when generating
10812 assembly code. Permissible names are: @samp{c62x},
10813 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
10816 @opindex mbig-endian
10817 Generate code for a big endian target.
10819 @item -mlittle-endian
10820 @opindex mlittle-endian
10821 Generate code for a little endian target. This is the default.
10825 Choose startup files and linker script suitable for the simulator.
10827 @item -msdata=default
10828 @opindex msdata=default
10829 Put small global and static data in the @samp{.neardata} section,
10830 which is pointed to by register @code{B14}. Put small uninitialized
10831 global and static data in the @samp{.bss} section, which is adjacent
10832 to the @samp{.neardata} section. Put small read-only data into the
10833 @samp{.rodata} section. The corresponding sections used for large
10834 pieces of data are @samp{.fardata}, @samp{.far} and @samp{.const}.
10837 @opindex msdata=all
10838 Put all data, not just small objets, into the sections reserved for
10839 small data, and use addressing relative to the @code{B14} register to
10843 @opindex msdata=none
10844 Make no use of the sections reserved for small data, and use absolute
10845 addresses to access all data. Put all initialized global and static
10846 data in the @samp{.fardata} section, and all uninitialized data in the
10847 @samp{.far} section. Put all constant data into the @samp{.const}
10852 @subsection CRIS Options
10853 @cindex CRIS Options
10855 These options are defined specifically for the CRIS ports.
10858 @item -march=@var{architecture-type}
10859 @itemx -mcpu=@var{architecture-type}
10862 Generate code for the specified architecture. The choices for
10863 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
10864 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
10865 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
10868 @item -mtune=@var{architecture-type}
10870 Tune to @var{architecture-type} everything applicable about the generated
10871 code, except for the ABI and the set of available instructions. The
10872 choices for @var{architecture-type} are the same as for
10873 @option{-march=@var{architecture-type}}.
10875 @item -mmax-stack-frame=@var{n}
10876 @opindex mmax-stack-frame
10877 Warn when the stack frame of a function exceeds @var{n} bytes.
10883 The options @option{-metrax4} and @option{-metrax100} are synonyms for
10884 @option{-march=v3} and @option{-march=v8} respectively.
10886 @item -mmul-bug-workaround
10887 @itemx -mno-mul-bug-workaround
10888 @opindex mmul-bug-workaround
10889 @opindex mno-mul-bug-workaround
10890 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
10891 models where it applies. This option is active by default.
10895 Enable CRIS-specific verbose debug-related information in the assembly
10896 code. This option also has the effect to turn off the @samp{#NO_APP}
10897 formatted-code indicator to the assembler at the beginning of the
10902 Do not use condition-code results from previous instruction; always emit
10903 compare and test instructions before use of condition codes.
10905 @item -mno-side-effects
10906 @opindex mno-side-effects
10907 Do not emit instructions with side-effects in addressing modes other than
10910 @item -mstack-align
10911 @itemx -mno-stack-align
10912 @itemx -mdata-align
10913 @itemx -mno-data-align
10914 @itemx -mconst-align
10915 @itemx -mno-const-align
10916 @opindex mstack-align
10917 @opindex mno-stack-align
10918 @opindex mdata-align
10919 @opindex mno-data-align
10920 @opindex mconst-align
10921 @opindex mno-const-align
10922 These options (no-options) arranges (eliminate arrangements) for the
10923 stack-frame, individual data and constants to be aligned for the maximum
10924 single data access size for the chosen CPU model. The default is to
10925 arrange for 32-bit alignment. ABI details such as structure layout are
10926 not affected by these options.
10934 Similar to the stack- data- and const-align options above, these options
10935 arrange for stack-frame, writable data and constants to all be 32-bit,
10936 16-bit or 8-bit aligned. The default is 32-bit alignment.
10938 @item -mno-prologue-epilogue
10939 @itemx -mprologue-epilogue
10940 @opindex mno-prologue-epilogue
10941 @opindex mprologue-epilogue
10942 With @option{-mno-prologue-epilogue}, the normal function prologue and
10943 epilogue that sets up the stack-frame are omitted and no return
10944 instructions or return sequences are generated in the code. Use this
10945 option only together with visual inspection of the compiled code: no
10946 warnings or errors are generated when call-saved registers must be saved,
10947 or storage for local variable needs to be allocated.
10951 @opindex mno-gotplt
10953 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
10954 instruction sequences that load addresses for functions from the PLT part
10955 of the GOT rather than (traditional on other architectures) calls to the
10956 PLT@. The default is @option{-mgotplt}.
10960 Legacy no-op option only recognized with the cris-axis-elf and
10961 cris-axis-linux-gnu targets.
10965 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
10969 This option, recognized for the cris-axis-elf arranges
10970 to link with input-output functions from a simulator library. Code,
10971 initialized data and zero-initialized data are allocated consecutively.
10975 Like @option{-sim}, but pass linker options to locate initialized data at
10976 0x40000000 and zero-initialized data at 0x80000000.
10979 @node Darwin Options
10980 @subsection Darwin Options
10981 @cindex Darwin options
10983 These options are defined for all architectures running the Darwin operating
10986 FSF GCC on Darwin does not create ``fat'' object files; it will create
10987 an object file for the single architecture that it was built to
10988 target. Apple's GCC on Darwin does create ``fat'' files if multiple
10989 @option{-arch} options are used; it does so by running the compiler or
10990 linker multiple times and joining the results together with
10993 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
10994 @samp{i686}) is determined by the flags that specify the ISA
10995 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
10996 @option{-force_cpusubtype_ALL} option can be used to override this.
10998 The Darwin tools vary in their behavior when presented with an ISA
10999 mismatch. The assembler, @file{as}, will only permit instructions to
11000 be used that are valid for the subtype of the file it is generating,
11001 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
11002 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
11003 and print an error if asked to create a shared library with a less
11004 restrictive subtype than its input files (for instance, trying to put
11005 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
11006 for executables, @file{ld}, will quietly give the executable the most
11007 restrictive subtype of any of its input files.
11012 Add the framework directory @var{dir} to the head of the list of
11013 directories to be searched for header files. These directories are
11014 interleaved with those specified by @option{-I} options and are
11015 scanned in a left-to-right order.
11017 A framework directory is a directory with frameworks in it. A
11018 framework is a directory with a @samp{"Headers"} and/or
11019 @samp{"PrivateHeaders"} directory contained directly in it that ends
11020 in @samp{".framework"}. The name of a framework is the name of this
11021 directory excluding the @samp{".framework"}. Headers associated with
11022 the framework are found in one of those two directories, with
11023 @samp{"Headers"} being searched first. A subframework is a framework
11024 directory that is in a framework's @samp{"Frameworks"} directory.
11025 Includes of subframework headers can only appear in a header of a
11026 framework that contains the subframework, or in a sibling subframework
11027 header. Two subframeworks are siblings if they occur in the same
11028 framework. A subframework should not have the same name as a
11029 framework, a warning will be issued if this is violated. Currently a
11030 subframework cannot have subframeworks, in the future, the mechanism
11031 may be extended to support this. The standard frameworks can be found
11032 in @samp{"/System/Library/Frameworks"} and
11033 @samp{"/Library/Frameworks"}. An example include looks like
11034 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
11035 the name of the framework and header.h is found in the
11036 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
11038 @item -iframework@var{dir}
11039 @opindex iframework
11040 Like @option{-F} except the directory is a treated as a system
11041 directory. The main difference between this @option{-iframework} and
11042 @option{-F} is that with @option{-iframework} the compiler does not
11043 warn about constructs contained within header files found via
11044 @var{dir}. This option is valid only for the C family of languages.
11048 Emit debugging information for symbols that are used. For STABS
11049 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
11050 This is by default ON@.
11054 Emit debugging information for all symbols and types.
11056 @item -mmacosx-version-min=@var{version}
11057 The earliest version of MacOS X that this executable will run on
11058 is @var{version}. Typical values of @var{version} include @code{10.1},
11059 @code{10.2}, and @code{10.3.9}.
11061 If the compiler was built to use the system's headers by default,
11062 then the default for this option is the system version on which the
11063 compiler is running, otherwise the default is to make choices which
11064 are compatible with as many systems and code bases as possible.
11068 Enable kernel development mode. The @option{-mkernel} option sets
11069 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
11070 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
11071 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
11072 applicable. This mode also sets @option{-mno-altivec},
11073 @option{-msoft-float}, @option{-fno-builtin} and
11074 @option{-mlong-branch} for PowerPC targets.
11076 @item -mone-byte-bool
11077 @opindex mone-byte-bool
11078 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
11079 By default @samp{sizeof(bool)} is @samp{4} when compiling for
11080 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
11081 option has no effect on x86.
11083 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
11084 to generate code that is not binary compatible with code generated
11085 without that switch. Using this switch may require recompiling all
11086 other modules in a program, including system libraries. Use this
11087 switch to conform to a non-default data model.
11089 @item -mfix-and-continue
11090 @itemx -ffix-and-continue
11091 @itemx -findirect-data
11092 @opindex mfix-and-continue
11093 @opindex ffix-and-continue
11094 @opindex findirect-data
11095 Generate code suitable for fast turn around development. Needed to
11096 enable gdb to dynamically load @code{.o} files into already running
11097 programs. @option{-findirect-data} and @option{-ffix-and-continue}
11098 are provided for backwards compatibility.
11102 Loads all members of static archive libraries.
11103 See man ld(1) for more information.
11105 @item -arch_errors_fatal
11106 @opindex arch_errors_fatal
11107 Cause the errors having to do with files that have the wrong architecture
11110 @item -bind_at_load
11111 @opindex bind_at_load
11112 Causes the output file to be marked such that the dynamic linker will
11113 bind all undefined references when the file is loaded or launched.
11117 Produce a Mach-o bundle format file.
11118 See man ld(1) for more information.
11120 @item -bundle_loader @var{executable}
11121 @opindex bundle_loader
11122 This option specifies the @var{executable} that will be loading the build
11123 output file being linked. See man ld(1) for more information.
11126 @opindex dynamiclib
11127 When passed this option, GCC will produce a dynamic library instead of
11128 an executable when linking, using the Darwin @file{libtool} command.
11130 @item -force_cpusubtype_ALL
11131 @opindex force_cpusubtype_ALL
11132 This causes GCC's output file to have the @var{ALL} subtype, instead of
11133 one controlled by the @option{-mcpu} or @option{-march} option.
11135 @item -allowable_client @var{client_name}
11136 @itemx -client_name
11137 @itemx -compatibility_version
11138 @itemx -current_version
11140 @itemx -dependency-file
11142 @itemx -dylinker_install_name
11144 @itemx -exported_symbols_list
11147 @itemx -flat_namespace
11148 @itemx -force_flat_namespace
11149 @itemx -headerpad_max_install_names
11152 @itemx -install_name
11153 @itemx -keep_private_externs
11154 @itemx -multi_module
11155 @itemx -multiply_defined
11156 @itemx -multiply_defined_unused
11159 @itemx -no_dead_strip_inits_and_terms
11160 @itemx -nofixprebinding
11161 @itemx -nomultidefs
11163 @itemx -noseglinkedit
11164 @itemx -pagezero_size
11166 @itemx -prebind_all_twolevel_modules
11167 @itemx -private_bundle
11169 @itemx -read_only_relocs
11171 @itemx -sectobjectsymbols
11175 @itemx -sectobjectsymbols
11178 @itemx -segs_read_only_addr
11180 @itemx -segs_read_write_addr
11181 @itemx -seg_addr_table
11182 @itemx -seg_addr_table_filename
11183 @itemx -seglinkedit
11185 @itemx -segs_read_only_addr
11186 @itemx -segs_read_write_addr
11187 @itemx -single_module
11189 @itemx -sub_library
11191 @itemx -sub_umbrella
11192 @itemx -twolevel_namespace
11195 @itemx -unexported_symbols_list
11196 @itemx -weak_reference_mismatches
11197 @itemx -whatsloaded
11198 @opindex allowable_client
11199 @opindex client_name
11200 @opindex compatibility_version
11201 @opindex current_version
11202 @opindex dead_strip
11203 @opindex dependency-file
11204 @opindex dylib_file
11205 @opindex dylinker_install_name
11207 @opindex exported_symbols_list
11209 @opindex flat_namespace
11210 @opindex force_flat_namespace
11211 @opindex headerpad_max_install_names
11212 @opindex image_base
11214 @opindex install_name
11215 @opindex keep_private_externs
11216 @opindex multi_module
11217 @opindex multiply_defined
11218 @opindex multiply_defined_unused
11219 @opindex noall_load
11220 @opindex no_dead_strip_inits_and_terms
11221 @opindex nofixprebinding
11222 @opindex nomultidefs
11224 @opindex noseglinkedit
11225 @opindex pagezero_size
11227 @opindex prebind_all_twolevel_modules
11228 @opindex private_bundle
11229 @opindex read_only_relocs
11231 @opindex sectobjectsymbols
11234 @opindex sectcreate
11235 @opindex sectobjectsymbols
11238 @opindex segs_read_only_addr
11239 @opindex segs_read_write_addr
11240 @opindex seg_addr_table
11241 @opindex seg_addr_table_filename
11242 @opindex seglinkedit
11244 @opindex segs_read_only_addr
11245 @opindex segs_read_write_addr
11246 @opindex single_module
11248 @opindex sub_library
11249 @opindex sub_umbrella
11250 @opindex twolevel_namespace
11253 @opindex unexported_symbols_list
11254 @opindex weak_reference_mismatches
11255 @opindex whatsloaded
11256 These options are passed to the Darwin linker. The Darwin linker man page
11257 describes them in detail.
11260 @node DEC Alpha Options
11261 @subsection DEC Alpha Options
11263 These @samp{-m} options are defined for the DEC Alpha implementations:
11266 @item -mno-soft-float
11267 @itemx -msoft-float
11268 @opindex mno-soft-float
11269 @opindex msoft-float
11270 Use (do not use) the hardware floating-point instructions for
11271 floating-point operations. When @option{-msoft-float} is specified,
11272 functions in @file{libgcc.a} will be used to perform floating-point
11273 operations. Unless they are replaced by routines that emulate the
11274 floating-point operations, or compiled in such a way as to call such
11275 emulations routines, these routines will issue floating-point
11276 operations. If you are compiling for an Alpha without floating-point
11277 operations, you must ensure that the library is built so as not to call
11280 Note that Alpha implementations without floating-point operations are
11281 required to have floating-point registers.
11284 @itemx -mno-fp-regs
11286 @opindex mno-fp-regs
11287 Generate code that uses (does not use) the floating-point register set.
11288 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
11289 register set is not used, floating point operands are passed in integer
11290 registers as if they were integers and floating-point results are passed
11291 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
11292 so any function with a floating-point argument or return value called by code
11293 compiled with @option{-mno-fp-regs} must also be compiled with that
11296 A typical use of this option is building a kernel that does not use,
11297 and hence need not save and restore, any floating-point registers.
11301 The Alpha architecture implements floating-point hardware optimized for
11302 maximum performance. It is mostly compliant with the IEEE floating
11303 point standard. However, for full compliance, software assistance is
11304 required. This option generates code fully IEEE compliant code
11305 @emph{except} that the @var{inexact-flag} is not maintained (see below).
11306 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
11307 defined during compilation. The resulting code is less efficient but is
11308 able to correctly support denormalized numbers and exceptional IEEE
11309 values such as not-a-number and plus/minus infinity. Other Alpha
11310 compilers call this option @option{-ieee_with_no_inexact}.
11312 @item -mieee-with-inexact
11313 @opindex mieee-with-inexact
11314 This is like @option{-mieee} except the generated code also maintains
11315 the IEEE @var{inexact-flag}. Turning on this option causes the
11316 generated code to implement fully-compliant IEEE math. In addition to
11317 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
11318 macro. On some Alpha implementations the resulting code may execute
11319 significantly slower than the code generated by default. Since there is
11320 very little code that depends on the @var{inexact-flag}, you should
11321 normally not specify this option. Other Alpha compilers call this
11322 option @option{-ieee_with_inexact}.
11324 @item -mfp-trap-mode=@var{trap-mode}
11325 @opindex mfp-trap-mode
11326 This option controls what floating-point related traps are enabled.
11327 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
11328 The trap mode can be set to one of four values:
11332 This is the default (normal) setting. The only traps that are enabled
11333 are the ones that cannot be disabled in software (e.g., division by zero
11337 In addition to the traps enabled by @samp{n}, underflow traps are enabled
11341 Like @samp{u}, but the instructions are marked to be safe for software
11342 completion (see Alpha architecture manual for details).
11345 Like @samp{su}, but inexact traps are enabled as well.
11348 @item -mfp-rounding-mode=@var{rounding-mode}
11349 @opindex mfp-rounding-mode
11350 Selects the IEEE rounding mode. Other Alpha compilers call this option
11351 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
11356 Normal IEEE rounding mode. Floating point numbers are rounded towards
11357 the nearest machine number or towards the even machine number in case
11361 Round towards minus infinity.
11364 Chopped rounding mode. Floating point numbers are rounded towards zero.
11367 Dynamic rounding mode. A field in the floating point control register
11368 (@var{fpcr}, see Alpha architecture reference manual) controls the
11369 rounding mode in effect. The C library initializes this register for
11370 rounding towards plus infinity. Thus, unless your program modifies the
11371 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
11374 @item -mtrap-precision=@var{trap-precision}
11375 @opindex mtrap-precision
11376 In the Alpha architecture, floating point traps are imprecise. This
11377 means without software assistance it is impossible to recover from a
11378 floating trap and program execution normally needs to be terminated.
11379 GCC can generate code that can assist operating system trap handlers
11380 in determining the exact location that caused a floating point trap.
11381 Depending on the requirements of an application, different levels of
11382 precisions can be selected:
11386 Program precision. This option is the default and means a trap handler
11387 can only identify which program caused a floating point exception.
11390 Function precision. The trap handler can determine the function that
11391 caused a floating point exception.
11394 Instruction precision. The trap handler can determine the exact
11395 instruction that caused a floating point exception.
11398 Other Alpha compilers provide the equivalent options called
11399 @option{-scope_safe} and @option{-resumption_safe}.
11401 @item -mieee-conformant
11402 @opindex mieee-conformant
11403 This option marks the generated code as IEEE conformant. You must not
11404 use this option unless you also specify @option{-mtrap-precision=i} and either
11405 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
11406 is to emit the line @samp{.eflag 48} in the function prologue of the
11407 generated assembly file. Under DEC Unix, this has the effect that
11408 IEEE-conformant math library routines will be linked in.
11410 @item -mbuild-constants
11411 @opindex mbuild-constants
11412 Normally GCC examines a 32- or 64-bit integer constant to
11413 see if it can construct it from smaller constants in two or three
11414 instructions. If it cannot, it will output the constant as a literal and
11415 generate code to load it from the data segment at runtime.
11417 Use this option to require GCC to construct @emph{all} integer constants
11418 using code, even if it takes more instructions (the maximum is six).
11420 You would typically use this option to build a shared library dynamic
11421 loader. Itself a shared library, it must relocate itself in memory
11422 before it can find the variables and constants in its own data segment.
11428 Select whether to generate code to be assembled by the vendor-supplied
11429 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
11447 Indicate whether GCC should generate code to use the optional BWX,
11448 CIX, FIX and MAX instruction sets. The default is to use the instruction
11449 sets supported by the CPU type specified via @option{-mcpu=} option or that
11450 of the CPU on which GCC was built if none was specified.
11453 @itemx -mfloat-ieee
11454 @opindex mfloat-vax
11455 @opindex mfloat-ieee
11456 Generate code that uses (does not use) VAX F and G floating point
11457 arithmetic instead of IEEE single and double precision.
11459 @item -mexplicit-relocs
11460 @itemx -mno-explicit-relocs
11461 @opindex mexplicit-relocs
11462 @opindex mno-explicit-relocs
11463 Older Alpha assemblers provided no way to generate symbol relocations
11464 except via assembler macros. Use of these macros does not allow
11465 optimal instruction scheduling. GNU binutils as of version 2.12
11466 supports a new syntax that allows the compiler to explicitly mark
11467 which relocations should apply to which instructions. This option
11468 is mostly useful for debugging, as GCC detects the capabilities of
11469 the assembler when it is built and sets the default accordingly.
11472 @itemx -mlarge-data
11473 @opindex msmall-data
11474 @opindex mlarge-data
11475 When @option{-mexplicit-relocs} is in effect, static data is
11476 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
11477 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
11478 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
11479 16-bit relocations off of the @code{$gp} register. This limits the
11480 size of the small data area to 64KB, but allows the variables to be
11481 directly accessed via a single instruction.
11483 The default is @option{-mlarge-data}. With this option the data area
11484 is limited to just below 2GB@. Programs that require more than 2GB of
11485 data must use @code{malloc} or @code{mmap} to allocate the data in the
11486 heap instead of in the program's data segment.
11488 When generating code for shared libraries, @option{-fpic} implies
11489 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
11492 @itemx -mlarge-text
11493 @opindex msmall-text
11494 @opindex mlarge-text
11495 When @option{-msmall-text} is used, the compiler assumes that the
11496 code of the entire program (or shared library) fits in 4MB, and is
11497 thus reachable with a branch instruction. When @option{-msmall-data}
11498 is used, the compiler can assume that all local symbols share the
11499 same @code{$gp} value, and thus reduce the number of instructions
11500 required for a function call from 4 to 1.
11502 The default is @option{-mlarge-text}.
11504 @item -mcpu=@var{cpu_type}
11506 Set the instruction set and instruction scheduling parameters for
11507 machine type @var{cpu_type}. You can specify either the @samp{EV}
11508 style name or the corresponding chip number. GCC supports scheduling
11509 parameters for the EV4, EV5 and EV6 family of processors and will
11510 choose the default values for the instruction set from the processor
11511 you specify. If you do not specify a processor type, GCC will default
11512 to the processor on which the compiler was built.
11514 Supported values for @var{cpu_type} are
11520 Schedules as an EV4 and has no instruction set extensions.
11524 Schedules as an EV5 and has no instruction set extensions.
11528 Schedules as an EV5 and supports the BWX extension.
11533 Schedules as an EV5 and supports the BWX and MAX extensions.
11537 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
11541 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
11544 Native toolchains also support the value @samp{native},
11545 which selects the best architecture option for the host processor.
11546 @option{-mcpu=native} has no effect if GCC does not recognize
11549 @item -mtune=@var{cpu_type}
11551 Set only the instruction scheduling parameters for machine type
11552 @var{cpu_type}. The instruction set is not changed.
11554 Native toolchains also support the value @samp{native},
11555 which selects the best architecture option for the host processor.
11556 @option{-mtune=native} has no effect if GCC does not recognize
11559 @item -mmemory-latency=@var{time}
11560 @opindex mmemory-latency
11561 Sets the latency the scheduler should assume for typical memory
11562 references as seen by the application. This number is highly
11563 dependent on the memory access patterns used by the application
11564 and the size of the external cache on the machine.
11566 Valid options for @var{time} are
11570 A decimal number representing clock cycles.
11576 The compiler contains estimates of the number of clock cycles for
11577 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
11578 (also called Dcache, Scache, and Bcache), as well as to main memory.
11579 Note that L3 is only valid for EV5.
11584 @node DEC Alpha/VMS Options
11585 @subsection DEC Alpha/VMS Options
11587 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
11590 @item -mvms-return-codes
11591 @opindex mvms-return-codes
11592 Return VMS condition codes from main. The default is to return POSIX
11593 style condition (e.g.@: error) codes.
11595 @item -mdebug-main=@var{prefix}
11596 @opindex mdebug-main=@var{prefix}
11597 Flag the first routine whose name starts with @var{prefix} as the main
11598 routine for the debugger.
11602 Default to 64bit memory allocation routines.
11606 @subsection FR30 Options
11607 @cindex FR30 Options
11609 These options are defined specifically for the FR30 port.
11613 @item -msmall-model
11614 @opindex msmall-model
11615 Use the small address space model. This can produce smaller code, but
11616 it does assume that all symbolic values and addresses will fit into a
11621 Assume that run-time support has been provided and so there is no need
11622 to include the simulator library (@file{libsim.a}) on the linker
11628 @subsection FRV Options
11629 @cindex FRV Options
11635 Only use the first 32 general purpose registers.
11640 Use all 64 general purpose registers.
11645 Use only the first 32 floating point registers.
11650 Use all 64 floating point registers
11653 @opindex mhard-float
11655 Use hardware instructions for floating point operations.
11658 @opindex msoft-float
11660 Use library routines for floating point operations.
11665 Dynamically allocate condition code registers.
11670 Do not try to dynamically allocate condition code registers, only
11671 use @code{icc0} and @code{fcc0}.
11676 Change ABI to use double word insns.
11681 Do not use double word instructions.
11686 Use floating point double instructions.
11689 @opindex mno-double
11691 Do not use floating point double instructions.
11696 Use media instructions.
11701 Do not use media instructions.
11706 Use multiply and add/subtract instructions.
11709 @opindex mno-muladd
11711 Do not use multiply and add/subtract instructions.
11716 Select the FDPIC ABI, that uses function descriptors to represent
11717 pointers to functions. Without any PIC/PIE-related options, it
11718 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
11719 assumes GOT entries and small data are within a 12-bit range from the
11720 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
11721 are computed with 32 bits.
11722 With a @samp{bfin-elf} target, this option implies @option{-msim}.
11725 @opindex minline-plt
11727 Enable inlining of PLT entries in function calls to functions that are
11728 not known to bind locally. It has no effect without @option{-mfdpic}.
11729 It's enabled by default if optimizing for speed and compiling for
11730 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
11731 optimization option such as @option{-O3} or above is present in the
11737 Assume a large TLS segment when generating thread-local code.
11742 Do not assume a large TLS segment when generating thread-local code.
11747 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
11748 that is known to be in read-only sections. It's enabled by default,
11749 except for @option{-fpic} or @option{-fpie}: even though it may help
11750 make the global offset table smaller, it trades 1 instruction for 4.
11751 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
11752 one of which may be shared by multiple symbols, and it avoids the need
11753 for a GOT entry for the referenced symbol, so it's more likely to be a
11754 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
11756 @item -multilib-library-pic
11757 @opindex multilib-library-pic
11759 Link with the (library, not FD) pic libraries. It's implied by
11760 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
11761 @option{-fpic} without @option{-mfdpic}. You should never have to use
11765 @opindex mlinked-fp
11767 Follow the EABI requirement of always creating a frame pointer whenever
11768 a stack frame is allocated. This option is enabled by default and can
11769 be disabled with @option{-mno-linked-fp}.
11772 @opindex mlong-calls
11774 Use indirect addressing to call functions outside the current
11775 compilation unit. This allows the functions to be placed anywhere
11776 within the 32-bit address space.
11778 @item -malign-labels
11779 @opindex malign-labels
11781 Try to align labels to an 8-byte boundary by inserting nops into the
11782 previous packet. This option only has an effect when VLIW packing
11783 is enabled. It doesn't create new packets; it merely adds nops to
11786 @item -mlibrary-pic
11787 @opindex mlibrary-pic
11789 Generate position-independent EABI code.
11794 Use only the first four media accumulator registers.
11799 Use all eight media accumulator registers.
11804 Pack VLIW instructions.
11809 Do not pack VLIW instructions.
11812 @opindex mno-eflags
11814 Do not mark ABI switches in e_flags.
11817 @opindex mcond-move
11819 Enable the use of conditional-move instructions (default).
11821 This switch is mainly for debugging the compiler and will likely be removed
11822 in a future version.
11824 @item -mno-cond-move
11825 @opindex mno-cond-move
11827 Disable the use of conditional-move instructions.
11829 This switch is mainly for debugging the compiler and will likely be removed
11830 in a future version.
11835 Enable the use of conditional set instructions (default).
11837 This switch is mainly for debugging the compiler and will likely be removed
11838 in a future version.
11843 Disable the use of conditional set instructions.
11845 This switch is mainly for debugging the compiler and will likely be removed
11846 in a future version.
11849 @opindex mcond-exec
11851 Enable the use of conditional execution (default).
11853 This switch is mainly for debugging the compiler and will likely be removed
11854 in a future version.
11856 @item -mno-cond-exec
11857 @opindex mno-cond-exec
11859 Disable the use of conditional execution.
11861 This switch is mainly for debugging the compiler and will likely be removed
11862 in a future version.
11864 @item -mvliw-branch
11865 @opindex mvliw-branch
11867 Run a pass to pack branches into VLIW instructions (default).
11869 This switch is mainly for debugging the compiler and will likely be removed
11870 in a future version.
11872 @item -mno-vliw-branch
11873 @opindex mno-vliw-branch
11875 Do not run a pass to pack branches into VLIW instructions.
11877 This switch is mainly for debugging the compiler and will likely be removed
11878 in a future version.
11880 @item -mmulti-cond-exec
11881 @opindex mmulti-cond-exec
11883 Enable optimization of @code{&&} and @code{||} in conditional execution
11886 This switch is mainly for debugging the compiler and will likely be removed
11887 in a future version.
11889 @item -mno-multi-cond-exec
11890 @opindex mno-multi-cond-exec
11892 Disable optimization of @code{&&} and @code{||} in conditional execution.
11894 This switch is mainly for debugging the compiler and will likely be removed
11895 in a future version.
11897 @item -mnested-cond-exec
11898 @opindex mnested-cond-exec
11900 Enable nested conditional execution optimizations (default).
11902 This switch is mainly for debugging the compiler and will likely be removed
11903 in a future version.
11905 @item -mno-nested-cond-exec
11906 @opindex mno-nested-cond-exec
11908 Disable nested conditional execution optimizations.
11910 This switch is mainly for debugging the compiler and will likely be removed
11911 in a future version.
11913 @item -moptimize-membar
11914 @opindex moptimize-membar
11916 This switch removes redundant @code{membar} instructions from the
11917 compiler generated code. It is enabled by default.
11919 @item -mno-optimize-membar
11920 @opindex mno-optimize-membar
11922 This switch disables the automatic removal of redundant @code{membar}
11923 instructions from the generated code.
11925 @item -mtomcat-stats
11926 @opindex mtomcat-stats
11928 Cause gas to print out tomcat statistics.
11930 @item -mcpu=@var{cpu}
11933 Select the processor type for which to generate code. Possible values are
11934 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
11935 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
11939 @node GNU/Linux Options
11940 @subsection GNU/Linux Options
11942 These @samp{-m} options are defined for GNU/Linux targets:
11947 Use the GNU C library. This is the default except
11948 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
11952 Use uClibc C library. This is the default on
11953 @samp{*-*-linux-*uclibc*} targets.
11957 Use Bionic C library. This is the default on
11958 @samp{*-*-linux-*android*} targets.
11962 Compile code compatible with Android platform. This is the default on
11963 @samp{*-*-linux-*android*} targets.
11965 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
11966 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
11967 this option makes the GCC driver pass Android-specific options to the linker.
11968 Finally, this option causes the preprocessor macro @code{__ANDROID__}
11971 @item -tno-android-cc
11972 @opindex tno-android-cc
11973 Disable compilation effects of @option{-mandroid}, i.e., do not enable
11974 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
11975 @option{-fno-rtti} by default.
11977 @item -tno-android-ld
11978 @opindex tno-android-ld
11979 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
11980 linking options to the linker.
11984 @node H8/300 Options
11985 @subsection H8/300 Options
11987 These @samp{-m} options are defined for the H8/300 implementations:
11992 Shorten some address references at link time, when possible; uses the
11993 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
11994 ld, Using ld}, for a fuller description.
11998 Generate code for the H8/300H@.
12002 Generate code for the H8S@.
12006 Generate code for the H8S and H8/300H in the normal mode. This switch
12007 must be used either with @option{-mh} or @option{-ms}.
12011 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
12015 Make @code{int} data 32 bits by default.
12018 @opindex malign-300
12019 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
12020 The default for the H8/300H and H8S is to align longs and floats on 4
12022 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
12023 This option has no effect on the H8/300.
12027 @subsection HPPA Options
12028 @cindex HPPA Options
12030 These @samp{-m} options are defined for the HPPA family of computers:
12033 @item -march=@var{architecture-type}
12035 Generate code for the specified architecture. The choices for
12036 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
12037 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
12038 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
12039 architecture option for your machine. Code compiled for lower numbered
12040 architectures will run on higher numbered architectures, but not the
12043 @item -mpa-risc-1-0
12044 @itemx -mpa-risc-1-1
12045 @itemx -mpa-risc-2-0
12046 @opindex mpa-risc-1-0
12047 @opindex mpa-risc-1-1
12048 @opindex mpa-risc-2-0
12049 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
12052 @opindex mbig-switch
12053 Generate code suitable for big switch tables. Use this option only if
12054 the assembler/linker complain about out of range branches within a switch
12057 @item -mjump-in-delay
12058 @opindex mjump-in-delay
12059 Fill delay slots of function calls with unconditional jump instructions
12060 by modifying the return pointer for the function call to be the target
12061 of the conditional jump.
12063 @item -mdisable-fpregs
12064 @opindex mdisable-fpregs
12065 Prevent floating point registers from being used in any manner. This is
12066 necessary for compiling kernels which perform lazy context switching of
12067 floating point registers. If you use this option and attempt to perform
12068 floating point operations, the compiler will abort.
12070 @item -mdisable-indexing
12071 @opindex mdisable-indexing
12072 Prevent the compiler from using indexing address modes. This avoids some
12073 rather obscure problems when compiling MIG generated code under MACH@.
12075 @item -mno-space-regs
12076 @opindex mno-space-regs
12077 Generate code that assumes the target has no space registers. This allows
12078 GCC to generate faster indirect calls and use unscaled index address modes.
12080 Such code is suitable for level 0 PA systems and kernels.
12082 @item -mfast-indirect-calls
12083 @opindex mfast-indirect-calls
12084 Generate code that assumes calls never cross space boundaries. This
12085 allows GCC to emit code which performs faster indirect calls.
12087 This option will not work in the presence of shared libraries or nested
12090 @item -mfixed-range=@var{register-range}
12091 @opindex mfixed-range
12092 Generate code treating the given register range as fixed registers.
12093 A fixed register is one that the register allocator can not use. This is
12094 useful when compiling kernel code. A register range is specified as
12095 two registers separated by a dash. Multiple register ranges can be
12096 specified separated by a comma.
12098 @item -mlong-load-store
12099 @opindex mlong-load-store
12100 Generate 3-instruction load and store sequences as sometimes required by
12101 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
12104 @item -mportable-runtime
12105 @opindex mportable-runtime
12106 Use the portable calling conventions proposed by HP for ELF systems.
12110 Enable the use of assembler directives only GAS understands.
12112 @item -mschedule=@var{cpu-type}
12114 Schedule code according to the constraints for the machine type
12115 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
12116 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
12117 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
12118 proper scheduling option for your machine. The default scheduling is
12122 @opindex mlinker-opt
12123 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
12124 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
12125 linkers in which they give bogus error messages when linking some programs.
12128 @opindex msoft-float
12129 Generate output containing library calls for floating point.
12130 @strong{Warning:} the requisite libraries are not available for all HPPA
12131 targets. Normally the facilities of the machine's usual C compiler are
12132 used, but this cannot be done directly in cross-compilation. You must make
12133 your own arrangements to provide suitable library functions for
12136 @option{-msoft-float} changes the calling convention in the output file;
12137 therefore, it is only useful if you compile @emph{all} of a program with
12138 this option. In particular, you need to compile @file{libgcc.a}, the
12139 library that comes with GCC, with @option{-msoft-float} in order for
12144 Generate the predefine, @code{_SIO}, for server IO@. The default is
12145 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
12146 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
12147 options are available under HP-UX and HI-UX@.
12151 Use GNU ld specific options. This passes @option{-shared} to ld when
12152 building a shared library. It is the default when GCC is configured,
12153 explicitly or implicitly, with the GNU linker. This option does not
12154 have any affect on which ld is called, it only changes what parameters
12155 are passed to that ld. The ld that is called is determined by the
12156 @option{--with-ld} configure option, GCC's program search path, and
12157 finally by the user's @env{PATH}. The linker used by GCC can be printed
12158 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
12159 on the 64 bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12163 Use HP ld specific options. This passes @option{-b} to ld when building
12164 a shared library and passes @option{+Accept TypeMismatch} to ld on all
12165 links. It is the default when GCC is configured, explicitly or
12166 implicitly, with the HP linker. This option does not have any affect on
12167 which ld is called, it only changes what parameters are passed to that
12168 ld. The ld that is called is determined by the @option{--with-ld}
12169 configure option, GCC's program search path, and finally by the user's
12170 @env{PATH}. The linker used by GCC can be printed using @samp{which
12171 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
12172 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
12175 @opindex mno-long-calls
12176 Generate code that uses long call sequences. This ensures that a call
12177 is always able to reach linker generated stubs. The default is to generate
12178 long calls only when the distance from the call site to the beginning
12179 of the function or translation unit, as the case may be, exceeds a
12180 predefined limit set by the branch type being used. The limits for
12181 normal calls are 7,600,000 and 240,000 bytes, respectively for the
12182 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
12185 Distances are measured from the beginning of functions when using the
12186 @option{-ffunction-sections} option, or when using the @option{-mgas}
12187 and @option{-mno-portable-runtime} options together under HP-UX with
12190 It is normally not desirable to use this option as it will degrade
12191 performance. However, it may be useful in large applications,
12192 particularly when partial linking is used to build the application.
12194 The types of long calls used depends on the capabilities of the
12195 assembler and linker, and the type of code being generated. The
12196 impact on systems that support long absolute calls, and long pic
12197 symbol-difference or pc-relative calls should be relatively small.
12198 However, an indirect call is used on 32-bit ELF systems in pic code
12199 and it is quite long.
12201 @item -munix=@var{unix-std}
12203 Generate compiler predefines and select a startfile for the specified
12204 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
12205 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
12206 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
12207 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
12208 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
12211 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
12212 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
12213 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
12214 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
12215 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
12216 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
12218 It is @emph{important} to note that this option changes the interfaces
12219 for various library routines. It also affects the operational behavior
12220 of the C library. Thus, @emph{extreme} care is needed in using this
12223 Library code that is intended to operate with more than one UNIX
12224 standard must test, set and restore the variable @var{__xpg4_extended_mask}
12225 as appropriate. Most GNU software doesn't provide this capability.
12229 Suppress the generation of link options to search libdld.sl when the
12230 @option{-static} option is specified on HP-UX 10 and later.
12234 The HP-UX implementation of setlocale in libc has a dependency on
12235 libdld.sl. There isn't an archive version of libdld.sl. Thus,
12236 when the @option{-static} option is specified, special link options
12237 are needed to resolve this dependency.
12239 On HP-UX 10 and later, the GCC driver adds the necessary options to
12240 link with libdld.sl when the @option{-static} option is specified.
12241 This causes the resulting binary to be dynamic. On the 64-bit port,
12242 the linkers generate dynamic binaries by default in any case. The
12243 @option{-nolibdld} option can be used to prevent the GCC driver from
12244 adding these link options.
12248 Add support for multithreading with the @dfn{dce thread} library
12249 under HP-UX@. This option sets flags for both the preprocessor and
12253 @node i386 and x86-64 Options
12254 @subsection Intel 386 and AMD x86-64 Options
12255 @cindex i386 Options
12256 @cindex x86-64 Options
12257 @cindex Intel 386 Options
12258 @cindex AMD x86-64 Options
12260 These @samp{-m} options are defined for the i386 and x86-64 family of
12264 @item -mtune=@var{cpu-type}
12266 Tune to @var{cpu-type} everything applicable about the generated code, except
12267 for the ABI and the set of available instructions. The choices for
12268 @var{cpu-type} are:
12271 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
12272 If you know the CPU on which your code will run, then you should use
12273 the corresponding @option{-mtune} option instead of
12274 @option{-mtune=generic}. But, if you do not know exactly what CPU users
12275 of your application will have, then you should use this option.
12277 As new processors are deployed in the marketplace, the behavior of this
12278 option will change. Therefore, if you upgrade to a newer version of
12279 GCC, the code generated option will change to reflect the processors
12280 that were most common when that version of GCC was released.
12282 There is no @option{-march=generic} option because @option{-march}
12283 indicates the instruction set the compiler can use, and there is no
12284 generic instruction set applicable to all processors. In contrast,
12285 @option{-mtune} indicates the processor (or, in this case, collection of
12286 processors) for which the code is optimized.
12288 This selects the CPU to tune for at compilation time by determining
12289 the processor type of the compiling machine. Using @option{-mtune=native}
12290 will produce code optimized for the local machine under the constraints
12291 of the selected instruction set. Using @option{-march=native} will
12292 enable all instruction subsets supported by the local machine (hence
12293 the result might not run on different machines).
12295 Original Intel's i386 CPU@.
12297 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
12298 @item i586, pentium
12299 Intel Pentium CPU with no MMX support.
12301 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
12303 Intel PentiumPro CPU@.
12305 Same as @code{generic}, but when used as @code{march} option, PentiumPro
12306 instruction set will be used, so the code will run on all i686 family chips.
12308 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
12309 @item pentium3, pentium3m
12310 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
12313 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
12314 support. Used by Centrino notebooks.
12315 @item pentium4, pentium4m
12316 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
12318 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
12321 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
12322 SSE2 and SSE3 instruction set support.
12324 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12325 instruction set support.
12327 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1
12328 and SSE4.2 instruction set support.
12330 Intel Core i7 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12331 SSE4.1, SSE4.2, AVX, AES and PCLMUL instruction set support.
12333 Intel Core CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
12334 SSE4.1, SSE4.2, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C instruction
12337 Intel Atom CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
12338 instruction set support.
12340 AMD K6 CPU with MMX instruction set support.
12342 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
12343 @item athlon, athlon-tbird
12344 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
12346 @item athlon-4, athlon-xp, athlon-mp
12347 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
12348 instruction set support.
12349 @item k8, opteron, athlon64, athlon-fx
12350 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
12351 MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit instruction set extensions.)
12352 @item k8-sse3, opteron-sse3, athlon64-sse3
12353 Improved versions of k8, opteron and athlon64 with SSE3 instruction set support.
12354 @item amdfam10, barcelona
12355 AMD Family 10h core based CPUs with x86-64 instruction set support. (This
12356 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
12357 instruction set extensions.)
12359 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
12362 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
12363 instruction set support.
12365 Via C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
12366 implemented for this chip.)
12368 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
12369 implemented for this chip.)
12371 Embedded AMD CPU with MMX and 3DNow!@: instruction set support.
12374 While picking a specific @var{cpu-type} will schedule things appropriately
12375 for that particular chip, the compiler will not generate any code that
12376 does not run on the i386 without the @option{-march=@var{cpu-type}} option
12379 @item -march=@var{cpu-type}
12381 Generate instructions for the machine type @var{cpu-type}. The choices
12382 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
12383 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
12385 @item -mcpu=@var{cpu-type}
12387 A deprecated synonym for @option{-mtune}.
12389 @item -mfpmath=@var{unit}
12391 Generate floating point arithmetics for selected unit @var{unit}. The choices
12392 for @var{unit} are:
12396 Use the standard 387 floating point coprocessor present majority of chips and
12397 emulated otherwise. Code compiled with this option will run almost everywhere.
12398 The temporary results are computed in 80bit precision instead of precision
12399 specified by the type resulting in slightly different results compared to most
12400 of other chips. See @option{-ffloat-store} for more detailed description.
12402 This is the default choice for i386 compiler.
12405 Use scalar floating point instructions present in the SSE instruction set.
12406 This instruction set is supported by Pentium3 and newer chips, in the AMD line
12407 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
12408 instruction set supports only single precision arithmetics, thus the double and
12409 extended precision arithmetics is still done using 387. Later version, present
12410 only in Pentium4 and the future AMD x86-64 chips supports double precision
12413 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
12414 or @option{-msse2} switches to enable SSE extensions and make this option
12415 effective. For the x86-64 compiler, these extensions are enabled by default.
12417 The resulting code should be considerably faster in the majority of cases and avoid
12418 the numerical instability problems of 387 code, but may break some existing
12419 code that expects temporaries to be 80bit.
12421 This is the default choice for the x86-64 compiler.
12426 Attempt to utilize both instruction sets at once. This effectively double the
12427 amount of available registers and on chips with separate execution units for
12428 387 and SSE the execution resources too. Use this option with care, as it is
12429 still experimental, because the GCC register allocator does not model separate
12430 functional units well resulting in instable performance.
12433 @item -masm=@var{dialect}
12434 @opindex masm=@var{dialect}
12435 Output asm instructions using selected @var{dialect}. Supported
12436 choices are @samp{intel} or @samp{att} (the default one). Darwin does
12437 not support @samp{intel}.
12440 @itemx -mno-ieee-fp
12442 @opindex mno-ieee-fp
12443 Control whether or not the compiler uses IEEE floating point
12444 comparisons. These handle correctly the case where the result of a
12445 comparison is unordered.
12448 @opindex msoft-float
12449 Generate output containing library calls for floating point.
12450 @strong{Warning:} the requisite libraries are not part of GCC@.
12451 Normally the facilities of the machine's usual C compiler are used, but
12452 this can't be done directly in cross-compilation. You must make your
12453 own arrangements to provide suitable library functions for
12456 On machines where a function returns floating point results in the 80387
12457 register stack, some floating point opcodes may be emitted even if
12458 @option{-msoft-float} is used.
12460 @item -mno-fp-ret-in-387
12461 @opindex mno-fp-ret-in-387
12462 Do not use the FPU registers for return values of functions.
12464 The usual calling convention has functions return values of types
12465 @code{float} and @code{double} in an FPU register, even if there
12466 is no FPU@. The idea is that the operating system should emulate
12469 The option @option{-mno-fp-ret-in-387} causes such values to be returned
12470 in ordinary CPU registers instead.
12472 @item -mno-fancy-math-387
12473 @opindex mno-fancy-math-387
12474 Some 387 emulators do not support the @code{sin}, @code{cos} and
12475 @code{sqrt} instructions for the 387. Specify this option to avoid
12476 generating those instructions. This option is the default on FreeBSD,
12477 OpenBSD and NetBSD@. This option is overridden when @option{-march}
12478 indicates that the target CPU will always have an FPU and so the
12479 instruction will not need emulation. As of revision 2.6.1, these
12480 instructions are not generated unless you also use the
12481 @option{-funsafe-math-optimizations} switch.
12483 @item -malign-double
12484 @itemx -mno-align-double
12485 @opindex malign-double
12486 @opindex mno-align-double
12487 Control whether GCC aligns @code{double}, @code{long double}, and
12488 @code{long long} variables on a two word boundary or a one word
12489 boundary. Aligning @code{double} variables on a two word boundary will
12490 produce code that runs somewhat faster on a @samp{Pentium} at the
12491 expense of more memory.
12493 On x86-64, @option{-malign-double} is enabled by default.
12495 @strong{Warning:} if you use the @option{-malign-double} switch,
12496 structures containing the above types will be aligned differently than
12497 the published application binary interface specifications for the 386
12498 and will not be binary compatible with structures in code compiled
12499 without that switch.
12501 @item -m96bit-long-double
12502 @itemx -m128bit-long-double
12503 @opindex m96bit-long-double
12504 @opindex m128bit-long-double
12505 These switches control the size of @code{long double} type. The i386
12506 application binary interface specifies the size to be 96 bits,
12507 so @option{-m96bit-long-double} is the default in 32 bit mode.
12509 Modern architectures (Pentium and newer) would prefer @code{long double}
12510 to be aligned to an 8 or 16 byte boundary. In arrays or structures
12511 conforming to the ABI, this would not be possible. So specifying a
12512 @option{-m128bit-long-double} will align @code{long double}
12513 to a 16 byte boundary by padding the @code{long double} with an additional
12516 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
12517 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
12519 Notice that neither of these options enable any extra precision over the x87
12520 standard of 80 bits for a @code{long double}.
12522 @strong{Warning:} if you override the default value for your target ABI, the
12523 structures and arrays containing @code{long double} variables will change
12524 their size as well as function calling convention for function taking
12525 @code{long double} will be modified. Hence they will not be binary
12526 compatible with arrays or structures in code compiled without that switch.
12528 @item -mlarge-data-threshold=@var{number}
12529 @opindex mlarge-data-threshold=@var{number}
12530 When @option{-mcmodel=medium} is specified, the data greater than
12531 @var{threshold} are placed in large data section. This value must be the
12532 same across all object linked into the binary and defaults to 65535.
12536 Use a different function-calling convention, in which functions that
12537 take a fixed number of arguments return with the @code{ret} @var{num}
12538 instruction, which pops their arguments while returning. This saves one
12539 instruction in the caller since there is no need to pop the arguments
12542 You can specify that an individual function is called with this calling
12543 sequence with the function attribute @samp{stdcall}. You can also
12544 override the @option{-mrtd} option by using the function attribute
12545 @samp{cdecl}. @xref{Function Attributes}.
12547 @strong{Warning:} this calling convention is incompatible with the one
12548 normally used on Unix, so you cannot use it if you need to call
12549 libraries compiled with the Unix compiler.
12551 Also, you must provide function prototypes for all functions that
12552 take variable numbers of arguments (including @code{printf});
12553 otherwise incorrect code will be generated for calls to those
12556 In addition, seriously incorrect code will result if you call a
12557 function with too many arguments. (Normally, extra arguments are
12558 harmlessly ignored.)
12560 @item -mregparm=@var{num}
12562 Control how many registers are used to pass integer arguments. By
12563 default, no registers are used to pass arguments, and at most 3
12564 registers can be used. You can control this behavior for a specific
12565 function by using the function attribute @samp{regparm}.
12566 @xref{Function Attributes}.
12568 @strong{Warning:} if you use this switch, and
12569 @var{num} is nonzero, then you must build all modules with the same
12570 value, including any libraries. This includes the system libraries and
12574 @opindex msseregparm
12575 Use SSE register passing conventions for float and double arguments
12576 and return values. You can control this behavior for a specific
12577 function by using the function attribute @samp{sseregparm}.
12578 @xref{Function Attributes}.
12580 @strong{Warning:} if you use this switch then you must build all
12581 modules with the same value, including any libraries. This includes
12582 the system libraries and startup modules.
12584 @item -mvect8-ret-in-mem
12585 @opindex mvect8-ret-in-mem
12586 Return 8-byte vectors in memory instead of MMX registers. This is the
12587 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
12588 Studio compilers until version 12. Later compiler versions (starting
12589 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
12590 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
12591 you need to remain compatible with existing code produced by those
12592 previous compiler versions or older versions of GCC.
12601 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
12602 is specified, the significands of results of floating-point operations are
12603 rounded to 24 bits (single precision); @option{-mpc64} rounds the
12604 significands of results of floating-point operations to 53 bits (double
12605 precision) and @option{-mpc80} rounds the significands of results of
12606 floating-point operations to 64 bits (extended double precision), which is
12607 the default. When this option is used, floating-point operations in higher
12608 precisions are not available to the programmer without setting the FPU
12609 control word explicitly.
12611 Setting the rounding of floating-point operations to less than the default
12612 80 bits can speed some programs by 2% or more. Note that some mathematical
12613 libraries assume that extended precision (80 bit) floating-point operations
12614 are enabled by default; routines in such libraries could suffer significant
12615 loss of accuracy, typically through so-called "catastrophic cancellation",
12616 when this option is used to set the precision to less than extended precision.
12618 @item -mstackrealign
12619 @opindex mstackrealign
12620 Realign the stack at entry. On the Intel x86, the @option{-mstackrealign}
12621 option will generate an alternate prologue and epilogue that realigns the
12622 runtime stack if necessary. This supports mixing legacy codes that keep
12623 a 4-byte aligned stack with modern codes that keep a 16-byte stack for
12624 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
12625 applicable to individual functions.
12627 @item -mpreferred-stack-boundary=@var{num}
12628 @opindex mpreferred-stack-boundary
12629 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
12630 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
12631 the default is 4 (16 bytes or 128 bits).
12633 @item -mincoming-stack-boundary=@var{num}
12634 @opindex mincoming-stack-boundary
12635 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
12636 boundary. If @option{-mincoming-stack-boundary} is not specified,
12637 the one specified by @option{-mpreferred-stack-boundary} will be used.
12639 On Pentium and PentiumPro, @code{double} and @code{long double} values
12640 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
12641 suffer significant run time performance penalties. On Pentium III, the
12642 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
12643 properly if it is not 16 byte aligned.
12645 To ensure proper alignment of this values on the stack, the stack boundary
12646 must be as aligned as that required by any value stored on the stack.
12647 Further, every function must be generated such that it keeps the stack
12648 aligned. Thus calling a function compiled with a higher preferred
12649 stack boundary from a function compiled with a lower preferred stack
12650 boundary will most likely misalign the stack. It is recommended that
12651 libraries that use callbacks always use the default setting.
12653 This extra alignment does consume extra stack space, and generally
12654 increases code size. Code that is sensitive to stack space usage, such
12655 as embedded systems and operating system kernels, may want to reduce the
12656 preferred alignment to @option{-mpreferred-stack-boundary=2}.
12685 @itemx -mno-fsgsbase
12721 These switches enable or disable the use of instructions in the MMX, SSE,
12722 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, F16C,
12723 FMA, SSE4A, FMA4, XOP, LWP, ABM, BMI, BMI2, LZCNT or 3DNow!
12724 @: extended instruction sets.
12725 These extensions are also available as built-in functions: see
12726 @ref{X86 Built-in Functions}, for details of the functions enabled and
12727 disabled by these switches.
12729 To have SSE/SSE2 instructions generated automatically from floating-point
12730 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
12732 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
12733 generates new AVX instructions or AVX equivalence for all SSEx instructions
12736 These options will enable GCC to use these extended instructions in
12737 generated code, even without @option{-mfpmath=sse}. Applications which
12738 perform runtime CPU detection must compile separate files for each
12739 supported architecture, using the appropriate flags. In particular,
12740 the file containing the CPU detection code should be compiled without
12745 This option instructs GCC to emit a @code{cld} instruction in the prologue
12746 of functions that use string instructions. String instructions depend on
12747 the DF flag to select between autoincrement or autodecrement mode. While the
12748 ABI specifies the DF flag to be cleared on function entry, some operating
12749 systems violate this specification by not clearing the DF flag in their
12750 exception dispatchers. The exception handler can be invoked with the DF flag
12751 set which leads to wrong direction mode, when string instructions are used.
12752 This option can be enabled by default on 32-bit x86 targets by configuring
12753 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
12754 instructions can be suppressed with the @option{-mno-cld} compiler option
12758 @opindex mvzeroupper
12759 This option instructs GCC to emit a @code{vzeroupper} instruction
12760 before a transfer of control flow out of the function to minimize
12761 AVX to SSE transition penalty as well as remove unnecessary zeroupper
12766 This option will enable GCC to use CMPXCHG16B instruction in generated code.
12767 CMPXCHG16B allows for atomic operations on 128-bit double quadword (or oword)
12768 data types. This is useful for high resolution counters that could be updated
12769 by multiple processors (or cores). This instruction is generated as part of
12770 atomic built-in functions: see @ref{Atomic Builtins} for details.
12774 This option will enable GCC to use SAHF instruction in generated 64-bit code.
12775 Early Intel CPUs with Intel 64 lacked LAHF and SAHF instructions supported
12776 by AMD64 until introduction of Pentium 4 G1 step in December 2005. LAHF and
12777 SAHF are load and store instructions, respectively, for certain status flags.
12778 In 64-bit mode, SAHF instruction is used to optimize @code{fmod}, @code{drem}
12779 or @code{remainder} built-in functions: see @ref{Other Builtins} for details.
12783 This option will enable GCC to use movbe instruction to implement
12784 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
12788 This option will enable built-in functions, @code{__builtin_ia32_crc32qi},
12789 @code{__builtin_ia32_crc32hi}. @code{__builtin_ia32_crc32si} and
12790 @code{__builtin_ia32_crc32di} to generate the crc32 machine instruction.
12794 This option will enable GCC to use RCPSS and RSQRTSS instructions (and their
12795 vectorized variants RCPPS and RSQRTPS) with an additional Newton-Raphson step
12796 to increase precision instead of DIVSS and SQRTSS (and their vectorized
12797 variants) for single precision floating point arguments. These instructions
12798 are generated only when @option{-funsafe-math-optimizations} is enabled
12799 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
12800 Note that while the throughput of the sequence is higher than the throughput
12801 of the non-reciprocal instruction, the precision of the sequence can be
12802 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
12804 Note that GCC implements 1.0f/sqrtf(x) in terms of RSQRTSS (or RSQRTPS)
12805 already with @option{-ffast-math} (or the above option combination), and
12806 doesn't need @option{-mrecip}.
12808 @item -mveclibabi=@var{type}
12809 @opindex mveclibabi
12810 Specifies the ABI type to use for vectorizing intrinsics using an
12811 external library. Supported types are @code{svml} for the Intel short
12812 vector math library and @code{acml} for the AMD math core library style
12813 of interfacing. GCC will currently emit calls to @code{vmldExp2},
12814 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
12815 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
12816 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
12817 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
12818 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
12819 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
12820 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
12821 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
12822 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
12823 function type when @option{-mveclibabi=svml} is used and @code{__vrd2_sin},
12824 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
12825 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
12826 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
12827 @code{__vrs4_log10f} and @code{__vrs4_powf} for corresponding function type
12828 when @option{-mveclibabi=acml} is used. Both @option{-ftree-vectorize} and
12829 @option{-funsafe-math-optimizations} have to be enabled. A SVML or ACML ABI
12830 compatible library will have to be specified at link time.
12832 @item -mabi=@var{name}
12834 Generate code for the specified calling convention. Permissible values
12835 are: @samp{sysv} for the ABI used on GNU/Linux and other systems and
12836 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
12837 ABI when targeting Windows. On all other systems, the default is the
12838 SYSV ABI. You can control this behavior for a specific function by
12839 using the function attribute @samp{ms_abi}/@samp{sysv_abi}.
12840 @xref{Function Attributes}.
12842 @item -mtls-dialect=@var{type}
12843 @opindex mtls-dialect
12844 Generate code to access thread-local storage using the @samp{gnu} or
12845 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
12846 @samp{gnu2} is more efficient, but it may add compile- and run-time
12847 requirements that cannot be satisfied on all systems.
12850 @itemx -mno-push-args
12851 @opindex mpush-args
12852 @opindex mno-push-args
12853 Use PUSH operations to store outgoing parameters. This method is shorter
12854 and usually equally fast as method using SUB/MOV operations and is enabled
12855 by default. In some cases disabling it may improve performance because of
12856 improved scheduling and reduced dependencies.
12858 @item -maccumulate-outgoing-args
12859 @opindex maccumulate-outgoing-args
12860 If enabled, the maximum amount of space required for outgoing arguments will be
12861 computed in the function prologue. This is faster on most modern CPUs
12862 because of reduced dependencies, improved scheduling and reduced stack usage
12863 when preferred stack boundary is not equal to 2. The drawback is a notable
12864 increase in code size. This switch implies @option{-mno-push-args}.
12868 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
12869 on thread-safe exception handling must compile and link all code with the
12870 @option{-mthreads} option. When compiling, @option{-mthreads} defines
12871 @option{-D_MT}; when linking, it links in a special thread helper library
12872 @option{-lmingwthrd} which cleans up per thread exception handling data.
12874 @item -mno-align-stringops
12875 @opindex mno-align-stringops
12876 Do not align destination of inlined string operations. This switch reduces
12877 code size and improves performance in case the destination is already aligned,
12878 but GCC doesn't know about it.
12880 @item -minline-all-stringops
12881 @opindex minline-all-stringops
12882 By default GCC inlines string operations only when destination is known to be
12883 aligned at least to 4 byte boundary. This enables more inlining, increase code
12884 size, but may improve performance of code that depends on fast memcpy, strlen
12885 and memset for short lengths.
12887 @item -minline-stringops-dynamically
12888 @opindex minline-stringops-dynamically
12889 For string operation of unknown size, inline runtime checks so for small
12890 blocks inline code is used, while for large blocks library call is used.
12892 @item -mstringop-strategy=@var{alg}
12893 @opindex mstringop-strategy=@var{alg}
12894 Overwrite internal decision heuristic about particular algorithm to inline
12895 string operation with. The allowed values are @code{rep_byte},
12896 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
12897 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
12898 expanding inline loop, @code{libcall} for always expanding library call.
12900 @item -momit-leaf-frame-pointer
12901 @opindex momit-leaf-frame-pointer
12902 Don't keep the frame pointer in a register for leaf functions. This
12903 avoids the instructions to save, set up and restore frame pointers and
12904 makes an extra register available in leaf functions. The option
12905 @option{-fomit-frame-pointer} removes the frame pointer for all functions
12906 which might make debugging harder.
12908 @item -mtls-direct-seg-refs
12909 @itemx -mno-tls-direct-seg-refs
12910 @opindex mtls-direct-seg-refs
12911 Controls whether TLS variables may be accessed with offsets from the
12912 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
12913 or whether the thread base pointer must be added. Whether or not this
12914 is legal depends on the operating system, and whether it maps the
12915 segment to cover the entire TLS area.
12917 For systems that use GNU libc, the default is on.
12920 @itemx -mno-sse2avx
12922 Specify that the assembler should encode SSE instructions with VEX
12923 prefix. The option @option{-mavx} turns this on by default.
12928 If profiling is active @option{-pg} put the profiling
12929 counter call before prologue.
12930 Note: On x86 architectures the attribute @code{ms_hook_prologue}
12931 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
12934 @itemx -mno-8bit-idiv
12936 On some processors, like Intel Atom, 8bit unsigned integer divide is
12937 much faster than 32bit/64bit integer divide. This option will generate a
12938 runt-time check. If both dividend and divisor are within range of 0
12939 to 255, 8bit unsigned integer divide will be used instead of
12940 32bit/64bit integer divide.
12942 @item -mavx256-split-unaligned-load
12943 @item -mavx256-split-unaligned-store
12944 @opindex avx256-split-unaligned-load
12945 @opindex avx256-split-unaligned-store
12946 Split 32-byte AVX unaligned load and store.
12950 These @samp{-m} switches are supported in addition to the above
12951 on AMD x86-64 processors in 64-bit environments.
12960 Generate code for a 32-bit or 64-bit environment.
12961 The @option{-m32} option sets int, long and pointer to 32 bits and
12962 generates code that runs on any i386 system.
12963 The @option{-m64} option sets int to 32 bits and long and pointer
12964 to 64 bits and generates code for AMD's x86-64 architecture.
12965 The @option{-mx32} option sets int, long and pointer to 32 bits and
12966 generates code for AMD's x86-64 architecture.
12967 For darwin only the @option{-m64} option turns off the @option{-fno-pic}
12968 and @option{-mdynamic-no-pic} options.
12970 @item -mno-red-zone
12971 @opindex mno-red-zone
12972 Do not use a so called red zone for x86-64 code. The red zone is mandated
12973 by the x86-64 ABI, it is a 128-byte area beyond the location of the
12974 stack pointer that will not be modified by signal or interrupt handlers
12975 and therefore can be used for temporary data without adjusting the stack
12976 pointer. The flag @option{-mno-red-zone} disables this red zone.
12978 @item -mcmodel=small
12979 @opindex mcmodel=small
12980 Generate code for the small code model: the program and its symbols must
12981 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
12982 Programs can be statically or dynamically linked. This is the default
12985 @item -mcmodel=kernel
12986 @opindex mcmodel=kernel
12987 Generate code for the kernel code model. The kernel runs in the
12988 negative 2 GB of the address space.
12989 This model has to be used for Linux kernel code.
12991 @item -mcmodel=medium
12992 @opindex mcmodel=medium
12993 Generate code for the medium model: The program is linked in the lower 2
12994 GB of the address space. Small symbols are also placed there. Symbols
12995 with sizes larger than @option{-mlarge-data-threshold} are put into
12996 large data or bss sections and can be located above 2GB. Programs can
12997 be statically or dynamically linked.
12999 @item -mcmodel=large
13000 @opindex mcmodel=large
13001 Generate code for the large model: This model makes no assumptions
13002 about addresses and sizes of sections.
13005 @node i386 and x86-64 Windows Options
13006 @subsection i386 and x86-64 Windows Options
13007 @cindex i386 and x86-64 Windows Options
13009 These additional options are available for Windows targets:
13014 This option is available for Cygwin and MinGW targets. It
13015 specifies that a console application is to be generated, by
13016 instructing the linker to set the PE header subsystem type
13017 required for console applications.
13018 This is the default behavior for Cygwin and MinGW targets.
13022 This option is available for Cygwin and MinGW targets. It
13023 specifies that a DLL - a dynamic link library - is to be
13024 generated, enabling the selection of the required runtime
13025 startup object and entry point.
13027 @item -mnop-fun-dllimport
13028 @opindex mnop-fun-dllimport
13029 This option is available for Cygwin and MinGW targets. It
13030 specifies that the dllimport attribute should be ignored.
13034 This option is available for MinGW targets. It specifies
13035 that MinGW-specific thread support is to be used.
13039 This option is available for mingw-w64 targets. It specifies
13040 that the UNICODE macro is getting pre-defined and that the
13041 unicode capable runtime startup code is chosen.
13045 This option is available for Cygwin and MinGW targets. It
13046 specifies that the typical Windows pre-defined macros are to
13047 be set in the pre-processor, but does not influence the choice
13048 of runtime library/startup code.
13052 This option is available for Cygwin and MinGW targets. It
13053 specifies that a GUI application is to be generated by
13054 instructing the linker to set the PE header subsystem type
13057 @item -fno-set-stack-executable
13058 @opindex fno-set-stack-executable
13059 This option is available for MinGW targets. It specifies that
13060 the executable flag for stack used by nested functions isn't
13061 set. This is necessary for binaries running in kernel mode of
13062 Windows, as there the user32 API, which is used to set executable
13063 privileges, isn't available.
13065 @item -mpe-aligned-commons
13066 @opindex mpe-aligned-commons
13067 This option is available for Cygwin and MinGW targets. It
13068 specifies that the GNU extension to the PE file format that
13069 permits the correct alignment of COMMON variables should be
13070 used when generating code. It will be enabled by default if
13071 GCC detects that the target assembler found during configuration
13072 supports the feature.
13075 See also under @ref{i386 and x86-64 Options} for standard options.
13077 @node IA-64 Options
13078 @subsection IA-64 Options
13079 @cindex IA-64 Options
13081 These are the @samp{-m} options defined for the Intel IA-64 architecture.
13085 @opindex mbig-endian
13086 Generate code for a big endian target. This is the default for HP-UX@.
13088 @item -mlittle-endian
13089 @opindex mlittle-endian
13090 Generate code for a little endian target. This is the default for AIX5
13096 @opindex mno-gnu-as
13097 Generate (or don't) code for the GNU assembler. This is the default.
13098 @c Also, this is the default if the configure option @option{--with-gnu-as}
13104 @opindex mno-gnu-ld
13105 Generate (or don't) code for the GNU linker. This is the default.
13106 @c Also, this is the default if the configure option @option{--with-gnu-ld}
13111 Generate code that does not use a global pointer register. The result
13112 is not position independent code, and violates the IA-64 ABI@.
13114 @item -mvolatile-asm-stop
13115 @itemx -mno-volatile-asm-stop
13116 @opindex mvolatile-asm-stop
13117 @opindex mno-volatile-asm-stop
13118 Generate (or don't) a stop bit immediately before and after volatile asm
13121 @item -mregister-names
13122 @itemx -mno-register-names
13123 @opindex mregister-names
13124 @opindex mno-register-names
13125 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
13126 the stacked registers. This may make assembler output more readable.
13132 Disable (or enable) optimizations that use the small data section. This may
13133 be useful for working around optimizer bugs.
13135 @item -mconstant-gp
13136 @opindex mconstant-gp
13137 Generate code that uses a single constant global pointer value. This is
13138 useful when compiling kernel code.
13142 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
13143 This is useful when compiling firmware code.
13145 @item -minline-float-divide-min-latency
13146 @opindex minline-float-divide-min-latency
13147 Generate code for inline divides of floating point values
13148 using the minimum latency algorithm.
13150 @item -minline-float-divide-max-throughput
13151 @opindex minline-float-divide-max-throughput
13152 Generate code for inline divides of floating point values
13153 using the maximum throughput algorithm.
13155 @item -mno-inline-float-divide
13156 @opindex mno-inline-float-divide
13157 Do not generate inline code for divides of floating point values.
13159 @item -minline-int-divide-min-latency
13160 @opindex minline-int-divide-min-latency
13161 Generate code for inline divides of integer values
13162 using the minimum latency algorithm.
13164 @item -minline-int-divide-max-throughput
13165 @opindex minline-int-divide-max-throughput
13166 Generate code for inline divides of integer values
13167 using the maximum throughput algorithm.
13169 @item -mno-inline-int-divide
13170 @opindex mno-inline-int-divide
13171 Do not generate inline code for divides of integer values.
13173 @item -minline-sqrt-min-latency
13174 @opindex minline-sqrt-min-latency
13175 Generate code for inline square roots
13176 using the minimum latency algorithm.
13178 @item -minline-sqrt-max-throughput
13179 @opindex minline-sqrt-max-throughput
13180 Generate code for inline square roots
13181 using the maximum throughput algorithm.
13183 @item -mno-inline-sqrt
13184 @opindex mno-inline-sqrt
13185 Do not generate inline code for sqrt.
13188 @itemx -mno-fused-madd
13189 @opindex mfused-madd
13190 @opindex mno-fused-madd
13191 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
13192 instructions. The default is to use these instructions.
13194 @item -mno-dwarf2-asm
13195 @itemx -mdwarf2-asm
13196 @opindex mno-dwarf2-asm
13197 @opindex mdwarf2-asm
13198 Don't (or do) generate assembler code for the DWARF2 line number debugging
13199 info. This may be useful when not using the GNU assembler.
13201 @item -mearly-stop-bits
13202 @itemx -mno-early-stop-bits
13203 @opindex mearly-stop-bits
13204 @opindex mno-early-stop-bits
13205 Allow stop bits to be placed earlier than immediately preceding the
13206 instruction that triggered the stop bit. This can improve instruction
13207 scheduling, but does not always do so.
13209 @item -mfixed-range=@var{register-range}
13210 @opindex mfixed-range
13211 Generate code treating the given register range as fixed registers.
13212 A fixed register is one that the register allocator can not use. This is
13213 useful when compiling kernel code. A register range is specified as
13214 two registers separated by a dash. Multiple register ranges can be
13215 specified separated by a comma.
13217 @item -mtls-size=@var{tls-size}
13219 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
13222 @item -mtune=@var{cpu-type}
13224 Tune the instruction scheduling for a particular CPU, Valid values are
13225 itanium, itanium1, merced, itanium2, and mckinley.
13231 Generate code for a 32-bit or 64-bit environment.
13232 The 32-bit environment sets int, long and pointer to 32 bits.
13233 The 64-bit environment sets int to 32 bits and long and pointer
13234 to 64 bits. These are HP-UX specific flags.
13236 @item -mno-sched-br-data-spec
13237 @itemx -msched-br-data-spec
13238 @opindex mno-sched-br-data-spec
13239 @opindex msched-br-data-spec
13240 (Dis/En)able data speculative scheduling before reload.
13241 This will result in generation of the ld.a instructions and
13242 the corresponding check instructions (ld.c / chk.a).
13243 The default is 'disable'.
13245 @item -msched-ar-data-spec
13246 @itemx -mno-sched-ar-data-spec
13247 @opindex msched-ar-data-spec
13248 @opindex mno-sched-ar-data-spec
13249 (En/Dis)able data speculative scheduling after reload.
13250 This will result in generation of the ld.a instructions and
13251 the corresponding check instructions (ld.c / chk.a).
13252 The default is 'enable'.
13254 @item -mno-sched-control-spec
13255 @itemx -msched-control-spec
13256 @opindex mno-sched-control-spec
13257 @opindex msched-control-spec
13258 (Dis/En)able control speculative scheduling. This feature is
13259 available only during region scheduling (i.e.@: before reload).
13260 This will result in generation of the ld.s instructions and
13261 the corresponding check instructions chk.s .
13262 The default is 'disable'.
13264 @item -msched-br-in-data-spec
13265 @itemx -mno-sched-br-in-data-spec
13266 @opindex msched-br-in-data-spec
13267 @opindex mno-sched-br-in-data-spec
13268 (En/Dis)able speculative scheduling of the instructions that
13269 are dependent on the data speculative loads before reload.
13270 This is effective only with @option{-msched-br-data-spec} enabled.
13271 The default is 'enable'.
13273 @item -msched-ar-in-data-spec
13274 @itemx -mno-sched-ar-in-data-spec
13275 @opindex msched-ar-in-data-spec
13276 @opindex mno-sched-ar-in-data-spec
13277 (En/Dis)able speculative scheduling of the instructions that
13278 are dependent on the data speculative loads after reload.
13279 This is effective only with @option{-msched-ar-data-spec} enabled.
13280 The default is 'enable'.
13282 @item -msched-in-control-spec
13283 @itemx -mno-sched-in-control-spec
13284 @opindex msched-in-control-spec
13285 @opindex mno-sched-in-control-spec
13286 (En/Dis)able speculative scheduling of the instructions that
13287 are dependent on the control speculative loads.
13288 This is effective only with @option{-msched-control-spec} enabled.
13289 The default is 'enable'.
13291 @item -mno-sched-prefer-non-data-spec-insns
13292 @itemx -msched-prefer-non-data-spec-insns
13293 @opindex mno-sched-prefer-non-data-spec-insns
13294 @opindex msched-prefer-non-data-spec-insns
13295 If enabled, data speculative instructions will be chosen for schedule
13296 only if there are no other choices at the moment. This will make
13297 the use of the data speculation much more conservative.
13298 The default is 'disable'.
13300 @item -mno-sched-prefer-non-control-spec-insns
13301 @itemx -msched-prefer-non-control-spec-insns
13302 @opindex mno-sched-prefer-non-control-spec-insns
13303 @opindex msched-prefer-non-control-spec-insns
13304 If enabled, control speculative instructions will be chosen for schedule
13305 only if there are no other choices at the moment. This will make
13306 the use of the control speculation much more conservative.
13307 The default is 'disable'.
13309 @item -mno-sched-count-spec-in-critical-path
13310 @itemx -msched-count-spec-in-critical-path
13311 @opindex mno-sched-count-spec-in-critical-path
13312 @opindex msched-count-spec-in-critical-path
13313 If enabled, speculative dependencies will be considered during
13314 computation of the instructions priorities. This will make the use of the
13315 speculation a bit more conservative.
13316 The default is 'disable'.
13318 @item -msched-spec-ldc
13319 @opindex msched-spec-ldc
13320 Use a simple data speculation check. This option is on by default.
13322 @item -msched-control-spec-ldc
13323 @opindex msched-spec-ldc
13324 Use a simple check for control speculation. This option is on by default.
13326 @item -msched-stop-bits-after-every-cycle
13327 @opindex msched-stop-bits-after-every-cycle
13328 Place a stop bit after every cycle when scheduling. This option is on
13331 @item -msched-fp-mem-deps-zero-cost
13332 @opindex msched-fp-mem-deps-zero-cost
13333 Assume that floating-point stores and loads are not likely to cause a conflict
13334 when placed into the same instruction group. This option is disabled by
13337 @item -msel-sched-dont-check-control-spec
13338 @opindex msel-sched-dont-check-control-spec
13339 Generate checks for control speculation in selective scheduling.
13340 This flag is disabled by default.
13342 @item -msched-max-memory-insns=@var{max-insns}
13343 @opindex msched-max-memory-insns
13344 Limit on the number of memory insns per instruction group, giving lower
13345 priority to subsequent memory insns attempting to schedule in the same
13346 instruction group. Frequently useful to prevent cache bank conflicts.
13347 The default value is 1.
13349 @item -msched-max-memory-insns-hard-limit
13350 @opindex msched-max-memory-insns-hard-limit
13351 Disallow more than `msched-max-memory-insns' in instruction group.
13352 Otherwise, limit is `soft' meaning that we would prefer non-memory operations
13353 when limit is reached but may still schedule memory operations.
13357 @node IA-64/VMS Options
13358 @subsection IA-64/VMS Options
13360 These @samp{-m} options are defined for the IA-64/VMS implementations:
13363 @item -mvms-return-codes
13364 @opindex mvms-return-codes
13365 Return VMS condition codes from main. The default is to return POSIX
13366 style condition (e.g.@ error) codes.
13368 @item -mdebug-main=@var{prefix}
13369 @opindex mdebug-main=@var{prefix}
13370 Flag the first routine whose name starts with @var{prefix} as the main
13371 routine for the debugger.
13375 Default to 64bit memory allocation routines.
13379 @subsection LM32 Options
13380 @cindex LM32 options
13382 These @option{-m} options are defined for the Lattice Mico32 architecture:
13385 @item -mbarrel-shift-enabled
13386 @opindex mbarrel-shift-enabled
13387 Enable barrel-shift instructions.
13389 @item -mdivide-enabled
13390 @opindex mdivide-enabled
13391 Enable divide and modulus instructions.
13393 @item -mmultiply-enabled
13394 @opindex multiply-enabled
13395 Enable multiply instructions.
13397 @item -msign-extend-enabled
13398 @opindex msign-extend-enabled
13399 Enable sign extend instructions.
13401 @item -muser-enabled
13402 @opindex muser-enabled
13403 Enable user-defined instructions.
13408 @subsection M32C Options
13409 @cindex M32C options
13412 @item -mcpu=@var{name}
13414 Select the CPU for which code is generated. @var{name} may be one of
13415 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
13416 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
13417 the M32C/80 series.
13421 Specifies that the program will be run on the simulator. This causes
13422 an alternate runtime library to be linked in which supports, for
13423 example, file I/O@. You must not use this option when generating
13424 programs that will run on real hardware; you must provide your own
13425 runtime library for whatever I/O functions are needed.
13427 @item -memregs=@var{number}
13429 Specifies the number of memory-based pseudo-registers GCC will use
13430 during code generation. These pseudo-registers will be used like real
13431 registers, so there is a tradeoff between GCC's ability to fit the
13432 code into available registers, and the performance penalty of using
13433 memory instead of registers. Note that all modules in a program must
13434 be compiled with the same value for this option. Because of that, you
13435 must not use this option with the default runtime libraries gcc
13440 @node M32R/D Options
13441 @subsection M32R/D Options
13442 @cindex M32R/D options
13444 These @option{-m} options are defined for Renesas M32R/D architectures:
13449 Generate code for the M32R/2@.
13453 Generate code for the M32R/X@.
13457 Generate code for the M32R@. This is the default.
13459 @item -mmodel=small
13460 @opindex mmodel=small
13461 Assume all objects live in the lower 16MB of memory (so that their addresses
13462 can be loaded with the @code{ld24} instruction), and assume all subroutines
13463 are reachable with the @code{bl} instruction.
13464 This is the default.
13466 The addressability of a particular object can be set with the
13467 @code{model} attribute.
13469 @item -mmodel=medium
13470 @opindex mmodel=medium
13471 Assume objects may be anywhere in the 32-bit address space (the compiler
13472 will generate @code{seth/add3} instructions to load their addresses), and
13473 assume all subroutines are reachable with the @code{bl} instruction.
13475 @item -mmodel=large
13476 @opindex mmodel=large
13477 Assume objects may be anywhere in the 32-bit address space (the compiler
13478 will generate @code{seth/add3} instructions to load their addresses), and
13479 assume subroutines may not be reachable with the @code{bl} instruction
13480 (the compiler will generate the much slower @code{seth/add3/jl}
13481 instruction sequence).
13484 @opindex msdata=none
13485 Disable use of the small data area. Variables will be put into
13486 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
13487 @code{section} attribute has been specified).
13488 This is the default.
13490 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
13491 Objects may be explicitly put in the small data area with the
13492 @code{section} attribute using one of these sections.
13494 @item -msdata=sdata
13495 @opindex msdata=sdata
13496 Put small global and static data in the small data area, but do not
13497 generate special code to reference them.
13500 @opindex msdata=use
13501 Put small global and static data in the small data area, and generate
13502 special instructions to reference them.
13506 @cindex smaller data references
13507 Put global and static objects less than or equal to @var{num} bytes
13508 into the small data or bss sections instead of the normal data or bss
13509 sections. The default value of @var{num} is 8.
13510 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
13511 for this option to have any effect.
13513 All modules should be compiled with the same @option{-G @var{num}} value.
13514 Compiling with different values of @var{num} may or may not work; if it
13515 doesn't the linker will give an error message---incorrect code will not be
13520 Makes the M32R specific code in the compiler display some statistics
13521 that might help in debugging programs.
13523 @item -malign-loops
13524 @opindex malign-loops
13525 Align all loops to a 32-byte boundary.
13527 @item -mno-align-loops
13528 @opindex mno-align-loops
13529 Do not enforce a 32-byte alignment for loops. This is the default.
13531 @item -missue-rate=@var{number}
13532 @opindex missue-rate=@var{number}
13533 Issue @var{number} instructions per cycle. @var{number} can only be 1
13536 @item -mbranch-cost=@var{number}
13537 @opindex mbranch-cost=@var{number}
13538 @var{number} can only be 1 or 2. If it is 1 then branches will be
13539 preferred over conditional code, if it is 2, then the opposite will
13542 @item -mflush-trap=@var{number}
13543 @opindex mflush-trap=@var{number}
13544 Specifies the trap number to use to flush the cache. The default is
13545 12. Valid numbers are between 0 and 15 inclusive.
13547 @item -mno-flush-trap
13548 @opindex mno-flush-trap
13549 Specifies that the cache cannot be flushed by using a trap.
13551 @item -mflush-func=@var{name}
13552 @opindex mflush-func=@var{name}
13553 Specifies the name of the operating system function to call to flush
13554 the cache. The default is @emph{_flush_cache}, but a function call
13555 will only be used if a trap is not available.
13557 @item -mno-flush-func
13558 @opindex mno-flush-func
13559 Indicates that there is no OS function for flushing the cache.
13563 @node M680x0 Options
13564 @subsection M680x0 Options
13565 @cindex M680x0 options
13567 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
13568 The default settings depend on which architecture was selected when
13569 the compiler was configured; the defaults for the most common choices
13573 @item -march=@var{arch}
13575 Generate code for a specific M680x0 or ColdFire instruction set
13576 architecture. Permissible values of @var{arch} for M680x0
13577 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
13578 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
13579 architectures are selected according to Freescale's ISA classification
13580 and the permissible values are: @samp{isaa}, @samp{isaaplus},
13581 @samp{isab} and @samp{isac}.
13583 gcc defines a macro @samp{__mcf@var{arch}__} whenever it is generating
13584 code for a ColdFire target. The @var{arch} in this macro is one of the
13585 @option{-march} arguments given above.
13587 When used together, @option{-march} and @option{-mtune} select code
13588 that runs on a family of similar processors but that is optimized
13589 for a particular microarchitecture.
13591 @item -mcpu=@var{cpu}
13593 Generate code for a specific M680x0 or ColdFire processor.
13594 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
13595 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
13596 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
13597 below, which also classifies the CPUs into families:
13599 @multitable @columnfractions 0.20 0.80
13600 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
13601 @item @samp{51} @tab @samp{51} @samp{51ac} @samp{51cn} @samp{51em} @samp{51qe}
13602 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
13603 @item @samp{5206e} @tab @samp{5206e}
13604 @item @samp{5208} @tab @samp{5207} @samp{5208}
13605 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
13606 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
13607 @item @samp{5216} @tab @samp{5214} @samp{5216}
13608 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
13609 @item @samp{5225} @tab @samp{5224} @samp{5225}
13610 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
13611 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
13612 @item @samp{5249} @tab @samp{5249}
13613 @item @samp{5250} @tab @samp{5250}
13614 @item @samp{5271} @tab @samp{5270} @samp{5271}
13615 @item @samp{5272} @tab @samp{5272}
13616 @item @samp{5275} @tab @samp{5274} @samp{5275}
13617 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
13618 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
13619 @item @samp{5307} @tab @samp{5307}
13620 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
13621 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
13622 @item @samp{5407} @tab @samp{5407}
13623 @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}
13626 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
13627 @var{arch} is compatible with @var{cpu}. Other combinations of
13628 @option{-mcpu} and @option{-march} are rejected.
13630 gcc defines the macro @samp{__mcf_cpu_@var{cpu}} when ColdFire target
13631 @var{cpu} is selected. It also defines @samp{__mcf_family_@var{family}},
13632 where the value of @var{family} is given by the table above.
13634 @item -mtune=@var{tune}
13636 Tune the code for a particular microarchitecture, within the
13637 constraints set by @option{-march} and @option{-mcpu}.
13638 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
13639 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
13640 and @samp{cpu32}. The ColdFire microarchitectures
13641 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
13643 You can also use @option{-mtune=68020-40} for code that needs
13644 to run relatively well on 68020, 68030 and 68040 targets.
13645 @option{-mtune=68020-60} is similar but includes 68060 targets
13646 as well. These two options select the same tuning decisions as
13647 @option{-m68020-40} and @option{-m68020-60} respectively.
13649 gcc defines the macros @samp{__mc@var{arch}} and @samp{__mc@var{arch}__}
13650 when tuning for 680x0 architecture @var{arch}. It also defines
13651 @samp{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
13652 option is used. If gcc is tuning for a range of architectures,
13653 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
13654 it defines the macros for every architecture in the range.
13656 gcc also defines the macro @samp{__m@var{uarch}__} when tuning for
13657 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
13658 of the arguments given above.
13664 Generate output for a 68000. This is the default
13665 when the compiler is configured for 68000-based systems.
13666 It is equivalent to @option{-march=68000}.
13668 Use this option for microcontrollers with a 68000 or EC000 core,
13669 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
13673 Generate output for a 68010. This is the default
13674 when the compiler is configured for 68010-based systems.
13675 It is equivalent to @option{-march=68010}.
13681 Generate output for a 68020. This is the default
13682 when the compiler is configured for 68020-based systems.
13683 It is equivalent to @option{-march=68020}.
13687 Generate output for a 68030. This is the default when the compiler is
13688 configured for 68030-based systems. It is equivalent to
13689 @option{-march=68030}.
13693 Generate output for a 68040. This is the default when the compiler is
13694 configured for 68040-based systems. It is equivalent to
13695 @option{-march=68040}.
13697 This option inhibits the use of 68881/68882 instructions that have to be
13698 emulated by software on the 68040. Use this option if your 68040 does not
13699 have code to emulate those instructions.
13703 Generate output for a 68060. This is the default when the compiler is
13704 configured for 68060-based systems. It is equivalent to
13705 @option{-march=68060}.
13707 This option inhibits the use of 68020 and 68881/68882 instructions that
13708 have to be emulated by software on the 68060. Use this option if your 68060
13709 does not have code to emulate those instructions.
13713 Generate output for a CPU32. This is the default
13714 when the compiler is configured for CPU32-based systems.
13715 It is equivalent to @option{-march=cpu32}.
13717 Use this option for microcontrollers with a
13718 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
13719 68336, 68340, 68341, 68349 and 68360.
13723 Generate output for a 520X ColdFire CPU@. This is the default
13724 when the compiler is configured for 520X-based systems.
13725 It is equivalent to @option{-mcpu=5206}, and is now deprecated
13726 in favor of that option.
13728 Use this option for microcontroller with a 5200 core, including
13729 the MCF5202, MCF5203, MCF5204 and MCF5206.
13733 Generate output for a 5206e ColdFire CPU@. The option is now
13734 deprecated in favor of the equivalent @option{-mcpu=5206e}.
13738 Generate output for a member of the ColdFire 528X family.
13739 The option is now deprecated in favor of the equivalent
13740 @option{-mcpu=528x}.
13744 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
13745 in favor of the equivalent @option{-mcpu=5307}.
13749 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
13750 in favor of the equivalent @option{-mcpu=5407}.
13754 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
13755 This includes use of hardware floating point instructions.
13756 The option is equivalent to @option{-mcpu=547x}, and is now
13757 deprecated in favor of that option.
13761 Generate output for a 68040, without using any of the new instructions.
13762 This results in code which can run relatively efficiently on either a
13763 68020/68881 or a 68030 or a 68040. The generated code does use the
13764 68881 instructions that are emulated on the 68040.
13766 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
13770 Generate output for a 68060, without using any of the new instructions.
13771 This results in code which can run relatively efficiently on either a
13772 68020/68881 or a 68030 or a 68040. The generated code does use the
13773 68881 instructions that are emulated on the 68060.
13775 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
13779 @opindex mhard-float
13781 Generate floating-point instructions. This is the default for 68020
13782 and above, and for ColdFire devices that have an FPU@. It defines the
13783 macro @samp{__HAVE_68881__} on M680x0 targets and @samp{__mcffpu__}
13784 on ColdFire targets.
13787 @opindex msoft-float
13788 Do not generate floating-point instructions; use library calls instead.
13789 This is the default for 68000, 68010, and 68832 targets. It is also
13790 the default for ColdFire devices that have no FPU.
13796 Generate (do not generate) ColdFire hardware divide and remainder
13797 instructions. If @option{-march} is used without @option{-mcpu},
13798 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
13799 architectures. Otherwise, the default is taken from the target CPU
13800 (either the default CPU, or the one specified by @option{-mcpu}). For
13801 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
13802 @option{-mcpu=5206e}.
13804 gcc defines the macro @samp{__mcfhwdiv__} when this option is enabled.
13808 Consider type @code{int} to be 16 bits wide, like @code{short int}.
13809 Additionally, parameters passed on the stack are also aligned to a
13810 16-bit boundary even on targets whose API mandates promotion to 32-bit.
13814 Do not consider type @code{int} to be 16 bits wide. This is the default.
13817 @itemx -mno-bitfield
13818 @opindex mnobitfield
13819 @opindex mno-bitfield
13820 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
13821 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
13825 Do use the bit-field instructions. The @option{-m68020} option implies
13826 @option{-mbitfield}. This is the default if you use a configuration
13827 designed for a 68020.
13831 Use a different function-calling convention, in which functions
13832 that take a fixed number of arguments return with the @code{rtd}
13833 instruction, which pops their arguments while returning. This
13834 saves one instruction in the caller since there is no need to pop
13835 the arguments there.
13837 This calling convention is incompatible with the one normally
13838 used on Unix, so you cannot use it if you need to call libraries
13839 compiled with the Unix compiler.
13841 Also, you must provide function prototypes for all functions that
13842 take variable numbers of arguments (including @code{printf});
13843 otherwise incorrect code will be generated for calls to those
13846 In addition, seriously incorrect code will result if you call a
13847 function with too many arguments. (Normally, extra arguments are
13848 harmlessly ignored.)
13850 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
13851 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
13855 Do not use the calling conventions selected by @option{-mrtd}.
13856 This is the default.
13859 @itemx -mno-align-int
13860 @opindex malign-int
13861 @opindex mno-align-int
13862 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
13863 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
13864 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
13865 Aligning variables on 32-bit boundaries produces code that runs somewhat
13866 faster on processors with 32-bit busses at the expense of more memory.
13868 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
13869 align structures containing the above types differently than
13870 most published application binary interface specifications for the m68k.
13874 Use the pc-relative addressing mode of the 68000 directly, instead of
13875 using a global offset table. At present, this option implies @option{-fpic},
13876 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
13877 not presently supported with @option{-mpcrel}, though this could be supported for
13878 68020 and higher processors.
13880 @item -mno-strict-align
13881 @itemx -mstrict-align
13882 @opindex mno-strict-align
13883 @opindex mstrict-align
13884 Do not (do) assume that unaligned memory references will be handled by
13888 Generate code that allows the data segment to be located in a different
13889 area of memory from the text segment. This allows for execute in place in
13890 an environment without virtual memory management. This option implies
13893 @item -mno-sep-data
13894 Generate code that assumes that the data segment follows the text segment.
13895 This is the default.
13897 @item -mid-shared-library
13898 Generate code that supports shared libraries via the library ID method.
13899 This allows for execute in place and shared libraries in an environment
13900 without virtual memory management. This option implies @option{-fPIC}.
13902 @item -mno-id-shared-library
13903 Generate code that doesn't assume ID based shared libraries are being used.
13904 This is the default.
13906 @item -mshared-library-id=n
13907 Specified the identification number of the ID based shared library being
13908 compiled. Specifying a value of 0 will generate more compact code, specifying
13909 other values will force the allocation of that number to the current
13910 library but is no more space or time efficient than omitting this option.
13916 When generating position-independent code for ColdFire, generate code
13917 that works if the GOT has more than 8192 entries. This code is
13918 larger and slower than code generated without this option. On M680x0
13919 processors, this option is not needed; @option{-fPIC} suffices.
13921 GCC normally uses a single instruction to load values from the GOT@.
13922 While this is relatively efficient, it only works if the GOT
13923 is smaller than about 64k. Anything larger causes the linker
13924 to report an error such as:
13926 @cindex relocation truncated to fit (ColdFire)
13928 relocation truncated to fit: R_68K_GOT16O foobar
13931 If this happens, you should recompile your code with @option{-mxgot}.
13932 It should then work with very large GOTs. However, code generated with
13933 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
13934 the value of a global symbol.
13936 Note that some linkers, including newer versions of the GNU linker,
13937 can create multiple GOTs and sort GOT entries. If you have such a linker,
13938 you should only need to use @option{-mxgot} when compiling a single
13939 object file that accesses more than 8192 GOT entries. Very few do.
13941 These options have no effect unless GCC is generating
13942 position-independent code.
13946 @node MCore Options
13947 @subsection MCore Options
13948 @cindex MCore options
13950 These are the @samp{-m} options defined for the Motorola M*Core
13956 @itemx -mno-hardlit
13958 @opindex mno-hardlit
13959 Inline constants into the code stream if it can be done in two
13960 instructions or less.
13966 Use the divide instruction. (Enabled by default).
13968 @item -mrelax-immediate
13969 @itemx -mno-relax-immediate
13970 @opindex mrelax-immediate
13971 @opindex mno-relax-immediate
13972 Allow arbitrary sized immediates in bit operations.
13974 @item -mwide-bitfields
13975 @itemx -mno-wide-bitfields
13976 @opindex mwide-bitfields
13977 @opindex mno-wide-bitfields
13978 Always treat bit-fields as int-sized.
13980 @item -m4byte-functions
13981 @itemx -mno-4byte-functions
13982 @opindex m4byte-functions
13983 @opindex mno-4byte-functions
13984 Force all functions to be aligned to a four byte boundary.
13986 @item -mcallgraph-data
13987 @itemx -mno-callgraph-data
13988 @opindex mcallgraph-data
13989 @opindex mno-callgraph-data
13990 Emit callgraph information.
13993 @itemx -mno-slow-bytes
13994 @opindex mslow-bytes
13995 @opindex mno-slow-bytes
13996 Prefer word access when reading byte quantities.
13998 @item -mlittle-endian
13999 @itemx -mbig-endian
14000 @opindex mlittle-endian
14001 @opindex mbig-endian
14002 Generate code for a little endian target.
14008 Generate code for the 210 processor.
14012 Assume that run-time support has been provided and so omit the
14013 simulator library (@file{libsim.a)} from the linker command line.
14015 @item -mstack-increment=@var{size}
14016 @opindex mstack-increment
14017 Set the maximum amount for a single stack increment operation. Large
14018 values can increase the speed of programs which contain functions
14019 that need a large amount of stack space, but they can also trigger a
14020 segmentation fault if the stack is extended too much. The default
14026 @subsection MeP Options
14027 @cindex MeP options
14033 Enables the @code{abs} instruction, which is the absolute difference
14034 between two registers.
14038 Enables all the optional instructions - average, multiply, divide, bit
14039 operations, leading zero, absolute difference, min/max, clip, and
14045 Enables the @code{ave} instruction, which computes the average of two
14048 @item -mbased=@var{n}
14050 Variables of size @var{n} bytes or smaller will be placed in the
14051 @code{.based} section by default. Based variables use the @code{$tp}
14052 register as a base register, and there is a 128 byte limit to the
14053 @code{.based} section.
14057 Enables the bit operation instructions - bit test (@code{btstm}), set
14058 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
14059 test-and-set (@code{tas}).
14061 @item -mc=@var{name}
14063 Selects which section constant data will be placed in. @var{name} may
14064 be @code{tiny}, @code{near}, or @code{far}.
14068 Enables the @code{clip} instruction. Note that @code{-mclip} is not
14069 useful unless you also provide @code{-mminmax}.
14071 @item -mconfig=@var{name}
14073 Selects one of the build-in core configurations. Each MeP chip has
14074 one or more modules in it; each module has a core CPU and a variety of
14075 coprocessors, optional instructions, and peripherals. The
14076 @code{MeP-Integrator} tool, not part of GCC, provides these
14077 configurations through this option; using this option is the same as
14078 using all the corresponding command line options. The default
14079 configuration is @code{default}.
14083 Enables the coprocessor instructions. By default, this is a 32-bit
14084 coprocessor. Note that the coprocessor is normally enabled via the
14085 @code{-mconfig=} option.
14089 Enables the 32-bit coprocessor's instructions.
14093 Enables the 64-bit coprocessor's instructions.
14097 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
14101 Causes constant variables to be placed in the @code{.near} section.
14105 Enables the @code{div} and @code{divu} instructions.
14109 Generate big-endian code.
14113 Generate little-endian code.
14115 @item -mio-volatile
14116 @opindex mio-volatile
14117 Tells the compiler that any variable marked with the @code{io}
14118 attribute is to be considered volatile.
14122 Causes variables to be assigned to the @code{.far} section by default.
14126 Enables the @code{leadz} (leading zero) instruction.
14130 Causes variables to be assigned to the @code{.near} section by default.
14134 Enables the @code{min} and @code{max} instructions.
14138 Enables the multiplication and multiply-accumulate instructions.
14142 Disables all the optional instructions enabled by @code{-mall-opts}.
14146 Enables the @code{repeat} and @code{erepeat} instructions, used for
14147 low-overhead looping.
14151 Causes all variables to default to the @code{.tiny} section. Note
14152 that there is a 65536 byte limit to this section. Accesses to these
14153 variables use the @code{%gp} base register.
14157 Enables the saturation instructions. Note that the compiler does not
14158 currently generate these itself, but this option is included for
14159 compatibility with other tools, like @code{as}.
14163 Link the SDRAM-based runtime instead of the default ROM-based runtime.
14167 Link the simulator runtime libraries.
14171 Link the simulator runtime libraries, excluding built-in support
14172 for reset and exception vectors and tables.
14176 Causes all functions to default to the @code{.far} section. Without
14177 this option, functions default to the @code{.near} section.
14179 @item -mtiny=@var{n}
14181 Variables that are @var{n} bytes or smaller will be allocated to the
14182 @code{.tiny} section. These variables use the @code{$gp} base
14183 register. The default for this option is 4, but note that there's a
14184 65536 byte limit to the @code{.tiny} section.
14188 @node MicroBlaze Options
14189 @subsection MicroBlaze Options
14190 @cindex MicroBlaze Options
14195 @opindex msoft-float
14196 Use software emulation for floating point (default).
14199 @opindex mhard-float
14200 Use hardware floating point instructions.
14204 Do not optimize block moves, use @code{memcpy}.
14206 @item -mno-clearbss
14207 @opindex mno-clearbss
14208 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
14210 @item -mcpu=@var{cpu-type}
14212 Use features of and schedule code for given CPU.
14213 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
14214 where @var{X} is a major version, @var{YY} is the minor version, and
14215 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
14216 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
14218 @item -mxl-soft-mul
14219 @opindex mxl-soft-mul
14220 Use software multiply emulation (default).
14222 @item -mxl-soft-div
14223 @opindex mxl-soft-div
14224 Use software emulation for divides (default).
14226 @item -mxl-barrel-shift
14227 @opindex mxl-barrel-shift
14228 Use the hardware barrel shifter.
14230 @item -mxl-pattern-compare
14231 @opindex mxl-pattern-compare
14232 Use pattern compare instructions.
14234 @item -msmall-divides
14235 @opindex msmall-divides
14236 Use table lookup optimization for small signed integer divisions.
14238 @item -mxl-stack-check
14239 @opindex mxl-stack-check
14240 This option is deprecated. Use -fstack-check instead.
14243 @opindex mxl-gp-opt
14244 Use GP relative sdata/sbss sections.
14246 @item -mxl-multiply-high
14247 @opindex mxl-multiply-high
14248 Use multiply high instructions for high part of 32x32 multiply.
14250 @item -mxl-float-convert
14251 @opindex mxl-float-convert
14252 Use hardware floating point conversion instructions.
14254 @item -mxl-float-sqrt
14255 @opindex mxl-float-sqrt
14256 Use hardware floating point square root instruction.
14258 @item -mxl-mode-@var{app-model}
14259 Select application model @var{app-model}. Valid models are
14262 normal executable (default), uses startup code @file{crt0.o}.
14265 for use with Xilinx Microprocessor Debugger (XMD) based
14266 software intrusive debug agent called xmdstub. This uses startup file
14267 @file{crt1.o} and sets the start address of the program to be 0x800.
14270 for applications that are loaded using a bootloader.
14271 This model uses startup file @file{crt2.o} which does not contain a processor
14272 reset vector handler. This is suitable for transferring control on a
14273 processor reset to the bootloader rather than the application.
14276 for applications that do not require any of the
14277 MicroBlaze vectors. This option may be useful for applications running
14278 within a monitoring application. This model uses @file{crt3.o} as a startup file.
14281 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
14282 @option{-mxl-mode-@var{app-model}}.
14287 @subsection MIPS Options
14288 @cindex MIPS options
14294 Generate big-endian code.
14298 Generate little-endian code. This is the default for @samp{mips*el-*-*}
14301 @item -march=@var{arch}
14303 Generate code that will run on @var{arch}, which can be the name of a
14304 generic MIPS ISA, or the name of a particular processor.
14306 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
14307 @samp{mips32}, @samp{mips32r2}, @samp{mips64} and @samp{mips64r2}.
14308 The processor names are:
14309 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
14310 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
14311 @samp{5kc}, @samp{5kf},
14313 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
14314 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
14315 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1},
14316 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
14317 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
14318 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
14322 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
14323 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
14324 @samp{rm7000}, @samp{rm9000},
14325 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
14328 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
14329 @samp{vr5000}, @samp{vr5400}, @samp{vr5500}
14331 The special value @samp{from-abi} selects the
14332 most compatible architecture for the selected ABI (that is,
14333 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
14335 Native Linux/GNU and IRIX toolchains also support the value @samp{native},
14336 which selects the best architecture option for the host processor.
14337 @option{-march=native} has no effect if GCC does not recognize
14340 In processor names, a final @samp{000} can be abbreviated as @samp{k}
14341 (for example, @samp{-march=r2k}). Prefixes are optional, and
14342 @samp{vr} may be written @samp{r}.
14344 Names of the form @samp{@var{n}f2_1} refer to processors with
14345 FPUs clocked at half the rate of the core, names of the form
14346 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
14347 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
14348 processors with FPUs clocked a ratio of 3:2 with respect to the core.
14349 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
14350 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
14351 accepted as synonyms for @samp{@var{n}f1_1}.
14353 GCC defines two macros based on the value of this option. The first
14354 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
14355 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
14356 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
14357 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
14358 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
14360 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
14361 above. In other words, it will have the full prefix and will not
14362 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
14363 the macro names the resolved architecture (either @samp{"mips1"} or
14364 @samp{"mips3"}). It names the default architecture when no
14365 @option{-march} option is given.
14367 @item -mtune=@var{arch}
14369 Optimize for @var{arch}. Among other things, this option controls
14370 the way instructions are scheduled, and the perceived cost of arithmetic
14371 operations. The list of @var{arch} values is the same as for
14374 When this option is not used, GCC will optimize for the processor
14375 specified by @option{-march}. By using @option{-march} and
14376 @option{-mtune} together, it is possible to generate code that will
14377 run on a family of processors, but optimize the code for one
14378 particular member of that family.
14380 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
14381 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
14382 @samp{-march} ones described above.
14386 Equivalent to @samp{-march=mips1}.
14390 Equivalent to @samp{-march=mips2}.
14394 Equivalent to @samp{-march=mips3}.
14398 Equivalent to @samp{-march=mips4}.
14402 Equivalent to @samp{-march=mips32}.
14406 Equivalent to @samp{-march=mips32r2}.
14410 Equivalent to @samp{-march=mips64}.
14414 Equivalent to @samp{-march=mips64r2}.
14419 @opindex mno-mips16
14420 Generate (do not generate) MIPS16 code. If GCC is targetting a
14421 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
14423 MIPS16 code generation can also be controlled on a per-function basis
14424 by means of @code{mips16} and @code{nomips16} attributes.
14425 @xref{Function Attributes}, for more information.
14427 @item -mflip-mips16
14428 @opindex mflip-mips16
14429 Generate MIPS16 code on alternating functions. This option is provided
14430 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
14431 not intended for ordinary use in compiling user code.
14433 @item -minterlink-mips16
14434 @itemx -mno-interlink-mips16
14435 @opindex minterlink-mips16
14436 @opindex mno-interlink-mips16
14437 Require (do not require) that non-MIPS16 code be link-compatible with
14440 For example, non-MIPS16 code cannot jump directly to MIPS16 code;
14441 it must either use a call or an indirect jump. @option{-minterlink-mips16}
14442 therefore disables direct jumps unless GCC knows that the target of the
14443 jump is not MIPS16.
14455 Generate code for the given ABI@.
14457 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
14458 generates 64-bit code when you select a 64-bit architecture, but you
14459 can use @option{-mgp32} to get 32-bit code instead.
14461 For information about the O64 ABI, see
14462 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
14464 GCC supports a variant of the o32 ABI in which floating-point registers
14465 are 64 rather than 32 bits wide. You can select this combination with
14466 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
14467 and @samp{mfhc1} instructions and is therefore only supported for
14468 MIPS32R2 processors.
14470 The register assignments for arguments and return values remain the
14471 same, but each scalar value is passed in a single 64-bit register
14472 rather than a pair of 32-bit registers. For example, scalar
14473 floating-point values are returned in @samp{$f0} only, not a
14474 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
14475 remains the same, but all 64 bits are saved.
14478 @itemx -mno-abicalls
14480 @opindex mno-abicalls
14481 Generate (do not generate) code that is suitable for SVR4-style
14482 dynamic objects. @option{-mabicalls} is the default for SVR4-based
14487 Generate (do not generate) code that is fully position-independent,
14488 and that can therefore be linked into shared libraries. This option
14489 only affects @option{-mabicalls}.
14491 All @option{-mabicalls} code has traditionally been position-independent,
14492 regardless of options like @option{-fPIC} and @option{-fpic}. However,
14493 as an extension, the GNU toolchain allows executables to use absolute
14494 accesses for locally-binding symbols. It can also use shorter GP
14495 initialization sequences and generate direct calls to locally-defined
14496 functions. This mode is selected by @option{-mno-shared}.
14498 @option{-mno-shared} depends on binutils 2.16 or higher and generates
14499 objects that can only be linked by the GNU linker. However, the option
14500 does not affect the ABI of the final executable; it only affects the ABI
14501 of relocatable objects. Using @option{-mno-shared} will generally make
14502 executables both smaller and quicker.
14504 @option{-mshared} is the default.
14510 Assume (do not assume) that the static and dynamic linkers
14511 support PLTs and copy relocations. This option only affects
14512 @samp{-mno-shared -mabicalls}. For the n64 ABI, this option
14513 has no effect without @samp{-msym32}.
14515 You can make @option{-mplt} the default by configuring
14516 GCC with @option{--with-mips-plt}. The default is
14517 @option{-mno-plt} otherwise.
14523 Lift (do not lift) the usual restrictions on the size of the global
14526 GCC normally uses a single instruction to load values from the GOT@.
14527 While this is relatively efficient, it will only work if the GOT
14528 is smaller than about 64k. Anything larger will cause the linker
14529 to report an error such as:
14531 @cindex relocation truncated to fit (MIPS)
14533 relocation truncated to fit: R_MIPS_GOT16 foobar
14536 If this happens, you should recompile your code with @option{-mxgot}.
14537 It should then work with very large GOTs, although it will also be
14538 less efficient, since it will take three instructions to fetch the
14539 value of a global symbol.
14541 Note that some linkers can create multiple GOTs. If you have such a
14542 linker, you should only need to use @option{-mxgot} when a single object
14543 file accesses more than 64k's worth of GOT entries. Very few do.
14545 These options have no effect unless GCC is generating position
14550 Assume that general-purpose registers are 32 bits wide.
14554 Assume that general-purpose registers are 64 bits wide.
14558 Assume that floating-point registers are 32 bits wide.
14562 Assume that floating-point registers are 64 bits wide.
14565 @opindex mhard-float
14566 Use floating-point coprocessor instructions.
14569 @opindex msoft-float
14570 Do not use floating-point coprocessor instructions. Implement
14571 floating-point calculations using library calls instead.
14573 @item -msingle-float
14574 @opindex msingle-float
14575 Assume that the floating-point coprocessor only supports single-precision
14578 @item -mdouble-float
14579 @opindex mdouble-float
14580 Assume that the floating-point coprocessor supports double-precision
14581 operations. This is the default.
14587 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
14588 implement atomic memory built-in functions. When neither option is
14589 specified, GCC will use the instructions if the target architecture
14592 @option{-mllsc} is useful if the runtime environment can emulate the
14593 instructions and @option{-mno-llsc} can be useful when compiling for
14594 nonstandard ISAs. You can make either option the default by
14595 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
14596 respectively. @option{--with-llsc} is the default for some
14597 configurations; see the installation documentation for details.
14603 Use (do not use) revision 1 of the MIPS DSP ASE@.
14604 @xref{MIPS DSP Built-in Functions}. This option defines the
14605 preprocessor macro @samp{__mips_dsp}. It also defines
14606 @samp{__mips_dsp_rev} to 1.
14612 Use (do not use) revision 2 of the MIPS DSP ASE@.
14613 @xref{MIPS DSP Built-in Functions}. This option defines the
14614 preprocessor macros @samp{__mips_dsp} and @samp{__mips_dspr2}.
14615 It also defines @samp{__mips_dsp_rev} to 2.
14618 @itemx -mno-smartmips
14619 @opindex msmartmips
14620 @opindex mno-smartmips
14621 Use (do not use) the MIPS SmartMIPS ASE.
14623 @item -mpaired-single
14624 @itemx -mno-paired-single
14625 @opindex mpaired-single
14626 @opindex mno-paired-single
14627 Use (do not use) paired-single floating-point instructions.
14628 @xref{MIPS Paired-Single Support}. This option requires
14629 hardware floating-point support to be enabled.
14635 Use (do not use) MIPS Digital Media Extension instructions.
14636 This option can only be used when generating 64-bit code and requires
14637 hardware floating-point support to be enabled.
14642 @opindex mno-mips3d
14643 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
14644 The option @option{-mips3d} implies @option{-mpaired-single}.
14650 Use (do not use) MT Multithreading instructions.
14654 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
14655 an explanation of the default and the way that the pointer size is
14660 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
14662 The default size of @code{int}s, @code{long}s and pointers depends on
14663 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
14664 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
14665 32-bit @code{long}s. Pointers are the same size as @code{long}s,
14666 or the same size as integer registers, whichever is smaller.
14672 Assume (do not assume) that all symbols have 32-bit values, regardless
14673 of the selected ABI@. This option is useful in combination with
14674 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
14675 to generate shorter and faster references to symbolic addresses.
14679 Put definitions of externally-visible data in a small data section
14680 if that data is no bigger than @var{num} bytes. GCC can then access
14681 the data more efficiently; see @option{-mgpopt} for details.
14683 The default @option{-G} option depends on the configuration.
14685 @item -mlocal-sdata
14686 @itemx -mno-local-sdata
14687 @opindex mlocal-sdata
14688 @opindex mno-local-sdata
14689 Extend (do not extend) the @option{-G} behavior to local data too,
14690 such as to static variables in C@. @option{-mlocal-sdata} is the
14691 default for all configurations.
14693 If the linker complains that an application is using too much small data,
14694 you might want to try rebuilding the less performance-critical parts with
14695 @option{-mno-local-sdata}. You might also want to build large
14696 libraries with @option{-mno-local-sdata}, so that the libraries leave
14697 more room for the main program.
14699 @item -mextern-sdata
14700 @itemx -mno-extern-sdata
14701 @opindex mextern-sdata
14702 @opindex mno-extern-sdata
14703 Assume (do not assume) that externally-defined data will be in
14704 a small data section if that data is within the @option{-G} limit.
14705 @option{-mextern-sdata} is the default for all configurations.
14707 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
14708 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
14709 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
14710 is placed in a small data section. If @var{Var} is defined by another
14711 module, you must either compile that module with a high-enough
14712 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
14713 definition. If @var{Var} is common, you must link the application
14714 with a high-enough @option{-G} setting.
14716 The easiest way of satisfying these restrictions is to compile
14717 and link every module with the same @option{-G} option. However,
14718 you may wish to build a library that supports several different
14719 small data limits. You can do this by compiling the library with
14720 the highest supported @option{-G} setting and additionally using
14721 @option{-mno-extern-sdata} to stop the library from making assumptions
14722 about externally-defined data.
14728 Use (do not use) GP-relative accesses for symbols that are known to be
14729 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
14730 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
14733 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
14734 might not hold the value of @code{_gp}. For example, if the code is
14735 part of a library that might be used in a boot monitor, programs that
14736 call boot monitor routines will pass an unknown value in @code{$gp}.
14737 (In such situations, the boot monitor itself would usually be compiled
14738 with @option{-G0}.)
14740 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
14741 @option{-mno-extern-sdata}.
14743 @item -membedded-data
14744 @itemx -mno-embedded-data
14745 @opindex membedded-data
14746 @opindex mno-embedded-data
14747 Allocate variables to the read-only data section first if possible, then
14748 next in the small data section if possible, otherwise in data. This gives
14749 slightly slower code than the default, but reduces the amount of RAM required
14750 when executing, and thus may be preferred for some embedded systems.
14752 @item -muninit-const-in-rodata
14753 @itemx -mno-uninit-const-in-rodata
14754 @opindex muninit-const-in-rodata
14755 @opindex mno-uninit-const-in-rodata
14756 Put uninitialized @code{const} variables in the read-only data section.
14757 This option is only meaningful in conjunction with @option{-membedded-data}.
14759 @item -mcode-readable=@var{setting}
14760 @opindex mcode-readable
14761 Specify whether GCC may generate code that reads from executable sections.
14762 There are three possible settings:
14765 @item -mcode-readable=yes
14766 Instructions may freely access executable sections. This is the
14769 @item -mcode-readable=pcrel
14770 MIPS16 PC-relative load instructions can access executable sections,
14771 but other instructions must not do so. This option is useful on 4KSc
14772 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
14773 It is also useful on processors that can be configured to have a dual
14774 instruction/data SRAM interface and that, like the M4K, automatically
14775 redirect PC-relative loads to the instruction RAM.
14777 @item -mcode-readable=no
14778 Instructions must not access executable sections. This option can be
14779 useful on targets that are configured to have a dual instruction/data
14780 SRAM interface but that (unlike the M4K) do not automatically redirect
14781 PC-relative loads to the instruction RAM.
14784 @item -msplit-addresses
14785 @itemx -mno-split-addresses
14786 @opindex msplit-addresses
14787 @opindex mno-split-addresses
14788 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
14789 relocation operators. This option has been superseded by
14790 @option{-mexplicit-relocs} but is retained for backwards compatibility.
14792 @item -mexplicit-relocs
14793 @itemx -mno-explicit-relocs
14794 @opindex mexplicit-relocs
14795 @opindex mno-explicit-relocs
14796 Use (do not use) assembler relocation operators when dealing with symbolic
14797 addresses. The alternative, selected by @option{-mno-explicit-relocs},
14798 is to use assembler macros instead.
14800 @option{-mexplicit-relocs} is the default if GCC was configured
14801 to use an assembler that supports relocation operators.
14803 @item -mcheck-zero-division
14804 @itemx -mno-check-zero-division
14805 @opindex mcheck-zero-division
14806 @opindex mno-check-zero-division
14807 Trap (do not trap) on integer division by zero.
14809 The default is @option{-mcheck-zero-division}.
14811 @item -mdivide-traps
14812 @itemx -mdivide-breaks
14813 @opindex mdivide-traps
14814 @opindex mdivide-breaks
14815 MIPS systems check for division by zero by generating either a
14816 conditional trap or a break instruction. Using traps results in
14817 smaller code, but is only supported on MIPS II and later. Also, some
14818 versions of the Linux kernel have a bug that prevents trap from
14819 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
14820 allow conditional traps on architectures that support them and
14821 @option{-mdivide-breaks} to force the use of breaks.
14823 The default is usually @option{-mdivide-traps}, but this can be
14824 overridden at configure time using @option{--with-divide=breaks}.
14825 Divide-by-zero checks can be completely disabled using
14826 @option{-mno-check-zero-division}.
14831 @opindex mno-memcpy
14832 Force (do not force) the use of @code{memcpy()} for non-trivial block
14833 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
14834 most constant-sized copies.
14837 @itemx -mno-long-calls
14838 @opindex mlong-calls
14839 @opindex mno-long-calls
14840 Disable (do not disable) use of the @code{jal} instruction. Calling
14841 functions using @code{jal} is more efficient but requires the caller
14842 and callee to be in the same 256 megabyte segment.
14844 This option has no effect on abicalls code. The default is
14845 @option{-mno-long-calls}.
14851 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
14852 instructions, as provided by the R4650 ISA@.
14855 @itemx -mno-fused-madd
14856 @opindex mfused-madd
14857 @opindex mno-fused-madd
14858 Enable (disable) use of the floating point multiply-accumulate
14859 instructions, when they are available. The default is
14860 @option{-mfused-madd}.
14862 When multiply-accumulate instructions are used, the intermediate
14863 product is calculated to infinite precision and is not subject to
14864 the FCSR Flush to Zero bit. This may be undesirable in some
14869 Tell the MIPS assembler to not run its preprocessor over user
14870 assembler files (with a @samp{.s} suffix) when assembling them.
14875 @opindex mno-fix-24k
14876 Work around the 24K E48 (lost data on stores during refill) errata.
14877 The workarounds are implemented by the assembler rather than by GCC.
14880 @itemx -mno-fix-r4000
14881 @opindex mfix-r4000
14882 @opindex mno-fix-r4000
14883 Work around certain R4000 CPU errata:
14886 A double-word or a variable shift may give an incorrect result if executed
14887 immediately after starting an integer division.
14889 A double-word or a variable shift may give an incorrect result if executed
14890 while an integer multiplication is in progress.
14892 An integer division may give an incorrect result if started in a delay slot
14893 of a taken branch or a jump.
14897 @itemx -mno-fix-r4400
14898 @opindex mfix-r4400
14899 @opindex mno-fix-r4400
14900 Work around certain R4400 CPU errata:
14903 A double-word or a variable shift may give an incorrect result if executed
14904 immediately after starting an integer division.
14908 @itemx -mno-fix-r10000
14909 @opindex mfix-r10000
14910 @opindex mno-fix-r10000
14911 Work around certain R10000 errata:
14914 @code{ll}/@code{sc} sequences may not behave atomically on revisions
14915 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
14918 This option can only be used if the target architecture supports
14919 branch-likely instructions. @option{-mfix-r10000} is the default when
14920 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
14924 @itemx -mno-fix-vr4120
14925 @opindex mfix-vr4120
14926 Work around certain VR4120 errata:
14929 @code{dmultu} does not always produce the correct result.
14931 @code{div} and @code{ddiv} do not always produce the correct result if one
14932 of the operands is negative.
14934 The workarounds for the division errata rely on special functions in
14935 @file{libgcc.a}. At present, these functions are only provided by
14936 the @code{mips64vr*-elf} configurations.
14938 Other VR4120 errata require a nop to be inserted between certain pairs of
14939 instructions. These errata are handled by the assembler, not by GCC itself.
14942 @opindex mfix-vr4130
14943 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
14944 workarounds are implemented by the assembler rather than by GCC,
14945 although GCC will avoid using @code{mflo} and @code{mfhi} if the
14946 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
14947 instructions are available instead.
14950 @itemx -mno-fix-sb1
14952 Work around certain SB-1 CPU core errata.
14953 (This flag currently works around the SB-1 revision 2
14954 ``F1'' and ``F2'' floating point errata.)
14956 @item -mr10k-cache-barrier=@var{setting}
14957 @opindex mr10k-cache-barrier
14958 Specify whether GCC should insert cache barriers to avoid the
14959 side-effects of speculation on R10K processors.
14961 In common with many processors, the R10K tries to predict the outcome
14962 of a conditional branch and speculatively executes instructions from
14963 the ``taken'' branch. It later aborts these instructions if the
14964 predicted outcome was wrong. However, on the R10K, even aborted
14965 instructions can have side effects.
14967 This problem only affects kernel stores and, depending on the system,
14968 kernel loads. As an example, a speculatively-executed store may load
14969 the target memory into cache and mark the cache line as dirty, even if
14970 the store itself is later aborted. If a DMA operation writes to the
14971 same area of memory before the ``dirty'' line is flushed, the cached
14972 data will overwrite the DMA-ed data. See the R10K processor manual
14973 for a full description, including other potential problems.
14975 One workaround is to insert cache barrier instructions before every memory
14976 access that might be speculatively executed and that might have side
14977 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
14978 controls GCC's implementation of this workaround. It assumes that
14979 aborted accesses to any byte in the following regions will not have
14984 the memory occupied by the current function's stack frame;
14987 the memory occupied by an incoming stack argument;
14990 the memory occupied by an object with a link-time-constant address.
14993 It is the kernel's responsibility to ensure that speculative
14994 accesses to these regions are indeed safe.
14996 If the input program contains a function declaration such as:
15002 then the implementation of @code{foo} must allow @code{j foo} and
15003 @code{jal foo} to be executed speculatively. GCC honors this
15004 restriction for functions it compiles itself. It expects non-GCC
15005 functions (such as hand-written assembly code) to do the same.
15007 The option has three forms:
15010 @item -mr10k-cache-barrier=load-store
15011 Insert a cache barrier before a load or store that might be
15012 speculatively executed and that might have side effects even
15015 @item -mr10k-cache-barrier=store
15016 Insert a cache barrier before a store that might be speculatively
15017 executed and that might have side effects even if aborted.
15019 @item -mr10k-cache-barrier=none
15020 Disable the insertion of cache barriers. This is the default setting.
15023 @item -mflush-func=@var{func}
15024 @itemx -mno-flush-func
15025 @opindex mflush-func
15026 Specifies the function to call to flush the I and D caches, or to not
15027 call any such function. If called, the function must take the same
15028 arguments as the common @code{_flush_func()}, that is, the address of the
15029 memory range for which the cache is being flushed, the size of the
15030 memory range, and the number 3 (to flush both caches). The default
15031 depends on the target GCC was configured for, but commonly is either
15032 @samp{_flush_func} or @samp{__cpu_flush}.
15034 @item mbranch-cost=@var{num}
15035 @opindex mbranch-cost
15036 Set the cost of branches to roughly @var{num} ``simple'' instructions.
15037 This cost is only a heuristic and is not guaranteed to produce
15038 consistent results across releases. A zero cost redundantly selects
15039 the default, which is based on the @option{-mtune} setting.
15041 @item -mbranch-likely
15042 @itemx -mno-branch-likely
15043 @opindex mbranch-likely
15044 @opindex mno-branch-likely
15045 Enable or disable use of Branch Likely instructions, regardless of the
15046 default for the selected architecture. By default, Branch Likely
15047 instructions may be generated if they are supported by the selected
15048 architecture. An exception is for the MIPS32 and MIPS64 architectures
15049 and processors which implement those architectures; for those, Branch
15050 Likely instructions will not be generated by default because the MIPS32
15051 and MIPS64 architectures specifically deprecate their use.
15053 @item -mfp-exceptions
15054 @itemx -mno-fp-exceptions
15055 @opindex mfp-exceptions
15056 Specifies whether FP exceptions are enabled. This affects how we schedule
15057 FP instructions for some processors. The default is that FP exceptions are
15060 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
15061 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
15064 @item -mvr4130-align
15065 @itemx -mno-vr4130-align
15066 @opindex mvr4130-align
15067 The VR4130 pipeline is two-way superscalar, but can only issue two
15068 instructions together if the first one is 8-byte aligned. When this
15069 option is enabled, GCC will align pairs of instructions that it
15070 thinks should execute in parallel.
15072 This option only has an effect when optimizing for the VR4130.
15073 It normally makes code faster, but at the expense of making it bigger.
15074 It is enabled by default at optimization level @option{-O3}.
15079 Enable (disable) generation of @code{synci} instructions on
15080 architectures that support it. The @code{synci} instructions (if
15081 enabled) will be generated when @code{__builtin___clear_cache()} is
15084 This option defaults to @code{-mno-synci}, but the default can be
15085 overridden by configuring with @code{--with-synci}.
15087 When compiling code for single processor systems, it is generally safe
15088 to use @code{synci}. However, on many multi-core (SMP) systems, it
15089 will not invalidate the instruction caches on all cores and may lead
15090 to undefined behavior.
15092 @item -mrelax-pic-calls
15093 @itemx -mno-relax-pic-calls
15094 @opindex mrelax-pic-calls
15095 Try to turn PIC calls that are normally dispatched via register
15096 @code{$25} into direct calls. This is only possible if the linker can
15097 resolve the destination at link-time and if the destination is within
15098 range for a direct call.
15100 @option{-mrelax-pic-calls} is the default if GCC was configured to use
15101 an assembler and a linker that supports the @code{.reloc} assembly
15102 directive and @code{-mexplicit-relocs} is in effect. With
15103 @code{-mno-explicit-relocs}, this optimization can be performed by the
15104 assembler and the linker alone without help from the compiler.
15106 @item -mmcount-ra-address
15107 @itemx -mno-mcount-ra-address
15108 @opindex mmcount-ra-address
15109 @opindex mno-mcount-ra-address
15110 Emit (do not emit) code that allows @code{_mcount} to modify the
15111 calling function's return address. When enabled, this option extends
15112 the usual @code{_mcount} interface with a new @var{ra-address}
15113 parameter, which has type @code{intptr_t *} and is passed in register
15114 @code{$12}. @code{_mcount} can then modify the return address by
15115 doing both of the following:
15118 Returning the new address in register @code{$31}.
15120 Storing the new address in @code{*@var{ra-address}},
15121 if @var{ra-address} is nonnull.
15124 The default is @option{-mno-mcount-ra-address}.
15129 @subsection MMIX Options
15130 @cindex MMIX Options
15132 These options are defined for the MMIX:
15136 @itemx -mno-libfuncs
15138 @opindex mno-libfuncs
15139 Specify that intrinsic library functions are being compiled, passing all
15140 values in registers, no matter the size.
15143 @itemx -mno-epsilon
15145 @opindex mno-epsilon
15146 Generate floating-point comparison instructions that compare with respect
15147 to the @code{rE} epsilon register.
15149 @item -mabi=mmixware
15151 @opindex mabi=mmixware
15153 Generate code that passes function parameters and return values that (in
15154 the called function) are seen as registers @code{$0} and up, as opposed to
15155 the GNU ABI which uses global registers @code{$231} and up.
15157 @item -mzero-extend
15158 @itemx -mno-zero-extend
15159 @opindex mzero-extend
15160 @opindex mno-zero-extend
15161 When reading data from memory in sizes shorter than 64 bits, use (do not
15162 use) zero-extending load instructions by default, rather than
15163 sign-extending ones.
15166 @itemx -mno-knuthdiv
15168 @opindex mno-knuthdiv
15169 Make the result of a division yielding a remainder have the same sign as
15170 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
15171 remainder follows the sign of the dividend. Both methods are
15172 arithmetically valid, the latter being almost exclusively used.
15174 @item -mtoplevel-symbols
15175 @itemx -mno-toplevel-symbols
15176 @opindex mtoplevel-symbols
15177 @opindex mno-toplevel-symbols
15178 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
15179 code can be used with the @code{PREFIX} assembly directive.
15183 Generate an executable in the ELF format, rather than the default
15184 @samp{mmo} format used by the @command{mmix} simulator.
15186 @item -mbranch-predict
15187 @itemx -mno-branch-predict
15188 @opindex mbranch-predict
15189 @opindex mno-branch-predict
15190 Use (do not use) the probable-branch instructions, when static branch
15191 prediction indicates a probable branch.
15193 @item -mbase-addresses
15194 @itemx -mno-base-addresses
15195 @opindex mbase-addresses
15196 @opindex mno-base-addresses
15197 Generate (do not generate) code that uses @emph{base addresses}. Using a
15198 base address automatically generates a request (handled by the assembler
15199 and the linker) for a constant to be set up in a global register. The
15200 register is used for one or more base address requests within the range 0
15201 to 255 from the value held in the register. The generally leads to short
15202 and fast code, but the number of different data items that can be
15203 addressed is limited. This means that a program that uses lots of static
15204 data may require @option{-mno-base-addresses}.
15206 @item -msingle-exit
15207 @itemx -mno-single-exit
15208 @opindex msingle-exit
15209 @opindex mno-single-exit
15210 Force (do not force) generated code to have a single exit point in each
15214 @node MN10300 Options
15215 @subsection MN10300 Options
15216 @cindex MN10300 options
15218 These @option{-m} options are defined for Matsushita MN10300 architectures:
15223 Generate code to avoid bugs in the multiply instructions for the MN10300
15224 processors. This is the default.
15226 @item -mno-mult-bug
15227 @opindex mno-mult-bug
15228 Do not generate code to avoid bugs in the multiply instructions for the
15229 MN10300 processors.
15233 Generate code which uses features specific to the AM33 processor.
15237 Do not generate code which uses features specific to the AM33 processor. This
15242 Generate code which uses features specific to the AM33/2.0 processor.
15246 Generate code which uses features specific to the AM34 processor.
15248 @item -mtune=@var{cpu-type}
15250 Use the timing characteristics of the indicated CPU type when
15251 scheduling instructions. This does not change the targeted processor
15252 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
15253 @samp{am33-2} or @samp{am34}.
15255 @item -mreturn-pointer-on-d0
15256 @opindex mreturn-pointer-on-d0
15257 When generating a function which returns a pointer, return the pointer
15258 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
15259 only in a0, and attempts to call such functions without a prototype
15260 would result in errors. Note that this option is on by default; use
15261 @option{-mno-return-pointer-on-d0} to disable it.
15265 Do not link in the C run-time initialization object file.
15269 Indicate to the linker that it should perform a relaxation optimization pass
15270 to shorten branches, calls and absolute memory addresses. This option only
15271 has an effect when used on the command line for the final link step.
15273 This option makes symbolic debugging impossible.
15277 Allow the compiler to generate @emph{Long Instruction Word}
15278 instructions if the target is the @samp{AM33} or later. This is the
15279 default. This option defines the preprocessor macro @samp{__LIW__}.
15283 Do not allow the compiler to generate @emph{Long Instruction Word}
15284 instructions. This option defines the preprocessor macro
15289 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
15290 instructions if the target is the @samp{AM33} or later. This is the
15291 default. This option defines the preprocessor macro @samp{__SETLB__}.
15295 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
15296 instructions. This option defines the preprocessor macro
15297 @samp{__NO_SETLB__}.
15301 @node PDP-11 Options
15302 @subsection PDP-11 Options
15303 @cindex PDP-11 Options
15305 These options are defined for the PDP-11:
15310 Use hardware FPP floating point. This is the default. (FIS floating
15311 point on the PDP-11/40 is not supported.)
15314 @opindex msoft-float
15315 Do not use hardware floating point.
15319 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
15323 Return floating-point results in memory. This is the default.
15327 Generate code for a PDP-11/40.
15331 Generate code for a PDP-11/45. This is the default.
15335 Generate code for a PDP-11/10.
15337 @item -mbcopy-builtin
15338 @opindex mbcopy-builtin
15339 Use inline @code{movmemhi} patterns for copying memory. This is the
15344 Do not use inline @code{movmemhi} patterns for copying memory.
15350 Use 16-bit @code{int}. This is the default.
15356 Use 32-bit @code{int}.
15359 @itemx -mno-float32
15361 @opindex mno-float32
15362 Use 64-bit @code{float}. This is the default.
15365 @itemx -mno-float64
15367 @opindex mno-float64
15368 Use 32-bit @code{float}.
15372 Use @code{abshi2} pattern. This is the default.
15376 Do not use @code{abshi2} pattern.
15378 @item -mbranch-expensive
15379 @opindex mbranch-expensive
15380 Pretend that branches are expensive. This is for experimenting with
15381 code generation only.
15383 @item -mbranch-cheap
15384 @opindex mbranch-cheap
15385 Do not pretend that branches are expensive. This is the default.
15389 Use Unix assembler syntax. This is the default when configured for
15390 @samp{pdp11-*-bsd}.
15394 Use DEC assembler syntax. This is the default when configured for any
15395 PDP-11 target other than @samp{pdp11-*-bsd}.
15398 @node picoChip Options
15399 @subsection picoChip Options
15400 @cindex picoChip options
15402 These @samp{-m} options are defined for picoChip implementations:
15406 @item -mae=@var{ae_type}
15408 Set the instruction set, register set, and instruction scheduling
15409 parameters for array element type @var{ae_type}. Supported values
15410 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
15412 @option{-mae=ANY} selects a completely generic AE type. Code
15413 generated with this option will run on any of the other AE types. The
15414 code will not be as efficient as it would be if compiled for a specific
15415 AE type, and some types of operation (e.g., multiplication) will not
15416 work properly on all types of AE.
15418 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
15419 for compiled code, and is the default.
15421 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
15422 option may suffer from poor performance of byte (char) manipulation,
15423 since the DSP AE does not provide hardware support for byte load/stores.
15425 @item -msymbol-as-address
15426 Enable the compiler to directly use a symbol name as an address in a
15427 load/store instruction, without first loading it into a
15428 register. Typically, the use of this option will generate larger
15429 programs, which run faster than when the option isn't used. However, the
15430 results vary from program to program, so it is left as a user option,
15431 rather than being permanently enabled.
15433 @item -mno-inefficient-warnings
15434 Disables warnings about the generation of inefficient code. These
15435 warnings can be generated, for example, when compiling code which
15436 performs byte-level memory operations on the MAC AE type. The MAC AE has
15437 no hardware support for byte-level memory operations, so all byte
15438 load/stores must be synthesized from word load/store operations. This is
15439 inefficient and a warning will be generated indicating to the programmer
15440 that they should rewrite the code to avoid byte operations, or to target
15441 an AE type which has the necessary hardware support. This option enables
15442 the warning to be turned off.
15446 @node PowerPC Options
15447 @subsection PowerPC Options
15448 @cindex PowerPC options
15450 These are listed under @xref{RS/6000 and PowerPC Options}.
15452 @node RS/6000 and PowerPC Options
15453 @subsection IBM RS/6000 and PowerPC Options
15454 @cindex RS/6000 and PowerPC Options
15455 @cindex IBM RS/6000 and PowerPC Options
15457 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
15464 @itemx -mno-powerpc
15465 @itemx -mpowerpc-gpopt
15466 @itemx -mno-powerpc-gpopt
15467 @itemx -mpowerpc-gfxopt
15468 @itemx -mno-powerpc-gfxopt
15471 @itemx -mno-powerpc64
15475 @itemx -mno-popcntb
15477 @itemx -mno-popcntd
15486 @itemx -mno-hard-dfp
15490 @opindex mno-power2
15492 @opindex mno-powerpc
15493 @opindex mpowerpc-gpopt
15494 @opindex mno-powerpc-gpopt
15495 @opindex mpowerpc-gfxopt
15496 @opindex mno-powerpc-gfxopt
15497 @opindex mpowerpc64
15498 @opindex mno-powerpc64
15502 @opindex mno-popcntb
15504 @opindex mno-popcntd
15510 @opindex mno-mfpgpr
15512 @opindex mno-hard-dfp
15513 GCC supports two related instruction set architectures for the
15514 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
15515 instructions supported by the @samp{rios} chip set used in the original
15516 RS/6000 systems and the @dfn{PowerPC} instruction set is the
15517 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
15518 the IBM 4xx, 6xx, and follow-on microprocessors.
15520 Neither architecture is a subset of the other. However there is a
15521 large common subset of instructions supported by both. An MQ
15522 register is included in processors supporting the POWER architecture.
15524 You use these options to specify which instructions are available on the
15525 processor you are using. The default value of these options is
15526 determined when configuring GCC@. Specifying the
15527 @option{-mcpu=@var{cpu_type}} overrides the specification of these
15528 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
15529 rather than the options listed above.
15531 The @option{-mpower} option allows GCC to generate instructions that
15532 are found only in the POWER architecture and to use the MQ register.
15533 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
15534 to generate instructions that are present in the POWER2 architecture but
15535 not the original POWER architecture.
15537 The @option{-mpowerpc} option allows GCC to generate instructions that
15538 are found only in the 32-bit subset of the PowerPC architecture.
15539 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
15540 GCC to use the optional PowerPC architecture instructions in the
15541 General Purpose group, including floating-point square root. Specifying
15542 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
15543 use the optional PowerPC architecture instructions in the Graphics
15544 group, including floating-point select.
15546 The @option{-mmfcrf} option allows GCC to generate the move from
15547 condition register field instruction implemented on the POWER4
15548 processor and other processors that support the PowerPC V2.01
15550 The @option{-mpopcntb} option allows GCC to generate the popcount and
15551 double precision FP reciprocal estimate instruction implemented on the
15552 POWER5 processor and other processors that support the PowerPC V2.02
15554 The @option{-mpopcntd} option allows GCC to generate the popcount
15555 instruction implemented on the POWER7 processor and other processors
15556 that support the PowerPC V2.06 architecture.
15557 The @option{-mfprnd} option allows GCC to generate the FP round to
15558 integer instructions implemented on the POWER5+ processor and other
15559 processors that support the PowerPC V2.03 architecture.
15560 The @option{-mcmpb} option allows GCC to generate the compare bytes
15561 instruction implemented on the POWER6 processor and other processors
15562 that support the PowerPC V2.05 architecture.
15563 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
15564 general purpose register instructions implemented on the POWER6X
15565 processor and other processors that support the extended PowerPC V2.05
15567 The @option{-mhard-dfp} option allows GCC to generate the decimal floating
15568 point instructions implemented on some POWER processors.
15570 The @option{-mpowerpc64} option allows GCC to generate the additional
15571 64-bit instructions that are found in the full PowerPC64 architecture
15572 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
15573 @option{-mno-powerpc64}.
15575 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
15576 will use only the instructions in the common subset of both
15577 architectures plus some special AIX common-mode calls, and will not use
15578 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
15579 permits GCC to use any instruction from either architecture and to
15580 allow use of the MQ register; specify this for the Motorola MPC601.
15582 @item -mnew-mnemonics
15583 @itemx -mold-mnemonics
15584 @opindex mnew-mnemonics
15585 @opindex mold-mnemonics
15586 Select which mnemonics to use in the generated assembler code. With
15587 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
15588 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
15589 assembler mnemonics defined for the POWER architecture. Instructions
15590 defined in only one architecture have only one mnemonic; GCC uses that
15591 mnemonic irrespective of which of these options is specified.
15593 GCC defaults to the mnemonics appropriate for the architecture in
15594 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
15595 value of these option. Unless you are building a cross-compiler, you
15596 should normally not specify either @option{-mnew-mnemonics} or
15597 @option{-mold-mnemonics}, but should instead accept the default.
15599 @item -mcpu=@var{cpu_type}
15601 Set architecture type, register usage, choice of mnemonics, and
15602 instruction scheduling parameters for machine type @var{cpu_type}.
15603 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
15604 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
15605 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
15606 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
15607 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
15608 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
15609 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{ec603e}, @samp{G3},
15610 @samp{G4}, @samp{G5}, @samp{titan}, @samp{power}, @samp{power2}, @samp{power3},
15611 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6}, @samp{power6x},
15612 @samp{power7}, @samp{common}, @samp{powerpc}, @samp{powerpc64}, @samp{rios},
15613 @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
15615 @option{-mcpu=common} selects a completely generic processor. Code
15616 generated under this option will run on any POWER or PowerPC processor.
15617 GCC will use only the instructions in the common subset of both
15618 architectures, and will not use the MQ register. GCC assumes a generic
15619 processor model for scheduling purposes.
15621 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
15622 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
15623 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
15624 types, with an appropriate, generic processor model assumed for
15625 scheduling purposes.
15627 The other options specify a specific processor. Code generated under
15628 those options will run best on that processor, and may not run at all on
15631 The @option{-mcpu} options automatically enable or disable the
15634 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
15635 -mnew-mnemonics -mpopcntb -mpopcntd -mpower -mpower2 -mpowerpc64 @gol
15636 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
15637 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx}
15639 The particular options set for any particular CPU will vary between
15640 compiler versions, depending on what setting seems to produce optimal
15641 code for that CPU; it doesn't necessarily reflect the actual hardware's
15642 capabilities. If you wish to set an individual option to a particular
15643 value, you may specify it after the @option{-mcpu} option, like
15644 @samp{-mcpu=970 -mno-altivec}.
15646 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
15647 not enabled or disabled by the @option{-mcpu} option at present because
15648 AIX does not have full support for these options. You may still
15649 enable or disable them individually if you're sure it'll work in your
15652 @item -mtune=@var{cpu_type}
15654 Set the instruction scheduling parameters for machine type
15655 @var{cpu_type}, but do not set the architecture type, register usage, or
15656 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
15657 values for @var{cpu_type} are used for @option{-mtune} as for
15658 @option{-mcpu}. If both are specified, the code generated will use the
15659 architecture, registers, and mnemonics set by @option{-mcpu}, but the
15660 scheduling parameters set by @option{-mtune}.
15662 @item -mcmodel=small
15663 @opindex mcmodel=small
15664 Generate PowerPC64 code for the small model: The TOC is limited to
15667 @item -mcmodel=medium
15668 @opindex mcmodel=medium
15669 Generate PowerPC64 code for the medium model: The TOC and other static
15670 data may be up to a total of 4G in size.
15672 @item -mcmodel=large
15673 @opindex mcmodel=large
15674 Generate PowerPC64 code for the large model: The TOC may be up to 4G
15675 in size. Other data and code is only limited by the 64-bit address
15679 @itemx -mno-altivec
15681 @opindex mno-altivec
15682 Generate code that uses (does not use) AltiVec instructions, and also
15683 enable the use of built-in functions that allow more direct access to
15684 the AltiVec instruction set. You may also need to set
15685 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
15691 @opindex mno-vrsave
15692 Generate VRSAVE instructions when generating AltiVec code.
15694 @item -mgen-cell-microcode
15695 @opindex mgen-cell-microcode
15696 Generate Cell microcode instructions
15698 @item -mwarn-cell-microcode
15699 @opindex mwarn-cell-microcode
15700 Warning when a Cell microcode instruction is going to emitted. An example
15701 of a Cell microcode instruction is a variable shift.
15704 @opindex msecure-plt
15705 Generate code that allows ld and ld.so to build executables and shared
15706 libraries with non-exec .plt and .got sections. This is a PowerPC
15707 32-bit SYSV ABI option.
15711 Generate code that uses a BSS .plt section that ld.so fills in, and
15712 requires .plt and .got sections that are both writable and executable.
15713 This is a PowerPC 32-bit SYSV ABI option.
15719 This switch enables or disables the generation of ISEL instructions.
15721 @item -misel=@var{yes/no}
15722 This switch has been deprecated. Use @option{-misel} and
15723 @option{-mno-isel} instead.
15729 This switch enables or disables the generation of SPE simd
15735 @opindex mno-paired
15736 This switch enables or disables the generation of PAIRED simd
15739 @item -mspe=@var{yes/no}
15740 This option has been deprecated. Use @option{-mspe} and
15741 @option{-mno-spe} instead.
15747 Generate code that uses (does not use) vector/scalar (VSX)
15748 instructions, and also enable the use of built-in functions that allow
15749 more direct access to the VSX instruction set.
15751 @item -mfloat-gprs=@var{yes/single/double/no}
15752 @itemx -mfloat-gprs
15753 @opindex mfloat-gprs
15754 This switch enables or disables the generation of floating point
15755 operations on the general purpose registers for architectures that
15758 The argument @var{yes} or @var{single} enables the use of
15759 single-precision floating point operations.
15761 The argument @var{double} enables the use of single and
15762 double-precision floating point operations.
15764 The argument @var{no} disables floating point operations on the
15765 general purpose registers.
15767 This option is currently only available on the MPC854x.
15773 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
15774 targets (including GNU/Linux). The 32-bit environment sets int, long
15775 and pointer to 32 bits and generates code that runs on any PowerPC
15776 variant. The 64-bit environment sets int to 32 bits and long and
15777 pointer to 64 bits, and generates code for PowerPC64, as for
15778 @option{-mpowerpc64}.
15781 @itemx -mno-fp-in-toc
15782 @itemx -mno-sum-in-toc
15783 @itemx -mminimal-toc
15785 @opindex mno-fp-in-toc
15786 @opindex mno-sum-in-toc
15787 @opindex mminimal-toc
15788 Modify generation of the TOC (Table Of Contents), which is created for
15789 every executable file. The @option{-mfull-toc} option is selected by
15790 default. In that case, GCC will allocate at least one TOC entry for
15791 each unique non-automatic variable reference in your program. GCC
15792 will also place floating-point constants in the TOC@. However, only
15793 16,384 entries are available in the TOC@.
15795 If you receive a linker error message that saying you have overflowed
15796 the available TOC space, you can reduce the amount of TOC space used
15797 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
15798 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
15799 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
15800 generate code to calculate the sum of an address and a constant at
15801 run-time instead of putting that sum into the TOC@. You may specify one
15802 or both of these options. Each causes GCC to produce very slightly
15803 slower and larger code at the expense of conserving TOC space.
15805 If you still run out of space in the TOC even when you specify both of
15806 these options, specify @option{-mminimal-toc} instead. This option causes
15807 GCC to make only one TOC entry for every file. When you specify this
15808 option, GCC will produce code that is slower and larger but which
15809 uses extremely little TOC space. You may wish to use this option
15810 only on files that contain less frequently executed code.
15816 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
15817 @code{long} type, and the infrastructure needed to support them.
15818 Specifying @option{-maix64} implies @option{-mpowerpc64} and
15819 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
15820 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
15823 @itemx -mno-xl-compat
15824 @opindex mxl-compat
15825 @opindex mno-xl-compat
15826 Produce code that conforms more closely to IBM XL compiler semantics
15827 when using AIX-compatible ABI@. Pass floating-point arguments to
15828 prototyped functions beyond the register save area (RSA) on the stack
15829 in addition to argument FPRs. Do not assume that most significant
15830 double in 128-bit long double value is properly rounded when comparing
15831 values and converting to double. Use XL symbol names for long double
15834 The AIX calling convention was extended but not initially documented to
15835 handle an obscure K&R C case of calling a function that takes the
15836 address of its arguments with fewer arguments than declared. IBM XL
15837 compilers access floating point arguments which do not fit in the
15838 RSA from the stack when a subroutine is compiled without
15839 optimization. Because always storing floating-point arguments on the
15840 stack is inefficient and rarely needed, this option is not enabled by
15841 default and only is necessary when calling subroutines compiled by IBM
15842 XL compilers without optimization.
15846 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
15847 application written to use message passing with special startup code to
15848 enable the application to run. The system must have PE installed in the
15849 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
15850 must be overridden with the @option{-specs=} option to specify the
15851 appropriate directory location. The Parallel Environment does not
15852 support threads, so the @option{-mpe} option and the @option{-pthread}
15853 option are incompatible.
15855 @item -malign-natural
15856 @itemx -malign-power
15857 @opindex malign-natural
15858 @opindex malign-power
15859 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
15860 @option{-malign-natural} overrides the ABI-defined alignment of larger
15861 types, such as floating-point doubles, on their natural size-based boundary.
15862 The option @option{-malign-power} instructs GCC to follow the ABI-specified
15863 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
15865 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
15869 @itemx -mhard-float
15870 @opindex msoft-float
15871 @opindex mhard-float
15872 Generate code that does not use (uses) the floating-point register set.
15873 Software floating point emulation is provided if you use the
15874 @option{-msoft-float} option, and pass the option to GCC when linking.
15876 @item -msingle-float
15877 @itemx -mdouble-float
15878 @opindex msingle-float
15879 @opindex mdouble-float
15880 Generate code for single or double-precision floating point operations.
15881 @option{-mdouble-float} implies @option{-msingle-float}.
15884 @opindex msimple-fpu
15885 Do not generate sqrt and div instructions for hardware floating point unit.
15889 Specify type of floating point unit. Valid values are @var{sp_lite}
15890 (equivalent to -msingle-float -msimple-fpu), @var{dp_lite} (equivalent
15891 to -mdouble-float -msimple-fpu), @var{sp_full} (equivalent to -msingle-float),
15892 and @var{dp_full} (equivalent to -mdouble-float).
15895 @opindex mxilinx-fpu
15896 Perform optimizations for floating point unit on Xilinx PPC 405/440.
15899 @itemx -mno-multiple
15901 @opindex mno-multiple
15902 Generate code that uses (does not use) the load multiple word
15903 instructions and the store multiple word instructions. These
15904 instructions are generated by default on POWER systems, and not
15905 generated on PowerPC systems. Do not use @option{-mmultiple} on little
15906 endian PowerPC systems, since those instructions do not work when the
15907 processor is in little endian mode. The exceptions are PPC740 and
15908 PPC750 which permit the instructions usage in little endian mode.
15913 @opindex mno-string
15914 Generate code that uses (does not use) the load string instructions
15915 and the store string word instructions to save multiple registers and
15916 do small block moves. These instructions are generated by default on
15917 POWER systems, and not generated on PowerPC systems. Do not use
15918 @option{-mstring} on little endian PowerPC systems, since those
15919 instructions do not work when the processor is in little endian mode.
15920 The exceptions are PPC740 and PPC750 which permit the instructions
15921 usage in little endian mode.
15926 @opindex mno-update
15927 Generate code that uses (does not use) the load or store instructions
15928 that update the base register to the address of the calculated memory
15929 location. These instructions are generated by default. If you use
15930 @option{-mno-update}, there is a small window between the time that the
15931 stack pointer is updated and the address of the previous frame is
15932 stored, which means code that walks the stack frame across interrupts or
15933 signals may get corrupted data.
15935 @item -mavoid-indexed-addresses
15936 @itemx -mno-avoid-indexed-addresses
15937 @opindex mavoid-indexed-addresses
15938 @opindex mno-avoid-indexed-addresses
15939 Generate code that tries to avoid (not avoid) the use of indexed load
15940 or store instructions. These instructions can incur a performance
15941 penalty on Power6 processors in certain situations, such as when
15942 stepping through large arrays that cross a 16M boundary. This option
15943 is enabled by default when targetting Power6 and disabled otherwise.
15946 @itemx -mno-fused-madd
15947 @opindex mfused-madd
15948 @opindex mno-fused-madd
15949 Generate code that uses (does not use) the floating point multiply and
15950 accumulate instructions. These instructions are generated by default
15951 if hardware floating point is used. The machine dependent
15952 @option{-mfused-madd} option is now mapped to the machine independent
15953 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
15954 mapped to @option{-ffp-contract=off}.
15960 Generate code that uses (does not use) the half-word multiply and
15961 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
15962 These instructions are generated by default when targetting those
15969 Generate code that uses (does not use) the string-search @samp{dlmzb}
15970 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
15971 generated by default when targetting those processors.
15973 @item -mno-bit-align
15975 @opindex mno-bit-align
15976 @opindex mbit-align
15977 On System V.4 and embedded PowerPC systems do not (do) force structures
15978 and unions that contain bit-fields to be aligned to the base type of the
15981 For example, by default a structure containing nothing but 8
15982 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
15983 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
15984 the structure would be aligned to a 1 byte boundary and be one byte in
15987 @item -mno-strict-align
15988 @itemx -mstrict-align
15989 @opindex mno-strict-align
15990 @opindex mstrict-align
15991 On System V.4 and embedded PowerPC systems do not (do) assume that
15992 unaligned memory references will be handled by the system.
15994 @item -mrelocatable
15995 @itemx -mno-relocatable
15996 @opindex mrelocatable
15997 @opindex mno-relocatable
15998 Generate code that allows (does not allow) a static executable to be
15999 relocated to a different address at runtime. A simple embedded
16000 PowerPC system loader should relocate the entire contents of
16001 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
16002 a table of 32-bit addresses generated by this option. For this to
16003 work, all objects linked together must be compiled with
16004 @option{-mrelocatable} or @option{-mrelocatable-lib}.
16005 @option{-mrelocatable} code aligns the stack to an 8 byte boundary.
16007 @item -mrelocatable-lib
16008 @itemx -mno-relocatable-lib
16009 @opindex mrelocatable-lib
16010 @opindex mno-relocatable-lib
16011 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
16012 @code{.fixup} section to allow static executables to be relocated at
16013 runtime, but @option{-mrelocatable-lib} does not use the smaller stack
16014 alignment of @option{-mrelocatable}. Objects compiled with
16015 @option{-mrelocatable-lib} may be linked with objects compiled with
16016 any combination of the @option{-mrelocatable} options.
16022 On System V.4 and embedded PowerPC systems do not (do) assume that
16023 register 2 contains a pointer to a global area pointing to the addresses
16024 used in the program.
16027 @itemx -mlittle-endian
16029 @opindex mlittle-endian
16030 On System V.4 and embedded PowerPC systems compile code for the
16031 processor in little endian mode. The @option{-mlittle-endian} option is
16032 the same as @option{-mlittle}.
16035 @itemx -mbig-endian
16037 @opindex mbig-endian
16038 On System V.4 and embedded PowerPC systems compile code for the
16039 processor in big endian mode. The @option{-mbig-endian} option is
16040 the same as @option{-mbig}.
16042 @item -mdynamic-no-pic
16043 @opindex mdynamic-no-pic
16044 On Darwin and Mac OS X systems, compile code so that it is not
16045 relocatable, but that its external references are relocatable. The
16046 resulting code is suitable for applications, but not shared
16049 @item -msingle-pic-base
16050 @opindex msingle-pic-base
16051 Treat the register used for PIC addressing as read-only, rather than
16052 loading it in the prologue for each function. The run-time system is
16053 responsible for initializing this register with an appropriate value
16054 before execution begins.
16056 @item -mprioritize-restricted-insns=@var{priority}
16057 @opindex mprioritize-restricted-insns
16058 This option controls the priority that is assigned to
16059 dispatch-slot restricted instructions during the second scheduling
16060 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
16061 @var{no/highest/second-highest} priority to dispatch slot restricted
16064 @item -msched-costly-dep=@var{dependence_type}
16065 @opindex msched-costly-dep
16066 This option controls which dependences are considered costly
16067 by the target during instruction scheduling. The argument
16068 @var{dependence_type} takes one of the following values:
16069 @var{no}: no dependence is costly,
16070 @var{all}: all dependences are costly,
16071 @var{true_store_to_load}: a true dependence from store to load is costly,
16072 @var{store_to_load}: any dependence from store to load is costly,
16073 @var{number}: any dependence which latency >= @var{number} is costly.
16075 @item -minsert-sched-nops=@var{scheme}
16076 @opindex minsert-sched-nops
16077 This option controls which nop insertion scheme will be used during
16078 the second scheduling pass. The argument @var{scheme} takes one of the
16080 @var{no}: Don't insert nops.
16081 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
16082 according to the scheduler's grouping.
16083 @var{regroup_exact}: Insert nops to force costly dependent insns into
16084 separate groups. Insert exactly as many nops as needed to force an insn
16085 to a new group, according to the estimated processor grouping.
16086 @var{number}: Insert nops to force costly dependent insns into
16087 separate groups. Insert @var{number} nops to force an insn to a new group.
16090 @opindex mcall-sysv
16091 On System V.4 and embedded PowerPC systems compile code using calling
16092 conventions that adheres to the March 1995 draft of the System V
16093 Application Binary Interface, PowerPC processor supplement. This is the
16094 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
16096 @item -mcall-sysv-eabi
16098 @opindex mcall-sysv-eabi
16099 @opindex mcall-eabi
16100 Specify both @option{-mcall-sysv} and @option{-meabi} options.
16102 @item -mcall-sysv-noeabi
16103 @opindex mcall-sysv-noeabi
16104 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
16106 @item -mcall-aixdesc
16108 On System V.4 and embedded PowerPC systems compile code for the AIX
16112 @opindex mcall-linux
16113 On System V.4 and embedded PowerPC systems compile code for the
16114 Linux-based GNU system.
16116 @item -mcall-freebsd
16117 @opindex mcall-freebsd
16118 On System V.4 and embedded PowerPC systems compile code for the
16119 FreeBSD operating system.
16121 @item -mcall-netbsd
16122 @opindex mcall-netbsd
16123 On System V.4 and embedded PowerPC systems compile code for the
16124 NetBSD operating system.
16126 @item -mcall-openbsd
16127 @opindex mcall-netbsd
16128 On System V.4 and embedded PowerPC systems compile code for the
16129 OpenBSD operating system.
16131 @item -maix-struct-return
16132 @opindex maix-struct-return
16133 Return all structures in memory (as specified by the AIX ABI)@.
16135 @item -msvr4-struct-return
16136 @opindex msvr4-struct-return
16137 Return structures smaller than 8 bytes in registers (as specified by the
16140 @item -mabi=@var{abi-type}
16142 Extend the current ABI with a particular extension, or remove such extension.
16143 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
16144 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
16148 Extend the current ABI with SPE ABI extensions. This does not change
16149 the default ABI, instead it adds the SPE ABI extensions to the current
16153 @opindex mabi=no-spe
16154 Disable Booke SPE ABI extensions for the current ABI@.
16156 @item -mabi=ibmlongdouble
16157 @opindex mabi=ibmlongdouble
16158 Change the current ABI to use IBM extended precision long double.
16159 This is a PowerPC 32-bit SYSV ABI option.
16161 @item -mabi=ieeelongdouble
16162 @opindex mabi=ieeelongdouble
16163 Change the current ABI to use IEEE extended precision long double.
16164 This is a PowerPC 32-bit Linux ABI option.
16167 @itemx -mno-prototype
16168 @opindex mprototype
16169 @opindex mno-prototype
16170 On System V.4 and embedded PowerPC systems assume that all calls to
16171 variable argument functions are properly prototyped. Otherwise, the
16172 compiler must insert an instruction before every non prototyped call to
16173 set or clear bit 6 of the condition code register (@var{CR}) to
16174 indicate whether floating point values were passed in the floating point
16175 registers in case the function takes a variable arguments. With
16176 @option{-mprototype}, only calls to prototyped variable argument functions
16177 will set or clear the bit.
16181 On embedded PowerPC systems, assume that the startup module is called
16182 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
16183 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
16188 On embedded PowerPC systems, assume that the startup module is called
16189 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
16194 On embedded PowerPC systems, assume that the startup module is called
16195 @file{crt0.o} and the standard C libraries are @file{libads.a} and
16198 @item -myellowknife
16199 @opindex myellowknife
16200 On embedded PowerPC systems, assume that the startup module is called
16201 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
16206 On System V.4 and embedded PowerPC systems, specify that you are
16207 compiling for a VxWorks system.
16211 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
16212 header to indicate that @samp{eabi} extended relocations are used.
16218 On System V.4 and embedded PowerPC systems do (do not) adhere to the
16219 Embedded Applications Binary Interface (eabi) which is a set of
16220 modifications to the System V.4 specifications. Selecting @option{-meabi}
16221 means that the stack is aligned to an 8 byte boundary, a function
16222 @code{__eabi} is called to from @code{main} to set up the eabi
16223 environment, and the @option{-msdata} option can use both @code{r2} and
16224 @code{r13} to point to two separate small data areas. Selecting
16225 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
16226 do not call an initialization function from @code{main}, and the
16227 @option{-msdata} option will only use @code{r13} to point to a single
16228 small data area. The @option{-meabi} option is on by default if you
16229 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
16232 @opindex msdata=eabi
16233 On System V.4 and embedded PowerPC systems, put small initialized
16234 @code{const} global and static data in the @samp{.sdata2} section, which
16235 is pointed to by register @code{r2}. Put small initialized
16236 non-@code{const} global and static data in the @samp{.sdata} section,
16237 which is pointed to by register @code{r13}. Put small uninitialized
16238 global and static data in the @samp{.sbss} section, which is adjacent to
16239 the @samp{.sdata} section. The @option{-msdata=eabi} option is
16240 incompatible with the @option{-mrelocatable} option. The
16241 @option{-msdata=eabi} option also sets the @option{-memb} option.
16244 @opindex msdata=sysv
16245 On System V.4 and embedded PowerPC systems, put small global and static
16246 data in the @samp{.sdata} section, which is pointed to by register
16247 @code{r13}. Put small uninitialized global and static data in the
16248 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
16249 The @option{-msdata=sysv} option is incompatible with the
16250 @option{-mrelocatable} option.
16252 @item -msdata=default
16254 @opindex msdata=default
16256 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
16257 compile code the same as @option{-msdata=eabi}, otherwise compile code the
16258 same as @option{-msdata=sysv}.
16261 @opindex msdata=data
16262 On System V.4 and embedded PowerPC systems, put small global
16263 data in the @samp{.sdata} section. Put small uninitialized global
16264 data in the @samp{.sbss} section. Do not use register @code{r13}
16265 to address small data however. This is the default behavior unless
16266 other @option{-msdata} options are used.
16270 @opindex msdata=none
16272 On embedded PowerPC systems, put all initialized global and static data
16273 in the @samp{.data} section, and all uninitialized data in the
16274 @samp{.bss} section.
16276 @item -mblock-move-inline-limit=@var{num}
16277 @opindex mblock-move-inline-limit
16278 Inline all block moves (such as calls to @code{memcpy} or structure
16279 copies) less than or equal to @var{num} bytes. The minimum value for
16280 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
16281 targets. The default value is target-specific.
16285 @cindex smaller data references (PowerPC)
16286 @cindex .sdata/.sdata2 references (PowerPC)
16287 On embedded PowerPC systems, put global and static items less than or
16288 equal to @var{num} bytes into the small data or bss sections instead of
16289 the normal data or bss section. By default, @var{num} is 8. The
16290 @option{-G @var{num}} switch is also passed to the linker.
16291 All modules should be compiled with the same @option{-G @var{num}} value.
16294 @itemx -mno-regnames
16296 @opindex mno-regnames
16297 On System V.4 and embedded PowerPC systems do (do not) emit register
16298 names in the assembly language output using symbolic forms.
16301 @itemx -mno-longcall
16303 @opindex mno-longcall
16304 By default assume that all calls are far away so that a longer more
16305 expensive calling sequence is required. This is required for calls
16306 further than 32 megabytes (33,554,432 bytes) from the current location.
16307 A short call will be generated if the compiler knows
16308 the call cannot be that far away. This setting can be overridden by
16309 the @code{shortcall} function attribute, or by @code{#pragma
16312 Some linkers are capable of detecting out-of-range calls and generating
16313 glue code on the fly. On these systems, long calls are unnecessary and
16314 generate slower code. As of this writing, the AIX linker can do this,
16315 as can the GNU linker for PowerPC/64. It is planned to add this feature
16316 to the GNU linker for 32-bit PowerPC systems as well.
16318 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
16319 callee, L42'', plus a ``branch island'' (glue code). The two target
16320 addresses represent the callee and the ``branch island''. The
16321 Darwin/PPC linker will prefer the first address and generate a ``bl
16322 callee'' if the PPC ``bl'' instruction will reach the callee directly;
16323 otherwise, the linker will generate ``bl L42'' to call the ``branch
16324 island''. The ``branch island'' is appended to the body of the
16325 calling function; it computes the full 32-bit address of the callee
16328 On Mach-O (Darwin) systems, this option directs the compiler emit to
16329 the glue for every direct call, and the Darwin linker decides whether
16330 to use or discard it.
16332 In the future, we may cause GCC to ignore all longcall specifications
16333 when the linker is known to generate glue.
16335 @item -mtls-markers
16336 @itemx -mno-tls-markers
16337 @opindex mtls-markers
16338 @opindex mno-tls-markers
16339 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
16340 specifying the function argument. The relocation allows ld to
16341 reliably associate function call with argument setup instructions for
16342 TLS optimization, which in turn allows gcc to better schedule the
16347 Adds support for multithreading with the @dfn{pthreads} library.
16348 This option sets flags for both the preprocessor and linker.
16353 This option will enable GCC to use the reciprocal estimate and
16354 reciprocal square root estimate instructions with additional
16355 Newton-Raphson steps to increase precision instead of doing a divide or
16356 square root and divide for floating point arguments. You should use
16357 the @option{-ffast-math} option when using @option{-mrecip} (or at
16358 least @option{-funsafe-math-optimizations},
16359 @option{-finite-math-only}, @option{-freciprocal-math} and
16360 @option{-fno-trapping-math}). Note that while the throughput of the
16361 sequence is generally higher than the throughput of the non-reciprocal
16362 instruction, the precision of the sequence can be decreased by up to 2
16363 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square
16366 @item -mrecip=@var{opt}
16367 @opindex mrecip=opt
16368 This option allows to control which reciprocal estimate instructions
16369 may be used. @var{opt} is a comma separated list of options, that may
16370 be preceded by a @code{!} to invert the option:
16371 @code{all}: enable all estimate instructions,
16372 @code{default}: enable the default instructions, equivalent to @option{-mrecip},
16373 @code{none}: disable all estimate instructions, equivalent to @option{-mno-recip};
16374 @code{div}: enable the reciprocal approximation instructions for both single and double precision;
16375 @code{divf}: enable the single precision reciprocal approximation instructions;
16376 @code{divd}: enable the double precision reciprocal approximation instructions;
16377 @code{rsqrt}: enable the reciprocal square root approximation instructions for both single and double precision;
16378 @code{rsqrtf}: enable the single precision reciprocal square root approximation instructions;
16379 @code{rsqrtd}: enable the double precision reciprocal square root approximation instructions;
16381 So for example, @option{-mrecip=all,!rsqrtd} would enable the
16382 all of the reciprocal estimate instructions, except for the
16383 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
16384 which handle the double precision reciprocal square root calculations.
16386 @item -mrecip-precision
16387 @itemx -mno-recip-precision
16388 @opindex mrecip-precision
16389 Assume (do not assume) that the reciprocal estimate instructions
16390 provide higher precision estimates than is mandated by the powerpc
16391 ABI. Selecting @option{-mcpu=power6} or @option{-mcpu=power7}
16392 automatically selects @option{-mrecip-precision}. The double
16393 precision square root estimate instructions are not generated by
16394 default on low precision machines, since they do not provide an
16395 estimate that converges after three steps.
16397 @item -mveclibabi=@var{type}
16398 @opindex mveclibabi
16399 Specifies the ABI type to use for vectorizing intrinsics using an
16400 external library. The only type supported at present is @code{mass},
16401 which specifies to use IBM's Mathematical Acceleration Subsystem
16402 (MASS) libraries for vectorizing intrinsics using external libraries.
16403 GCC will currently emit calls to @code{acosd2}, @code{acosf4},
16404 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
16405 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
16406 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
16407 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
16408 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
16409 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
16410 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
16411 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
16412 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
16413 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
16414 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
16415 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
16416 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
16417 for power7. Both @option{-ftree-vectorize} and
16418 @option{-funsafe-math-optimizations} have to be enabled. The MASS
16419 libraries will have to be specified at link time.
16424 Generate (do not generate) the @code{friz} instruction when the
16425 @option{-funsafe-math-optimizations} option is used to optimize
16426 rounding a floating point value to 64-bit integer and back to floating
16427 point. The @code{friz} instruction does not return the same value if
16428 the floating point number is too large to fit in an integer.
16430 @item -mpointers-to-nested-functions
16431 @itemx -mno-pointers-to-nested-functions
16432 @opindex mpointers-to-nested-functions
16433 Generate (do not generate) code to load up the static chain register
16434 (@var{r11}) when calling through a pointer on AIX and 64-bit Linux
16435 systems where a function pointer points to a 3 word descriptor giving
16436 the function address, TOC value to be loaded in register @var{r2}, and
16437 static chain value to be loaded in register @var{r11}. The
16438 @option{-mpointers-to-nested-functions} is on by default. You will
16439 not be able to call through pointers to nested functions or pointers
16440 to functions compiled in other languages that use the static chain if
16441 you use the @option{-mno-pointers-to-nested-functions}.
16443 @item -msave-toc-indirect
16444 @itemx -mno-save-toc-indirect
16445 @opindex msave-toc-indirect
16446 Generate (do not generate) code to save the TOC value in the reserved
16447 stack location in the function prologue if the function calls through
16448 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
16449 saved in the prologue, it is saved just before the call through the
16450 pointer. The @option{-mno-save-toc-indirect} option is the default.
16454 @subsection RX Options
16457 These command line options are defined for RX targets:
16460 @item -m64bit-doubles
16461 @itemx -m32bit-doubles
16462 @opindex m64bit-doubles
16463 @opindex m32bit-doubles
16464 Make the @code{double} data type be 64-bits (@option{-m64bit-doubles})
16465 or 32-bits (@option{-m32bit-doubles}) in size. The default is
16466 @option{-m32bit-doubles}. @emph{Note} RX floating point hardware only
16467 works on 32-bit values, which is why the default is
16468 @option{-m32bit-doubles}.
16474 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
16475 floating point hardware. The default is enabled for the @var{RX600}
16476 series and disabled for the @var{RX200} series.
16478 Floating point instructions will only be generated for 32-bit floating
16479 point values however, so if the @option{-m64bit-doubles} option is in
16480 use then the FPU hardware will not be used for doubles.
16482 @emph{Note} If the @option{-fpu} option is enabled then
16483 @option{-funsafe-math-optimizations} is also enabled automatically.
16484 This is because the RX FPU instructions are themselves unsafe.
16486 @item -mcpu=@var{name}
16488 Selects the type of RX CPU to be targeted. Currently three types are
16489 supported, the generic @var{RX600} and @var{RX200} series hardware and
16490 the specific @var{RX610} CPU. The default is @var{RX600}.
16492 The only difference between @var{RX600} and @var{RX610} is that the
16493 @var{RX610} does not support the @code{MVTIPL} instruction.
16495 The @var{RX200} series does not have a hardware floating point unit
16496 and so @option{-nofpu} is enabled by default when this type is
16499 @item -mbig-endian-data
16500 @itemx -mlittle-endian-data
16501 @opindex mbig-endian-data
16502 @opindex mlittle-endian-data
16503 Store data (but not code) in the big-endian format. The default is
16504 @option{-mlittle-endian-data}, i.e.@: to store data in the little endian
16507 @item -msmall-data-limit=@var{N}
16508 @opindex msmall-data-limit
16509 Specifies the maximum size in bytes of global and static variables
16510 which can be placed into the small data area. Using the small data
16511 area can lead to smaller and faster code, but the size of area is
16512 limited and it is up to the programmer to ensure that the area does
16513 not overflow. Also when the small data area is used one of the RX's
16514 registers (@code{r13}) is reserved for use pointing to this area, so
16515 it is no longer available for use by the compiler. This could result
16516 in slower and/or larger code if variables which once could have been
16517 held in @code{r13} are now pushed onto the stack.
16519 Note, common variables (variables which have not been initialised) and
16520 constants are not placed into the small data area as they are assigned
16521 to other sections in the output executable.
16523 The default value is zero, which disables this feature. Note, this
16524 feature is not enabled by default with higher optimization levels
16525 (@option{-O2} etc) because of the potentially detrimental effects of
16526 reserving register @code{r13}. It is up to the programmer to
16527 experiment and discover whether this feature is of benefit to their
16534 Use the simulator runtime. The default is to use the libgloss board
16537 @item -mas100-syntax
16538 @itemx -mno-as100-syntax
16539 @opindex mas100-syntax
16540 @opindex mno-as100-syntax
16541 When generating assembler output use a syntax that is compatible with
16542 Renesas's AS100 assembler. This syntax can also be handled by the GAS
16543 assembler but it has some restrictions so generating it is not the
16546 @item -mmax-constant-size=@var{N}
16547 @opindex mmax-constant-size
16548 Specifies the maximum size, in bytes, of a constant that can be used as
16549 an operand in a RX instruction. Although the RX instruction set does
16550 allow constants of up to 4 bytes in length to be used in instructions,
16551 a longer value equates to a longer instruction. Thus in some
16552 circumstances it can be beneficial to restrict the size of constants
16553 that are used in instructions. Constants that are too big are instead
16554 placed into a constant pool and referenced via register indirection.
16556 The value @var{N} can be between 0 and 4. A value of 0 (the default)
16557 or 4 means that constants of any size are allowed.
16561 Enable linker relaxation. Linker relaxation is a process whereby the
16562 linker will attempt to reduce the size of a program by finding shorter
16563 versions of various instructions. Disabled by default.
16565 @item -mint-register=@var{N}
16566 @opindex mint-register
16567 Specify the number of registers to reserve for fast interrupt handler
16568 functions. The value @var{N} can be between 0 and 4. A value of 1
16569 means that register @code{r13} will be reserved for the exclusive use
16570 of fast interrupt handlers. A value of 2 reserves @code{r13} and
16571 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
16572 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
16573 A value of 0, the default, does not reserve any registers.
16575 @item -msave-acc-in-interrupts
16576 @opindex msave-acc-in-interrupts
16577 Specifies that interrupt handler functions should preserve the
16578 accumulator register. This is only necessary if normal code might use
16579 the accumulator register, for example because it performs 64-bit
16580 multiplications. The default is to ignore the accumulator as this
16581 makes the interrupt handlers faster.
16585 @emph{Note:} The generic GCC command line @option{-ffixed-@var{reg}}
16586 has special significance to the RX port when used with the
16587 @code{interrupt} function attribute. This attribute indicates a
16588 function intended to process fast interrupts. GCC will will ensure
16589 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
16590 and/or @code{r13} and only provided that the normal use of the
16591 corresponding registers have been restricted via the
16592 @option{-ffixed-@var{reg}} or @option{-mint-register} command line
16595 @node S/390 and zSeries Options
16596 @subsection S/390 and zSeries Options
16597 @cindex S/390 and zSeries Options
16599 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
16603 @itemx -msoft-float
16604 @opindex mhard-float
16605 @opindex msoft-float
16606 Use (do not use) the hardware floating-point instructions and registers
16607 for floating-point operations. When @option{-msoft-float} is specified,
16608 functions in @file{libgcc.a} will be used to perform floating-point
16609 operations. When @option{-mhard-float} is specified, the compiler
16610 generates IEEE floating-point instructions. This is the default.
16613 @itemx -mno-hard-dfp
16615 @opindex mno-hard-dfp
16616 Use (do not use) the hardware decimal-floating-point instructions for
16617 decimal-floating-point operations. When @option{-mno-hard-dfp} is
16618 specified, functions in @file{libgcc.a} will be used to perform
16619 decimal-floating-point operations. When @option{-mhard-dfp} is
16620 specified, the compiler generates decimal-floating-point hardware
16621 instructions. This is the default for @option{-march=z9-ec} or higher.
16623 @item -mlong-double-64
16624 @itemx -mlong-double-128
16625 @opindex mlong-double-64
16626 @opindex mlong-double-128
16627 These switches control the size of @code{long double} type. A size
16628 of 64bit makes the @code{long double} type equivalent to the @code{double}
16629 type. This is the default.
16632 @itemx -mno-backchain
16633 @opindex mbackchain
16634 @opindex mno-backchain
16635 Store (do not store) the address of the caller's frame as backchain pointer
16636 into the callee's stack frame.
16637 A backchain may be needed to allow debugging using tools that do not understand
16638 DWARF-2 call frame information.
16639 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
16640 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
16641 the backchain is placed into the topmost word of the 96/160 byte register
16644 In general, code compiled with @option{-mbackchain} is call-compatible with
16645 code compiled with @option{-mmo-backchain}; however, use of the backchain
16646 for debugging purposes usually requires that the whole binary is built with
16647 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
16648 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16649 to build a linux kernel use @option{-msoft-float}.
16651 The default is to not maintain the backchain.
16653 @item -mpacked-stack
16654 @itemx -mno-packed-stack
16655 @opindex mpacked-stack
16656 @opindex mno-packed-stack
16657 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
16658 specified, the compiler uses the all fields of the 96/160 byte register save
16659 area only for their default purpose; unused fields still take up stack space.
16660 When @option{-mpacked-stack} is specified, register save slots are densely
16661 packed at the top of the register save area; unused space is reused for other
16662 purposes, allowing for more efficient use of the available stack space.
16663 However, when @option{-mbackchain} is also in effect, the topmost word of
16664 the save area is always used to store the backchain, and the return address
16665 register is always saved two words below the backchain.
16667 As long as the stack frame backchain is not used, code generated with
16668 @option{-mpacked-stack} is call-compatible with code generated with
16669 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
16670 S/390 or zSeries generated code that uses the stack frame backchain at run
16671 time, not just for debugging purposes. Such code is not call-compatible
16672 with code compiled with @option{-mpacked-stack}. Also, note that the
16673 combination of @option{-mbackchain},
16674 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
16675 to build a linux kernel use @option{-msoft-float}.
16677 The default is to not use the packed stack layout.
16680 @itemx -mno-small-exec
16681 @opindex msmall-exec
16682 @opindex mno-small-exec
16683 Generate (or do not generate) code using the @code{bras} instruction
16684 to do subroutine calls.
16685 This only works reliably if the total executable size does not
16686 exceed 64k. The default is to use the @code{basr} instruction instead,
16687 which does not have this limitation.
16693 When @option{-m31} is specified, generate code compliant to the
16694 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
16695 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
16696 particular to generate 64-bit instructions. For the @samp{s390}
16697 targets, the default is @option{-m31}, while the @samp{s390x}
16698 targets default to @option{-m64}.
16704 When @option{-mzarch} is specified, generate code using the
16705 instructions available on z/Architecture.
16706 When @option{-mesa} is specified, generate code using the
16707 instructions available on ESA/390. Note that @option{-mesa} is
16708 not possible with @option{-m64}.
16709 When generating code compliant to the GNU/Linux for S/390 ABI,
16710 the default is @option{-mesa}. When generating code compliant
16711 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
16717 Generate (or do not generate) code using the @code{mvcle} instruction
16718 to perform block moves. When @option{-mno-mvcle} is specified,
16719 use a @code{mvc} loop instead. This is the default unless optimizing for
16726 Print (or do not print) additional debug information when compiling.
16727 The default is to not print debug information.
16729 @item -march=@var{cpu-type}
16731 Generate code that will run on @var{cpu-type}, which is the name of a system
16732 representing a certain processor type. Possible values for
16733 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
16734 @samp{z9-109}, @samp{z9-ec} and @samp{z10}.
16735 When generating code using the instructions available on z/Architecture,
16736 the default is @option{-march=z900}. Otherwise, the default is
16737 @option{-march=g5}.
16739 @item -mtune=@var{cpu-type}
16741 Tune to @var{cpu-type} everything applicable about the generated code,
16742 except for the ABI and the set of available instructions.
16743 The list of @var{cpu-type} values is the same as for @option{-march}.
16744 The default is the value used for @option{-march}.
16747 @itemx -mno-tpf-trace
16748 @opindex mtpf-trace
16749 @opindex mno-tpf-trace
16750 Generate code that adds (does not add) in TPF OS specific branches to trace
16751 routines in the operating system. This option is off by default, even
16752 when compiling for the TPF OS@.
16755 @itemx -mno-fused-madd
16756 @opindex mfused-madd
16757 @opindex mno-fused-madd
16758 Generate code that uses (does not use) the floating point multiply and
16759 accumulate instructions. These instructions are generated by default if
16760 hardware floating point is used.
16762 @item -mwarn-framesize=@var{framesize}
16763 @opindex mwarn-framesize
16764 Emit a warning if the current function exceeds the given frame size. Because
16765 this is a compile time check it doesn't need to be a real problem when the program
16766 runs. It is intended to identify functions which most probably cause
16767 a stack overflow. It is useful to be used in an environment with limited stack
16768 size e.g.@: the linux kernel.
16770 @item -mwarn-dynamicstack
16771 @opindex mwarn-dynamicstack
16772 Emit a warning if the function calls alloca or uses dynamically
16773 sized arrays. This is generally a bad idea with a limited stack size.
16775 @item -mstack-guard=@var{stack-guard}
16776 @itemx -mstack-size=@var{stack-size}
16777 @opindex mstack-guard
16778 @opindex mstack-size
16779 If these options are provided the s390 back end emits additional instructions in
16780 the function prologue which trigger a trap if the stack size is @var{stack-guard}
16781 bytes above the @var{stack-size} (remember that the stack on s390 grows downward).
16782 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
16783 the frame size of the compiled function is chosen.
16784 These options are intended to be used to help debugging stack overflow problems.
16785 The additionally emitted code causes only little overhead and hence can also be
16786 used in production like systems without greater performance degradation. The given
16787 values have to be exact powers of 2 and @var{stack-size} has to be greater than
16788 @var{stack-guard} without exceeding 64k.
16789 In order to be efficient the extra code makes the assumption that the stack starts
16790 at an address aligned to the value given by @var{stack-size}.
16791 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
16794 @node Score Options
16795 @subsection Score Options
16796 @cindex Score Options
16798 These options are defined for Score implementations:
16803 Compile code for big endian mode. This is the default.
16807 Compile code for little endian mode.
16811 Disable generate bcnz instruction.
16815 Enable generate unaligned load and store instruction.
16819 Enable the use of multiply-accumulate instructions. Disabled by default.
16823 Specify the SCORE5 as the target architecture.
16827 Specify the SCORE5U of the target architecture.
16831 Specify the SCORE7 as the target architecture. This is the default.
16835 Specify the SCORE7D as the target architecture.
16839 @subsection SH Options
16841 These @samp{-m} options are defined for the SH implementations:
16846 Generate code for the SH1.
16850 Generate code for the SH2.
16853 Generate code for the SH2e.
16857 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
16858 that the floating-point unit is not used.
16860 @item -m2a-single-only
16861 @opindex m2a-single-only
16862 Generate code for the SH2a-FPU, in such a way that no double-precision
16863 floating point operations are used.
16866 @opindex m2a-single
16867 Generate code for the SH2a-FPU assuming the floating-point unit is in
16868 single-precision mode by default.
16872 Generate code for the SH2a-FPU assuming the floating-point unit is in
16873 double-precision mode by default.
16877 Generate code for the SH3.
16881 Generate code for the SH3e.
16885 Generate code for the SH4 without a floating-point unit.
16887 @item -m4-single-only
16888 @opindex m4-single-only
16889 Generate code for the SH4 with a floating-point unit that only
16890 supports single-precision arithmetic.
16894 Generate code for the SH4 assuming the floating-point unit is in
16895 single-precision mode by default.
16899 Generate code for the SH4.
16903 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
16904 floating-point unit is not used.
16906 @item -m4a-single-only
16907 @opindex m4a-single-only
16908 Generate code for the SH4a, in such a way that no double-precision
16909 floating point operations are used.
16912 @opindex m4a-single
16913 Generate code for the SH4a assuming the floating-point unit is in
16914 single-precision mode by default.
16918 Generate code for the SH4a.
16922 Same as @option{-m4a-nofpu}, except that it implicitly passes
16923 @option{-dsp} to the assembler. GCC doesn't generate any DSP
16924 instructions at the moment.
16928 Compile code for the processor in big endian mode.
16932 Compile code for the processor in little endian mode.
16936 Align doubles at 64-bit boundaries. Note that this changes the calling
16937 conventions, and thus some functions from the standard C library will
16938 not work unless you recompile it first with @option{-mdalign}.
16942 Shorten some address references at link time, when possible; uses the
16943 linker option @option{-relax}.
16947 Use 32-bit offsets in @code{switch} tables. The default is to use
16952 Enable the use of bit manipulation instructions on SH2A.
16956 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
16957 alignment constraints.
16961 Comply with the calling conventions defined by Renesas.
16965 Comply with the calling conventions defined by Renesas.
16969 Comply with the calling conventions defined for GCC before the Renesas
16970 conventions were available. This option is the default for all
16971 targets of the SH toolchain.
16974 @opindex mnomacsave
16975 Mark the @code{MAC} register as call-clobbered, even if
16976 @option{-mhitachi} is given.
16980 Increase IEEE-compliance of floating-point code.
16981 At the moment, this is equivalent to @option{-fno-finite-math-only}.
16982 When generating 16 bit SH opcodes, getting IEEE-conforming results for
16983 comparisons of NANs / infinities incurs extra overhead in every
16984 floating point comparison, therefore the default is set to
16985 @option{-ffinite-math-only}.
16987 @item -minline-ic_invalidate
16988 @opindex minline-ic_invalidate
16989 Inline code to invalidate instruction cache entries after setting up
16990 nested function trampolines.
16991 This option has no effect if -musermode is in effect and the selected
16992 code generation option (e.g. -m4) does not allow the use of the icbi
16994 If the selected code generation option does not allow the use of the icbi
16995 instruction, and -musermode is not in effect, the inlined code will
16996 manipulate the instruction cache address array directly with an associative
16997 write. This not only requires privileged mode, but it will also
16998 fail if the cache line had been mapped via the TLB and has become unmapped.
17002 Dump instruction size and location in the assembly code.
17005 @opindex mpadstruct
17006 This option is deprecated. It pads structures to multiple of 4 bytes,
17007 which is incompatible with the SH ABI@.
17011 Optimize for space instead of speed. Implied by @option{-Os}.
17014 @opindex mprefergot
17015 When generating position-independent code, emit function calls using
17016 the Global Offset Table instead of the Procedure Linkage Table.
17020 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
17021 if the inlined code would not work in user mode.
17022 This is the default when the target is @code{sh-*-linux*}.
17024 @item -multcost=@var{number}
17025 @opindex multcost=@var{number}
17026 Set the cost to assume for a multiply insn.
17028 @item -mdiv=@var{strategy}
17029 @opindex mdiv=@var{strategy}
17030 Set the division strategy to use for SHmedia code. @var{strategy} must be
17031 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
17032 inv:call2, inv:fp .
17033 "fp" performs the operation in floating point. This has a very high latency,
17034 but needs only a few instructions, so it might be a good choice if
17035 your code has enough easily exploitable ILP to allow the compiler to
17036 schedule the floating point instructions together with other instructions.
17037 Division by zero causes a floating point exception.
17038 "inv" uses integer operations to calculate the inverse of the divisor,
17039 and then multiplies the dividend with the inverse. This strategy allows
17040 cse and hoisting of the inverse calculation. Division by zero calculates
17041 an unspecified result, but does not trap.
17042 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
17043 have been found, or if the entire operation has been hoisted to the same
17044 place, the last stages of the inverse calculation are intertwined with the
17045 final multiply to reduce the overall latency, at the expense of using a few
17046 more instructions, and thus offering fewer scheduling opportunities with
17048 "call" calls a library function that usually implements the inv:minlat
17050 This gives high code density for m5-*media-nofpu compilations.
17051 "call2" uses a different entry point of the same library function, where it
17052 assumes that a pointer to a lookup table has already been set up, which
17053 exposes the pointer load to cse / code hoisting optimizations.
17054 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
17055 code generation, but if the code stays unoptimized, revert to the "call",
17056 "call2", or "fp" strategies, respectively. Note that the
17057 potentially-trapping side effect of division by zero is carried by a
17058 separate instruction, so it is possible that all the integer instructions
17059 are hoisted out, but the marker for the side effect stays where it is.
17060 A recombination to fp operations or a call is not possible in that case.
17061 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
17062 that the inverse calculation was nor separated from the multiply, they speed
17063 up division where the dividend fits into 20 bits (plus sign where applicable),
17064 by inserting a test to skip a number of operations in this case; this test
17065 slows down the case of larger dividends. inv20u assumes the case of a such
17066 a small dividend to be unlikely, and inv20l assumes it to be likely.
17068 @item -maccumulate-outgoing-args
17069 @opindex maccumulate-outgoing-args
17070 Reserve space once for outgoing arguments in the function prologue rather
17071 than around each call. Generally beneficial for performance and size. Also
17072 needed for unwinding to avoid changing the stack frame around conditional code.
17074 @item -mdivsi3_libfunc=@var{name}
17075 @opindex mdivsi3_libfunc=@var{name}
17076 Set the name of the library function used for 32 bit signed division to
17077 @var{name}. This only affect the name used in the call and inv:call
17078 division strategies, and the compiler will still expect the same
17079 sets of input/output/clobbered registers as if this option was not present.
17081 @item -mfixed-range=@var{register-range}
17082 @opindex mfixed-range
17083 Generate code treating the given register range as fixed registers.
17084 A fixed register is one that the register allocator can not use. This is
17085 useful when compiling kernel code. A register range is specified as
17086 two registers separated by a dash. Multiple register ranges can be
17087 specified separated by a comma.
17089 @item -madjust-unroll
17090 @opindex madjust-unroll
17091 Throttle unrolling to avoid thrashing target registers.
17092 This option only has an effect if the gcc code base supports the
17093 TARGET_ADJUST_UNROLL_MAX target hook.
17095 @item -mindexed-addressing
17096 @opindex mindexed-addressing
17097 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
17098 This is only safe if the hardware and/or OS implement 32 bit wrap-around
17099 semantics for the indexed addressing mode. The architecture allows the
17100 implementation of processors with 64 bit MMU, which the OS could use to
17101 get 32 bit addressing, but since no current hardware implementation supports
17102 this or any other way to make the indexed addressing mode safe to use in
17103 the 32 bit ABI, the default is -mno-indexed-addressing.
17105 @item -mgettrcost=@var{number}
17106 @opindex mgettrcost=@var{number}
17107 Set the cost assumed for the gettr instruction to @var{number}.
17108 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
17112 Assume pt* instructions won't trap. This will generally generate better
17113 scheduled code, but is unsafe on current hardware. The current architecture
17114 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
17115 This has the unintentional effect of making it unsafe to schedule ptabs /
17116 ptrel before a branch, or hoist it out of a loop. For example,
17117 __do_global_ctors, a part of libgcc that runs constructors at program
17118 startup, calls functions in a list which is delimited by @minus{}1. With the
17119 -mpt-fixed option, the ptabs will be done before testing against @minus{}1.
17120 That means that all the constructors will be run a bit quicker, but when
17121 the loop comes to the end of the list, the program crashes because ptabs
17122 loads @minus{}1 into a target register. Since this option is unsafe for any
17123 hardware implementing the current architecture specification, the default
17124 is -mno-pt-fixed. Unless the user specifies a specific cost with
17125 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
17126 this deters register allocation using target registers for storing
17129 @item -minvalid-symbols
17130 @opindex minvalid-symbols
17131 Assume symbols might be invalid. Ordinary function symbols generated by
17132 the compiler will always be valid to load with movi/shori/ptabs or
17133 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
17134 to generate symbols that will cause ptabs / ptrel to trap.
17135 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
17136 It will then prevent cross-basic-block cse, hoisting and most scheduling
17137 of symbol loads. The default is @option{-mno-invalid-symbols}.
17140 @node Solaris 2 Options
17141 @subsection Solaris 2 Options
17142 @cindex Solaris 2 options
17144 These @samp{-m} options are supported on Solaris 2:
17147 @item -mimpure-text
17148 @opindex mimpure-text
17149 @option{-mimpure-text}, used in addition to @option{-shared}, tells
17150 the compiler to not pass @option{-z text} to the linker when linking a
17151 shared object. Using this option, you can link position-dependent
17152 code into a shared object.
17154 @option{-mimpure-text} suppresses the ``relocations remain against
17155 allocatable but non-writable sections'' linker error message.
17156 However, the necessary relocations will trigger copy-on-write, and the
17157 shared object is not actually shared across processes. Instead of
17158 using @option{-mimpure-text}, you should compile all source code with
17159 @option{-fpic} or @option{-fPIC}.
17163 These switches are supported in addition to the above on Solaris 2:
17168 Add support for multithreading using the POSIX threads library. This
17169 option sets flags for both the preprocessor and linker. This option does
17170 not affect the thread safety of object code produced by the compiler or
17171 that of libraries supplied with it.
17175 This is a synonym for @option{-pthreads}.
17178 @node SPARC Options
17179 @subsection SPARC Options
17180 @cindex SPARC options
17182 These @samp{-m} options are supported on the SPARC:
17185 @item -mno-app-regs
17187 @opindex mno-app-regs
17189 Specify @option{-mapp-regs} to generate output using the global registers
17190 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
17193 To be fully SVR4 ABI compliant at the cost of some performance loss,
17194 specify @option{-mno-app-regs}. You should compile libraries and system
17195 software with this option.
17201 With @option{-mflat}, the compiler does not generate save/restore instructions
17202 and uses a ``flat'' or single register window model. This model is compatible
17203 with the regular register window model. The local registers and the input
17204 registers (0--5) are still treated as ``call-saved'' registers and will be
17205 saved on the stack as needed.
17207 With @option{-mno-flat} (the default), the compiler generates save/restore
17208 instructions (except for leaf functions). This is the normal operating mode.
17211 @itemx -mhard-float
17213 @opindex mhard-float
17214 Generate output containing floating point instructions. This is the
17218 @itemx -msoft-float
17220 @opindex msoft-float
17221 Generate output containing library calls for floating point.
17222 @strong{Warning:} the requisite libraries are not available for all SPARC
17223 targets. Normally the facilities of the machine's usual C compiler are
17224 used, but this cannot be done directly in cross-compilation. You must make
17225 your own arrangements to provide suitable library functions for
17226 cross-compilation. The embedded targets @samp{sparc-*-aout} and
17227 @samp{sparclite-*-*} do provide software floating point support.
17229 @option{-msoft-float} changes the calling convention in the output file;
17230 therefore, it is only useful if you compile @emph{all} of a program with
17231 this option. In particular, you need to compile @file{libgcc.a}, the
17232 library that comes with GCC, with @option{-msoft-float} in order for
17235 @item -mhard-quad-float
17236 @opindex mhard-quad-float
17237 Generate output containing quad-word (long double) floating point
17240 @item -msoft-quad-float
17241 @opindex msoft-quad-float
17242 Generate output containing library calls for quad-word (long double)
17243 floating point instructions. The functions called are those specified
17244 in the SPARC ABI@. This is the default.
17246 As of this writing, there are no SPARC implementations that have hardware
17247 support for the quad-word floating point instructions. They all invoke
17248 a trap handler for one of these instructions, and then the trap handler
17249 emulates the effect of the instruction. Because of the trap handler overhead,
17250 this is much slower than calling the ABI library routines. Thus the
17251 @option{-msoft-quad-float} option is the default.
17253 @item -mno-unaligned-doubles
17254 @itemx -munaligned-doubles
17255 @opindex mno-unaligned-doubles
17256 @opindex munaligned-doubles
17257 Assume that doubles have 8 byte alignment. This is the default.
17259 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
17260 alignment only if they are contained in another type, or if they have an
17261 absolute address. Otherwise, it assumes they have 4 byte alignment.
17262 Specifying this option avoids some rare compatibility problems with code
17263 generated by other compilers. It is not the default because it results
17264 in a performance loss, especially for floating point code.
17266 @item -mno-faster-structs
17267 @itemx -mfaster-structs
17268 @opindex mno-faster-structs
17269 @opindex mfaster-structs
17270 With @option{-mfaster-structs}, the compiler assumes that structures
17271 should have 8 byte alignment. This enables the use of pairs of
17272 @code{ldd} and @code{std} instructions for copies in structure
17273 assignment, in place of twice as many @code{ld} and @code{st} pairs.
17274 However, the use of this changed alignment directly violates the SPARC
17275 ABI@. Thus, it's intended only for use on targets where the developer
17276 acknowledges that their resulting code will not be directly in line with
17277 the rules of the ABI@.
17279 @item -mcpu=@var{cpu_type}
17281 Set the instruction set, register set, and instruction scheduling parameters
17282 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
17283 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
17284 @samp{leon}, @samp{sparclite}, @samp{f930}, @samp{f934}, @samp{sparclite86x},
17285 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
17286 @samp{ultrasparc3}, @samp{niagara} and @samp{niagara2}.
17288 Native Solaris and Linux toolchains also support the value @samp{native},
17289 which selects the best architecture option for the host processor.
17290 @option{-mcpu=native} has no effect if GCC does not recognize
17293 Default instruction scheduling parameters are used for values that select
17294 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
17295 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
17297 Here is a list of each supported architecture and their supported
17302 v8: supersparc, hypersparc, leon
17303 sparclite: f930, f934, sparclite86x
17305 v9: ultrasparc, ultrasparc3, niagara, niagara2
17308 By default (unless configured otherwise), GCC generates code for the V7
17309 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
17310 additionally optimizes it for the Cypress CY7C602 chip, as used in the
17311 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
17312 SPARCStation 1, 2, IPX etc.
17314 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
17315 architecture. The only difference from V7 code is that the compiler emits
17316 the integer multiply and integer divide instructions which exist in SPARC-V8
17317 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
17318 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
17321 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
17322 the SPARC architecture. This adds the integer multiply, integer divide step
17323 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
17324 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
17325 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
17326 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
17327 MB86934 chip, which is the more recent SPARClite with FPU@.
17329 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
17330 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
17331 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
17332 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
17333 optimizes it for the TEMIC SPARClet chip.
17335 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
17336 architecture. This adds 64-bit integer and floating-point move instructions,
17337 3 additional floating-point condition code registers and conditional move
17338 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
17339 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
17340 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
17341 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
17342 @option{-mcpu=niagara}, the compiler additionally optimizes it for
17343 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
17344 additionally optimizes it for Sun UltraSPARC T2 chips.
17346 @item -mtune=@var{cpu_type}
17348 Set the instruction scheduling parameters for machine type
17349 @var{cpu_type}, but do not set the instruction set or register set that the
17350 option @option{-mcpu=@var{cpu_type}} would.
17352 The same values for @option{-mcpu=@var{cpu_type}} can be used for
17353 @option{-mtune=@var{cpu_type}}, but the only useful values are those
17354 that select a particular CPU implementation. Those are @samp{cypress},
17355 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{f930}, @samp{f934},
17356 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, @samp{ultrasparc3},
17357 @samp{niagara}, and @samp{niagara2}. With native Solaris and Linux
17358 toolchains, @samp{native} can also be used.
17363 @opindex mno-v8plus
17364 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
17365 difference from the V8 ABI is that the global and out registers are
17366 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
17367 mode for all SPARC-V9 processors.
17373 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
17374 Visual Instruction Set extensions. The default is @option{-mno-vis}.
17377 These @samp{-m} options are supported in addition to the above
17378 on SPARC-V9 processors in 64-bit environments:
17381 @item -mlittle-endian
17382 @opindex mlittle-endian
17383 Generate code for a processor running in little-endian mode. It is only
17384 available for a few configurations and most notably not on Solaris and Linux.
17390 Generate code for a 32-bit or 64-bit environment.
17391 The 32-bit environment sets int, long and pointer to 32 bits.
17392 The 64-bit environment sets int to 32 bits and long and pointer
17395 @item -mcmodel=medlow
17396 @opindex mcmodel=medlow
17397 Generate code for the Medium/Low code model: 64-bit addresses, programs
17398 must be linked in the low 32 bits of memory. Programs can be statically
17399 or dynamically linked.
17401 @item -mcmodel=medmid
17402 @opindex mcmodel=medmid
17403 Generate code for the Medium/Middle code model: 64-bit addresses, programs
17404 must be linked in the low 44 bits of memory, the text and data segments must
17405 be less than 2GB in size and the data segment must be located within 2GB of
17408 @item -mcmodel=medany
17409 @opindex mcmodel=medany
17410 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
17411 may be linked anywhere in memory, the text and data segments must be less
17412 than 2GB in size and the data segment must be located within 2GB of the
17415 @item -mcmodel=embmedany
17416 @opindex mcmodel=embmedany
17417 Generate code for the Medium/Anywhere code model for embedded systems:
17418 64-bit addresses, the text and data segments must be less than 2GB in
17419 size, both starting anywhere in memory (determined at link time). The
17420 global register %g4 points to the base of the data segment. Programs
17421 are statically linked and PIC is not supported.
17424 @itemx -mno-stack-bias
17425 @opindex mstack-bias
17426 @opindex mno-stack-bias
17427 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
17428 frame pointer if present, are offset by @minus{}2047 which must be added back
17429 when making stack frame references. This is the default in 64-bit mode.
17430 Otherwise, assume no such offset is present.
17434 @subsection SPU Options
17435 @cindex SPU options
17437 These @samp{-m} options are supported on the SPU:
17441 @itemx -merror-reloc
17442 @opindex mwarn-reloc
17443 @opindex merror-reloc
17445 The loader for SPU does not handle dynamic relocations. By default, GCC
17446 will give an error when it generates code that requires a dynamic
17447 relocation. @option{-mno-error-reloc} disables the error,
17448 @option{-mwarn-reloc} will generate a warning instead.
17451 @itemx -munsafe-dma
17453 @opindex munsafe-dma
17455 Instructions which initiate or test completion of DMA must not be
17456 reordered with respect to loads and stores of the memory which is being
17457 accessed. Users typically address this problem using the volatile
17458 keyword, but that can lead to inefficient code in places where the
17459 memory is known to not change. Rather than mark the memory as volatile
17460 we treat the DMA instructions as potentially effecting all memory. With
17461 @option{-munsafe-dma} users must use the volatile keyword to protect
17464 @item -mbranch-hints
17465 @opindex mbranch-hints
17467 By default, GCC will generate a branch hint instruction to avoid
17468 pipeline stalls for always taken or probably taken branches. A hint
17469 will not be generated closer than 8 instructions away from its branch.
17470 There is little reason to disable them, except for debugging purposes,
17471 or to make an object a little bit smaller.
17475 @opindex msmall-mem
17476 @opindex mlarge-mem
17478 By default, GCC generates code assuming that addresses are never larger
17479 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
17480 a full 32 bit address.
17485 By default, GCC links against startup code that assumes the SPU-style
17486 main function interface (which has an unconventional parameter list).
17487 With @option{-mstdmain}, GCC will link your program against startup
17488 code that assumes a C99-style interface to @code{main}, including a
17489 local copy of @code{argv} strings.
17491 @item -mfixed-range=@var{register-range}
17492 @opindex mfixed-range
17493 Generate code treating the given register range as fixed registers.
17494 A fixed register is one that the register allocator can not use. This is
17495 useful when compiling kernel code. A register range is specified as
17496 two registers separated by a dash. Multiple register ranges can be
17497 specified separated by a comma.
17503 Compile code assuming that pointers to the PPU address space accessed
17504 via the @code{__ea} named address space qualifier are either 32 or 64
17505 bits wide. The default is 32 bits. As this is an ABI changing option,
17506 all object code in an executable must be compiled with the same setting.
17508 @item -maddress-space-conversion
17509 @itemx -mno-address-space-conversion
17510 @opindex maddress-space-conversion
17511 @opindex mno-address-space-conversion
17512 Allow/disallow treating the @code{__ea} address space as superset
17513 of the generic address space. This enables explicit type casts
17514 between @code{__ea} and generic pointer as well as implicit
17515 conversions of generic pointers to @code{__ea} pointers. The
17516 default is to allow address space pointer conversions.
17518 @item -mcache-size=@var{cache-size}
17519 @opindex mcache-size
17520 This option controls the version of libgcc that the compiler links to an
17521 executable and selects a software-managed cache for accessing variables
17522 in the @code{__ea} address space with a particular cache size. Possible
17523 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
17524 and @samp{128}. The default cache size is 64KB.
17526 @item -matomic-updates
17527 @itemx -mno-atomic-updates
17528 @opindex matomic-updates
17529 @opindex mno-atomic-updates
17530 This option controls the version of libgcc that the compiler links to an
17531 executable and selects whether atomic updates to the software-managed
17532 cache of PPU-side variables are used. If you use atomic updates, changes
17533 to a PPU variable from SPU code using the @code{__ea} named address space
17534 qualifier will not interfere with changes to other PPU variables residing
17535 in the same cache line from PPU code. If you do not use atomic updates,
17536 such interference may occur; however, writing back cache lines will be
17537 more efficient. The default behavior is to use atomic updates.
17540 @itemx -mdual-nops=@var{n}
17541 @opindex mdual-nops
17542 By default, GCC will insert nops to increase dual issue when it expects
17543 it to increase performance. @var{n} can be a value from 0 to 10. A
17544 smaller @var{n} will insert fewer nops. 10 is the default, 0 is the
17545 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
17547 @item -mhint-max-nops=@var{n}
17548 @opindex mhint-max-nops
17549 Maximum number of nops to insert for a branch hint. A branch hint must
17550 be at least 8 instructions away from the branch it is effecting. GCC
17551 will insert up to @var{n} nops to enforce this, otherwise it will not
17552 generate the branch hint.
17554 @item -mhint-max-distance=@var{n}
17555 @opindex mhint-max-distance
17556 The encoding of the branch hint instruction limits the hint to be within
17557 256 instructions of the branch it is effecting. By default, GCC makes
17558 sure it is within 125.
17561 @opindex msafe-hints
17562 Work around a hardware bug which causes the SPU to stall indefinitely.
17563 By default, GCC will insert the @code{hbrp} instruction to make sure
17564 this stall won't happen.
17568 @node System V Options
17569 @subsection Options for System V
17571 These additional options are available on System V Release 4 for
17572 compatibility with other compilers on those systems:
17577 Create a shared object.
17578 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
17582 Identify the versions of each tool used by the compiler, in a
17583 @code{.ident} assembler directive in the output.
17587 Refrain from adding @code{.ident} directives to the output file (this is
17590 @item -YP,@var{dirs}
17592 Search the directories @var{dirs}, and no others, for libraries
17593 specified with @option{-l}.
17595 @item -Ym,@var{dir}
17597 Look in the directory @var{dir} to find the M4 preprocessor.
17598 The assembler uses this option.
17599 @c This is supposed to go with a -Yd for predefined M4 macro files, but
17600 @c the generic assembler that comes with Solaris takes just -Ym.
17604 @subsection V850 Options
17605 @cindex V850 Options
17607 These @samp{-m} options are defined for V850 implementations:
17611 @itemx -mno-long-calls
17612 @opindex mlong-calls
17613 @opindex mno-long-calls
17614 Treat all calls as being far away (near). If calls are assumed to be
17615 far away, the compiler will always load the functions address up into a
17616 register, and call indirect through the pointer.
17622 Do not optimize (do optimize) basic blocks that use the same index
17623 pointer 4 or more times to copy pointer into the @code{ep} register, and
17624 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
17625 option is on by default if you optimize.
17627 @item -mno-prolog-function
17628 @itemx -mprolog-function
17629 @opindex mno-prolog-function
17630 @opindex mprolog-function
17631 Do not use (do use) external functions to save and restore registers
17632 at the prologue and epilogue of a function. The external functions
17633 are slower, but use less code space if more than one function saves
17634 the same number of registers. The @option{-mprolog-function} option
17635 is on by default if you optimize.
17639 Try to make the code as small as possible. At present, this just turns
17640 on the @option{-mep} and @option{-mprolog-function} options.
17642 @item -mtda=@var{n}
17644 Put static or global variables whose size is @var{n} bytes or less into
17645 the tiny data area that register @code{ep} points to. The tiny data
17646 area can hold up to 256 bytes in total (128 bytes for byte references).
17648 @item -msda=@var{n}
17650 Put static or global variables whose size is @var{n} bytes or less into
17651 the small data area that register @code{gp} points to. The small data
17652 area can hold up to 64 kilobytes.
17654 @item -mzda=@var{n}
17656 Put static or global variables whose size is @var{n} bytes or less into
17657 the first 32 kilobytes of memory.
17661 Specify that the target processor is the V850.
17664 @opindex mbig-switch
17665 Generate code suitable for big switch tables. Use this option only if
17666 the assembler/linker complain about out of range branches within a switch
17671 This option will cause r2 and r5 to be used in the code generated by
17672 the compiler. This setting is the default.
17674 @item -mno-app-regs
17675 @opindex mno-app-regs
17676 This option will cause r2 and r5 to be treated as fixed registers.
17680 Specify that the target processor is the V850E2V3. The preprocessor
17681 constants @samp{__v850e2v3__} will be defined if
17682 this option is used.
17686 Specify that the target processor is the V850E2. The preprocessor
17687 constants @samp{__v850e2__} will be defined if this option is used.
17691 Specify that the target processor is the V850E1. The preprocessor
17692 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
17693 this option is used.
17697 Specify that the target processor is the V850ES. This is an alias for
17698 the @option{-mv850e1} option.
17702 Specify that the target processor is the V850E@. The preprocessor
17703 constant @samp{__v850e__} will be defined if this option is used.
17705 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
17706 nor @option{-mv850e2} nor @option{-mv850e2v3}
17707 are defined then a default target processor will be chosen and the
17708 relevant @samp{__v850*__} preprocessor constant will be defined.
17710 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
17711 defined, regardless of which processor variant is the target.
17713 @item -mdisable-callt
17714 @opindex mdisable-callt
17715 This option will suppress generation of the CALLT instruction for the
17716 v850e, v850e1, v850e2 and v850e2v3 flavors of the v850 architecture. The default is
17717 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
17722 @subsection VAX Options
17723 @cindex VAX options
17725 These @samp{-m} options are defined for the VAX:
17730 Do not output certain jump instructions (@code{aobleq} and so on)
17731 that the Unix assembler for the VAX cannot handle across long
17736 Do output those jump instructions, on the assumption that you
17737 will assemble with the GNU assembler.
17741 Output code for g-format floating point numbers instead of d-format.
17744 @node VxWorks Options
17745 @subsection VxWorks Options
17746 @cindex VxWorks Options
17748 The options in this section are defined for all VxWorks targets.
17749 Options specific to the target hardware are listed with the other
17750 options for that target.
17755 GCC can generate code for both VxWorks kernels and real time processes
17756 (RTPs). This option switches from the former to the latter. It also
17757 defines the preprocessor macro @code{__RTP__}.
17760 @opindex non-static
17761 Link an RTP executable against shared libraries rather than static
17762 libraries. The options @option{-static} and @option{-shared} can
17763 also be used for RTPs (@pxref{Link Options}); @option{-static}
17770 These options are passed down to the linker. They are defined for
17771 compatibility with Diab.
17774 @opindex Xbind-lazy
17775 Enable lazy binding of function calls. This option is equivalent to
17776 @option{-Wl,-z,now} and is defined for compatibility with Diab.
17780 Disable lazy binding of function calls. This option is the default and
17781 is defined for compatibility with Diab.
17784 @node x86-64 Options
17785 @subsection x86-64 Options
17786 @cindex x86-64 options
17788 These are listed under @xref{i386 and x86-64 Options}.
17790 @node Xstormy16 Options
17791 @subsection Xstormy16 Options
17792 @cindex Xstormy16 Options
17794 These options are defined for Xstormy16:
17799 Choose startup files and linker script suitable for the simulator.
17802 @node Xtensa Options
17803 @subsection Xtensa Options
17804 @cindex Xtensa Options
17806 These options are supported for Xtensa targets:
17810 @itemx -mno-const16
17812 @opindex mno-const16
17813 Enable or disable use of @code{CONST16} instructions for loading
17814 constant values. The @code{CONST16} instruction is currently not a
17815 standard option from Tensilica. When enabled, @code{CONST16}
17816 instructions are always used in place of the standard @code{L32R}
17817 instructions. The use of @code{CONST16} is enabled by default only if
17818 the @code{L32R} instruction is not available.
17821 @itemx -mno-fused-madd
17822 @opindex mfused-madd
17823 @opindex mno-fused-madd
17824 Enable or disable use of fused multiply/add and multiply/subtract
17825 instructions in the floating-point option. This has no effect if the
17826 floating-point option is not also enabled. Disabling fused multiply/add
17827 and multiply/subtract instructions forces the compiler to use separate
17828 instructions for the multiply and add/subtract operations. This may be
17829 desirable in some cases where strict IEEE 754-compliant results are
17830 required: the fused multiply add/subtract instructions do not round the
17831 intermediate result, thereby producing results with @emph{more} bits of
17832 precision than specified by the IEEE standard. Disabling fused multiply
17833 add/subtract instructions also ensures that the program output is not
17834 sensitive to the compiler's ability to combine multiply and add/subtract
17837 @item -mserialize-volatile
17838 @itemx -mno-serialize-volatile
17839 @opindex mserialize-volatile
17840 @opindex mno-serialize-volatile
17841 When this option is enabled, GCC inserts @code{MEMW} instructions before
17842 @code{volatile} memory references to guarantee sequential consistency.
17843 The default is @option{-mserialize-volatile}. Use
17844 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
17846 @item -mforce-no-pic
17847 @opindex mforce-no-pic
17848 For targets, like GNU/Linux, where all user-mode Xtensa code must be
17849 position-independent code (PIC), this option disables PIC for compiling
17852 @item -mtext-section-literals
17853 @itemx -mno-text-section-literals
17854 @opindex mtext-section-literals
17855 @opindex mno-text-section-literals
17856 Control the treatment of literal pools. The default is
17857 @option{-mno-text-section-literals}, which places literals in a separate
17858 section in the output file. This allows the literal pool to be placed
17859 in a data RAM/ROM, and it also allows the linker to combine literal
17860 pools from separate object files to remove redundant literals and
17861 improve code size. With @option{-mtext-section-literals}, the literals
17862 are interspersed in the text section in order to keep them as close as
17863 possible to their references. This may be necessary for large assembly
17866 @item -mtarget-align
17867 @itemx -mno-target-align
17868 @opindex mtarget-align
17869 @opindex mno-target-align
17870 When this option is enabled, GCC instructs the assembler to
17871 automatically align instructions to reduce branch penalties at the
17872 expense of some code density. The assembler attempts to widen density
17873 instructions to align branch targets and the instructions following call
17874 instructions. If there are not enough preceding safe density
17875 instructions to align a target, no widening will be performed. The
17876 default is @option{-mtarget-align}. These options do not affect the
17877 treatment of auto-aligned instructions like @code{LOOP}, which the
17878 assembler will always align, either by widening density instructions or
17879 by inserting no-op instructions.
17882 @itemx -mno-longcalls
17883 @opindex mlongcalls
17884 @opindex mno-longcalls
17885 When this option is enabled, GCC instructs the assembler to translate
17886 direct calls to indirect calls unless it can determine that the target
17887 of a direct call is in the range allowed by the call instruction. This
17888 translation typically occurs for calls to functions in other source
17889 files. Specifically, the assembler translates a direct @code{CALL}
17890 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
17891 The default is @option{-mno-longcalls}. This option should be used in
17892 programs where the call target can potentially be out of range. This
17893 option is implemented in the assembler, not the compiler, so the
17894 assembly code generated by GCC will still show direct call
17895 instructions---look at the disassembled object code to see the actual
17896 instructions. Note that the assembler will use an indirect call for
17897 every cross-file call, not just those that really will be out of range.
17900 @node zSeries Options
17901 @subsection zSeries Options
17902 @cindex zSeries options
17904 These are listed under @xref{S/390 and zSeries Options}.
17906 @node Code Gen Options
17907 @section Options for Code Generation Conventions
17908 @cindex code generation conventions
17909 @cindex options, code generation
17910 @cindex run-time options
17912 These machine-independent options control the interface conventions
17913 used in code generation.
17915 Most of them have both positive and negative forms; the negative form
17916 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
17917 one of the forms is listed---the one which is not the default. You
17918 can figure out the other form by either removing @samp{no-} or adding
17922 @item -fbounds-check
17923 @opindex fbounds-check
17924 For front-ends that support it, generate additional code to check that
17925 indices used to access arrays are within the declared range. This is
17926 currently only supported by the Java and Fortran front-ends, where
17927 this option defaults to true and false respectively.
17931 This option generates traps for signed overflow on addition, subtraction,
17932 multiplication operations.
17936 This option instructs the compiler to assume that signed arithmetic
17937 overflow of addition, subtraction and multiplication wraps around
17938 using twos-complement representation. This flag enables some optimizations
17939 and disables others. This option is enabled by default for the Java
17940 front-end, as required by the Java language specification.
17943 @opindex fexceptions
17944 Enable exception handling. Generates extra code needed to propagate
17945 exceptions. For some targets, this implies GCC will generate frame
17946 unwind information for all functions, which can produce significant data
17947 size overhead, although it does not affect execution. If you do not
17948 specify this option, GCC will enable it by default for languages like
17949 C++ which normally require exception handling, and disable it for
17950 languages like C that do not normally require it. However, you may need
17951 to enable this option when compiling C code that needs to interoperate
17952 properly with exception handlers written in C++. You may also wish to
17953 disable this option if you are compiling older C++ programs that don't
17954 use exception handling.
17956 @item -fnon-call-exceptions
17957 @opindex fnon-call-exceptions
17958 Generate code that allows trapping instructions to throw exceptions.
17959 Note that this requires platform-specific runtime support that does
17960 not exist everywhere. Moreover, it only allows @emph{trapping}
17961 instructions to throw exceptions, i.e.@: memory references or floating
17962 point instructions. It does not allow exceptions to be thrown from
17963 arbitrary signal handlers such as @code{SIGALRM}.
17965 @item -funwind-tables
17966 @opindex funwind-tables
17967 Similar to @option{-fexceptions}, except that it will just generate any needed
17968 static data, but will not affect the generated code in any other way.
17969 You will normally not enable this option; instead, a language processor
17970 that needs this handling would enable it on your behalf.
17972 @item -fasynchronous-unwind-tables
17973 @opindex fasynchronous-unwind-tables
17974 Generate unwind table in dwarf2 format, if supported by target machine. The
17975 table is exact at each instruction boundary, so it can be used for stack
17976 unwinding from asynchronous events (such as debugger or garbage collector).
17978 @item -fpcc-struct-return
17979 @opindex fpcc-struct-return
17980 Return ``short'' @code{struct} and @code{union} values in memory like
17981 longer ones, rather than in registers. This convention is less
17982 efficient, but it has the advantage of allowing intercallability between
17983 GCC-compiled files and files compiled with other compilers, particularly
17984 the Portable C Compiler (pcc).
17986 The precise convention for returning structures in memory depends
17987 on the target configuration macros.
17989 Short structures and unions are those whose size and alignment match
17990 that of some integer type.
17992 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
17993 switch is not binary compatible with code compiled with the
17994 @option{-freg-struct-return} switch.
17995 Use it to conform to a non-default application binary interface.
17997 @item -freg-struct-return
17998 @opindex freg-struct-return
17999 Return @code{struct} and @code{union} values in registers when possible.
18000 This is more efficient for small structures than
18001 @option{-fpcc-struct-return}.
18003 If you specify neither @option{-fpcc-struct-return} nor
18004 @option{-freg-struct-return}, GCC defaults to whichever convention is
18005 standard for the target. If there is no standard convention, GCC
18006 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
18007 the principal compiler. In those cases, we can choose the standard, and
18008 we chose the more efficient register return alternative.
18010 @strong{Warning:} code compiled with the @option{-freg-struct-return}
18011 switch is not binary compatible with code compiled with the
18012 @option{-fpcc-struct-return} switch.
18013 Use it to conform to a non-default application binary interface.
18015 @item -fshort-enums
18016 @opindex fshort-enums
18017 Allocate to an @code{enum} type only as many bytes as it needs for the
18018 declared range of possible values. Specifically, the @code{enum} type
18019 will be equivalent to the smallest integer type which has enough room.
18021 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
18022 code that is not binary compatible with code generated without that switch.
18023 Use it to conform to a non-default application binary interface.
18025 @item -fshort-double
18026 @opindex fshort-double
18027 Use the same size for @code{double} as for @code{float}.
18029 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
18030 code that is not binary compatible with code generated without that switch.
18031 Use it to conform to a non-default application binary interface.
18033 @item -fshort-wchar
18034 @opindex fshort-wchar
18035 Override the underlying type for @samp{wchar_t} to be @samp{short
18036 unsigned int} instead of the default for the target. This option is
18037 useful for building programs to run under WINE@.
18039 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
18040 code that is not binary compatible with code generated without that switch.
18041 Use it to conform to a non-default application binary interface.
18044 @opindex fno-common
18045 In C code, controls the placement of uninitialized global variables.
18046 Unix C compilers have traditionally permitted multiple definitions of
18047 such variables in different compilation units by placing the variables
18049 This is the behavior specified by @option{-fcommon}, and is the default
18050 for GCC on most targets.
18051 On the other hand, this behavior is not required by ISO C, and on some
18052 targets may carry a speed or code size penalty on variable references.
18053 The @option{-fno-common} option specifies that the compiler should place
18054 uninitialized global variables in the data section of the object file,
18055 rather than generating them as common blocks.
18056 This has the effect that if the same variable is declared
18057 (without @code{extern}) in two different compilations,
18058 you will get a multiple-definition error when you link them.
18059 In this case, you must compile with @option{-fcommon} instead.
18060 Compiling with @option{-fno-common} is useful on targets for which
18061 it provides better performance, or if you wish to verify that the
18062 program will work on other systems which always treat uninitialized
18063 variable declarations this way.
18067 Ignore the @samp{#ident} directive.
18069 @item -finhibit-size-directive
18070 @opindex finhibit-size-directive
18071 Don't output a @code{.size} assembler directive, or anything else that
18072 would cause trouble if the function is split in the middle, and the
18073 two halves are placed at locations far apart in memory. This option is
18074 used when compiling @file{crtstuff.c}; you should not need to use it
18077 @item -fverbose-asm
18078 @opindex fverbose-asm
18079 Put extra commentary information in the generated assembly code to
18080 make it more readable. This option is generally only of use to those
18081 who actually need to read the generated assembly code (perhaps while
18082 debugging the compiler itself).
18084 @option{-fno-verbose-asm}, the default, causes the
18085 extra information to be omitted and is useful when comparing two assembler
18088 @item -frecord-gcc-switches
18089 @opindex frecord-gcc-switches
18090 This switch causes the command line that was used to invoke the
18091 compiler to be recorded into the object file that is being created.
18092 This switch is only implemented on some targets and the exact format
18093 of the recording is target and binary file format dependent, but it
18094 usually takes the form of a section containing ASCII text. This
18095 switch is related to the @option{-fverbose-asm} switch, but that
18096 switch only records information in the assembler output file as
18097 comments, so it never reaches the object file.
18098 See also @option{-grecord-gcc-switches} for another
18099 way of storing compiler options into the object file.
18103 @cindex global offset table
18105 Generate position-independent code (PIC) suitable for use in a shared
18106 library, if supported for the target machine. Such code accesses all
18107 constant addresses through a global offset table (GOT)@. The dynamic
18108 loader resolves the GOT entries when the program starts (the dynamic
18109 loader is not part of GCC; it is part of the operating system). If
18110 the GOT size for the linked executable exceeds a machine-specific
18111 maximum size, you get an error message from the linker indicating that
18112 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
18113 instead. (These maximums are 8k on the SPARC and 32k
18114 on the m68k and RS/6000. The 386 has no such limit.)
18116 Position-independent code requires special support, and therefore works
18117 only on certain machines. For the 386, GCC supports PIC for System V
18118 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
18119 position-independent.
18121 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18126 If supported for the target machine, emit position-independent code,
18127 suitable for dynamic linking and avoiding any limit on the size of the
18128 global offset table. This option makes a difference on the m68k,
18129 PowerPC and SPARC@.
18131 Position-independent code requires special support, and therefore works
18132 only on certain machines.
18134 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
18141 These options are similar to @option{-fpic} and @option{-fPIC}, but
18142 generated position independent code can be only linked into executables.
18143 Usually these options are used when @option{-pie} GCC option will be
18144 used during linking.
18146 @option{-fpie} and @option{-fPIE} both define the macros
18147 @code{__pie__} and @code{__PIE__}. The macros have the value 1
18148 for @option{-fpie} and 2 for @option{-fPIE}.
18150 @item -fno-jump-tables
18151 @opindex fno-jump-tables
18152 Do not use jump tables for switch statements even where it would be
18153 more efficient than other code generation strategies. This option is
18154 of use in conjunction with @option{-fpic} or @option{-fPIC} for
18155 building code which forms part of a dynamic linker and cannot
18156 reference the address of a jump table. On some targets, jump tables
18157 do not require a GOT and this option is not needed.
18159 @item -ffixed-@var{reg}
18161 Treat the register named @var{reg} as a fixed register; generated code
18162 should never refer to it (except perhaps as a stack pointer, frame
18163 pointer or in some other fixed role).
18165 @var{reg} must be the name of a register. The register names accepted
18166 are machine-specific and are defined in the @code{REGISTER_NAMES}
18167 macro in the machine description macro file.
18169 This flag does not have a negative form, because it specifies a
18172 @item -fcall-used-@var{reg}
18173 @opindex fcall-used
18174 Treat the register named @var{reg} as an allocable register that is
18175 clobbered by function calls. It may be allocated for temporaries or
18176 variables that do not live across a call. Functions compiled this way
18177 will not save and restore the register @var{reg}.
18179 It is an error to used this flag with the frame pointer or stack pointer.
18180 Use of this flag for other registers that have fixed pervasive roles in
18181 the machine's execution model will produce disastrous results.
18183 This flag does not have a negative form, because it specifies a
18186 @item -fcall-saved-@var{reg}
18187 @opindex fcall-saved
18188 Treat the register named @var{reg} as an allocable register saved by
18189 functions. It may be allocated even for temporaries or variables that
18190 live across a call. Functions compiled this way will save and restore
18191 the register @var{reg} if they use it.
18193 It is an error to used this flag with the frame pointer or stack pointer.
18194 Use of this flag for other registers that have fixed pervasive roles in
18195 the machine's execution model will produce disastrous results.
18197 A different sort of disaster will result from the use of this flag for
18198 a register in which function values may be returned.
18200 This flag does not have a negative form, because it specifies a
18203 @item -fpack-struct[=@var{n}]
18204 @opindex fpack-struct
18205 Without a value specified, pack all structure members together without
18206 holes. When a value is specified (which must be a small power of two), pack
18207 structure members according to this value, representing the maximum
18208 alignment (that is, objects with default alignment requirements larger than
18209 this will be output potentially unaligned at the next fitting location.
18211 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
18212 code that is not binary compatible with code generated without that switch.
18213 Additionally, it makes the code suboptimal.
18214 Use it to conform to a non-default application binary interface.
18216 @item -finstrument-functions
18217 @opindex finstrument-functions
18218 Generate instrumentation calls for entry and exit to functions. Just
18219 after function entry and just before function exit, the following
18220 profiling functions will be called with the address of the current
18221 function and its call site. (On some platforms,
18222 @code{__builtin_return_address} does not work beyond the current
18223 function, so the call site information may not be available to the
18224 profiling functions otherwise.)
18227 void __cyg_profile_func_enter (void *this_fn,
18229 void __cyg_profile_func_exit (void *this_fn,
18233 The first argument is the address of the start of the current function,
18234 which may be looked up exactly in the symbol table.
18236 This instrumentation is also done for functions expanded inline in other
18237 functions. The profiling calls will indicate where, conceptually, the
18238 inline function is entered and exited. This means that addressable
18239 versions of such functions must be available. If all your uses of a
18240 function are expanded inline, this may mean an additional expansion of
18241 code size. If you use @samp{extern inline} in your C code, an
18242 addressable version of such functions must be provided. (This is
18243 normally the case anyways, but if you get lucky and the optimizer always
18244 expands the functions inline, you might have gotten away without
18245 providing static copies.)
18247 A function may be given the attribute @code{no_instrument_function}, in
18248 which case this instrumentation will not be done. This can be used, for
18249 example, for the profiling functions listed above, high-priority
18250 interrupt routines, and any functions from which the profiling functions
18251 cannot safely be called (perhaps signal handlers, if the profiling
18252 routines generate output or allocate memory).
18254 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
18255 @opindex finstrument-functions-exclude-file-list
18257 Set the list of functions that are excluded from instrumentation (see
18258 the description of @code{-finstrument-functions}). If the file that
18259 contains a function definition matches with one of @var{file}, then
18260 that function is not instrumented. The match is done on substrings:
18261 if the @var{file} parameter is a substring of the file name, it is
18262 considered to be a match.
18267 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
18271 will exclude any inline function defined in files whose pathnames
18272 contain @code{/bits/stl} or @code{include/sys}.
18274 If, for some reason, you want to include letter @code{','} in one of
18275 @var{sym}, write @code{'\,'}. For example,
18276 @code{-finstrument-functions-exclude-file-list='\,\,tmp'}
18277 (note the single quote surrounding the option).
18279 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
18280 @opindex finstrument-functions-exclude-function-list
18282 This is similar to @code{-finstrument-functions-exclude-file-list},
18283 but this option sets the list of function names to be excluded from
18284 instrumentation. The function name to be matched is its user-visible
18285 name, such as @code{vector<int> blah(const vector<int> &)}, not the
18286 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
18287 match is done on substrings: if the @var{sym} parameter is a substring
18288 of the function name, it is considered to be a match. For C99 and C++
18289 extended identifiers, the function name must be given in UTF-8, not
18290 using universal character names.
18292 @item -fstack-check
18293 @opindex fstack-check
18294 Generate code to verify that you do not go beyond the boundary of the
18295 stack. You should specify this flag if you are running in an
18296 environment with multiple threads, but only rarely need to specify it in
18297 a single-threaded environment since stack overflow is automatically
18298 detected on nearly all systems if there is only one stack.
18300 Note that this switch does not actually cause checking to be done; the
18301 operating system or the language runtime must do that. The switch causes
18302 generation of code to ensure that they see the stack being extended.
18304 You can additionally specify a string parameter: @code{no} means no
18305 checking, @code{generic} means force the use of old-style checking,
18306 @code{specific} means use the best checking method and is equivalent
18307 to bare @option{-fstack-check}.
18309 Old-style checking is a generic mechanism that requires no specific
18310 target support in the compiler but comes with the following drawbacks:
18314 Modified allocation strategy for large objects: they will always be
18315 allocated dynamically if their size exceeds a fixed threshold.
18318 Fixed limit on the size of the static frame of functions: when it is
18319 topped by a particular function, stack checking is not reliable and
18320 a warning is issued by the compiler.
18323 Inefficiency: because of both the modified allocation strategy and the
18324 generic implementation, the performances of the code are hampered.
18327 Note that old-style stack checking is also the fallback method for
18328 @code{specific} if no target support has been added in the compiler.
18330 @item -fstack-limit-register=@var{reg}
18331 @itemx -fstack-limit-symbol=@var{sym}
18332 @itemx -fno-stack-limit
18333 @opindex fstack-limit-register
18334 @opindex fstack-limit-symbol
18335 @opindex fno-stack-limit
18336 Generate code to ensure that the stack does not grow beyond a certain value,
18337 either the value of a register or the address of a symbol. If the stack
18338 would grow beyond the value, a signal is raised. For most targets,
18339 the signal is raised before the stack overruns the boundary, so
18340 it is possible to catch the signal without taking special precautions.
18342 For instance, if the stack starts at absolute address @samp{0x80000000}
18343 and grows downwards, you can use the flags
18344 @option{-fstack-limit-symbol=__stack_limit} and
18345 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
18346 of 128KB@. Note that this may only work with the GNU linker.
18348 @item -fsplit-stack
18349 @opindex fsplit-stack
18350 Generate code to automatically split the stack before it overflows.
18351 The resulting program has a discontiguous stack which can only
18352 overflow if the program is unable to allocate any more memory. This
18353 is most useful when running threaded programs, as it is no longer
18354 necessary to calculate a good stack size to use for each thread. This
18355 is currently only implemented for the i386 and x86_64 backends running
18358 When code compiled with @option{-fsplit-stack} calls code compiled
18359 without @option{-fsplit-stack}, there may not be much stack space
18360 available for the latter code to run. If compiling all code,
18361 including library code, with @option{-fsplit-stack} is not an option,
18362 then the linker can fix up these calls so that the code compiled
18363 without @option{-fsplit-stack} always has a large stack. Support for
18364 this is implemented in the gold linker in GNU binutils release 2.21
18367 @item -fleading-underscore
18368 @opindex fleading-underscore
18369 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
18370 change the way C symbols are represented in the object file. One use
18371 is to help link with legacy assembly code.
18373 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
18374 generate code that is not binary compatible with code generated without that
18375 switch. Use it to conform to a non-default application binary interface.
18376 Not all targets provide complete support for this switch.
18378 @item -ftls-model=@var{model}
18379 @opindex ftls-model
18380 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
18381 The @var{model} argument should be one of @code{global-dynamic},
18382 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
18384 The default without @option{-fpic} is @code{initial-exec}; with
18385 @option{-fpic} the default is @code{global-dynamic}.
18387 @item -fvisibility=@var{default|internal|hidden|protected}
18388 @opindex fvisibility
18389 Set the default ELF image symbol visibility to the specified option---all
18390 symbols will be marked with this unless overridden within the code.
18391 Using this feature can very substantially improve linking and
18392 load times of shared object libraries, produce more optimized
18393 code, provide near-perfect API export and prevent symbol clashes.
18394 It is @strong{strongly} recommended that you use this in any shared objects
18397 Despite the nomenclature, @code{default} always means public; i.e.,
18398 available to be linked against from outside the shared object.
18399 @code{protected} and @code{internal} are pretty useless in real-world
18400 usage so the only other commonly used option will be @code{hidden}.
18401 The default if @option{-fvisibility} isn't specified is
18402 @code{default}, i.e., make every
18403 symbol public---this causes the same behavior as previous versions of
18406 A good explanation of the benefits offered by ensuring ELF
18407 symbols have the correct visibility is given by ``How To Write
18408 Shared Libraries'' by Ulrich Drepper (which can be found at
18409 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
18410 solution made possible by this option to marking things hidden when
18411 the default is public is to make the default hidden and mark things
18412 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
18413 and @code{__attribute__ ((visibility("default")))} instead of
18414 @code{__declspec(dllexport)} you get almost identical semantics with
18415 identical syntax. This is a great boon to those working with
18416 cross-platform projects.
18418 For those adding visibility support to existing code, you may find
18419 @samp{#pragma GCC visibility} of use. This works by you enclosing
18420 the declarations you wish to set visibility for with (for example)
18421 @samp{#pragma GCC visibility push(hidden)} and
18422 @samp{#pragma GCC visibility pop}.
18423 Bear in mind that symbol visibility should be viewed @strong{as
18424 part of the API interface contract} and thus all new code should
18425 always specify visibility when it is not the default; i.e., declarations
18426 only for use within the local DSO should @strong{always} be marked explicitly
18427 as hidden as so to avoid PLT indirection overheads---making this
18428 abundantly clear also aids readability and self-documentation of the code.
18429 Note that due to ISO C++ specification requirements, operator new and
18430 operator delete must always be of default visibility.
18432 Be aware that headers from outside your project, in particular system
18433 headers and headers from any other library you use, may not be
18434 expecting to be compiled with visibility other than the default. You
18435 may need to explicitly say @samp{#pragma GCC visibility push(default)}
18436 before including any such headers.
18438 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
18439 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
18440 no modifications. However, this means that calls to @samp{extern}
18441 functions with no explicit visibility will use the PLT, so it is more
18442 effective to use @samp{__attribute ((visibility))} and/or
18443 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
18444 declarations should be treated as hidden.
18446 Note that @samp{-fvisibility} does affect C++ vague linkage
18447 entities. This means that, for instance, an exception class that will
18448 be thrown between DSOs must be explicitly marked with default
18449 visibility so that the @samp{type_info} nodes will be unified between
18452 An overview of these techniques, their benefits and how to use them
18453 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
18455 @item -fstrict-volatile-bitfields
18456 @opindex fstrict-volatile-bitfields
18457 This option should be used if accesses to volatile bitfields (or other
18458 structure fields, although the compiler usually honors those types
18459 anyway) should use a single access of the width of the
18460 field's type, aligned to a natural alignment if possible. For
18461 example, targets with memory-mapped peripheral registers might require
18462 all such accesses to be 16 bits wide; with this flag the user could
18463 declare all peripheral bitfields as ``unsigned short'' (assuming short
18464 is 16 bits on these targets) to force GCC to use 16 bit accesses
18465 instead of, perhaps, a more efficient 32 bit access.
18467 If this option is disabled, the compiler will use the most efficient
18468 instruction. In the previous example, that might be a 32-bit load
18469 instruction, even though that will access bytes that do not contain
18470 any portion of the bitfield, or memory-mapped registers unrelated to
18471 the one being updated.
18473 If the target requires strict alignment, and honoring the field
18474 type would require violating this alignment, a warning is issued.
18475 If the field has @code{packed} attribute, the access is done without
18476 honoring the field type. If the field doesn't have @code{packed}
18477 attribute, the access is done honoring the field type. In both cases,
18478 GCC assumes that the user knows something about the target hardware
18479 that it is unaware of.
18481 The default value of this option is determined by the application binary
18482 interface for the target processor.
18488 @node Environment Variables
18489 @section Environment Variables Affecting GCC
18490 @cindex environment variables
18492 @c man begin ENVIRONMENT
18493 This section describes several environment variables that affect how GCC
18494 operates. Some of them work by specifying directories or prefixes to use
18495 when searching for various kinds of files. Some are used to specify other
18496 aspects of the compilation environment.
18498 Note that you can also specify places to search using options such as
18499 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
18500 take precedence over places specified using environment variables, which
18501 in turn take precedence over those specified by the configuration of GCC@.
18502 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
18503 GNU Compiler Collection (GCC) Internals}.
18508 @c @itemx LC_COLLATE
18510 @c @itemx LC_MONETARY
18511 @c @itemx LC_NUMERIC
18516 @c @findex LC_COLLATE
18517 @findex LC_MESSAGES
18518 @c @findex LC_MONETARY
18519 @c @findex LC_NUMERIC
18523 These environment variables control the way that GCC uses
18524 localization information that allow GCC to work with different
18525 national conventions. GCC inspects the locale categories
18526 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
18527 so. These locale categories can be set to any value supported by your
18528 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
18529 Kingdom encoded in UTF-8.
18531 The @env{LC_CTYPE} environment variable specifies character
18532 classification. GCC uses it to determine the character boundaries in
18533 a string; this is needed for some multibyte encodings that contain quote
18534 and escape characters that would otherwise be interpreted as a string
18537 The @env{LC_MESSAGES} environment variable specifies the language to
18538 use in diagnostic messages.
18540 If the @env{LC_ALL} environment variable is set, it overrides the value
18541 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
18542 and @env{LC_MESSAGES} default to the value of the @env{LANG}
18543 environment variable. If none of these variables are set, GCC
18544 defaults to traditional C English behavior.
18548 If @env{TMPDIR} is set, it specifies the directory to use for temporary
18549 files. GCC uses temporary files to hold the output of one stage of
18550 compilation which is to be used as input to the next stage: for example,
18551 the output of the preprocessor, which is the input to the compiler
18554 @item GCC_COMPARE_DEBUG
18555 @findex GCC_COMPARE_DEBUG
18556 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
18557 @option{-fcompare-debug} to the compiler driver. See the documentation
18558 of this option for more details.
18560 @item GCC_EXEC_PREFIX
18561 @findex GCC_EXEC_PREFIX
18562 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
18563 names of the subprograms executed by the compiler. No slash is added
18564 when this prefix is combined with the name of a subprogram, but you can
18565 specify a prefix that ends with a slash if you wish.
18567 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
18568 an appropriate prefix to use based on the pathname it was invoked with.
18570 If GCC cannot find the subprogram using the specified prefix, it
18571 tries looking in the usual places for the subprogram.
18573 The default value of @env{GCC_EXEC_PREFIX} is
18574 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
18575 the installed compiler. In many cases @var{prefix} is the value
18576 of @code{prefix} when you ran the @file{configure} script.
18578 Other prefixes specified with @option{-B} take precedence over this prefix.
18580 This prefix is also used for finding files such as @file{crt0.o} that are
18583 In addition, the prefix is used in an unusual way in finding the
18584 directories to search for header files. For each of the standard
18585 directories whose name normally begins with @samp{/usr/local/lib/gcc}
18586 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
18587 replacing that beginning with the specified prefix to produce an
18588 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
18589 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
18590 These alternate directories are searched first; the standard directories
18591 come next. If a standard directory begins with the configured
18592 @var{prefix} then the value of @var{prefix} is replaced by
18593 @env{GCC_EXEC_PREFIX} when looking for header files.
18595 @item COMPILER_PATH
18596 @findex COMPILER_PATH
18597 The value of @env{COMPILER_PATH} is a colon-separated list of
18598 directories, much like @env{PATH}. GCC tries the directories thus
18599 specified when searching for subprograms, if it can't find the
18600 subprograms using @env{GCC_EXEC_PREFIX}.
18603 @findex LIBRARY_PATH
18604 The value of @env{LIBRARY_PATH} is a colon-separated list of
18605 directories, much like @env{PATH}. When configured as a native compiler,
18606 GCC tries the directories thus specified when searching for special
18607 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
18608 using GCC also uses these directories when searching for ordinary
18609 libraries for the @option{-l} option (but directories specified with
18610 @option{-L} come first).
18614 @cindex locale definition
18615 This variable is used to pass locale information to the compiler. One way in
18616 which this information is used is to determine the character set to be used
18617 when character literals, string literals and comments are parsed in C and C++.
18618 When the compiler is configured to allow multibyte characters,
18619 the following values for @env{LANG} are recognized:
18623 Recognize JIS characters.
18625 Recognize SJIS characters.
18627 Recognize EUCJP characters.
18630 If @env{LANG} is not defined, or if it has some other value, then the
18631 compiler will use mblen and mbtowc as defined by the default locale to
18632 recognize and translate multibyte characters.
18636 Some additional environments variables affect the behavior of the
18639 @include cppenv.texi
18643 @node Precompiled Headers
18644 @section Using Precompiled Headers
18645 @cindex precompiled headers
18646 @cindex speed of compilation
18648 Often large projects have many header files that are included in every
18649 source file. The time the compiler takes to process these header files
18650 over and over again can account for nearly all of the time required to
18651 build the project. To make builds faster, GCC allows users to
18652 `precompile' a header file; then, if builds can use the precompiled
18653 header file they will be much faster.
18655 To create a precompiled header file, simply compile it as you would any
18656 other file, if necessary using the @option{-x} option to make the driver
18657 treat it as a C or C++ header file. You will probably want to use a
18658 tool like @command{make} to keep the precompiled header up-to-date when
18659 the headers it contains change.
18661 A precompiled header file will be searched for when @code{#include} is
18662 seen in the compilation. As it searches for the included file
18663 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
18664 compiler looks for a precompiled header in each directory just before it
18665 looks for the include file in that directory. The name searched for is
18666 the name specified in the @code{#include} with @samp{.gch} appended. If
18667 the precompiled header file can't be used, it is ignored.
18669 For instance, if you have @code{#include "all.h"}, and you have
18670 @file{all.h.gch} in the same directory as @file{all.h}, then the
18671 precompiled header file will be used if possible, and the original
18672 header will be used otherwise.
18674 Alternatively, you might decide to put the precompiled header file in a
18675 directory and use @option{-I} to ensure that directory is searched
18676 before (or instead of) the directory containing the original header.
18677 Then, if you want to check that the precompiled header file is always
18678 used, you can put a file of the same name as the original header in this
18679 directory containing an @code{#error} command.
18681 This also works with @option{-include}. So yet another way to use
18682 precompiled headers, good for projects not designed with precompiled
18683 header files in mind, is to simply take most of the header files used by
18684 a project, include them from another header file, precompile that header
18685 file, and @option{-include} the precompiled header. If the header files
18686 have guards against multiple inclusion, they will be skipped because
18687 they've already been included (in the precompiled header).
18689 If you need to precompile the same header file for different
18690 languages, targets, or compiler options, you can instead make a
18691 @emph{directory} named like @file{all.h.gch}, and put each precompiled
18692 header in the directory, perhaps using @option{-o}. It doesn't matter
18693 what you call the files in the directory, every precompiled header in
18694 the directory will be considered. The first precompiled header
18695 encountered in the directory that is valid for this compilation will
18696 be used; they're searched in no particular order.
18698 There are many other possibilities, limited only by your imagination,
18699 good sense, and the constraints of your build system.
18701 A precompiled header file can be used only when these conditions apply:
18705 Only one precompiled header can be used in a particular compilation.
18708 A precompiled header can't be used once the first C token is seen. You
18709 can have preprocessor directives before a precompiled header; you can
18710 even include a precompiled header from inside another header, so long as
18711 there are no C tokens before the @code{#include}.
18714 The precompiled header file must be produced for the same language as
18715 the current compilation. You can't use a C precompiled header for a C++
18719 The precompiled header file must have been produced by the same compiler
18720 binary as the current compilation is using.
18723 Any macros defined before the precompiled header is included must
18724 either be defined in the same way as when the precompiled header was
18725 generated, or must not affect the precompiled header, which usually
18726 means that they don't appear in the precompiled header at all.
18728 The @option{-D} option is one way to define a macro before a
18729 precompiled header is included; using a @code{#define} can also do it.
18730 There are also some options that define macros implicitly, like
18731 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
18734 @item If debugging information is output when using the precompiled
18735 header, using @option{-g} or similar, the same kind of debugging information
18736 must have been output when building the precompiled header. However,
18737 a precompiled header built using @option{-g} can be used in a compilation
18738 when no debugging information is being output.
18740 @item The same @option{-m} options must generally be used when building
18741 and using the precompiled header. @xref{Submodel Options},
18742 for any cases where this rule is relaxed.
18744 @item Each of the following options must be the same when building and using
18745 the precompiled header:
18747 @gccoptlist{-fexceptions}
18750 Some other command-line options starting with @option{-f},
18751 @option{-p}, or @option{-O} must be defined in the same way as when
18752 the precompiled header was generated. At present, it's not clear
18753 which options are safe to change and which are not; the safest choice
18754 is to use exactly the same options when generating and using the
18755 precompiled header. The following are known to be safe:
18757 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
18758 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
18759 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
18764 For all of these except the last, the compiler will automatically
18765 ignore the precompiled header if the conditions aren't met. If you
18766 find an option combination that doesn't work and doesn't cause the
18767 precompiled header to be ignored, please consider filing a bug report,
18770 If you do use differing options when generating and using the
18771 precompiled header, the actual behavior will be a mixture of the
18772 behavior for the options. For instance, if you use @option{-g} to
18773 generate the precompiled header but not when using it, you may or may
18774 not get debugging information for routines in the precompiled header.